Comparative EST analysis provides insights into the basal aquatic fungus Blastocladiella emersonii

Background Blastocladiella emersonii is an aquatic fungus of the Chytridiomycete class, which is at the base of the fungal phylogenetic tree. In this sense, some ancestral characteristics of fungi and animals or fungi and plants could have been retained in this aquatic fungus and lost in members of late-diverging fungal species. To identify in B. emersonii sequences associated with these ancestral characteristics two approaches were followed: (1) a large-scale comparative analysis between putative unigene sequences (uniseqs) from B. emersonii and three databases constructed ad hoc with fungal proteins, animal proteins and plant unigenes deposited in Genbank, and (2) a pairwise comparison between B. emersonii full-length cDNA sequences and their putative orthologues in the ascomycete Neurospora crassa and the basidiomycete Ustilago maydis. Results Comparative analyses of B. emersonii uniseqs with fungi, animal and plant databases through the two approaches mentioned above produced 166 B. emersonii sequences, which were identified as putatively absent from other fungi or not previously described. Through these approaches we found: (1) possible orthologues of genes previously identified as specific to animals and/or plants, and (2) genes conserved in fungi, but with a large difference in divergence rate in B. emersonii. Among these sequences, we observed cDNAs encoding enzymes from coenzyme B12-dependent propionyl-CoA pathway, a metabolic route not previously described in fungi, and validated their expression in Northern blots. Conclusion Using two different approaches involving comparative sequence analyses, we could identify sequences from the early-diverging fungus B. emersonii previously considered specific to animals or plants, and highly divergent sequences from the same fungus relative to other fungi.


Background
Since the sequencing of the first complete fungal genome, the budding yeast Saccharomyces cerevisiae [1], fungal genomics and the specific area of comparative genome analysis in fungi have experienced a recent but impressive advance. Following sequencing of the genomes of two other ascomycetes, Schyzosaccharomyces pombe and Neu-rospora crassa [2,3], efforts have focused on species throughout the fungal kingdom that represent diverse scientific interests. In this sense, genomes from plant and human pathogenic basidiomycetes have been sequenced [4,5] and there are drafts or genome projects in progress of other fungi. Likewise, the sequencing of one zygomycete genome has been completed and there are two chytrid genome projects underway (see [6] for an overview of fungal genome sequencing projects).
Expressed sequence tag (EST) data from fungi, even though less numerous than genome sequences, are also showing to be useful to specific and diverse aims, such as mapping previously characterized genes [7], investigation of patterns of fungal genome evolution [8], prediction of novel genes [9], prediction of pathogenicity determinants [10], identification of disease-related sequences [11], improvement of functional assignments [12], and identification of alternatively spliced mRNA species [13].
Recently, we reported a sequencing program of nearly 17,000 ESTs corresponding to different developmental stages of Blastocladiella emersonii life cycle, an early diverging fungus that belongs to the Chytridiomycete class [14,15]. Approximately 52% of the uniseqs presented similarity to sequences deposited in public data banks. Interestingly, several of these ESTs revealed similarity with known genes not previously reported in fungi, and which had been recognized as animal or plant specific proteins. Despite the fact that a consensus phylogenetic tree seems to resolve fungi and animals as sister groups [16], we wondered if some ancestral characteristics of fungi and animals or of fungi and plants could have been retained in this basal fungus and have been lost or become highly divergent in members of the late-branching group of fungi.
Our previous study provided the functional identification of putative unique transcripts based on sequence comparison, and contributed to linking in silico expression profile data with previous information about biological processes occurring throughout the fungal life cycle [14]. In this sense, the survey increased the knowledge about this interesting biological model. However, our approach did not provide a direct link between expressed sequences in B. emersonii and gene expression in major groups, like animals and plants.
In the present work, we carried out a large-scale comparative analysis of B. emersonii ESTs against protein and transcript sequences of fungi, animals and plants, using databases constructed ad hoc. Our goals were to identify putative orthologues in B. emersonii of genes previously classified as specific to animals and/or plants, as well as to find B. emersonii sequences common to fungi but which have evolved at a lower rate in this chytrid than in latediverging fungi. Based on our results, we discuss possible relationships between expressed sequences and structures and/or biological processes occurring in animals and/or plants and B. emersonii, including a metabolic pathway previously reported only in animals and bacteria.

B. emersonii-animal shared sequences
To uncover sequences shared by B. emersonii and animals, we carried out a comparison between B. emersonii ESTs and an animal database constructed ad hoc, and assigned a putative identification to these sequences (see Methods section below). We then classified B. emersonii sequences that matched with animal data (named B. emersonii-animal shared sequences) as follows: hits only found in animals; hits found in animals and protists (including flagellated and ciliated organisms and green algae not filtered as plants on purpose, some hits also included bacteria); hits only found in animals (when using an Evalue ≤ 10 -5 as cut-off) but with protein family members also described in plants and/or fungi; hits found in animals and bacteria ( Table 1).
As the most important result, matches only with animal proteins revealed two consensus sequences encoding enzymes involved in coenzyme B 12 -dependent propionyl-CoA metabolism: DL-methylmalonyl-CoA racemase (EC 5.1.99.1) and methylmalonyl-CoA mutase (EC 5.4.99.2). We also found ESTs encoding the alpha and beta chains of propionyl-CoA carboxylase (EC 6.4.1.3), the enzyme that catalyzes the first step of this metabolic route (Table 1 and Figure 1). These enzymes give the capacity to metabolise propionate through propionyl-CoA and methylmalonyl-CoA in the TCA cycle and they seem to be present in most animal species and prokaryotes [17,18] but there are no sequences or activities described in fungi. Furthermore, methylmalonyl-CoA mutase needs adenosylcobalamin (coenzyme B 12 ) as a cofactor and we wondered whether sequences encoding enzymes involved in biosynthesis of coenzyme B 12 would be expressed in B. emersonii. Interestingly, we found another assembled sequence, among the matches with animal and bacteria proteins, encoding an ATP:Cob(I)alamin adenosyltransferase (EC 2.5.1.17), the enzyme that catalyses the last step of coenzyme B 12 biosynthesis. Afterwards, we proceeded to do an experimental validation of the expression of these sequences in the fungus. As shown in Northern blot assays (Figure 2A-H), these genes are expressed during B. emersonii sporulation. In addition, as cobalt is necessary for the pathway to function, we also evaluated expression of these genes in cells exposed to cobalt and all four genes appeared to be induced by this cation (Figure 2E,H).
Moreover, approximately 10% of the hits obtained through this approach were related to flagella and significant alignments appeared with animals and Chlamydomonas reinhardti, suggesting that these sequences are conserved among different organisms. Finally, 13% of the matches included animals and euglenozoos, as Trypanosoma cruzi, and 10% included the ciliated Tetrahymena thermophila. A small proportion of two hits revealed  Scheme of the pathway of cobalamin-dependent propionyl-CoA metabolism Figure 1 Scheme of the pathway of cobalamin-dependent propionyl-CoA metabolism. Enzymes mentioned in the text and in Table 1 are underlined. propionyl-CoA sequences that retrieved significant matches only with animal proteins, but which belong to protein families with members among plants and fungi. However, both of them are not well characterized.

B. emersonii-plant shared sequences
After removal of contaminants, as carried out for B. emersonii-animal shared sequences, we classified the hits found against the plant database (B. emersonii-plant shared sequences) as follows: hits only found in plants; hits found in plants and protists; hits only found in plants but with protein family members also described in animals and/or fungi; hits found in plants and bacteria ( Table 2).
The first important difference observed in B. emersonii-animal and B. emersonii-plant sequence comparison was the number of uniseqs with matches in each group: the number of matches with plants was one fifth of the number obtained with animal proteins (20 vs. 105).
However, some noteworthy information could be obtained. Three putative protein receptors: a phytosulfokine receptor, an ethylene receptor CS_ETR2 and a protein kinase receptor (the first two mentioned in [14]), which are plant receptors not previously found among fungi, were found in B. emersonii.
On the other hand, two B. emersonii assembled sequences presented significant matches only with plants but encode proteins that belong to families with members in animals and fungi. One of them encodes a putative Isp4 protein, which represents a family of transporters of small oligopeptides (OPT family), initially characterized only in three different species of yeast [19][20][21]. A set of related proteins from Arabidopsis thaliana, characterized as oligopeptide transporters, was later described as an outgroup to the yeast set by neighbor joining analysis [22]. The B. emersonii assembled sequence aligns with a significant score only to sequences of the plant OPT family and not Northern blot analysis of B. emersonii genes encoding enzymes involved in propionyl-CoA metabolism to the fungal sequences. Nevertheless, although the assembled sequence presents the conserved regions characteristic of the protein family of both animals and fungi, the region of alignment comprises less than 60% of the total length of the best matching sequence. Thus, the assignment of a putative function for the protein should be taken with caution.
The other assembled sequence matched with a member of the syntaxin family of soluble N-ethylmaleimide-sensitive factor adaptor protein receptors (SNAREs) superfamily, which is known to play an important role in the fusion of transport vesicles with specific organelles [23]. In a general sense, animals and plants have syntaxins that are orthologues to one of the yeast syntaxins. However, whereas some classes of yeast and mammalian syntaxin genes appear to be absent in Arabidopsis, its genome presents syntaxin gene families not found in other eukaryotes [24]. The SYP7 family (with three members) does not appear to have an ortholog among yeast or animal syntaxins, and this group may be unique to plants. The B. emersonii assembled sequence mentioned above matched with a putative syntaxin 71, a member of SYP7 family. We also looked for B. emersonii ESTs encoding other syntaxin fam-ily members and found representatives for all except one (SYP8) of the families categorized according to Sanderfoot et al. [24] (Table 3).

B. emersonii-animal-plant shared sequences
Following the same procedure carried out for the two previous analyses, we classified the hits found against the animal and plant databases (B. emersonii-animal-plant shared sequences) as follows: hits found in animals and plants; hits found in animals, plants and protists (some hits also included bacteria); hits found in animals and plants but with protein family members also described in fungi; hits found in animals, plants and bacteria (some hits also included protists) ( Table 4).
A noteworthy identification was a putative urocanate hydratase, urocanase or imidazolone-propionate hydrolase (EC 4.2.1.49), the second enzyme involved in the catabolism of histidine by conversion of this amino acid to glutamate [25]. We also found an EST encoding an imidazolonepropionase (EC 3.5.2.7), among the matches with animal and bacteria proteins, the third enzyme in the same pathway. The first enzyme of the pathway is the histidase or histidine-ammonia lyase, which converts his- tidine into urocanate, the substrate of the urocanase. Urocanase has been found in bacteria, in the liver of mammals, in the land plant white clover, and also in Chlamydomonas reinhardtii (see [26] and ref. therein; [27]). This activity is probably present in protists and other plants as Medicago sativa, according to sequences deposited in Genbank protein database. In bacteria, the degradation of histidine to glutamate provides the organism with a source of carbon and nitrogen (see [28] and ref. therein). Fungi apparently lack urocanase activity, as revealed by the absence of genes encoding the enzyme in fungal sequence resources. The enzyme activity has been specifically searched in Aspergillus nidulans [28]. This fungus synthesizes an active histidase enzyme but cannot use histidine as the sole carbon source, which has been attributable to the lack of an active urocanase; histidine is quantitatively converted to urocanate, which accumulates in the extracellular medium.
Among the sequences recovered, we also observed a type C fructose-bisphosphate aldolase (FBA). This type of enzyme belongs to the Class I aldolase family, whose members have been observed mainly in higher eukaryotes. Fungi FBAs belong to the Class II aldolase family, presenting little similarity with proteins from Class I (Rutter, 1964 in [29]). In addition, we did not find another FBA in B. emersonii transcript database. We also identified a novel sequence not previously identified in fungi: a singlet encoding a gamma-SNAP protein (soluble N-ethylmaleimide-sensitive factor-attachment protein). Whereas alpha-SNAP homologues have been identified in yeast, plant, mollusk and insect cells, gamma-SNAP homologues have been found only in mammals, plants and more recently Dictyostelium discoideum [30]. In addition, a cDNA encoding an alpha-SNAP homologue was also observed in B. emersonii database, showing that this fungus has the two different types of SNAP proteins.
Finally, among B. emersonii-animal, B. emersonii-plant and B. emersonii-animal-plant shared sequences, the highest percentage of matches (approximately 65%) was achieved for sequences encoding proteins classified in unknown processes. In fact, the functional characterization of these sequences remains one of the most important challenges in post-transcriptome research.

Sequence divergence comparison between B. emersonii and N. crassa or U. maydis
We also carried out a comparative analysis to identify B. emersonii putative genes with a higher degree of similarity to animal or plant genes than to their fungal counterparts. The S' values, obtained for pairs of putative orthologues from N. crassa/U. maydis and B. emersonii, were plotted with their best matches in animal or plant sequences. Pairs of hits with highest differences in S' values in two or more comparisons were chosen for further investigation ( Figure 3). Four apparently divergent sequences were identified and three of these were, unexpectedly, more divergent in B. emersonii than in N. crassa and U. maydis ( Table 5). None of the four sequences appeared related by biological process, function or localization. In addition, Rbj-like or Rjl proteins are members of Ras-related GTPbinding proteins. Rjl proteins have recently been identified as a new family, independent of the Rab family, to which they were initially linked [31]. There is no evidence for a role for these proteins in other organisms, except chordates [31]. As no Rjl sequences were identified in other fungi, we looked for family signatures in B. emersonii deduced protein sequence, and we also checked N. crassa, animal and plant data, which had been collected as orthologues.
In the putative protein from B. emersonii, four of the family characteristics identified by Nepomuceno-Silva et al. [31] were observed: 1) the substitution of the canonical glutamine residue in the third GTP binding domain; 2) the alteration of the DTAGQE motif to DMAGDR (it is the first motif with E to R substitution); 3) the percent identity with other Rjl proteins (between 37 and 40%), with only one exception; 4) the absence of a prenylation motif. Using a hidden Markov Model [32,33], a signal peptide prediction was made but with low probability (58 %), and no signal anchor was predicted, as is expected for Rjl proteins.
The apparent N. crassa ortholog [Genbank: EAA33910] and its best matches among animal and plant data were GTPases from the Rab family. Likewise, the best match of B. emersonii Rjl in the plant database was a Rab protein [Genbank: AK062838], in agreement with the absence of Rjl records in land plants. In contrast, when compared Pairwise score comparison between fungal orthologues and animal and plant sequences Figure 3 Pairwise score comparison between fungal orthologues and animal and plant sequences. Black dots on the plots represent score-pairs with a difference in bit score equal or higher than 150.  The elongation factor 1-α (EF1-α), a core member of the protein biosynthesis machinery, is ubiquitous in eukaryotes and in prokaryotes, where it is named EF-Tu [34]. B. emersonii presents an EF-like or EFL protein, which is different from the canonical EF1-α identified in the majority of the organisms [14]. Due to this fact, we expected to sample the B. emersonii divergent sequence encoding the EFL protein during this procedure. Although EFL and EF1α probably perform similar roles, they are clearly different proteins, and EFL proteins form a completely separated branch in molecular phylogenies. Moreover, taxa genomes with EF1-α lack EFL, suggesting that EFL has replaced eEF-1 α several times independently [34]. However to our surprise, a first data processing revealed no significant S' difference (ΔS') between the pair B. emersonii EFL/plant protein match, and N. crassa/U. maydis EF1-α/ plant protein matches, even though these two fungi present the canonical form of EF1-α. The explanation for this unexpected result is that Oryza sativa genome apparently contains two different genes [Genbank: AK110624 and AK107366], one that matched with the sequence encoding B. emersonii EFL, and another that matched with the fungal N. crassa and U. maydis EF1-α. In addition, O. sativa genome also presents a third gene encoding the canonical EF1-α [Genbank: AK103738] usually found in plants. The first two rice sequences do not seem to be contaminant products, as no positive results were obtained using blastn against Genbank non-redundant or dbEST databases. Altogether, O. sativa genome seems to contain three genes encoding divergent EF1-α. Whether or not all three sequences actually represent rice genes requires clarification.
Another assembled sequence shown to be divergent in B. emersonii encodes a mitochondrial ADP/ATP translocase. This translocase, also known as a mitochondrial adenine nucleotide translocator (ANT), catalyses the exchange of ATP and ADP between mitochondria and cytosol, and seems to participate in mitochondrial events that control cell death (see [35]    Finally, we found sequences encoding a putative 2-amino-3-carboxymuconate-6-semialdehyde decarboxylase or ACMSD (EC 4.1.1.45), an enzyme involved in the tryptophan-niacin pathway in eukaryotes. There are few eukaryotic and even fewer prokaryotic sequences known. Fungal sequences are poorly characterized, and they are apparently absent from plants. Moreover, we could observe that some bacterial sequences diverged with the eukaryotic counterparts in a molecular phylogeny constructed with the same method used for the putative ANT (data not shown). This observation suggests the possibility of lateral gene transfer, as proposed by Muraki et al. [36]. B. emersonii deduced protein appears to have the two conserved motifs described for ACMSD proteins by the same authors, even though the motifs possess clear differences from those found in fungal homologues.

Discussion
We carried out a comprehensive comparative EST analysis that identified one hundred and sixty-six expressed sequences from the aquatic fungus B. emersonii encoding putative proteins not previously reported in other fungi.
Among the ESTs with significant similarity to animal sequences, we found assembled sequences encoding enzymes involved in coenzyme B 12 -dependent propionyl-CoA metabolism. Propionate, the second most abundant fatty acid in soil, is formed by fermentative processes from carbohydrates and several amino acids [37]. Propionate is converted to propionyl-CoA, which is also formed by oxidation of odd-chain fatty acids and several amino acids, and then converted to succinyl-CoA that then enters the central metabolism. This pathway is used in diverse metabolic processes and homologues of intervening enzymes were found within archaeal, bacterial and eukaryal genomes, but not in plants or fungi. In mammals, this route is employed in the catabolism of valine, isoleucine, methionine, threonine, thymine, cholesterol, as well as odd-chain fatty acids [38], and defects in some of the enzymes involved lead to the rare but severe inherited disease methylmalonyl aciduria [17].
Propionate is generally toxic to fungi and bacteria, and this is the reason why it is widely used as a preservative [39]. Despite its toxicity, many bacteria and fungi are able to use propionate as carbon and energy sources under aerobic conditions, using an alternative pathway to that mentioned above, the "methyl citrate cycle" that catalyses the oxidation of propionate to pyruvate.
Fungi such as S. cerevisiae and Aspergillus nidulans seem to lack cobalamin-dependent functions and therefore cannot use the methyl-malonyl-CoA pathway [18]. Leadley et al. [18] suggested that the maintenance of this pathway in proto-eukaryotes would have meant a high evolutionary cost, due to the need to preserve also the enzymes capable of producing coenzyme B 12 , and at the same time, the existence of other pathways for propionate utilization may have superceded the selective pressure for preserving this metabolic route.
The question is why B. emersonii would express coenzyme B 12 -dependent enzymes under the conditions tested. The presence in archaea, eubacteria and animals of coenzyme B 12 and coenzyme B 12 -dependent enzymes seems to indicate that the conservation of these functions is important to diverse processes and this principle can also be applied to fungi. Despite the absence of genes encoding cobalamin-dependent proteins and enzymes of coenzyme B 12 biosynthesis in the fungal genomes sequenced, pathways involving this coenzyme could be active under conditions not frequently tested in other fungi whose genomes have not been sequenced yet. In fact, some of B. emersonii ESTs encoding these enzymes were isolated from a cDNA library constructed with mRNA isolated from cells exposed to high concentrations of cadmium. Differently from cadmium, several transition metals, such as cobalt, play a role as catalysts in a variety of enzymatic reactions. These metals, which are normally useful to the cells, can be toxic when in excess. Thus, many molecular mechanisms for cell detoxification have been developed. Some of these mechanisms are promiscuous, being responsible for detoxification of more than one of these heavy metals [40]. In this sense, we cannot rule out the possibility that some B. emersonii genes induced by exposure to cadmium could be involved in cobalt metabolism.
Our analysis has also shown eleven sequences associated to flagella-related proteins expressed in B. emersonii and animals, nine of them also present in green algae. The absence of these sequences from fungi and plant databases was expected, as a consequence of the bias in the most investigated species, which mainly belong to latediverging fungi and land plants. Thus, B. emersonii could be a good model to study processes related to flagella structure and movement, probably contributing to the characterization of differences between animals and other flagellated cells.
Comparison of B. emersonii ESTs with plant sequences revealed two assembled sequences with high similarity to genes found only in plants, but encoding proteins that belong to families with members also in animals and fungi: a putative Isp4 protein (an oligopeptide transporter) and a putative syntaxin 71 (a member of SYP7 family of protein receptors). Oligopeptides can be used as source of amino acids, nitrogen and carbon, and their transporters have been documented in bacteria, fungi and plants. The identification of multiple OPTs in Arabidopsis, with tissue-specific expression patterns, supports the idea of different functional roles for these transporters, e. g., regulators of hormone activity in hormone-peptide conjugates [22]. There is also evidence indicating that members of other peptide transporter family, the PTR, have a role in plant growth and development. Thus, it is possible that the putative OPT found in B. emersonii has a specific function, different from those described in other fungi, as regulation of growth and differentiation.
In this same context, we can include the matches of B. emersonii ESTs with two plant receptors associated with the control of proliferation and development in plants.
Even though the alignments extend over a conserved region in the plant sequences, domains characteristic of the assignments do not overlap. Consequently, B. emersonii proteins could be involved in completely different processes. Further studies will be necessary to clarify this hypothesis.
The syntaxin family of proteins is well represented in B. emersonii transcriptome. We found representatives for all except one (SYP8) of the defined families [24] ( Table 3). The group included syntaxin 71, a member of SYP7 family, which seems exclusive of plants. Such broad representation suggests that syntaxin 71 could have specific functions in B. emersonii, perhaps related to functions developed in plants.
Members of the syntaxin family are known to play an important role in the fusion of transport vesicles with specific organelles [23], and specifically SYP7 proteins seem to be involved in transport between the ER and the Golgi apparatus [41]. Interestingly, membrane transport and vesicle rearrangement have critical importance during the sporulation stage of B. emersonii life cycle [42], and several ESTs related to this function were exclusively isolated from sporulating cells (see [43] GO:0006886 intracellular protein transport), which includes the EST encoding the possible syntaxin 71.
B. emersonii-animal-plant common sequences included an urocanase, an enzyme with no records in sequenced fungal genomes and which could be indicative of B. emersonii's ability to use histidine as a carbon source. The presence of sequences encoding enzymes possibly involved in the catabolism of valine, isoleucine, methionine and threonine (such as the enzymes that are active in coenzyme B12-dependent propionyl-CoA metabolism), and enzymes possibly involved in the catabolism of histidine, suggest that B. emersonii metabolism might be directed towards amino acid catabolism, as a source of carbon. Early studies in chytrids indicated distinct roles for some amino acids, other than serving as nitrogen source or protein building blocks. For instance, certain amino acids have been shown to be effective in initiating growth on sugars different from glucose, such as mannose and fructose, in Allomyces macrogynus cultures, presumably supplying both carbon and nitrogen sources [44].
An assembled sequence encoding a FBA type C, a member of class I FBA, was also found in our analyses. FBAs are divided into two non-related protein classes: Class I FBA, not found in fungi but with widespread distribution in other eukaryotes and also found in prokaryotes, and Class II FBA, identified mainly in eubacteria and also in eukaryotes, including fungi [29,45]. Although the scattered taxonomic distribution of FBA classes does not have a consensual evolutionary explanation yet, gene duplication events and replacement of one paralog by the other are events that could have occurred. For instance, there is some evidence for the existence of an ancestral class II aldolase, from the endosymbiosis with a cyanobacterium, which could have been replaced by a class I aldolase in red and green algae, as well as in higher plants [29,46]. Class II FBA genes of ascomycetes are also of eubacterial origin, and probably consequence of endosymbiosis with mitochondria ancestors [47]. Thus, a gene replacement event such as the proposed for red and green algae and land plants could similarly be proposed for the origin of B. emersonii FBA gene.
Even more interesting than the presence of a member of class I FBA in B. emersonii, could be the type observed, the C type, which is supposed to have evolved after divergence of the B type [48]. In fact, no B type FBA sequences were observed among B. emersonii ESTs. One possible explanation would be that the C type FBA could have replaced the B type. Another explanation, perhaps the simplest one, is that the B type sequence was not found among B. emersonii sequenced ESTs, but the gene is present in the genome. However, why both types of FBAs would be expressed in B. emersonii is not clear yet. In vertebrates, Class I comprises three types of isozymes expressed in different tissues: aldolase A (muscle type, also expressed in brain), B (liver type) and C (brain type). As described for other enzymes of the glycolytic pathway, aldolases A and C display activities different from that observed during glucose metabolism, as they regulate the stability of the light neurofilament mRNA through their ribonuclease activity [49]. A specialized function for the C type aldolase in B. emersonii should not be ruled out.
In addition, two putative genes encoding the alpha and gamma-SNAPs were observed in B. emersonii. Until now, no gamma-SNAPs have been described in fungi, B. emer-sonii being the first fungus in which this gene has been identified. However, the presence of both alpha and gamma-SNAPs in eukaryotic cells seems to be the rule, with fungi being the exception, considering that five phylogenetically distant species are known to possess both alpha and gamma-SNAP: D. melanogaster, B. taurus, H. sapiens, A. thaliana and D. discoideum [30]. The protein alpha-SNAP is essential for membrane traffic because it allows efficient NSF/SNARE interaction. Instead of this direct function, gamma-SNAP could have a regulatory role in membrane fusion. It was also suggested a role for gamma-SNAP in mitochondrial dynamics, contributing as an adaptor in the attachment of mitochondria to the cytoskeleton [50]. Our results in B. emersonii indicate that D. discoideum is not the only simple eukaryote containing both alpha and gamma-SNAPs.
As a second approach to discover non-typical fungal genes in B. emersonii, we carried out a comparative analysis with the expressed sequences of this aquatic fungus and other fungal sequences. We intended to be conservative at the time of selection and very few sequences were identified. Likewise, several difficulties arose due to the complexity of dealing with large multigene families. Indeed, two of the four selected sequences initially collected were not orthologues, and the relationship between the other two is not evident, but we decided to include these sequences in our analysis because the information extracted was also relevant. In fact, even though the sequences encoding the Rjl protein were not reported in fungi or plants, we did not detect them among the 105 B. emersonii-animalshared sequences selected in our first approach. The divergence found for the other three cases (EF1α, ADP/ATP translocase and aminocarboxymuconate semialdehyde decarboxylase) is also noteworthy, because it could reflect high evolutionary rates, gene duplication and replacement (as suggested for EF1α in [34]), gene conversion, or horizontal gene transfer from prokaryotes to eukaryotes or among eukaryotes, which seems to be more common that previously thought (see [31]).
This collection of selected B. emersonii assembled sequences represents the result of approaches that use comparative EST analyses to address differences and similarities between chytrids and other eukaryotes (other fungi, animals and plants). The results of such analyses will probably suffer modifications when more fungal sequences are available. Specifically, other chytrid and zygomycete sequences will contribute to define retained, lost and divergent genes. Moreover, at least part of the borderline sequences (with an Evalue between 10 -3 and 10 -5 against ad hoc fungal database), which could be true divergent homologues, could constitute a group of interest to help understand phylogenetic relationships among fungi.

Conclusion
Through two different approaches involving comparative sequence analyses, and using computational tools and manual revision, we identified 162 protein-coding sequences from B. emersonii previously described in animals (such as coenzyme B 12 -dependent propionyl-CoA pathway members, and proteins related to flagella structure or movement), in plants (such as protein receptors, a putative member of small olipeptide transporter, and a SYP7 family member of syntaxins), and in animals and plants (such as an urocanase, a fructose-bisphosphate aldolase (FBA) type C, members of the MtN3 family and a gamma-SNAP representative). We also found 4 sequences from B. emersonii, which were identified in a fungal sequence comparison as not found or highly divergent from other fungal species: a Rbj-like protein (similar to animal proteins), an EF-like protein (dispersely distributed in taxa, already described in [14]), an ADP/ATP translocase (similar neither to plant nor to animal sequences) and a 2-amino-3-carboxymuconate-6-semialdehyde decarboxylase (different from fungal sequences, poorly characterized). When the selected ESTs were classified according to the biological processes in which they could be involved, cell growth and maintenance, signal transduction and metabolism resulted as the biological processes most represented. Some sequences selected were expected, based on the knowledge about chytrids, like those associated to specific structures not found in other fungi (e. g., flagellar-associated ESTs). Thus, B. emersonii seems to be an interesting model to study flagellaassociated structures or functions.
Among the ESTs exclusively isolated from sporulating cells, we collected sequences associated to membrane transport, such as syntaxin 71. Membrane fusion and vesicle rearrangement are crucial events in B. emersonii sporulation, when cytokinesis occurs. A set of core SNAREs is apparently sufficient to mediate most intracellular vesicle fusion events, although multicellular organisms would express additional SNARE proteins for specific functions associated with the body complexity [51]. Thus, proteins like syntaxin 71 are good candidates to function as additional SNARE proteins in the transition of unicellular multinucleated zoosporangia to zoospores during B. emersonii life cycle.
Other collected ESTs were unexpected, like those involved in specific metabolic pathways, such as sequences involved in conversion of propionate and histidine to glutamate. We hypothesize that alternative pathways leading to the use of amino acids and other substrates as carbon and nitrogen sources could have been lost in latediverging fungi and retained in basal fungi.
Finally, a large number of sequences selected by the first approach were not classified in a known process, which suggests that other structures or biological processes not identified yet can be shared by B. emersonii, animals and plants.

B. emersonii EST database
All the information concerning B. emersonii ESTs, such as construction and nucleotide sequencing of cDNA libraries, removal of contaminant sequences, and the annotation process were previously described [14]. to 138 million residues. When choosing species to incorporate into the datasets, the number of sequences deposited in Genbank and the biological representation into the group were considered. Searches throughout databases were carried out using the NCBI stand-alone blastall program. BLASTX and tBLASTX algorithms [53] were used for searching against protein databases and unigene databases, respectively. Linux tools and scripts were used to deal with data sets and blast outputs, and extract specific text/data lines of interest. The pipeline is summarized in Figure 4. Database sizes ranged from ~40 million to 138 million of amino acids and we used an Evalue ≤ 10 -5 as the cut-off to assign significance to best hit in the alignments. Final data to be analysed (indicated as "B. emersonii-animal shared sequences", "B. emersonii-plant shared sequences" and "B. emersonii-animal-plant shared sequences") were obtained after their filtration against species not included in our databases (fungi, plants and animals) to remove those sequences initially considered as not found in these groups, named as contaminants in this study. Hits found also in bacteria were specially checked for the presence of a poly A+ tail. Taking account that after the initial construction of the ad hoc databases several new fungal genomes became publicly available [6], we constructed two new fungal databases (protein and nucleotide bases) to proceed with the filtration. Data were downloaded from four of the several centres that have released genome sequences [6]: the Joint Genome Institute (JGI) [54], the Broad Institute [55] the University of Oklahoma [56] and the NCBI [57]. The complete list of species used to construct fungal databases is in Table 6. A local search against the new fungal bases was made using BLASTX or tBLASTX algorithms. We also used a client server program (blastcl3 program) and BLASTX or tBLASTX algorithms for remote search against non-redundant (nr) and dbEST-others databases from Genbank [58], respectively. Considering that databases at Genbank are larger than our ad hoc databases, we adjusted the E threshold to a less stringent value (~1 to 6E-4), maintaining S' constant (~50) and following the equation E = mn2 -S' [53]. Standalone blast and client server blast packages were downloaded from NCBI BLAST ftp site [59].

Approach 2. Database source and construction, and pipeline for comparative sequence analysis
Two databases were constructed using N. crassa and U. maydis protein sequences downloaded from NCBI database. N. crassa and U. maydis were chosen in this study as representatives of ascomycetes and basidiomycetes with completely sequenced genomes, respectively. Putative orthologues of B. emersonii in N. crassa or U. maydis were obtained by comparing B. emersonii full-length sequences against the two fungal databases using BLASTX program. The pipeline is summarized in Figure 5. We chose not to proceed with a bidirectional best hit (BBH) comparison to select orthologous sequences because it could produce equivocal results, since B. emersonii transcriptome data are incomplete. Instead, we carried out a final manual revision of the resulting divergent sequences to exclude paralogues from our analysis. We accepted as homologues B. emersonii sequences that presented at least 80% of overlap with the corresponding protein sequences in N. crassa or U. maydis, and an Evalue ≤ 10 -5 as the cut-off to assign significance to best hit in the alignments. Full-length coding sequences were estimated as previously reported [14]. We based our analysis on the procedure adopted by Braun et al. [8] for comparing the amount of divergence. Pairs of putative orthologues from N. crassa/U. maydis and translated B. emersonii putative unique sequences were compared using BLASTP or tBLASTP against animal or plant databases, respectively, and the obtained bit scores (S') were recorded. A score difference equal or higher than 150 (ΔS' ≥ 150) was chosen to consider proteins as divergent. Divergent bit scores were re-evaluated by comparing them to other fungal vs. animal/plant scores to exclude divergences only proper to the two fungi initially considered.

Sequence annotation
To assign a putative identification to B. emersonii uniseqs, we took into account BLASTX best-hit descriptions, or subsequent alignments with an Evalue below the assumed cut-off, resulting from sequence comparison against the nr and dbEST-others databases at NCBI. We also considered the biological process categories from Gene Ontology Consortium (GO) [60] attributed to uniseqs after comparison with sequences from curated databases (Swiss-Prot and TrEMBL) available at ExPASy proteomics server of the Swiss Institute of Bioinformatics (SIB) [61]. We maintained the GO structure we used in [14] for the classification of B. emersonii ESTs. This classification is available at [43]. However, GO classification is being upgraded continuously; upgrades can be checked in [60]. Other information sources were also consulted (mainly InterPro [62] and linked references, MIPS [63], Fantom3 [64] and Fly-Base [65]) to refine the annotation.
Overview of the pipeline used in the EST comparative study, approach 1

B -contaminants
Overview of the pipeline used in the EST comparative study, approach 2 Figure 5 Overview of the pipeline used in the EST comparative study, approach 2. B. emersonii putative full-length uniseqs were compared against fungal protein sequences from N. crassa and U. maydis, and orthologous pairs were compared against animal and plant databases. See Methods section for details.