Fast evolutionary rates associated with functional loss in class I glucose transporters of Schistosoma mansoni

The trematode parasite, Schistosoma mansoni, has evolved to switch from oxidative phosphorylation to glycolysis in the presence of glucose immediately after invading the human host. This metabolic switch is dependent on extracellular glucose concentration. Four glucose transporters are encoded in the genome of S. mansoni, however, only two were shown to facilitate glucose diffusion. By modeling the phase of human host infection, we showed that transporter transcript expression profiles of recently transformed schistosomula have two opposing responses to increased glucose concentrations. Concurring with the transcription profiles, our phylogenetic analyses revealed that S. mansoni glucose transporters belong to two separate clusters, one associated with class I glucose transporters from vertebrates and insects, and the other specific to parasitic Platyhelminthes. To study the evolutionary paths of both groups and their functional implications, we determined evolutionary rates, relative divergence times, genomic organization and performed structural analyses with the protein sequences. We finally used the modelled structures of the S. mansoni glucose transporters to biophysically (i) analyze the dynamics of key residues during glucose binding, (ii) test glucose stability within the active site, and (iii) demonstrate glucose diffusion. The two S. mansoni Platyhelminthes-specific glucose transporters, which seem to be younger than the other two, exhibit slower rates of molecular evolution, are encoded by intron-poor genes, and transport glucose. Interestingly, our molecular dynamic analyses suggest that S. mansoni class I glucose transporters are not able to transport glucose. The glucose transporter family in S. mansoni exhibit different evolutionary histories. Our results suggested that S. mansoni class I glucose transporters lost their capacity to transport glucose and that this function evolved independently in the Platyhelminthes-specific glucose transporters. Finally, taking into account the differences in the dynamics of glucose transport of the Platyhelminthes-specific transporters of S. mansoni compared to that of humans, we conclude that S. mansoni glucose transporters may be targets for rationally designed drugs against schistosomiasis.


Background
Upon contact with host mammalian skin, the free-living cercaria of Schistosoma mansoni undergoes dramatic biological [1] and metabolic transformations [2]. The metabolic transformation is a switch from an oxidative metabolism to a glycolytic metabolism that is reversibly dependent on external glucose concentrations [2]. In the mammalian host, glucose transport is facilitated by diffusion through the S. mansoni tegument [3]. Two closely related (61 % identity) schistosome glucose transporter proteins (SGTPs) have been functionally characterized in S. mansoni, named SGTP1 and SGTP4 [4,5]. Both show stereospecificity for glucose, relaxed specificity for different hexoses and sodium-independent activity [4]. Differential protein expression profiles show that SGTP4 is expressed in the host-interactive outer tegument [6], while SGTP1 is mainly located in the basal membranes of the tegument [7]. Furthermore, Western blot analysis showed that SGTP1 is expressed in the egg, sporocyst, cercaria, schistosomula, and in adult male and female worms, but SGTP4 is only detected in mammalian-stage parasites (schistosomula and adults) [6]. Studies using RNAi showed that both SGTP1 and SGTP4 were critical for exogenous glucose uptake in schistosomula in vitro and for parasite development in vivo [5].
A third glucose transporter gene, sgtp2, has also been identified. The sgtp2 open reading frame was interrupted by a missing base~260 bp downstream the first initiator ATG codon. The hypothetical addition of a base at this downstream position resulted in a translation product homologous to glucose transporters [4]. Functional assays using Xenopus oocytes, however, showed that this hypothetical SGTP2 does not transport glucose (at 1 mM and 10 mM of substrate). In addition, transcripts of the sgtp2 gene were only found in adult females [4], possibly due to a specific biological function at this stage. Although expressed sequence tags exist for all S. mansoni sgtp genes [5,8], a fourth glucose transporter, sgtp3, is rarely mentioned in the literature.
The human glucose transporter protein family is divided into three classes: class I comprises GLUT1 to GLUT4 and GLUT14; class II comprises GLUT5, GLUT7, GLUT9, and GLUT11; and, class III comprises GLUT6, GLUT8, GLUT10 and GLUT12 (Augustin [9]). These three human glucose transporter classes are phylogenetically differentiated and show distinct molecular properties [9]. Among the human glucose transporters, GLUT1 and GLUT4 have been well studied and are transcriptionally upregulated under glucose deprivation conditions [10]. In terms of primary amino acid sequence, SGTP1 and SGTP4 were reported to be homologous to human GLUT1 [11], while SGTP2 is more similar to GLUT4 [4].
Glucose transporters undergo conformational shifts during glucose diffusion [12]. The resolved crystal structure of the proton symporter, XylE, of Escherichia coli has a glucose-bound, occluded structural conformation towards the intracellular compartment [13]. This occluded conformation disallows glucose diffusion into the cell. The crystal structure of human GLUT1, possesses a glucose-bound, open-inward structural conformation permitting glucose diffusion within the cell [12]. Comparing these two homologous transporter structures, Deng et al. [12] were able to hypothesize four intermediate conformational shifts during glucose binding and diffusion. XylE and GLUT1 represent the two respective, intermediate and sequential, glucose-bound conformations. The other two are glucose-free that represent conformations before glucose binding and after glucose diffusion [12].
In this study, the evolution of S. mansoni glucose transporters is explored and the effects of glucose-induced transcriptional regulation of the encoded genes were determined. Our findings led us to investigate glucose stability and migration with S. mansoni glucose transporters using a publicly available, state-of-the-art algorithm. These results provide insights into the molecular properties, evolution and biophysics of glucose transport in trematodes. Unveiling the structural and dynamic differences in glucose transport by parasitic worms can establish the basis for the rational design of schistosome-specific glucose transporter inhibitors.

Results and discussion
Glucose induces transcriptional changes in S. mansoni glucose transporter genes in schistosomula larvae Regulation of nutrient transporters by nutrient availability is a well-known phenomenon in microorganisms, such as yeast [14] and bacteria [15]. The regulation of mammalian transporters by their substrates, however, is less understood [10]. Nevertheless, the effect of glucose on the transcriptional regulation of glucose transporters has been studied both in vitro and in vivo [10]. From these studies it is clear that mRNA levels of class I glucose transporters glut1 and glut4, the most extensively studied in this regard, change in response to the glucose concentration in the medium. As a consensus, glut1 and glut4 mRNA levels are higher under glucose deprivation conditions than in the presence of glucose [10]. Despite extensive characterization of SGTP1 and SGTP4 from S. mansoni [5], no study on transcriptional regulation in response to glucose has been conducted. In addition, knowledge of the properties of SGTP2 and SGTP3 is limited.
Upon infection, changes in glucose concentration have been shown to be essential for S. mansoni [2]. Here we used an infection relevant model to evaluate the transcriptional regulation of the four glucose transporter genes from S. mansoni in presence of glucose. All glucose transporters were expressed in schistosomula 3 h after mechanical transformation in the presence of 0.05 mM glucose. A significant reduction in expression was observed after 8 h incubation in glucose deprivation conditions (p-value <0.01). However, when the medium was supplemented with 10 mM glucose we observed a significant change in expression pattern. While the transcriptional levels of sgtp1 and sgtp4 were downregulated (p-value <0.025), the transcriptional levels of sgtp2 and sgtp3 were upregulated (p-value <0.025) in the presence of 10 mM glucose (Fig. 1). The downregulation of sgtp1 and sgtp4 transcripts was specific to glucose, as the addition of galactose or maltose to the medium resulted in a significant increase (p-value <0.025) in the transcriptional levels of both genes (Fig. 1).
Our results complement those obtained by Krautz-Peterson et al. [5] since they showed that RNA interference (RNAi) of both SGTP1 and SGTP4 more strongly affected schistosomula when they were maintained in glucose-poor medium than in glucose-rich medium. In glucose deprivation conditions the parasite may increase the expression of these two glucose transporters to maximize the uptake of glucose. Thus, inhibiting the glucose transporters in glucose deprivation conditions will be more detrimental to the parasite than in the presence of glucose. The transcriptional expression levels of sgtp2 significantly increased in the presence of galactose (p-value = 0.010), but not maltose (p-value = 0.163). In contrast, the transcriptional expression levels of sgtp3 significantly decreased in the presence of galactose (p-value = 0.004), but not maltose (p-value = 0.125). These observations suggest that the transcriptional regulation of sgtp2 and sgtp3 may not be glucose-specific.
S. mansoni glucose transporters cluster separately and evolve at different rates Maximum likelihood (ML), maximum parsimony (MP) and neighbor joining (NJ) analyses were performed to phylogenetically characterize the S. mansoni glucose transporters SGTP1, SGTP2, SGTP3 and SGTP4. Amino acid sequences from glucose and trehalose transporters from the taxonomic classes Insecta, Mammalia, Teleostei, Cestoda and Trematoda were included in the phylogenetic analyses. All phylogenetic trees had the same topology with each of the above methods. Therefore, only one topology (ML) is displayed (Fig. 2).
Three classes of glucose transporters have been described in humans that form different phylogenetic clusters [9]. We found that S. mansoni SGTP2 and SGTP3 clustered together with vertebrate and D. melanogaster class I glucose transporters (Fig. 2, roman numeral I), while SGTP1 and SGTP4 formed a Platyhelminthesspecific glucose transporter cluster. The phylogeny presented in Fig. 2 is consistent with the monophyly of both Cestoda and Digenea [16] since the two clusters Fig. 1 Relative expression of sgtp1, sgtp2, sgtp3 and sgtp4 after transformation and adding glucose. The relative expression of sgtp1, sgtp4 (Panel a), sgtp2 and sgtp3 (Panel b) in schistosomula was determined using quantitative RT-PCR. Cercariae were mechanically transformed into schistosomula. A total of 10 000 schistosomula were incubated for 3 h with traces of glucose (0.05 mM). Subsequently, glucose, maltose and galactose were added to a final concentration of 10 mM. After 8 h of incubation, total RNA was extracted, cDNA was synthesized and quantitative RT-PCR was performed. Transcripts of sgtp1 (p-value = 0.0006) and sgtp4 (p-value = 0.02) were significantly downregulated after adding 10 mM glucose, while sgtp2 (p-value = 0.002) and sgtp3 (p-value = 0.02) were significantly upregulated after adding 10 mM glucose. Results shown are the means of three independent experiments. Gene expression data is presented relative to 8 h with 0.05 mM of glucose which was used as "calibrator sample" containing glucose transporters from Platyhelminthes, cestodes and trematodes clustered separately (Fig. 2). The phylogenetic analyses allowed us to classify SGTP2 and SGTP3 together with some glucose transporters from the flukes S. japonicum, Clonorchis sinensis and the cestodes Taenia solium, Echinococcus granulosus, E. multilocularis and Hymenolepis microstoma as class I glucose transporters. In particular, we found homologs of SGTP1, SGTP2, SGTP3 and SGTP4 in both S. japonicum and S. haematobium (Additional file 1).
Mammalia, Platyhelminthes and Insecta shared a common ancestor circa 695 MYA [17]. The fact that class I glucose transporters included sequences from Platyhelminthes, insects and mammals suggested that this cluster is ancestral to the Platyhelminthes-specific cluster. To test this hypothesis we determined the relative divergence times for selected nodes in the 100 most parsimonious phylogenetic trees (see Methods). Indeed, the results from this analysis showed that the cluster containing class I glucose transporters is older than the Platyhelminthes-specific cluster (Fig. 3a, node 2 vs. node 10) with p-value <0.0001. In agreement with this, SGTP2 seems to be the oldest glucose transporter in S. mansoni (Fig. 3b, p-value <0.0001). The relative divergence time of SGTP3 and SGTP4 suggested that these phylogenetic events coincide in time, and they are recently evolved glucose transporters compared to SGTP2 and SGTP1. The phylogenetic position of SGTP3 and SGTP4 (Fig. 2), and the relative divergence times (Fig. 3b) of these genes support the idea that they originated from SGTP2 and SGTP1 respectively, possibly due to gene duplication.
Gene duplications were reported in glucose transporters from Homo sapiens and Saccharomyces cerevisiae [18,19]. After gene duplication, relaxation of selective constraints on duplicated genes has been observed, resulting in copies that evolve faster than their original orthologs [20,21]. To test variations in evolutionary rates, we estimated these along branches of the glucose transporters phylogeny using the RelTime method [22]. Figure 3c shows that SGTP2 and SGTP3 evolved faster than SGTP1 and SGTP4 (p-value = 0.0001). In addition, SGTP3 and SGTP4 were found to evolve faster than SGTP2 and SGTP1, respectively with p-value <0.0001.
We further compared the evolutionary rates of SGTP3/ SGTP2 and SGTP4/SGTP1 using Tajima's relative rate test [23]. The χ2 test statistic was 6.50 (p-value = 0.01079) for SGTP4/SGTP1 and 0 (p-value = 1.0000) for SGTP3/ SGTP2. Thus, the null hypothesis of equal rates was rejected for SGTP4/SGTP1, but not for SGTP3/SGTP2. This suggests that Tajima's relative rate test may not be suited for distinguishing different rates of evolution in fast-evolving sequences like SGTP3/SGTP2 or simply that after divergence SGTP3 and SGTP2 evolved at the same rates. In agreement with the latter, other sequences from S. japonicum and C. sinensis, in the cluster where SGTP2 and SGTP3 are located, are also fast-evolving sequences (Fig. 3). Interestingly, the transcriptional regulation of SGTP1 and SGTP4 compared with SGTP2 and SGTP3 in the presence of glucose ( Fig. 1) is in agreement with these phylogenetic and evolutionary observations.
The rapidly-evolving glucose transporters SGTP2 and SGTP3 are encoded by intron-rich genes The cDNA sequence encoding SGTP2 originally described by Skelly et al. [4] was constructed from a cloned sequence by the insertion of a single nucleotide in order to restore the open reading frame. The authors speculated that sgtp2 might therefore be an expressed pseudogene and not encoded as a functional protein. Comparison of the sgtp2 (L25066) cDNA sequence to the genome sequence allowed the reconstruction of a "genomic" predicted cDNA sequence. Three base changes were found within the encoding region, one leading to an amino acid change (Asn88Lys). Since the cDNA sequence encodes a full-length protein reconstructed by assembling the exons Fig. 3 Relative divergence times and evolutionary rates of S. mansoni glucose transporters. The figure displays the relative divergence times of selected nodes from the ML tree topology in Fig. 1. Class I glucose transporters (node 2) and Platyhelminthes-specific cluster of glucose transporters (node 10) (Panel a). Divergence times for selected branching points were calculated in the 100 most parsimonious trees using the ML method based on the LG model (see Methods). Node 2 (average of relative time of divergence: 0.68, Std: 0.02) is older than node 10 (average of relative time of divergence: 0.28, Std: 0.009) with p-value < 0.0001. The colored scale shows the relative evolutionary rates among selected branches (Panels a and b). The evolutionary rate (evol rate) measured as the relative number of amino acid substitutions per site was calculated for selected nodes in the 100 most parsimonious trees using RelTime method [22]. The glucose transporters SGTP2 (average of evol rate: 3.6, Std: 0.42) and SGTP3 (average of evol rate: 4.9, Std: 0.81) evolved faster than SGTP1 (average of evol rate: 0.13, Std: 0.028) and SGTP4 (average of evol rate: 1.6, Std: 0.067) with a p-value < 0.0001 (Panel c) of the genomic sequence, we speculated that the missing nucleotide in the cDNA clone obtained by Skelly et al. [4] might represent a cloning artifact. We therefore amplified the full-length coding sequence of SGTP2 from cDNA derived from adult worms and miracidia and sequenced the clones obtained. However, the examination of twenty such sequences failed to identify a single clone corresponding to the predicted genomic sequence (not shown). A number of these contain unspliced intronic sequences, whilst others contained short indels. None of the clones encoded a full-length SGTP2 protein sequence. We are therefore unable to verify our speculation concerning the status of the sgtp2 gene as a pseudogene or a viable coding gene. One possibility is that the functional transcript might only be produced at specific stages during the life cycle. To compare the sgtp2 gene organization and the potential structure and function of the SGTP2 protein, we used the transcript and protein sequence predicted by the genome. In contrast we confirmed that the sequence of SGTP3 predicted from the genome corresponds to viable transcripts present in adult worms. The transcript sequences were found to be identical to the genome prediction, but have a small deletion at the N-terminal end and to encode a further 86 amino acids at the C-terminal end. However, these sequence modifications have no effect on the core glucose transporter structure. The transcript sequence was submitted to GenBank [GenBank: KT895372].
The schistosome genome sequence and the derived exon/intron structure shown in Fig. 4 and Table 1 are aligned with published transcript sequences of SGTP1 and SGTP4, the transcript sequence of SGTP2 predicted from the genome, and the predicted transcript sequence of SGTP3. We analyzed the gene structure of human glut1, and collected that of glut4 from the literature [24], for comparison, since our phylogenetic analyses placed the human GLUT1 and GLUT4 within the same clade as SGTP2 and SGTP3. The analysis shows that the gene structures of sgtp1 and sgtp4 are similar to each other, but different from those of sgtp2 and sgtp3, which are also similar to each other. Both sgtp1 and sgtp4 have 5 introns, four of which are large (>1 000 bp), the size of the fifth was not evaluated due to missing sequence data. In contrast, sgtp2 has 10 introns and sgtp3 has 7 introns. Apart from the first intron of sgtp2, which is large and not conserved in sgtp3, the first introns of these two genes are small, between 31 bp and 41 bp, whereas the  Table 1 for complementary information) introns at the 3' end of both genes are large. Moreover, the sizes of exons 4 to 9 of sgtp2 and 2 to 7 of sgtp3 are identical and the positions of these introns with respect to the encoded amino acid sequence of the corresponding proteins are conserved. Comparison of sgtp1, sgtp4, sgtp2, sgtp3 and glut1 genes (Table 1) shows that 3 intron positions are perfectly conserved between all, and that the position of introns 1 and 2 is conserved between all but sgtp3. In addition, glut1 shares an intron position (intron 5) with sgtp2 (intron 6) and sgtp 3 (intron 4) that is not present in sgtp1 or sgtp4. Overall, this analysis underlines the close relationship between sgtp2 and sgtp3, and sgtp1 and sgtp4 that correlates well with our phylogenetic analysis.

S. mansoni glucose transporters show typical molecular signatures
Despite SGTP1 and SGTP4 being clustered separately from class I and II glucose transporters, they share several molecular properties with these classes (Fig. 5). Human class I glucose transporters comprise GLUT1, GLUT2, GLUT3, GLUT4 and GLUT14 with GLUT1 being the first isoform cloned described by Mueckler et al. [28]. All members of GLUT family isolated from humans possess 12 transmembrane (TM) α-helices [9], structural patterns that are also present in some glucose transporters from Insecta [29]. Using the TMHMM server we predicted the presence of the 12 TM α-helices in SGTP1-4 and compared their positions with human class I glucose transporters (Fig. 5). The positions of these TMs were conserved in the four glucose transporters from S. mansoni, except for the lack of TM7 in SGTP3. The only remarkable difference between the amino acid sequence of SGTP3 and the other glucose transporters at TM7 is the presence of two phenylalanine residues (Fig. 5). However, the modelling analysis (see below) showed that SGTP3 has the typical 12 TM α-helices. The discrepancy between TMHMM and modelling predictions regarding TMs in SGTP3 can only be solved through the experimental determination of the structure of this transporter. Probably due to differences in the algorithm used, the positions of our predicted TM differ slightly from those reported by Skelly et al. [4] for SGTP1 and SGTP4. We then compared the sugar transporter signatures [9,30] of human and S. mansoni glucose transporters. Some glycine (Gly) residues critical for structure stabilization in the GLUT family [28] were also conserved in the TM α-helices 1, 4, 5, 7 and 10 of S. mansoni glucose transporters (Fig. 5). The glutamic acid (Glu) and arginine (Arg) residues implicated in conformational alteration [31] and membrane topology [32] in human GLUTs were also present in cytoplasmic loops 2, 4, 8, and 10 of S. mansoni glucose transporters (Fig. 5).
Some exclusive characteristics of class I/II glucose transporters in humans are conserved in S. mansoni. For example, the N-linked glycosylation site (Fig. 5) in the first extracellular loop between TM1 and TM2 [33] and the motif PESPR in the sixth intracellular loop between TM6 and TM7 (Fig. 5). SGTP2 and SGTP3 showed an amino acid change in the last Arg of the motif PESPR to threonine (Thr) and serine (Ser), respectively. Despite the improbability of deducing substrate specificity or transport kinetics from the primary amino acid structure of class I glucose transporters, GLUT1, 3 and 4 are known to transport glucose, and not fructose, via a QLS sequence present in TM 7 [34]. This triad sequence has been implicated in substrate binding and is conserved in S. mansoni SGTP1, SGTP4 and SGTP3 (Fig. 5), but not Table 1 Genomic structure and organization of S. mansoni glucose transporters Exon and intron sizes are in base pairs In bold: Exons for which the exon-intron junction at the 3' end is conserved between different glucose transporter genes In blue: Exons for which the exon-intron junction at the 3' end is conserved among all selected glucose transporter genes In red: Exons for which the exon-intron junction at the 3' end is conserved only between human glut1, sgtp2 and sgtp3 In green: Exons for which the exon-intron junction at the 3' end is conserved only between sgtp2 and sgtp3 n.e. not evaluated (information missing from genome sequences) in SGTP2. In SGTP2 there is a substitution of leucine (Leu) by phenylalanine (Phe) and the same substitution was found in the human GLUT2. Human GLUT2 has low affinity for glucose (K m = 17 mM), which is uncommon among the known members of the GLUT family [35], but has a high affinity for glucosamine (K m ≈ 0.8 mM) [36]. S. mansoni SGTP2 does not seem to transport glucose in Xenopus oocytes at 10 mM glucose [4], however, further studies should clarify whether this molecule shows a higher K m for glucose or whether it has an affinity for glucosamine as does GLUT2. The presence of glucose transporters with different affinities for glucose is extremely important for the regulation of glucose uptake. For instance, under high glucose concentration conditions an increase in intracellular glucose reduces the glucose influx 50 % in yeast cells expressing high-affinity glucose transporters [37]. Thus, by regulating the expression profile of the different glucose transporters with different kinetic properties, yeasts avoid affecting their glucose uptake, which is directly dependent on extracellular glucose concentration [38]. Despite the fact that the hypothesis of glucose transporters with different affinities for glucose in S. mansoni is appealing and matches the expression profile in Fig. 1, our biophysical analysis (below) suggests that neither SGTP2 nor SGTP3 transport glucose.
The predicted tertiary structural conformations of S. mansoni glucose transporters are homologous to GLUT1 and XylE Upon substrate binding, the tertiary structure of most proteins undergoes conformational changes that coordinate their function. By comparing the structural differences of XylE and GLUT1, Deng et al. [12] were able to hypothesize on the conformational changes of glucose transporters during glucose binding and diffusion. This working model of glucose diffusion predicts that transporters prefer an outward (towards the extracellular), open, glucose-free conformation (i.e., the initial conformation). Glucose association (i.e., conformation of XylE) and disassociation (i.e., conformation of GLUT1) trigger the migration of glucose towards an environment with a lower concentrationi.e., the intracellular compartment [12]. The final conformation, glucose-free occluded (by the intracellular α-helices), remains to be structurally resolved.
As proof of principle for our simulations, we submitted the modelled S. mansoni glucose transporters to the DALI server [39] to find homologous crystal structures with the best matching conformation. The crystal structures GLUT1 of H. sapiens [PDB: 4PYP] and XylE of E. coli [PDB: 4GBZ] were among the top candidates. This provided two opportunities to control our simulations since, (i) glucose is not the substrate of XylE, it is a competitive inhibitor and (ii) XylE and GLUT1 are the two sequential intermediate structural conformations during glucose binding and diffusion, respectively [12]. The conformational changes during binding occur mainly in TM α-helices 1, 4, 7 and 10 that cause the occlusion by the intracellular α-helices to open, thereby permitting diffusion [12]. Root mean square deviation (RMSD) is frequently used in structural bioinformatics to measure the average distances between atoms from two or more protein structures. The superposition in Fig. 6a graphically shows these overall structural differencesthe RMSD between XylE [PDB: 4GBZ] and GLUT1 [PDB: 4PYP] is 3.2 Å. Figure 6b shows the predicted tertiary structures of S. mansoni glucose transporters SGTP1-SGTP4 and their structural similarity to GLUT1 and XylE. Both SGTP1 (RMSD 1.0 Å) and SGTP4 (RMSD 1.3 Å) are structurally homologous to XylE. SGTP2 (RMSD 2.3 Å) and SGTP3 (RMSD 0.6 Å) are structurally homologous to GLUT1. The higher RMSD for SGTP2 is mainly caused by differences in TM α-helices 9 and 10, intracellular αhelices and the C-terminus. These opposing conformations allow us to examine the stability of glucose within the active site of SGTP1 and SGTP4 and the inability of glucose migration (or diffusion) by SGTP2 and SGTP3.

Residue dynamics reveal how glucose affects invariant residues of glucose transporters involved in binding
The dynamics of key residues during substrate binding vary depending on the substrate and the protein. Since 1958, the key-lock theory to explain protein-substrate complexes has been revised by Koshland [40] who proposed an induced-fit process. The induced-fit theory demonstrates that during protein-substrate interactions, conformational changes occur between both the protein and substrate to accommodate the complex. These molecular interactions subsequently cause the global conformational changes aforementioned in Fig. 6. To understand how glucose will interact with key residues of glucose transporters, we performed an induced-fit docking simulation using the PELE server (see Methods).

Biophysical properties of S. mansoni glucose transporters provide insights into glucose affinity
The induced-fit simulations revealed how invariant residues behave upon binding (since the algorithm constrains glucose migration within the active site), but what would happen if glucose were unconstrained? To answer this question, we simulated an unconstrained glucose migration starting from the active site. Due the importance of Tyr292/298 in maintaining the openinward conformation of glucose transporters [12], we used it to triangulate glucose migration from its initial position in the active site. Additional file 2 shows distinct glucose migration patterns for GLUT1 and XylE. In accordance with the structural similarities (Fig. 6), transcript expression (Fig. 1) and evolution (Figs. 2 and 3), S. mansoni glucose transporters group together in glucose migration patterns showing that in SGTP2 and SGTP3 glucose migrates similarly to GLUT1, while SGTP1 and SGTP4 are similar to XylE. These migration patterns , and SGTP2 (magenta) and SGTP3 (cyan) with GLUT1 of H. sapiens (red). Structures are color-coded from the amino-terminus (N-ter; red) to the carboxyl-terminus (C-ter; blue). The structural alignment of the α-carbon backbone was performed using the default tool of Schrodinger's Maestro program seem to show glucose either slowly diffusing away from the active site (i.e., XylE and SGTP4) or being ejected from it (i.e., GLUT1, SGTP2 and SGTP3). In SGTP1, however, glucose seems to stay within 10 Å of the active site (Additional file 2). This concentrated positioning within the active site agrees with the higher affinity that SGTP1 has for glucose compared with other SGTPs [4].
In order to determine the biophysics of glucose affinity and the direction of migration patterns (towards the extracellular or intracellular compartments) we performed two types of analyses: (i) a cluster analysis to visualize the direction, and, (ii) an incorporation of 3D energy mapping of glucose migration to localize energy favorable clusters. Figure 8 shows that XylE and GLUT1 possess conversely energy favorable glucose migration patterns. This is strictly caused by the structural conformation of these transporters since both are occluded by the extracellular TMs, but differ in the conformation of intracellular α-helices (closed for XylE and open for GLUT1). The open-inward conformation of GLUT1 is Fig. 7 Protein side chain dynamics during glucose induced-fit docking of XylE, GLUT1 and S. mansoni glucose transporters. The line graphs depict the conformational changes of protein side chains in proximity to glucose for glucose transporters GLUT1 (red), XylE (green), SGTP1 (blue), SGTP2 (magenta), SGTP3 (cyan) and SGTP4 (brown). The accepted PELE steps (x-axis) for all simulations were used to calculate the root mean square deviation (RMSD; given in Å) of side chain dynamics (y-axis)see Methods. Next to each line graph are the respective initial positions for each residue (sticks) of GLUT1 (red) and XylE (green) crystal structures with the number indicating the parent TM α-helix (tubes). The title of each line graph is color-coded respectively to GLUT1 and XylE. The boxed-in panel is the negative control since it does not directly affect in substrate binding [13]. (Note: The graphical structural representations are a bird's-eye view of GLUT1 and XylE.) the characteristic that permits the diffusion of glucose within the cell. The distinct migration patterns and energy signatures of S. mansoni glucose transporters, however, may explain their individual response to glucose (Fig. 8).
Although SGTP2 and SGTP3 are structurally similar to GLUT1, the energy signatures for both are not comparable to GLUT1. The energy signature difference between SGTP2 and SGTP3 may be due to their slight structural deviations (as mentioned above and shown in Fig. 6). Functional studies showed that SGTP2 does not transport glucose, even at 10 mM concentration [4]. Our transcriptional analysis shows that SGTP2 and SGTP3 have opposing responses to glucose compared with SGTP1 and SGTP4 (Fig. 1), and phylogenetically group together (Fig. 2). These functional and evolutionary insights may explain glucose migration for SGTP2 and SGTP3 (Fig. 8), since it migrates from an energy favorable state (i.e., the active site) to an unfavorable energy state (i.e., towards the intracellular compartment). Our results showed that transcriptional regulation of SGTP2 and SGTP3 respond differently to maltose and galactose than do SGTP1 and SGTP4 (Fig. 1). All these results suggest that glucose is not the native substrate for SGTP2 and SGTP3, but may act as an inhibitor (as in XylE) since both show energy favorable states within the active site (Fig. 8).
Glucose affinity is higher for SGTP1 (K m = 1.3 mM) than for SGTP4 (K m = 2 mM) [4] and this affinity is Fig. 8 Biophysics of glucose migration from the active sites of XylE, GLUT1 and S. mansoni glucose transporters. All 3D plots show the mapping of glucose binding energy (kcal/mol; z-axis) compared to the triangulation of the glucose migratory distance (given in Å) from the Tyr292/298 homologous residue (x-axis) and the active site (y-axis) for GLUT1, XylE, SGTP1, SGTP2, SGTP3 and SGTP4. The spectrum indicates the energy (kcal/mol) from a favorable glucose binding state (decreasing value; violet-blue) to an unfavorable binding state (increasing value; red-yellow). The contour base on the Cartesian coordinates is a 2D representation of the topology for the binding energy map. Next to each 3D plot are the graphical structural representations from the cluster analysis (see Methods) depicting glucose migration from favorable binding energies (blue) to unfavorable ones (red). One oxygen atom (O5) was used to represent each glucose migration step away from the active site. The final position of Tyr292/298 is shown for each structure and each are color-coded from the N-ter (red) to the C-ter (blue). The lines indicating the position of the membrane is an approximation based on the position in Fig. 5 only for demonstration purposes clearly demonstrated in Fig. 8. For SGTP4 glucose migrates from an energy favorable state (i.e., the active site) to an unfavorable energy state (i.e., towards the extracellular compartment). In contrast, glucose spends the majority of the simulation within the active site of SGTP1 with unfavorable energy states. We do see, however, that a smaller glucose cluster migrates towards the intracellular compartment with energy favorable states (Fig. 8). This may be due to the similarity in structural conformation between SGTP1 and XylE (with closed intracellular α-helices), thereby impeding its diffusion by trapping glucose within the active site. The fact that SGTP1 differs from XylE and SGTP4 in energy and migration patterns may be due to the fact that SGTP1 possesses a higher affinity for glucose [4] and that xylose, not glucose, is the native substrate for XylE [13].

Conclusions
SGTP2/3, not SGTP1/4, are relatively close to human GLUT1/4. This conclusion is based on our observations. The phylogenetic analysis shows that SGTP1/4 belong to a Platyhelminthes-specific glucose transporter class, while SGTP2/3 clearly belong to class I glucose transporters together with vertebrates and insect glucose transporters. Both SGTP2 and SGTP3 show a genome organization similar to that of GLUT1/4 and different to SGTP1/4. SGTP2 seems to be the ancestral glucose transporter in S. mansoni and GLUT1 is a structural homolog of SGTP2/3. Thus, the capacity of SGTP1/4 to transport glucose may have evolved independently, while SGTP2/3 apparently lost this capacity after the divergence from the last common ancestor with the other glucose transporters of class I. Overall, these observations may permit the development of specific inhibitors for S. mansoni glucose transporters.
A recent review [8] suggested that blocking glucose uptake, as the first step of glucose metabolism, represents an appealing strategy for developing drugs to produce starvation, energy supply deficiency, and mortality in parasitic worms. This is particularly necessary since praziquantel is currently the only treatment available for schistosomiasis and its massive use, notably in sub-Saharan Africa, has led to concerns about the potential for the development of resistance [41]. Glucosides, plant-derived compounds with glucose bound to a functional, non-sugar group (aglycone), have been shown to inhibit glucose diffusion in S. mansoni [4,42] and could therefore represent lead compounds for developing drugs targeting glucose transport. With the advent of computational biology, the ability to rationally design drugs for specific protein targets has greatly advanced. For example, using the S. mansoni glucose transporter structures, the aglycone could be altered to computationally test for stability and affinity of the designed glucoside(s). The top glucoside candidate(s) could then be experimentally analyzed for effectiveness. Since glucosides are non-transportable [43] they could "trap" themselves in the active site to maintain the structural conformation similar to SGTP1 in Fig. 8i.e., the glucose-bound, occluded conformation of XylE. One of the caveats of the PELE algorithm is that it cannot perform large protein conformational changes, as those described by Deng et al. [12]. Such simulations to rationally design drugs will benefit from other molecular dynamics software such as Desmond [44].

Parasite material
A Puerto Rican strain of S. mansoni was maintained in the laboratory using the snail, Biomphalaria glabrata as its intermediate host and the golden hamster, Mesocricetus auratus as a definitive host. Schistosomula were obtained by mechanical transformation from cercariae released from infected snails as previously reported [45]. Schistosomula were washed three times in serum and glucose-free DMEM medium (Life sciences, Carlsbad, USA) supplemented with traces of glucose (0.05 mM) then incubated at 37°C, in an atmosphere of 5 % CO 2 . After three hours of incubation, glucose, maltose or galactose were added to the medium at a concentration of 10 mM each, and the schistosomula were incubated for another eight hours. A total of 10 000 parasites were used per condition and experiments were carried out in triplicate.

RNA isolation and quantitative RT-PCR
Total RNA was isolated from the different stages of S. mansoni with TRIzol reagent (Invitrogen, Carlsbad, USA) according to the manufacturer's instructions. Complementary DNAs (cDNA) were obtained by reverse transcription of total RNA using the Thermoscript RT-PCR System (Invitrogen, Carlsbad, USA). The cDNAs were then used as templates in triplicate assays for RT-PCR amplification using the KAPA SYBR FAST ABI Prism kit (Kapa Biosystems, Boston, USA), and ABI PRISM 7000 sequence detection system. We used previously reported primers for S. mansoni sgtp1 and sgtp4 [5]. The primers for sgtp2 and sgtp3 were designed for this study: sgtp2F 5' TTTACCTTCGAGGGCAAGAT 3' and sgtp2R 5' CACCGCAAGTATGGAATACG 3' , sgtp3F 5' GCAGCAACTCTCAGGAATCA 3' and sgtp3R 5' ACACAATAACCGCTCCAACC 3'. The ratios of relative expression were calculated using the 2 -ΔΔCt ratio [46] with S. mansoni α-tubulin as the endogenous control gene [47]. The statistical significance between groups was evaluated using the unpaired non-parametric Mann Whitney's test in the GraphPad 6 Prism program (GraphPad Software Inc.). Differences were considered significant when p-value <0.05.

Phylogenetic analyses
Sequences were collected from GenBank for phylogenetic analyses of sugar transporter proteins from mammals (3 species), fish (teleosts, 7 species), insects (8 species), cestodes (6 species) and flukes (4 species). A Drosophila melanogaster sucrose transporter was introduced as an outgroup. The species, sugar specificity and accession numbers of each sequence is provided in Fig. 2. Glucose transporter amino acid sequences were aligned using MUSCLE (v3.7) configured for highest accuracy [48] and non-aligned regions were removed with Gblocks (v 0.91b) [49]. Thus, the final alignment contained 333 gap-free amino acid positions. The best-fit model of the sequence evolution was selected based on Akaike Information Criterion (AIC), Corrected Akaike Information Criterion (cAIC) and Bayesian Information Criterion (BIC) implemented in Datamonkey [50]. The LG [51] model, which had the lowest values of AIC, cAIC and BIC, was chosen for subsequent phylogenetic analyses. Maximum likelihood (ML), maximum parsimony (MP) and neighbor joining (NJ) methods, implemented in MEGA 6 [52], were used to obtain the best tree topologies for each method. A proportion of Gamma distributed and invariants sites (G + I) were estimated in MEGA for each phylogenetic method. Reliability of internal branches was assessed using the bootstrapping method (1 000 bootstrap replicates) and the approximate likelihood ratio test (aLRT -SH-Like) implemented in MEGA and PhyML, respectively [52,53]. Graphical representation and editing of the phylogenetic tree was performed with EvolView [54].

Molecular clock and evolutionary rates
First, the likelihood ratio test (LRT) [55] was used to test the molecular clock hypothesis (i.e., all tips of the tree are equidistant from the root of the tree) on the aforementioned phylogenetic trees [52]. The LRT rejected the null hypothesis of equal evolutionary rate throughout the tree at a 5 % significance level (p-value = 0) for both topologies (ML and MP). Therefore, relative time divergence was determined [56]. Time trees were generated using the RelTime method [22], as implemented in MEGA 6 [52]. RelTime is a useful method to estimate relative lineage-specific evolutionary rates and relative divergence times without requiring the pre-specification of statistical distribution of lineage rates and clock calibrations [22]. To calculate evolutionary rates, RelTime assumes that the elapsed time of two sister lineages from their most recent common ancestor is equal [22]. Divergence times for all branching points were calculated using the Maximum Likelihood method based on the LG model. The MP tree(s) were used to estimate the relative divergence times of glucose transporters in different taxa with special emphasis on S. mansoni. To account for statistical errors in our analysis, the 100 most parsimonious trees were searched using the Min-mini heuristic algorithm [57], and the relative divergence times of the selected nodes was determined in all the trees. The Shapiro-Wilk normality test [58] rejected normal distribution of the obtained relative divergence times (P <0.0001), therefore a paired Wilcoxon test was used to test whether the difference between divergence times of selected nodes were significant [56]. The GraphPad 6 Prism programme (GraphPad Software Inc.) was used to perform both Shapiro-Wilk and Wilcoxon tests. Evolutionary rates (measured as the relative number of amino acid substitutions per site) were calculated using the RelTime method [22], implemented in MEGA 6 [52]. To further compare the evolutionary rates of SGTP3/SGTP2 and SGTP4/SGTP1 we used Tajima's relative rate test [23]. This method tests whether two sequences have equal rates of evolution using a third sequence as outgroup. The equality is tested using the chi-square (χ2) test. When the observed χ2 is significantly higher (p-value <0.05) than expected, the null hypothesis of equal rates can be rejected. Thus, to perform Tajima's relative rate test, Fhep.-G.AAS94013 and Tsol.G2.AAB05920 sequences (see Fig. 2) were used as outgroups to compare the rates of SGTP3/SGTP2 and SGTP4/SGTP1, respectively.
The top candidate model structures were refined via minimization to remove steric clashes and optimization of the hydrogen-bond network by means of side chain sampling using the Schrödinger's Maestro Protein Preparation Wizard [69]. Briefly, the Protein Preparation Wizard analyzes the structure to build a cluster of hydrogen bonds and with the highest degree of sampling, the algorithm then performs 100 000 Monte Carlo orientations for each cluster. Based on the electrostatic and geometric scoring functions, the algorithm then determines an optimized structure. Glucose and the other glucose transporter crystal structures (see below) were also prepared and optimized in this manner.

Glucose induced-fit docking and migration
For the biophysical simulations, namely induced-fit docking and substrate migration, we used the state-of-the-art Protein Energy Landscape Exploration server (PELE) [70]. The PELE server provides ready-made scripts for substrate binding refinement (induced-fit docking) and unconstraint substrate binding site search (migration). The many uses of PELE can be accessed at https:// pele.bsc.es/pele.wt and its algorithm is thoroughly explained in references [70,71]. Briefly, the PELE algorithm performs three stages. First, localized substrate perturbations are performed. Protein perturbations of the αcarbon backbone are also performed using an anisotropic network model (ANM) [72]. Second, the amino acid side chains in proximity to the substrate are optimized using steric filters and a rotamer library [73]. Finally, a truncated Newton minimizer and a surface generalized Born implicit solvent for minimization [74] is used to achieve a local minimum after the initial perturbation. These three stages are performed for a number of steps and are analyzed in parallel with several computer-processing units. The results are a series of trajectories that represent protein side chain conformational changes and substrate migration. Based on the calculated energies of each step, a Monte Carlo Metropolis criterion implemented in PELE either accepts (if they are equal to and/or less than the initial energy) or rejects the steps (greater than the initial energy) [71]. The energy is calculated by using a standard force field to describe the potential energy of a particular molecular system, known as the optimized potentials for liquid simulations (OPLS-2005) [75].
Several parameters were optimized from the readymade scripts provided by the PELE server to facilitate sampling. For the substrate binding refinement (induced-fit docking), (i) steric_tr ('100') and tries ('25') were reduced for the simulations to be less computationally expensive, (ii) translation (tra_r '2') and rotation (rot_r '0.3') of glucose were increased to permit more exploration of the active site, (iii) minimization radius (mirad) to allow minimization of the whole system, (iv) an ANM type to '4' was added and ANM mode changed to '5' for favorable protein perturbations; and, (v) an increased number of steps (1000) was used for sufficient overall sampling. For the unconstrained substrate binding site search (migration), (i) waitfor was increased to '4' to allow sufficient sampling of the active site, (ii) translation (tra_r) was reduced to '4' to avoid large translation of glucose, (iii) an ANM type to '4' was added since the default ANM mode produced favorable protein perturbations; (iv) an increased number of steps (1000) was used for sufficient overall sampling; and, (v) a focused ANM (anmrad 15 Å) and minimization (mirad 10 Å) in proximity of the substrate was used for localized interpretation of glucose transport.