We previously constructed a linkage map by surveying the segregation patterns of 534 SNPs detected in 190 first-generation backcross (BC₁) individuals from a single pair mating between a p50T female and an F₁ male (p50T female × C108T male) 18. Based on the analysis of additional data with Mapmaker/exp (version 3.0 25; LOD [log of the odds] score 3.0) using the same mapping panel, we successfully positioned a total of 1,755 SNPs on an expanded linkage map. The SNP markers segregated into 28 linkage groups, with a total recombination length of 1,413 cM. We assigned 26 of the SNP linkage groups to classical silkworm chromosomes 1 to 26, defined by morphologic markers (for example, cocoon color and larval markings) and protein polymorphisms (for instance, hemolymph proteins), as reported previously 18. Unambiguous morphologic markers are not yet available for the remaining two classical linkage groups; therefore, we anchored linkage group 27 to a reference gene (vitellogenin), and arbitrarily defined linkage group 28.

The extended SNP linkage map is illustrated in Figures 1 to 526 (Additional data file 1 for details of each marker, including BAC accession number). Basic map parameters are significantly improved in this version. The number of markers per linkage group varies from 20 (group 2) to 105 (group 4; Figure 1), and the recombination length for each linkage group ranges from 42.2 cM (group 15; Figure 3) to 68 cM (group 24; Figure 4). In the previously reported SNP map 18, the minimum and maximum number of markers were seven (groups 26 and A; Figure 5) and 32 (group 10; Figure 2), and the linkage map lengths ranged from 27 cM (group 20; Figure 4) to 64 cM (group 11; Figure 2). The number of markers for individual linkage groups in the revised map increased proportionally to the threefold rise in the total number of markers, but the extension of the linkage maps remained relatively small, because of a higher marker density. The average distance between the markers is 0.81 cM, which is much improved compared with that of the previous map (2.5 cM). The markers are not evenly distributed throughout the linkage map, and so different regions are more densely or sparsely populated. The number of gaps with lengths exceeding 10 cM decreased to five from 14 in the previous map, and the largest gap length decreased to 12.4 cM from 21.3 cM.

We fingerprinted a total of 81,024 BACs from three BAC libraries, each made with a different restriction enzyme (Table 1), using the large-scale agarose gel-based restriction fingerprinting method 2728. We used the computer program FPC V6.0 2930 to assemble BAC contigs from the BAC fingerprints. We performed preliminary tests to determine appropriate tolerance and cut-off value parameters, and adopted values of 3 and 1 × e^-12, respectively. A lower stringency condition produced larger contigs, but it also increased the risk for false contigs because of a high density of repetitive sequences.

Of the clones fingerprinted, we deleted 2,246 BACs during fingerprint editing because of insert-empty clones, having too many bands because of possible contamination, or too few bands (fewer than five). In addition, we removed 1,152 BACs from contigs assembled by the FPC program because they formed an extensive false contig, in which several constituting BACs were assigned to different chromosomes by independent BAC-fluorescence in situ hybridization experiments (data not shown). The false BAC contig is likely to have formed from transposon-rich regions with similar restriction fingerprints, which were difficult to remove even by the use of high-stringency assembly conditions (namely, with a cut-off below 1 × e^-12).

The resulting physical map contained 6,221 contigs, among which 782 major contigs included eight or more BAC clones, as summarized in Table 2 (Additional data file 2). The total length of the 782 major contigs was 376 megabases (Mb), which corresponds to 79% of the silkworm genome (476 Mb measured by flow cytometry; Johnston JS, personal communication).

We evaluated the reliability of the predicted BAC contigs using two approaches. First, using BAC end sequences 31 we designed primer sets for the BAC clones belonging to a putative contig to determine the presence or absence of the sequence in each BAC clone independently by PCR. Using this method, we found mis-assemblies in six contigs out of the 782 major BAC contigs; we could not determine whether nine additional contigs, which we denote as 'doubtful', were correctly formed. In most cases of doubtful contigs, only one BAC clone bridged two well defined clusters of overlapped BACs. Such a BAC clone is likely to be chimeric. Therefore, we removed such chimeric and doubtful BAC clones from the 15 BAC contigs involved, which no longer represented a contradiction in our evaluation.

The second approach was a comparison with SNP markers. If a BAC contig included more than two SNPs, then those markers should be positioned around the same locus. A total of 128 BAC contigs contained more than two SNP markers, among which we found contradictions in 11 BAC contigs. We checked the 11 contigs by PCR using primers designed from BAC end sequences, as described above. However, we were unable to resolve the contradictions because most of the BAC end sequences in question corresponded to repetitive transposon sequences, and PCR provided similar bands even if BACs in the false contig belonged to different chromosomes. This situation reinforced the supposition that BAC clones forming all or most of the false contigs were derived from transposon-rich domains that produced similar restriction band patterns.

Altogether we mapped 581 BAC contigs containing 6,061 BACs onto 28 chromosomes through BAC clones containing SNP markers common to the linkage map. The length of mapped singletons and BAC contigs calculated for each chromosome is shown in Table 3. A total of 361.1 Mb are covered by BACs, which corresponds to 76% coverage of the genome.

Screening of mapped BACs harboring functional genes by HDR filter hybridization using EST probes is a powerful tool for positional cloning. A large EST database is available in silkworm 222324. Therefore, in addition to using DNA fingerprinting for the construction of BAC contigs, we employed the 'overlapping' method with EST markers, whereby BAC clones arrayed on HDR filters are subjected to large-scale screening by hybridization with individual, nonredundant ESTs to identify clones carrying single copy sequences. This approach helped confirm BAC contigs and allowed us to identify functional genes on the combined physical-genetic map. For screening we used the RPCI-96 BAC library, consisting of 36,864 clones with an average insert size of 168 kb, which corresponds to 13× redundancy.

From the number of positive BAC clones screened by HDR filter hybridization, we could roughly estimate whether the probe cDNA was a single-copy or multiple-copy gene, or contained a repetitive sequence. Table 4 summarizes the results of the EST hybridization experiments. For EST mapping, we employed only putative single-copy genes, based on filter hybridization characteristics and the number of positive hits per filter. For a single-copy gene, approximately 13 BACs should give hybridization signals. Of 692 putative single-copy genes identified by this procedure, we were able to assign 523 EST markers to chromosomes by identifying BACs common to fingerprinted contigs that had been integrated with the genetic map via SNP markers (Additional data file 1). We identified an additional 353 ESTs on the mapped contigs by BLAST search of BAC end sequences. Of these 152 were duplicates, yielding a total of 724 mapped single-copy genes. For confirmation of the initial map assignments, we designed specific primer sets to amplify expected ESTs on HDR filter-screened BACs by PCR (Additional data file 1).

In parallel with the HDR filter hybridization experiments, we carried out RFLP analysis of segregants from a backcross between an F₁ female (p50T × C108T) and a C108T male using a common set of ESTs as hybridization probes. The lack of meiotic crossing over in females results in complete linkage, enabling fast and efficient chromosome assignment of large numbers of markers using small segregant populations 13. We assigned a total of 1,082 ESTs to linkage groups by RFLP analysis (Table 5 and Additional data file 3); of these, 118 ESTs were mapped in duplicate using HDR filter hybridization or BAC end SNPs. In addition to providing independent confirmation of linkage assignments on the integrated SNP-physical map, the linkage assignment of 964 new ESTs will enable future annotation and evaluation of mapped scaffolds in the WGS assembly now in progress 78 (Mita K, Xia Q, personal communication).

Altogether we assigned 1,688 independent silkworm genes to 28 linkage groups (Table 5). Of these, there is positional information for 724 genes, whereas 964 ESTs are simply assigned to a chromosome. We tested these 1,688 genes for orthology between silkworm and other model holometabolous insects for which complete genome data were available, notably A. mellifera, T. castaneum, A. gambiae, and Drosophila melanogaster, in order to compare the syntenic relationships among them. We did not conduct an analysis of synteny relative to dipteran chromosomes because of their small number, which would produce many false connections, and their high rate of chromosome rearrangement 323334, which would probably reduce the number of significant syntenic blocks within chromosome arms or segments. Among 1,688 silkworm genes, we found 769 orthologs for A. mellifera and 790 for T. castaneum that could be used for a test of synteny. We then checked the distribution of silkworm gene orthologs in honey bee and beetle chromosomes (Additional data file 3).

Figure 6 presents Oxford grids showing the number of shared orthologs mapped in honeybee (555 total) and beetle (628 total) on silkworm chromosomes. Based on the simplifying assumption that chromosomes are of equal length and shared orthologs are uniformly distributed throughout the genome, the ratios of observed versus expected silkworm orthologs per chromosome should be 1.0. The ratios of shared orthologs per chromosome fell in the range of 0.75 to 1.3 in the 16 honey bee chromosomes 35 and 0.8 to 1.2 in the ten beetle chromosomes 36, which is consistent with the possibility that the mapped silkworm genes were randomly sampled for both genomes.

A striking feature of this analysis is that a very poor syntenic relationship exists between silkworm and honey bee (Figure 6a) compared with that of silkworm and beetle (Figure 6b). Thus, only six chromosomes of the silkworm genome, namely linkage groups (LGs) 10, 12, 15, 17, 22 and 28 (shaded in Figure 6a), share high syntenic conservation with honey bee chromosomes (eight or more shared orthologs). In contrast, 20 silkworm chromosomes, namely LGs 1, 2, 3, 4, 6, 8, 9, 10, 11, 12, 13, 15, 16, 17, 18, 19, 20, 22, 23 and 25 (shaded in Figure 6b), exhibit high correspondence of synteny conservation relative to beetle chromosomes. Consistent with these findings, honey bee chromosomes had fewer shared silkworm orthologs than did beetle (average 9 versus 12.7 orthologs per chromosome), and the fraction of shared orthologs in syntenic groups was much smaller (0.07 in honey bee versus 0.35 in beetle). For most of the orthologous genes, homology scores between silkworm and honey bee were similar to those between silkworm and beetle, indicating that the rate of evolutionary change was comparable, whereas those for Diptera were frequently lower (data not shown).

The inbred silkworm strains p50T and C108T, maintained at the University of Tokyo, were used as parental strains for the mapping panels. For linkage map construction, the same population of 190 segregants of a single-pair backcross (BC₁) between a p50T female and an F₁ male (p50T female × C108T male) was used as the first-generation SNP map 18.

Genomic DNA of parental strains, F₁ individuals and segregants from the female-informative backcross was isolated from whole bodies of fifth instar larvae after removing midguts and hemolymph, as described in a previous report 14. Genomic DNA of individual BC₁ segregants was isolated from whole pupae using DNAzol (Invitrogen Japan K.K., Tokyo, Jpn) after freezing in liquid nitrogen and homogenization with stainless steel beads.

For the linkage map construction, SNPs, including small base insertions and deletions, were used. The SNPs were identified in a large number of PCR amplicons that were synthesized using primers designed from the data obtained by BAC end sequencing, as reported previously 18. Briefly, for each end sequence, we designed a PCR primer pair using Primer3 45, and performed PCR amplification of the genomic DNA of the parental (p50T and C108T) and F₁ strains with ExTaq (TaKaRa Bio Inc., Otsu, Shiga, Jpn), using the manufacturer's instructions. We detected the presence of SNPs in these amplicons by sequencing all three genotypes and analyzing the resulting traces using PolyPhred 46.

Linkage map construction was carried out as previously reported 18. Genomic DNA of the same 190 BC₁ segregants was amplified with primer sets corresponding to the SNPs detected between parent strains, and polymorphism in the segregants was determined (Additional data file 1). Segregation patterns were analyzed using Mapmaker/exp (version 3.0 25) with the Kosambi mapping function 47. For genotyping polymorphisms, we directly sequenced PCR amplicons from BAC end regions and used a fluorescent polarization dye terminator SNP detection assay 1848 in parallel.

Three BAC libraries prepared using different restriction enzymes were emplyed in the present study; their characteristics are summarized in Table 1. A BAC library was constructed from genomic DNA of strain p50T fifth instar day 3 posterior silk glands partially digested with EcoRI 21, designated RPCI-96 (RP96), and is available from BACPAC Resources of the Children's Hospital Oakland Research Institute 49. The HindIII-BAC library was prepared using HindIII partial digestion of the same p50T strain DNA 20. The BamHI library, similarly prepared with BamHI using DNA from a strain derived from the same line as p50T (Wu C, Goldsmith M, personal communication) was purchased from the Laboratory for Plant Genomics and GENEfinder Genomic Resource of Texas A&M University 50. All three libraries were fingerprinted for the construction of genome-wide BAC contigs. The RPCI-96 and BamHI-BAC libraries were used for SNP analysis; the EcoRI-BAC library was used for EST mapping by HDR filter hybridization.

RPCI-96-BAC clones arrayed in duplicate in a specific pattern onto nylon membranes (HDR filters) were obtained from BACPAC Resources of the Children's Hospital Oakland Research Institute 49. Labeling, hybridization, and detection were performed using the ECL Direct Nucleic Acid Labeling and Detection System (GE Heathcare UK Ltd., Little Chalfont, Buckinghamshire, UK), in exact accordance with the manufacturer's instructions 21. Probes derived from ESTs were obtained by PCR amplification of plasmid cDNA inserts using universal sequencing primers corresponding to the plasmid vectors (see below). EST library construction and analysis have been reported 22.

The BLASTN 51 search of 1,755 BAC end sequences was carried out against an in-house collection of B. mori cDNA/ESTs containing 185,765 sequences. In this search, the e value (a probability cut-off value) was set to 1 × e^-50 and no complexity filter was used. The cDNA/ESTs of B. mori were retrieved from NCBI-GenBank Flat Files (release 159.0) with a custom Perl script (Additional data file 1).

DNA fingerprinting analysis was used to construct BAC contigs. The detailed protocol is reported in Marra and coworkers 27, and we followed the steps described by Osoegawa and coworkers 28 exactly. BAC DNA was isolated from 1.2 ml cultures of LB medium containing chloramphenicol using an automated method (PI-1100; Kurabo Industries Ltd, Osaka, Japan) and dissolved in 50 μl of TRIS-EDTA buffer. Approximately 6 ng (3 μl) of DNA was digested in a 20 μl reaction containing five units of EcoRI at 37°C overnight. When the digestion was completed, 4 μl of 6× loading buffer (15% Ficoll [Sigma-Aldrich Japan K. K., Tokyo, Jpn], 0.25% bromophenol blue, and 0.25% xylene cyanol) was added and 3 μl of EcoRI-digested DNA was separated on a 1% agarose gel in 1ξ Tris-Acetate-EDTA buffer at 90 V for 15 minutes, then at 45 V for 12 hours. Electrophoresis was carried out in Horizon 20-25 electrophoresis tanks (Life Technologies, Gaithersburg, MD, USA) at 16°C. Size markers containing 12.5 ng/μl Analytical Marker DNA, Wide Range Ladder (Promega K. K., Tokyo, Jpn), and 2 ng/μl Marker V (Roche Diagnostics K.K., Tokyo, Jpn) were loaded in every fifth lane. After electrophoresis, the gel was stained in 500 ml of 1:1,000 dilution of SYBR Green 1 (FMC BioProducts, Rockland, ME, USA) in 1× TAE for 30 minutes and scanned using a Molecular Imager FX (Bio-Rad Laboratories Inc., Tokyo, Jpn). The gel image was analyzed using Image 3.9d software 5253, followed by FPC v.6.0 software 2930.

Southern blotting was used to assign EST clones to linkage groups. After sequencing and homology search 22, unique clones were amplified by PCR with plasmid-specific primers to encompass the inserts such as M13 reverse or T3/M13M3, M13M4, or T7 primer sets. Amplified EST DNA was labeled using the ECL Direct Nucleic Acid Labeling and Detection System (Amersham-Pharmacia Biotech). BC₁ segregants from a single female informative pair mating ([p50T × C108T] female × C108T male) were used for analysis of linkage, using 16 samples per probe. When the DNA of a sample was used up a new sample was taken from the same family and assayed with a reference set of anchor loci to determine the inheritance pattern for each linkage group 13. To test polymorphism, digestion was carried out on 2.4 μg parental DNA with one of six restriction enzymes, BamHI, BglII, EcoRI, HindIII, KpnI, or SacI, and subjected to 0.8 % agarose gel electrophoresis for 16 hours at 14°C. Gel treatment for denaturation, depurination, neutralization, blotting onto nylon membranes (nylon membranes, positively charged; Roche Diagnostics K. K., Tokyo, Jpn), and probe labeling and hybridization were performed in exact accordance with the manufacturer's instructions. Selection of suitable enzymes to detect the RFLPs between the two parents and the analysis of linkage in the BC₁ segregants was carried out using the method of Kadono-Okuda and coworkers 13.