Introduction

Aspergillus fumigatus is the most common mold pathogen of humans and causes both invasive disease in immunocompromised patients and allergic disease in patients with atopic immune systems (Denning et al. 2002). Although A. fumigatus only makes up a small proportion of all aerial spores, around 0.3% in the air of one particular hospital, it causes roughly 90% of invasive aspergillosis cases (Brakhage and Langfelder 2002). This suggests that A. fumigatus possesses certain factors that permit it to become an opportunistic human pathogen in immunocompromised patients.

To systematically identify genes and their products in this fungus, including potential targets for chemotherapy, diagnostics, and vaccine development, an international group of scientists initiated the sequencing of the A. fumigatus genome in 2001. The group selected the clinical isolate, Af293, as the strain to be sequenced (Denning et al. 2002).

There is urgent medical need for novel fungicidal agents that have high efficacy, lack of cross-resistance with existing agents, and low host toxicity. Compounds that target enzymes that are essential in fungi but absent in the mammalian host are attractive candidates. Such an enzyme is inositol phosphorylceramide (IPC) synthase, Aur1p, which is involved in the fungal sphingolipid biosynthetic pathway (Fig. 1). It has become increasingly evident that sphingolipids, once thought to be only structural components of cell membranes, are important molecules in cell regulation (Hannun and Obeid 1997). They have important roles in cell stress responses whereby they mediate diverse biological responses such as cell growth, apoptosis, angiogenesis, differentiation, and senescence. For example, sphingosine-1-phosphate plays a role in Ca2+-mediated guard-cell closure, and a sphingosine transfer protein is involved in ceramide synthesis (Ng and Hetherington 2001). Functional studies of sphingolipids in Saccharomyces cerevisiae mutants have provided groundwork for the identification of genes required for sphingolipid biosynthesis in other species because many of the enzymes have been conserved throughout evolution.

Fig. 1
figure 1

Diagram of sphingolipid metabolism in S. cerevisiae. Gene names are shown in italics and the functions of their products are described in Table 2. Dihydrosphingosine-1-P dihydrosphingosine-1-phosphate, Phytosphingosine-1-P phytosphingosine-1-phosphate, IPC inositol phosphoceramide, MIPC mannose inositol phosphoceramide, M(IP)2 C mannose-(inositol-P)2-ceramide

Full size image

Sphingolipids are found in eukaryotic membranes that contain a hydrophobic segment (ceramide), which is a long-chain base (LCB) that is N-acylated with a very long-chain α-hydroxy fatty acid, linked to various polar head groups. The sphingolipids in mammals are sphingosine, while phytosphingosin (PHS) is the primary LCB in plants and fungi on a quantity basis. But plants and fungi contain a diversity of different LCBs in their glycosylceramides (Warnecke and Heinz 2003). Glycosylceramides are involved in host/pathogen interactions (Koga et al. 1998), play a role fungal development (Levery et al. 2002). Many toxins target sphingolipid metabolism (Table 1).

Table 1 Toxins that target sphingolipid metabolism of fungi (adapted from Obeid et al. 2002)
Full size table

The execution of various genetic screens in S. cerevisiae with the availability of its complete genomic sequence could give clues to the identification of the genes or proteins for sphingolipid biosynthesis in the opportunistic fungal pathogen A. fumigatus for which an unannotated genome sequence is available. In this study, we carried out a comparative genome analysis of the sphingolipid biosynthesis pathway in four fungal species including S. cerevisiae, Schizosaccharomyces pombe, Candida albicans, and Neurospora crassa, and attempted to identify sphingolipid biosynthesis genes in the A. fumigatus genome, estimated to be 30 Mb (http://www.sanger.ac.uk).

Materials and methods

Homologue search for sphingolipid biosynthesis gene(s) in A. fumigatus

Sequence data from the A. fumigatus genome were analyzed to identify S. cerevisiae homologues for sphingolipid biosynthesis. S. cerevisiae protein sequences were used as query sequences in BLAST (tBLASTn) searches (Gish and States 1993). The tenfold whole genome shotgun sequence assembly of A. fumigatus was obtained from The Institute for Genome Research (TIGR) website at http://www.tigr.org. The BLAST search engine runs the WU-BLAST 2.0 program (http://blast.wustl.edu). The tBLASTn program was used for the homologue search, and the statistical significance threshold for reporting matches against database sequences was 10. The e-value cutoff used to assign homologues was 1-e6. For comparative analysis of the sphingolipid biosynthetic pathway of A. fumigatus with that of S. cerevisiae, S. pombe, C. albicans, and N. crassa, we retrieved the genes for sphingolipid biosynthesis from several databases (geneDB, http://www.genedb.org; Candida DB, http://genolist.pasteur.fr/CandidaDB; NeurosporaBD, http://www-genome.wi.mit.edu/annotation/fungi/neurospora; SPTR, http://srs.ebi.ac.uk).

Gene prediction by GlimmerM

Regions of the A. fumigatus genome that possessed a high sequence similarity with S. cerevisiae genes for sphingolipid biosynthesis in the BLAST search were used as input for the GlimmerM program, which is specifically designed for small eukaryotic genomes with a gene density of around 20% (http://www.tigr.org) (Salzberg et al. 1999). GlimmerM contains a dynamic programming algorithm that considers all combinations of possible exons for inclusion in a gene model and chooses the best of these combinations. GlimmerM builds interpolated Markov models from a set of DNA sequences chosen for training. For coding regions, it builds three separate interpolated Markov models, one for each codon position. GlimmerM has been trained by the eukaryotic annotation team at TIGR for use with Aspergilli using a training set of 210 Aspergillus genomic sequences validated by mRNA or protein matches. This set contained 190 complete genes and 76 partial genes (13 were from A. fumigatus), consisting of 39 intronless genes, 722 exons and 532 introns, and 558 acceptor sites and 555 donor sites. The genes predicted by GlimmerM were edited manually, if necessary, according to the statistical data of A. fumigatus genes produced by Anderson et al. (2001).

Validation of identified genes

For validation of the identified A. fumigatus genes, a bi-directional best-hit analysis was performed using the polypeptide sequence of the predicted A. fumigatus ORFs as queries for BLASTp searches of protein databases at the Swiss Institute of Bioinformatics (http://www.ch.embnet.org; http://SwissProt/TrEMBL/TREMBL_NEW) and Saccharomyces protein sequences in the Saccharomyces genome database (http://genome-www.stanford.edu/Saccharomyces). A further analysis of the predicted genes was conducted using the polypeptide sequences of the predicted A. fumigatus ORFs as query sequences for InterProScan (http://www.ebi.ac.uk/interpro).

Results

Comparative analysis of the biosynthetic pathway for sphingolipids in four fungal species

We surveyed various databases for the enzymes or genes from four fungal species (S. cerevisiae, S. pombe, C. albicans, and N. crassa) for comparative analysis of the sphingolipid pathway. Knowledge of S. cerevisiae genes necessary for sphingolipid metabolism has increased many-fold since the genome sequence was released in 1996. At least one gene is known for most steps in S. cerevisiae sphingolipid metabolism, which is outlined in Fig. 1. The functions of gene products are summarized in Table 2. The defining feature of sphingolipids is an LCB that is amide-linked to a fatty acid to form a ceramide. The type of LCB and fatty acid differs considerably between organisms and is not uniform even within an organism (Warnecke and Heinz 2003).

Table 2 Genes for metabolism of S. cerevisiae sphingolipids (modified from Daum et al. 1998)
Full size table

Saccharomyces cerevisiae makes two types of LCB, dihydrosphingosine (DHS) and phytosphingosin (PHS), with an additional hydroxyl on C-4 (Fig. 1). Mammals contain small amounts of both DHS and PHS, however their primary LCB is sphingosine, which is DHS with a 4, 5-double bond. S. cerevisiae does not synthesize sphingosine but does respond in several ways to exogenous sphingosine (Birchwood et al. 2001).

Sphingolipid synthesis begins with the condensation of palmitol-CoA and serine to yield 3-ketodihydrosphingosine (3-ketosphinganine), which is reduced to yield DHS (Fig. 1). The condensation reaction is catalyzed by serine palmitoyltransferase (SPT), a membrane-bound enzyme that is encoded by two essential genes in S. cerevisiae, LCB1 and LCB2 (reviewed in Obeid et al. 2002). These two genes are also found in S. pombe, C. albicans,and N. crassa (Table 3). Recently, TSC3, a third gene required for optimal SPT activity, was identified (Gable et al. 2000). This gene belongs to the TSC family of temperature-sensitive suppressors of calcium sensitivity.

Table 3 Sphingolipid enzymes in four fungal species (in C. albicans and N. crassa, the gene identified by a best bi-directional blast hit to the yeast protein is represented by their database accession number with the prefix of CA and NCU, respectively. Databases: C. albicans, http://genolist.pasteur.fr/CandidaDB; N. crassa, http://www-genome.wi.mit.edu/annotation/fungi/neurospora)
Full size table

In the second step in S. cerevisiae sphingolipid synthesis, 3-ketodihydrosphingosine is reduced in a reaction utilizing NADPH to produce the LCB. TSC10 encodes 3-ketosphingosine reductase, a member of the short-chain dehydrogenase and reductase family, which is designated adh_short in the protein families database (Pfam). The adh_short domain was well conserved in the TSC10 homologue of S. pombe, C. albicans, and N. crassa (data not shown). LCBs are phosphorylated to LCB phosphates (LCBPs) by Lcb4p kinase. S. cerevisiae cells contain two types of LCBPs, dihydrosphingosine-1-phosphate (DHS-P) and phytosphingosine-1-phosphate (PHS-P) (Fig. 1). These LCBPs can be degraded by the phosphatase Lcb3p to yield LCBs. LCBPs can also be degraded by DPL1 lyase (Zhang et al. 2001). S. cerevisiae, C. albicans, and N. crassa, have the proteins necessary for LCBP metabolism (Lcb3p, Lcb4p, and Dpl1p), whereas S. pombe lacks DPL1. An S. cerevisiae deletion mutant of this enzyme displayed unregulated proliferation in the stationary phase, and was more resistant to killing by heat (Gottlieb et al. 1999; Zhang et al. 2001). This suggests that LCBPs regulate cell proliferation and are likely to play a role in heat stress. Zhang et al. (2001) reported that elimination of the DPL1 lyase and LCB3 phosphatase pathways by gene deletion in S. cerevisiae was lethal, indicating that these enzymes regulate LCBP levels to prevent accumulation. They also suggest that the C18 and C20 species of LCBPs are preferentially degraded by LCB3 phosphatase, whereas the DPL1 lyase prefers C16 DHS-P as a substrate. We propose that the breakdown of LCBPs in S. pombe occurs mainly through the LCB3 phosphatase pathway because it lacks DPL1.

In the next step in S. cerevisiae sphingolipid synthesis, DHS is converted to PHS. SUR2 is required for the hydroxylation of DHS at C-4 while SCS7 is required for hydroxylation of the very long chain fatty acid (VLCFA) (Haak et al. 1997)

Additionally, SUR2 deletion mutants are resistant to the fungal toxin syringomycin E (Grilley et al. 1998). The mechanism by which this occurs is unknown. However, Obeid et al. (2002) suggested that 4-hydroxylation could either directly modulate syringomycin binding by influencing sphingolipid exposure on the membrane surface, or indirectly affect syringomycin action by perturbing the lipid bilayer or by creating microdomains that facilitate ion channel formation. Both SUR2 and SCS7 are also found in S. pombe, C. albicans, and N. crassa (Table 3). This is consistent with the fact that the hydroxylation of C4 in the DHS is common in most fungal ceramides (Lester and Dickson 1993).

The next step in S. cerevisiae sphingolipid synthesis involves acylation of the LCB, PHS, to phytoceramide. This acylation involves VLCFAs, which are formed by enzymes encoded by the FEN1, SUR4 genes (Oh et al. 1997). Fenlp elongates fatty acids by up to 24 carbons, and Sur4p is essential for the conversion of C24 fatty acids to C26 fatty acids. There are three members of the family in S. cerevisiae, ELO1, FEN1/ELO2, and SUR4/ELO3. The Fen1p and Sur4p proteins are required for VLCFA synthesis, with Fen1p mediating elongation of C16 or C18 to C22 or C24, but unable to direct the synthesis of C26. Sur4p displays overlapping activity with Fen1p and is also able to take C24 to C26. It is impossible, on the basis of amino acid homology, to predict the substrate specificities of the Elop proteins. However, any fatty acid with a chain length greater than C18 is generated by fatty acid elongation, and while the Elop proteins display distinct specificities for the acyl-CoAs to be elongated, the other known components of the elongating system are used to process the product of the Elop-mediated step (a 3-keto intermediate) through a reduction, dehydration and second reduction to form the final product. Kohlwein et al. (2001) indicated that TSC13 gene encodes a protein required for elongation, possibly the enoyl reductase that catalyzes the last step in each cycle of elongation.

Although S. pombe, C. albicans, and N. crassa have the elongase complex including FEN1, SUR4 and TSC13 like S. cerevisiae (Table 3), C. albicans and N. crassa predominantly produce sphingolipids with C24 fatty acids (Wells et al. 1996). In fact, the C26 fatty acid in sphingolipids is unique to S. cerevisiae and is formed by a series of fatty acid synthesis and elongation steps with ACB1, ACC1, FAS1, FAS2, FEN1, SUR4, and TSC13 (Sims et al. 2004). Therefore, S. pombe, C. albicans, and N. crassa might be missing one or more of these genes.

The enzyme involved in acylating these VLCFAs onto the sphingolipids, ceramide synthase or sphingolipid base N-acyl transferase, requires either of two redundant genes, LAG1 or LAC1 (Guillas et al. 2001). The LAG1 gene was initially identified as a gene whose deletion endowed a longer lifespan phenotype on S. cerevisiae (D’Mello et al. 1994). Although there is little overall sequence similarity between LAG1 proteins among various species, all these proteins possess a stretch of 52 amino acids of high sequence similarity, which has been dubbed the LAG1 motif (Jiang et al. 1998). Comparison of the amino acid sequence of LAG1 homologue proteins from S. cerevisiae, S. pombe, C. albicans, and N. crassa showed that the N-terminal region had low similarity, but the LAG1 motif was detected in all four fungi following the central sequence region (data not shown).

Ceramides in S. cerevisiae are incorporated into complex sphingolipids that differ from mammalian sphingolipids in that the head group is comprised of an inositol phosphate instead of a choline phosphate. S. cerevisiae uses ceramides to make three types of complex sphingolipids: IPC, mannose-inositol-phosphorylceramide (MIPC), and mannose-(inositol-P)2-ceramide (M(IP)2C) (Table 4). This simplicity and the identification of many genes necessary for sphingolipid metabolism make S. cerevisiae an excellent organism for unraveling the roles of sphingolipids in signaling and in membrane structure and function. Ceramide is converted to IPC, the first of three so-called complex sphingolipids. Inositol phosphate is transferred from phosphatidylinositol to the C-1 OH group of ceramide (Fig. 1). This reaction is catalyzed by phosphatidylinositol:ceramide phosphoinositol transferase (IPC synthase), a membrane-bound enzyme (Becker and Lester 1980). The AUR1 gene encodes IPC synthase or a subunit of the enzyme (Nagiec et al. 1997). IPC synthase is a promising target for the development of antifungal drugs because this enzyme is not found in mammals and its inhibition leads to fungal cell death. The homologue for AUR1 is found in S. pombe, C. albicans, and N. crassa (Table 3), supporting the notion that the modification to C-1 of ceramide in fungi is common (Lester and Dickson 1993). IPC is mannosylated to yield MIPC, a reaction that requires the SUR1 and CSG2 genes (Beeler et al. 1997). The terminal step of sphingolipid synthesis in S. cerevisiae requires the IPT1 gene (Dickson et al. 1997) whose product presumably catalyzes the addition of inositol phosphate to MIPC to yield M(IP)2C. All the genes required for the biosynthesis of IPC, MIPC, and M(IP)2C from ceramide are only found in S. cerevisiae and C. albicans. This supports the idea that C. albicans has the same type of IPCs as S. cerevisiae (Table 4). The lack of IPT1 in S. pombe and N. crassa suggests that they have different types of IPC from those of S. cerevisiae.

Table 4 Fungal sphingolipids (modified from Dickson and Lester 1999)
Full size table

In summary, we propose that the sphingolipid biosynthetic pathway is well conserved in the four fungal species analyzed, although there are a few exceptions (Table 3). This suggests that the other pathogenic fungus, A. fumigatus, might have similar sphingolipid biosynthetic pathways.

Identification of A. fumigatus genes for sphingolipid biosynthesis

The S. cerevisiae genes for sphingolipid biosynthesis were used as query sequences in a BLAST analysis to identify homologous genomic regions in A. fumigatus. The next step was to identify the precise protein-coding regions within these homologous regions. The prediction of protein-coding regions within DNA sequences in eukaryotes is not easy, possibly because of the low coding density (probably as low as 2% in humans, although the protein-coding regions within these fungal genomes have densities of at least 40%) and the presence of introns within relatively short coding regions. Various weak signals are combined, such as GC bias, splice sites, and translational start and stop sites. Different knowledge-based methods complimented by similarity searches are applied to utilize them (Guigo 1997). Commonly used methods for eukaryotic gene prediction depend on training a computer program to recognize sequences that are characteristic of known exons in genomic DNA sequences. The GlimmerM program (Salzberg et al. 1999) was trained with the Aspergillus training set for Glimmer provided by TIGR to enable the prediction of the positions of exons in genomic sequences and their arrangement into a predicted gene structure. We used the trained GlimmerM program to predict the positions of genes within regions of the A. fumigatus genome that were identified by BLAST searching with S. cerevisiae enzymes for sphingolipid biosynthesis.

Our results show that both CSG2 and IPT1 homologues of S. cerevisiae were not detected in the A. fumigatus genome (Table 5). The function of CSG2 is not obvious, but it is required for the mannosylation of IPC to MIPC together with SUR1. Daum et al. (1999) suggested that these two genes are functionally redundant or that there is an unidentified pathway for making M(IP)2C because deletion of either gene reduces the level of MIPC and M(IP)2C in S. cerevisiae. Therefore, it is possible that the mannosylation of IPC to MIPC in A. fumigatus is catalyzed by SUR1 alone, or an alternative enzyme exists that has a novel sequence compared to other fungal CSG2 genes.

Table 5 The predicted genes involved in sphingolipid metabolism in the A. fumigatus genome and the results of their best hit using BLASTp against the Saccharomyces proteome and InterProScan
Full size table

Another gene, IPT1, was also not identified in the A. fumigatus genome, nor in S. pombe and N. crassa (Tables 3, 5). M(IP)2C has been reported in S. cerevisiae (Steiner et al. 1969) and C. albicans (Wells et al. 1996) but not in N. crassa and A. fumigatus. This could be due to the lack of M(IP)2C synthase, encoded by IPT1. Ipt1p shows 27% amino acid identity to Aur1p, IPC synthase, over a region of 365 amino acids and both proteins use phosphatidylinositol as a substrate. Ipt1p and Aur1p are inhibited in S. cerevisiae by aureobasidin A, one of the strongest known fungicides, produced by Aureobasidium pullulans (Dickson et al. 1997). Aureobasidin A shows strong fungicidal activity against many pathogenic fungi including Candida spp., Cryptococcus neoformans, and some Aspergillus spp., but not A. fumigatus (Takesako et al. 1993). Zhong et al. (2000) suggests that the resistance in A. fumigatus is due to increased efflux of aureobasidin A by some transporter(s).

The fatty acid elongase complex including FEN1, SUR4 and TSC13 was identified in the A. fumigatus genome (Table 5). However, Levery et al. (2001) reported that A. fumigatus has mainly C24 fatty acids in its ceramides, as do C. albicans and N. crassa.

The glycosylceramides produced by fungi were represented in Table 4. The majority of fungi synthesize only glucosylceramide (Cer-Glc), but among Aspergillus species the expression of galactosylceramide (Cer-Gal) (A. niger), glucosylceramide (A. oryzae; A. versicolor; A. fumigatus), or mixture of both (A. fumigatus) has been reported (Toledo et al. 1999). This suggests that A. fumigatus might have the enzymes responsible for the synthesis of glucosylceramides, for example, sphingolipid-delta4-desaturase, sphingolipid-delata8-desaturase, glucosylceramide synthase. We could find the homologue of the C. albicans sphingolipid-delta4-desaturase in contig 4826 (742,151–744,788), but there is no homologue for sphingolipid-delta8-desaturase of C. albicans. Each homologue of Arabidopsis thaliana sphingolipid-delta8-desaturase and glucosylceramide synthase was found in contig 4917 (98,876–100,236) and contig 4938 (742,151–744,788), respectively.

Identification of the predicted genes of A. fumigatus for sphingolipid biosynthesis was validated by performing a reciprocal BLASTp search against the Saccharomyces proteome database. Using the predicted polypeptide sequence encoded by A. fumigatus genes as the query sequence, the predicted proteins for sphingolipid biosynthesis were identified (Table 5).

Each gene that was found to be involved in A. fumigatus sphingolipid biosynthesis was further characterized by computational functional analysis using InterProScan. All the identified genes had a significant hit for a protein family or domain and could be classified by this function or functional domain (Table 5).

Discussion

The complete sequence of the S. cerevisiae genome has advanced our knowledge of sphingolipid metabolism. At least one gene has now been identified for each step in yeast sphingolipid metabolism (Fig. 1). We have carried out the first trial to identify the genes necessary for sphingolipid biosynthesis in A. fumigatus using comparative pathway analysis and the gene-predicting program GlimmerM.

Most of the genes involved in S. cerevisiae sphingolipid biosynthesis are present in the A. fumigatus genome, except for CSG2 and IPT1 (Table 5). The CSG2 is involved in the mannosylation of IPC to MIPC, and the IPT1 is involved in the synthesis of M(IP)2C from MIPC. Levery et al. (2001) demonstrated that A. fumigatus and A. blazei contain dimannosylinositol phosphorylceramide (MMIPC) and MIPC, respectively. Thus, the mannosylation of IPC in A. fumigatus may be required and could occur via Sur4p or/with an unidentified enzyme having same or similar function to Csg2p.

A. fumigatus also lacks IPT1, inositol phosphoryl transferase, which is required for the synthesis of M(IP)2C from MIPC. IPT1 is also absent in S. pombe and N. crassa (Table 3), which may explain why M(IP)2C has not been reported in A. fumigatus, S. pombeor N. crassa.

Even though the fatty acid elongase complex including FEN1, SUR4 and TSC13 has been identified in S. cerevisiae, S. pombe, C. albicans, N. crassa, and A. fumigatus, the C26 fatty acids are mainly produced in S. cerevisiae ceramides. This suggests that only S. cerevisiae has the complete set of enzymes required for the synthesis of C26 fatty acids.

Our results show that genes in A. fumigatus could reliably be detected using GlimmerM, which is in good agreement with our previous work (Do et al. 2004). This study was limited in computational analysis; however, it would be useful as support in the verification of genes in the sphingolipid pathway in A. fumigatus.