To compare Hps gene structure among soybean cultivars or lines that differ in seed lustre, a DNA blot analysis was performed using an Hps cDNA probe. Figure 1 shows results from a representative analysis of ten different soybean lines, after digestion of genomic DNA with the restriction enzyme Bgl II. Polymorphisms were noted in both the number and intensity of hybridizing genomic DNA fragments among the different cultivars and lines. The most intensely hybridizing fragment was estimated to be 2.4 kb in size. This fragment could produce strong hybridization signals even after short exposure times, indicating that multiple copies may be present in genomes of selected soybean cultivars or lines. The presence of this hybridizing fragment was associated with seed phenotypes that were dull or intermediate in lustre. This fragment was absent from shiny seeded phenotypes. Two different soybean lines with a bloom phenotype showed contrasting patterns, with the hybridizing band present in Clark B1 but absent from Sooty.

This analysis was extended to compare soybean DNA samples to genomic DNA samples isolated from six related legume species, including five species from the same genus Glycine. The results shown in Figure 2 indicate that hybridizing bands to Hps cDNA could only be detected in two species, soybean (Glycine max) and wild soybean (Glycine soja), at least under the high-stringency hybridization conditions that were performed here. The genomic DNA samples from the five different soybean lines displayed greater differences in their hybridization patterns and signal intensities than did the four Glycine soja lines. The strong hybridization signals present in Glycine max cv Harosoy 63 were not detected in any of the Glycine soja lines, although the patterns of hybridization were similar. The hybridization results shown in Figure 2 also demonstrate that copies of the Hps gene are present in shiny seeded phenotypes.

The restriction enzyme fragments producing strong hybridization signals in the DNA blot analyses suggested that Hps may occur as a multi-copy gene in certain Glycine max lines, such as cv Harosoy 63. To estimate the number of copies of the Hps gene present in soybean cv Harosoy 63, hybridization signals were compared between samples of soybean genomic DNA and a plasmid standard carrying a single copy of the Hps gene. Results from this 'reconstruction hybridization' are shown in Figure 3. Measurement of the band intensities by image analysis results in a calculated value of 27 ± 5 copies of Hps per haploid genome, for the 2.4 kb Bgl II fragment, assuming a haploid genome size of 1.212 pg DNA 12. This is a minimum value, since additional bands hybridizing to the Hps cDNA were also present in the Bgl II digestions of cv. Harosoy 63.

To determine whether differences in Hps gene copy number among soybean lines could be detected by other methods, we performed a comparative genomic DNA hybridization to cDNA microarrays. This method has been shown to be effective to distinguish changes in gene copy number in other species 13. A total of six hybridizations were performed in three experiments, using a cDNA microarray of 18613 soybean cDNAs. Two experiments compared genomic DNA from OX281 to Mukden, and one experiment compared genomic DNA from Harosoy 63 to Sooty, as shown in Table 1. The results from Experiment II are shown in more detail in Figure 4.

The absolute hybridization values, and therefore the signal-to-noise ratios, were low compared to microarrays that were conventionally probed with cDNA derived from mRNA samples. Nonetheless, the normalized ratios for the ~18,000 genes on the array were tightly clustered around unity (~1), as expected, but the cDNA on the array encoding HPS displayed exceptional hybridization ratios in these experiments in every case. Thus, in each experiment that compared hybridization ratios of OX281:Mukden or of Harosoy 63:Sooty, the cDNA encoding HPS always displayed a hybridization ratio >2, ranging from 4.2 to 6.4. These results indicate that the Hps copy number differences detected by conventional Southern analysis are also detectable by comparative genomic DNA hybridization to cDNA microarrays.

To determine whether the Hps copy number polymorphisms cosegregate with seed lustre phenotype, seed surface protein (HPS), and associated genetic markers 2, we analyzed 30 F₃ families from a cross of OX281 and Mukden. A total of 8 of the F₃ families were shiny in phenotype, without surface HPS, and displayed a non-repetitive, low-copy Hps restriction fragment length polymorphism (RFLP) pattern. The remaining 22 F₃ families were dull in phenotype, with abundant surface HPS, and displayed a repetitive, high-copy Hps RFLP pattern. A representative hybridization is shown in Figure 5. As expected, results from this analysis indicate that Hps copy number polymorphisms absolutely cosegregate with seed lustre phenotype. This analysis additionally shows that the multiple copies of Hps that are present in OX281 segregate as a genetic unit, indicating that the copies occur together at a single locus.

To isolate the Hps gene(s), soybean genomic libraries were screened with an Hps cDNA probe. Additional probes, derived from the sequences of the genomic clones identified from the first round of screening, were used to isolate overlapping or flanking clones. In all, more than 30 genomic clones were isolated. The size, the library source, and the restriction enzyme digestion patterns of the various clones were compared. Using these criteria, the genomic clones could be classified into six different types. A representative of each type was chosen for complete sequencing, as shown in Table 2. Analysis of the DNA sequences revealed all of the clones shared regions of high sequence identity, and that three of the clones could be aligned to produce a repetitive motif, as shown in Figure 6. This hypothetical or model Hps repetitive motif was tested by DNA blot hybridization using three distinct probes derived from different regions of the repetitive unit. The results show that for each probe the most intensely hybridizing DNA fragments, representing most of the copies of the repetitive unit, could be accounted for by reconstructing the restriction enzyme fragments from the aligned genomic clones. A single Hps gene is present in each unit. A putative matrix attachment region was predicted to occur 3 kb upstream from the Hps open reading frame but no other genes were detected. Thus, most copies of Hps occur in a reiterated array of 8.6 kb units.

Other copies of Hps are also present in the soybean genome, as evidenced by the additional hybridizing bands on the DNA blots and by the genomic clones that do not exactly match the repetitive pattern. There are closely related genes, such as HPS2.1, that may correspond to paralogue pair-mates that arose from whole genome duplication, since soybean is considered an ancient tetraploid. Additional Hps genomic copies may also represent flanking regions or sequence variations occurring within the tandem array. For example, clones HPS1.5 (DQ208939) and HPS1.6 (DQ208940) share 97.5% sequence identity over their 8 kb length, but HPS1.5 has a single Hind III site whereas HPS1.6 possess two Hind III sites. These two copies of Hps will produce different patterns of hybridization after Hind III digestion. Although both patterns appear to be visible in the DNA blots, the smaller sized fragments produce much stronger signals, indicating that clone HPS1.6 with two Hind III sites better represents the majority of copies of Hps within the genome.

Seeds of soybean (Glycine max) and Lotus japonicus were from collections at Agriculture and Agri-Food Canada, and were provided by Dr. Vaino Poysa and Dr. Krzysztof Szczyglowski, respectively. Seeds of Glycine soja and Glycine tabacina were from the USDA Soybean Germplasm Collection. Seeds of Glycine canescens, Glycine curvata, and Glycine tomentella were from the Australian National Herbarium. Plants were grown in field plots outdoors or in glass enclosed greenhouses. The cross of soybean line OX281 to cv Mukden and the generation of F₃ families has been described 2. Seed lustre was determined by visual inspection.

Soybean genomic DNA was purified from frozen tissues using a modified CTAB (hexadecyltrimethyl ammonium bromide) method 25. Restriction enzyme digestion, electrophoretic separation on agarose gels, and blotting to Nylon membranes followed standard protocols 26. Agarose gels were stained with ethiduim bromide and examined prior to transfer to ensure equal DNA loading and digestion. To prepare probes, the Hps cDNA was excised by restriction enzyme digestion from a plasmid clone 5. Other probes were prepared by polymerase chain reaction (PCR) using cloned genomic fragments from soybean as DNA template (primer sequences and PCR conditions are available upon request). The DNA probes were isolated by excision from agarose electrophoresis gels, purified, and labelled with ³²P dCTP using a random primer labelling system (RediprimeII, Amersham Biosciences, Baie d'Urfé, Canada). Unincorporated ³²P dCTP was removed by gel filtration (Microspin S-200, Amersham Biosciences, Baie d'Urfé, Canada). Probes were denatured by heating to 95°C for 10 min. Hybridization was carried out at 65°C for 16 h in 0.25 mM Na₂HPO₄, pH 7.2, 7% (w/v) SDS, 1% BSA, and 1 mm EDTA. Filters were then washed four times for 15 min each at 22°C in high stringency wash solution (20 mM Na₂HPO₄, pH 7.2, 1% [w/v] SDS, and 1 mM EDTA), followed by three 15-min washes in the same solution at 68°C.

Microarray slides (soybean 18 K – A series) were purchased from Dr. Lila Vodkin, University of Illinois, Urbana, IL. The slides contain 18613 soybean cDNAs of low redundancy, sourced from a variety of tissues and organs 27. A total of six slides were hybridized in three separate experiments, using independent samples of genomic DNA for each experiment. Slides were prehybridized for 45 min in 5× SSC, 0.1% SDS and 1% BSA at 42°C, followed by two washes in 0.1× SSC at 22°C. The slides were dipped in water and dried by centrifugation. Genomic DNA purified from soybean tissues was digested with Dpn II and precipitated with ethanol prior to labelling with Cy3- or Cy5-dCTP (Amersham Biosciences, Baie d'Urfé, Canada) using published methods 13. Labelled DNA was purified by column chromatography, quantitated, and the amount of dye incorporation was determined 28. For hybridization mixtures, proportional amounts of Cy3- and Cy5-labelled DNA were dried by vacuum centrifugation, then re-dissolved and combined in a solution containing 22 ng uL^-1 mouse COT-1 (Invitrogen, Rockville, MD), 4.5 ng uL^-1 polyA DNA, 1× hybridization buffer (Amersham Biosciences, Baie d'Urfé, Canada), and 50% formamide (v/v). The hybridization solution (44 uL total volume) was denatured for 2 min at 100 °C, cooled for 30 min at 22°C, then applied to the prehybridized microarray slide. A 24 × 60 mm cover slip was placed over the slide and hybridization was carried out for 45 h in darkness, at 42°C. Post hybridization washes were as follows: one wash of 2× SSC and 0.1% SDS at 42°C for 5 min; two washes of 0.1× SSC and 0.1% SDS at 42°C for 2 min; and two washes of 0.1× SSC at 22°C for 1 min. Slides were dipped in water and dried under a stream of nitrogen gas prior to laser scanning (BioRad ChipReader with VersArray ChipReader v3.0 software, BioRad Corporation, Hercules, CA). Spot intensities were quantified and corrected for background signals (Array Vision v6.0 software, St. Catherines, ON, Canada), then values imported into GeneSpring v7.2 (Silicon Genetics, Redwood City, CA) and normalized by per spot, per chip, intensity dependant LOWESS. A dye swap was included in each experiment, and the final normalized hybridization ratios were averaged from separate values calculated for each slide. Data was MIAMI validated and deposited to the Gene Expression Omnibus 15 series GSE3810; samples GSM87407-GSM87412.

Methods for construction and screening of genomic DNA libraries of soybean cv Harosoy 63 have been described 29. Additional genomic libraries were prepared for this study, using commercially available λ-phage vectors and packaging extracts (Stratagene, La Jolla, CA), and by following the manufacture's instructions. Positive clones were plaque purified and sub-cloned into a plasmid vector (pBluescript, Strategene, La Jolla, CA) for sequence analysis. Automated sequencing of DNA was accomplished using dye-labelled terminators and fragments separated in acrylamide gels (model 377, Applied Biosystems, Foster City, CA). Genomic clones were shotgun sequenced by random transposon insertion (GPS-1, New England Biolabs, Beverly, MA) to an average of 10-fold coverage, and gaps were closed by primer walking. Finished sequences were assembled and edited using a computer program (Lasergene, DNAStar, Inc., Madison, WI).