•Brown algae exhibit a unique carbon (C) storage metabolism. The photoassimilate d-fructose 6-phosphate is not used to produce sucrose but is converted into d-mannitol. These seaweeds also store C as β-1,3-glucan (laminarin), thus markedly departing from most living organisms, which use α-1,4-glucans (glycogen or starch).
•Using a combination of bioinformatic and phylogenetic approaches, we identified the candidate genes for the enzymes involved in C storage in the genome of the brown alga Ectocarpus siliculosus and traced their evolutionary origins.
•Ectocarpus possesses a complete set of enzymes for synthesis of mannitol, laminarin and trehalose. By contrast, the pathways for sucrose, starch and glycogen are completely absent.
•The synthesis of β-1,3-glucans appears to be a very ancient eukaryotic pathway. Brown algae inherited the trehalose pathway from the red algal progenitor of phaeoplasts, while the mannitol pathway was acquired by lateral gene transfer from Actinobacteria. The starch metabolism of the red algal endosymbiont was entirely lost in the ancestor of Stramenopiles. In light of these novel findings we question the validity of the ‘Chromalveolate hypothesis’.
Brown algae (Phaeophyceae) are photosynthetic, multicellular organisms that dominate rocky coastal environments. These macroalgae are members of the Stramenopiles, a eukaryotic phylum which also includes diatoms, Oomycetes and various protists (Fig. 1) (Baldauf, 2008). Stramenopiles are characterized by the occurrence in their life history of cells with two unequal flagella, and the monophyly of this group has been confirmed by molecular phylogenies (Ben Ali et al., 2001). Stramenopile plastids arose via a secondary endosymbiotic event, in which a unicellular red alga was engulfed by an ancestral protist (Reyes-Prieto et al., 2007). Other eukaryotic lineages also possess secondary plastids derived from red algae, particularly the Alveolates, Haptophytes and Cryptophytes (Fig. 1). Cavalier-Smith (1999) proposed that a single secondary endosymbiosis with a red alga gave rise to the plastid ancestor of all these eukaryotic groups (the Chromalveolate hypothesis) and, therefore, that the host lineages form a monophyletic supergroup designated as Chromalveolates. The Chromalveolate hypothesis has been intensely debated in the last decade and is still a highly contentious issue (Bodyl et al., 2009; Keeling, 2009).
This complex evolutionary history of brown algae is reflected by the uniqueness of their carbohydrate metabolism. With respect to the outflow of the Calvin cycle, the photoassimilate d-fructose-6-phosphate (F6P) is not used by brown algae to produce sucrose as in higher plants, but it is mainly converted into d-mannitol. This alcohol sugar is localized in the cytosol and is synthesized in two steps: F6P is reduced by mannitol-1-phosphate 5-dehydrogenase (M1PDH) into mannitol-1P, which is then converted into mannitol by mannitol-1-phosphatase (M1Pase). These enzymatic activities have been measured in several brown algae (Ikawa et al., 1972), but the corresponding genes have not been described yet. Mannitol is thought to be recycled by mannitol-2-dehydrogenase (M2DH) and hexokinase (Iwamoto & Shiraiwa, 2005).
Brown algae and other Stramenopiles also possess unique carbon (C) storage polysaccharides. Most living organisms store C as linear or branched α-1,4-glucans: glycogen in Opisthokonta (animals and fungi) and in most bacteria, or starch in the diazotrophic cyanobacteria and Archaeplastida (i.e. red algae, green algae and plants; Deschamps et al., 2008). By contrast, the storage polysaccharide of brown algae is laminarin, a vacuolar β-1,3-glucan with occasional β-1,6-linked branches (Percival & Ross, 1951). This polysaccharide is polydisperse, consisting of a minor G-series with polymers containing only glucose residues, and a more abundant M-series with glucans terminated with a 1-linked d-mannitol residue (Read et al., 1996). Diatoms and Oomycetes also produce vacuolar β-1,3-glucans, known as chrysolaminarin (Beattie et al., 1961) and mycolaminarin (Wang & Bartnicki-Garcia, 1974), respectively. These storage polysaccharides have a branched structure similar to that of laminarin, but they do not contain mannitol residues. In the sieve elements of Laminariales, plates are made of microfibrillar β-1,3-glucans deposits, reminiscent of plant callose (Parker & Huber, 1965). Self-assembling linear β-1,3-glucans also occur as structural components in the cell wall of Oomycetes (Bartnicki-Garcia, 1968).
Experiments with radioactive C demonstrated that laminarin and mannitol are interchangeable storage compounds in brown algae, as are sucrose and starch in higher plants (Yamaguchi et al., 1966). In kelps, mannitol can be remobilized and translocated via the sieve tubes from mature tissues to supply the rapidly growing parts of the alga with C (Schmitz & Lobban, 1976; Lobban & Harrison, 1994).
The molecular bases of carbohydrate metabolism in brown algae are essentially uncharacterized. Until recently, expressed sequence tag (EST) libraries were the only molecular data available for inferences on central and storage C metabolisms in brown algae. Analysis of a cDNA library produced from sporophytes of Laminaria digitata retrieved six partial open reading frames corresponding to genes potentially involved in central C metabolism (Moulin et al., 1999). The complete genome sequence of the brown alga Ectocarpus siliculosus, from the order Ectocarpales (Charrier et al., 2008), is now available (Cock et al., 2010), providing, for the first time, a comprehensive view of brown algal carbohydrate metabolism.
Here we unravel the candidate proteins for central sugar metabolism and C storage in brown algae and reconstruct the phylogenetic relationships of key enzymes in these metabolic pathways. This analysis provides novel insights into the origin and evolution of carbohydrates in Eukaryotes and the results raise doubts about the Chromalveolate hypothesis.
Materials and Methods
Identification and bioinformatic analyses of carbohydrate-related proteins
The proteins involved in carbohydrate metabolism encoded by the E. siliculosus (Dillwyn) Lyngbye genome were identified by homology with biochemically characterized proteins selected in the CAZY (http://www.cazy.org/, Cantarel et al., 2009) and UniProt databases. For each identified Ectocarpus protein, evidence of conserved protein modules was queried using the Pfam database (Bateman et al., 2004). The presence of additional, orphan modules was detected by BlastP searches against the UniProt database. Signal peptides and transmembrane helices were predicted using HECTAR (Gschloessl et al., 2008) and TMHMM (Krogh et al., 2001), respectively. Numerous proteins involved in carbohydrate metabolism belong to families encompassing several enzymatic activities and/or substrate specificities. To clarify their function, these proteins were further analyzed by a phylogenetic approach. For each different activity of a polyspecific family, a set of experimentally characterized proteins was selected in the UniProt database. These representative proteins were aligned with their homologs from E. siliculosus using MAFFT with the iterative refinement method and the scoring matrix Blosum62 (Katoh et al., 2002). Phylogenetic trees were derived from this refined alignment using the maximum likelihood method with the program PhyML (Guindon & Gascuel, 2003). The reliability of the trees was always tested by bootstrap analysis using 100 resamplings of the dataset. The trees were displayed with MEGA (Kumar et al., 2004). The functional annotation of Ectocarpus proteins was based on the proximity to specific characterized proteins in the phylogenetic trees. Genomic comparisons were performed using the genomic BLAST server at NCBI (http://www.ncbi.nlm.nih.gov/sutils/genom_table.cgi).
From C fixation to carbohydrate biosynthesis: an overview of the carbohydrate active enzymes in E. siliculosus
As in most photosynthetic organisms, Rubisco is responsible for C fixation in brown algae, releasing two molecules of glycerate-3-phosphate (Assali et al., 1991). The enzymes converting this triose-phosphate into F6P are well conserved in E. siliculosus (Supporting Information, Table S1). Based on bioinformatic predictions of signal peptides (Gschloessl et al., 2008), the corresponding isoenzymes mainly differ in their localization (cytosol, mitochondrial or plastid matrices). Interestingly, the gene Esi0187_0027 encodes a modular protein encompassing a triose-phosphate isomerase (TPI) fused to a glyceraldehyde-3-phosphate dehydrogenase (GAPDH). Homologs of this translational fusion are only found in diatoms (Armbrust et al., 2004; Bowler et al., 2008) and Oomycetes (Tyler et al., 2006; Haas et al., 2009), suggesting that this bifunctional enzyme is characteristic of Stramenopiles. The TPI/GAPDH fusion protein from the diatom Phaeodactylum tricornutum has been shown to be active and is localized in mitochondria (Liaud et al., 2000). The enzymes that equilibrate the hexose-phosphate pool (F6P, d-glucose-6P and d-glucose-1P) are also highly conserved, with two isoforms each of glucose-6 isomerase (GPI, Esi0060_0128 and Esi0266_0033) and of phosphoglucomutase (PGM, Esi0002_0317 and Esi0430_0005). Moreover, Esi0430_0005 displays an additional uridine diphosphate (UDP)-glucose-pyrophosphorylase (UGP) module at the N-terminus. Again, homologs of this UGP/PGM fusion protein are only found in the genomes of Oomycetes and of the diatom P. tricornutum. In Thalassiosira pseudonana, these activities are, however, encoded by two distinct genes (Armbrust et al., 2004). The putative bifunctional enzymes TPI/GAPDH and UGP/PGM catalyze consecutive reactions, suggesting that these fusion events may have increased the efficiency of the metabolic steps, leading to the activated form of glucose (UDP-glucose) in most Stramenopiles.
Synthesis of polysaccharides and glycoconjugates is catalyzed by glycosyltransferases (GTs), which use activated sugar donors, or by transglycosylases. The cleavage of glycosidic linkages is performed by glycoside hydrolases (GHs) or by polysaccharide lyases (PLs). Carbohydrate esterases (CEs) remove methyl or acetyl groups from substituted polysaccharides. Collectively these enzymes are termed carbohydrate active enzymes (CAZYmes). Based on sequence similarities (Henrissat, 1991), CAZYmes have been classified into > 200 protein families (http://www.cazy.org/; Cantarel et al., 2009). The genome of Ectocarpus encodes 41 GHs and 88 GTs, but surprisingly lacks genes homologous to known PLs or CEs (Table S2). This seaweed possesses a slightly greater number of GH/GT genes than the marine green microalgae Micromonas and Ostreococcus (Worden et al., 2009), but at least six times fewer than terrestrial plants. Arabidopsis thaliana, for example, has 730 GH/GT genes (Henrissat et al., 2001). However, looking at the number of CAZY families, the difference between brown algae and land plants is less pronounced: Ectocarpus has members of 18 GH and 32 GT families, while Arabidopsis contains members of 34 GH and 40 GT families. The impressive number of CAZymes in plants is mainly explained by the presence of large multigenic families. For instance, Arabidopsis has 67 polygalacturonases (GH28), which participate in pectin recycling, and 121 GT1s that are mainly involved in secondary metabolite biosynthesis (Henrissat et al., 2001). By con-trast, Ectocarpus features less functional redundancy, with fewer genes in each CAZY family. Interestingly, this seaweed contains some families which are absent from plants but are conserved with other phyla, such as bacteria (GH88, Δ-4,5 unsaturated β-glucuronyl hydrolase), fungi (GT15, biosynthesis of cell wall mannoproteins), animals and fungi (GH30, β-glucosidase; GT23, GT49 and GT54, N-glycosylation) and Amoebozoa (GT60 and GT74, O-glycosylation of Skp1 subunits of E3 ubiquitin-protein ligase).
Three Ectocarpus proteins (Esi0017_0062, Esi0020_0181 and Esi0080_0017) share significant sequence identity (c. 30%) with the M1PDH from the Apicomplexa Eimeria tenella. The only known M1Pase gene was also cloned from this parasite (Liberator et al., 1998), but no homolog has been found in Ectocarpus. Recycling of mannitol is probably performed by Esi0135_0010, which is similar to bacterial and fungal M2DH. However, no ortholog of eukaryotic hexokinases was found in E. siliculosus. Instead, this seaweed possesses a gene, Esi0139_0025, which is closely related to cyanobacterial fructokinases. Thus, brown algae are apparently an exception to the trend that broad-specificity hexokinases are typical of multicellular Eukaryotes, whereas sugar-specific kinases are distinctive of bacteria and unicellular Eukaryotes (Cardenas et al., 1998).
Putative M1PDH genes have been identified in cDNA libraries from the brown algae L. digitata (Roeder et al., 2005) and Sargassum binderi (Wong et al., 2007), but are absent from the genome of diatoms and Oomycetes. Homologs of M1PDH from Apicomplexa and brown algae have been found in only a few organisms so far, namely Gram-positive bacteria, some fungi and two species of Micromonas (Worden et al., 2009). The presence of M1PDH in Micromonas is surprising, since this gene is absent from the genomes of two other green algae, namely Ostreococcus (Derelle et al., 2006) and Chlamydomonas (Merchant et al., 2007), from the red alga Cyanidioschyzon merolae (Matsuzaki et al., 2004) and from terrestrial plants. However, mannitol has been reported in some other Prasinophytes (Kremer, 1980) and M1PDH activity was detected in the green alga Platymonas subcordiformis (Richter & Kirst, 1987). Phylogenetic analyses suggest that bacterial M1PDH genes were acquired independently by fungi and brown algae (Fig. 2a). By contrast, the M1PDH sequences of E. tenella and of the Micromonas spp. are nested within the brown algal clade. The Micromonas M1PDH genes (MICPUN_62892 and MICPUC_48208) both encode fusion proteins that include a haloacid dehalogenase-like domain (HAD). Interestingly the HAD superfamily contains various phosphatases, including sucrose-phosphate and trehalose-phosphate phosphatases (Ridder & Dijkstra, 1999). The two Micromonas spp. possess a second HAD-like enzyme existing as a standalone protein (MICPUC_47598 and MICPUN_101665). These M1PDH/HAD proteins are reminiscent of the fusion sucrose-phosphate synthase/sucrose-phosphate phosphatases (SPS/SPP) and trehalose-phosphate synthase/trehalose-phosphate phosphatases (TPS/TPPs), suggesting that the HAD modules provide the missing M1Pase activity. This assumption is further strengthened by the identification in Ectocarpus of two standalone proteins (Esi0080_0016 and Esi0100_0020) highly similar to the Micromonas putative M1Pases (63% sequence identity with MICPUC_47598). In addition, the putative M1Pase Esi0080_0016 is located next to the M1PDH gene Esi0080_0017. This would be a rare example of functional clustering in Ectocarpus (Cock et al., 2010). A conserved putative M1Pase was also found in the EST library of the brown alga Fucus serratus (Pearson et al., 2009), while additional significant homologs occur in some bacteria and fungi only. Comparison with crystal structures of HAD-like phosphatases confirm that the catalytic machinery (Ridder & Dijkstra, 1999) is strictly conserved in the putative M1Pases (data not shown). Phylogenetic analyses indicate that brown algal putative M1Pases are also basal to the Micromonas standalone enzymes, while putative MP1ases fused to M1PDH form a sister group (Fig. 2b).
Sucrose and trehalose are the most commonly occurring nonreducing disaccharides. Their metabolic pathways are biochemically similar, but involve enzymes which are not homologous (Paul et al., 2008). Phylogenetic analyses of sucrose-related proteins indicate that plant sucrose metabolism was acquired from the cyanobacterial progenitor of chloroplasts (Salerno & Curatti, 2003). Sucrose metabolism is completely absent in Ectocarpus, as deduced from the lack of genes encoding SPS (family GT4), sucrose synthase (GT4), SPP and invertases (families GH32 and GH100). These enzymes are also absent in diatoms and Oomycetes, with the exception of the GH32 family of invertases present in Phytophthora spp. However, these latter enzymes are related to fungal orthologs, suggesting that Phytophthora uses invertases to feed on plant sucrose. By contrast, Ectocarpus likely possesses a complete trehalose pathway. This sugar is synthesized by a family of six bifunctional enzymes, including a TPS (family GT20) fused to a TPP, while the recycling of trehalose is probably assured by a single trehalase (family GH37).
Trehalose-phosphate synthases are widespread in bacteria, suggesting a prokaryotic origin for these glycosyltransferases. Bacterial TPS exists in two forms, either as standalone protein or as a fusion with TPP. Both types of TPS exist in fungi and they form two distinct clades in phylogenetic trees (data not shown). By contrast, TPSs are only present as fusion proteins in Dictyostelium discoideum (Amoebozoa), insects, red algae, green algae, plants, Alveolates and Stramenopiles. Phylogenetic analyses of the TPS and TTP domains of these fusion proteins result in similar topologies, with similar clades observed for both domains (Fig. 3), even though the TPS domains are more divergent and score weaker bootstrap values. Therefore the TPS and TPP domains have co-evolved, indicating that all these eukaryotic phyla have acquired TPSs directly as fusion proteins. Bacterial and archaeal TPS domains constitute a well-supported clade, which was chosen as the outgroup (Fig. 3a). The TPSs from insects emerge as an isolated cluster, while the fungal bifunctional TPSs are distantly related to the TPS domain of D. discoideum. The TPSs from Archaeplastida, Alveolates and Stramenopiles form three distinct clades, which group together. The red alga C. merolae possesses three TPS isoforms (CMI293C, CMO053C and CMP219C). Green algal and plant TPSs cluster together into two well-supported clades, which are sister groups of CMI293C and CMO053C, respectively. However, CMP219C-like TPS genes are absent from the green lineage (Fig. 3a), indicating either gene loss in the ancestor of green algae and plants or a specific gene duplication in red algae. Red algal TPSs are basal to the TPSs from Stramenopiles and Alveolates, indicating that these eukaryotic phyla acquired TPSs from ancestral red algal endosymbiont(s). Ectocarpus possesses the three types of red algal TPS and has even evolved a complex CMI293-like TPS family. Oomycetes conserved the CMP219C- and CMO053C-like TPS but they lost the CMI293C-like TPS. By contrast, diatoms and Apicomplexa have conserved only this latter type of TPS (Fig. 3a).
Laminarin and other β-1,3-glucans
The biosynthetic pathway of laminarin is essentially unknown, but we identified several genes which are likely to be involved in this metabolism in the Ectocarpus genome. This seaweed contains two cytosolic isoforms of UDP-glucose pyrophosphorylase (UGP, Esi0144_0004 and Esi0430_0005), supporting the assumption that UDP-glucose is the activated sugar needed for the production of laminarin. Beta-1,3-glucan synthases fall into two different GT families: the GT2, a polyspecific family which includes bacterial β-1,3-glucan synthases, and the GT48, which only contains eukaryotic β-1,3-glucan synthases (Cantarel et al., 2009). Ectocarpus harbors 11 GT2 homologous to cellulose synthases. Phylogenetic analysis confirms that these enzymes are not significantly related to bacterial β-1,3-glucan synthases (Michel et al., 2010). By contrast, the three members of the GT48 family (Esi0033_0138, Esi0193_0029 and Esi0338_0032) display significant similarities with plant callose synthases (c. 35% sequence identity). The phylogenetic tree of the GT48 family (Fig. 4) is congruent with the currently accepted phylogeny of the Eukaryotes (Fig. 1), with the β-1,3-glucan synthases from plants, fungi, Apicomplexa and Stra-menopiles consistently emerging as distinct clades. The putative β-1,3-glucan synthases of Stramenopiles are further divided into three groups. Clade A, which includes Esi0338_0032, is common to brown algae, diatoms and Oomycetes, suggesting that these glycosyltransferases are responsible for the polymerization of the backbones of laminarin, chrysolaminarin and mycolaminarin, respectively. The β-1,3-glucan synthases of clade B are only found in Phytophthora, and therefore they are likely to be involved in the biosynthesis of the Oomycete cell wall β-1,3-glucans. Clade C is unique to brown algae, but the exact role of Esi0033_0138 and Esi0193_0029 is unclear. These β-1,3-glucan synthases could specifically catalyze the production of laminarin M-series, which is a distinctive feature of brown algae (Read et al., 1996). Alternatively, they might be involved in callose biosynthesis, this molecule having been found in the sieve plates of Laminariales (Parker & Huber, 1965). In addition, Ectocarpus possesses two proteins (Esi0100_0034 and Esi0243_0020) which are homologous to KRE6, a GH16 family transglycosylase involved in β-1,6-branching of cell wall β-1,3-glucans in yeasts (Montijn et al., 1999). Therefore, these two proteins represent good candidates for the synthesis of β-1,6-linked branches of laminarin. This hypothesis is strengthened by the conservation of KRE6-like proteins in diatoms and Oomycetes. Remarkably, the KRE6-like protein PITG_03335 from P. infestans (Haas et al., 2009) is fused to a GT48 family β-1,3-glucan synthase. This GT48 module belongs to the clade A (Fig. 4), the very subgroup that we suggest to be responsible for laminarin polymerization.
The degradation of laminarin is potentially catalyzed by 10 endo-1,3-β-glucanases belonging to three different families (GH16, four genes; GH17, one gene; GH81, five genes), and by two exo-1,3-beta-glucanases (family GH5). These numerous laminarinases have homologs in bacteria (family GH16), fungi (families GH5 and GH81) and plants (family GH17), underlining the complexity of laminarin metabolism in brown algae. Laminarin oligosaccharides would be further hydrolyzed by β-glucosidases of the GH1 (Esi0061_0010, Esi0176_0045 and Esi0212_0019) or GH3 families (Esi0010_0226). The end-product, glucose, would be subsequently phosphorylated by a glucokinase (Esi0000_0270) before entering glycolysis.
Ectocarpus siliculosus has not retained any trace of the starch metabolism of its red algal endosymbiont
The uptake of a red alga by the heterotrophic ancestor of brown algae resulted in the acquisition of photosynthesis. In plants and red algae, the acquisition of photosynthetic capacity following the capture of a cyanobacteria was accompanied by the transformation of the glycogen metabolism of the ancestral host cell into a starch biosynthetic pathway (Ball & Morell, 2003; Deschamps et al., 2008). In contrast to plants, red algae and dinoflagellates synthesize starch in their cytosol, and not in their plastids. Biochemical analyses suggested that these organisms use UDP-glucose as the activated sugar instead of ADP-glucose (Nyvall et al., 1999; Viola et al., 2001). This hypothesis has been recently demonstrated through mutant selection and analysis in dinoflagellates (Dauvillee et al., 2009). The question then arises as to whether there is any remnant of this red algal starch metabolism in extant brown algae.
Starch and glycogen are synthesized and recycled by homologous enzymes. The α-1,4-glucan backbone is produced by starch synthase (GT5) or glycogen synthase (GT3 and GT5), while branching enzymes (GH13) are responsible for the creation of α-1,6 branches (Ball & Morell, 2003). Recycling is mainly performed by the combined action of glycogen/starch phosphorylases (GT35), debranching enzymes (GH13) and α-1,4-glucanotransferases (GH77). Additional hydrolytic enzymes are also involved in α-1,4-glucan catabolism: α- and β-amylases (GH13 and GH14) and α-glucosidases of the GH31 family (Ball & Morell, 2003). Starch degradation is specifically initiated by glucan water dikinases (GWDs) and phosphoglucan water dikinases (PWDs), which loosen the crystalline starch granules (Edner et al., 2007). A systematic search for all these enzymes in Ectocarpus did not retrieve any corresponding gene. This seaweed does contain two isoforms of UGP, but these enzymes are not specific for glycogen/starch metabolism as they could also be involved in other biosynthetic pathways (cellulose, laminarin, glycosylation). The ADP-glucose pyrophosphorylase (AGP), which is responsible for ADP-glucose synthesis in plants, is also absent from brown algae. Diatom and Oomycete genomes also completely lack starch- or glycogen-related genes.
The central and storage carbohydrate metabolism in brown algae
Fig. 5 shows the putative central and storage C metabolism of E. siliculosus, as reconstructed from the whole-genome analyses reported earlier. Further experimental analyses will be needed to confirm the biochemical function of the identified candidate genes. In brown algae, this metabolism is profoundly different from what is known for fungi, animals and plants, a uniqueness which is further reinforced in the phylogenomic reconstruction of the metabolic routes for the biosynthesis of cell wall polysaccharides (Michel et al., 2010). With the exception of guanosine diphosphate (GDP)-mannose pyrophosphorylase (Michel et al., 2010), the enzymes involved in the equilibration of the hexose-phosphate pool and in the generation of activated sugars are conserved with other Eukaryotes. However, these key sugars are then used by specific enzymes to generate two unique storage compounds, mannitol and laminarin. Fructose-6-phosphate plays a central role in these pathways. As in the other photosynthetic lineages, this photoassimilate is the precursor of all activated sugars. In brown algae, however, F6P can be also directly transformed into a soluble storage compound (mannitol), without prior conversion into an activated sugar (Fig. 5). Two other activated sugars assume pivotal functions: UDP-glucose, as the starting point for the biogenesis of trehalose, laminarin and cellulose; and GDP-mannose, which is the direct precursor of alginates and the indirect precursor of sulfated fucans (Michel et al., 2010).
Particularly attractive targets are the putative M1Pases and the KRE6-like proteins, which are potentially responsible for the final steps of mannitol and laminarin biosynthesis, respectively. Our findings also open several novel questions as to the nature of the central and storage carbohydrate metabolism in brown algae. To the best of our knowledge, the potential ability of brown algae to produce trehalose was never reported before this study, even in recent reviews (Elbein et al., 2003; Iturriaga et al., 2009). In plants, trehalose and trehalose-6-P (T6P) are in low abundance and difficult to assay. The role of trehalose in green plants is uncertain, but it may regulate starch breakdown (Paul et al., 2008). T6P is considered to be a signal of glucose-6-P (G6P) and UDP-glucose pool size and, hence, to acts as an effective indicator of sucrose. This signal metabolite is thus thought to coordinate C metabolism with plant development in response to C availability and stress (Paul et al., 2008). By analogy with plants, and since F6P, G6P and UDP-glucose are also interconnected to T6P in Ectocarpus (Fig. 5), trehalose and T6P might act as metabolic regulators in brown algae.
Another unresolved question is the biochemical route which connects mannitol and laminarin and the reason why the majority of the laminarin chains are terminated by a mannitol residue at their reducing end (Read et al., 1996). An attractive working hypothesis is that the addition of a mannitol residue is probably catalyzed by specific glycosyltransferases, for instance the GT48 isoforms which are unique to brown algae (Fig. 4), or by a new, undiscovered family of glycosyltransferases. Considering the catalytic mechanism of glycosyltransferases (Lairson et al., 2008), transfer of activated glucose onto mannitol is likely to be the first step in the biosynthesis of the laminarin M-chains. Indeed, the simplest way to form a covalent bond between mannitol and glucose residues would be for an as yet undescribed GT to use mannitol as an acceptor molecule while the donor sugar substrate would be UDP-glucose. The newly formed glucose-mannitol disaccharide would be subsequently used as an acceptor molecule by a GT48 which would further elongate the laminarin chain.
Origin of central and storage carbohydrates in brown algae
Brown algae and the other Stramenopiles have a rich evolutionary history. This lineage arose from the symbiosis between a protist and a red alga, which gave rise to plastids. As red algae had themselves originated from a primary endosymbiosis, between a protist and a cyanobacterium, the emergence of Stramenopiles is described as a secondary endosymbiosis, which involved two eukaryotic cells (Reyes-Prieto et al., 2007). Based on molecular clock analyses calibrated by paleontological constraints, Stramenopiles are thought to have diverged from other major eukaryotic groups c. 1 billion yr ago and it is also thought that the secondary endosymbiosis happened shortly after the primary endosymbiosis (Douzery et al., 2004).
A recent genomic study has further complicated this already complex picture by suggesting that a cryptic endosymbiosis with a green alga, related to extant Prasinophytes such as Micromonas, had occurred in Stramenopiles before the capture of the red algal secondary endosymbiont. This hypothesis is based on the presence, in Stramenopile genomes, of a high number of genes phylogenetically related to those of green microalgae (Moustafa et al., 2009). However, the only available whole-genome resource for the red algal lineage so far is the genome sequence of C. merolae, an unusual red microalga which lacks cell wall, lives in acidic hot water and has a reduced genome (Matsuzaki et al., 2004). Therefore, the interpretation of Moustafa et al. is somewhat weakened from the lack of genomic data from archetypal red algae, which may also possess many of the proposed ‘green algal’ genes (Dagan & Martin, 2009). In this intricate context we discuss in the following sections the possible origins of the various carbohydrates from brown algae.
Trehalose was acquired from the red algal endosymbiont
Sucrose is mainly limited to plants, green algae and cyanobacteria, while trehalose is found in a large range of bacteria and Eukaryotes (Salerno & Curatti, 2003). The widespread occurrence of trehalose led to the hypothesis that this nonreducing disaccharide is a more ancient metabolite than sucrose (Goddijn & van Dun, 1999). Sucrose metabolism in green algae and land plants (Chloroplastida) originated from the cyanobiont (Salerno & Curatti, 2003). Consequently, red algae probably acquired sucrose metabolism during primary endosymbiosis as well, even though these photosynthetic organisms are not currently known to produce sucrose (Kremer, 1980). Inspection of the genome of C. merolae (Matsuzaki et al., 2004) indicates that this red alga does not contain any sucrose-related gene. Similarly, genes involved in sucrose metabolism are completely absent from Stramenopiles genomes, as deduced from the lack of sucrose synthase, SPS and of invertase. Therefore, it is very likely that the red alga that gave birth to the phaeoplasts of extant Stramenopiles (Reyes-Prieto et al., 2007) had already lost the cyanobacterial sucrose metabolism when the secondary endosymbiosis took place. Nevertheless, this hypothesis needs additional genomic data on red algae to be definitively confirmed.
In the case of trehalose, the question also arises as to the origin of this potential C storage disaccharide in brown algae; whether it was originated from the host or from the red algal endosymbiont. Trehalose synthesis requires two enzyme activities: TPS and TPP. Here we show that the six putative TPS/TPPs of Ectocarpus are closely related to their orthologs from C. merolae (Fig. 3). Therefore, it is likely that trehalose metabolism was transmitted into Stramenopiles by the endosymbiont and that the red algal genes supplanted pre-existing host orthologs, if any existed. The assimilation of the red algal pathway by the host was probably facilitated by the fact that this pathway already consisted of a single bifunctional enzyme, with both TPP and TPS activities, in ancestral red algae (Fig. 5).
Mannitol: evidence for a major horizontal gene transfer event from Actinobacteria
Brown algae differ from other Eukaryotes, including diatoms and Oomycetes, by their capacity to synthesize an additional C storage sugar, mannitol. The conversion of F6P into mannitol involves two enzymatic steps (Fig. 5), catalyzed by M1PDH and M1Pase. These enzyme activities are indeed rather rare among living organisms and here we show that the putative M1PDH and M1Pase genes from Ectocarpus are closely related to those of Gram-positive bacteria (Fig. 2). Based on this close phylogenetic relationship, we propose that, after the divergence of brown algae from diatoms and Oomycetes, these two enzymes were imported into brown algae by a horizontal gene transfer involving an ancestral Gram-positive bacterium related to extant Actinobacteria. It is noteworthy that this bacterial phylum is common and diversified in the marine environment (Bull et al., 2005). As reported in the second part of this study (Michel et al., 2010), this horizontal gene transfer (HGT) event extended beyond the acquisition of mannitol metabolism. The actinobacterium would also have contributed to the enzyme machinery that synthesizes alginate and hemicelluloses, two major cell wall polysaccharides which are absent from diatoms and Oomycetes. Therefore, this bacterial HGT was probably instrumental in the emergence of complex multicellularity in brown algae, by providing two essential elements in this evolutionary process: extracellular matrix polysaccharides for the construction of multicellular tissues and organs; and mannitol for the long-distance translocation of photoassimilates (Parker & Huber, 1965).
Micromonas acquired the capacity to synthesize mannitol from brown algae
In the phylogenetic analyses of putative M1PDH and M1PAses, the Stramenopile genes cluster with their orthologs in Micromonas (Prasinophytes), suggesting that these microalgae feature a mannitol metabolism of brown algal origin. Beyond the robustness and the congruence of the phylogenetic trees (Fig. 2), several pieces of evidence support the direction of this potential HGT, from brown algae to these Prasinophytes: putative M1PDH and M1Pases are standalone genes in Ectocarpus and the closest orthologs of brown algal M1Pases in Micromonas are also standalone proteins, whereas the M1Pases fused to M1PDH are more divergent (Fig. 2b). This suggests that a gene duplication specifically occurred in Micromonas followed by a gene-fusion event, between M1PDH and the putative M1Pase domains; putative M1PDH and M1Pase genes are not conserved in other Prasinophyte genomes and are absent from red algae, chlorophytes, mosses and higher plants. By contrast, orthologs of these genes are present in brown algal EST libraries (Roeder et al., 2005; Wong et al., 2007; Pearson et al., 2009); and the alginate- and hemicellulose- related genes of actinobacterial origin in Ectocarpus are not present in any available genomes within the Archaeplastida supergroup (Michel et al., 2010), indicating that these genes were not acquired through a ‘green’ or ‘red’ secondary endosymbiosis. Altogether these findings suggest that the mannitol-related genes in Micromonas were acquired relatively recently from a brown alga by these Prasinophytes.
Laminarin biosynthesis is ancestral in Stramenopiles
The β-1,3-glucan synthases from brown algae and other Stramenopiles belong to the GT48 family, which is conserved in most eukaryotic phyla and absent from bacteria and Archaea. The phylogenetic tree of the GT48 family is congruent with the currently accepted phylogeny of Eukaryotes (Fig. 4). This finding suggests that β-1,3-glucans were present in the last eukaryotic common ancestor (LECA) and that laminarin is an ancestral metabolite in Stramenopiles. The conservation between fungi (Unikonts) and Stramenopiles (Bikonts) of KRE6-like proteins (GH16) and some laminarinases (GH5 and GH81) is also consistent with the ancestral character of laminarin metabolism. It is likely that intracellular β-1,3-glucans were the ancestral forms of these carbohydrates and that they subsequently evolved to be secreted as extracellular polysaccharides as complex multicellularity emerged. Interestingly, such an evolution occurred independently in fungi and in plants, as well as within the Stramenopiles, with Oomycetes and Laminariales producing both storage and cell wall β-1,3-glucans (Parker & Huber, 1965; Bartnicki-Garcia, 1968).
Insights into the origin and evolution of storage polysaccharides in Eukaryotes
Biosynthesis and remobilization of C stores comprise a fundamental process in all living cells. Extant Eukaryotes exhibit a striking diversity in the nature of their storage polysaccharides, which probably reflects their complex evolutionary history. We discuss in the following various evolutionary scenarios which would account for such a high diversity.
The starch metabolism from the rhodobiont was not retained in Stramenopiles Starch metabolism is a complex process, involving a minimal set of 10 different enzymes in Rhodophyceae (Deschamps et al., 2008). From the seminal work of Steven Ball and colleagues on the origin of starch metabolism, it is now well recognized that this biosynthetic pathway, characteristic of Archaeplastida, results from the merging of the glycogen metabolism of the eukaryotic host cell with the starch metabolism of the cyanobacterial endosymbiont (Ball & Morell, 2003; Deschamps et al., 2008; Plancke et al., 2008). In particular, all the genes required for the complete pathway of glycogen metabolism in heterotrophic Eukaryotes are maintained in red algae. The transition from glycogen to starch metabolism was the result of the transfer of a small number of cyanobacterial genes, namely isoamylase (GH13), granule-bound starch synthase I (GBSSI, GT5) and α-(1,4)-glucanotransferase (GH77), to the host nuclear genome (Deschamps et al., 2008). This fate contrasts with the usual loss of storage polysaccharides by bacterial parasites and obligatory symbionts (Henrissat et al., 2002; Gil et al., 2004). The success of the acquisition of starch metabolism in the Archaeplastida following plastid primary endosymbiosis was probably the result of compatibility between biosynthetic pathways in the eukaryotic host cell and in its cyanobiont. The limited number of transferred genes needed for this metabolic transformation was also a favorable factor (Deschamps et al., 2008).
In contrast to Unikonts and Archaeplastida, the genomes of brown algae, diatoms and Oomycetes completely lack starch- or glycogen-related genes. It follows that the starch metabolism of the rhodobiont was not retained in Stramenopiles. The question then arises of which evolutionary mechanism led to such a specific loss in this lineage. The storage compounds of Stramenopiles consist of laminarin, a β-1,3-glucan. Here we demonstrate that the metabolism of β-1,3-glucans involves ancient eukaryotic enzymes, which were probably present in the LECA. As to the absence of glycogen metabolism in extant Stramenopiles, several scenarios may have occurred. The simplest situation is that their common ancestor never possessed glycogen metabolism. The C storage machinery of the protistean ancestor of Stramenopiles was thus exclusively dedicated to the synthesis of β-1,3-glucans, a metabolism totally unrelated to the biosynthesis of α-1,4-glucans in its red algal endosymbiont. Based on the reasoning developed by Ball and coworkers for Archaeplastida, and as observed for obligatory symbionts (Henrissat et al., 2002; Gil et al., 2004), this metabolic incompatibility resulted in a rapid loss of the starch metabolism of the engulfed red alga (scenario 1). An alternative hypothesis is that the LECA possessed both β-1,3-glucans and glycogen as storage compounds (Fig. 6). This assumption is plausible because glycogen is also a very ancient polysaccharide. Most of the enzymes involved in this metabolism are conserved in bacteria, in Unikonts and in most Bikonts. An early loss of the glycogen metabolism in the protistean ancestor of Stramenopiles (scenario 2) would account for the subsequent loss of the rhodobiont starch metabolism by the same mechanism as described earlier. The last theoretical possibility is that the Stramenopile host cell possessed both β-1,3-glucans and glycogen at the time of the secondary endosymbiosis. If this hypothesis is correct, it would imply the subsequent loss of both the host glycogen metabolism and the rhodobiont starch metabolism (scenario 3), in spite of the compatibility of these biosynthetic routes. However, it is difficult to accept the idea that the uptake of an additional set of genes involved in α-1,4-glucan synthesis (the red algal starch metabolism) would have resulted in the loss of all the α-1,4-glucan-related genes, from both the host and the rhodobiont. Although it cannot be formally excluded, we therefore do not favor scenario 3.
Synapomorphies of Stramenopiles with respect to C storage metabolism question the Chromalveolate hypo-thesis In comparison to the other major eukaryotic groups, Stramenopiles and Alveolates are rather closely related lineages (Baldauf, 2008). They are thought to belong to a monophyletic supergroup, known as the Chromoalveolates (Cavalier-Smith, 1999), distinct from the Opisthokonts and the Archaeplastida (Fig. 1). In spite of their phylogenetic relatedness with Stramenopiles, Alveolates such as Dinoflagellates and Apicomplexa accumulate starch in their cytosol (Coppin et al., 2005). As observed in the Archaeplastida, the starch pathway of Alveolates is a hybrid metabolism. The indirect debranching enzymes from Apicomplexa have homologs in animals and fungi, but not in Archaeplastida, indicating that the aveolate host cell contained its own pathway for glycogen synthesis (Coppin et al., 2005). Phylogenetic analyses of other key enzymes for starch synthesis demonstrate that, following the secondary endosymbiosis, the ancestral host-cell glycogen pathway evolved into a genuine starch metabolism through the transfer of red algal starch-related genes (Coppin et al., 2005).
Altogether, Stramenopiles and Alveolates have evolved completely different pathways for C storage, β-1,3-glucan vs starch, respectively. The fate of starch metabolism from the red algal endosymbiont(s) was also radically different: complete loss of the rhodobiont starch pathway in Stramenopiles as opposed to transition from ancestral glycogen to red algal-like starch metabolism in Alveolates. Under the Chromal-veolate hypothesis, these two lineages had the same original genetic background, that is, genomes from the same ancestral host and rhodobiont cells. If this is true, it is difficult to speculate how the Stramenopiles and Alveolates have evolved two so strikingly different C storage metabolisms. The only compatible evolutionary route is scenario 3, which postulates the existence of a glycogen metabolism in the Stramenopile ancestor, the maintenance of this pathway up to the secondary endosymbiosis and its loss in the Stramenopiles after the engulfment of the red algal symbiont. As discussed earlier, in the absence of any remnant of glycogen metabolism in extant Stramenopiles, this scenario appears very speculative. It follows that extant Stramenopiles and Alveolates more likely arose from related, yet distinct, eukaryotic host cells. The occurrence of two such independent, secondary endosymbiotic events contradicts the Chromalveolate hypothesis (Cavalier-Smith, 1999).
Through an extensive phylogenomic approach, we have here reconstructed the putative pathway for C storage in the model brown alga E. siliculosus. Nonetheless, the exact function of the diverse candidate genes needs to be rigorously tested by future functional approaches. Compared with the other eukaryotic phyla, Stramenopiles feature very distinctive traits in their storage C metabolism. Long-term C storage is based on laminarin, a soluble vacuolar β-1,3-glucan, whereas most other Eukaryotes use polysaccharides based on α-1,4-glucan (glycogen or starch) form. As suggested in this study, the polymerization of β-1,3-glucans is catalyzed by GT48, a protein family conserved in most eukaryotic groups, but absent in Prokaryotes. Therefore, β-1,3-glucans are an ancient storage polysaccharide in Eukaryotes, which probably coexisted with glycogen in the LECA (Fig. 6). In most of the extant eukaryotic lineages, glycogen has supplanted β-1,3 glucan as storage compound. However, the enzymes responsible for β-1,3-glucan synthesis were conserved in several eukaryotic groups (e.g. fungi, plants, Oomycetes, brown algae), where they evolved to produce extracellular β-1,3-glucans. In Archaeplastida and Alveolates, the ancestral glycogen metabolism of the host cell evolved into a starch metabolism, after the plastid primary and secondary endosymbioses, respectively (Coppin et al., 2005; Deschamps et al., 2008). If the common ancestor of Stramenopiles ever possessed a glycogen metabolism, we propose that this pathway was lost very early in the evolution of this lineage, thus preventing the incorporation of the red algal starch metabolism into the host cell (Fig. 6). Finally, in brown algae only, a significant HGT likely happened with an actinobacterium, leading to the de novo acquisition of an additional pathway for C storage, based on mannitol. This putative HGT also provided several crucial cell wall polysaccharides, alginates and some hemicelluloses (Michel et al., 2010). Considering also that they have inherited an ancestral trehalose metabolism from their red algal endosymbiont, brown algae have evolved a very unique central and storage C metabolism.