The genome of the laboratory reference S mat+ strain was sequenced using a whole-genome shotgun approach (see Materials and methods for a detailed explanation of the sequencing and assembly strategies). Ten-fold coverage permitted complete assembly of the mitochondrial genome as a single circular contig of about 95 kb and most of the nuclear genome (Table 3). The latter was assembled in 1,196 contigs clustered into 33 large scaffolds, comprising nearly all unique sequences, and 45 small scaffolds composed almost exclusively of transposon sequences, collectively totaling 35 Mb. Based on the frequency of sequence runs corresponding to the rDNA compared to that of unique sequences, we estimated that 75 rDNA units are present in the genome. With this assumption, the total sequence length of the genome is 35.5-36 Mb, a value somewhat superior to pulse field estimates 2829. Presently, all large scaffolds are assigned to a chromosome as defined by the genome map that now includes over 300 markers (see Materials and methods; Additional data file 1).

The annotation strategy, described in the Materials and methods section, identified 10,545 putative coding sequences (CDSs), including two inteins 30. 5S rRNA, tRNA, as well as several small nuclear RNAs (snRNAs) and small nucleolar RNAs (snoRNAs) were also identified. Statistics concerning the protein coding capacity of the P. anserina genome and the main features of the CDSs are indicated in Table 3. The present estimates of the coding capacity of N. crassa are 9,826 CDSs at the Broad Institute 31 and 9,356 CDSs at the Munich Information Center for Protein Sequences (MIPS) 32. It remains to be established whether the higher coding capacity of P. anserina is real or due to the differences in strategies used to annotate the genomes of these fungi. We have searched for orthologous genes between P. anserina, N. crassa, M. grisea and A. nidulans by the best reciprocal hit method and found that these four fungi share a common core of 2,876 genes (Figure 2a). Comparison of the P. anserina CDSs with N. crassa orthologues (Figure 2b) indicates that they are, on average, 60.5 ± 16.0 percent identical, a value similar to the one calculated previously on a small sample 24. The P. anserina CDSs were 54.7 ± 15.8% identical to M. grisea and 47.9 ± 15.1% to A. nidulans orthologues. The identities reflect the known phylogenetic relationship between these four pezizomycotina and are comparable to those found between species of saccharomycotina 9.

In addition to genomic DNA sequencing, a collection of 51,759 cDNAs was sequenced. These originate from libraries constructed at different stages of the P. anserina life cycle (Table 4). The resulting expressed sequence tags (ESTs) were mapped on the genomic sequence to help with the annotation but also to gain insight into the transcriptional ability of P. anserina. As seen in Table 4, these cDNAs confirmed 5,848 genes. However, we detected alternative splicing events in 3.8% of the clusters. This suggests that the P. anserina proteome might be more complex than concluded from the present annotation. Of interest is the presence of 668 transcribed regions without obvious protein-coding capacity (designated here as 'non-coding transcripts'). Some of these produce ESTs that are spliced, poly-adenylated or present in multiple copies, suggesting that they originate from true transcription units. Although some genes may have been miss-called during annotation, these transcription units may correspond to transcriptional noise, code for catalytic/regulatory RNA or reflect polycistronic units coding for small peptides as described recently 3334. Finally, we detected 45 antisense transcripts corresponding to 36 different CDSs. These transcripts might potentially be involved in proper gene regulation, as described for the S. cerevisiae PHO5 gene 35. In large scale analyses of Fusarium verticilloides 36 and S. cerevisiae 37 ESTs, similar arrays of alternatively spliced, 'non-coding' and antisense transcripts were detected, suggesting that the production of these 'unusual' transcripts is, in fact, a normal situation in ascomycete fungi, as described for other eukaryotes 38.

During the manual annotation of the genome, we detected 14 genes possibly requiring a frame-shift or a read-through to be properly expressed (Additional data file 2). In all cases, sequencing errors were discounted. In addition, ESTs covering putative read-through or frame-shift sites confirm six of them. Some of the putative frame-shifts and read-throughs detected could correspond to first mutations that will lead to pseudogene formation. However, four sites seem conserved during evolution, arguing for a physiological role. One of the putative -1 frame-shift sequences is located in the Yeti retrotransposon, a classic feature of this type of element. The 13 others affect genes coding for cellular proteins. Factors involved in the control of translation fidelity and affecting rates of frame-shift and read-through have been studied in P. anserina and shown to strongly impact physiology 39404142. To date, the reasons for these effects are not known. None of the components responsible for insertion of selenocysteine are found in the P. anserina genome, excluding a role in the observed phenotypes of the non-conventional translation insertion of this amino acid, which takes place at specific UGA stop codons 43. Similarly, no obvious suppressor tRNA was discovered in the genome.

We have explored in more detail the synteny between orthologous genes in the P. anserina and N. crassa genomes (Figures 3 and 4). Synteny was defined as orthologous genes that have the same order and are on the same DNA strand. As observed for other fungal genomes 1844, extensive rearrangements have occurred since the separation of the two fungi. However, most of them seem to happen within chromosomes since a good correlation exists between the gene contents of many chromosomes, even though a few translocations are detected (Figure 3). For example, most of P. anserina chromosome 1 corresponds to N. crassa chromosome I except for a small part, which is translocated to the N. crassa chromosome IV. Within the chromosomes, numerous rearrangements have occurred, compatible with the prevalence of small inversions in fungal genome evolution as observed previously between genes of saccharomycotina (hemiascomycetous) yeasts 45. The size of the synteny blocks loosely follows an exponential decrease (Figure 4), compatible, therefore, with the random breakage model 46, suggesting that most breaks occur randomly, as observed for genome evolution in Aspergilli 18. However, in both Aspergilli and saccharomycotina yeasts, blocks of synteny have been dispersed among the various chromosomes 1847, unlike what is observed between P. anserina and N. crassa. This discrepancy of genome evolution between the three groups of fungi might stem from the fact that P. anserina and N. crassa have likely had a long history of heterothallism, whereas Aspergilli and saccharomycotina yeasts are either homothallic, undergo a parasexual cycle or switch mating types. In heterothallics, the presence of interchromosomic translocation results in chromosome breakage during meiosis and, hence, reduced fertility. On the contrary, homothallism, parasexualilty or mating-type switching may allow translocation to be present in both partners during sexual reproduction and, therefore, have fewer consequences on fertility. Additionally, meiotic silencing of unpaired DNA (MSUD), an epigenetic gene silencing mechanism operating in N. crassa 48, abolishes fertility in crosses involving rearranged chromosomes in one of the partners.

Interestingly, the largest synteny block between P. anserina and N. crassa, with 37 orthologous genes, encompasses the mating type, a region involved in sexual incompatibility. A similar trend in conserved synteny in the mating-type region has been observed in the genus Aspergillus 18. This suggests that recombination may be inhibited in this region on an evolutionary scale. In both P. anserina and N. crassa, the mating-type regions are known to display peculiar properties. In P. anserina, meiotic recombination is severely repressed around the mating-type locus 49, as also described in Neurospora tetrasperma 50. In N. crassa, MSUD is inhibited in the mat region 48. However, recombination is not completely abolished around this locus. Indeed, between pairs of orthologous genes, a few species-specific CDSs were detected. These genes may come from de novo insertion or, alternatively, these species-specific genes have been lost in the other species. This lends credit to the hypothesis put forward to explain the mating-type region of Cryptococcus neoformans 51, in which the genetic incompatibility is driven by two genetically different sequences of 100 kb. In these regions, not only the mating-type regulatory genes are different, but also housekeeping genes. Inhibition of recombination at this locus may have driven the differential acquisition of genes by the two haplotypes within the same species. Note that on a longer evolutionary scale, inhibition of recombination cannot be detected because the synteny of the mating-type region of P. anserina with that of M. grisea or A. nidulans is absent or limited to very few genes.

The pilot project that sequenced about 500 kb around the centromere of chromosome 5 revealed an apparent paucity in repeated sequences in P. anserina 24. The draft sequence reported here confirms a paucity of repeats but not as much as suggested by the pilot project. In fact, repeats cover about 5% of the P. anserina genome (omitting the rDNA cluster). They can be divided into four categories: RNA genes (Table 3; see Materials and methods), true transposons (Additional data file 3), repetitive elements of unknown origin (Additional data file 3) and segmental duplications (Additional data file 4). Collectively, the transposons occupy about 3.5% of the genome. However, as many transposons border the sequence gaps present in the draft assembly, the actual percentage in the complete genome may be higher. This is about three times less than in the genomes of M. grisea 11 and N. crassa 16, close relatives of P. anserina. Most segmental amplifications are small (Additional data file 4), although one is 20 kb large. They occupy about 1.5% of the genome. An interesting feature of all these repeated sequences (except for the 5S RNA and tRNA genes) is that they are nested together (Figure 5), as previously described for Fusarium oxysporum transposons 52. In particular, large parts of many chromosomes are almost devoid of these repeated sequences whereas chromosome 5 is enriched in repeats. Ironically, the pilot project sequenced a region of this chromosome 5 almost devoid of repeated sequences.

Nearly all copies of these repeated elements differ by polymorphisms, many of which appear to be caused by repeat induced point mutation (RIP). RIP is a transcriptional gene silencing and mutagenic process that occurs during the sexual dikaryotic stage of many pezizomycotina 53. P. anserina displays a very weak RIP process 5455. It results, as in N. crassa, in the accumulation of C●G to T●A transitions in duplicated sequences present in one nucleus, and, therefore, 'ripped' sequences present a higher than average T/A content. However, although the RIP process acts in the P. anserina genome, it does not account for all the mutations found in these inactivated paralogues. For example, the copies of 'rainette', the last transposon to have invaded the P. anserina genome (Additional data file 3), differ by 30 polymorphic sites. Twenty-five of them (83%) were C●G versus T●A polymorphisms and may, therefore, be accounted for by RIP, while the five others (17%) cannot. A reciprocal ratio was observed in other instances as seen for the largest segmental triplication with two copies present on chromosome 5 and one on chromosome 1. The three members share a common region of about 9 kb. In this region they differ by numerous indels and in about 20% of their nucleotides. More precisely, in the 4,000 nucleotide-long core region where the three sequences can unambiguously be aligned, there are 1,341 polymorphic sites in which at least one sequence differs from the others. For 418 of them (31%), two members have a C●G polymorphism whereas the other has a T●A polymorphism, strongly suggesting that these polymorphisms may originate from RIP, whereas for the remaining 923 (69%), the variations are small indels or single nucleotide variations not accounted for by RIP. Therefore, in the case of rainette, RIP polymorphisms are foremost, whereas for the triplication, non-RIP polymorphisms are more frequent. This is compatible with a model in which RIP occurs first and is then followed by accumulation of other types of mutations.

Overall, these data suggest that P. anserina has experienced a fairly complex history of transposition and duplications, although it has not accumulated as many repeats as N. crassa. P. anserina possesses all the orthologues of N. crassa factors necessary for gene silencing (Additional data file 5), including RIP, meiotic MSUD 48 and also vegetative quelling, a post transcriptional gene silencing mechanism akin to RNA interference 56. However, to date, no MSUD or quelling has been described in P. anserina, despite the construction of numerous transgenic strains since transformation was first performed 57. Surprisingly, the DIM-2 DNA methyltransferase 58, the RID DNA methyltransferase-related protein 59 and the HP1 homolog necessary for DNA methylation 60 described in N. crassa are present in the genome of P. anserina. Although the P. anserina orthologues of these two proteins seem functional based on the analysis of the conserved catalytic motifs, no cytosine methylation has been reported to occur in this fungus 54. A possibility would be that methylation is restricted to a specific developmental stage or genomic region that has not yet been investigated. Overall, the apparent absence (quelling and MSUD) or lack of efficiency (RIP) of these genome protection mechanisms in P. anserina questions their true impact on genome evolution, especially since this fungus contains less repeated sequences than N. crassa. Maybe the life strategy of P. anserina makes it less exposed to incoming selfish DNA elements, therefore diminishing the requirement of highly efficient gene silencing mechanisms. Supporting this assumption is the fact that, although heterothallic, formation of ascospores makes P. anserina pseudo-homothallic (Figure 1), with seemingly very little out-crossing 61, whereas N. crassa is strictly heterothallic and presents a low fertility in crosses between closely related strains 62.

The detection of segmental duplications raised the question of whether new genes evolved through duplication in the lineage that gave rise to P. anserina. It is known that creating new genes through duplication in N. crassa, in which RIP is very efficient, is almost impossible 16. On the contrary, RIP is much less efficient in P. anserina; in particular, RIP is absent in progeny produced early during the maturation of the fructifications 55. In addition, the mutagenic effect of RIP is very slight since it has been estimated that at most 2% of cytosines are mutated when RIP affects duplicated sequences present on two different chromosomes 63. We previously reported that some thioredoxin isoforms were encoded by a triplicated gene set in P. anserina as compared to N. crassa 64, showing that gene duplications can indeed generate new genes in P. anserina. However, thioredoxins are small proteins encoded by small genes. To test if large genes were duplicated, we performed a three-way comparison between the P. anserina, N. crassa and M. grisea putative CDSs and screened for P. anserina CDSs that show a best hit with another P. anserina CDS to the exclusion of proteins from N. crassa and M. grisea. Such CDSs may originate from duplication that occurred in the P. anserina lineage after its divergence from N. crassa. In this analysis, small genes were excluded because the putative candidates were selected on the basis of an e-value of less than 10^-190 in Blast comparison against the database containing the three predicted proteomes (as a consequence, the thioredoxin genes were not included in the set).

To confirm that the candidates recovered indeed originated from recent duplications, phylogenetic trees were constructed with the CDSs from P. anserina, N. crassa, M. grisea and additional fungal CDSs. In some instances, the trees confirmed a recent duplication event in the P. anserina lineage after the split between P. anserina and N. crassa, because the phylogenetic analysis clustered the P. anserina paralogues with high statistical confidence. Figure 6 shows the trees obtained for three such couples of paralogues, for example, genes coding for putative alkaline phosphatase D precursors (Pa_4_1520 and Pa_6_8120; Figure 6a), putative HC-toxin efflux carrier proteins related to ToXA from Cochliobolus carbonum (Pa_2_7900 and Pa_6_8600; Figure 6b) and putative chitinases related to the killer toxin of Kluyveromyces lactis (Pa_4_5560 and Pa_5_1570; Figure 6c). Overall, our analysis detected an initial set of 33 putative duplicated gene families, including the het-D/E gene family, whose evolutionary history has been reported elsewhere 65. Among these, at least nine (including the het-D/E genes) have duplicated recently. However, some additional recent duplication events may have occurred but are not supported with sufficient statistical confidence to differentiate between recent duplications followed by rapid divergence, and ancient duplications (see Figure 6c for an example of such duplications with putative chitinases). The fact that large genes may duplicate in P. anserina is not contradictory to the presence of RIP, since if RIP may inactivate genes when efficient, it can accelerate gene divergence when moderately efficient, as described for the het-D/E family 65.

The phylogenetic analyses of the multigene families suggest that gene loss may also have occurred during fungal evolution. The putative chitinases related to the killer toxin of K. lactis provide a clear example of this situation. N. crassa and M. grisea have two paralogues, whereas P. anserina has eight. The phylogenetic tree including the ten paralogues present in A. nidulans (Figure 6c) suggests that these proteins can be grouped into two families. Surprisingly, the P. anserina proteins cluster in one subfamily, whereas the M. grisea proteins cluster in the other, indicating differential gene losses. In P. anserina, even if Pa_4_5560 and Pa_5_1570 seem to have duplicated recently, this is not as clear for the other members since they are not very similar. They may result from ancient gene duplications or from recent duplications followed by rapid evolution, possibly thanks to RIP. Evolution of this family seems thus to proceed by a complex set of gain and loss at various times. The same holds true for polyketide synthase (PKS) genes. Seven PKSs were reported for N. crassa 16, while M. grisea has 23 11, and we identified 20 PKS genes for P. anserina. A comparison of all these PKSs (data not shown) indicates a complex evolution process in which N. crassa has probably lost most of its PKSs and the two other fungi present several duplications yielding very different copies. Again, this does not permit us to establish whether the duplications are ancient or recent but followed by intense divergence. See also below for additional examples of losses and amplifications of genes involved in carbon source degradation.

Such gene losses may be frequent events in filamentous ascomycete. As seen in Figure 2a, P. anserina, M. grisea and A. nidulans share 1,624 genes that seem to be lacking in N. crassa (among these, 449 are present in the three fungi, 630 in both P. anserina and M. grisea, and 545 in both P. anserina and A. nidulans), even though M. grisea and A. nidulans are more distantly related to P. anserina than is N. crassa (Figure 2b). Although some genes may have evolved beyond recognition specifically in N. crassa, the most parsimonious explanation is that P. anserina has retained many genes that N. crassa has lost. Similarly, N. crassa, M. grisea and A. nidulans share 1,050 genes that are absent in P. anserina. Therefore, we tentatively suggest that genomes from sordariomycetes may be shaped by more gene loss and gene duplications than anticipated by the presence of RIP. Similar rates of gene loss in filamentous ascomycetes have recently been demonstrated 66.

In nature, P. anserina lives exclusively on dung of herbivores. In this biotope, a precise succession of fungi fructifies 67. An explanation put forward to account for this succession is nutritional. The first fungi to appear feed preferably on simple sugars, which are easy to use, followed by species able to digest more complex polymers that are not easily degraded. Indeed, the mucormycotina zygomycetes, which are usually the first ones to fructify on dung, prefer glucose and other simple sugars as carbon sources. They are followed by ascomycetes that use more complex carbohydrates such as (hemi)cellulose but rarely degrade lignin. The succession ends with basidiomycetes, some of which can degrade lignin to reach the cellulose fiber not available to other fungi 686970.

Usually, P. anserina fructifies in the late stage of dung decomposition 67. This late appearance of the P. anserina fruiting body is hard to correlate with slow growth of the mycelium and delay in fructification since in laboratory conditions ascospore germination occurs overnight and fruit body formation takes less than a week. However, P. anserina harbors unexpected enzymatic equipment, suggesting that it may be capable of at least partly degrading lignin, which concurs with the nutritional hypothesis (Table 5). It includes a large array of glucose/methanol/choline oxidoreductases 71, many of which are predicted to be secreted, two cellobiose dehydrogenases, a pyranose oxidase, a galactose oxidase, a copper radical oxidase, a quinone reductase, several laccases and one putative Lip/Mn/Versatile peroxidase. Enzymes homologous to these CDSs are known to produce or use reactive oxygen species during lignin degradation 686970. This ascomycete fungus may thus be able to access carbon sources normally available mainly to basidiomycetes. Interestingly, P. anserina is closely related to xylariales, a group of ascomycete fungi that seems to contain true white rot fungi capable of degrading lignin 72; also, P. anserina has the most complete enzymatic toolkit involved in lignin degradation when compared to the three other ascomycetes included in Table 5. The comparison with N. crassa is particularly striking. This is in line with the fact that N. crassa in its less competitive biotope may have access to more easily digestible carbon sources.

As mentioned above, P. anserina is considered a late growing ascomycete on herbivorous dung. This suggests that the fungus is likely to target lignocellulose as a carbon source, since most hemicellulose and pectin would probably be consumed by zygomycetes and early ascomycetes. A close examination of the genome sequence of P. anserina for the presence of carbohydrate active functions (Additional data file 6) and a comparison with the genome sequence of other fungi confirmed the adaptation capacity of P. anserina to growth on lignocellulose. The total number of putative glycoside hydrolases (GHs), glycoside transferases, polysaccharide lyases (PLs) and carbohydrate esterases (CEs) are similar to those of other ascomycetes, such as A. niger 20 and M. grisea 73, but P. anserina has the highest number of carbohydrate-binding modules (CBMs) of all the fungal genomes sequenced to date. Despite possessing similar numbers of putative enzymes, the distribution of the possible enzyme functions related to plant cell wall degradation (Table 6) is significantly different in P. anserina from that of other fungi. P. anserina has the largest fungal set of candidate enzymes for cellulose degradation described to date. This is particularly remarkable in GH family 61 (GH61) with 33 members, two-fold higher than the phytopathogen ascomycete M. grisea and the white rot basidiomycete P. chrysosporium. Similar patterns are visible for other cellulose-degrading families (for example, GH6, GH7, GH45) and in the high number of CBM1 (possibly cellulose-binding) modules found, which are only equivalent to the sets of P. chrysosporium and M. grisea.

Strikingly, P. anserina also has an increased potential for xylan degradation, with abundant enzyme sets in families GH10 and GH11, together with a relative abundance of exo-acting enzymes in families GH3 and GH43. Interestingly, no α-fucosidases of families GH29 and GH95 are found, suggesting a depletion of xyloglucan prior to growth of P. anserina. During the stage at which P. anserina grows in dung, significant amounts of cellulose, but also xylan, are still available. Xylan can be cross-linked to lignin through ferulic acid 74 or 4-O-methyl-glucuronic acid 75. In light of the potential of P. anserina for lignin degradation, it is conceivable that this fungus particularly consumes lignin-linked xylan that could not be degraded by 'earlier' growing organisms that lack a lignin-degradation system. The relatively high number of putative CE1 acetyl xylan and feruloyl esterases found in P. anserina by comparison with other fungi correlates with this hypothesis.

In contrast to the increased potential for cellulose and xylan degradation, a significantly weak potential for pectin degradation was observed for P. anserina. No members of GH28 (containing pectin hydrolases) were detected in the genome and only a single α-rhamnosidase (GH78). In comparison, A. niger contains 21 GH28 members and 8 GH78 members 20. The number of putative pectin lyases is also much smaller than that observed for A. niger. The auxiliary activities of GH88 and GH105, likely to act on pectin lyase degradation products, are equally absent from P. anserina while present in all pectin-degrading organisms (Table 6). The absence of the potential to degrade sucrose and inulin is concluded from the lack of enzymes in the GH32 family. This also correlates with the low capacity of P. anserina to grow on rapidly degradable carbohydrates that are most likely depleted by 'earlier' organisms. Furthermore, the large number of GH18 and CBM18 modules, 20 and 30 respectively, could indicate that P. anserina has the ability to degrade exogenous chitin and possibly to depend on available fungal cell material (derived from the set of fungi that grow earlier on dung of herbivores and that P. anserina may kill by hyphal interference 76).

To evaluate whether the enzymatic potential reflects the ability of P. anserina to degrade plant polymeric substrates, growth was monitored on minimal medium plates containing lignin, cellulose, beech wood xylan, apple pectin, inulin and 25 mM sucrose, D-glucose, D-fructose or D-xylose (Figure 7). P. anserina did grow on lignin, indicating that it is able to degrade lignin. However, it is suspected that in nature lignin degradation, an energy consuming process, may not be to obtain a carbon source, but mainly to gain access to the (hemi-)cellulose. Growth on cellulose, xylan and D-xylose was significantly faster than on pectin, which agrees with the enzymatic potential based on the genome sequence as described above. No growth was observed on inulin or sucrose, while efficient growth was observed on D-fructose and D-glucose. This is in agreement with the absence of genes required to degrade sucrose and inulin from the genome of P. anserina. Overall, these data suggest that P. anserina has all the enzymatic complement necessary to efficiently scavenge the carbohydrates it encounters in its natural biotope. Selection has in fact evolved its genome to deal efficiently with these carbon sources, first by duplicating genes involved in cellulose degradation, as shown by the high number of GH61 CDSs, and second by deleting genes required to use carbon sources not commonly encountered (for example, pectin, inulin, and sucrose). This demonstrates the high environmental pressure on evolution as well as the high level of specialization that occurs in the fungal kingdom.

The sequenced strain is the S mat+ homokaryotic strain 79. Culture conditions for this organism were described 61, and currently used methods and culture media can be accessed at the Podospora anserina Genome Project web site 80.

Nuclear genomic DNA was extracted and separated from mitochondrial DNA as described 81. Residual mitochondrial DNA present in the preparation was sufficient to allow sequencing of the full mitochondrial DNA circular chromosome. Construction of plasmid DNA libraries was made at Genoscope. The construction of the bacterial artificial chromosome (BAC) library is described in 24.

Two strategies were used to construct the cDNA libraries. First, a mycelium library was constructed in the yeast expression vector pFL61 82. Total RNA was extracted from the s wild-type strain (mat-) and polyA⁺ RNA was purified twice on oligo (dT)-cellulose columns (mRNA purification kit, Amersham Pharmacia Biotech, GE Healthcare Bio-Sciences AB, Uppsala, Sweden). Anchored dT25 primers were used to obtain double-stranded DNA (cDNA kit, Amersham Pharmacia Biotech, GE Healthcare Bio-Sciences AB, Uppsala, Sweden). Three cDNA libraries, corresponding to three ranges of molecular weight cDNA (0.2-1 kb, 1-2.5 kb, > 2.5 kb) were cloned using BstX1 adaptators in the pFL61 vector between the 5' (promoter) and 3' (terminator) sequences of the S. cerevisiae pgk1 gene as described previously 82.

Second, total RNA obtained under various physiological conditions (Table 4) was extracted as described 83, using the 'RNeasy Maxi Kit' (Qiagen, Germantown, MD, USA). PolyA⁺ mRNAs were extracted with the 'Oligotex mRNA Maxi Kit' (Qiagen), reverse transcribed and cloned with the 'cloneMiner cDNA library construction Kit' into plasmid pDONR222 (Invitrogen, Carlsbad, CA, USA).

The genome of P. anserina was sequenced using a 'whole genome shotgun and assembly' strategy. We generated 510,886 individual sequences from two plasmid libraries of 3.3 and 12 kb insert sizes, and from one BAC library of about 90 kb insert size. This corresponds to genome coverage of 9.7-fold. The reads were automatically assembled using Arachne 84, and the initial assembly was improved by eliminating small redundant scaffolds. Additionally, in cases when the genetic map indicated the proximity of two scaffolds (see below), we joined them if there was some additional read pair information between them that was not used by Arachne. Some inter-contig gaps were also filled by placing a contig between two other contigs when matches and read pair information existed and were coherent. The final automatic assembly consisted of 2,784 contigs of N50 size 43 kb, grouped in 728 scaffolds of N50 size 638 kb, for a total genome size (without gaps) of 35.7 Mb. Manual sequence gap filling and removal of contigs corresponding to rDNA genes permitted the decrease of scaffolds and contig numbers to 1,196 contigs clustered into 78 scaffolds.

To connect the genome sequence with the genetic map 85, two approaches were followed. First, sequenced genes, whose positions on the genetic map were known, were mapped by searching the corresponding sequence in the scaffolds, enabling the attribution of some scaffolds to known chromosomes. Second, potential molecular polymorphic markers (microsatellites, minisatellites and indels) were searched and their polymorphisms were assessed in geographic isolates D, E M, T and U. It rapidly appeared that strain T was the genetically most distant strain from strain S, since about three-quarters of tested markers were actually polymorphic between the two strains. A cross between the T and S strains was set up and 51 homokaryotic progenies from this cross were assayed for 120 polymorphic sites scattered onto the 36 largest scaffolds that represented all the coding parts of the genome (except for one putative CDS). Linkage analysis made it possible to define seven linkage groups that were matched with the chromosomes thanks to the already known genes mapped on the sequence by the first approach. Additional polymorphic markers were then used to confirm local assembly, resulting in the new genome map, which contain 325 markers (Additional data file 1). No discrepancy was observed between the established genetic map, the newly defined linkage groups and the sequence assembly. Presently, all but one CDS-containing scaffold are attributed to a chromosome position, although in a few cases orientation of some scaffolds within the chromosome could not be accurately defined because of their small size. One 33 kb scaffold containing one predicted CDS as well as small scaffolds exclusively made up of repeated sequences are presently not mapped. Collectively, they represent about 1% of the genome.

Chromosome 1: CU633438; CU633901; CU633867; CU633899; CU633445; CU633897. Chromosome 2: CU633446; CU640366; CU633447. Chromosome 3: CU633448; CU633447; CU633453. Chromosome 4: CU633454; CU633455; CU633456; CU633895. Chromosome 5: CU633457; CU633458; CU633459; CU633866;