The completed Assembly 21 contains 15.845 Mb of DNA, organized into the 8 C. albicans chromosomes. The assembly does not include the complete telomeric sequences for every chromosome, and includes only one copy of the normally repeated rDNA on chromosome R. All but two chromosomes end with the subtelomeric repeat CARE-2 (Rel-2) 1920 at each telomere. This repeat was originally shown to be telomere-associated on chromosome 7 8. Macro-restriction maps locate MRSs on all chromosomes but chromosome 3, but since the MRSs are highly repeated and more than 16 kb in size, they are represented but not included in the assembly. On chromosomes 7 and 6, where the MRSs have been sequenced, they are inserted into the sequence. The finished sequence of chromosome 7 has been published elsewhere 21; this assembly includes a slightly revised version. In addition to the missing MRS regions there is one gap, on chromosome 3. Assembly 21 is a haploid assembly; in cases where Assembly 19 detected heterozygosity, allelic contigs 19-1XXXX and 19.2XXXX were assembled. In regions where there were allelic contigs in Assembly 19, only the 19-1XXXX contigs were used to construct Assembly 21, so it provides no information about heterozygosity.

Table 1 shows the sizes of the individual chromosomes in Assembly 21 and compares the size of the sequence with the size of each chromosome as determined by the optical map. The Assembly 21 size in Table 1 does not include the MRS if one or more are present, and for chromosome R, only one copy of the rDNA repeat is included. The chromosomes range in size from 3,190,598 bases for chromosome 1 to 943,480 bases for chromosome 7. For chromosome R, the actual number of rDNA repeat adds 350 kb to one homologue and 800 kb to the other. Where chromosome homologues are of different sizes, as with chromosome R, the optical map software will choose one homologue. The optical map thus gives the size, including the rDNA, of the smaller homologue of chromosome R. The larger homologue is very close to chromosome 1 in size, about 3.1 Mb. The fact that the subtelomeric repeat CARE-2 is missing from one telomere may indicate that the sequence does not extend to the end of the chromosome. On chromosomes 2 and 7 both the TLO gene and the CARE-2 sequences are missing. We were able to map 233,091 out of 250,884 (93%) of the original sequence traces on Assembly 21. The remaining 17,793 sequences probably represent some of these missing sequences as well as allelic variants but it was impossible to map them to the current assembly.

Assembly 21 demonstrates that the sub-telomeric repeats found on chromosome 7 21 are characteristic of all the chromosomes. The common factor is all or part of the repeat CARE-2 19, which includes the long terminal repeat (LTR) kappa (AF041469) and shares some sequence with Rel-2 20. There are several other LTRs at the telomeres, and on chromosome 1R there is an intact transposon pCa1 (AF007776). Although one telomere on each of chromosomes 2 and 7 in Assembly 21 is missing CARE-2 (and the TLO gene (see below)), the fosmid map shows these repeats on every telomere. Sequence near the telomere is hard to clone, and the most likely interpretation of this discrepancy is that the appropriate clones for these regions were underrepresented in the Stanford library. Since the repeats tend to be telomere-proximal to the TLO gene, the fact that TLO is missing on chromosomes 2 and 7 also suggests that some sequence is missing. The detailed organization of the sub-telomeric repeats is complex and differs at each telomere.

Sequences that bind the Cse4p protein (the CENP-A orthologue) from C. albicans have been identified on each chromosome by Sanyal et al. 22, who suggested that they are at least part of the centromere. This interpretation is supported by the observation that the sequence identified on chromosome 7 contributes to its mitotic stability. Table 1 gives the address (distance in bases from the left-hand end of the chromosome) of each centromere and shows that the chromosomes are generally metacentric, with the centromere sequences located near the center of the chromosomes. However, chromosomes 2 and 6 are acrocentric. On chromosome 2 the centromere is about 85% of the way toward the right telomere, and on 6 it is more than 95%. In contrast to Saccharomyces cerevisiae and Candida glabrata, there are no sub-telomeric gene families other then the TLO genes in C. albicans.

The ORF analysis of Assembly 21 is being carried out at the Candida Genome Database (47). The present analysis, based on the previous assembly, 20, differs slightly from that of Assembly 19. The human-curated annotation of Assembly 19 identified 6,354 genes. The Candida Genome Database contains 12,015 genes, including allelic variants. Assembly 20 contains 6,090 genes. Of these, the identity and sequence of 6,065 have not changed significantly from Assembly 19 (>98% sequence identity). Fourteen new ORFs have been added to Assembly 20. Thus, 290 genes from the annotation of Assembly 19 are not in Assembly 20. Of these, 192 have been found to be identical to other Assembly 20 genes and have >90% of their sequence incorporated in these genes, while an additional 19 have >50% of their sequence incorporated into another ORF. The 79 remaining Assembly 19 ORFs have no strong Blast hits and include 55 hypotheticals, 15 gene family members, 7 that were truncated, 1 that was overlapping, and 2 putative sequencing errors. The Candida Genome Database provides an extensive analysis of the changes in ORF classifications between the two assemblies. Some discrepancies between their numbers and ours arise from the fact that they started from a greater number of orf19 genes.

C. albicans was shown to have eight chromosomes by pulse-field electrophoresis. Early studies demonstrated not only that the organism was diploid but that it contained several large blocks of repeated DNA, the MRS, with a complete or partial copy on each of the eight chromosomes 18232425.

Analysis of the emerging sequence demonstrated the existence of a large number of LTRs and other repeated sequences related to transposons 26. Gene families also constitute a significant source of repeated DNA. Some of these repeated sequences, especially the MRSs, are too large to be crossed in a single sequencing run, so that assembly using bioinformatics is blocked. The sequences of the MRSs on chromosomes should be considered unreliable due to the repeated nature of MRSs, which attract traces from many different other MRSs in the genome. The smaller repeated DNA regions are subject to the same potential problem and led to erroneous assembly in some of the Assembly 19 contigs. The fosmid map and the optical map were very helpful in identifying these errors and correcting them.

The putative transcription factor gene CTA2 was identified by Kaiser et al. 27 in a one-hybrid screen in S. cerevisiae. Goodwin and Poulter 26 first noticed that this or a related sequence was often telomere-associated. The CTA2 sequence has homology to members of a gene family found at 14 of the 16 telomeres. Since CTA2 is a gene name, we renamed the family TLO for TeLOmere-associated genes. On this assembly, they are numbered so that the left arm of the chromosome has an odd-numbered TLO gene and the right TLO gene has an even number, ascending from chromosome R through chromosome 7. Thus, chromosome R has TLO1 on the left and TLO2 on the right. Although the original member of this gene family was isolated because it has transactivating activity in S. cerevisiae, the function of these genes in C. albicans has not been determined. There are 15 members of this gene family in Assembly 21; 14 are located within 14 kb of an end of a chromosome contig, and all are oriented in a 5'-3' direction toward the centromere and away from the telomere (Figure 1). In addition, all but one (chromosome 1R) have the LTR kappa 5' to the ORF, usually immediately adjacent but sometimes a few kb away. One member of the TLO gene family (ORF19.2661) is found in the interior of chromosome 1, 1.29 Mb from the left end of this 3.19 Mb chromosome. Like the other family members, this ORF points toward the centromere and away from the telomere and has a kappa sequence from the subtelomeric repeat CARE-2 in its 5' region. In this case, the kappa sequence is the only part of the CARE-2 sequence present. Thus, this particular copy resembles all the telomere-associated family members but is located in the middle of chromosome 1, and we have numbered it TLO34, since it is located between TLO3 and TLO4. In order to keep the numbering system consistent, we have saved the names TLO6 and TLO15 for the genes on 2R and 7L that we expect will eventually be identified. In addition, Figure 2a demonstrates that TLO14, on the right arm of chromosome 6, has been confirmed by PCR but this gene has yet to be sequenced.

An alignment of the amino acid sequences encoded by 13 TLO genes is shown in Figure 2c. While we initially noticed that six of them (TLO5, 7, 8, 11, 13, and 16) had a different carboxy-terminal domain, an astute reviewer remarked that adding an intron to the annotations of these genes would allow them to code for a protein with a carboxy-terminal end that is similar to the other TLO genes. We used rtPCR with a universal 5' primer and gene-specific 3' primers to show that both types of transcripts are expressed in Candida cells. Although the amplicons nevertheless contained a mixture of two or more alleles or family members, end sequencing clearly showed that the long form contained the common carboxy-terminal sequences while the short form encoded the unique sequences first noticed in TLO5. Although the annotation assumes that all six members of this TLO family sub-group yield transcripts in both spliced and unspliced forms, confirmation would require screening of an expressed sequence tag library. Finally, TLO4 (orf19.7276.1) lies almost at the edge of a sequence contig and is obviously missing an upstream exon while the additional amino-terminal domain encoded by TLO34 may or may not be part of the actual transcript.

As noted by Braun and coworkers 28, C. albicans has a number of gene families. These range in size from 2 members to as many as 26 (the ABC transporter superfamily) 29. Several are clustered on one or two chromosomes. For example, chromosome 6 contains members of six families, with five represented by more than one member, and one, ALS, with five members. Most of the families with more than two members have representatives on more than one chromosome. In addition, some two-member families are on different chromosomes. Examination of the genes surrounding family members on different chromosomes for the most part gives no hint as to the mechanism by which homologous sequences arrived at different locations.

The arrangement of these families on the chromosomes is not random. Figure 3a shows the ORFs from Assembly 21 aligned along the eight chromosomes. ORFs with 80% or more similarity are connected by lines. There are some areas that appear to be tandem duplications. An example from chromosome 3 is shown in Figure 3b. The sequence containing the gene DTD2 and those for two hypothetical proteins seem to have undergone a duplication accompanied by an inversion. A similar event seems to have occurred 40 kb away involving the genes DAL5, ECM18, and FUR4. There are no sequences such as LTRs near these duplications. Although the most plausible mechanism for the generation of gene families whose members are close together on one chromosome is tandem duplication, in many cases of gene families that contain tandemly repeated genes, the most closely related members are dispersed. Table 2 shows this for the SAP (

S

A

P

LIP

The first step in Assembly 21 was to assign contigs from Assembly 19 to the appropriate chromosomes. One hundred contigs were anchored by probes on the physical map. To assign the rest, chromosomes were separated by pulse-field gel electrophoresis. Chromosomal bands were eluted from the gel, labeled with either Cy3 of Cy5 dyes, and hybridized to a microarray 10 based on ORFs identified from Assembly 4. As a control, total genomic DNA was also labeled with the reciprocal dye and co-incubated along with the partially purified chromosomes. At least two individual hybridizations (including dye swap controls) were conducted for each of the eight chromosomes. Results and details from these experiments can be seen on our web page 39. Fluorescence ratios were interpreted visually and it was very easy for us to assign chromosomal localization for 183 Assembly 19 contigs representing 98.4% of DNA sequences and 97.3% of known genes. This analysis also identified nine misassembled contigs by the fact that genes assigned to them hybridized to different chromosomes. The emerging sequence from the closely related species C. dubliniensis was then aligned, using megablast, to the Assembly 19 contigs. Phrap was used to connect pairs of contigs that were assigned to the same chromosome and were found to be adjacent based on the C. dubliniensis overlap 48. After the release of the C. albicans WO-1 traces, Phrap and these traces were also used to locate misassemblies and to correct them.

The emerging alignments were compared with the physical map and areas of disparity mapped with more probes. Attempts to assemble the contigs into whole chromosomes with Phrap failed because of the large number of repeated elements in the C. albicans genome. A custom-made stitcher script (which utilized a 100 nucleotide pure text string starting from the previous alignment) was used to assemble the contig-alignments into chromosomes. The resulting assembly was checked for the presence of all the ORFs from the community annotation 28, and very few ORFs were missing. The penultimate assembly was compared with the optical genome map and the physical map and disparities resolved.

One major discrepancy was the orientation of the sequence between the MRS regions on chromosome 4. The optical map and the physical map showed it in one conformation, while the assembly showed it in the other. The final orientation, consistent with the physical and optical maps, was confirmed by PCR of the border fragments of one MRS and by restriction digestion of genomic DNA and hybridization of a Southern blot with probes from the regions that flank the two MRSs on both sides.

The remaining gaps were filled by PCR followed by sequencing of the products. One gap proved to be the result of a misassembly in a contig. The insertion of a partially overlapping contig closed it. Two gaps were filled by using the draft sequence of Broad Institute's WO-1 assembly 40. One gap was filled by a sequence from Selmecki et al. 5. The result was called Assembly 20 and it was composed of eight chromosomes with one unresolved (not connected) contig pair. Two gaps were introduced due to the flip of the area between two MRSs on chromosome 4. Two gaps are located around the rDNA region. In addition, most of the MRS regions have not been resolved and can be considered gaps, with the exception of the two MRSs in chromosome 7 and the one MRS in chromosome 6, which were cloned and sequenced at Chiba University.

Following the completion of Assembly 20, we discovered that sequence traces that were used for the assembly and that we had thought were from strain SC5314 were in fact coming from the WO-1 sequencing project. Additional experiments and analysis were thus required to produce Assembly 21 in which the C. albicans WO-1 sequences from the contig junctions are replaced with sequences from SC5314. The original traces from SC5314 were obtained from the Stanford Genome Technology Centre and aligned over the junction areas of Assembly 20 using Sequencher 4.7 41. Data that were contaminated with WO-1 sequences were tagged as 'Reference sequences', which means that they were not used to construct the consensus sequence for the new alignments. Any missing bases not covered by SC5314 traces were filled with Ns. The resulting alignments can be viewed on our web page 39. All but 85 junctions could be confidently filled with the SC5314 traces. PCRs were then performed on gaps, overlaps and rearranged regions that had low SC5314 trace coverage. As shown in Additional data file 1, a first round of 85 PCRs was successful in producing 80 amplicons of the expected size and efforts will continue to produce the missing SC5314 sequence data over the next few months. Additional data file 2 includes the sequences of the primers used for these PCR reactions.

A fosmid library was constructed from strain 1161 42 by M Strathman. Sau3A-digested genomic DNA was inserted into the fosmid vector pFOS1 43. The library consists of 3,840 clones with an average insert size of 40 kb (10× genome coverage). For probing, the library was arrayed in 10 384-well plates. Two plates each were printed on a 12 × 16 cm nylon membrane (Hybond-N+, Amersham Pharmacia Biotech Inc Piscataway, NJ, USA), giving a complete library set on five membranes. For printing, freshly thawed fosmid clones were transferred, using a 384-point replicator, to the membrane overlaid on Luria agar containing 20 μg chloramphenicol. The plates were grown overnight at 37°C and the membranes were processed according to the manufacturer's instructions for colony lifts.

Probes were variously clones of genomic DNA, T- and S-end probes from fosmids 44, and PCR products generated from SC5314 DNA using primers based on the public genomic sequences. Probes were randomly labeled with ³²P. The mapping was carried out as described in Chibana et al. 8. Briefly, membranes containing the library were hybridized with probes and fosmids hybridizing with the same probe were considered to overlap. Probes were also hybridized to Southern blots of pulse-field separations of chromosomes and genomic SfiI digests. The overlapping fosmids were arranged in a linear fashion with chromosome markers like the MRS used to orient the array. The complete set of overlapping fosmids was trimmed to make a minimum tiling set. The physical map is available 45.

The optical map was prepared by OpGen (Madison, WI, USA) using a proprietary technique. The technique involves preparation of large (>500 kb) molecules of genomic DNA, and spreading them on a microfluidic device that causes them to elongate and adhere to the substratum. They are then treated with a restriction enzyme (in this case, XhoI). Since the molecules are under slight tension, restriction sites appear as breaks in the molecules. The molecules are then stained with a fluorescent dye, scanned, and the positions of the restriction sites determined, using the amount of fluorescence to calculate the mass of the fragments. The results are averaged over 500 molecules, allowing the preparation of a macro-restriction map. This can be compared with the restriction map derived from the assembly. A more complete description of the technique and more information can be found at 46. Since the optical map software searches the image database for matches and then constructs the map based on the consensus pattern, the optical map of a chromosome describes only one of the two homologues, and no data on heterozygosity are included.

Although in most cases the three methods were in agreement on the sequence assembly, in cases where disagreements could not be resolved, concurrence of two of the three approaches was deemed sufficient.

The complete Assembly 21 has been submitted to the Candida Genome Database 47, which will be in charge of its maintenance and curation.