The type, frequency and relative distribution of the microsatellites in the unigene sequences of sugarcane are given in Table 1. The perfect microsatellite (excluding the mononucleotides) frequency in the unigenes of sugarcane was one in every 10.9 kb and the proportion of microsatellite carrying unigenes was 3.7% (584 out of 15,582). When 1,871 (12%) mononucleotide microsatellites were included, the proportion of unigenes carrying microsatellites increased to 17.4%. The mononucleotides in sugarcane showed a strong bias (84.6%) towards A/T repeat, with the majority (89%) being 9 to 30 bases long and the remaining (11%) extending up to 69 bases (T₆₉). A total of 841 perfect microsatellites were identified in 584 unigene sequences of sugarcane. One hundred sixty-seven (28.6%) of these unigenes contained multiple microsatellites (that accounted for 424 microsatellites) which were interrupted by more than 100 nucleotides and the remaining 417 (71.4%) unigenes had a single microsatellite each. In addition to the perfect microsatellites, we identified 183 compound microsatellites, of which 74.9% were interrupting types and the rest being non-interrupting types (Table 1). The trinucleotide repeat motifs were the most prevalent (73.1%) class of microsatellites followed by dinucleotide (23.8%), tetranucleotide (1.8%), pentanucleotide (0.83%) and hexanucleotide (0.47%) repeat motifs (Table 1). The GC rich repeat motifs GCA/GCC/GCG/GCT (23%) coding for alanine were most abundant followed by arginine (AGA/AGG/CGA/CGC/CGG/CGT, 22%) and glycine (GGA/GGC/GGT, 11%) (see Additional file 1). Among the dinucleotide repeats, the most and least frequent motifs were GA and CG, which accounted for 21% and 1.4% of all the microsatellites, respectively (see Additional file 2).

Out of 841 perfect microsatellites identified in sugarcane, 587 (69.8%) were found in the ORFs and the remaining were present either in the 3'UTRs (102, 12.1%) or in the 5'UTRs (152, 18.1%). The trinucleotide repeat motifs were significantly more frequent (about 86%) in the ORFs. In contrast, the GA-rich dinucleotide repeat motifs were more in the 5' (49%) and 3' (32%) UTRs. The density of longer motif containing perfect class I microsatellites was one in every 44.3 kb sequences, which accounted for 24.6% (207) of the total 841 microsatellites identified (Table 1, Figure 1).

The primer pairs could be designed for 810 perfect microsatellites that was 96.3% of the total microsatellites (841) identified in the present investigation. The primer sequences flanking all the perfect UGMS including 207 class I microsatellites with their Tm values and product sizes are given in the Additional file 3. Besides, the primer sequences for 151 compound class I UGMS were designed and provided in the Additional file 4. To validate the UGMS markers, 176 primer-pairs designed from different microsatellite containing unigenes were used in PCR amplification (see Additional file 3). One hundred sixty seven (94.9%) of these produced clear and reproducible amplicons, whereas remaining nine (5.1%) did not give amplification in the S. officinarum from which sequence the primers were designed. To verify that the primers did amplify the expected microsatellite repeat-motifs, the amplified products obtained with 19 of the primers in all the Saccharum species and related genera as well as cereals were sequenced and the presence of the target microsatellite motifs as well the flanking sequences was observed in all the cases (see Additional file 5).

In silico polymorphism analysis was confined to Saccharum officinarum and five cereal species namely, rice, wheat, maize, Sorghum and barley for which unigene sequences were available in the database. The polymorphism based on variation in microsatellite repeat length was observed for a maximum of 163 primers (46.6%) between sugarcane and barley followed by 161 (42.5%) between sugarcane and wheat and least (92, 17.8%) between sugarcane and Sorghum (see Additional file 6). The actual level of polymorphism detected by automated fragment analysis using 47 fluorescent dye labeled primers was much higher than that based on in silico analysis although the trend was maintained. S. officinarum had maximum polymorphism with barley (92.7%) followed by wheat (90.6%), rice (85.8%), maize (67.7%) and Sorghum (61.4%). Forty-three (91.5%) of the 47 primers detected polymorphism (mean PIC of 0.85) among the 41 genotypes belonging to Saccharum species, related genera, tropical and sub-tropical Indian sugarcane varieties and five cereal species (Table 2). This included 21 (95.5%, mean PIC of 0.81) amplifying dinucleotide repeats, 16 (84.2%, PIC of 0.73) trinucleotide repeats, two (100%, PIC of 0.87) tetranucleotide repeats, two (100%, PIC of 0.86) pentanucleotide repeats and two (100%, PIC of 0.83) amplifying hexanucleotide repeats. All the 38 (100%, mean PIC of 0.82) primers that targeted amplification of class I repeats and five (55.6%, PIC of 0.55) of the nine that targeted class II repeats showed polymorphism among Saccharum species, genera, varieties and cereal species. The microsatellites in the 5' and 3'UTR sequences showed higher potential for polymorphism (15 out of 16, 94%, mean PIC of 0.83) as compared to that from the coding regions (28 out of 31, 90%, mean PIC of 0.77).

Due to high polyploidy and heterozygosity of the Saccharum genome, most (28, 65.1%) of the primers amplified two to three different loci (each locus being designated as an UGMS marker) yielding a total of 60 loci with the number of alleles per locus ranging from 2 to 22 and the allele size varying from 82 to 408 bp across sugarcane species, genera and varieties (Figure 2) as well as five cereal species. The remaining 15 (34.9%) primers gave single locus amplification (Table 2) with 1 to 23 alleles each. Overall, the 43 polymorphic primers amplified an average of 7.42 alleles per UGMS marker locus with a total of 722 alleles across 74 loci. Forty-two (73 UGMS marker loci with 720 alleles, 97.7% polymorphic, mean PIC of 0.81) of the 43 informative primers revealed polymorphism between sugarcane species and related genera, whereas 37 (66 UGMS marker loci with 678 alleles, 86% polymorphic, PIC of 0.74) primers detected polymorphism among the Indian commercial sugarcane varieties (Figure 3). Twenty-six (34.7%) of the 75 UGMS marker loci showed heterozygosity in different genotypes used. Maximum heterozygous loci was detected in five sugarcane species (14, 18.7%) followed by commercial varieties (9, 12%) and minimum (4, 5.3%) in the three related genera and five cereal species.

Comparison of fragment size (bp) of variant alleles amplified at 75 polymorphic UGMS loci with changes in number of repeats among Saccharum complex, varieties and five cereal species revealed purely step-wise allelic distribution for 51 (68%) loci, while remaining 24 (32%) loci showed mixed type of allele size distribution (Figure 2). To elucidate the distribution pattern of UGMS alleles showing fragment length polymorphism, 10 amplified size variant alleles each of two microsatellite primers namely, UGSuM2 and UGSuM27 showing both step-wise and mixed type of allele distribution were sequenced. High quality sequence alignment revealed that the size variant alleles showing step-wise distribution contained the expected microsatellite repeat sequences with conserved primer binding sites, but corresponded exactly to the step-wise multiples of the number of repeat units. The size variation of sequenced alleles was explained by differences in the number of repeat units. Mixed type of allele distribution resulted from both variation in the number of repeat-units and insertions/deletions in the sequences flanking the UGMS repeats. For instance, UGSuM2 (Figure 4A) and UGSuM27 (Figure 4B) showing both step-wise and mixed type allele size distribution had the expected (AT)_n and (GGC)_n UGMS motifs, respectively with varying number of repeat-units in different sugarcane species, genera, varieties and cereals along with a small stretch of nucleotide insertion/deletion in the flanking regions of target microsatellites.

In order to assess the functional significance of the UGMS, gene ontology classification of 767 unigenes carrying perfect and compound microsatellites was carried out. Six hundred seventy-two (87.6%) of these (see Additional files 3 and 4) could be functionally annotated and were shown to encode enzymes for sugar metabolism (39%), structural proteins (26%), transcription and translation factors (18%), signal transduction pathway proteins (8%), cell growth and development factors (5%) and disease resistance proteins (4%) (see Additional file 7). Three hundred sixty-four (54.2%) of 672 unigenes contained microsatellite repeats in 89 different functional domains of proteins encoded by these unigenes (see Additional file 8) which included cytochrome b/c oxidase, chlorophyll A/B binding, FAD and ubiquitin binding, sucrose synthase, alpha-amylase, acetyl COA, aldehyde dehydrogenase, glyceraldehyde-3-phosphate, lipase, S-adenosylmethionine, leucine rich repeat (LRR), ferritin, protein kinase, chitinase, basic leucine zipper (bZIP), zinc finger, TATA box, Myb and WRKY DNA binding domains (Figure 5). Fifteen primers designed from 15 different unigenes targeting such functional domains (see Additional file 9) that gave single locus amplification and step-wise allele distribution (Table 2) were selected for validation. All of these primers gave fragment length polymorphism among Saccharum complex, varieties and five cereals based on variation in the number of UGMS repeat-units within the functional domains.

One of these primers (UGSuM26) showing polymorphism targeted the amylase catalytic domain of β-amylase (Figure 6). To understand the possible biological significance of the variable UGMS repeat in the amylase domain, the structure predicted from the aminoacid sequences of the functional domain was analyzed in silico. It revealed variation in the active binding site involved in formation of the Ca²⁺ ligand complexes due to the presence of variable number of repeats encoding Leucine-Serine aminoacid residues (Figure 6). This would most likely alter the function of amylase gene in respect of carbohydrate metabolism in sugarcane. Similarly, the expansion/contraction of microsatellite motifs (AG)_n encoding Arginine-Glutamine aminoacid residues was observed in the protein kinase functional domain of another sugarcane unigene which was amplified by the primer UGSuM17. In silico analysis suggested a novel secondary protein structure and catalytic domain binding site in the variant form that was different from the native form which contained (AG)₁₈ microsatellite repeats (see Additional file 10).

The pair-wise similarity among 36 genotypes belonging to Saccharum species, related genera and 28 tropical and sub-tropical Indian varieties of sugarcane based on 43 fluorescent dye labeled polymorphic primers revealed a broad range from 0.12 to 0.91 with an average of 0.32 similarity index. The similarity among the sugarcane species varied from 0.24 (IJ-76-3-1-9 and 1151) to 0.88 (IJ-76-3-1-9 and IM-76-256) with an average of 0.36. Among the species, S. officinarum showed maximum similarity (0.88) with S. robustum followed by S. barberi with S. sinense (similarity of 0.82). S. spontaneum was most divergent from the rest of the Saccharum species. The similarity of the three genera of sugarcane, namely, Erianthus, Narenga and Sclerostachya with the sugarcane species varied from 0.22 to 0.90. Erianthus had the least similarity with the Saccharum species. Similarity among the 28 Indian tropical and sub-tropical sugarcane varieties varied from 0.33 (Bo 91 and Co 8021) to 0.84 (Co 8371 and Co 8021) with an average of 0.40. The pair-wise similarity between the sugarcane species and varieties varied from 0.20 to 0.78 with average of 0.35, while between the three related genera and varieties, it ranged from 0.13 to 0.80 with an average of 0.24. The average similarity between the sub-tropical and tropical varieties (0.40) was slightly lower than the average pair-wise similarity measures within the sub-tropical (0.47) and tropical varietal groups (0.43).

The genetic relationship among the Saccharum species, related genera, and tropical and sub-tropical varieties is depicted in Figure 7. The UGMS markers clearly discriminated all the 36 genotypes from each other and resulted in a definitive grouping among different genera and species of Saccharum with high bootstrap values (81 to 100) that corresponded well with their known pedigree relationships. All the clones of five Saccharum species were included in the major cluster I in which S. officinarum and S. robustum clones were sub-clustered (Ia) with 98% occurrence in boot strap analysis; S. barberi and S. sinense clones grouped together in a separate sub-cluster (Ib) with 100% occurrence and S. spontaneum grouped distinctly from these sub-clusters with 92% occurrence. The three related genera of Saccharum being highly divergent from the Saccharum species and varieties, grouped in a separate cluster (III) with 98% occurrence.

All the tropical and sub-tropical varieties were included in a distinct and separate cluster (II) with seven distinct sub-clusters (Figure 7) with high (88) bootstrap value. The tropical variety Co 62175 grouped in a major sub-cluster IIa with its tropical male parent Co 419 (similarity 0.71) with a bootstrap value of 99%. In this sub-cluster (IIa), the other tropical varieties namely, Co 8021, Co 8371, Co 7704 and Co 6304 were also grouped together with average similarity of 0.77 (supported by 93% occurrences in bootstrap) possibly because of their common ancestry involving Co 419. The two sub-tropical varieties, Co 8347 and Co 87263, and three other sub-tropical varieties, CoPant 84211, Co 89003 and Co 7717, in spite of having a common ancestor Co 775, were included in two distantly placed major sub-clusters; IIb and IIg, respectively. It could be due to inclusion of two diverse sub-tropical varieties, Co 87268 and CoLk 8102 having common parent Black Cheribon within the sub-cluster IIb. Similarly, the clustering of two tropical varieties namely, CoC 671 and Co 87025 in the sub-cluster IIc with average similarity of 0.69 (supported by 100% occurrences) and five tropical varieties, Co 7219, Co 740, Co 86010, Co 86032 and Co 85002 together in another separate large sub-cluster (IIf) with average similarity of 0.64 (supported by 92% occurrences) was observed even though they had one common distant progenitor parent Black Cheribon. The grouping of sub-tropical varieties namely, Co 1148, CoS 8436 and CoPant 84212 in cluster IId with an average similarity of 0.70 (88% occurrences in bootstrap) was clearly influenced by a common ancestry involving Co 1148. The sub-tropical varieties, CoJ 64, Co 1158 and CoS 88230 included in the cluster IIe had common progenitor parent PoJ 2878 that could have influenced their clustering with a high bootstrap value (94.3%).

Fifteen thousand five hundred ninty-four sugarcane (Saccharum sp.) unigenes (Build 13.0, 19^th Feb' 2008) comprising of 9.17 million base sequences were acquired in FASTA format from the recent NCBI FTP UniGene repository database of S. officinarum [32] in batches and searched for microsatellites using a perl scripting language based program MISA (MIcroSAtellite) [33] considering complementary sequence of repeat-motifs (like AG, GA, TC and CT) in the same class [18]. The identified microsatellites were characterized as perfect (monomers to hexamers repeated up to 100 times without any interruption), compound (non-interrupting) with at least two different repeat-motifs without any interruption, compound (interrupting) with a maximum of 100 nucleotides interposing two microsatellite repeat-motifs, class I (≥20 nucleotides) and class II (12 to 20 nucleotides) types. The microsatellite (excluding monomers) containing unigenes were used in a batch module for designing forward and reverse primers employing the microsatellite primer discovery tool of BatchPrimer3 [34]. All default options except for the optimum and maximum primer sizes of 22 and 24 nucleotides, respectively, were used. The putative functions for these unigenes were determined using NCBI BLASTX search tool [35] as against the nr-protein database. The BLASTX output was annotated into four functional categories viz, exact, putative, unknown and hypothetical, and then extracted to Excel sheets. Different structural components namely, coding and untranslated regions (5'UTRs and 3'UTRs) in the unigene sequences was determined using the software tools FGENESH [36] and UTRScan [37], respectively. The aminoacid sequences encoded by the predicted coding nucleotide regions of the microsatellite carrying unigenes were analyzed using the software Pfam [38] to determine the presence of functional domain/protein family within the UGMS. The aminoacid sequences of the functional domain carrying microsatellites was analyzed further using the I-TASSER automated web server [39,40] for prediction of three dimensional protein structure and active binding sites with ligands. Five different protein models and active binding sites were predicted at significant cut-off confidence (C ≥ -1.5) and binding site (BS ≥ 0.5) scores. The high quality protein model of correct topology and protein-ligand complex active binding sites was identified based on high C- and BS- scores.

To assess the amplification success rate of microsatellite markers designed from sugarcane unigenes, 176 primers were used to amplify one genotype of S. officinarum. Forty-seven, including thirty-eight class I and nine class II markers of these, labeled with 6-carboxyflourescein (6-FAM) dye phosphoramidites, were used to genotype one accession each belonging to five Saccharum species and three related genera, and twenty-eight commercial Indian tropical and sub-tropical sugarcane varieties (Table 3) for evaluating their polymorphic potential and assessment of genetic diversity. Five cereal species namely, rice, wheat, maize, Sorghum and barley were included in the experiment for comparison. Standard PCR constituents and cycling conditions except for annealing temperature, which varied from 55°C to 64°C depending on the primers, were used for PCR amplification. The amplified products were purified, mixed with 3.75 μl of MegaBACE formamide loading buffer and 0.25 μl of internal-lane MegaBACE™ET 550-R ROX size standard, denatured, cooled and resolved in automated MegaBACE 1000 DNA sequencer (Amersham Biosciences, Piscataway NJ, USA). The electropherogram containing trace files were analyzed and automated allele calling was carried out using the Binning Peak Post Processor tool of MegaBACE™Fragment Profiler Software Version 1.2 (Amersham Biosciences, Piscataway NJ, USA). The actual allele size (bp) was determined and fragment length polymorphism among the sugarcane genotypes and cereal species was identified. The average peak height and peak quality of alleles generated for all the UGMS loci were graphed and alleles exhibiting average peak height and quality of ≥ 1500 and ≥ 6.0 fluorescence units [41,42] respectively were considered for sizing and base-pair estimation. Allele binning was performed for individual UGMS loci for precise and accurate allele size determination and discrimination of homozygous and heterozygous allele types [42,43] for each genotype.

Variation in the fragment size (bp) of amplified alleles at each polymorphic UGMS marker locus was compared with changes in the number of microsatellite repeat-units at that target locus, and the occurrence of "stepwise" and "mixed" type of allele size distribution was inferred. When the allele size differences strictly corresponded to the variation in the number of repeat-units, it was considered as stepwise distribution. A mixed allele distribution was assumed when the allele size differences could partly be explained by the stepwise model. Multiple amplicons obtained by a primer-pair with peaks ≥ 1500 fluorescence units showing ≥ 100 bp allele size differences were binned into different loci. To confirm that the primers amplified the target microsatellite repeat motifs in different species and genera, the amplified products were purified using Micropon PCR purification kits (Millipore, Bedford, MA, USA) and sequenced two times in both forward and reverse directions using a capillary-based Automated DNA Sequencer (MegaBACE 1000, Amersham Biosciences, Piscataway NJ, USA). The trace files were base called, checked for quality and assembled into contigs [44]. The high quality sequences thus obtained were used for interspecies comparison using CLUSTALW multiple sequence alignment tool employing BIOEDIT software [45]. Ten size variant amplicons each of three microsatellite markers (UGSuM2, UGSuM26 and UGSuM27) showing both step-wise and mixed type of allele size distribution in sugarcane species, genera, varieties and cereals were eluted, purified, cloned in pGEM-T Easy Vector (Promega, USA) and sequenced as described above.

The polymorphic information content (PIC) was calculated using the formula, PIC = 1- ∑P_ij² [46], where P_ij is the frequency of the j^th allele for i^th locus summed across all alleles for the locus. Cluster analysis among the 36 genotypes of Saccharum species, varieties and related genera was based on Nei and Li similarity coefficient [47] using the un-weighted pair group method (UPGMA) in PowerMarker Version 3.0 [48,49] software. The confidence limits of UPGMA based dendrogram was determined by 1000 bootstrap replicates and bootstrap of 50% majority rule consensus unrooted phylogenetic tree was constructed.