The chromosome number of N. furzeri is 2n = 38 and includes four pairs of metacentric, three pairs of acrocentric and twelve pairs of subtelo-/submetacentric chromosomes (Figure 1a). Based on morphology, there do not seem to be clearly differentiated sex chromosomes. Specific heterochromatin staining indicates that the cytological organization of the N. furzeri genome is highly structured as is evident from the presence of large blocks of C-banding positive heterochromatin in the centromeric region of most chromosomes (Figure 1b). Accumulation of heterochromatin in other chromosomal sites cannot be detected. To evaluate the composition of heterochromatin, we performed base specific fluorochrome staining on metaphase chromosomes. Staining with the A+T specific dye DAPI resulted in poor fluorescence of centromeric regions (Figures 1a and 2a). Conversely, staining with mithramycin, which shows high affinity for G+C-rich DNA, generated bright fluorescence in most centromeric regions that are DAPI dull (Figure 1c). This indicates that constitutive heterochromatin in N. furzeri is G+C-rich. The analysis of closely related African Nothobranchii, including the sympatric species N. orthonotus as well as two allopatric species from Tanzania, N. hengstleri and N. eggersi, does not indicate a comparably structured genome organization (Figure 2b–d). Thus, at the cytological level, a compartmentalization of the N. furzeri genome is apparent which is likely caused by substantial differences in DNA composition. This seems unusual since fish genomes are generally characterized by a very limited compositional heterogeneity 2526.

We sequenced 5.4 Mb of the short-lived strain N. furzeri GRZ, the long-lived, recently wild-derived strain N. furzeri MZM-0403, and the long-lived closely related species N. kunthae using a whole genome shotgun approach and Sanger sequencing (Table 1). To assess the genome size of N. furzeri based on the sequences, we assumed that the number of its protein coding genes does not differ significantly from that of other fish species, as was previously suggested for vertebrates 27. We identified sequences containing protein coding information in the 5.4 Mb of both N. furzeri strains by BLASTX searches in Swiss-Prot/TrEMBL, and did the same for medaka, stickleback, tetraodon and zebrafish, for which we extracted adequate genomic samples from public databases. Based on the reported genome sizes of the latter four fish species, we then deduced the genome size of N. furzeri. In detail, 444 (8%) of the 5,540 N. furzeri GRZ sequences and 443 (8%) of the 5,686 MZM-0403 sequences show significant similarity (p < 10^-10) to protein coding genes. Respective sequences comprise 31%, 20%, 13% and 11% of the tetraodon, stickleback, medaka and zebrafish genomic samples, respectively, which corresponds to a genome size of 1.59-1.92 Gb for N. furzeri (Additional data files 1 and 2). For experimental confirmation, we performed flow cytometry measurements using DAPI and propidium iodide (PI). DAPI, which preferentially stains A+T rich DNA segments, yields a DNA content of 2.33 pg/diploid cell while the non-base-specific PI staining results in 3.11 pg/diploid cell (Additional data file 3). In light of the large blocks of G+C-rich heterochromatin observed in our cytogenetic studies, the value obtained with PI is likely the correct one since it is based on a dye that does not depend on DNA composition, whereas DAPI staining most probably results in an underestimation of DNA content 28. The value obtained with PI corresponds well with our sequence based estimate; that is, it is equal to a genome size of approximately 1.5 Gb. The 5.4 Mb thus represents roughly 0.3-0.5% of the N. furzeri genome.

Based on these data, the N. furzeri genome is likely at least half the size of the human genome, bigger than the four other fish genomes, and has less chromosomes. At 1.4 Gb (25 chromosomes) 29, the zebrafish genome is slightly smaller, while medaka (1 Gb, 24 chromosomes 20), stickleback (0.7 Gb, 21 chromosomes 30) and tetraodon (0.4 Gb, 21 chromosomes 31) genomes are considerably more compact.

The G+C content of the 5.4 Mb sample of N. furzeri GRZ is 44.9%. Interestingly, approximately 10% of the sequences have a G+C content of 60% or higher; in Figure 3a this is indicated by a second peak at approximately 62% in a plot of number of sequences against G+C content. The same unusual G+C content distribution is seen in the recently wild-derived N. furzeri strain MZM-0403 (Additional data file 4). To test whether this represents an artifact introduced by preferential propagation of G+C-rich sequences in Escherichia coli 32 during library preparation, we performed whole genome shotgun sequencing using Roche/454 Life Sciences technology, which does not involve cloning and amplification in bacterial systems 33, for N. furzeri GRZ. In this second GRZ genomic sequence sample, the G+C content is 44.3% (111.6 Mb; Table 1). Similar to the Sanger sequences, approximately 10% of the 454 sequences show a G+C content of at least 60%, which indicates that there is not a strong cloning bias in the genomic libraries we initially sequenced. Taken together, our sequence data suggest that a distinct G+C-rich fraction is present in the N. furzeri genome. A similar fraction does not seem to exist in medaka, stickleback, tetraodon and zebrafish (Figure 3b) and is absent as well in the closely related species N. kunthae (Additional data file 4).

To assess the extent to which our sample-based estimation reflects the G+C content of the N. furzeri genome, we analyzed entire genomes as well as adequate genomic samples of medaka, stickleback, tetraodon and zebrafish. We found that the G+C content estimates of genomes and genomic samples are essentially the same (Table 2). Our calculations are in agreement with previously published data; we estimated a G+C content of 46.4% for tetraodon and reported values are 45.5% 31 and 46.4% 20; similarly, 40.3% was reported for medaka 20 and we found it to be 40.5%. Thus, the G+C content of N. furzeri is likely 44-45%, which is similar to stickleback (44.6%), slightly lower than in tetraodon (46.4%) and considerably higher than in medaka (40.5%) and zebrafish (36.6%). Based on genomic sequences, an inverse correlation of genome size and G+C content was recently found for the latter four fish species 34, which confirmed previous experimental results showing that small genomes are generally associated with high G+C content and vice versa 25. N. furzeri would seem an exception as its G+C content is nearly as high as that of tetraodon, which has a four times smaller genome. The sequence fraction in the N. furzeri genome with the high G+C content (≥60%) increases the global G+C content rather slightly - the value is 43.1% if these sequences are excluded - and it seems that these sequences occupy defined chromosomal regions (see below).

While to our knowledge there is no evidence for a direct influence of nuclear genome composition on the life span of a vertebrate species, some reports correlate life span with the composition of proteins encoded by the mitochondrial genome. For example, Rottenberg 35 found that rates of amino acid substitution per site in mitochondrial DNA of mammals are positively correlated with longevity of a genus and suggested that the evolution of longevity drove the accelerated evolution of peptides encoded by mitochondrial DNA. Moosmann and Behl 36 showed that the frequency with which cysteine is encoded by mitochondrial DNA is a specific indicator for longevity; that is, longevity is associated with a depletion of mitochondrial cysteine in aerobic species. It will be very interesting to analyze the cysteine content in mitochondrially encoded proteins in comparison to nuclear proteins in N. furzeri strains/Nothobranchius species with different life spans.

Repetitive elements can be grouped into the main classes 'tandem repeats', 'transposon-derived interspersed repeats', 'processed pseudogenes' and 'segmental duplications'. Because our genomic samples comprise rather small, fragmented fractions of the N. furzeri genome, it is impossible to identify pseudogenes and large complex repeat structures. Also, the short average sequence length in the 111.6 Mb generated by the Roche/454 technology (100 nucleotides; Table 1) practically rules out a meaningful repeat analysis. We therefore concentrated on the mere identification of tandem repeats and transposon-derived interspersed repeats in the 5.4 Mb of both N. furzeri strains and N. kunthae generated by Sanger sequencing and, for comparison, analyzed our samples of medaka, stickleback, tetraodon and zebrafish genomes.

We considered as tandem repeats microsatellites, minisatellites and satellites composed of 1-5 nucleotides, 6-99 nucleotides, and over 100 nucleotides per repeat unit, respectively. About 1% of the N. furzeri DNA is composed of microsatellites, which is comparable with tetraodon (1.1%) and stickleback (0.8%), about half as much as in zebrafish (2%) and five times more than in medaka (0.2%) (Table 2). Roest Crollius et al. 31 reported that microsatellites comprise 3.21% of the tetraodon genome. The higher number is most probably due to the different algorithms and motif sizes applied; that is, Roest Crollius et al. used a Smith and Waterman algorithm-based approach previously applied to Takifugu rubripes 37 and a motif size of 1-6 nucleotides, while we used the program Sputnik 38, a specific tool for the detection of microsatellites, and a motif size of 1-5 nucleotides.

In N. furzeri, dinucleotide repeats are the most common type of repeat (37%; Additional data file 5) and cumulatively occupy the third largest amount of sequence compared to the other tandem repeats (Table 3). The repeat motif AC is the most frequent (26%), which is slightly less than in tetraodon and stickleback (30% each), and considerably more than in medaka (10%) and zebrafish (15%; Additional data file 5).

Minisatellites are far more abundant in N. furzeri than in the other four fish species. In particular, a 77-nucleotide minisatellite is most frequent. It comprises approximately 10% of the 5.4 Mb (Table 3) and its consensus sequence has a G+C content (63.6%) well above the genome average. Sequence conservation is high, for example, in an alignment of 189 repeat monomers; 65 of 77 positions (84%) are identical in at least 90% of monomers (Figure 4). Using in situ hybridization we found that this minisatellite localizes to centromeric regions of many chromosomes (Figure 5a). Also, the 77-nucleotide minisatellite is N. furzeri specific as we did not detect this or similar tandem repeats in available genomic sequences of other fishes and vertebrates. We identified several other abundant and G+C-rich minisatellites (Table 3), which also localize to centromeric regions. For example, a 49-nucleotide minisatellite is also found in centromeric regions of many chromosomes, whereas a 24-nucleotide minisatellite specifically localizes to centromeres of only two chromosomes (Figure 5b, c). Based on these analyses, we conclude that the large blocks of heterochromatin in centromeric regions, which are visualized at the cytological level by G+C-specific staining methods (Figure 1c), are mainly composed of N. furzeri-specific and abundant G+C-rich tandem repeats. We plan to isolate additional minisatellites to be used in a more elaborate, multi-color fluorescence in situ hybridization (FISH) study to assess chromosome specificity and derive chromosome-specific probes.

Hundreds of repetitions of the minisatellite motif 5'-TTAGGG-3' are found at telomeric ends of vertebrate chromosomes 39 and with associated proteins keep telomeres in homeostasis 40. The telomeric repeat is not contained in the 5.4 Mb of N. furzeri. Most likely this is due to the limited sample size and so we searched for the repeat motif in the 111.6 Mb sample. We found 50 sequences, 14 of which are entirely composed of [TTAGGG]_n. Correspondingly, in FISH experiments using the 5'-TTAGGG-3' motif as probe, specifically the terminal ends of chromosomes were labeled (I Nanda, unpublished results). Thus, characteristic vertebrate telomeric structures are present in N. furzeri and will be described elsewhere 15.

In addition to the G+C-rich minisatellites reported above, there is a prominent satellite repeat in N. furzeri. It comprises approximately 6% of our genomic sequence samples and has a monomer length of 348 nucleotides (Table 3). The G+C content of its consensus sequence (41.1%) is below the genome average. This satellite accounts for a second anomaly in the G+C distribution of N. furzeri; it causes an excess of sequences at approximately 41% G+C (Figure 3a). We found that this repeat also localizes to centromeric regions of many chromosomes (Figure 5d). It is currently not clear if higher order repeat structures are formed by the two types of tandem repeats and if these are specific to certain chromosomes, or if and how this impacts on the structural organization of the N. furzeri genome 4142. A direct influence on N. furzeri life span seems unlikely, since the tandem repeat content of the GRZ and MZM-0403 strains is identical while their life spans differ by a factor of two. It is tempting, however, to speculate that these tandem repeats play a role in cell division, and that both the exceptional size and composition of N. furzeri centromeres might be functionally linked to the extremely fast growth of young fish. In mammals, repetitive DNA becomes demethylated over age, which might affect chromosome structure in constitutive heterochromatin composed of such repeats. Since DNA methylation is also observed in fishes 43444546, it will be interesting to analyze whether methylation occurs in the G+C-rich tandem repeats of N. furzeri and whether methylation changes with age, or differs between strains exhibiting different life spans.

In summary, tandem repeats, comprising a total of 20.6% of the genome, are exceptionally abundant in N. furzeri. They are 4-12 times more frequent than in zebrafish (5.0%), tetraodon (3.6%), stickleback (2.1%) and medaka (1.7%; Table 2). In tetraodon, two major satellites are reported 31. One is a subtelocentric, highly variable ten-nucleotide minisatellite, which is also the most prominent tandem repeat in this fish, while the other is a centromeric, somewhat variable 118-nucleotide satellite also found in Takifugu rubripes.

Interspersed transposon-derived repeats are mobile genetic elements and common to many eukaryotic genomes. Using RepeatMasker and Repbase Update, the reference library of vertebrate repeats 47, we found that 7-9% of the N. furzeri 5.4 Mb samples is composed of known interspersed repeats (Table 2). Of these, approximately 6.3% are retrotransposons and approximately 1.5% DNA transposons. To identify novel transposon-derived repeats, we used the ab initio repeat identification program RepeatScout 48, which was recently found to be suitable for analyzing short sequences, although originally developed to analyze longer sequences 49. Accordingly, 16-18% of our N. furzeri genomic samples are composed of novel interspersed repeats. Unfortunately, a detailed classification of the novel repeats is currently not feasible, again due to the limited/fragmented sample size. However, we assume that the fraction of these repeats identified in the 5.4 Mb closely resembles the overall fraction present in the N. furzeri genome, because our estimate for the medaka genomic sample corresponds well with the estimate given for the medaka draft genome 20. In total, interspersed repeats comprise 24.7% of the N. furzeri genome, which isconsiderably more than in tetraodon (3.4%), stickleback (4.5%), and medaka (13.6%), and less than in zebrafish (35.4%). The overall repeat content in N. furzeri thus amounts to approximately 45%, which is the highest among the analyzed fish species. We note that this value might still be an underestimate, as we can not exclude that the interspersed repeat content might be found to be higher upon analysis of larger genomic samples since it is conceivable that we have missed 'rare' repeats in our samples.

We attempted to explore the protein coding fraction contained in the 5.4 Mb of N. furzeri GRZ genomic sequence with respect to conservation in medaka, stickleback, tetraodon, zebrafish and human. As indicated above, we found that 444 of the N. furzeri GRZ sequences (473 kb of the 5.4 Mb) bear fragments of protein coding genes (Additional data file 2). Roughly one-third (152 kb) of the sequences are actually coding, and most of these (399, 94.3%) contain one to three exons. The G+C content of the 444 sequences (44.1%) fits the genome average and is considerably higher (50.3%) in the coding portions. Similar observations were made in tetraodon and medaka 1920.

Of the 444 sequences, 310 (70%) show best matches to fishes, while 32 (7%) and 23 (5%) match best to human and mouse, respectively (Additional data file 2). Furthermore, 410 (92%) of the 444 N. furzeri gene fragments have homologs in at least one of the other four fish species with amino acid identities of 22-100%. For a subset of 180 (40%) N. furzeri gene fragments we identified homologs in all four fish species. In those, amino acid conservation is highest in medaka (77.4%) and stickleback (77.1%) followed by tetraodon (75.2%) and zebrafish (70.1%). Lastly, of the 34 (8%) N. furzeri gene fragments without a counterpart in currently available sequences of the four fish species, four match best to other fish species (58-100% amino acid identity), 13 to mammalian/vertebrate proteins (26-67% amino acid identity), and 17 to plant, fungal, invertebrate and other proteins (27-72% amino acid identity) (Additional data file 6).

In a complementary approach, we cloned complete coding sequences of five N. furzeri genes, Cdkn2b, Cdkn2d, Msra, Sirt1 and Tp53. Their gene products are involved in different aspects of aging and functionally conserved across taxa 15051. Sirtuin1 is a NAD(+)-dependent histone deacetylase considered to mediate the life span extending effects of calorie restriction 52, and its aging regulating function is conserved from yeast to mammals. Msra repairs and protects proteins from oxidation 53, and mice with mutations in it show a reduced life span 54. Cdkn2b, Cdkn2d and Tp53 control cell cycle progression and mediate senescence 555657. A comparison of respective N. furzeri and human proteins should give an indication as to the potential use of this fish as a model for human biology. The five N. furzeri proteins are 40-65% conserved in humans (Table 4), which is in the range reported for orthologous proteins of teleost fishes and humans (60-70%) 19. For comparison with zebrafish, which is perhaps the best established fish model for human biomedical phenotypes 58, we deduced zebrafish Cdkn2b, Cdkn2d, Msra, Sirt1 and Tp53 genes based on database searches and alignments of expressed sequence tags to genomic DNA (Additional data file 7). Conservation of inferred zebrafish proteins to human orthologs is in the same range (41-65%) as for N. furzeri and humans (Cdkn2b, 41%; Msra, 65.3%; Sirt1, 48.9%; Tp53, 48%). Conservation of the five N. furzeri aging relevant genes with respect to the other four fish species is highest in medaka, followed by stickleback, tetraodon and zebrafish (Table 4), the same as for the 180 random gene fragments described above.

Finally, we attempted to relate the 444 gene fragments and 5 aging related genes with respect to conservation in humans. We analyzed exons encompassing at least 33 amino acids, which number 366 in the former data set and 21 in the latter. In these exons, conservation with human counterparts is considerably higher for random gene fragments than for aging related genes: 67% versus 62%. Currently, we can only speculate about the reasons for this difference. One possibility might be that it reflects a bias introduced by the sampling procedure of the 444 gene fragments. On the other hand, the difference might be related to the huge difference in life span between N. furzeri and human.

In an initial attempt at phylogenetic reconstruction, we used the N. furzeri 28S rRNA gene (Additional data file 8) in comparison with 15 fish and two tetrapod sequences. However, standard parsimony analysis resulted in tree topologies much different from widely accepted taxonomic groupings (data not shown), indicating that the phylogenetic information contained in the 28S rRNA locus does not resolve the evolutionary history of fishes and tetrapods. We thus attempted to infer the phylogenetic positioning of N. furzeri with respect to the four fish model species and humans based on the 444 partial protein coding sequences and aging related genes identified in this work. A phylogenetic tree obtained by the maximum parsimony approach (Figure 6) is fully consistent with the taxonomic classification of the analyzed fish species. Accordingly, Nothobranchius is most closely related to medaka. Zebrafish and humans are both distantly related to the euteleost fish division consisting of the Atherinomorpha Nothobranchius and medaka on the one hand, and the Percomorpha stickleback and tetraodon on the other. The same tree topology was obtained using the neighbor-joining method, giving slightly lower bootstrap confidence (data not shown). We conclude that the coding fragments are sufficient to create a stable phylogeny, as previously suggested 59. Our data are in agreement with phylogenetic analyses performed for ten fish species, which, based on expressed sequence tag analyses, found both killifish and medaka to be more distant from pufferfish than from each other 59.

The N. furzeri GRZ strain studied in this work most probably originates from a single pair, which stems from wild fish captured in 1968, and was used for breeding more than 30 years ago 11. If we assume a generation time of 6 months - comprising 12-16 weeks following hatching and about the same time for embryonic development - GRZ fish would have been inbred for approximately 80 generations until today. Thus, it was generally assumed that GRZ fish are highly homogeneous 17, although experimental proof was lacking.

To assess the genetic variability of the GRZ strain, we genotyped eight microsatellites (four di-, three tri- and one tetranucleotide repeat, identified in the 5.4 Mb), in 20 GRZ specimens. All microsatellites are monomorphic; we observed only one allele in the 20 specimens, respectively (Additional data file 9). To extend the analysis to gene associated markers, we resequenced parts of 19 protein coding genes encompassing 4 exonic and 15 intronic regions (10.2 kb in total). Again, there was no sequence variation in the 20 GRZ fish. In a third approach, we performed multi-locus fingerprinting in ten GRZ specimens using the 5'-GGAT-3' repeat as a marker 60. As for the other markers, all specimens were homozygous at all 5'-GGAT-3' marker loci. Thus, we provide here the first experimental confirmation that the GRZ strain is highly inbred. In contrast, N. furzeri strain MZM-0403, bred in the sixth captive generation in our facility, exhibits 2-6 alleles in the microsatellites and has single nucleotide variations in 14 of the 19 genes. Also, a laboratory strain of the closely related species N. kunthae is highly heterogeneous in all markers that could be analyzed (Additional data files 9 and 10). While the inbreeding status of the GRZ strain is advantageous for further genetic analyses, it is currently unclear to what extent its short life span results from inbreeding depression 61. A reduction of fitness can be caused by the accumulation of deleterious alleles 62, and reduced genetic diversity often affects birth weight, survival, reproduction and resistance to disease 63. In this respect is noteworthy that the N. furzeri GRZ strain is derived from one 'normal' breeding pair that, 35 years ago, lived "a good 4 and a half months" (16-18 weeks), and that "healthy young fish" were used for spawning 11. In our facility, the 10% survivorship (that is, the maximum life span) of direct descendants of these fish is 16 weeks and the fish can live up to 21 weeks 15, which is well in the range reported for the founders. Furthermore, there are differences in age-related histological damage between GRZ and age matched MZM-0403, which argue for accelerated aging in the GRZ strain rather than inbreeding-related early death. For example, measurement of lipofuscin, which is an autofluorescing pigment that accumulates over age in a number of species 64, shows accelerated aging in brain and liver of the GRZ strain compared to age matched MZM-0403 18. To assess the degree of inbreeding depression on the life span of N. furzeri GRZ, access to new N. furzeri wild isolates from the GRZ game reserve in Zimbabwe or a nearby location is highly desirable. These fish could be cross-bred with the inbred GRZ strain and traits related to fitness, such as growth rate, reproduction, and survival, could be recorded and compared to the parental GRZ strain/wild isolate. On the other hand, it has to be noted that inbred lines are extremely powerful tools for biomedical research; if not subjected to severe inbreeding depression, they are valuable in many respects as they provide a uniform and stable genetic background and respective phenotype(s). Our results prove the feasibility of establishing N. furzeri inbred lines, which in the future may become instrumental in dissecting complex traits like life span.

For genomic sequencing, N. furzeri GRZ obtained in 2002 from Marc Bellemans (Hove, Belgium) were used. Karyotyping and chromosome stainings were done using: two male and female N. furzeri GRZ specimens obtained from Werner Krammer (Deutsche Killifisch Gesellschaft, Ausgburg, Germany); one male N. orthonotus from Nisa (Mozambique) collected by MaS and SS in 2006; one male N. hengstleri collected in 2005 in Mozambique from location MZHL-2005-14; and one male N. eggersi collected in 1997 in Tanzania from location TZ 97/55. Fish were kept in 40 liter tanks at 26°C and room temperature with sponge filters driven by an air supply. Water changes were performed every other day.

To obtain mitotic chromosomes, healthy fish were exposed to a 0.03% colchicine solution for 8-12 h in a well ventilated container. Then fish were sedated on ice, killed by caput dislocation, and gills, spleen and kidney carefully removed. Preparation of cell suspension, hypotonic treatment, fixation of cells, and preparation of slides were as described 65. Conventional chromosome staining, fluorescence staining with DAPI (4',6-diamidino-2-phenylindole) and mithramycin were as described 66. To enhance the mithramycin fluorescence, slides were first pre-incubated with distamycin A. The location of constitutive heterochromatin on chromosomes was visualized as reported 67.

Fin clips were taken from three N. furzeri GRZ females and at least 10,000 cells per sample measured on a Cell Analyzer CAII (Partec, Muenster, Germany). As reference, similar numbers of erythrocytes of a female chicken were used. Preparation of cell suspensions, fixation, DAPI staining and measurements were as described 28. PI staining was performed at a final concentration of 50 μg/ml for 30 minutes at 37°C as described 60. For each dye, three independent measurements were performed.

FISH mapping of repetitive DNA on N. furzeri GRZ chromosomes was performed as described 68. Briefly, a biotin labeled plasmid clone bearing a specific tandem repeat was denatured and hybridized overnight at 37°C to denatured metaphase chromosomes. After post-hybridization washes at low (2× SSC, 50°C) and moderately (1× SSC, 60°C) stringent conditions, hybridization sites were detected with fuorescein isothiocyanate (FITC) conjugated avidin (Vector Laboratories, Burlingame, CA, USA) followed by signal amplification through incubation of slides with biotinylated anti-avidin and FITC conjugated avidin. Slides were examined on a microscope with a CCD camera (Zeiss, Goettingen, Germany); chromosomes showing specific FITC signals were displayed on counter-stained DAPI metaphases using appropriate software (Applied Spectral Imaging, Neckarhausen, Germany).

For genomic library construction, DNA of one male specimen of each of N. furzeri GRZ, N. furzeri MZM-0403, and N. kunthae, was isolated from 30 mg frozen tissue with the Blood and Cell Culture DNA Mini kit (Qiagen, Hilden, Germany). For marker validation, genomic DNA from 20 N. furzeri GRZ, 10 MZM-0403 and 10 N. kunthae specimens was isolated using 30 mg frozen tissue and AquaGenomicSolution (MobiTec, Goettingen, Germany), respectively. For GRZ, 50 ng DNA from ten specimens each were combined in a pool. For cloning of known aging related genes, RNA was isolated from one N. furzeri GRZ male using the RNeasy Mini kit (Qiagen). Primer sequences are given in Additional data file 11.

For shotgun sequencing using Sanger technology, randomly sheared, end-repaired 1-3 kb DNA fragments were ligated into pUC18. Recombinant plasmids were amplified in E. coli, purified, sequenced from both ends (BigDye Terminator v.1.1 Cycle Sequencing Kit, ABI, Weiterstadt, Germany) and separated on ABI 3730xl capillary sequencers. After quality clipping, sequences were assembled based on overlaps and read pair information 69. Contaminations were identified by BLASTN searches in GenBank sections bacteria, protozoa and phage (version 161, 15.08.2007 70) and removed. Then, a database comprising all sequences was built and each sequence searched against the database using WU-BLASTN. A sequence was regarded redundant if: in addition to the best hit against itself another hit was found (p ≤ 10^-30, ≥95% identitiy); and no other matches were found. Redundant sequences were excluded. Remaining data comprise 5,540 contigs and 5.36 Mb for N. furzeri GRZ, 5,686 contigs and 5.39 Mb for N. furzeri MZM-0403, 6,273 contigs and 5.36 Mb for N. kunthae (Table 1). Contigs are referred to as sequences in the text. All Sanger sequences were submitted to the National Center for Biotechnology Information (NCBI). Also, a database was created and is available for BLAST searches at 71. For genomic sequencing using Roche/454 Life Sciences GS20 technology, 5 μg of N. furzeri GRZ genomic DNA were used and 111.6 Mb sequenced (Table 1). These sequences were not assembled prior to analyses and submitted to the trace archive at NCBI.

Primers were designed using the GAP4 module of the Staden Sequence Analysis Package 72 and ordered from Metabion (Martinsried, Germany). For PCR, 50 ng DNA, 10 pmol of each primer and PuReTaq Ready-To-Go PCR beads (GE Healthcare, Munich, Germany) were used; initial denaturation was at 94°C for 30 s followed by 35 cycles of 60 s at 94°C, 60 s at primer annealing temperature, 60 s extension at 72°C, and a final extension at 72°C for 300 s. PCR products were sequenced as described above. Sequences were visualized, edited and assembled in GAP4.

Microsatellites were identified in the 5.4 Mb of N. furzeri GRZ using Sputnik (see Repeat identification). Flanking primers were designed in sequences with: one microsatellite; at least 20 repeat units for dinucleotide repeats, or at least 10 repeat units for tri- and tetranucleotide repeats; at least 100 nucleotides of flanking sequence; perfect repeats. PCR was performed in two pools of GRZ DNA (see DNA/RNA) with one primer being 6-FAM-labeled. For genotyping, ABI 3730xl sequencers and GeneMapper software Version 4 was used. Allele calling was performed independently by two individuals. Discrepancies were resolved by retyping. For resequencing of gene-associated markers, primers were designed in 19 sequences with significant homology to known genes in at least one of the four fish reference genomes. PCR and sequencing was performed in two DNA pools of GRZ, and ten individual N. kunthae fish as described above.

In order to compare our genomic data of N. furzeri strains GRZ and MZM-0403, and N. kunthae, which we obtained by random genome-wide sample sequencing, we simulated this approach in silico for the other four fish species. That is, we extracted random genomic data from the publicly available genome assemblies for each. In detail, these were 20 samples of random genomic sequence sets of medaka, stickleback, tetraodon and zebrafish from genome assemblies at Ensembl 21 (Tetraodon nigroviridis update release (2007-05-29), Gasterosteus aculeatus update release (2007-05-30), Oryzias latipes update release (2007-05-24), Danio rerio update release (2007-07-30)). For ten samples, each reference data set represented 0.4% of the respective genome and had a sequence length distribution corresponding to the 5.4 Mb of N. furzeri generated by Sanger sequencing. For the other ten samples, sequence sets comprised 6% of the respective genome with a length distribution corresponding to the 111.6 Mb of N. furzeri GRZ sequenced by 454/GS20 technology. For genome size estimation, random genomic sequences of medaka, stickleback, tetraodon, and zebrafish were extracted in the same way; each comprised 5.4 Mb and had a sequence length distribution similar to the N. furzeri 5.4 Mb.

To determine the coding fraction, a BLASTX search in Swiss-Prot/TrEMBL (version 50.6) was performed. Hits with p < 10^-10 were regarded significant. Reference data sets of the four fish species were analyzed in the same way. Then, N. furzeri GRZ and MZM-0403 sequences containing coding portions were used for TBLASTX searches in medaka, stickleback, tetraodon and zebrafish cDNA collections downloaded from Ensembl 21 (see 'Reference data sets' above for release dates). A p-value < 10^-10 was set as the threshold for significance.

First, tandem repeats were identified and masked (replaced by Ns) using Tandem Repeats Finder, version 4.00 73 with the following parameters: Match 2, Mismatch 7, Delta 7, PM 80, PI 10, Minscore 50, MaxPeriod 1,300 (half the length of the longest sequence in the N. furzeri 5.4 Mb). Second, known interspersed repeats - that is, sequences matching entries from Repbase Update 47 - were identified and masked using RepeatMasker (version open-3.1.8; AFA Smit, R Hubley and P Green, unpublished data) 74, with standard parameters. Third, for de novo repeat identification, RepeatScout, version 1.0.2 48 with standard parameters, except for a minimum repeat copy number of 3, was used on the pre-masked sequences. Finally, a species specific repeat consensus library was built using both the identified known repeat families and the previously unknown repeat consensus sequences.

To detect microsatellites (1-5 nucleotides) the program Sputnik 38 with parameters -v 1 -u 5 was used.

Of the 444 gene fragments of N. furzeri GRZ, we first identified 44 for which we could unambiguously assign human orthologs based on BLASTX searches in Swiss-Prot. Based on these assignments, we searched for orthologous regions in medaka, stickleback, tetraodon, and zebrafish genome assemblies at University of California Santa Cruz (UCSC) 75 (oryLat1, tetNig1, gasAcu1, danRer5, hg18) by inspection of MAF tracks. We extracted exons, which we regarded as true orthologs if: no duplicated gene was detected in the respective species; splice sites were conserved; and no frameshifts were detected if consecutive exons were concatenated. In total, this resulted in 71 orthologous exons, identified in all six species (average length is 142 bp for human exons), which represent fragments of 26 protein coding genes (sequences are listed in Additional data file 12, including exons of N. furzeri GRZ). Five of the gene fragments represent one exon, while the others span at least one splice site. For further analyses, sequences were concatenated, translated, aligned by ClustalW v1.81 with standard parameters 76 and trimmed at the ends if necessary. Coding sequences of the aging related genes Msra, Sirt1, and Tp53 were also translated and aligned. All alignments were concatenated, resulting in 4,828 amino acid positions, and analyzed by MEGA4 77 with different methods (alignment available on request). Positions containing gaps and missing data were eliminated, leaving a total of 4,245 positions in the final dataset, of which 508 were parsimony informative. A maximum parsimony tree was obtained using the Close-Neighbor-Interchange algorithm in which the initial trees were obtained with the random addition of sequences (100 replicates).

Nucleotide sequences of putative orthologs of the four analyzed fish species were identified as described above and are listed in Additional data file 7. Nucleotide and translated sequences were aligned using needle (EMBOSS, version 3.0.0 78).

The 5.4 Mb of N. furzeri GRZ genomic DNA sequence generated by the Sanger technique have been deposited under the project accession [GenBank:ABLO00000000]. The version described in this paper is the first version ABLO01000000. The 111.6 Mb of N. furzeri GRZ genomic sequence generated by Roche/454 technology have been deposited at GenBank under project id (PID) 29535. The 5.4 Mb of N. furzeri MZM-0403 genomic DNA sequence generated by the Sanger technique have been deposited under the project accession [GenBank:ACCZO00000000]. The version described in this paper is the first version [GenBank:ACCZ01000000]. The 5.4 Mb of N. kunthae genomic sequence generated by the Sanger technique have been deposited under the project accession [GenBank:ACDA00000000]. The version described in this paper is the first version [GenBank:ACDA01000000]. N. furzeri GRZ major rRNA gene cluster [GenBank:EU780557]; Cdkn2b [GenBank: EU271680]; Cdkn2d [GenBank: EU400615]; Msra [GenBank EU400617]; Sirt1 [GenBank:EU271679]; Tp53 [GenBank:EU271681].