S. hermonthica is likely to be a diploid species with a chromosome number of n = 19 10 . First, we estimated the genome size of S. hermonthica to gain information about its genome contents. Leaves of S. hermonthica plants parasitizing to rice were harvested and the DNA contents were measured with a flow cytometer. Arabidopsis thaliana, whose genome size is 128 Mbp, was used as a control. Five individual plants were used for the measurements with two or more replicates for each plant. The genome size of S. hermonthica was estimated to be 1,801 Mbp (± 321 Mbp) (Fig. 1), which is approximately 14 times that of Arabidopsis, 4 times those of rice and poplar, and 2 times that of sorghum.

Next, we sequenced both the 5'- and 3'-ends of 37,710 clones from the S. hermonthica full-length enriched cDNA library. The sequence chromatograms were analyzed using the EST2uni package 15 , which is an automated analysis tool for the clean-up, clustering, and annotation of EST sequences. Among the 75,330 raw sequence reads, we found that 67,814 were of good quality and were deposited in the DNA Databank of Japan [DDBJ: FS438984-FS506797]. The sequences are clustered into 17,137 non-redundant unigenes (10,319 contigs and 6,818 singletons) (Table 3). The average GC content among the unigene sequences is 44.5%. The lengths of the unigenes are distributed between 82 and 3,949 bp, and most of them (11,546 unigenes) have sequence lengths between 601 and 900 bp (Additional file 1), with an average of 810.3 bp. Most (84%) of the unigenes are comprised of fewer than 6 ESTs (Additional file 1), suggesting that the redundancy rate is relatively low in this normalized library.

For the functional annotation of the 17,137 unigene sequences, we carried out a blastx analysis against the UniRef90 database 16 17 . About 79% of the S. hermonthica unigenes were annotated as homologs of known proteins. For further functional annotations of the structural domains, the Pfam database 18 was searched using the HMMER program (ver. 2.3.2, 19 20 ), and 31% (5367) of the unigenes contained Pfam hits. Then the S. hermonthica unigenes were classified into Gene ontology (GO) groups based on their similarities with the corresponding Arabidopsis genes (Fig. 2). In the classification of genes according to their cellular components, we found that 16% of the unigenes encode putative membrane proteins and 10% encode putative plastid proteins. In the classification of molecular functions, 12% were assigned to catalytic activity. These percentages are similar to those in Arabidopsis 21 , indicating that there was no functional bias among the predicted proteins encoded in the S. hermonthica library.

The S. hermonthica unigenes were compared with genes in other plant genomes, including A. thaliana, poplar (Populus trichocarpa), grape (Vitis vinifera), soybean (Glycine max), rice (Oryza sativa), sorghum (Sorghum bicolor), a moss (Physcomitrella patens), and an algae (Chlamydomonas reindardtii) 22 23 24 25 26 . Seventy-seven to seventy-nine percent of the S. hermonthica unigenes showed similarities with genes from other dicotyledonous plants (Arabidopsis, grape, soybean, and poplar), as detected by blastx (e_value < e-10). Approximately 75% of the unigenes have homologs in monocotyledonous plants (rice and sorghum), and approximately 65% and 38% showed blastx hits in the P. patens and C. reindardtii databases, respectively. These lower percentages of blast hits are consistent with the greater evolutionary distances from those organisms.

We plotted the percentages of S. hermonthica unigenes against levels of amino acid sequence identity with homologs in the other plant genomes (Fig. 3). Larger percentages of S. hermonthica unigenes showed higher levels of identity with poplar and grape sequences than with sequences from the other plant species. The identity scores corresponding to half the population of S. hermonthica unigenes were 0.68 for grape and poplar, 0.65 for Arabidopsis, 0.62 for rice, and 0.56 for P. patens. These numbers roughly reflect the evolutionary distances between S. hermonthica and these species.

Large scale EST sequence datasets have previously been reported for Triphysaria versicolor 8 and Triphysaria pusilla 27 , which are hemiparasitic plants belonging to the Orobanchaceae. The assembled EST sequences are available at the plantGDB web site 28 . Althoughthe genus Triphysaria is closely related taxonomically to S. hermonthica, only 74% of the S. hermonthica unigenes showed similarity to Triphysaria sequences (including both T. pusilla and T. versicolor), when analyzed with the tblastx program (Table 4). This is significantly lower than percentages of similarity found with the other dicotyledonous plants, but this is likely due to the lack of saturation of the Triphysaria EST datasets.

The conservation of the genes between S. hermonthica and Arabidopsis, grape, poplar, or Triphysaria spp. is shown in a Venn diagram (Fig. 4). Among the 17,137 unigenes, 11,711 (68%) are conserved among all five groups. Only 19, 36, and 58 of the S. hermonthica unigenes are conserved specifically in Arabidopsis, grape, and poplar, respectively. Interestingly, we found that 662 (3.9%) of the S. hermonthica unigenes are conserved in Triphysaria spp. but not in Arabidopsis, grape, or poplar.

Of these 662 sequences, 73 show similarities to sequences in other databases such as rice, sorghum, soybean, Physcomitrella, UniRef90 or nr (the non-redundant peptide database from NCBI). We found no other homologs for the remaining 589 unigenes (Additional file 2). Since T. pusilla and T. versicolor are hemiparasitic plants, these 589 might include genes specific to parasitic plants. The ongoing project to sequence the genome of Mimulus spp. may help to narrow down the number of candidate genes that are involved in parasitism, because Mimulus spp. are non-parasitic members of the family Scrophulariaceae, which is taxonomically close to Orobanchaceae. The 2,389 unigenes (14%) that did not show significant hits with any known peptide sequences in the tested databases (including nr) are also listed in Additional file 2. These unigenes may include sequences that are specific to Striga.

S. hermonthica is an obligate outcrossing plant with high levels of morphological and genetic variation 29 . The EST2uni program detected 9,299 putative single nucleotide polymorphisms (SNPs) among the S. hermonthica unigenes. To exclude the misidentification of sequencing errors as SNPs, only polymorphisms confirmed by at least 2 independent sequences were counted, although there is still the possibility that those polymorphisms occurred during cDNA synthesis. The average frequency of SNPs in the unigene sequences is 0.67%, or approximately 1 SNP per 1.5 kbp. Although these SNPs will need to be confirmed, these data will be useful for developing EST-SNP markers for S. hermonthica 30 .

We found 1,445 di-, tri- or tetra-nucleotide microsatellites (or SSRs) among the S. hermonthica unigenes. The most frequent of these are the tri-nucleotide repeats (Additional file 3), which is in agreement with previous studies of other plant species 31 32 33 . The most frequent individual microsatellite repeat is AG (including TC, GA, and TC) (283, 19.6%) and the second most frequent is AC (including TG, CA, and GT) (218, 15.1%). The most frequent tri-nucleotide repeat is ATC (including TCA and CAT) (157, 11.0%) (Additional file 4).

The EST-SSR sequences are good candidates for genetic markers, which can be used for molecular diagnosis, for biotyping weeds, and for investigating the genetic diversity and population structures of S. hermonthica. To investigate whether the SSRs that we identified can be used as such markers, we designed primers using sequences flanking the putative SSRs and looked for polymorphisms by PCR. First, we pooled DNA samples extracted from the leaves of several plants in the same field and used the DNA pools as PCR templates. Of the 64 primer sets tested, 44 successfully amplified DNA bands. However, 26 primer sets (59%) produced smears or multiple bands that were not countable and only 18 primer pairs (41%) amplified clear separate bands (Additional file 5). The smeared bands may indicate heterozygosity and genetic diversity among the individual plants harvested from the same field. Therefore, we tested the individual plants for polymorphisms. Several markers that showed smear patterns from the pooled DNA templates actually amplified clear polymorphic bands from individual plants in the same population (Additional file 6). These data verify that S. hermonthica is a highly adaptable weed that has maintained a high degree of genetic variation and plasticity, to survive in various ecosystems 34 .

Although individual S. hermonthica plants possess highly diversified genomes, 18 of the primer sets we tested showed countable band patterns when using pooled DNA templates. Using those primer sets, we investigated the relationships between different S. hermonthica populations from 6 fields growing sorghum, maize, or pearl millet in various locations in Sudan or Kenya 35 . Of the 18 primer sets, 10 showed clear polymorphisms for different S. hermonthica populations (Table 5, Additional file 5). The analysis of PCR products was carried out using MultiNa^® (Shimadzu, Japan), a microchip electrophoresis system that permits the separation of small fragments and that can detect 5 bp differences. The average polymorphism information content (PIC) was 0.463, which confirms that the SSR markers used in this study were highly informative The lowest PIC value was 0.305 for SSR57, and the highest was 0.545 for SSR26 (Table 5). The analyzed loci included 3 di-, 3 tri-, and 4 tetra-nucleotide repeats. A total of 27 alleles were detected, with an average number of alleles per locus of 2.7. The genetic diversity among the six populations was revealed by the gene diversity values, which ranged from 0.375 to 0.625, with an average of 0.549. These results suggest a high level of diversity among the surveyed populations, as was expected for this obligate outcrossing plant 36 37 38 .

We also looked for correlations between host species and S. hermonthica biotypes, using the Unweighted Pair Group Method with Arithmetic mean (UPGMA) clustering analysis. The populations from El Obeid (host: sorghum), Dirweesh (host: sorghum), and Kenya (host: maize) clustered in one group, while the population from Elkaraiba (host: sorghum) was in a distant branch of the same group. Those from Tandalti (host: pearl millet) and Agadi (host: maize) formed another cluster (Fig. 5). Thus, we did not detect any correlations between genetic distance and host specificity in this study. This result is consistent with previous epidemiological reports 35 38 39 40 . In summary, our results suggest that the SSRs found in our study could be useful tools for further investigations of genetic diversity in S. hermonthica.

The results of the sequencing and analysis of the S. hermonthica ESTs are freely available online from our web-based database 9 . The web interface was based on the original EST2uni web site 15 . The database contains features for complex query searches and a blast search. A page for each unigene consists of its sequence, contig images, results of blast similarity searches, lists of detected SSRs and SNPs, and GO categorizations. In addition, the homologs of each unigene are linked to outside databases such as The Arabidopsis Information Resource (TAIR) 41 . This web-based database will be a powerful tool for the detailed analysis of S. hermonthica genes.

S. hermonthica seeds collected from a sorghum field in 1994 in Kenya were provided by Dr. A. G. Babiker (Univ. of Sudan, Khartoum, Sudan). Rice seeds (Oryza sativa L. subspecies japonica, cultivar Koshihikari) were originally obtained from the National Institute of Agricultural Sciences (NIAS, Tsukuba, Japan). S. hermonthica plants parasitizing rice were grown in rhizotrons as described previously 42 or in soil (1:1 mixture of vermiculite: clay). For the axenic culture of S. hermonthica, seeds were sterilized with 20% bleach solution (approx. 6% NaOCl) for 5 min and washed thoroughly with sterile water. The sterile seeds were preconditioned on MS medium with 1% sucrose and 0.5% phytagel (Sigma) at 26°C for 7 to 10 days in the dark and germination was stimulated by the exogenous application of 5 μl 1 μM Strigol per plate. Sterile S. hermonthica plants were grown on the same medium at 26°C with a 16-h photoperiod, and the medium was renewed every 3 weeks.

The nuclear DNA content was analyzed with a flow cytometer (Partec PA, Tokyo, Japan). Soil-grown S. hermonthica (host: rice) leaves were chopped with a razor blade into small pieces and analyzed according to the previously published method 43 . Leaves of Arabidopsis (ecotype Col -0) were used as the control.

The S. hermonthica tissues and developmental stages used for RNA extraction are listed in Table 1. S. hermonthica RNAs were extracted using a modified cetyl trimethylammonium bromide (CTAB) method. Briefly, plant tissues were ground under liquid nitrogen and suspended in 5 × volumes of CTAB solution (2% CTAB, 2% polybinylpyrrolidone (PVP), 25 mM ethylenediaminetetraacetic acid(EDTA), 2 M NaCl, 1% beta-mercaptoethanol, 100 mM Tris-HCl (pH 8.0)) and phenol:chloroform (5:1, pH 4.7, Sigma). The mixtures were shaken at 55°C for 5 min. After 10 min centrifugation, the aqueous phase was extracted with an equal volume of phenol:chloroform, and subsequently with chloroform. The RNAs were precipitated by adding 0.25 volumes of 10 M LiCl. The RNA pellet was washed with 70% ethanol and then dissolved in nuclease-free water. Samples were subsequently purified using the PureLink RNA mini kit (Invitrogen) according to the manufacture's instructions. To obtain mRNA for library construction, total RNAs from each tissue and developmental stage were mixed and purified using an mRNA purification kit (GE) according to the manufacture's instructions. The quality and quantity of the total RNA and the mRNA were assessed by measurements of OD₂₃₀, OD₂₆₀, and OD₂₈₀, followed by visual checking by electrophoresis.

The construction of the normalized, full-length enriched library was carried out in Evrogen (Russia). The cDNA normalization was conducted using a Duplex-specific nuclease (DSN)-based method, and full-length cDNAs were enriched using the SMART™ technology (Clontech). Each cDNA was inserted into the pAL17.3 vector. Sequencing of randomly picked clones was performed in the Genome Center at Washington University using the ABI3730 capillary sequencer.

The EST sequences were automatically trimmed, clustered and annotated using the EST2uni analysis pipeline 15 . Sequence assembly was performed using the CAP3 program with the default parameter settings 44 . Blast searches were performed with NCBI blast program against the databases shown in Table 4. The S. hermonthica online database was constructed based on the EST2uni web program with slight modifications.

Genomic DNA was extracted from about 10 g of S. hermonthica seeds using the modified CTAB method described previously 35 . Primers flanking the microsatellites were designed using the PRIMER 3 program 45 . The PCRs were performed in 10 μl volumes with one initial denaturation step of 1 min at 95°C, followed by 40 cycles of 15 sec at 94°C, 30 sec at 60°C and 30 sec at 72°C, anda final extension step of 5 min at 72°C. The PCR products were analyzed either by 4% agarose gel electrophoresis (Additional file 6) or using the MCE-202 MultiNa Microchip Electrophoresis System for DNA/RNA analysis (Shimadzu, Japan) using the DNA-500 kit (Table 5 and Fig. 5). The data were analyzedusing the PowerMarker program version 3.25 46 , and the genetic diversity was estimated based on allelic numbersand the gene diversity value:

where n is the number of populations sampled, p _lu is the frequency of uth allele at the lth locus, and f is the inbreeding coefficient (association between alleles) at the lth locus. The Polymorphism Information Content (PIC) was estimated as , where the p _lv is the frequency of the vth allele at the lth locus. The phylogenetic UPGMA tree was generated based on a matrix of the frequencies and distances using the LogSharedAllele algorithm with the PowerMarker v.3.25 program. Bootstrap analysis was performed using the software package WINBOOT 47 .