Construction of a plant-transformation-competent BIBAC library and genome sequence analysis of polyploid Upland cotton (Gossypium hirsutum L.)

Background Cotton, one of the world’s leading crops, is important to the world’s textile and energy industries, and is a model species for studies of plant polyploidization, cellulose biosynthesis and cell wall biogenesis. Here, we report the construction of a plant-transformation-competent binary bacterial artificial chromosome (BIBAC) library and comparative genome sequence analysis of polyploid Upland cotton (Gossypium hirsutum L.) with one of its diploid putative progenitor species, G. raimondii Ulbr. Results We constructed the cotton BIBAC library in a vector competent for high-molecular-weight DNA transformation in different plant species through either Agrobacterium or particle bombardment. The library contains 76,800 clones with an average insert size of 135 kb, providing an approximate 99% probability of obtaining at least one positive clone from the library using a single-copy probe. The quality and utility of the library were verified by identifying BIBACs containing genes important for fiber development, fiber cellulose biosynthesis, seed fatty acid metabolism, cotton-nematode interaction, and bacterial blight resistance. In order to gain an insight into the Upland cotton genome and its relationship with G. raimondii, we sequenced nearly 10,000 BIBAC ends (BESs) randomly selected from the library, generating approximately one BES for every 250 kb along the Upland cotton genome. The retroelement Gypsy/DIRS1 family predominates in the Upland cotton genome, accounting for over 77% of all transposable elements. From the BESs, we identified 1,269 simple sequence repeats (SSRs), of which 1,006 were new, thus providing additional markers for cotton genome research. Surprisingly, comparative sequence analysis showed that Upland cotton is much more diverged from G. raimondii at the genomic sequence level than expected. There seems to be no significant difference between the relationships of the Upland cotton D- and A-subgenomes with the G. raimondii genome, even though G. raimondii contains a D genome (D5). Conclusions The library represents the first BIBAC library in cotton and related species, thus providing tools useful for integrative physical mapping, large-scale genome sequencing and large-scale functional analysis of the Upland cotton genome. Comparative sequence analysis provides insights into the Upland cotton genome, and a possible mechanism underlying the divergence and evolution of polyploid Upland cotton from its diploid putative progenitor species, G. raimondii.


Background
Cottons (Gossypium L. species) are the leading fiber and an important oilseed crop in the world. Cotton fibers sustain the world's textile industry and are an alternative of the synthetic fibers whose production annually consumes billions of barrels of fossil oil worldwide. Cottonseeds are traditionally used to produce food oil and currently have been used as the feedstock of biodiesel production. Furthermore, cottons are also a model system for studies of plant polyploidization, cellulose biosynthesis and cell wall biogenesis. The cultivated cottons, Upland cotton (G. hirsutum L.) and Sea Island cotton (G. barbadanse L.), are allotetraploids containing two homoeologous genomes defined A-and D-subgenomes. It was estimated that they originated from their diploid progenitor species about 1-2 million years ago [1][2][3]. Therefore, the cotton polyploiddiploid complex has been widely used as a model system for study of plant polyploidization and its impacts on speciation, biology and evolution [4,5]. Cotton fibers, usually 20-40 mm long and 15 μm thick, are derived from individual epidermal cells of developing seeds and more than 90% of their content is cellulose. Cellulose is a major component of plant cell walls and constitutes the largest portion of plant biomass, with an estimated annual world production of billions of metric tons. Therefore, cotton research is of significance not only for the world's textile and energy industries, but also for understanding the mechanism underlying cellulose biosynthesis and cell wall biogenesis [6][7][8][9] that are applicable to the entire plant kingdom.
Cottons have been subjected to extensive research in modern genomics and genetics [10]. A number of molecular genetic maps and several thousands of DNA markers have been developed, hundreds of genes and QTLs (quantitative trait loci) of agronomic importance have been mapped, and a large collection of expressed sequence tags (ESTs) have been generated from a variety of tissues collected at different growth and development stages [10]. Bacterial artificial chromosome (BAC) libraries have been demonstrated to be crucial for different aspects of advanced genomics and genetics research, and have been constructed for some genotypes of Upland cotton [11][12][13][14] and Sea Island cotton [15]. A draft physical map [16] and draft genome sequences (http://www.ncbi.nlm.nih.gov/ sra/SRA024364?report = full) have been generated recently for G. raimondii. Gossypium raimondii contains a D genome (D 5 ) and was proposed to be the closest diploid progenitor of the D-subgenome of allotetraploid cottons, including Upland cotton [4]. Nevertheless, no large-insert plant-transformation-competent binary BAC (BIBAC) library has been reported yet for any of the cotton species. BIBAC libraries have all of the functionality of BAC libraries, but also can be used as a vehicle of large DNA fragments containing a number of genes and QTLs for direct transformation in plants via either Agrobacterium [17][18][19][20][21] or biolistic bombardment [22] because they have their own selection markers for plant transformation. The direct transformation competence of BIBACs streamlines their uses for gene and QTL cloning [23], large-scale genome functional analysis [19,22,24] and molecular breeding. For instance, using BIBACs spanning the interval of a QTL (e.g., 5 -10 cM), the DNA fragment containing the QTL could be identified by BIBAC transformation, without the need of high-resolution mapping used in the traditional but laborious process of map-based cloning. Using BIBACs with an average insert size of about 150 kb as a vehicle, the Upland cotton genome could be directly transformed into another genome by approximately 16,000 transformations, which is within the realm of current transformation technologies [17,19,22]. These utilities of BIBACs will dramatically facilitate many aspects of current genomics and genetics research such as gene and QTL cloning, targeted marker development and molecular breeding. Moreover, transformation of BIBACs with an insert size of about 150 kb potentially containing 5-30 genes allows transferring in a single event a cluster of genes that are likely involved in a particular biological process, thus improving the efficiency of molecular breeding [22,24]. Finally and importantly, transformation of large-insert BIBACs is particularly desirable for engineering genes from a species to an unrelated species, such as from cotton to Arabidopsis, because the native regulatory sequences and neighboring genes of the donor species are important to the transgene expression in the recipient species [22,24]. Therefore, a BIBAC library constructed from a cultivated cotton is of significance for cotton gene and QTL cloning [23], genome functional analysis and molecular breeding [22,24].
The genome origin and polyploidization of polyploid cottons, including two cultivated cottons (G. hirsutum and G. barbadense) and three wild species (G. tomentosum, G. mustelinum and G. darwinii), have been studied extensively [1][2][3][4][5]. A few extant diploid species containing an A or a D genome, including G. raimondii, have been proposed to be the most closest reluted of the polyploid cotton A-and D-subgenomes [4]. Nevertheless, debate widely exists about the genome origin and evolution of the polyploid cotton A-and D-subgenomes because such knowledge is of significance for enhanced cotton genetic improvement. The recent release of the G. raimondii draft genome sequence (http://www.ncbi.nlm.nih.gov/sra/SRA024364?report = full) provides useful tools for advanced studies of the origin and evolution of the polyploid cotton A-and Dsubgenomes.
In this study, we constructed a large-insert BIBAC library from Upland cotton (G. hirsutum) cv. "Texas Marker-1" (TM-1) having a genome size of 2,425 Mb/1C [25], sequenced the ends of a large number of the BIBACs and preliminarily compared the genome of Upland cotton with the D genome of G. raimondii. Upland cotton represents approximately 95% of cotton growing in the world, and the genotype TM-1 has been widely used in cotton genomics and genetics research [14]. Therefore, the BIBAC library constructed from the genotype provides tools and resources for many aspects of cotton genomics, genetics and molecular biology research. From the BIBAC library, we isolated the BIBAC clones containing a number of genes of agronomic importance, such as those involved in fiber development (MYBB, MYBT2 and RDL1), fiber cellulose biosynthesis (CelA1, CelA3, CelA6, GhCesA2, GhIRX3, GhCesA3, and an unnamed Ces), seed fatty acid metabolism (FADO6) and host-nematode interaction (MIC3 and MIC1-15). Moreover, we also isolated the BIBAC clones potentially containing the genes conferring resistance to cotton bacterial blight pathogen, Xanthomonas campestris pv. malvacearum. Finally, we sequenced approximately 10,000 BIBAC ends (BESs) randomly selected from the library, with a sequence sample of 400 bp in an average of every 250 kb along the Upland cotton genome, to provide some insights into the Upland cotton genome and its relationships with G. raimondii. Using the BESs, we estimated the composition of the Upland cotton cv. TM-1 genome, identified over 1,269 SSRs, of which 1,006 were new, and compared the genome sequence of Upland cotton with that of G. raimondii. These results significantly promote genomics and genetics research of Upland cotton, including integrative physical and genetic mapping, genome sequencing and functional analysis of the Upland cotton genome, and provide deeper insights into the mechanisms underlying the genome origin, variation and evolution of Upland cotton.

Construction of the BIBAC library
We successfully constructed a high-quality BIBAC library for Upland cotton using the TM-1 megabase-sized nuclear DNA partially digested with BamH I (Invitrogen, Carlsbad, CA) in pCLD04541 [26,27], a widely-used, Agrobacteriummediated plant-transformation-competent BIBAC vector [22,[28][29][30][31][32][33][34][35]. Cottons are extremely abundant in polyphenolics [36,37] that may interact with DNA, thus interfering with DNA digestion and cloning [36,38,39]. Therefore, we prepared the DNA plugs from the cotyledons collected from the seedlings growing in a controllable growth chamber and treated in dark for 48 h (see the Methods). These measures minimized the contents of metabolites in the cells, especially polyphenolics and polysaccharides that may significantly influence the cloning of large DNA fragments. Furthermore, we washed the nuclei two more times than the original procedure for megabase-sized DNA preparation [37,39,40]. This measure further minimized the chloroplasts, mitochondria, polyphenolics and polysaccharides contained in the cytoplasm. Therefore, the resulting agarose plugs of the cotton nuclear DNA were transparent and pale in color, which is generally considered to be well suited for large-insert DNA library construction [38]. The third measure that we took to enhance the quality of the BIBAC library was the preparation of the DNA plugs at a proper concentration of DNA per plug. This is because the concentration of DNA in the plugs would significantly affect the methods and results of size selection on agarose gels, thus the insert sizes of the BIBAC clones resulted from the DNA [38]. Therefore, the use of the tissues containing reduced amount of polyphenolics and polysaccharides plus additional washes of nuclei before embedding into LMP agarose plugs is an effective and economical method to enhance the quality of megabase-sized DNA and ultimately the quality of the resultant BIBAC library. Preparation of the source DNA with an appropriate concentration in the agarose plugs further improved the average insert size of the resultant BIBAC library.

Characterization of the BIBAC library
The clones of the Upland cotton cv. TM-1 BIBAC library were individually arrayed into 200 384-well microtiter plates, containing a total of 76,800 clones. To determine the insert sizes of the clones and the portion of insertempty clones in the library, we selected a random sample of 122 clones from the library, isolated DNA from the clones, digested with Not I (Invitrogen) and analyzed on pulsed-field gels ( Figure 1A). Four of the 122 clones (3.3%) had no insert and the remaining 118 (96.7%) all had inserts. The distribution of the clones according to their insert sizes is shown in Figure 1B. The clones had insert sizes ranging from 60 to 250 kb, with an average insert size of 135 kb. Of the clones, nearly 80% had insert sizes larger than 100 kb. Furthermore, we estimated the percentage of the clones in the library derived from chloroplast DNA. A total of 691 positive clones were identified, accounting for approximately 0.9% of the library clones. Therefore, when the insert-empty clones and the chloroplast DNA-derived clones were excluded, approximately 95.8% (73,575) of the library clones contained inserts derived from cotton nuclear DNA. Since Upland cotton cv. TM-1 was estimated to have a genome size of 2,425 Mb/1C [25], the BIBAC library had a 4.1× coverage of the TM-1 haploid genome, with a 98.4% probability of obtaining at least one positive clone from the library using a single-copy probe [41,42].

Identification of BIBACs containing genes important in cotton
To further characterize the library and test its utility for cotton genome research, we screened the library with 13 genes of agronomic importance (Additional file 1: Table S1), and four RGAs (resistance gene analogues) closely linked to the cotton bacterial blight resistance ( Table 1). The library screening with the 13 genes resulted in a total of 60 positive clones, with each gene probe having 2-8 positive clones and an average number of 4.6 positive clones ( Table 2), suggesting that the BIBAC library is suitable for map-based cloning and characterization of genes and QTLs important in cotton. The library screening with the four RGAs yielded 1, 72 and 104 positive clones for 1C08, 2D14 and 2D03/2B19, respectively (Table 1). Since the genome coverage of the library was 4.1 ×, the sequences of 2D14 and 2D03/2B19 are possibly dispersed in as many as 17 and 25 loci, respectively, in the Upland cotton genome.

BES analysis
A total of 10,752 BIBAC ends randomly selected from the Upland cotton BIBAC library were sequenced. After base calling and trimming the E. coli and vector sequences, 9,711 (90.3% successful rate) Q20 BESs with a minimum length of 50 bp were generated (dbGSS JY253441-JY262664). The total length of BESs was 3,842,009 bp, with each BES having a sequence length ranging from 50 to 842 bp, with an average length of 395 bp. Of these 9,711 BESs, 4,441 BIBAC clones (84.3%) were successfully sequenced at both ends, generating mate-paired reads. Sequence analysis of the BESs indicated that the Upland cotton cv. TM-1 genome, as many other dicotyledonous plant genomes, is A/T-rich with 64.50% A/T and 35.50% G/C (Table 3). From the 9,711 BESs, we identified a total of 2,912 or 3,022 exons, depending on the computer programs used (Table 3), and a total of 374,019 base pairs of repeated sequences. Of the transposable elements identified, the retroelement Gypsy/DIRS1 family alone accounted for 77.86% (4.22/ 5.42) ( Table 4).

Identification of microsatellites from the BESs
To identify new SSR markers from the BESs, we analyzed the BESs for SSRs and compared the SSRs identified in this study with the existing cotton SSR database (www.cottonmarker.org). We identified a total of 1,269, 313 and 103 SSRs having motifs di-, tri-, tetra-, pentaand hexa-mers under the settings of ≥ 4, ≥ 5 and ≥ 6 repeat units, respectively (Table 3). Analysis of the 1,269 SSRs showed that dimer SSRs absolutely predominate in the Upland cotton genome ( Figure 2). Among the dimer SSR motifs, (AT) n (where n ≥ 4) was the most abundant in the genome, accounting for nearly 50% of all SSRs in the genome, and (CT) n , (AG) n , (AC) n and (GT) n were after (AT) n in descending order, with each accounting for from 12.5% down to 5.5%. The other types of SSR motifs together only accounted for approximately 15% of the SSRs in the genome. To determine how many of them were new, we downloaded the cotton SSR database that contains 17,343 SSRs from the Cotton Marker Database (www.cottonmarker.org). The 1,269 SSRs identified in this study were then searched against the cotton SSR database. As a result, the locus sequences of 263 of the 1,269 SSRs matched the cotton SSR database; therefore, 1,006 or 79.28% of the 1,269 SSRs were novel.
Comparative sequence analysis of the polyploid Upland cotton genome with the diploid G. raimondii genome The origin of the D-subgenome of tetraploid cottons, including Upland cotton and Sea Island cotton, remains obscure, even though G. raimondii was previously proposed to be the closest diploid progenitor [4,45]. Because the whole genome sequence of G. raimondii has recently become available to the public, we compared the Upland cotton genome with the G. raimondii genome using the 9,711 BESs generated in this study as inquiries (Table 5). Surprisingly, only 57 of the 9,711 (0.59%) Upland cotton BESs were aligned to the raimondii genome using the criteria of a continuous minimal match of 100 bp and a sequence identity of 100%. This number corresponded to only 0.41% of the total sequence length of the BESs attempted in the alignment. When the sequence identity was reduced to 90% or lower while the criterion of continuous minimal match of 100 bp was maintained, the percentage of the Upland cotton BESs aligned to the raimondii genome reached and stabilized at approximately 70%, corresponding to approximately 51% of the total inquired Upland cotton BES sequence length.
Moreover, we extracted the Upland cotton A-and Dsubgenome specific BESs from the 9,711 BESs using the A-or D-subgenome-derived BIBACs identified by Zhang et al. [46]. We were able to identify a total of 418 BESs derived from the A-subgenome BIBACs and 184 BESs derived from the D-subgenome BIBACs of Upland cotton (Table 5). When the A-and D-subgenome specific Table 1 Positive clones of the Upland cotton TM-1 BIBAC library identified by library screening with RGAs closely linked with the resistance to bacterial blight pathogen, Xanthomonas campestris pv. malvacearum [43,44] 2D03 (AY600401) and 2B19 (AY600394) (104 positive clones): 2D14 (AY600383) (72 positive clones): 1C08 (AY600376) (1 positive clone):

B125E12
BESs were separately aligned to the G. raimondii genome sequence using the same criterion as above, the percentages of aligned BESs were not different from those obtained using the 9,711 BESs as inquiries. This result suggested that the genomic regions of Upland cotton from which the A-and D-subgenome specific BIBACs were derived have similar sequence identities to the other genomic regions of the A-and D-subgenomes with the G. raimondii genome. Since G. raimondii only has a D genome (D 5 ), it was expected that the percentage of the D-subgenome specific BESs aligned to the G. rainmodii genome would be much higher than that of the A-subgenome specific BESs. Surprisingly, no significant difference was obtained between the percentages of the A-and D-subgenome specific BESs aligned to the G. raimondii genome.

Discussion
We have developed a large-insert BIBAC library for Upland cotton in a BIBAC vector (pCLD04541) that is competent for direct plant transformation via both Agrobacterium [29,30] and biolistic bombardment [22]. The library contains 76,800 clones, with an average insert size of 135 kb. Therefore, the Upland cotton BIBAC library has a reasonably large average insert size. The 76,800 clones of the BIBAC library, when the insert-empty clones and the clones derived from chloroplast DNA are excluded, provide a 4.1× coverage of the Upland cotton haploid genome, with an approximate 99% probability of obtaining at least one positive clone from the library using a single-copy sequence probe. The results from the library screening with 13 gene-specific overgo probes designed from the unique sequences of genes, in which we obtained 2-8 positive clones for every probe, support the estimated genome coverage of the library, even though some of the probes might hybridize to two or more copies of the sequences because Upland cotton is an allotetraploid containing two homoeologous genomes, A-and Dsubgenomes. Importantly, differing from the existing cotton BAC libraries [11][12][13][14][15], this BIBAC library is the first one competent for direct plant transformation via both Agrobacterium [29,30] and biolistic bombardment [22]. The plant transformability of the BIBAC library streamlines the map-based cloning of cotton genes and QTLs (Introduction) and large-scale functional analysis of the cotton genome through high-throughput BIBAC transformation [22,24]. Therefore, the BIBAC library reported here will promote many aspects of cotton genome  We have used the BIBAC library to isolate BIBACs containing or closely linked to the genes controlling fiber development, fiber cellulose biosynthesis, seed fatty acid metabolism, cotton-nematode interaction, and cotton bacterial blight resistance. This experiment has not only further verified the quality and demonstrated the utility of the BIBAC library for cotton genomics research, but also provided the tools essential for characterization of the genes at the genomic level and promoted the use of the genes in molecular breeding through BIBAC transformation. Using the BIBACs, we have developed a highthroughput system for functional analysis of the entire Upland cotton genome in the plant model species, Arabidopsis thaliana, with which the entire Upland Furthermore, the BESs generated in this study provide 9,711 STSs (sequence-tagged sites), with one STS in approximately every 250 kb along the Upland cotton genome. These STSs will facilitate integrative physical and genetic mapping, and sequencing of the Upland cotton genome using the next-generation sequencing technology [47,48]. Using the BIBAC library, we have constructed separate genome-wide physical maps for both the A-and D-subgenomes of Upland cotton cv. TM-1 and identified the minimal tiling path (MTP) clones (15,277 clones) spanning the physical map [46]. We have also initiated a project of sequencing the Upland cotton cv. TM-1 genome based on the BIBAC library using the nextgeneration sequencing technology (D.M. S., H.-B. Z.). We Motif type Figure 2 Frequency of the ten most abundant SSR motifs in the Upland cotton genome. The SSR motifs are di-, tri-, tetra-, penta-and hexa-mers, with four or more repeats.
are developing the integrated physical and genetic maps of individual chromosomes of the Upland cotton genome, with 1-10 contigs per chromosome, from this BIBAC library and a BAC library, which will be necessary for assembling the chromosome-sized pseudomolecules of the Upland cotton genome sequence. Analysis of the BESs has shown that the 9,711 BESs, with a length range from 50 -842 bp and an average length of approximately 400 bp, contain about 3,000 exons. If the exons of cotton genes have an average length of 200 -500 bp [49,50], the 3,000 exons are likely from approximately 3,000 genes, suggesting that approximately 30% of the 9,711 BESs are parts of genes. Among the transposable elements, the LTR retroelement family, Gypsy/DIRS1, is the most predominant in the Upland cotton genome, representing over 77% of the transposable elements in the genome. This result is consistent with that of Hawkins et al. [51] estimated for the genomes of its diploid putative progenitor species, G. raimondii and G. herbaceum, by random shotgun clone sequencing. The more than 1,000 new SSRs (79% of the SSRs) identified from the BESs indicate the abundance of SSR loci in the Upland cotton genome and provide additional tools for cotton genome analysis, and gene and QTL mapping.
The comparative BES analysis has shown that the Upland cotton genome has significantly diverged from the G. raimondii genome. The unexpectedly high genome divergence between the two species could be attributed to the rapid evolution of the G. raimondii genome, the polyploidization and post-polyploidization evolution of Upland cotton, or both. However, the hypothesis of the rapid evolution of the G. raimondii genome does not appear to be compatible with the finding of this study that the A-subgenome of Upland cotton has the same level of similarity as its D-subgenome when compared to the G. raimondii genome, even though G. raimondii has only a D genome. Therefore, the rapid genome evolution after polyploidization must have occurred in Upland cotton, or both Upland cotton and G. raimondii. Moreover, it appears that an extensive element exchange has occurred between the A-and D-subgenomes of Upland cotton during the process of and/or after polyploidization, thus leading to a higher similarity between the two subgenomes at the genomic element sequence level. Therefore, when the A-and D-subgenome specific BESs were aligned to a third genome -the G. raimondii genome in this study, similar alignment results were obtained. Zhang et al. [5] studied the gene number variation of nucleotide-binding site (NBS)-encoding gene family and receptor-like kinase (RLK)-encoding gene family between the tetraploid cottons including Upland cotton and their diploid putative progenitor species including G. raimondii and G. herbaceum. They found that the tetraploid cottons have similar numbers of NBS and RLK genes as their diploid putative progenitor species, suggesting that a large number of genes in the families were lost during the process of and/or after the cotton polyploidizaiton. It has been found recently that most genome-constituent fundamental function elements of rice, including genes, DNA transposable elements, simple sequence repeats and low complexity repeats, have a very low content variation among different chromosomes even though they are nonhomologous, suggesting the existence of genomic element exchanges among chromosomes within a genome. These results provide indirect, but strong, support for the above hypothesis. (Liu Y-H, Zhang MP, Wu C, Huang JJ, Zhang H-B: DNA is structured as a linear "Jigsaw Puzzle" in the genomes of Arabidopsis, rice and budding yeast, submitted for publication). However, further studies remain to determine the molecular mechanisms underlying the rapid genome evolution and genomic element exchanges between subgenomes in the polyploid cotton cells in the process of post-polyploidization.

Conclusions
We have constructed a high-quality large-insert planttransformation-competent BIBAC library for Upland cotton, isolated BIBACs containing several important cotton genes, sequenced nearly 10,000 BESs from BIBACs randomly selected from the cotton BIBAC library and identified over 1,000 new SSR markers from the BESs. These results provide resources and tools useful for advanced research of cotton genomics, genetics and breeding in numerous aspects such as Upland cotton genome integrative physical and genetic mapping, genome sequencing, gene and QTL cloning and characterization, and molecular breeding. Moreover, this study provides some insights into the genome composition and organization of the tetraploid cotton. The retroelement Gypsy/DIRS1 family and the SSRs with (AT) motifs are the most abundant in the Upland cotton genome. Finally, comparative sequence analysis has shown that the genome of the tetraploid Upland cotton has diverged much more than expected from that of its D-subgenome diploid putative progenitor species, G. raimondii. The divergence has likely resulted from the rapid evolution of tetraploid cotton and an apparent extensive genome-constituent element exchange or "balancing" between the A-and D-subgenomes of the tetraploid cotton during the process of postpolyploidization, even though further studies remain to test this hypothesis.

Plant materials
Upland cotton cv. TM-1 was used as the source DNA of the BIBAC library. The seeds of the cultivar were obtained from USDA/ARS, College Station, Texas, germinated and grown in a growth chamber at 25°C, 16 h light/8 h dark. Plant seedlings were treated in continuous dark for 48 h immediately before DNA isolation to minimize polyphenolics and polysaccharides. Cotyledons were harvested, frozen in liquid nitrogen and stored at −80°C before use.

Preparation of megabase-sized nuclear DNA
Megabase-sized DNA was prepared according to Zhang et al. [37] with minor modifications [39,40,42,52]. Nuclei were washed for two additional times to minimize the contamination of polyphenolics, polysaccharides, chloroplasts and mitochondria contained in cytoplasm. The nuclei suspension of approximately 2 × 10 7 nuclei /ml was prepared and used to make low-melting-point (LMP) agarose plugs. Therefore, each 100 μl-plug contained approximately 5 μg nuclear DNA.

Preparation of BIBAC vector
The BIBAC vector pCLD04541 [27] was used for the construction of the BIBAC library. This vector has been widely used in the construction of BIBAC libraries for a number of species [22,[28][29][30][31][32][33][34][35]. Vector DNA was isolated by the alkaline lysis method and purified by two rounds of the cesium chloride gradient centrifugation [38,53]. The vector DNA was completely digested with BamH I and prepared according to a procedure that we developed previously [38,42].

Library construction
The nuclear DNA of Upland cotton was partially digested with BamH I and cloned in the BIBAC vector pCLD04541 to construct the BIBAC library [38]. To determine the optimal condition for partial digestion of megabase-sized DNA in LMP agarose plugs, particularly the amount of restriction enzyme (BamH I) per reaction and digestion incubation time, we first conducted a partial digestion test using a series of BamH I concentrations per reaction. From the results of the partial digestion test, 2.4 units of BamH I per reaction containing one-third of a 100-μl megabase-sized DNA plug and 8-min incubation at 37°C were selected for large-scale partial digestion for the library construction. Ten 100-μl megabase-sized DNA plugs were used for partial digestion of the library construction using the condition predetermined by the partial digestion test (2.4 units of BamH I per reaction and 8-min incubation at 37°C). The partially digested megabase-sized DNA was subjected to one size selection on a 1% agarose gel by pulsed-field gel electrophoresis and DNA fragments ranging from 100-250 kb were selected [38]. The selected DNA fragments contained in the agarose gel were electroeluted, dialyzed against 0.5 × TE (5 mM Tris-HCl, 0.5 mM EDTA, pH 8.0) and ligated into the dephosphorylated pCLD04541 vector. The ligation was performed at a molar ratio of 3 vectors: 1 insert, 1.5 ng of insert DNA /μl of ligation reaction and 1.0 unit T4 DNA ligase (Invitrogen) /50 μl of ligation reaction at 16°C for 8 h.
The ligated DNA was transformed into Escherichia coli strain DH10B (Invitrogen) by electroporation using the Cell Porator™ Device (Gibco BRL, Carlsbad, CA) consisting of Power Supply (cat. 1600), Chamber Safe (cat. 1600) and Voltage Booster (cat. 1612). The electroporation settings were as Zhang et al. [38], and 375 volts were used for electroporation. The transformed cells were selected for transformants and recombinant BIBACs on selective LB agar plates containing 15 μg of tetracycline, 15 μg of IPTG (isopropylthiogalactoside) and 60 μg of X-Gal (5-bromo-4chloro-3-indolyl-β-D-galactoside) per ml of medium. Ligation selection and large-scale transformation for library construction were described as Zhang et al. [38]. The recombinant white clones were arrayed as individual BIBACs in 384-well microtiter plates, duplicated and kept at −80°C.

Library characterization
A random sample of the clones arrayed in the 384-well microtiter plates was analyzed [38] to estimate the insert sizes of the library clones and the percentage of clones having no inserts (insert-empty clones). BIBAC DNA was isolated, digested with Not I to release the insert of cotton DNA from the cloning vector pCLD04541 and subjected to pulsed-field gel electrophoresis. The insert size of each clone was estimated by the sum of the sizes of all insert band(s) in each gel lane using the lambda ladder PGE marker (New England BioLabs, Ipswich, MA) as the molecular weight marker.
The BIBAC library was printed in a 4 × 4 format using the GeneTAC Robotic Workstation (Genomic Solutions Inc., Ann Arbor, MI) onto 22.5 × 22.5 cm Hybond N + membrane (Amersham-Pharmacia, Piscataway, NJ) laid on LB agar medium containing 15 μg tetracycline/ml medium. To facilitate reading the positive clones accurately and minimize the false positive clones, each of the library clones was printed in duplicate on the membrane. Therefore, each 22.5 × 22.5 membrane contained a total of 18,432 (48 × 384) double-printed clones and the entire library was printed on four 22.5 × 22.5 cm and one 12 × 8 cm membranes. The clones on the membranes were grown at 37°C overnight, the membranes were processed and the DNA of each clone was crosslinked on the membranes [41,54]. The membranes were hybridized with a probe prepared from a mixture of three chloroplast genes [41], ndhA, rbcL and psbA, to identify the clones of the library that were derived from chloroplast DNA. To ensure unambiguously reading the positive clones identified on the library membranes, we added 1-2 ng [1 -2% (w/w) of the probe DNA] of the pCLD04541 vector DNA to the labeling reaction to create appropriate background when hybridized. The hybridization was done at 65°C overnight. After hybridization, the library membranes were washed in 1 × SSC, 0.1% (w/v) SDS [54] at 65°C twice, 10 min each time, followed by 0.5 × SSC, 0.1% SDS at 65°C twice, 10 min each time.

Identification of BIBACs containing genes important to cotton
The membranes of the BIBAC library were screened to further verify its quality, demonstrate its utility and to identify the BIBACs containing cotton genes of interest. We first hybridized the library membranes using genespecific overgo probes to identify the BIBACs containing the genes involved in fiber development (MYBB, MYBT2 and RDL1), fiber cellulose biosynthesis (CelA1, CelA3, CelA6, GhCesA2, GhIRX3, GhCesA3 and an unnamed Ces), seed fatty acid metabolism (FADO6) and cottonnematode interaction (MIC3 and MIC1-15; Additional file 1: Table S1). The gene-specific overgos were designed from the unique sequences of the target genes, synthesized, pooled at equal amounts (100 ng per overgo pair) and used as probes to hybridize the library membranes [34]. The positive clones resulted from the primary hybridization were then re-arrayed into a new microtiter plate, double-printed onto a 12 × 8 cm Hybond N + membrane in a 2 × 2 format and sorted according to individual gene overgos using a 4 × 4 pooling hybridization strategy [55,56]. To ensure unambiguously reading the positive clones identified on the library membranes, we added 1-2 ng [1 -2% (w/w) of the gene overgos] of pCLD04541specific overgos (forward oligo 5'-TTAAGTTGGGTAAC GCCAGGGTTT-3 and reverse oligo 5'-CAACGTCGT GACTGGGAAAACCCT-3') to the gene-specific overgos. The gene and vector overgo mixture was labeled with [ 32 P]-dATP and [ 32 P]-dCTP to increase the hybridization signals, and the hybridization was carried out at 60°C for overnight. After hybridization, the library membranes were washed in 1 × SSC, 0.1% (w/v) SDS at 60°C twice, 10 min each time, followed by 0.5 × SSC, 0.1% SDS at 60°C twice, 10 min each time.
Moreover, we also hybridized the library membranes to identify the BIBACs containing the disease resistance analogues (RGAs) from three loci, 1C08 (GenBank No.: AY600376), 2D14 (AY600383), and 2D03 (AY600401)/ 2B19 (AY600394) [43]. The 1C08 RGA shares a significant similarity with the tomato root-knot nematode resistance gene, Mi-1.2 [43,57] and is closely linked to the B2 gene of the cotton bacterial blight (X. campestris pv. malvacearum) resistance [43,44]. The 2D14 and 2D03/ 2B19 RGAs flank the Qb 6a QTL of the bacterial blight resistance [43,44]. The inserts of the RGA clones were excised by enzymatic digestion and gel purification, labeled with [ 32 P]-dCTP using the random priming method and used as probes to hybridize the library membranes as the above library hybridization with the chloroplast gene probes.
Software Phred [60,61] was used for sequence base calling, with a quality score of Q20. The vector sequences were removed using Cross_match (http://www.phrap.org/ phredphrapconsed.html#block_phrap). Sequence trimming was conducted by Sequencher v. 3.7 (Gene Codes Corp., Ann Arbor, MI). The repeat elements in the BESs were identified using RepeatMasker (A. Smit, G. Glusman and R. Hubley at http://repeatmasker.org) against the latest repeat library (as of 04/19/2011) of Arabidopsis thaliana that is the phylogenetically closest to cotton in the database, with the default parameters. We also predicted the genes/ exons contained in the BESs using Genscan [62] and GeneMark [63] against Arabidopsis model/smat using the default settings. The BESs were searched against the recently released G. raimondii genome sequence assembly (DOE Joint Genome Institute: Cotton D V1.0, http://www. phytozome.net/cotton.php) by a local Blat server [64] with a continuously minimal match of 100 bp and a sequence identity of 100%, 95%, 90%, 80%, 70% and 60%, respectively. Microsatellites were identified in non-redundant BESs using Msatfinder v2.0 that was specifically designed to identify and characterize microsatellites [65]. Only the microsatellites of 2-6 nucleotide motifs with at least 4, 5 and 6 repeat units were collected.