Introduction

Sugarcane is an important industrial crop cultivated for its sucrose which accumulates in the plant stem internodes. This clonally propagated large grass belonging to the Poaceae family is also characterized by high content of valuable chemicals, very high photosynthetic efficiency and one of the highest efficiencies of water usage among plants. These traits provide a very attractive option of using sugarcane as a biomass producer and biofactory of desirable compounds (Botha 2006). Currently, sugarcane is grown on 13 million hectares in more than 100 countries across tropical and subtropical zones. Of almost 150 million tons of sugar consumed worldwide in 2005, 74% was made from sugarcane cultivated primarily in Brazil and India (FAO/FAOSTAT website). The world demand for sugarcane is growing rapidly in recent years due to increasing interest in utilising sugarcane sucrose to produce bioethanol as a biofuel alternative to petrol. Additionally, sugarcane’s efficient biomass production attracts a lot of interest for its potential to be converted to electricity as a response to the requirement to limit the world’s dependence on fossil fuels and to decrease greenhouse gas emissions (Goldemberg 2007).

Improving the efficiency of sugarcane breeding programs is an important component in addressing this growing demand for sugarcane. The conventional improvement of sugarcane varieties takes on average 12 years from an initial cross to the release of improved cultivars (Silva and Bressiani 2005). It is anticipated that sugarcane molecular genetics tools can accelerate the rate of genetic gain in breeding programs. However, sugarcane, as one of the most genetically complex crops, presents a considerable challenge for genetic and molecular analysis. The large genome (2C) of species belonging to the Saccharum genus (7,440 Mbp for S. officinarum L., Grivet and Arruda 2001; 10,000 Mbp for a modern cultivar R570, D’Hont 2005) is characterized by a high degree of polyploidy and frequent aneuploidy (D’Hont et al. 1996). Most sugar-producing modern cultivars are derived from crosses between S. officinarum (x = 10, 2n = 8x = 80) and S. spontaneum L. (x = 8, 2n = 5–16x = 40–128), both highly polyploid, resulting in a high number of chromosomes (100–130) which pair and recombine in many combinations (Jannoo et al. 2004). The genome of current cultivars is derived in ~70 to 80% from S. officinarum, in ~10 to 20% from S. spontaneum, while about 10% are recombinant chromosomes of these two species (D’Hont et al. 1996).

In spite of this genome complexity, a significant progress has been made in genomics of sugarcane, including advances in molecular cytogenetics, diversity analysis, and genetic mapping (reviewed in Grivet and Arruda 2001; D’Hont et al. 2008). Additionally, a sugarcane Bacterial Artificial Chromosome (BAC) library (Tomkins et al. 1999) and over 237,000 sugarcane Expressed Sequence Tags (EST) (Vettore et al. 2006) have become available. Several types of molecular markers have been established for sugarcane including RFLPs (Da Silva et al. 1993), RAPDs (Mudge et al. 1996), SSRs (Cordeiro et al. 1999) and AFLPs (Hoarau et al. 2001). Markers developed so far have been used for molecular investigation of the sugarcane genome structure (Grivet et al. 1996; Jannoo et al. 1999a) and for tracking genetic diversity and relationships between species and cultivars (Jannoo et al. 1999b; Lima et al. 2002; Aitken et al. 2006). Genetic maps have also been constructed for S. spontaneum, S. officinarum and modern cultivars. The first genetic maps were generated with one type of marker, such as RFLPs (Grivet et al. 1996), RAPDs (Mudge et al. 1996) and more recently AFLPs (Hoarau et al. 2001), providing a genome coverage ranging from 2,008 cM for RFLPs to 5,849 cM for AFLPs. To expand genome coverage and to improve identification of linkage groups, various types of markers have been integrated. Recently, a combination of AFLP, RAF and SSR markers revealed 136 linkage groups with a total map length of 9,058 cM in the Q165 Australian cultivar using a segregating F1 population (Aitken et al. 2005).

A significant factor limiting the application of established molecular markers in breeding programs is the speed and cost at which sufficient numbers of markers can be screened across large populations of genotypes. Moreover, some PCR-based marker systems such as SNPs or SSRs rely on sequence information for their initial development, limiting their use to the species with sufficient sequence information. This especially affects polyploid species, sequencing of which is technically difficult. As modern sugarcane cultivars have approximately 110 chromosomes corresponding to a genetic map of about 17,000 cM, and given the species’ high ploidy level, 5,000–7,000 markers would be needed to achieve an adequate coverage for genetic mapping and breeding applications. We therefore attempted to develop Diversity Arrays Technology (Jaccoud et al. 2001) for this crop. DArT combines the low cost and high throughput of the DNA microarray platform with the ability to identify various types of DNA polymorphisms (INDELs, SNPs, methylation changes) (Kilian et al. 2005). Hundreds to thousands (depending on the polymorphism level) of polymorphic features (DArT markers) distributed across the genome are discovered in a single hybridisation-based assay with an accuracy of 99.8% (Wenzl et al. 2004; Akbari et al. 2006). DArT markers do not require further assay development once they are discovered and they are sequence-independent but sequence-ready. The DArT technology proof-of-concept was originally developed on rice (Jaccoud et al. 2001). Subsequently, the technology was established for plant species with bigger and more complex genomes such as barley (Wenzl et al. 2004), polyploid wheat (Akbari et al. 2006) and polyploid oats (Tinker et al. 2009) as well as for another fifty plant species, two species of plant pathogenic fungi, a bacterium and two animal species (Lezar et al. 2004; Wittenberg et al. 2005; Xia et al. 2005; Yang et al. 2006; Sessittsch et al. 2006; Mace et al. 2008; James et al. 2008; Bonin et al. 2008; Mantovani et al. 2008). An increasing number of published studies indicate a successful deployment of DArT markers in a variety of applications, for example construction of high-density consensus genetic maps providing a framework for transferring genetic information between different marker systems, identification of quantitative trait loci (QTL) for variety of phenotypic traits and studies of genetic diversity in crops such as wheat, barley, sorghum, oats and others, as well as genomic evolutionary studies in non-model organisms (www.DiversityArrays.com/publications).

Here, we describe the effective development of DArT technology for sugarcane. Numerous methods of genome complexity reduction were evaluated for their efficiency in detecting DArT markers. The best methods were established and successfully tested in diversity analysis of sugarcane ancestral accessions and modern cultivars. The majority of identified sugarcane DArT markers segregated in a Mendelian fashion and were incorporated into a framework genetic map. The successful validation of DArT genotyping in sugarcane demonstrates that the technology can serve as a high-throughput genome scan for plants with high levels of genome complexity, cost-effectively providing enough markers for a range of applications such as whole-genome-based molecular breeding, diversity analysis or genetic identity testing for a broad spectrum of plant species.

Materials and methods

Plant material

Leaf samples were supplied by the Cooperative Research Centre for Sugar Industry Innovation through Biotechnology (CRCSIIB, Australia). Table 1 lists 16 sugarcane genotypes used for testing of the genome complexity reduction methods, developing DArT libraries and arrays, and generating genomic representations to test performance of the constructed arrays.

Table 1 Sugarcane genotypes used for testing the methods of genome complexity reduction and developing DArT arrays
Full size table

The Q813 (S. officinarum) and a cultivar F172 were used to estimate reproducibility and robustness of a PstI/TaqI genome complexity reduction method.

The Q813 was a driver in the construction of the PstI/TaqI Suppression Subtractive Hybridization (SSH) DArT library, while a tester was a mixture of 16 sugarcane genotypes listed in Table 1.

The IJ76-514 (S. officinarum) and Q165 (an Australian cultivar), parents of a mapping population, were used to develop the PstI/TaqI SSH DArT library constructed for an assessment of sugarcane DArT markers segregation patterns in the genotyping experiment of the 94 progeny clones and the parents of this mapping population.

DNA extraction

Genomic DNA was extracted using a modified cetyltrimethylammonium bromide (CTAB)/chloroform/isoamylalcohol method (Doyle and Doyle 1987).

Testing the methods of genome complexity reduction

Various combinations of 12 restriction enzymes were tested in a total of 10 methods (Electronic Supplementary Material: Online Resource 1) using 16 sugarcane genotypes representing the Saccharum genetic diversity (Table 1). For each method, ~100 ng of DNA of each sample listed in Table 1 was used. Digestion and ligation reactions were carried out simultaneously with 2 units of each of restriction enzymes (NEB), 80 units of T4 DNA ligase (NEB) and 0.05 μM of adaptor(s) (Online Resource 1, or specified individually). One microlitre of the digestion/ligation product was used as template for amplification with 2 units of RedTaq DNA polymerase (Sigma) and 0.4 μM of primer if one primer was used or 0.2 μM of primer if two primers were used. The eight methods are summarized in Online Resource 1a.

For method 9, 50 μl of PstI/TaqI amplicons generated as described for method 1 (Online Resource 1a) were redigested with 1 unit each of DpnII, HpyCH4IV, MseI and NlaIII (NEB) restriction enzymes to further decrease the genome complexity. A 25 μl digestion product was ligated to 0.05 μM of equimolar mixture of two sets of adaptors (sets A and B, Online Resource 1b) with 160 units of T4 DNA ligase (NEB). Five microlitres of the resulting ligation product was used as a template for amplification with 2 units of RedTaq DNA polymerase (Sigma); primers and amplification conditions used were as specified in Online Resource 1b.

Finally, a method based on the display of Miniature Inverted-repeat Transposable Elements (MITEs) was tested (method 10, Online Resource 1c). In this method the genomic representation comprised the fragments amplified between the Terminal Inverted Repeats (TIRs) present at both ends of MITEs and an adaptor ligated to overhangs created by Bsp1286I restriction enzyme (S. Patarapuwadol, personal communication). Restriction with Bsp1286I (NEB) with simultaneous ligation of Bsp1286I adaptors (Online Resource 1c) was carried out as described for other methods. A 0.5 μl aliquot of digestion/ligation product was used as template for the first round of amplification with 0.04 μM of primer I (Online Resource 1c) complementary to adaptor ligated to the Bsp1286I site, 0.4 μM of each of the primers (II and III, Online Resource 1c) amplifying from TIRs and 1 unit of RedTaq DNA polymerase (Sigma). Amplification conditions were as follows: 94°C for 1 min, 15 cycles of 94°C for 30 s, 50°C for 40 s, followed by 72°C for 1 min. A 0.5 μl aliquot of the resulting amplification product served as a template for the second round of amplification performed as in the first round except that the concentration of primer I was increased tenfold and 10 more cycles were added.

An approach to enrich for polymorphic clones

In an attempt to enrich for polymorphic clones, a Suppression Subtractive Hybridization (SSH) technique was tested. The guidelines for SSH were as published by Diatchenko et al. (1996). Sugarcane ancestral species Q813 (S. officinarum) was used as a driver to subtract the shared component of the genome from a mix of 16 genotypes used as a tester (listed in Table 1). At first, the genome complexity of driver and tester was reduced as described above for method 9 (Online Resource 1b). Subsequently, driver and tester DNA was extracted with phenol/chloroform and precipitated with isopropanol according to the standard methods (Sambrook et al. 1989). Two hundred nanograms of purified tester DNA was ligated to 10 μM of two sets of adaptors (set A and set B, Online Resource 1b) in two separate ligation reactions, each reaction with 80 units of T4 DNA ligase (NEB). Subtraction was done in two quantitative ratios of driver to tester (30:1 and 10:1) and carried out in one and two rounds of subtractive hybridization. For one round subtraction, DNA quantities of driver and tester appropriate to the specified ratios were mixed separately for each of the two sets of adaptors. The mixed DNA was precipitated with isopropanol, dissolved in 1.5 μl Subtraction Hybridization Buffer (Diatchenko et al. 1996), overlaid with 5 μl mineral oil, denatured at 98°C for 1.5 min and allowed to hybridize for 5 h at 68°C. Once the hybridization was completed, the subtraction product was diluted with 200 μl Subtraction Dilution Buffer (Diatchenko et al. 1996), and then heated at 72°C for 7 min. The two rounds subtraction was done as one round but after 5 h incubation at 68°C, the hybridization products with set A of adaptors were mixed with hybridization products with set B of adaptors, separately for different ratios. Subsequently, 300 ng of freshly denatured (98°C for 1.5 min) driver was added to each ratio variant. The mixtures were allowed to hybridize for an additional 5 h at 68°C. Once the hybridization was completed, each subtraction product was diluted with 200 μl Subtraction Dilution Buffer, and then heated at 72°C for 7 min.

One, 5 or 10 μl aliquots of each variant of the hybridization product were amplified with 0.2 μM of each of primers 5′-CTGAGTAGTGCCAGAACGGTC-3′ and 5′-TCGTAGACTGCGTATCCG-3′ and 2 units of RedTaq DNA polymerase (Sigma). The reaction conditions were as follows: 72°C for 7 min, 95°C for 3 min, 30 cycles of 94°C for 20 s, 58°C for 40 s, 72°C for 1 min followed by 72°C for 7 min.

The steps of DArT technology downstream of the genome complexity reduction method were as previously published (Jaccoud et al. 2001; Wenzl et al. 2004; Akbari et al. 2006) with some modifications and will be described briefly.

Preparations of DArT arrays

Each method of the genome complexity reduction was used to build a library comprising from 768 to 4,608 individual clones. For each method, genomic representations of the 16 genotypes listed in Table 1 were mixed and cloned using a “TOPO TA Cloning” kit (Invitrogen) according to the methods recommended by the manufacturer. Individual recombinant colonies were grown overnight in 384-well plates in a “freezing medium” (LB with 100 μg/ml ampicillin and mix of salts eliminating LB inhibitory effects on subsequent PCR amplification; unpublished observation). A 0.5 μl aliquot of bacterial culture was used as a template for insert amplification with 0.2 μM of each M13 Forward and M13 Reverse primers (Invitrogen). The cycling conditions were as follows: 95°C for 4 min, 57°C for 35 s, 72°C for 1 min followed by 35 cycles of 94°C for 35 s, 52°C for 35 s and 72°C for 1 min and finally 72°C for 7 min. The amplicons were dried, washed with 70% ethanol and dissolved in a spotting buffer developed especially for poly-l-lysine-coated microarray slides (P. Wenzl, personal communication). The amplified clones suspended in the spotting buffer were printed in duplicate on the poly-l-lysine-coated slides (Erie Scientific) using a MicroGridII arrayer (Biorobotics). After printing, the DNA deposited onto slides was denatured by incubation in hot water (95°C) for 2 min, followed by dipping in MQ water supplemented with 0.1 mM DTT and 0.1 mM EDTA and drying by centrifugation at 500g for 7 min.

Genotyping using DArT arrays

Preparation of genomic representations of individual sugarcane genotypes (=targets) for hybridization to the DArT arrays was done using the same methods as for target preparation for DArT arrays generation. Genetic representations were precipitated with isopropanol and dissolved in molecular grade water (Sigma) before labelling. Denatured targets were labelled with 2.5 units of exo-Klenow fragment of E. coli Polymerase I (NEB), 25 μM random decamers and 2.5 nmoles of either Cy3-dUTP or Cy5-dUTP (Amersham Bioscience). Labelling was carried out for at least 3 h at 37°C. Labelled targets were denatured for 2 min at 95°C, mixed with 60 μl of hybridization solution and deposited onto the microarrays. The hybridization solution was composed of a 50:5:1 mixture of, respectively, ExpressHyb (Clonetech), herring sperm DNA (Promega), FAM-labelled polylinker of pCR 2.1 vector used for cloning (Invitrogen), and 2 mM EDTA (pH 8.0). Hybridization was carried out for 18 h at 60°C. Following hybridization, the slides were washed at room temperature in four solutions with increasing salt stringency (solution 1—1 × SSC, 0.1% SDS; solution 2—1 × SSC; solution 3—0.2 × SSC; solution 4—0.02 × SSC) and dried by centrifugation at 500g for 7 min. Finally, the microarrays were scanned using a Tecan LS300 (Grödig, Austria) confocal laser scanner. Image analysis and polymorphism identification and scoring was done with the programme DArTSoft (version 7.4.3) specifically developed for that purpose by Diversity Arrays Technology Pty. Ltd. (www.DiversityArrays.com/software.html) and as described in Wenzl et al. (2004) and Akbari et al. (2006). Briefly, DArTSoft analyzed the microarray images, extracted the hybridization intensity data from the hybridized arrays, identified and scored polymorphism and calculated a range of quality parameters for each marker. The polymorphism analysis was performed with DArTSoft default settings which select markers of high quality and reproducibility based on extensive experience with several other genomes (Wenzl et al. 2004; Wittenberg et al. 2005; Xia et al. 2005), including polyploid wheat (Akbari et al. 2006).

Evaluation of the genome complexity reduction methods

The genetic representations of 16 samples representative of the Saccharum diversity and used to build the arrays (Table 1) were hybridized to the microarrays printed for each of the tested complexity reduction methods. Each sample was analyzed with replication to allow evaluation of technical reproducibility. DArTSoft analysis with the default settings was run for each hybridization experiment/method of genome complexity reduction. The most important criterion taken into consideration for choosing the best genome complexity reduction methods was the frequency of polymorphic markers. The average Call Rate (the percentage of successful allelic assignments), the average value of function “P” (measuring the level of relative target signal bimodality/polymorphism for a particular spot) and reproducibility of scores obtained from independent technical replicates were considered as supporting selection criteria.

Validation of the PstI/TaqI method in reducing the genome complexity and testing method reproducibility

To validate the PstI/TaqI method in reducing the genome complexity, a hybridization experiment of genomic DNA of Tabongo (S. spontaneum), Badila (S. officinarum) and Q124 (cultivar) was carried out on a PstI/TaqI array containing 1,536 clones. Each sample was hybridized in two replicates. One hundred nanograms of genomic DNA was labeled with Cy3-dUTP, hybridized and analyzed as described in the section “Genotyping using the DArT arrays”.

In order to assess reproducibility of the PstI/TaqI method, a genotyping experiment on two sugarcane accessions (Q813 and F172) was carried out on the PstI/TaqI array with 1,536 clones printed in triplicate. For each genotype, two independent digestion/ligation steps were performed and from each reaction six targets were prepared using two different methods of amplification in triplicate. One amplification method used a single 50 μl volume reaction while the second method was based on bulking three independent reactions with 17 μl each. The targets were labeled with Cy3-dUTP, hybridized and analyzed as described in the section “Genotyping using the DArT arrays”.

The reproducibility of replicated targets was estimated by calculating a correlation (R 2) between the genomic representations’ relative signal intensities (log2 of ratio derived from fluorescence signal of genomic representation divided by reference signal).

Assessment of Mendelian behavior of DArT markers

Construction of a DArT array for the assessment of Mendelian behavior of DArT markers

The libraries built during testing of the methods of genome complexity reduction were constructed using germplasm comprising sugarcane genetic diversity, not the best suited for the characterization of DArT markers’ segregation patterns in a specific genetic mapping population. In order to enrich the clone pool for clones segregating in the specific progeny, the library was constructed from the parents of the IJ76-514 (S. officinarum) × Q165 (an Australian cultivar) mapping population. We used the Suppression Subtractive Hybridization technique as described in the section “An approach to enrich for polymorphic clones”. Each parent was used both as a tester and a driver, respectively, in a 30:1 ratio of driver to tester in one or two rounds of subtraction. The constructed library comprised 3,072 clones, 1,536 clones each for one-round and two-rounds subtraction methods. The final array used in the experiment to assess the Mendelian behavior of DArT markers consisted of 6,144 clones. The array composition is shown in Online Resource 2.

Marker segregation

Each marker was scored as present (1) or absent (0) in the 94 progenies of the IJ76-514 × Q165 mapping population. A marker was considered reliable if scored successfully in at least 75% of individuals, a threshold significantly lower compared to the thresholds used for species with less complex genomes. Markers were tested for deviation from the segregation ratios expected for single-dose (1:1) and bi-parental single-dose markers or double-dose markers present in only one parent (3:1) by the Chi-square (χ2) test and declared as significantly distorted at P ≤ 0.05.

Sequence analysis of the sugarcane candidate polymorphic DArT markers

The 384 polymorphic clones were chosen from the hybridization experiments performed using the PstI/TaqI/SSH and PstI/TaqI arrays. The criteria for choosing the candidate polymorphic markers were as follows: (1) if the given sample was hybridized in more than one experiment, the obtained marker scores were compared and markers with up to one discordant score were chosen; (2) if the sample was hybridized in one experiment only and thus marker score comparison could not be carried out, markers with 100% reproducibility and P > 80 were chosen.

Altogether, 332 PstI/TaqI/SSH clones and 52 PstI/TaqI clones were selected. Bacterial cultures of the chosen clones were rearrayed from the original libraries to the 384-well microtiter plate filled with 50 μl of the “freezing medium” using a MicroGridII arrayer (Biorobotics). After rearraying, the cultures were allowed to grow for 24 h at 37°C. A 0.5 μl aliquot of each culture was used as a template for insert amplification in the conditions as described in “Preparation of DArT arrays”. The amplified inserts (markers) were sequenced in the Biomolecular Resource Facility (BRF) of the Australian Cancer Research Foundation (ACRF), Australian National University, Canberra, Australia. The sequencing reactions were carried out in-house according to the methods recommended by BRF (www.brf.jcs.anu.edu.au/services/DNAsequencing). Amplified PCR products were agarose-gel purified and sequenced on an ABI 3730 sequencer (ACRF BRF, ANU) following the manufacturer’s protocol (Applied Biosystems 2002). The traces were viewed in Sequence Scanner software v. 1.0 (Applied Biosystems). The 31 sequences of either poor quality (the basecaller quality value less than 20 for majority of chromatogram) or too short (less than 50 bp) were rejected. A Perl script developed in-house (Diversity Arrays Technology Pty. Ltd, unpublished) was used to trim the adaptor sequences, or up to 30 bp upstream of EcoRI cloning site from both the 3′ and 5′ ends if adaptor sequences could not be read. Trimmed sequences were subjected to NBLAST and TBALSTX searches against the nonredundant (nr) and the EST section of GenBank with an E value threshold of 1e-04. The searches were done in batches using a custom-written Perl script based on NCBI CGI BLAST interface (Diversity Arrays Technology Pty. Ltd, unpublished). The sequences were also analyzed for redundancy by using pairwise comparisons to detect sequence homology over a minimum of 100 bp. This was performed using stand alone BLAST v. 2.2 (NCBI) as well as scripts written in Perl and Mathcad v. 11.2a (Mathsoft) (D. Jaccoud, personal communication).

Results

Evaluation of the methods for genome complexity reduction

The method of genome complexity reduction is the first step in DArT technology. Resulting genomic representations are the starting point for the downstream steps in DArT genotyping. A number of approaches to reduce the genome complexity can be utilized. We based our methods on either solely restriction-enzyme (RE) digestion of DNA, adaptor ligation and amplification of adaptor-ligated fragments (methods 1–8 and method 9, Online Resource 1a and 1b, respectively) or a combination of this approach with exploiting naturally occurring Miniature Inverted-repeat Transposable Elements (MITEs) in plants (method 10, Online Resource 1c). In summary, we used various combinations of single or double primary cutters (6 bp) with one to five secondary cutters (4 bp) followed by the ligation of one to five adaptors and amplification of the adaptors’ ligated fragments. Each method was tested on a set of 16 sugarcane genotypes representing old cultivars present in the genealogy of most cultivars worldwide, important cultivars or elite parents in Australia and including two genotypes of each of the ancestral species S. officinarum and S. spontaneum (Table 1). The results of polymorphism analysis for each of the tested methods are summarized in Table 2. The polymorphism frequencies varied almost 20-fold between the methods, in the range from 0.52% (method 7) to 9.78% (method 10) with an average of 4.09% (P < 0.0001 as determined by a Pearson’s χ2 statistic). Based on the results presented in Table 2 and criteria described in “Methods” in the section “Evaluation of the genome complexity reduction methods”, three methods of genome complexity reduction, PstI/TaqI, PstI/MseI, and MITE/Bsp1286I, were the most promising with 7.03, 6.64, and 9.78% polymorphism frequency, respectively (Table 2; methods 1, 2 and 10).

Table 2 Performances of the tested methods of genome complexity reduction
Full size table

In order to validate the performance of the PstI/TaqI method in reducing the genome complexity, genomic DNAs of one S. officinarum, one S. spontaneum, and one modern cultivar were hybridized to the PstI/TaqI array without reducing their genome complexity. For each sample tested, approximately only 5% of the genomic DNA hybridized to this array (Fig. 1, panels a and c), indicating the effectiveness of PstI/TaqI clones in simplifying the genome complexity. We estimate that the PstI/TaqI genomic representation has in excess of 10,000 individual fragments representing 0.05–0.1% of the genome and reduces its complexity by filtering away the repetitive sequences (see “Discussion” for the details).

Fig. 1
figure 1

Effectiveness of the genomic representations based on the PstI/TaqI method in reducing the genome complexity of sugarcane. The PstI/TaqI based representations effectively filter away the repetitive sequences: ~95% of the PstI/TaqI clones contain low copy sequences. a A PstI/TaqI array hybridized with genomic DNA of sugarcane cultivar Q124 labelled with Cy3 dye; b same array hybridized with the reference DNA labelled with FAM dye; c a graph representing median signal intensity of genomic DNA of three sugarcane genotypes hybridized to the same array

Full size image

We tested the PstI/TaqI method’s reproducibility and robustness by analyzing two sugarcane accessions (S. officinarum Q813 and cultivar F172), as described in “Methods”, on the array containing 1,536 clones printed in triplicate. Analysis of the 24 arrays (12 for each genotype) for a relative signal intensity at each probe on the array indicated that DArT assays based on PstI/TaqI complexity reduction method are highly reproducible with an R 2 of approximately 0.98 (Online Resource 3). In addition, this experiment identified 55 DArT markers among 1,536 probes (3.6%), discriminating Q813 and F172 with 100% consistency (data not presented).

Enrichment for the polymorphic markers

We attempted Suppression Subtractive Hybridization (SSH) (Diatchenko et al. 1996) to enrich for the polymorphic markers (see “Discussion” for rationale). The results of the SSH experiment are shown in Table 3. The average polymorphism frequency of all subtraction variants (one or two rounds of subtraction, 30:1 or 10:1 quantitative ratio of driver to tester—method 10 A, B, C, D in Table 3) was only slightly better than the PstI/TaqI method only, 7.63 and 7.03%, respectively (Tables 2, 3, method 1, respectively). However, one-round subtraction with a 30:1 quantitative ratio of driver to tester resulted in 13.28% polymorphism frequency, almost doubling the result obtained for the PstI/TaqI method only (P < 0.0001 based on a Pearson’s χ2 statistic). To compare these two methods directly, another experiment was performed on the expanded PstI/TaqI array (3,456 clones) with 2,688 clones of a PstI/TaqI/SSH array (method 10 A, B, C, Table 3) by hybridization of genomic representations from two genotypes of each ancestral species S. spontaneum and S. officinarum, and 12 cultivars (Table 1). Each representation was hybridized in four independent replications. The polymorphism frequencies obtained were similar to the previous results with 7.71 and 6.34%, respectively for the PstI/TaqI/SSH-based array and the PstI/TaqI-only array. However, because sugarcane ancestral species, especially the “wild” species S. spontaneum, are strongly differentiated from modern cultivars on the genetic level (Grivet and Arruda 2001), they contribute the most to the detectable polymorphism. Accordingly, once four ancestral species were removed from the polymorphism analysis, the PstI/TaqI/SSH-based array was better in identifying polymorphism among modern cultivars with 6.32% polymorphism frequencies compared to 3.76% for PstI/TaqI array (P < 0.0001 based on a Pearson’s χ2 statistic).

Table 3 Summaries of the results of suppression subtractive hybridization experiment to enrich for polymorphic clones
Full size table

To test the performance of the PstI/TaqI versus PstI/TaqI/SSH polymorphic markers, the markers generated in the described experiment were used in a Factorial Analysis of Correspondence (FAC). The analysis was based on a Jaccard dissimilarity matrix computed from the presence/absence of alleles at 125 and 103 markers, respectively, for PstI/TaqI and PstI/TaqI/SSH, using the DARwin5 software (Perrier and Jacquemoud-Collet 2006). Figure 2 illustrates the improved performance of PstI/TaqI DArT markers subjected to SSH in better determination of a genetic relationship pattern within the population of analyzed sugarcanes, especially in separation of S. officinarum ancestral species from the modern cultivars. As expected, both PstI/TaqI and PstI/TaqI/SSH markers revealed the “wild” ancestral species S. spontaneum (Mandalay and Tabongo) as the most distant. For PstI/TaqI markers (Fig. 2a), Axis 1 (73.5% of total variance) separated S. spontaneum (Mandalay and Tabongo) from S. officinarum (Korpi and Badila), which grouped together with the cultivars (11.3% of variation detected, Axis 2). Once a part of the common component of the S. officinarum genome was removed by subtraction, the genetic differences between S. spontaneum, S. officinarum, and the cultivars were much better resolved (Fig. 2b). The separation between S. spontaneum and other genotypes was along Axis 1 (with 41.7% of total variance explained) while S. officinarum separated from the modern cultivars along Axis 2 (23.2% of total variance).

Fig. 2
figure 2

Comparison of the performance of the PstI/TaqI DArT markers with and without subtraction (SSH) in revealing the genetic relationships in sugarcane. The PstI/TaqI SSH markers perform better in determining the pattern of genetic relationships among sugarcane genotypes representative of Saccharum diversity, as revealed by two-dimensional plots of factorial analysis (FAC) based on a Jaccard dissimilarity matrix. FAC was computed from presence/absence of 125 PstI/TaqI (a) and 103 SSH DArT markers (b) for four ancestral species Mandalay and Tabongo (S. spontaneum), Korpi and Badila (S. officinarum), and twelve cultivars: 1—F172, 2—QN58-829, 3—CP74-2005, 4—Tellus, 5—Q201, 6—Trojan, 7—Mida, 8—CO475, 9—Trojan, 10—QN67-3184, 11—Q124, 12—Q208

Full size image

Genetic relationships among sugarcane accessions revealed by DArT markers

The DArT markers of the two best complexity reduction methods MITE/Bsp1286I and PstI/TaqI/SSH (method 10 in Tables 2, 3, respectively) were tested for their performance in revealing the genetic relationship between 16 sugarcane genotypes used for researching the genome complexity reduction methods and building the arrays (Table 1). The library created with the PstI/TaqI/SSH method was first expanded to 6,144 clones and the genotyping experiment for 16 samples was repeated on this new array with three independent replications of each sample. A total of 364 markers were scored by DArTSoft with the default settings. These markers, along with 303 markers obtained for the same set of 16 genotypes with the MITE/Bsp1286I method, were used to construct two UPGMA dendrograms based on the Nei/Li restriction-fragment-distance index using the PHYLIP 3.6 software package (Felsenstein 2002). As shown in Fig. 3, the relationships revealed among the assayed sugarcane genotypes appear consistent with expectations and results obtained with other technologies: the ancestral species of S. spontaneum and S. officinarum separated from the rest of the samples with both methods of genome complexity reduction. As expected, S. spontaneum samples Tabongo and Mandalay were the most distant and formed individual clusters on the dendrogram. The S. officinarum Korpi and Badila, although separated from modern clones and cultivars on a separate branch, still formed one clade with them, in agreement with the notion that ~80% of the modern sugarcane genome is derived from S. officinarum species (D’Hont et al. 1996). The consistency of genetic relationship revealed by the complexity reduction methods surveyed is noteworthy since they were targeting different types of DNA sequences and use different enzymatic methods. However, maximum bootstrap values (1,000 replicates) obtained for the PstI/TaqI/SSH method, i.e. 100% for both ancestral species (Fig. 3a) versus 100% for S. spontaneum but only 66% for S. officinarum in the MITE/Bsp1286I method (Fig. 3b), showed that the PstI/TaqI/SSH method provided a higher level of confidence in defining the genetic relationship. The UPGMA dendrograms supported the result obtained already in FAC for the PstI/TaqI/SSH versus the PstI/TaqI method (Fig. 2).

Fig. 3
figure 3

Genetic relationships between sugarcane genotypes representative of Saccharum diversity revealed by the DArT markers. UPGMA dendrograms based on the Nei/Li restriction-fragment-distance index for four ancestral sugarcane species and twelve cultivars based on 364 polymorphic markers obtained with the PstI/TaqI SSH genome complexity reduction method (a) and 303 polymorphic markers obtained with the MITE/Bsp1286I genome complexity reduction method (b). The numbers at the branches are confidence values based on Felsenstein’s bootstrap method shown only if greater than 50% for 1,000 replicates

Full size image

Genetic behavior of DArT markers in sugarcane

The segregation of DArT markers in sugarcane was assessed on 94 progeny genotypes from the cross IJ76-514 (S. officinarum) × Q165 (an Australian cultivar). The genotyping experiment was carried out on the array consisting of 6,144 clones (Online Resource 2). Individual progenies (in duplicate) were hybridized to this array and DArTSoft analysis with the default settings resulted in 241 polymorphic markers giving 3.9% polymorphism frequency. In the absence of segregation distortion, the expected segregation ratios for the tested mapping population were 1:1 for single-dose and 3:1 for both bi-parental single-dose markers and double-dose markers present only in one parent. The majority of the polymorphic markers (204, 84.7%) segregated as single-dose markers or double-dose markers with either 1:1 or 3:1 ratios. Of these 204 markers, 73% segregated as single-dose markers, while 27% segregated as bi-parental single-dose markers (or double-dose markers present only in one parent). The parents scored differentially for 215 markers (89.2%) with the majority of scoring assignments “1” coming from the Q165 (an Australian cultivar) parent. The IJ76-514 (S. officinarum) parent had only 21% of markers scored representing “1” allele. The segregation patterns of the polymorphic markers are summarized in Table 4, while Fig. 4 shows an example of DArT markers incorporation into linkage groups of homo(eo)logy group II of a framework genetic map of the Q165 sugarcane cultivar (Aitken et al. 2005).

Table 4 Summary of the segregation patterns of 241 polymorphic DArT markers scored for the IJ76-514 × Q165 mapping population
Full size table
Fig. 4
figure 4

Location of DArT markers on linkage groups from homo(eo)logy group II on a genetic linkage map of the Q165 sugarcane cultivar. The DArT markers were incorporated into the Q165 linkage map generated by Aitken et al. (2005) using Joinmap v 3.0 as in Aitken et al. (2007). The DArT markers are shown in boldface and annotated according to the method of genome complexity reduction used to generate them: BstNI followed by the 6-digit number represents the PstI/TaqI/BstNI method, HaeIII followed by the 6-digit number represents the PstI/TaqI/BstNI method, and the 6-digit number represents the PstI/TaqI method. The annotations of the other types of markers are as in Aitken et al. (2005)

Full size image

Analysis of sugarcane candidate polymorphic DArT markers’ sequences

We sequenced 384 candidate polymorphic markers chosen as described in “Methods” and analyzed 353 high-quality sugarcane sequences with a phred quality value greater than or equal to 20, average sequence length of 314 bp, minimum length of 57 bp and maximum length of 797 bp. These sequences were subjected to searches in NCBI nucleotide databases. NBLAST and TBALSTX searches against the nonredundant (nr) and the EST section of GenBank were performed. The results of the searches are summarized in Table 5. The search in the nucleotide nonredundant database returned 25 and 29% similarities found for untranslated and translated sequences, respectively. This proportion went up to more than 40% in both cases, when the search was narrowed to the non-mouse and non-human EST database. It is worth pointing out that approximately 29 and 31% similarities for untranslated and translated searches, respectively, were found for sugarcane EST entries.

Table 5 Results of BLAST searches performed for the sugarcane DArT markers against NCBI nucleotide databases
Full size table

The detailed analysis of sequenced sugarcane DArT markers is in progress, though a brief glimpse of initial results revealed some interesting similarities. For example, for one of the clones a very strong similarity (10−39 ≤ E≤10−70) was found (nr, NBLAST, TBALSTX) with receptor-like kinase (rlk) gene of other Poaceae such as maize, rice, oat, barley and wheat; in the last species the rlk gene is implicated in leaf rust resistance (Feuillet et al. 1997). Strong similarities were also identified with sequences of wild relatives of modern sugarcane cultivars from the so-called “Saccharum complex”: Erianthus arundinaceus Retz. and several representatives of Miscanthus genus. For example, a strong similarity was found to the 18S rRNA gene and a ribosomal DNA external transcribed spacer of E. arundinaceus (E ≤ 10−43 NBLAST, E ≤ 10−48 TBLASTX) and to the 25S rRNA and 17S rRNA genes and an internal transcribed spacer 1 of Miscanthus sinensis Anderss. (10−8 ≤E ≤ 10−10 NBLAST, 10−10 ≤ E≤10−15 TBLASTX).

The redundancy of sequenced markers was analyzed by sequence pairwise comparison. In total, there were 36 redundant markers: one present in five copies, five in three copies and 22 in two copies. Twenty-seven (75%) redundant markers originated from PstI/TaqI/SSH libraries generated from IJ76-514 and Q165, the parents of the mapping population (PstI/TaqI/SSH mapping); six redundant clones came from PstI/TaqI/SSH libraries constructed from 16 sugarcane genotypes representing sugarcane diversity (PstI/TaqI/SSH diversity), and only three were from the PstI/TaqI library constructed from 16 sugarcane genotypes representing the Saccharum diversity (PstI/TaqI diversity). Taking into account the origin of sequenced clones resulted in 15.6, 4.6 and 6.0% redundancy for PstI/TaqI/SSH mapping, PstI/TaqI/SSH diversity and PstI/TaqI diversity libraries, respectively, and 10.2% on average.

Discussion

Diversity Arrays Technology genotyping is a well-established technology for a number of species providing hundreds to thousands of high-quality polymorphic markers. The technology finds an increasingly broad range of applications, for example in whole-genome-based molecular breeding, genetic identity testing, diversity analysis, and molecular systematics of natural populations (www.DiversityArrays.com/publications). We report here on a development of DArT technology for sugarcane. The most critical step in DArT is the method of genome complexity reduction, as it substantially determines the efficiency of polymorphism identification. Capturing the polymorphic fraction of the genome is especially challenging in the polyploid species, where copies of homologous chromosomes “dilute” the polymorphism. Therefore, our primary effort in developing DArT for sugarcane relied on an extensive testing of various methods to reduce genome complexity of a pool of sugarcane accessions representing the Saccharum genetic diversity.

The successful performance of DArT in identifying polymorphism relies largely on high-fidelity restriction enzymes (RE) detecting SNPs in RE recognition sequences (Akbari et al. 2006; Wittenberg et al. 2005). Therefore we evaluated 12 restriction enzymes in 10 various methods for their effectiveness in polymorphism identification. As shown in the “Results” section, the methods based on PstI restriction enzyme as a primary cutter performed well in finding polymorphism in sugarcane, as was the case in other plant species for which DArT genotyping was developed. For example, the PstI/TaqI method resulted in 10.4% polymorphism in barley (Wenzl et al. 2004), 14.6% in cassava (Xia et al. 2005), and 9.4% in wheat (Akbari et al. 2006). The good performance of PstI RE can be attributed to the methylation susceptibility of the enzyme. Its recognition sequence contains two CXG groups (X = A or T) and methylation of cytosine blocks the enzyme activity. Since the repetitive DNA blocks in plant genomes are usually methylated at 5′-CG-3′ and 5′-CXG-3′ cytosine (Gruenbaum et al. 1981), PstI preferentially cuts non-methylated, non-repetitive DNA. Thus, the PstI-based complexity reduction methods filter away repetitive sequences and the markers obtained from such methods are predominantly located in low-copy, gene-rich areas of the genome which are typically hypomethylated or non-methylated at CG and CXG sites (Rabinowicz et al. 2003).

MITE/Bsp1286I was another effective method of reducing genome complexity and identifying polymorphism in sugarcane. This method was also successfully used for DArT genotyping in rice and sorghum (S. Patarapuwadol, personal communication; unpublished data). It is not surprising as MITEs are widespread in plants, small in size but high in copy number and polymorphism, stable and located preferentially in the genic regions of the genomes. These features offer a good basis from which to exploit MITEs as molecular markers and indeed they are becoming popular as useful tools in plant genetics and genomics (Casa et al. 2000; Park et al. 2003; reviewed in Feng 2003). An application of Bsp1286I restriction enzyme increased further the MITE-identified polymorphism, since the enzyme recognition site 5′-GDGCHC-3′ (D = A or G or T; H = A or C or T) allows for a variety of sequence combinations. Three of these combinations were captured by the design of adaptors annealing to the products of Bsp1286I digestion.

With polymorphism frequencies of 7.03 and 9.78% for the two best methods of genome complexity reduction PstI/TaqI and MITE/Bsp1286I, respectively, we considered a polymorphism enrichment technique to facilitate development of a final array comprising 5,000–7,000 polymorphic markers. As ~70 to 80% of the genome of modern sugarcane cultivars originates from S. officinarum with eight copies of the basic chromosome number (x = 10, 2n = 8x = 80) (D’Hont et al. 1996), the approach was to remove this repetitive part of the genome by means of subtraction, thus revealing more of the genome polymorphism originating from S. spontaneum. We decided on Suppression Subtractive Hybridization (SSH), a PCR-based DNA subtraction method originally developed to generate differentially regulated or tissue-specific cDNA libraries (Diatchenko et al. 1996). We used the same principle, which was to equalize the quantities of DNA fragments by hybridization kinetics and subsequent selective amplification of target DNA fragments. Although the average polymorphism frequency for all subtraction variants tested was only slightly better than for the PstI/TaqI method without subtraction, twofold enrichment in the number of polymorphic markers was achieved in one of the subtraction variants. Thus, SSH seemed to be a promising technique for polymorphism enrichment in sugarcane. However, a redundancy of markers has to be taken into consideration before committing to this technique as a method of choice for building the final array (discussed below).

The subtraction showed an additional advantage apart from improving the polymorphism frequency. The subtracted PstI/TaqI clones were better than PstI/TaqI clones without subtraction in determining the genetic relationships between surveyed sugarcane accessions as revealed by Factorial Analysis of Correspondence. This result was further confirmed when the PstI/TaqI/SSH markers obtained in the independent experiment, as well as the MITE/Bsp1286I markers from the second best method of genome complexity reduction, were used to draw the UPGMA dendrograms. The genetic relationships revealed among sixteen genotypes based on these two sets of markers are consistent with the previous findings for sugarcane genetic relationships obtained with other molecular markers and pedigree information. Assessments of genetic diversity and similarity of various numbers of sugarcane accessions were performed with molecular markers such as rDNAs (Glaszmann et al. 1990), RFLPs (Jannoo et al. 1999b) and AFLPs (Besse et al. 1998; Lima et al. 2002; Aitken et al. 2006). The findings of these studies demonstrated a strong genetic differentiation between S. officinarum and S. spontaneum and both ancestral species from the modern cultivars, just as was shown with both types of the best sugarcane DArT markers.

The segregation ratios of sugarcane DArT markers demonstrated that, in spite of high polyploidy and the bi-specific nature of the sugarcane genome, the majority of DArT markers segregated in the Mendelian fashion. The estimated proportion of single-dose markers obtained for the IJ76-514 × Q165 mapping population in this study is similar to the proportions reported for the same, as well as other, sugarcane mapping populations. There were 71% of polymorphisms detected with a combination of AFLPs, SSRs and RAFs markers presenting as single-dose markers in the IJ76-514 × Q165 mapping population (Aitken et al. 2005), while 73 and 79% were estimated, respectively, for a mapping population derived from a cross between two pre-commercial cultivars, using a combination of RFLPs, SSRs and AFLPs (Garcia et al. 2006) and the selfed progeny of R570 cultivar with AFLP markers (Hoarau et al. 2001).

From the markers scored differently between the parents, the majority of scoring assignments “1” originated from modern cultivar Q165 parent. The bi-specific nature of modern sugarcane cultivars may be the contributing factor to this finding. The higher redundancy levels of the S. officinarum part of the genome (Aitken et al. 2005) and known lower diversity of this species (Aitken et al. 2006) could have contributed to the under-representation of S. officinarum polymorphic markers. Thus, the majority of markers identified from Q165 seemed to originate from the S. spontaneum part of the genome due to its known more polymorphic nature. Better coverage of S. spontaneum than of S. officinarum chromosomes is a typical feature of sugarcane mapping (Grivet and Arruda 2001).

In order to obtain extensive coverage of each chromosome for a species with such a large and complex genome, higher numbers of single-dose markers would be required. The dominance of single-dose markers is a desirable characteristic in out-crossing polyploid species such as sugarcane, since alleles present as a single dose are much more informative in the construction of genetic maps and marker-assisted selection (MAS) than any others. Single-dose markers in the coupling phase are used to construct the map and in the repulsion phase to determine chromosome assortment. Duplex or multiplex markers can only be incorporated into the framework map constructed based on single-dose markers. Therefore, experiments aiming to enrich for single-dose sugarcane DArT markers are well advanced. Incorporating the third frequently cutting restriction enzyme BstNI or HaeIII into the PstI/TaqI method of genome complexity reduction resulted in a 25% increase in the frequency of single-dose markers (data not presented). While the detailed genetic map of the Q165 cultivar comprising the single-dose DArT markers resulting from the described enrichment will be published separately, Fig. 4 shows an example of incorporation of these DArT markers into the Q165 linkage map generated by Aitken et al. (2005). The majority of the markers mapped into the existing linkage groups, indicating that they map to similar regions of the genome as AFLP and SSR markers.

The sequence analysis of sugarcane DArT markers revealed some interesting findings. Close to one-third of these markers were found to be similar to sugarcane EST sequences originating from the transcribed portion of the genome. As these markers are derived from expressed genes, they may facilitate functional genomics studies and identification of the genetic basis of desirable traits (Varshney et al. 2005). It is especially of interest in sugarcane, the species where until 2006 only three major genes were identified: two rust resistance genes and a gene controlling stalk color (Asnaghi et al. 2004; Raboin et al. 2006). The similarities found for some DArT markers and entries for Erianthus sp. and Miscanthus sp. also deserve highlighting. Both plants are related genera within the Sacchararinea tribe, though it was recently suggested that lineages leading to the genus Saccharum were independent from those leading to genera Erianthus and Miscanthus (Grivet et al. 2004). However, there is considerable interest in incorporating E. arundinaceus Retz., characterized by high fiber content, vigor, and resistance to disease and environmental stresses, into sugarcane breeding programs, as fertile hybrids of E. arundinaceus and sugarcane have been established (Cai et al. 2005). Thus DArT markers bearing this similarity could be of use in assessing the phylogenetic relationship of sugarcane and related species and in facilitating marker-assisted incorporation of parts of the E. arundinaceus genome into modern sugarcane cultivars.

There was 10.2% redundancy observed for sequenced sugarcane markers contributed mainly (75%) by clones originating from two subtracted parents of the mapping population. This finding may indicate that the SSH-mediated polymorphism enrichment would be better suited for the more diverse genetic pool. Indeed, the subtracted clones coming from 16 sugarcane genotypes contributed only 17% of redundant markers. Overall, the redundancy level found for sugarcane was significantly lower than the redundancy noted for barley and Arabidopsis thaliana L. DArT markers (Wenzl et al. 2004; Wittenberg et al. 2005) and almost identical to that recorded for wheat DArT markers (Akbari et al. 2006). The redundancy estimation in barley and wheat was carried out on the mapping data based on identity of marker scores, which could lead to an overestimation since the closely linked loci would score the same but could still be different sequences. The more accurate redundancy assessment based on the sequence analysis carried out for A. thaliana revealed approximately 1.78-fold redundancy (42%) (Wittenberg et al. 2005). The difference observed for Arabidopsis and sugarcane could be attributed to the extreme differences in these species genome sizes, ploidy levels and the use of a single accession for construction of the A. thaliana array.

Sequencing of all DArT markers discovered (ongoing) will offer not only an ability to remove the excessive redundancy from the final genotyping array, but will also facilitate relating the results of sugarcane genetic analysis to the genome sequence resources of sorghum. The sorghum genome sequencing project has been completed (Paterson et al. 2009), and therefore the resulting sequence assembly will provide a useful framework for sorghum and sugarcane genetics, as the two crops have highly related genomes (Guimarães et al. 1997). A recent development of DArT for sorghum (Mace et al. 2008) and sequencing of over 2,000 sorghum markers (Bouchet, personal communication) will further facilitate the transfer of genetic and genomic information between these two important crops. However, any strongly reductionist attempts to “translate” sorghum genome information into genetics of sugarcane are likely to fail, as the two organisms vary greatly in their ploidy levels, genome stability and level of DNA sequence variation within genotypes and breeding populations. Indeed, each sugarcane cultivar could be viewed genetically as a “bulk” with many alleles representing a substantial fraction of total diversity segregating in breeding populations. In this context it is worth noting that the DArT platform has been deployed successfully for quantitative Bulked Segregant Analysis (qBSA) in barley and that the relative signal intensity of the representations prepared from bulks was shown to monitor precisely allele frequencies in complex samples (Wenzl et al. 2007). It is our view that highly reproducible sugarcane DArT assays open the possibility of using relative signal intensity of DArT arrays (in addition to binary scored markers) in allele frequency estimation and effective marker–trait association analysis in sugarcane. We will present results in support of this notion in our next paper.

Conclusions

The presented work demonstrates that DArT genotyping effectively discovers and scores hundreds of polymorphisms in the large, aneupolyploid genome of sugarcane. Sugarcane DArT fingerprints can be used for both investigation of genetic relationships and construction of genetic maps. An expansion of libraries generated with the best genome complexity reduction methods will enable us to construct a “sugarcane DNA chip” capable of providing genome-wide scans with many thousands of markers for hundreds of genotypes timely and cost-effectively. This could open up a gateway for whole-genome—profile-assisted breeding programs of sugarcane, providing superior varieties in a much shorter time than via the currently used conventional breeding methods. The methods of genome complexity reduction utilizing the restriction enzymes PstI/TaqI and incorporating the third frequently cutting RE either BstNI or HaeIII, due to their best performance, became our method of choice to generate genomic representations for routine sugarcane genotyping.

This new capability of utilizing DArT-based whole genome information in practical breeding is in fact already at the early stages of implementation by the Australian sugarcane breeding program. While it will take a few years before the anticipated advantages of genome-based selections can be validated, our modeling based on large volume of phenotypic and DArT marker data suggests that sugarcane is well positioned to capture the value of genome profiling technology presented in this paper.