We examined 1440 AFLP H. erato polymorphisms using 23 primer combinations (EcoCN/MseCNN). The number of AFLP bands per gel ranged between 26 and 132 with a mean of 72 bands per primer combination. Of these, approximately 84% were polymorphic in our outbred F2 cross. The experiment-wide error rate for our screen was approximately 1.0%, as inferred from discrepancies among female informative (FI) markers. In total, we scored 490 Male Informative (MI) and 470 backcross informative (BI) loci. Assuming an estimated H. erato genome size of 395 Mb 14, and assuming that AFLP markers are distributed randomly, suggests that we surveyed polymorphisms at approximately 362 kb intervals across the genome. This would suggest a resolution of 1.3 cM assuming that the relationship between physical and recombination distance is 276 kb/cM 9.

Our genome scan identified several AFLP markers 1–3 cM away from D. For the other gene, Cr, previous work using an identical strategy on crosses of H. melpomene provided markers within one cM of this gene in H. erato 15. In total, we identified five AFLP loci within a 3 cM target window around the D locus. Across our H. himera × H. erato notabilis mapping family for D, three loci (MI_EcoCA-MseCAA-114 bp, BI_EcoCC-MseCAG-155 bp, BI_EcoCT-MseCCG-139 bp) were perfectly linked and two loci (MI_EcoCC-MseCAC-527 bp, MI_EcoCC-MseCAC-485) showed only one recombinant. We cloned and sequenced several of these AFLP loci and developed PCR primers to amplify them from genomic DNA. Interestingly, two AFLP bands tightly linked to the D gene, MI_EcoCC-MseCAC-527 bp, MI_EcoCC-MseCAC-485, were allelic variants of the same locus.

We screened the H. erato BAC library with D-linked (MI_EcoCC-MseCAC-527 bp) and Cr-linked ("βggt-II" – Rab geranylgeranyl transferase beta subunit, βggt-II gene 15) probes. In our screens with these and other probes (15 probes in total, with an average of 12 positives per probe), we consistently observed between 9 and 15 PCR-confirmed positives per probe, suggesting the H. erato BAC libraries have approximately 10× genome coverage. The largest clones identified from the D-linked and Cr-linked probing experiments were sequenced at 8× coverage. The D-linked clone (BBAM-25K4) was composed of two large sequences that could easily be orientated to produce an approximately 180 kb genomic fragment (Figure 2). Similarly, sequence of the Cr-linked clone (BBAM-38A20) was composed of two large sequences that together spanned approximately 165 kb (Figure 3). The probe sequences were clearly identifiable in the D-linked and Cr-linked BACs and linkage to color pattern genes was confirmed by mapping (see below).

From the sequence of the first Cr-linked BAC, identified with the βggt-II gene, we designed additional probes to use for a second round of BAC library screening. Specifically we generated two more probes, corresponding to the genes Trehalase1 and B9, to expand our walk on both directions. With this strategy we identified new BACs on the 3' end that were positive for B9 and negative for Trehalase1 and others on the 5' end positive for βggt-II and negative for Trehalase1. After fingerprinting we selected one BAC to sequence from each end. Ultimately, BBAM-27D18 extended the overall contig by 211 kb on the 3' end, while BBAM-12K4 extended the contig on the 5' for 55 kb (Figure 4).

The D- and Cr-linked BAC sequences were AT-rich (65% and 66% AT content, respectively) and contained many repeats (Table 1). Although all three major classes of transposable elements (DNA transposons, LTR, and non-LTR retrotransposons) were present in the genomic sequences (Table 1A), the vast majority of repetitive sequences showed no significant BLAST similarity (i.e. e-value < 0.001) to any of the insect genomes currently available NCBI databases nor to any arthropod transposable elements listed in RepBase 19. Heliconius-specific repetitive sequences corresponded to the nine core motifs identified with RepeatFinder (Table 1B; Additional file 1: Novel repetitive elements in Heliconius (row sequences)). Of the nine motifs, six are present in both H. erato and H. melpomene, two are unique to H. erato, and one is unique to H. melpomene (Table 1; Additional file 1: Novel repetitive elements in Heliconius (row sequences)).

Gene density appeared to be relatively low across both the Cr- and the D-linked genomic regions. Although there were a moderate number of predicted open reading frames (ORFs) over 60 amino acids long (Figures 2 and 3), few showed any similarity to known proteins or lepidopteran ESTs, including our own collection of nearly 20,000 Heliconius ESTs (Table 2). For example, across the D-linked BAC we identified 75 hypothetical proteins using the Kaikogaas annotation tool and our own BLAST analysis. Over 90%, however, were less than 150 amino acids in length, only one of which showed similarity to any known or predicted protein. Across the entire ~190 kb region near the D locus in H. erato there were only two hypothetical proteins that showed significant homology to a known protein or contained a known structural element. One was similar to a sequence in our EST collection (HEC00815), while the other showed strong homology to a lepidopteran methionine-rich larval storage protein (Table 2). Although there was a smaller absolute number of predicted proteins across the Cr-linked BAC, there were more than twice as many predicted proteins of amino-acid length greater than 150 across, all but one of which showed strong homology to known proteins or to Heliconius ESTs (Table 2, Figure 2). Overall, the gene density we observed in H. erato is similar to what has been observed in the repeat-rich heterochromatin domains of Drosophila melanogaster, which averages 2.9 genes per 100 kb (versus 12.6 genes per 100 kb in euchromatin) 20.

VISTA analysis 21 (70% identity, 30 bp window – see Methods) showed a 35% conservation between the 164 kb H. erato Cr-linked BAC clone (BBAM-38A20) and the homologous sequence from H. melpomene (Figure 3). All of the predicted genes showed strong similarity (80–95% identity) between the two species and perfect overall synteny (Figure 3 and 4). Furthermore, a significant portion of 57 kb of presumed non-coding sequence (i.e. did not show notable open reading frames) was highly conserved between the two species (Figure 3). Despite the overall conservation between H. erato and H. melpomene sequences, two ESTs did show a difference between the species. Firstly, HEC03006 was found only in H. erato, and corresponded to a large indel sequence. Also, HEC01402 was found in the H. melpomene BAC sequence but not the H. erato sequence, although it shared some similarity with an exon of the Forkhead gene in the H. erato BAC sequence. For HEC01402, it is likely that the H. erato BAC sequence did not extend far enough to cover the homologus genomic region containing the gene in H. melpomene.

We found evidence for fine-scale synteny between H. erato and B. mori in the 420 kb genomic region linked to the Cr color pattern gene (Figure 5). No evidence for microsynteny was observed between D-linked H. erato genomic regions and B. mori, due to the lack of conserved genes in the D-linked clone. B. mori scaffold sequence (nscaf2829), downloaded from SilkDB 22, contained all of the major genes annotated on the Cr-linked BAC clones (Table 2, Figure 5). All genes were unambiguously identified: βggt-II (nscaf2829, position 2933018–2938173); Glucose dehydrogenase (nscaf2829, position 2929161–2931568); Forkhead Box (nscaf2829, position 2921212–2923967); Trehalase1 (nscaf2829, position 2866386–2868122); Trehalase2 (nscaf2829, position 2861182–2862921); B9 (nscaf2829, position 2830788–2833491); Unkempt (nscaf2829, position 2748736–2755694); Beta fructosidase FruA (nscaf2829, position 2704806–2706326). With the exception of the DNA helicase (nscaf2829, position 2870771–2872660), which appears to have been translocated, all of the gene orders and distances in Heliconius and B. mori are highly conserved. For example, Glucose dehydrogenase and Forkhead Box were separated by 5300 bp in H. erato and 5193 bp in B. mori, while Trehalase1 and Trehalase2 were separated by 3110 bp in H. erato and 3464 bp in B. mori. All seven genes showed 70–85% nucleotide acid sequence similarity between species (Figure 5). In addition to the major genes, there were many other genomic regions between Heliconius and B. mori with a nucleotide acid sequence similarity higher than 85% that did not show BLAST similarities to any known proteins.

We have previously demonstrated that the Cr locus in H. erato and the Yb gene of H. melpomene map to homologous areas of the genome 15. BAC genome sequence data for a region tightly linked to the Yb gene was obtained from H. melpomene using methodology similar to that described here. A single gene marker developed from the H. melpomene sequence mapped close to the Cr locus in H. erato. Here we provide the first genomic sequence evidence that, across a broad region around this gene, gene order and gene content is conserved. Across a 420 kb overlapping region all putative proteins showing high similarity to known proteins were in the same order in the two co-mimics (Figure 3 and 4). This further supports the hypothesis that a homologous gene, or set of genes, is responsible for color pattern variation in the two species. Indeed, with the exception of two large ORFs with strong sequence similarity to a reverse transcriptase (Table 2, Figure 3), gene order in the Cr-linked H. erato region and the N/Yb/Sb-linked H. melpomene region was nearly perfectly preserved. Furthermore, many of the smaller ORFs, as well as some non-coding sequence, were highly conserved both in the relative order and sequence. Generally, the H. erato and H. melpomene genomes appear to be structurally very similar. Because of this, linkage analyses and positional cloning efforts in each individual species should be highly informative for the co-mimetic species, and probably for the genus as a whole.

One of the most obvious differences between the H. erato and H. melpomene genomic sequences was the larger physical distance between homologous anchor points in H. erato relative to H. melpomene. This observation was probably related to the fact that the genome of H. erato is about 30% larger than that of H. melpomene 910. In this respect, it was notable that the difference in genome sizes between the two species was roughly proportional to the size of a number of sequence blocks that are absent in H. melpomene relative to H. erato in our genomic sequences (Figure 3). Many of these blocks were comprised of Heliconius-specific repetitive sequences, or showed strong similarity to known mobile genetic elements. These indel blocks appeared to be primarily noncoding sequences because none of them contained protein-coding sequences, with the exception of one EST (HEC03006) and a reverse transcriptase.

Differences in the genome size between closely related species is common and is usually associated with differences in abundance of different classes of noncoding DNA 23. It has been suggested that the insertion and replication of repetitive elements, as well as indel biases, can lead to profound differences in genome size 2425. Our observations suggest that both of these effects might be relevant in Heliconius, however, the differences in amount of repetitive DNA sequences in indels, at least over the small region that we examined, were not large enough to completely account for the differences in the genome size of the two species.

Using RepeatFinder 26 and RepeatMasker 27, we identified 16 different repeated elements in the Heliconius genomic sequences (Table 1). In total, repetitive sequences accounted for about 20% of the H. melpomene and about 26% of the H. erato genomic sequence. Seven of these repeated elements corresponded to previously described sequences (Table 1A). Because we were unable to detect any of the remaining nine repeats (all identified via RepeatFinder) in public sequence databases we assume these nine repeats represent novel repetitive elements unique to the Heliconius genus (Table 1B, Additional file 1: Novel repetitive elements in Heliconius (row sequences)). The seven previously described elements were larger (~1–5 kb) relative to the nine novel elements (100–600 bp) and occurred much less frequently. Most instances of these novel repeats observed in the BAC sequences were intact, full-length, highly similar versions of the core motifs. However, a wide range of fragmentation and divergence relative to the core motifs was also observed among repetitive regions. Motif #7 in H. erato best exemplifies this pattern, as it was the most abundant repeat observed, with 310 BAC regions showing significant similarity to the 266 bp core motif. These regions ranged continuously in size from 37 to 286 bp and in divergence from 2% to 32%. All other motifs showed a qualitatively similar pattern where BAC regions corresponding to a novel element ranged from being highly similar to the core motif sequence to being fragmented and divergent. This pattern is consistent with a process of motif replication and insertion followed by mutational degradation, suggesting these novel repeat motifs likely represent some sort of transposable element such as short interspersed nuclear elements (SINEs) or miniature inverted-repeat terminal elements (MITEs) 28. Both SINEs and MITEs have been reported from Lepidoptera 2930. More work is required to confidently determine the origin of these novel repetitive sequences in Heliconius. We are hopeful that future genomic sequences from other butterfly species will allow a better understanding the phylogenetic distribution and evolutionary origins of some of the observed repeats.

The observed fine scale synteny within Heliconius complements a recent study showing that patterns of macrosynteny are strongly conserved across the Macrolepidoptera 11. This previous study demonstrated that a total of 70 markers mapped in H. melpomene and B. mori showed large-scale patterns of synteny across the genome (taking into account a number of putative chromosomal fusions that explain the difference in chromosome number between these species). Our data further suggests that synteny has been preserved between Heliconius and B. mori on a much finer scale. Specifically, we show here that seven predicted genes in the Cr region have a similar order in the homologous B. mori genomic sequence (Figure 5). Although this is a very small sampling of the genome as a whole, it is still notable that gene order, as well as intergenic distances, has been preserved over such a long time scale.

The goal of this study was to identify and characterize regions of the genome linked to wing pattern polymorphism in Heliconius butterflies. We did not necessarily expect these initial BAC sequences to contain the color pattern genes themselves, however, these sequences provide important genomic "anchors" for ongoing positional cloning work. Fine-scale mapping experiments imply that we are very near the D and Cr color pattern loci. A microsatellite marker at the 5' end of the D-linked BAC showed 7 recombinants across 444 individuals, suggesting that this end is about 1.9 cM from the gene. There were two fewer recombinants for a marker developed from exon sequence of the Methionine Rich Storage Protein (MRSP) gene at the 3' end of the BAC. This marker is about 150 kb from our 5' microsatellite marker, suggesting that D is a further 1 cM in the 3' direction. We appear to have come down even closer to the Cr gene. In this case, a marker developed from the βggt-II gene near the 5' of the BBAM-38A20 BAC clone showed no recombinants across 430 individuals. A similar result was obtained with a marker developed from the Putative DNA helicase Rec-Q in the middle of the BBAM-27D18 BAC clone, suggesting that the zero recombinant interval might be somewhat large. For this reason, given an expected relationship of physical to recombination distance of 275 kb/cM 9, an expectation consistent with our initial mapping results, we are optimistic to identify the genomic recombinant interval containing both genes with the next one (Cr) to three (D) BAC steps.

In terms of identifying color pattern genes, there are obvious practical benefits of a fine-scale preservation of gene order between species. Foremost, conservation should greatly facilitate the identification of functional genes and the discovery of some regulatory elements associated with these genes 3132. Indeed, there were numerous conserved regions that were not simply protein coding regions (Figure 3). Even though this analysis covers only a small portion of the Heliconius genome, it confirms, for the first time, at a fine scale level what has been seen in comparative mapping projects concerning gene order conservation between different species in the genus 91033.

A comparative approach will be particularly important for pinpointing the regions responsible for pattern variation in Heliconius. Pattern formation in Heliconius probably involves discrete changes in conserved protein coding or regulatory regions 212. There is little precedent for what to expect, however, variation in pattern formation could be controlled by a number of cis-regulatory elements of a single gene, clusters of duplicated genes with divergent function, or clusters of non-paralogous but functionally-related genes.

We generated large F2 mapping families by crossing two different geographic races of H. erato to the same stock of H. himera (Figure 1). We followed segregating variation at both the D and the Cr loci in crosses between H. erato cyrbia and H. himera and the D and Sd loci using crosses between H. erato notabilis and H. himera (Figure 1). All crosses were carried out in the Heliconius insectary at the University of Puerto Rico from stocks originally gathered from the wild: H. himera was collected in Vilcabamba, Ecuador (79.13 W, 4.6 S), H. erato notabilis was collected from an area near Puyo, Ecuador (78.0 W, 1.5 S), and H. erato cyrbia was collected near Guayquichuma Glen, Ecuador (79.6 W, 3.9 S). After eclosion butterflies were euthanized, their wings removed for later morphological analysis, and their bodies frozen at -80 °C for later molecular analysis.

Genomic DNA was extracted following procedures described in 14 and approximately 500 ng of DNA was used as template of our AFLP reactions. Genomic DNA restriction digestions and the ligation of oligonucleotide adapters were performed using the Core Reagent Kit (Invitrogen Life Technologies) following manufacture's instructions. For all the other steps of the AFLP analysis (e.g. serial dilutions and PCR amplification protocols), we followed the modifications of the original protocol described by Vos et al. 34 as outlined in Papa et al. 35.

To efficiently identify primer combinations that contained loci tightly linked to color pattern genes, we used a modification of the "bulk segregant analysis" method 36. Specifically, we screened 23 AFLP primer combinations across an initial panel of 48 individuals arranged by color pattern genotype. Each reaction was run on a 10% polyacrylamide gel and visualized using fluorescently labeled (IRDye 700 or 800) EcoCN primers on a NEN^® Global Edition IR2 DNA Analyzer (LI-COR^® Biosciences, Lincoln, NE). In our screening gels, individuals with distinctive color-pattern genotypes were grouped and these groups were run side-by-side. For example, in the H. erato notabilis × H. himera cross we structured the gel with the following four genotypic groups: 1) D^himD^not, Sd^himSd^him; 2) D^himD^not, Sd^not Sd^not; 3) D^himD^him, Sd^himSd^not; 4)D^notD^not, Sd^himSd^not (Figure 1). In this way, we easily identified marker loci linked to the color pattern genes and those primer combinations showing tightly linked markers were further assayed on all 120 individuals in the brood.

Because of the nature of our outbred F2 cross design 910 we observed AFLP loci in all possible phases and segregation patterns. For each AFLP primer combination, we scored four different AFLP marker types: (1) monomorphic, (2) female-informative (FI) (AFLP band present in the mother and absent in the father), (3) male-informative markers (MI) (AFLP band present in the father but absent in the mother), and (4) markers that were present in both parents but segregated in offspring (BI). Both MI and FI markers were expected to segregate in a 1:1 ratio, whereas the BI markers, which were heterozygous in the parents, segregate in a 3:1 ratio. There is no crossing over during the oogenesis in Lepidoptera 3738 and FI markers on the same chromosome are inherited as a linkage block 10. Thus, only the MI and BI provided information on recombination distance between AFLP polymorphisms and color pattern genes. Nonetheless, the FI markers were extremely useful because they allowed us to unambiguously identify chromosomal linkage groups and estimate background error rate of the AFLP technique 910.

Genotypes for the color pattern genes were inferred from wing pattern phenotypes. Based on the amount of red, white and yellow in the forewing and hindwing, nine genotypes were observed among both groups of F2 progeny. For the D gene, shared by all three grand-parental types: homozygotes for the H. himera allele (D^hiD^hi) show only red present on the hind-wing bar, when D homozygotes for H. erato cyrbia or H. erato notabilis alleles (D^cyr D^cyr; D^notD^not) have red scales on the forewings, while D is heterozygotes between H. himera and the two H. erato races (D^hi D^cyr; D^hi D^not) have red pigmented scales on both forewings and hindwings. The Cr gene segregates only in the H. erato cyrbia × H. himera cross, and has an epistatic effect on the D gene by modulating the amount of red. Homozygotes with H. himera alleles (Cr^hiCr^hi) show a typical yellow forewing band when D is pure H. himera (Cr^hiCr^hi; D^hiD^hi) and a red/white forewing band (total amount of red pigments < 50%) when D is either homozygous for H. erato cyrbia alleles (Cr^hiCr^hi; D^cyrD^cyr) or heterozygous (Cr^hiCr^hi; D^hiD^cyr). When the Cr color pattern gene is heterozygous (Cr^hiCr^cyr) a yellow forewing shadow band is present when the D gene is homozyous for H. himera alleles (Cr^hiCr^cyr; D^hiD^hi), and an almost all red forewing band (red >50%) when D homozygote for H. erato cyrbia alleles (Cr^hiCr^cyr; D^cyrD^cyr) or heterozygote (Cr^hiCr^cyr; D^hiD^cyr). A white trailing dorsal edge and a yellow bar on the ventral hindwing identify the pure H. erato cyrbia form (Cr^cyrCr^cyr) with a totally black forewing when D is homozygous for H. himera alleles (Cr^cyrCr^cyr; D^hiD^hi), or totally red when the D gene is pure for H. erato cyrbia alleles (Cr^cyrCr^cyr; D^cyrD^cyr) or heterozygous (Cr^cyrCr^cyr; D^hiD^cyr).

To score color patterns, one must also understand activity of the Sd gene. Sd segregates in the H. erato notabilis × H. himera cross, where it controls the shape of the melanic window in the middle forewing of H. himera and shows a clear interaction with the distal forewing patch of H. erato notabilis. In the pure H. himera (Sd^hiSd^hi) a typical forewing band is shown, while in the pure H. erato notabilis (Sd^notSd^not) a shortened basal forewing patch, as well as a smaller distal forewing patch are evident. The heterozygotes (Sd^hiSd^not) present an intermediate form lacking a distal patch (as found in H. erato notabilis) and have a shortened basal patch showing melanin anterior to the costal vein (this area is normally light colored in H. himera).

We screened all primer combinations that generated BI or MI AFLP markers strongly associated with particular color pattern genes in our initial "bulked" sample across all 120 individuals from our mapping family. We excised and cloned those tightly linked AFLP markers larger than 160 base-pairs using a three-step strategy. First, the band was isolated from a polyacrylamide gel using a LI-COR^® Biosciences Odyssey^® Infrared Imaging System. We followed methods outlined in 39 and used a grid to position and excise specific fragments with a scalpel. We validated each excision by re-scanning the gel to confirm that we had removed the correct fragment. All gel fragments were placed in 15 ml of 1× TE and frozen at -80°C. Next, we re-amplified the AFLP using the original selective primer combination and the original PCR conditions. We generated template for this reaction by freeze/thawing the excised band three times. In each freeze/thaw cycle, we collected the resulting supernatant after the band was frozen for one hour at -80°C, heated at 55°C for 15 min and centrifuged for 15 min at 15,000 rpm. We checked amplification success by running PCR products on polyacrylamide gels adjacent to the positive control original AFLP reactions. Finally, we cloned the PCR product using Invitrogen's TOPO TA^® Cloning kit. PCR amplified inserts from 10–15 positive clones were again verified size on a polyacrylimide gel and those of the correct size were sequenced using DYEnamicT ET Terminator Cycle Sequencing Kit (Amersham Biosciences). Sequencing reactions were run on a MegaBACE 500 (Amersham Bioscience) or a 3130 DNA PRISM (ABI) at the Sequencing Facilities of the University of Puerto Rico, Rio Piedras. Resulting sequences were aligned by eye and PCR primers were designed using OLIGO version 4.0, and tested in genomic extracts from a panel of H. erato individuals.

Two H. erato BAC libraries, one partially restricted with EcoRI and one with BamHI, were created from a line of H. erato petiverana collected in Gamboa, Panama and inbred for several generations. The H. erato BAC library was constructed by C. Zhang (TAMU) and M. R. Goldsmith (URI) following the procedure outlined in Wu et al. 40. Both libraries contain 19,200 clones arrayed in 384-well plates and the average insert size for the EcoRI and BamHI libraries was estimated to be 153 kb and 175 kb, respectively. Libraries were gridded onto nylon membranes using a strategy where each clone is spotted twice to facilitate the identification of "true" positives. AFLP probes were labeled with P³² using the Prime-It II Random Primer Labeling Kit (Stratagene, CA, USA). The resulting radioactive labeled products were cleaned using a sephadex purification column and hybridized to the filters overnight at 65 degrees in Church Buffer (0.5 M NaHPO4 pH 7.2, 7%SDS, 1 mM EDTA, and 1%BSA) with rotation. The filters were washed twice with 2× SSC +0.1%SDS, then once or twice with 1× SSC + 0.1%SDS. The washed filters were placed on film for 1–5 days at -80 degrees, depending on signal strength.

All positive clones identified in our screen of the H. erato BAC libraray were PCR-confirmed using probe specific primers. Clones were then grown overnight on agar plate and single colony was used to inoculate TB media. Insert DNA was extracted using the Qiagen Maxi prep kit. This insertion size of each BAC clone is estimated by summing all fragments after restriction enzyme digestion using EcoRI or BamHI. The largest clone identified from our fingerprinting experiments was sequenced and assembled by the Baylor Genomics Center. Clones were first sheared to create 4–6 kb fragments and subcloned into pUC19. Approximately 8× sequence coverage of each BAC was then generated in paired 600–800 bp reads. Data were assembled using PHRAP 41 and edited in a GAP4 42 database.

The BAC sequences were analyzed using a variety of sequence annotation programs. First we used Kaikogaas 43, an automated annotation package for gene prediction 44, to identify possible genes. Kaikogaas integrates a variety of programs for gene prediction and structural analysis of genomic sequence including software for coding region and splice-site prediction, sequence homology analysis, protein localization site prediction, and protein classification and secondary structure prediction. All putative open reading frames were also compared directly to our database of Heliconius Expressed Sequence Tags (ESTs) located in ButterflyBase 45 using BLAST 46. Any EST that showed highly significant similarity (e-value of ≥ 10^-40) to a putative ORF was then directly compared to the full genomic sequence. In this way, we could better assess possible homology versus similarity due to repetitive DNA elements (see below). To identify the B. mori sequences homologous to the H. erato BAC sequences, we performed a tBLASTn against the whole-genome shotgun reads of B. mori using as a query the translated protein sequence predicted by Kaikogaas.

A linkage strategy similar to ours was previously used to identify BACs near the Yb gene complex in H. melpomene 15. Sequencing and finishing of three contiguous H. melpomene BACs across this region was carried out by the Wellcome Trust Sanger Institute (accession numbers: CR974474, CT955980, CU367882). We used VISTA 21 to identify similarities between H. erato, H. melpomene, and B. mori genomic sequences. Pairwise genomic alignments were performed on the mVISTA server using the Avid alignment algorithm and the results were displayed together with the position of the annotations (ORFs, genes, mobile DNA, microsatellites). Each annotation or new genomic region identified from the comparative analysis that showed a strong similarity (≥ 80% conserved) was verified by aligning the sequences from both species using the LAGAN program 47 to get an accurate DNA sequence assembly.

We also searched the H. erato and H. melpomene BAC sequences for novel repetitive sequence as well as previously described transposable elements. To identify novel repetitive sequences in the BACs, we used the RepeatFinder software 4248. RepeatFinder is explicitly designed to find complex repeated motifs in large contiguous blocks genomic DNA sequence. Its algorithm uses BLAST to iteratively query segments of the input sequence against the intact input sequence. It then combines subsequences with high BLAST similarity into groups, which are further refined by considering the variability in length and divergence among constituent subsequences. The final output is a list of groups of aligned, highly similar subsequences. It is important to note that because the algorithm uses local alignments (generated by BLAST), different groups may overlap in part or in whole. This overlap allows the groups to be clustered into just a few contigs representing a set of "core motifs" to which each group can be uniquely assigned. Thus, each core motif is a consensus of consenses.

We submitted the concatenated BAC sequences from each species to RepeatFinder using default parameters except for the following: Block Size = 2000, Minimum Repeat Size = 20, Maximum Repeat Size = 700. Larger values for the Maximum Repeat Size parameter were tried, but no groups larger than ~600 bp were ever identified. Core motifs were assembled from group consensus sequences >35 bp in length using the default parameters in CodonCode Aligner software (CodonCode Corp., Dedham, MA). For both species' sets of core motifs, all-vs-all BLASTn searches were performed 1) between species to identify motifs shared between species and 2) within species to verify that motifs within species were unrelated. BLASTn was also used to search two additional databases for sequences similar to the core motifs. First we searched a combined database of the completed genomic or whole genome shotgun sequences from all 20 insect genome projects currently available at NCBI. We also searched all arthropod transposable elements available in the Repbase library of transposable elements 19.

We simultaneously identified and masked BAC regions corresponding to both the novel core motifs and known repetitive sequences using RepeatMasker 27. We combined core motif sequences with sequences of all arthropod transposable elements from RepBase into a single database. We queried this database with the concatenated BAC sequences using RepeatMasker and the CrossMatch search algorithm on the 'slow' setting to maximize sensitivity. The resulting data was summarized using custom scripts implemented in the R statistics package 49.