Main

Encephalitozoon cuniculi infects various mammals, including humans, and can cause digestive and nervous clinical syndromes in HIV-infected or cyclosporine-treated people1. Its reproduction proceeds as a sequence of two major stages: merogony, involving the multiplication of large, wall-lacking cells (meronts); and sporogony, leading to small, thick-walled spores. The sporal invasive apparatus is characterized by a long polar tube that can be quickly extruded then used for transferring the sporoplasm into the target cell. Consisting of 11 linear chromosomes ranging from 217 to 315 kb, the E. cuniculi genome is remarkably reduced (∼2.9 Mb)10. The nucleotide sequence of the smallest chromosome has been recently reported11. The full sequencing of this minimal genome among eukaryotes was expected to provide insight into the metabolism and general biology of microsporidia and to help in the understanding of the evolutionary history of amitochondriate eukaryotes currently considered ‘curious fungi’9 (see Table 1).

Table 1 General features of the E. cuniculi genome
Full size table

The chromosome sequences were determined through a plasmid library (∼3 kb inserts) and a miniBAC library (∼20–25 kb inserts) totalling 15 genome equivalents (∼46 Mb). All chromosomes possess a ‘unique sequences’ core region flanked by two 28-kb divergently oriented regions, each including one ribosomal DNA unit11,12. The subtelomeric repeats upstream from rDNA are mostly degenerated minisatellites, whereas downstream repeats consist essentially of non-polymorphic microsatellite arrays. The chromosome cores lack simple sequence repeats, minisatellite arrays and known transposable elements. No imprint of retrogenes or pseudoretrogenes of either polymerase (pol) II or pol III type was found. General features of genome organization are indicated in Table 1. The gradual increase in G + C content at the core centre described in chromosome I (chrI)11 exists in all the chromosomes (maximum 51.0%). The 1,997 protein-coding DNA sequences (CDSs) represent about 90% of the chromosome cores, as a result of generally short intergenic regions (see Supplementary Information). Gene density is slightly lower than that observed in the nucleomorph genome of the cryptomonad Guillardia theta13. Only about 44% of the CDSs are assigned to functional categories and about 6% to conserved hypothetical proteins (Fig. 1). In contrast to the nucleomorph genome13, no overlapping of CDSs with predicted functions was revealed. Structural or functional clusters are rare and never composed of more than two CDSs (for example, histones H3 and H4 on chrIX). Genome compaction can also be related to gene shortening, as indicated by the length distribution of all potential proteins (Fig. 2a). The mean and median lengths of all potential E. cuniculi proteins are only 359 and 281 amino acids, respectively. We compared the lengths of 350 proteins with Saccharomyces cerevisiae homologues (Fig. 2b). More than 85% of these proteins are shorter than in yeast, with a mean relative size difference of 14.6%. From the analysis of the protein size distributions derived from sequenced genomes, it has been suggested that the lengthening of proteins in eukaryotes (non-parasitic species) allows for more complex regulation networks14. Thus, protein shortening in E. cuniculi may reflect reduced protein–protein interactions as a result of various gene losses linked to the intracellular parasitic nature (Fig. 2a, b).

Figure 1: Distribution of predicted E. cuniculi proteins among functional groups.
figure 1

The corresponding gene list is given in the Supplementary Information.

Full size image
Figure 2: Sizes of E. cuniculi (Ec) proteins and comparison with S. cerevisiae (Sc) homologues.
figure 2

a, Distribution of the lengths of all the potential E. cuniculi protein chains. Only six have more than 2,000 amino acids (maximum 3,456 amino acids). b, Degrees of reduction in length of E. cuniculi proteins (n = 350) relative to those of S. cerevisiae, expressed as a percentage: 100(Sc protein length - Ec protein length)/(Sc protein length). The positive classes representative of shorter E. cuniculi proteins are in grey. The dynein heavy chain, the largest protein chain with a clearly predicted function (3,151 amino acids), is ∼23% shorter than the yeast homologue (4,092 amino acids). Mean values associated with major functional categories are 7.3% for ‘protein synthesis’, 11.0% for ‘protein destination’, 11.4% for ‘metabolism/energy’, 14.4% for ‘intracellular transport’, 15.3% for ‘transcription’ and 20.1% for ‘cell growth, cell division and DNA synthesis’.

Full size image

Perfect segmental duplications of 0.5–10-kb coding and non-coding sequences represent about 3.7% of the core region in average (from ∼1% in chrII, III, IV and VI to ∼7% in chrIX; see Supplementary Information). A segment carrying four enzyme-coding genes is perfectly duplicated in the extreme part of the chrI core11 but partially duplicated near the end of chrVIII (truncated serine hydroxymethyltransferase gene). The genome core sequences exhibit a very low base polymorphism, restricted to a few positions in eight chromosomes. A duplicated gene homologous to CTP synthases (chrXI) revealed a rare polymorphic tract of 189 base pairs (bp) (116,847–117,036). Hybridization experiments indicate that each polymorphic sequence is present in about 50% of the DNA molecules (data not shown), suggesting that they are alleles and supporting the diploidy hypothesis10,12.

A large proportion of CDSs is assigned to the conservation and transmission of genetic information as well as to protein modification and intracellular transport processes (Fig. 1), but with a significant degree of simplification, mainly related to the lack of DNA-containing organelles. For example, subunits for DNA pol I (δ), pol II (ε) and pol III (α but not β) are present while there is no candidate for the mitochondrial DNA polymerase γ. Apart from the telomerase catalytic subunit, no RNA-dependent reverse transcriptase was found. This and the lack of any retrotransposition elements may explain the absence of pseudoretrogenes. The potential transcription machinery includes RNA pol I, II and III, more than 70 messenger RNA transcription-associated proteins and a virtually complete set of genes for splicing, 5′ and 3′ processing. An (A + T)-rich consensus transcription initiation sequence is revealed in the 120-bp region upstream from the initiation codon of numerous genes, suggesting short 5′ leaders. Seventy different eukaryotic-type ribosomal proteins (40 from the large subunit, 30 from the small subunit) are predicted, which is slightly lower than in the amitochondriate protist Giardia lamblia (74)15. Compared with the cytoplasmic ribosome of S. cerevisiae, the missing proteins are LP1, L14, L29, L38, L40, L41, S27 and S27A. Small putative spliceosomal introns with usual GT–AG boundaries were detected in 11 ribosomal protein genes (S8, S17, S24, S26, L5, L19, L27A, L37, L37A and L39). They start either in the initiator ATG or the next codon, as often observed in yeast16 or nucleomorph13 genomes. Two more internal introns create frame shifts within a CDP-diacylglycerol serine phosphatidyltransferase gene (chrXI). Two of the 44 transfer RNA genes (tDNAIle and tDNATyr) also harbour a small intron.

Reduced metabolic capacities and low diversity of transporters can be inferred from the genome sequence, as illustrated in Fig. 3. The repertoire for the biosynthesis of amino acids is restricted to asparagine synthetase and serine hydroxymethyltransferase genes. Genes for de novo biosynthesis of purine and pyrimidine nucleotides are absent but several nucleotide interconversions are predicted. Genes encoding a fatty acid synthase complex are lacking, which supports the uptake of host-derived fatty acids17. The E. cuniculi spore membrane contains cholesterol. This sterol might also be of host origin, as no gene for the conversion of farnesyl-PP into cholesterol was detected. In contrast, E. cuniculi seems to be capable of synthesizing usual membrane phospholipids. Genes for principal enzymes for the synthesis and degradation of trehalose confirm that this disaccharide could be the major sugar reserve in microsporidia18, as in other fungi. A complete glycolytic glucose-to-pyruvate pathway is predicted. In contrast, genes required for the tricarboxylic acid cycle, fatty acid β-oxidation, respiratory electron-transport chain and the F0F1-ATPase complex are absent. Thus, ATP production in microsporidia would be possible by substrate-level phosphorylation only. As proliferating microsporidia recruit host mitochondria near their plasma membrane, it has been proposed that these parasites could use host-derived ATP18. This is reinforced by the finding of four genes encoding ADP/ATP carrier proteins that are homologous to ADP/ATP translocases from chloroplasts and obligate intracellular bacteria (Rickettsia and Chlamydia) capable of importing host ATP19. The fate of pyruvate remains difficult to predict because of the lack of genes for lactate and ethanol fermentation as well as for glycolysis reversal. A potential cytosolic glycerol-3P dehydrogenase (GPDH) might serve to reoxidize the NADH produced during glycolysis. Surprisingly, two CDSs have significant similarity to the subunits of the E1 component of the mitochondrial pyruvate dehydrogenase complex. Pyruvate decarboxylation could be inferred, but, in the absence of evidence for E2 and E3 components, a subsequent production of acetyl coenzyme A cannot be concluded.

Figure 3: An overview of metabolism and transport in E. cuniculi, as deduced from genome sequence analysis.
figure 3

Pathways for nucleotide biosynthesis, energy production and chitin biosynthesis are indicated. Endocytosis and vesicular transport involving a cis–trans polarized Golgi apparatus are also illustrated. Potential transporters associated with the plasma membrane are shown with indications on substrate specificity. Question marks correspond to major uncertainties about the fate of pyruvate and the production of second messengers for signal transduction. The parasite is represented within a parasitophorous vacuole (PV) of the host cytoplasm.

Full size image

Microsporidia have a presumably simplified Golgi apparatus in which a cis–trans polarity is not cytologically distinguishable but that is central in sporogony-specific secretion processes1. The spore wall protein SWP1 (ref. 20) is encoded by a unique gene on chrX whereas two genes for the polar tube proteins PTP1 and PTP2 are arranged in tandem on chrVI21. The set of chaperones for protein folding includes a complete oligomeric TCP-1 complex, four members of the HSP70 system but no chaperonin CPN60. The mitochondrial-type HSP70 has been previously characterized in three different microsporidian species3,4,5. Initial steps of N-glycosylation using UDP–N-acetylglucosamine (UDP-GNAc) and GDP-mannose (GDP-Man) may occur, but further trimming by mannosidases associated with endoplasmic reticulum or Golgi apparatus and formation of a complex N-linked oligosaccharide are not supported. The major sugar used for O-linked glycosylations would be mannose (two mannosyltransferases of the fungal PMT family). The lack of genes for the two enzymes involved in phosphorylation of mannose residues argues for the absence of sorting of lysosomes. Membrane fusion and recognition of some target membrane processes are sustained by a restricted set of potential proteins for Golgi and post-Golgi trafficking. The constitutive secretion pathway leading to the plasma membrane may involve some characteristic Rab proteins. Several potential partners for trans-Golgi and endosome transport include β1-adaptin, Vps1-like dynamin and vacuolar protein sorting-associated proteins, confirming that the Golgi-like apparatus is functionally polarized. Endocytosis of certain macromolecules was previously shown in Spraguea lophii sporoplasms18, when maintained in vitro in a cell culture medium. Several genes are also suggestive of an endocytosis pathway in E. cuniculi. This might drive the internalization of macromolecular ligands representing sources of fatty acids, cholesterol or iron (see Fig. 4).

Figure 4: Conceptual scheme of a mitochondrion-derived organelle (‘mitosome’) in E. cuniculi suggested by the detection of several homologues of mitochondrial proteins.
figure 4

Under this hypothesis, pyruvate decarboxylation may occur through a heterotetrameric form (α2β2) of the pyruvate dehydrogenase E1 component (PDH-E1) and transfer of reducing power towards the organelle transits through a glycerol-3-phosphate shuttle involving both cytosolic (GPDH-C) and mitochondrial (GPDH-M) glycerol-3-phosphate dehydrogenases. By analogy with hydrogenosomal pyruvate:ferredoxin oxidoreductase, a system based on ferredoxin (Fdx) and NAD(P)H ferredoxin:oxidoreductase (FOR) is assumed to be used for acetate production. A cytosolic acetyl coenzyme A (CoA) synthetase (ACS1) may catalyse acetate activation. Considering an aerobic environment and the lack of hydrogenase, a simplified but specific electron transport towards molecular oxygen remains a possibility. The predicted manganese superoxide dismutase (Mn-SOD) would ensure protection against oxygen radicals. A homologue of yeast ERV1 (a small protein required for mitochondrial biogenesis) is indicated. In the lower part of the scheme are depicted some of the major potential factors required for targeting of proteins to the organelle and for biosynthesis of Fe–S clusters and transport of Fe–S proteins towards the cytosol. Iron is predicted to be essential for controlling the expression of mitosomal proteins.

Full size image

The evolutionary origin of microsporidia has been much debated but strong evidence supporting a fungal origin of these organisms has recently accumulated6,7,8,9. The present genome sequence extends this evidence: phylogenetic analyses of putative genes for seryl-tRNA synthetase, transcription initiation factor IIB, subunit A of vacuolar ATPase, and a GTP-binding protein place microsporidia as a sister group of fungi with bootstrap supports ranging from 70% to 92% (see Supplementary Information; a systematic phylogenetic analysis of the genome will be presented elsewhere). Genes of putative mitochondrial evolutionary origin in the E. cuniculi genome were systematically sought by comparison with the 423 recently surveyed yeast mitochondrial proteins encoded by the nuclear and the mitochondrial genomes22. Twenty-two genes with significant similarity to the yeast genes were identified and phylogenetic analysis showed that six of them are closely related to homologues from α-proteobacteria, the bacterial group from which mitochondria are believed to derive23. The yeast homologues of these six proteins are ATM1 (ABC transporter), ISU1/ISU2 (NIFU-like protein), NFS1 (unique homologue of bacterial ISC-S and NIF-S), SSQ1 (heat-shock protein of relative molecular mass 70,000), YAH1 (ferredoxin) and PDB1 (β-subunit of pyruvate dehydrogenase component E1). The first five proteins are typically involved in the Fe–S cluster assembly machinery, an essential function of mitochondria24.

The presence of characteristic domains and key amino acids suggests that the potential mitochondrial-type proteins are functional. Moreover, PSORT analysis predicts amino-terminal presequences for the targeting of five of these proteins (see Supplementary Information). A common feature is an arginine residue at -2 relative to the cleavage site, similar to presequences of mitochondrial and hydrogenosomal proteins25. The amitochondriate protozoan Entamoeba histolytica has recently been shown to contain a residual mitochondrion-derived organelle26,27,28 that some authors have called mitosome28. Considering the set of potential E. cuniculi proteins usually associated with mitochondria, we propose that a cryptic organelle is also present in microsporidia (Fig. 4). This mitosome would be significantly different from hydrogenosomes found in several anaerobic unicellular eukaryotes (type II anaerobes)29. Hydrogen production through pyruvate catabolism seems unlikely in microsporidia (because of a lack of a hydrogenase gene). The development of microsporidia in various aerobic host cells is suggestive of a rather high O2 tolerance and therefore of an efficient protection against oxidative stress. No catalase gene is identified, in agreement with the lack of peroxisomes. Thus, in addition to glutathione and thioredoxin-based systems, E. cuniculi might use its unique manganese superoxide dismutase as an antioxidant.

This first report of the genome sequence of a eukaryotic parasite should stimulate proteomic approaches to identify gene products of interest for diagnosis and therapy of microsporidioses, as well as to test the mitosome hypothesis. In addition, we expect the E. cuniculi genome to provide a useful reference for comparative genomics of microbial eukaryotes, particularly to identify the relative importance of shared genes among various evolutionarily distant intracellular parasites, including major human-infecting parasites such as Plasmodium and Leishmania.

Methods

The reference mouse isolate of Encephalitozoon cuniculi (GB-M1), cloning and library construction, nucleotide sequencing, sequence validation and sequence analysis have been described in detail previously11 (see Supplementary Information). Further information on recombinant DNA preparation, sequencing and sequence analysis can be found elsewhere30. Annotation was manually performed with the help of AceDB and Artemis graphic interfaces. Genes were characterized by Glimmer prediction of coding DNA sequences, combined with BLAST all-homology results (BLAST X, BLAST N against ‘nr’, BLAST P against SwissProt, PSI-BLAST). Transfer RNA genes were detected with the tRNA Scan program. Spliceosomal-type introns were manually detected. Phylogenetic analyses were done as in Keeling et al.8. Gamma-corrected ML distances with eight rate categories and invariant sites were computed with TREE-PUZZLE version 5.0 and trees were derived with BioNJ. Bootstrapping was on 500 replicates with alpha parameters and the fraction of invariant sites estimated once from the original data. Sequences of individual chromosomes were submitted to EMBL under the accession numbers AL39173 for chrI and AL590442–AL590451 for chrII–chrXI, respectively.