The GA02-15 clone from the SRVR germplasm 19, one of the parental plants of a mapping population 18, was used to construct the SSR-enriched libraries. An initial PCR reaction in a 25 μl reaction contained 250 ng of genomic DNA, 25 μM of NaCl, 1 unit of T4 ligase (Promega, Madison, WI, USA), 1× T4 DNA ligase buffer (Promega, Madison, WI, USA), 25 μg/ml BSA (New England Biolabs, Boston, MA, USA), 2.5 units of MseI, and 2 pmol of adapter. To make the adapters, 1 μM of each of two oligonucleotides (5' GACGATGAGTCCTGAG 3' and 5' TACTCAGGACTCAT 3') were combined and heated at 95°C for 5 min in a MJ Research thermal cycler. These were allowed to cool at room temperature for 15 min to allow the two sequences to anneal together and form the adapters. An initial digestion-ligation of genomic DNA of the SRVR clone with MseI (New England Biolabs, Boston, MA, USA) and the adapters was incubated at 37°C for 3 hours, and then directly amplified using the adapter-specific primers MseI-N (5'-GATGAGTCCTGAGTAAN-3'), which contained equal amounts of all four possible selective bases (N = A, T, C, or G). This permits the amplification of all fragments flanked by MseI sites. PCR reactions were performed in a total volume of 20 μl: 1.5 mM MgCl₂ (Applied Biosystems, Foster City, CA, USA), 30 ng of each of the four MseI-N primers, 200 μM each of dNTPs (GE Healthcare, Piscataway, NJ, USA), 1× Taq DNA polymerase buffer (Promega, Madison, WI, USA), 0.4 units of Taq DNA polymerase (Promega, Madison, WI, USA), and 5 μL of a 1:10 dilution of digested-ligated DNA. PCR conditions used were: 94°C 30 s, 53°C 1 min, 72°C 1 min, and 18, 20 and 23 cycles were tested to determine the optimal number of amplification cycles.

A total of 500 ng of amplified DNA using 20 PCR cycles were added to 200 pmol of HPLC-purified biotinylated oligonucleotide (AG)₁₇ or (ATG)₁₂ and combined in a total volume of 100 μL of SSC 4.2×, SDS 0.07%. The DNA mixture was denatured at 95°C for 3 min and allowed to anneal at room temperature for 15 min.

One mg of streptavidin-coated magnetic particle beads (Roche Applied Science, Indianapolis, IN, USA) was washed in TEN₁₀₀ buffer (10 mM Tris-HCl, 1 mM EDTA, 100 mM NaCl, pH 7.5) and re-suspended in 40 μl of the same buffer. Ten μl (corresponding to approximately 1 μg of DNA) of tall fescue PCR product using primers from the actin gene, were combined with the beads to minimize nonspecific binding of white clover genomic DNA. DNA-probe hybrid molecules were diluted with 300 μL of TEN₁₀₀, and combined with the prepared beads. A 450 μl DNA-probe-bead solution was incubated for 30 min at room temperature with constant gentle agitation. The hybridization buffer was separated from the beads-probe-DNA complex using a Dynabeads magnetic separation stand (Invitrogen, Carlsbad, CA, USA) and discarded. Washing of DNA-probe and DNA collection were performed as previously described 2.

Recovered DNA fragments from three separate wash steps were precipitated with one volume of isopropanol and sodium acetate (0.15 M final concentration), and re-suspended in 50 μl of water. Two μl of the recovered fraction were amplified with 30 cycles of PCR using the MseI-N primer using the same conditions described above. Four separate PCR reactions were conducted, each one with one primer (MseI-A, or T, or G, or C) removed from the primer mixture to select the DNA populations with the least amount of duplicated amplification. The amplified fragments with the selected primers produced a smear when visualized in an ethidium-bromide-stained agarose gel. To check for complete removal of nonspecifically bound DNA, 2 μl of the last fraction recovered was amplified with the same primer sets and did not yield any product. The resulting PCR products were cloned into TOPO TA cloning vector and transformed with One Shot TOP10F' chemically competent cells (Invitrogen, Carlsbad, CA, USA) following the manufacturer's instructions. Colony PCR was conducted on 16 clones per library using universal M13 primers to examine the inserts.

Sequencing and data analysis were conducted at the Samuel Roberts Noble Foundation. Briefly, white colonies were cultured in 96-well culture blocks containing 1.5 ml TB and a salt supplement (TE-RNase A+T1, SDS/NaOH, 3 M NaOAc, pH 4.8) together with 100 μg/ml ampicillin for 22 hours at 37°C while shaking at 350 rpm. Double-stranded DNA template was isolated as described http://www.noble.org/PlantBio/Genomics/ProtocolBiomek.htm. All sequencing reactions were performed using the flanking M13 reverse primer site and analyzed in an ABI 3730 capillary sequencer using the BigDye^® Terminator Sequencing Kit (Applied Biosystems, Foster City, CA, USA). Base calling of the ABI sequencer trace files was done with the vendor program which was essentially based on Phred http://www.phrap.org. Each sequence was then processed using an in-house semi-automated pipeline for overall base quality screening and removal of any contaminating vector, mitochondrial, ribosomal and E. coli sequences. High quality sequences that passed this screen and having a minimum length of 100 bp were assembled into contigs and used for SSR detection. All assembled contigs were also examined for homology to a variety of public databases through batch BLASTN 20 searches.

The high quality sequence contigs were analyzed using the phpSSRMiner web server http://bioinfo.noble.org/phpssrminer/index.php, developed at The Samuel Roberts Noble Foundation. PhpSSRMiner provides a user-friendly interface which integrates its back-end pipeline to streamline the process of perfect SSR identification by SSRIT 21, imperfect SSR identification by Sputnik http://abajian.net/sputnik developed by 22, PCR primer design using Primer3 23, and in silico amplification of the designed primers via isPCR 24. SSR identification, PCR primer design and in silico amplification of the designed primers were performed and hosted by phpSSRMiner. The sequence information, both from the contig sequences and GenBank, feeds directly into the software and the output is a tab-delimited text file with separate columns for the positions and sequences of the forward and reverse primers, type of SSR repeat (perfect vs. imperfect), SSR motif, and length.

Forward and reverse primers were synthesized by Qiagen/Operon Biotechnologies (Los Angeles, CA, USA) with an additional 18 nucleotides from the M13 universal primer appended to the 5' end of the forward primer 25. PCR amplification and genotyping in the GA02-15 genotype from SRVR and the GA02-56 genotype from 'Durana' 26, as well as in six progeny from the resulting F₁ mapping population, were performed following standard procedures previously described 18. SSR fragments amplified were analyzed on the ABI PRISM 3730 Genetic Analyzer (Applied Biosystems, Foster City, CA, USA) and visualized and scored using GeneMapper 3.7 software (Applied Biosystems, Foster City, CA, USA).

In most plant species the AT/TA and GA/CT are the most common motifs 27. However, 28 found that CA and ATG were the most frequent repeat motifs in white clover and therefore, (CA)₁₇ and (ATG)₁₂ were used as the di- and tri-nucleotide repeat probes for constructing the libraries. To remove duplicate PCR products generated during amplification, the best combination of three primers (minus one of four MseI-A, -T, -C, or -G) was tested to amplify SSR-containing fragments in the last step of the DNA precipitation procedure. To avoid sequencing duplicated DNA fragments, only the PCR products showing smears and not distinguishable bands were used for cloning (Figure 1). Primers without MseI-G or MseI-C were used to produce the PCR products of the (CA)₁₇, and (ATG)₁₂ probe libraries, respectively (Figure 1). In a study with wheat, the mixture of primers minus MseI-A produced the most specific amplification (Mahmoud Zeid, personal communication) suggesting that the optimal primer combination might be species-specific. After removing the priming sites, the size range of the inserts for the (CA)₁₇ library was 200–300 bp and 200–350 for (ATG)₁₂. No double inserts were detected (data not shown).

Over 6,800 clones were generated from the two SSR-enriched libraries and sequenced. The (CA)₁₇ library contained 2,208 clones and the (ATC)₁₂ library contained 4,608 clones (Table 1). Approximately 56% and 66% of these clones generated successful sequencing reactions, with an average length of 238 bp and 263 bp for the two libraries, respectively. After assembling the sequences from the two libraries, 826 and 876 contigs were found in each library representing more than 61% and 43% good quality, unique sequences, indicating that the strategy of using selected primer combinations in PCR reduced duplication. A total of 1,698 contigs were found after assembling all sequences from both libraries, slightly fewer than the total number of contigs from the two libraries (Table 1). This indicates the uniqueness of the sequences selected with different probes using the FIASCO procedure, with very little overlap existing between the (CA)₁₇ and (ATG)₁₂ library sequences. A previous study used the Edwards method 3 to construct SSR-enriched libraries with six different restriction enzymes prior to enrichment of the libraries and generated 1,123 white clover clones with readable DNA sequences 28. Our study generated a total of 2,689 white clover genomic DNA sequences which have been submitted to GenBank http://www.ncbi.nlm.nih.gov. These include 1,224 sequences from the (CA)₁₇ library and 1,465 from (ATG)₁₂ library. The accession numbers [GenBank: EF690813-EF691565, EU416331 – EU416801] and [GenBank: EF691566-EF692395, EU416802 – EU417436] were assigned to the sequences from the (CA)₁₇, and (ATG)₁₂ libraries, respectively.

The repeat sequences identified using the phpSSRMiner software include both perfect and imperfect SSRs with a minimal size of 18 bp. A similar number of sequences containing SSRs were found in the two libraries, although these are based on different numbers of contigs (Table 2). The percentage of sequences with SSRs was higher in the (ATG)₁₂ than the (CA)₁₇library. Up to three repeat motifs were present in each contig and 260 and 318 SSRs were identified for the (ATG)₁₂ and the (CA)₁₇ library, respectively. The most frequent motif in the (CA)₁₇ library was TG, followed by CA, AC, and GT (Table 2). The most common and distinguishable SSR motifs identified in the (ATG)₁₂ library were GAT, TGA, CAT, and ATC. As expected, the most frequent repeats were di-nucleotides in the (CA)₁₇ library and tri-nucleotides in the (ATG)₁₂ library. Only 21 tri-nucleotide repeats were identified in the (CA)₁₇ library and only four di-nucleotide repeats were found in the (ATG)₁₂ library, indicating the high selection efficiency of FIASCO for either di- or tri-nucleotide repeats. These results suggest that the probing procedure was able to target the specific motif and their complementary strings, producing SSR-enriched libraries with the expected repeat motifs. The length of the repeat motifs ranged from 2–10 bp, with 2–34 copies of repeats. Both the di- and the tri-nucleotides were the most abundant type of repeat and represented approximately 45% of the total number of SSRs identified. Hexa-nucleotide repeats identified in both libraries were the second most frequent type of repeat detected. A total of 575 SSRs were detected in the assembled sequences. The top motifs GAT, TGA, and TG, CA in the (ATG)₁₂, and (CA)₁₇ libraries remain the same due to the unique sequences between the two libraries (Table 2). The motifs containing hexa-, tri-, and di-nucleotide repeats were the most common ones, accounting for 76% of all SSR motifs identified (Table 2).

BLASTN searches of the 1,698 contig sequences against 1,983 white clover nucleotide sequences available in GenBank using an e-value threshold of 1e^-20 indicated that only 123 sequences generated in this study had homology to GenBank white clover nucleotide sequences (Table 3). These sequences accounted for only 7.2% of the 1,698 sequences obtained from the libraries, indicating that more than 92% of the identified white clover sequences were unique. BLASTN searches against the NCBI non-redundant nucleotide library (NCBI NT) using a less stringent threshold of 1e^-10 indicate that approximately 68% of our white clover sequences have matches and the remaining one third of the contigs represent novel white clover sequences. BLASTN and TBLASTX searches against the Medicago truncatula databases (BAC and EST) resulted in matches for only 17% and 9% of the identified sequences in the IMGAG Medicago Pseudo Genome http://www.medicago.org/genome and M. truncatula Gene Index from TIGR (MtGI, http://compbio.dfci.harvard.edu/tgi/cgi-bin/tgi/gimain.pl?gudb=medicago), respectively. The low percent of sequence similarity between M. truncatula and these white clover sequences, as well as results from a previous study indicating the relatively low amplification rate of white clover genome sequences using M. truncatula EST-SSRs 18 further enhances the importance of these primer sequences.

As of Jan 22 '08, 4,672 white clover nucleotide sequences were available in GenBank. A total of 2,689 (58%) of them were generated in this study from the SSR-enriched genomic libraries. The remaining 1,983 white clover sequences were downloaded from GenBank and loaded into phpSSRMiner for microsatellite primer design. The most frequent motif repeats from GenBank sequences were CA and TG. These were also the most frequent motifs in the white clover sequence from the libraries developed in this study (Table 2). After removing the 102 GenBank sequences having high homology with the sequences obtained from the two genomic libraries, 1,881 sequences were used to design 597 primers that included 230 perfect and 367 imperfect SSR primers. The phpSSRMiner analysis also identified 262 SSR primers from the 1,698 white clover sequences generated in this study. From those, 22% (58) were perfect primers, indicating that they amplify perfect repeats of the motif, and 78% (204) were imperfect primers. Imperfect primers allow a one or two base pair insertion in the repeat motif. In total, 859 primer pairs were identified (see Additional file 1: Characteristics of 859 SSR primers developed using GenBank white clover nucleotide sequences and two genomic SSR-enriched libraries generated in this study) and their location is currently being mapped in a white clover population. A smaller number of SSR primers were generated from the two genomic libraries compared to GenBank database sequences even though a similar number of sequences were contained in each data set. The main reason for this difference was the length of the sequences. The average length of the 1,881 white clover sequences from GenBank was 813 bp compared to the 254 bp average of the contigs from the two genomic libraries. This difference in the length of the sequences affects the likelihood of identifying potential priming sites.

PCR amplification in the laboratory with a subset of randomly selected 191 primer pairs flanking 63 perfect and 128 imperfect SSRs designed using the phpSSRMiner software was used to verify amplification products of the expected size. DNA from the parents of a mapping population (GA43 and SRVR), as well as six F₁ progeny from a double-pseudo testcross 18 was used to determine the polymorphisms of these SSR primers (Table 4). A total of 176 (92%) of the primers screened produced amplicons, 66% of which were polymorphic between the two parents.

Primers were designed to amplify sequences containing both perfect and compound SSRs. For the 63 perfect and 128 imperfect SSRs, the percentage of primers amplifying polymorphic fragments was higher for the perfect (75%) vs. the imperfect SSRs (54%) (Table 4). Among the 191 primers, 33% of them were from hexa-nucleotide SSRs, followed by 25%, and 19% from tri- and di-nucleotide SSRs, respectively (Figure 2). Even though primers amplifying di-nucleotide SSRs were not the most abundant, they gave the highest percentage of polymorphism (Figure 2). Our results suggest that the FIASCO procedure can be utilized to develop molecular markers in allopolyploid genomes with homoeologs of high sequence identity, such as white clover.