A first generation physical map of the medaka genome in BACs essential for positional cloning and clone-by-clone based genomic sequencing

Mechanisms of Development 121 (2004) 903–913 www.elsevier.com/locate/modo

Research paper

A first generation physical map of the medaka genome in BACs essential for positional cloning and clone-by-clone based genomic sequencingq Maryam Zadeh Khorasania, Steffen Henniga, Gabriele Imrea, Shuichi Asakawab, Stefanie Palczewskia, Anja Bergera, Hiroshi Horic, Kiyoshi Narused, Hiroshi Mitanie, Akihiro Shimae, Hans Lehracha, Jochen Wittbrodtf, Hisato Kondohg, Nobuyoshi Shimizub, Heinz Himmelbauera,* a

Department of Vertebrate Genomics, Max-Planck-Institute of Molecular Genetics, Ihnestr. 73, 14195 Berlin, Dahlem, Germany Department of Molecular Biology, Keio University School of Medicine, 35 Shinanomachi, Shinjuku-ku, Tokyo 160-8582, Japan c Division of Biological Sciences, Graduate School of Science, Nagoya University, Nagoya 464-8602, Japan d Department of Biological Sciences, Graduate School of Science, The University of Tokyo, Bunkyo-ku, Tokyo 113-0033, Japan e Department of Integrated Biosciences, Graduate School of Frontier Sciences, The University of Tokyo, Kashiwa, Chiba 277-8562, Japan f European Molecular Biology Laboratory, Developmental Biology Programme, Meyerhofstr. 1, 69012 Heidelberg, Germany g Laboratory of Developmental Biology, Graduate School of Frontier Biosciences, Osaka University, 1-3 Yamadaoka, Suita, Osaka 565-0871, Japan b

Received 15 January 2004; received in revised form 29 March 2004; accepted 30 March 2004

Abstract In order to realize the full potential of the medaka as a model system for developmental biology and genetics, characterized genomic resources need to be established, culminating in the sequence of the medaka genome. To facilitate the map-based cloning of genes underlying induced mutations and to provide templates for clone-based genomic sequencing, we have created a first-generation physical map of the medaka genome in bacterial artificial chromosome (BAC) clones. In particular, we exploited the synteny to the closely related genome of the pufferfish, Takifugu rubripes, by marker content mapping. As a first step, we clustered 103,144 public medaka EST sequences to obtain a set of 21,121 non-redundant sequence entities. Avoiding oversampling of gene-dense regions, 11,254 of EST clusters were successfully matched against the draft sequence of the fugu genome, and 2363 genes were selected for the BAC map project. We designed 35mer oligonucleotide probes from the selected genes and hybridized them against 64,500 BAC clones of strains Cab and Hd-rR, representing 14-fold coverage of the medaka genome. Our data set is further supplemented with 437 results generated from PCR-amplified inserts of medaka cDNA clones and BAC end-fragment markers. Our current, edited, first generation medaka BAC map consists of 902 map segments that cover about 74% of the medaka genome. The map contains 2721 markers. Of these, 2534 are from expressed sequences, equivalent to a non-redundant set of 2328 loci. The 934 markers (724 different) are anchored to the medaka genetic map. Thus, genetic map assignments provide immediate access to underlying clones and contigs, simplifying molecular access to candidate gene regions and their characterization. q 2004 Elsevier Ireland Ltd. All rights reserved. Keywords: Medaka; Bacterial artificial chromosome; Positional cloning; Physical map; Genome sequencing

1. Introduction The medaka Oryzias latipes is a small, 2 –3 cm in size teleost fish of the order Cypriniformes that is native to East Asia, primarily to Japan, but also to Korea and Eastern q Supplementary data associated with this article can be found, in the online version, at doi: 10.1016/j.mod.2004.03.024. * Corresponding author. Tel.: þ49-30-8413-1354; fax: þ 49-30-84131380. E-mail address: [email protected] (H. Himmelbauer).

0925-4773/$ - see front matter q 2004 Elsevier Ireland Ltd. All rights reserved. doi:10.1016/j.mod.2004.03.024

China, where its preferential habitats are brooks and rice fields. Currently, 22 species account for the genus Oryzias, with O. latipes as the only species distributed in the temperate zone. The remainder of species inhabit tropical lakes and rivers in South East Asia. Analysis of protein polymorphisms showed that wild populations of O. latipes can be divided into four groups which are, respectively, known as the North Japan, South Japan, East Korea and China/West Korea populations. Allozymic variations and differences in mitochondrial DNA characterize these four

904

M. Zadeh Khorasani et al. / Mechanisms of Development 121 (2004) 903–913

groups. Of the four populations, the North and South Japan populations are the most closely related. The average sequence divergence between inbred strains derived from the northern and southern medaka population is approximately 0.8% in the coding regions and 2.6% in the intron regions (Naruse et al., 2000). The northern population is genetically homogeneous and only very few genetic variations of protein and mitochondrial sequences have been observed (Sakaizumi, 1986). In contrast, the southern populations are genetically variable. Examples of strains from the south are Cab and Hd-rR, both used for physical mapping described in this work. However, strains Cab and Hd-rR are quite closely related (0.035% divergence in coding regions; own unpublished data). In Japan, breeding of medaka color variants for ornamental purposes has tradition that goes back several centuries. The popularity of medaka as a model species in developmental biology and genetics is the result of medaka’s relatively short life cycle, its ease of breeding, and its resistance to common fish diseases. Also, males and females are distinguished by a dimorphic anal and dorsal fin. Since the early 20th century, the physiology, embryology and genetics of medaka have been investigated extensively. In fact it was the study of the inheritance of body color in medaka by Toyama (1916) which demonstrated that Mendelian laws also applied to fish. Another milestone achievement, the discovery of Y-chromosome linked inheritance, was accomplished by Aida (1921). The past five years have seen rapid advances in the generation of genomic resources to study genes and gene functions in the medaka (Wittbrodt et al., 2002; Shima et al., 2003). EST projects carried out in Japan and Germany have jointly compiled sequence information on the majority of medaka genes, by sampling genes expressed in various embryonic stages and adult tissues (Kimura et al., 2004; Berger et al., in preparation). A gene-based genetic linkage map for the medaka has been constructed, resolving into 24 linkage groups, covering 1400 cM in males (Naruse et al., 2000, 2004). There is open access to linkage map data through the medaka database MBase (http://mbase.bioweb. ne.jp). Genomic libraries in bacterial artificial chromosome (BAC) vectors have been generated from inbred strains HNI (northern strain) as well as Cab and Hd-rR (Kondo et al., 2002; Matsuda et al., 2001; Amemiya et al., unpublished). Most importantly, screens for gene functions have been implemented in the medaka Cab strain by mutagenesis using ethylnitrosourea (ENU), to uncover genes that are essential for vertebrate development. The Medaka Genome Initiative (MGI) (http://gardy.dsp. jst.go.jp/MedGenIn) is a consortium of mostly Japanese or German research groups, dedicated to the creation and characterization of resources for an improved utilization of the medaka as a model species in functional genomics. Embedded within MGI, we have initiated a project to generate a physical map of the medaka genome in BACs. In the current absence of a contiguous sequence of the medaka

genome, a BAC map is an essential tool for positional cloning of genes identified in ENU mutagenesis screens. Different approaches have been suggested for BAC map construction. BAC fingerprinting is implemented by restriction digestion of BAC DNA, followed by electrophoretic separation of restriction fragments and band size estimation. If two BAC clones overlap, they will have restriction fragments in common, and thus will be placed into the same contig. BAC maps by fingerprinting have been constructed for multiple species, including man, mouse, rat and zebrafish (McPherson et al., 2001; Gregory et al., 2002; Krzywinski et al., 2004; http://www.sanger.ac.uk/Projects/D_rerio/ WebFPC/zebra/small.shtml), to name a few. There is several major drawbacks of the fingerprinting approach. First, clone contigs that are produced will initially be to a large extent anonymous. Therefore, in a second step, markers with known location in the genome will need to be typed on the clones, to provide anchoring points. Second, fingerprinting will only be successful if BAC resources generated from genotypes are used that do not differ to drastically. Polymorphisms that alter the size of restriction fragments will complicate contig construction and local clone order. Finally, small stretches of overlap between clones will remain undetected by fingerprinting, to prevent an unacceptably high background of falsely assembled contigs. On the positive side, contigs assembled by fingerprinting allow the very accurate size determination of target regions, because the final product of a BAC fingerprinting project essentially is the restriction map of a genome. Marker-content based approaches do not suffer from the drawbacks described above. Markers can be selected on the basis that contig location is known early on, BAC libraries from very different genotypes can be used and small overlaps between clones are readily detected. Technically, marker content mapping is conducted by PCR on pooled BAC DNA templates, prepared according to pooling schemes that reduce the amplification effort. Alternatively, hybridization assays of labeled PCR products or oligonucleotide probes against arrayed BAC pools or BAC colonies are performed. STS content mapping will result in a BAC map that is useful already at an early project stage, because of the advantage to have markers with known features (sequence, location) on the map. When final maps prepared by fingerprinting or STS mapping, respectively, are compared, both resources will be equally useful. By locating markers on fingerprinted clones, or by providing integration points via BAC end sequencing, BACs can readily be selected for further experimentation. The whole genome shotgun (WGS) approach has gained wide spread popularity for genome sequencing, because the generation of a large set of sequence data can easily be achieved (Venter et al., 1998). However, the assembly of WGS sequences into the context of large sequence scaffolds is not trivial (Batzoglou et al., 2002). As a compromise, a combination of WGS and BAC clone-based sequencing has been suggested. In practice, low-coverage sequencing of BACs is accompanied by parallel WGS read production.

M. Zadeh Khorasani et al. / Mechanisms of Development 121 (2004) 903–913

Subsequently, skimmed BACs are used as baits to identify matching sequence reads in the WGS data set, to obtain high sequence coverage of BAC inserts. This strategy has been successfully implemented to obtain a high-quality draft sequence of the rat genome (Rat genome sequencing project consortium, 2004). Low coverage sequencing from ordered medaka BACs, clone by clone (hence dubbed the CBC approach), will therefore significantly improve genome sequence quality, because assembly problems are essentially reduced to BAC-sized genome intervals. The map presented here will be of central importance to the CBCbased sequencing of the medaka genome. Our BAC map is to a large extent based on gene-derived markers, and encompasses a large proportion of markers currently employed for whole genome scans. Thus, genetic map assignments provide immediate access to underlying clones and contigs, simplifying molecular access to candidate gene regions and their characterization.

2. Results and discussion For the construction of a medaka BAC map by STScontent mapping, we took advantage of two available

905

resources, first, medaka cDNA sequences deposited in GenBank and second, the draft sequence of the pufferfish Takifugu rubripes genome (Fig. 1). Phylogenetic analysis places medaka in close relationship to the pufferfish, with an assumed divergence time of 60– 80 Myr, less than the estimated evolutionary distance between man and mouse (Wittbrodt et al., 2002). This lead us to assume that syntenies should have remained largely conserved between pufferfish and medaka, even though genome sizes differ twofold between the two species. Our assumption is supported by an analysis of sequence data generated from BAC clones located on medaka linkage group (LG) 22, revealing conservation of synteny, despite the expected size discrepancies that distinguish the two genomes (Sasaki et al., 2004). Published in 2002, the fugu genomic sequence can be regarded as a milestone achievement, as it constitutes the first almost complete genomic sequence from a teleost fish (Aparicio et al., 2002). The fugu genome v3.0 assembly from August 2002 encompasses 20,379 sequence scaffolds, that collectively cover 329.7 Mb. In the fugu assembly, scaffolds are sequentially ordered with respect to their size in base pairs. The large number of scaffolds does not adequately reflect the true quality of the sequence assembly.

Fig. 1. Strategy outline. (a,b) In silico cluster generation of public medaka EST data. (c) Alignment of clusters with the fugu genome sequence. Red and yellow circles are EST clusters from which probes are generated. For yellow clusters, genetic mapping information is available. (d) Probe hybridization against BAC arrays to generate the data set. (e) BAC contig construction. In the present example, one fugu scaffold translates into two medaka contigs, both of which are linked to same medaka chromosome by genetic map data. We refer to such adjacent contigs as map segments.

906

M. Zadeh Khorasani et al. / Mechanisms of Development 121 (2004) 903–913

For instance, scaffolds 1 – 300 together contain 30% and the first 1500 scaffolds encompass 70% of the total assembly. At the same time, a large number of EST sequences from medaka cDNA clones have been generated and deposited in the public domain. For the analysis presented here, 103,144 public medaka EST sequences were downloaded from the NCBI website (http://www.ncbi.nlm.nih.gov/). In order to generate a representative non-redundant medaka EST set we used a clustering procedure developed in-house (Poustka et al., 2003). In a first step, the ESTs were grouped into preclusters based on an all-against-all comparison by blastn (Altschul et al., 1990), followed by a hierarchical clustering algorithm. The pre-clusters were then assembled into the final sequence clusters based on consensus by the cap3 program (Huang and Madan, 1999). Following this procedure, the original 103,144 ESTs were merged into 21,121 sequence clusters of which 9379 were true clusters containing more than one EST, while 11,742 remained singletons. The number of 21,121 clusters will certainly still be inflated with respect to the number of different genes represented in the data set for several reasons. First, we adopted a conservative approach in cluster generation such that true sequence overlap was necessary for a decision whether two ESTs belonged to the same cluster or not. However, many cDNA clones had been sequenced from both ends and sequences did not overlap, and therefore were assigned to different clusters. Second, cap3 will split preclusters that contain an internal region of non-homology, observed, for instance, in the case of alternative splicing of transcripts. Third, the 30 end of a gene may vary due to the usage of alternative polyadenylation sites. Since the bulk of cDNA libraries had been constructed from oligo (dT) primed mRNA, the variation in the length of 30 UTR sequences could have a profound impact on clustering statistics. The last assumption is substantiated by the observation that only 38% of clusters matched a known protein in nrprot at a stringency cutoff of 1.0e210. Genes are valuable as markers on a map. They can be used as reference points for comparison with the maps and genome sequences of other organisms, because they remain conserved during evolution. However, gene density is nonrandom and varies greatly with respect to the chromosomal region concerned. To prevent oversampling of gene-dense regions, we aligned the set of 21,121 medaka EST clusters described above against the fugu genome draft. The 11,254 of clusters (53%) successfully matched a fugu scaffold. In total, 3397 fugu scaffolds were hit by at least one medaka cluster. These 3397 scaffolds cumulatively encompass 249 Mb of fugu sequence, corresponding to 75% of the total assembly. The ‘matched’ fraction of medaka clusters is vastly enriched in protein-coding entities. Sixty-two percent of ‘matched’ clusters contained protein homology to an nrprot entry at a stringency threshold of 1.0e210. Conversely, only 10% of ‘non-matched’ EST clusters found a match in nrprot. The fugu genome sequence is not entirely complete, for instance, 4.1% (13.7 Mb) of scaffold sequence

is composed of unspecified nucleotides, bridging known gaps within the scaffolds. Protein-coding genes represented in the medaka EST set that do not match a fugu scaffold could therefore be localized within gaps in the present sequence assembly. The high proportion of medaka EST sequences without protein homology in the ‘non-matched’ EST fraction, however, is highly indicative for an enrichment of sequences that are less well conserved during evolution, e.g. extended 30 UTR sequences. For most of medaka genes it is the sequences that are publicly available, but not the clones, as physical entities. We therefore devised a strategy that allowed us to fully exploit the wealth of existing medaka cDNA sequence data, by designing 35mer oligonucleotide probes. Our approach maximizes the flexibility to convert sequences to markers on the map, because the strategy is of course not restricted to probe design based on cDNA data. In fact, any STS or genetically mapped marker or BAC end-fragment, for which sequence information exists, can be converted into a marker for the mapping project. In addition, 35mer probes are free of vector sequences and can be designed such that known repeat sequences are avoided. For the current project, we designed 2363 35mer probes based on the alignment of medaka EST clusters against the fugu genome sequence (Table S1). Since the exon structures of genes remain conserved during evolution, this strategy prevents the inadvertent design of 35mer probes that are disrupted by intronic sequences in genomic DNA. As expected, the density of matched medaka EST clusters varied greatly between fugu scaffolds. On average, we observed one hit per 24 kb of fugu genomic sequence. As some clusters hit more than one location, 9586 cluster hits, corresponding to 8588 different clusters, are contained within the 1500 largest fugu scaffolds. These scaffolds range in size between 1145 kb (scaffold 1) and 52 kb (scaffold 1499). We aimed at selecting medaka clusters for probe design that matched 100 kb apart from each other in the fugu genome. Assuming the fugu genome as half the size of medaka and knowing that BAC clones in the Hd-rR library exceed 200 kb average size (see below), we anticipated that our probe design strategy would be most economical in terms of translating the fugu genome sequence into medaka BAC contigs. To accelerate data production, 35mer probes were labeled and hybridized in sets of six or eight probes, respectively, according to a 2D pooling scheme (Fig. 2). Subsequently, hybridization results were deconvoluted to obtain a list of positive clones for each probe (Table S5). We hybridized the probes against colony macroarrays prepared from clones of two different medaka BAC libraries, generated from inbred strains Cab and Hd-rR, both of them South Japan strains. The strains are important, because Hd-rR is the genotype that has been selected for generating the reference sequence of the medaka genome. The Cab strain is utilized for the ongoing ENU mutagenesis screens. Despite their different history, both strains are remarkably similar, with an estimated divergence of 0.035% in coding

M. Zadeh Khorasani et al. / Mechanisms of Development 121 (2004) 903–913

907

Fig. 2. Hybridization with pooled 35mer oligonucleotide probes against BAC colony macroarray. Hybridizations were carried out against a pair of identical filters containing 36,864 Hd-rR BAC clones spotted in duplicate. (a) Probe prepared from vertical pool of eight oligonucleotides (96-well plate positions A5– H5, see inset); (b) Probe prepared from horizontal pool of six oligonucleotides (96-well plate positions B1 –B6). The intersection of both pools is probe FS44923-OLc58.12d on LG21, Map segment 16. Matching positive clones obtained with either probe pool are indicated in red circles.

regions, corresponding to one single nucleotide polymorphism in 2800 bp. For the map project, we used 27,648 Cab clones and 36,864 Hd-rR clones. The average insert size in the Cab library is 150 kb and the average size of an Hd-rR BAC clone is 210 kb. Assuming a size of 800 Mb for the medaka genome, the BAC resources utilized corresponded to fivefold coverage for Cab and ninefold coverage for HdrR. Close to what is expected from the genome coverage of either library, we found on average 4.8 and 10.5 positive clones per probe in Cab and Hd-rR libraries, respectively. In total, 41,882 clones (65% of clones used, assuming no empty plate positions) were hit by at least one probe. The efficiency of our approach is supported by the observation that only 86 of 2363 35mer probes (3.6%) did not hybridize to a clone in neither library (Fig. 3). These probes may represent either 35mers with poor hybridization performance, or, alternatively, the regions they localize to are genuinely underrepresented in both BAC libraries. Of fugu scaffolds number 1– 1500, 1072 scaffolds are represented by at least one 35mer probe in our data set. The total number of 35mers for scaffolds 1 –1500 is 2086. In addition, our data set contains 191 further 35mer probes that localized to fugu scaffolds ranked with numbers higher than 1500. These probes were included, because most of them corresponded to genetically mapped medaka genes according to MBase. Our data set is further supplemented with 250 results generated from PCR-amplified inserts of medaka cDNA clones, produced at an early stage of the project on a subset of the available BAC clone resources (Hd-rR plates

145 – 192). In total, primary data are available for 2534 probes derived from expressed sequences, corresponding to a non-redundant set of 2323 loci (Table 1). Of these, 934 markers (724 different) are anchored to the genetic map according to MBase (Tables S3,S4). For map construction, we utilized the wprobeorder software (Mott et al., 1993; Grigoriev et al., 1998). Wprobeorder uses a simulated annealing algorithm to define an optimal order of markers, based on clones that are shared between the markers. The more clones two markers have in common, the closer they are located in the genome. Alternatively, wprobeorder accepts a given list of probes, in the format of a contig file, as input. For the purpose of identifying segments of the medaka genome syntenic to fugu scaffolds, we employed the latter option. Using this approach, a large fraction of the probes immediately fell into contigs, having clones in common with their nearest neighbors (Fig. 4; Table S2). Adopting this strategy allowed us to also take advantage of the probes that had hybridized to more clones than expected on average, for map construction. For instance, there is 572 probes in the data set that hybridized to 20 clones or more (Fig. 3). These results could be due to the presence of closely related gene sequences, e.g. pseudogenes, on other chromosomes that cross-hybridized with the probes used. However, when positive clones were considered only in their local context, multi-match probes could be successfully used, avoiding false links. In a second step we then refined the probe order manually, if our medaka data suggested a probe path

908

M. Zadeh Khorasani et al. / Mechanisms of Development 121 (2004) 903–913

Fig. 3. Number of clones per probe. We used two different BAC libraries, prepared from medaka inbred strains Cab and Hd-rR. The clone resources together represented a 14-fold coverage of the medaka genome. With 2359 gene-derived oligonucleotide probes used, on average 15 positive clones were detected (data from both libraries combined). In total, 41,882 clones were positive for at least on probe (65% of all clones used under the assumption that plates did not contain empty wells). Eighty-six probes did not hybridize wth any clone in either library and 153 probes hybridized with more than 30 clones.

deviating from the marker order in fugu. In regions with low probe densities, gaps in contigs were observed. We kept these contigs together and referred to them as map segments, if genetically mapped markers located within adjacent contigs suggested co-localization (Figs. 1,5a). Once map segments had been established, we then proceeded to inspect links between different map segments and connected them, if supported by genetic map data. This process of map refinement was repeated several times, until all possible links had been evaluated. Finally, we used the increment program (Schalkwyk et al., 2001) to integrate a small set of BAC end-fragment hybridization data into the map. Our final, edited, first generation medaka BAC map consists of 902 map segments, containing a total of 2534 markers from expressed sequences, 133 end-fragment markers from Hd-rR BAC clones and 54 end-fragments from Cab BACs. 462 of map segments are uniquely anchored to a single medaka chromosome (Fig. 5a,b). In a small number of cases, a map segment was anchored to a particular medaka chromosome by multiple genetically mapped markers, but also contained a marker derived from a different linkage group (Fig. 5b). The current data set contains 11 map segments with markers derived from two different medaka linkage groups and one map segment that has markers from three chromosomes integrated. An example is map segment 16, encompassing seven different markers that map to LG21 (Table S2). In addition, probe v3FS10-603-s000478, corresponding to MBase marker OLa2812c on LG22, is contained within map segment 16. The integration of probe v3FS10-603-s000478 into this map segment is very well supported by several BAC links to adjacent markers. We cannot fully resolve such data

conflicts at present. A possible explanation could be the presence of closely related gene copies on different medaka chromosomes, with one copy showing a polymorphism between the mapping strains, resulting in the genetic map assignment. We have also observed fugu scaffolds that cannot readily be translated into medaka BAC contigs. An example is map segment 1, arranged from fugu scaffold 1. In the present map display, six markers localized within 440 kb of fugu scaffold 1 do not have any BAC clones in common. A possible reason for this result could be very different chromosomal organization of such regions in medaka and fugu, for instance, regional expansion in medaka. To clarify such non-trivial situations, we have selected further probes and will incorporate results into more advanced versions of the map. Of the 902 map segments, 439 (49%) are equivalent to singleton contigs, currently not connected to the other map segments. On the other hand, our data set also contains five map segments consisting of more than 20 markers and 35 segments encompassing between 10 and 19 markers. These 40 largest segments alone will cumulatively encompass at least 10% of the medaka genome. We estimate that the average map segment that contains two or more markers has Table 1 Marker content of BAC map EST-derived, using 35mer probes EST-derived, cDNA clone inserts Hd-rR BAC clone end fragments Cab BAC clone end fragments Sum Sum excluding redundancies

2284 250 133 54 2721 2510

M. Zadeh Khorasani et al. / Mechanisms of Development 121 (2004) 903–913

909

Fig. 4. Map segment 65 from medaka LG21, containing 17 different markers. Probe v3FS651-100-s008943 (identical to AU169483), OLa1304f and XRCC5/ Ku80 are markers that have been genetically mapped to interval 24 of LG21 (see MBase). The segments containing probes v3FS651-100-s008943 and OLa1304f are not connected by BAC clones, but the medaka genetic map and fugu synteny data support co-localization in medaka. The map segment integrates probes that localize to different sequence scaffolds in fugu (scaffolds 651, 51, 319, 487, 541, and 146), and contains one pair of end-probes generated from Hd-rR BAC clone Hd-146N22.

an average marker content of 5.4 and an approximate length of 1 Mb, resulting in a cumulative length of 463 Mb. At the same time, 131 Mb will be contained within the 439 singleton contigs. We, therefore, estimate that the total map length is 594 Mb and approximately corresponds to 74% of the medaka genome.

3. Concluding remarks We here present the first BAC map draft for the medaka genome. Our map currently encompasses 2534 gene-derived markers and 187 end-fragments (Table 1). Each marker on the map can be addressed by its sequence, facilitating

910

M. Zadeh Khorasani et al. / Mechanisms of Development 121 (2004) 903–913

Fig. 5. Map segment and probe number statistics. (a) Number of map segments located on medaka LG1–LG24 and number of probes contained therein. LG02 map segments are underrepresented, because less markers exist for this chromosome. (b) Number of map segments that are mapped (M), unmapped (Un), or which are uncertain with respect to their location (ambiguous, A), and total number of probes associated with them. Ambiguous map segments contain markers that map to different chromosomes in medaka (see text for details).

the distributed use of our mapping information. Continued effort will be directed into resolving ambiguities present in the current map version, the bridging of gaps within map segments and to emphasize long-range connectivity. However, even in the current draft version the map has already been greatly useful, testified by multiple interactions with groups aiming at the positional cloning of mutations affecting formation of heart and thymus, as well as defects in eye and CNS development. Most importantly, clone-by-clone

sequencing of the medaka genome has been initiated in the framework of the Medaka Genome Initiative, using the map presented here as a backbone. CBC sequencing currently focuses on LG22 as a medaka model chromosome, with expansion planned for the remainder of the medaka genome (Sasaki et al., 2004). In conclusion, the data presented here will have a profound impact on the efficient use of medaka as a model system, to promote the translation of genomical information into biological knowledge.

M. Zadeh Khorasani et al. / Mechanisms of Development 121 (2004) 903–913

4. Experimental procedures 4.1. BAC libraries and colony filter production Plates number 1 –72 containing 27,648 clones from RZPD Lib. 756 (Cab BAC) (Amemiya et al., unpublished) were obtained from RZPD in Berlin, Germany with permission from J. Wittbrodt. Plates 145– 240 (36,864 clones) from the Hd-rR BAC library (Matsuda et al., 2001) were contributed by Hiroshi Hori. Colony macroarrays were produced in-house on custombuilt robotic equipment using a 384-pin spotting head. Clones were spotted in duplicate onto Hybond N þ membranes (Amersham), following standard procedures (Nizetic et al., 1991; Himmelbauer et al., 1998). We used different spotting densities for the two library segments, 5 £ 5 for Cab and 6 £ 6 for Hd-rR, in order to accommodate each segment onto a single macroarray. Spotting pins had a diameter of 400 mm for the 5 £ 5 pattern and 250 mm for 6 £ 6, respectively. Once colony plates had been spotted, they were immediately removed from the stacker and shockfrozen on dry ice. Thus, three to four spotting runs that resulted in high-quality macroarrays could be carried out for each library spotting copy. Following the spotting run, filters were incubated on top of 22 £ 22 cm2 agar dishes (2YT agar supplemented with 32 mg/ml chloramphenicol) for up to 24 h at 37 8C. Each membrane was inspected for uniform colony growth, with colony diameters approximating 0.5 mm. In situ colony lysis was carried out as described (Nizetic et al., 1991). Finally, filters were left to dry for two days and thereafter UV-fixed in a Stratagene crosslinking device. 4.2. PCR amplification of cDNA inserts, probe labeling and hybridization Genetically mapped cDNA clones (Naruse et al., 2000, 2004) were obtained in three different vectors, pME18s, pBluescript SK(þ ), and pUC118 (see http://mbase.bioweb. ne.jp/, dclust/lib_info.htm). PCR amplification was carried out with one of the following primer pairs: pM18F (GCTGCGGAATTCCTCGAGCAC) plus pM18R (CGCGACCTGCAGCTCGAGCAC); SK-F (TGGATCCCCCGGGCTGCAG) plus SK-R (TACCGGGCCCCCCCTCGAG); and pUC118-F (ACAGCTATGACCATGATTACG) plus pUC118-R (CCAAGCTTGCATGCCTGCAGG). Reactions in MJResearch PTC100 thermocyclers contained, in a reaction volume of 30 ml, 10 ng template DNA, 10 pmol of each primer, 125 mM of each dNTP and three units Taq DNA Polymerase. Reactions were heat denatured for 3 min, and then processed by 35 cycles of 94 8C (30 s)/annealing (30 s)/72 8C (90 s), followed by a final incubation at 72 8C for 10 min. The annealing temperature was 65 8C for clones in Bluescript and pM18s and 60 8C for pUC118. Radioactively labeled probes from amplified inserts (100 ng) were prepared with

911

the random priming method (Feinberg and Vogelstein, 1983), using 30 mCi of a32PdCTP. Reactions were incubated at room temperature over night and heatdenatured in a PCR-machine for 10 min at 94 8C. Hybridizations were carried out in a hybridization oven set to 65 8C in rotisserie bottles, containing one filter of each library, separated by a nylon mesh. Before adding probes, colony filters were pre-hybridized for a minimum of 1 h at 65 8C using 50 ml hybridization buffer (Church and Gilbert, 1984). Thereafter, the buffer was replaced with fresh 15 ml of fresh hybridization buffer containing the probe and incubation was continued for 16 h. Following hybridization, filters were washed twice for up to 30 min each at 65 8C in 2 £ SSC/0.1% SDS and 0.1 £ SSC/0.1% SDS. Exposure against Kodak XAR hyperfilm was carried out for up to two weeks at 2 80 8C. 4.3. Selection of 35mer oligonucleotide probes A non-redundant set of medaka cDNA sequences obtained by clustering public cDNA data available from NCBI (http://www.ncbi.nlm.nih.gov) was aligned against the Takifugu rubripes genomic sequence (v3.0 assembly from August 2002), that was downloaded from the JGI fugu genome project web site (http://genome.jgi-psf.org/fugu6/ fugu6.home.html). cDNA matches on fugu scaffolds were identified by blastn (Altschul et al., 1990) and exon alignments were refined using CrossMatch (http://www. genome.washington.edu/UWGC/analysistools/Swat.cfm). Results were tabulated into html format to facilitate data access. Oligonucleotide probes were chosen from matches 100 kb apart. A single representative oligonucleotide was selected for scaffolds smaller than 100 kb in size. Oligonucleotides were preferentially designed from scaffold matches corresponding to genetically mapped medaka genes listed in MBase (http://mbase.bioweb.ne.jp). Thirtyfive-mers were ordered from Metabion (Martinsried, Germany) and shipped lyophilized in 96-well plates. 4.4. Oligonucleotide labeling and hybridization Data production was accelerated by hybridizing oligonucleotides in pools. For each 96 well plate, 12 column pools of eight oligonucleotides each and 16 half-row pools consisting of six oligonucleotides each were generated. Pools were prepared from mixing oligonucleotides at a concentration of 10 pmol/ml. Twenty-five pmol of pooled oligonucleotides were radioactively labeled using 25 mCi of g32P-dATP and 10 units polynucleotide kinase (New England Biolabs) in a volume of 15 ml. Reactions were incubated at 37 8C for 1 h. Unincorporated g32P-dATP was removed using G-50 MicroColumns (Amersham). The hybridization procedure was the same as for PCR-generated probe fragments, except that the incubation and posthybridization washes were done at 58 8C. Two consecutive

912

M. Zadeh Khorasani et al. / Mechanisms of Development 121 (2004) 903–913

washing steps using 2 £ SSC/0.1%SDS and 0.5 £ SSC/0.1%SDS were carried out. 4.5. Data analysis Autoradiograms were scanned using a CCD camera connected to a PC, and image files were saved as Windows bitmap (BMP). For analysis, image files were imported into the program VisualGrid (Implemented by Markus Kietzmann with contributions from David Bancroft and Igor Ivanov. Copyrighted and Licensed by GPC Biotech AG). A defined grid with a spotting format, either for Hd-rR or Cab, was placed onto the colony array image. Positive signals were identified and selected by clicking. After selection, the image file was saved as Tiff (TIF). Selected positives were exported into file as Textfile (TXT) and saved again as Nonstandard Residue (RST) file. Scripts written in perl deconvoluted pool results into probe/clone files suitable for import into the contig building software wprobeorder (Mott et al., 1993; Grigoriev et al., 1998). 4.6. Generation of BAC end sequences BAC DNA was prepared from cultures grown in 96deep-well plates using standard alkaline lysis procedure (Birnboim and Doly, 1979). Ten nanograms of BAC DNA was digested to completion with one unit RsaI in a volume of 20 ml at 37 8C over night. Digested DNA was ligated into EcoRV cut Bluescript vector. The amplification of BAC end-fragments was accomplished using one primer specific for the BAC vector and another primer matching Bluescript. The following primers were used: For Hd-rR clones (vector: pBAC-lac): BAC-specific primer Reverse-HindIII (AGA GTC GAC CTG CAG GCA TGC AAG) or Forward-HindIII (AAA ACG ACG GCC AGT GCC AAG), in combination with Bluescript-specific primer pKS-lac (GAC GGT ATC GAT AAG CTT GA). For Cab clones (vector: pBAC3.6), we used primers Sp6-long (ATT TAG GTG ACA CTA TAG AA) plus T3-long (AAT TAA CCC TCA CTA AAG GG), and T7-Bac-Sp (TCA CTA TAG GGA GAG GAT CCG) plus T3-pKS-Sp (AAA GGG AAC AAA AGC TGG GTA CC). PCR reactions were set up as described above for cDNA inserts. Annealing was carried out at 52 8C for Sp6long/T3-long; 65 8C for T7-Bac-Sp/T3-pKS; and 588 for both ends of Hd-rR clones. The fragments were sequenced using standard ABI sequencing technology. 4.7. Data access Contig and probe data are available from H. Himmelbauer and can be accessed at the web site of the Mouse/Medaka/MHC group at www.molgen.mpg.de/ , rodent.

Acknowledgements We thank Susanne Blachut and Andreas Hirsch for excellent technical assistance, Elham Alishirkouhi for data entry, and Detlef Groth and Gu¨nther Teltow for data management and script implementation. We are grateful to Drs Masaru Nonaka and Hiroshi Kimura for providing unpublished mapping information on probe DRP2. H.H., H.K. and J.W. received support from HFSP grant RGP0040/2001-M (coordinator: J.W.). We gratefully acknowledge Dr Makoto Furutani-Seiki for continued discussions, helpful comments, and valuable advice.

References Aida, T., 1921. On the inheritance of color in a freshwater fish Aplocheilus latipes Temminck and Schlegel, with special reference to sex-linked inheritance. Genetics 6, 554–573. Altschul, S.F., Gish, W., Miller, W., Myers, E.W., Lipman, D.J., 1990. Basic local alignment search tool. J. Mol. Biol. 215, 403 –410. Aparicio, S., Chapman, J., Stupka, E., Putnam, N., Chia, J.M., Dehal, P., et al., 2002. Whole-genome shotgun assembly and analysis of the genome of Fugu rubripes. Science 297, 1301– 1310. Batzoglou, S., Jaffe, D.B., Stanley, K., Butler, J., Gnerre, S., Mauceli, E., et al., 2002. ARACHNE: a whole-genome shotgun assembler. Genome Res. 12, 177 –189. Birnboim, H.C., Doly, J., 1979. A rapid alkaline extraction procedure for screening recombinant plasmid DNA. Nucleic Acids Res. 7, 1513– 1523. Church, G.M., Gilbert, W., 1984. Genomic sequencing. Proc. Natl. Acad. Sci. USA 81, 1991–1995. Feinberg, A.P., Vogelstein, B., 1983. A technique for radiolabeling DNA restriction endonuclease fragments to high specific activity. Anal. Biochem. 132, 6– 13. Gregory, S.G., Sekhon, M., Schein, J., Zhao, S., Osoegawa, K., Scott, C.E., et al., 2002. A physical map of the mouse genome. Nature 418, 743 –750. Grigoriev, A., Levin, A., Lehrach, H., 1998. A distributed environment for physical map construction. Bioinformatics 14, 252–258. Himmelbauer, H., Wedemeyer, N., Haaf, T., Wanker, E.E., Schalkwyk, L.C., Lehrach, H., 1998. IRS-PCR-based genetic mapping of the huntingtin interacting protein gene (HIP1) on mouse chromosome 5. Mamm. Genome 9, 26–31. Huang, X., Madan, A., 1999. CAP3: a DNA sequence assembly program. Genome Res. 9, 868 –877. Kimura, T., Jindo, T., Narita, T., Naruse, K., Kobayashi, D., Shin-I, T., et al., 2004. Large-scale isolation of ESTs from medaka embryos and its application to medaka developmental genetics. Mech. Dev. 121, 915 –932. Kondo, M., Froschauer, A., Kitano, A., Nanda, I., Hornung, U., Volff, J.N., et al., 2002. Molecular cloning and characterization of DMRT genes from the medaka Oryzias latipes and the platyfish Xiphophorus maculatus. Gene 295, 213 –222. Krzywinski, M., Wallis, J., Go¨sele, C., Bosdet, I., Chiu, R., Graves, T., et al., 2004. Integrated and sequence-ordered BAC and YAC-based physical maps of the rat genome. Genome Res. 14, 766 –779. Matsuda, M., Kawato, N., Asakawa, S., Shimizu, N., Nagahama, Y., Hamaguchi, S., et al., 2001. Construction of a BAC library derived from the inbred Hd-rR strain of the teleost fish, Oryzias latipes. Genes Genet. Syst. 76, 61–63. McPherson, J.D., Marra, M., Hillier, L., Waterston, R.H., Chinwalla, A., Wallis, J., et al., 2001. A physical map of the human genome. Nature 409, 934–941.

M. Zadeh Khorasani et al. / Mechanisms of Development 121 (2004) 903–913 Mott, R., Grigoriev, A., Maier, E., Hoheisel, J., Lehrach, H., 1993. Algorithms and software tools for ordering clone libraries: application to the mapping of the genome of Schizosaccharomyces pombe. Nucleic Acids Res. 21, 1774– 1965. Naruse, K., Fukamachi, S., Mitani, H., Kondo, M., Matsuoka, T., Kondo, S., et al., 2000. A detailed linkage map of Medaka (Oryzias latipes): comparative genomics and genome evolution. Genetics 154, 1773–1784. Naruse, K., Tanaka, M., Mita, K., Shima, A., Postlethwait, J., Mitani, H., 2004. A medaka gene map: the trace of ancestral vertebrate protochromosomes revealed by comparative gene mapping. Genome Res. in press. Nizetic, D., Drmanac, R., Lehrach, H., 1991. An improved bacterial colony lysis procedure enables direct DNA hybridization using short (10, 11 bases) oligonucleotides to cosmids. Nucleic Acids Res. 19, 182. Poustka, A.J., Groth, D., Hennig, S., Thamm, S., Cameron, A., Beck, A., et al., 2003. Generation, annotation, evolutionary analysis, and database integration of 20,000 unique sea urchin EST clusters. Genome Res. 13, 2736–2746. Rat genome sequencing project consortium, 2004. Evolution of the mammalian genome: sequence of the genome of the Brown Norway rat. Nature 428, 493 –521.

913

Sakaizumi, M., 1986. Genetic divergence in wild populations of the fish Oryzias latipes (Pisces: Oryziatidae) from Japan and China. Genetica 69, 119–125. Sasaki, T., Asakawa, S., Shimizu, A., Ishikawa, S.K., Imai, S., Himmelbauer, H., et al., 2004. Medaka genome mapping and sequencing: toward complete genome sequence. Mar. Biotechnol. in press. Schalkwyk, L.C., Cusack, B., Dunkel, I., Hopp, M., Kramer, M., Palczewski, S., et al., 2001. Advanced integrated mouse YAC map including BAC framework. Genome Res. 11, 2142–2150. Shima, A., Himmelbauer, H., Mitani, H., Furutani-Seiki, M., Wittbrodt, J., Schartl, M., 2003. Fish genomes flying. Symposium on Medaka genomics. EMBO Rep. 4, 121–125. Toyama, K., 1916. Some examples of Mendelian characters. Nihon Ikushugaku Kaiho 1, 1–9. Venter, J.C., Adams, M.D., Sutton, G.G., Kerlavage, A.R., Smith, H.O., Hunkapiller, M., 1998. Shotgun sequencing of the human genome. Science 280, 1540–1542. Wittbrodt, J., Shima, A., Schartl, M., 2002. Medaka: a model organism from the far east. Nat. Rev. Genet. 3, 53–64.