Rapid microsatellite development for tree peony and its implications

  • Zhimin Gao2Email author,

    Affiliated with

    • Jie Wu1, 3,

      Affiliated with

      • Zheng’an Liu1,

        Affiliated with

        • Liangsheng Wang1,

          Affiliated with

          • Hongxu Ren1 and

            Affiliated with

            • Qingyan Shu1Email author

              Affiliated with

              BMC Genomics201314:886

              DOI: 10.1186/1471-2164-14-886

              Received: 3 June 2013

              Accepted: 4 December 2013

              Published: 16 December 2013

              Abstract

              Background

              Microsatellites are ubiquitous in genomes of various organisms. With the realization that they play roles in developmental and physiological processes, rather than exist as ‘junk’ DNA, microsatellites are receiving increasing attention. Next-generation sequencing allows acquisition of large-scale microsatellite information, and is especially useful for plants without reference genome sequences.

              Results

              In this study, enriched DNA libraries of tree peony, a well-known ornamental woody shrub, were used for high-throughput microsatellite development by 454 GS-FLX Titanium pyrosequencing. We obtained 675,221 reads with an average length of 356 bp. The total size of examined sequences was 240,672,018 bp, from which 237,134 SSRs were identified. Of these sequences, 164,043 contained SSRs, with 27% featuring more than one SSR. Interestingly, a high proportion of SSRs (43%) were present in compound formation. SSRs with repeat motifs of 1–4 bp (mono-, di-, tri-, and tetra-nucleotide repeats) accounted for 99.8% of SSRs. Di-nucleotide repeats were the most abundant. As in most plants, the predominant motif in tree peony was (A/T)n, with (G/C)n less common. The lengths of SSRs were classified into 11 groups. The shortest SSRs (10 bp) represented 1% of the total number, whereas SSRs 21–30 and 101–110 bp long accounted for 26% and 29%, respectively, of all SSRs. Many sequences (42,111) were mapped to CDS (coding domain sequence) regions using Arabidopsis as a reference. GO annotation analysis predicted that CDSs with SSRs performed various functions associated with cellular components, molecular functions, and biological processes. Of 100 validated primer pairs, 24 were selected for polymorphism analysis among 23 genotypes; cluster analysis of the resulting data grouped genotypes according to known relationships, confirming the usefulness of the developed SSR markers.

              Conclusions

              The results of our large-scale SSR marker development using tree peony are valuable for investigating plant genomic structural evolution and elucidating phenotypic variation in this species during its evolution and artificial selection. The newly identified SSRs should be useful for genetic linkage map construction, QTL mapping, gene location and cloning, and molecular marker-assisted breeding. In addition, the genome-wide marker resources generated in this study should aid genomic studies of tree peony and related species.

              Keywords

              zzzMicrosatellite Next-generation sequencing Tree peozzzzny Ornamental SSR marker

              Background

              Microsatellites, or simple sequence repeats (SSRs), are tandemly repeated 1–6-bp DNA regions ubiquitous in prokaryotes and eukaryotes. As components of genomes, they can be found both in protein-coding and non-coding regions. SSRs have been universally utilized as genetic markers because of their abundance and inherent potential for variation [1]. The functions of SSRs were previously unclear, and until recently they were regarded as ‘junk’ (i.e., having no significant genomic role). At present, much progress has been achieved in regard to elucidation of SSR function. SSR locations appear to determine the types of functional roles that SSRs play, and alterations in SSR lengths at different locations can lead to changes in organismal phenotypes [2, 3]. SSRs in different gene positions (i.e., promoter regions, 5′ untranslated regions (UTRs), 3′ UTRs, exons, and introns) may play important roles in determining protein function, genetic development, and regulation of gene expression. For example, expansion of CAG repeats in the HD gene coding region can lead to Huntington’s disease in humans, possibly through activation of some so-called ‘toxic’ proteins [1]. With expanding knowledge of SSR functions in terms of development, gene regulation, and evolution, SSRs are receiving increasing attention. Because genomic information is lacking for most species, however, it is difficult to study microsatellite origin, distribution, and evolution, or even to develop new SSR-based molecular markers.

              Traditional SSR development is time-consuming, and involves laborious iterations of genomic DNA library screening with SSR probes required to isolate microsatellite-containing sequences [4]. Next-generation sequencing technologies are remarkably well-developed, and are widely used for genome sequencing, transcriptome sequencing, and genome deep-sequencing in plants [5, 6]. It has been successfully used for identifying molecular markers, including SSRs and simple nucleotide polymorphisms (SNPs), in organisms such as the water strider [7], copperhead snake [8], blue duck [9], pine pathogen fungus [10], and scuttle fly [11]. Because of the complicated structure of plant genomes, however, molecular marker development using next-generation sequencing has had limited application, especially in non-model plants lacking genomic information.

              Among next generation-sequencing approaches, Roche 454 pyrosequencing (R454) holds great promise with respect to the long reads obtained as well as acquisition of sufficient genetic information of interest within single reads. The large amount of generated data facilitates sequence assembly without genomic information [5], and increases the likelihood that a single read contains microsatellite repeats along with suitable flanking regions of unique sequences. Another approach to mining molecular markers, involving in silico methods, has also been successful; examples include the derivation of markers from a draft genome [12] and the mining of existing expressed sequence tag (EST) libraries [13]. Compared with traditional library-based and in silico methods, R454 offers great advantages, being faster, less costly, and less dependent on existing genetic resources [14]. Another advantage is the huge amount of genetic information produced, with the possibility of future use. This is greatly beneficial for studies of plants without genomic information, especially woody plants; in such species, no established systems exist for in vitro culture or transformation for genetic manipulation, hampering new cultivar breeding. Molecular marker-assisted breeding is efficient for such organisms. DNA markers developed via next-generation sequencing are also increasingly being used for genetic diagnostics, drug discovery, gene cloning, genome analysis, comparative genomics, and molecular evolution studies.

              The purpose of this study was to apply next-generation sequencing, such as R454, to SSR development in tree peony (Paeonia suffruticosa Andrews). Such an approach was expected to drastically shorten the time required for effective marker development and utilization. Tree peony belongs to sect. Moutan DC. of the genus Paeonia L. (Paeoniaceae). It is a well-known ornamental plant enjoying worldwide popularity on account of its large, showy, colorful and fragrant flowers. Little genomic information is currently available for this species. In a previous study, we constructed a cDNA library from flower buds and obtained 2,241 ESTs, from which 167 SSRs were derived and a dataset of 185 putative SNPs obtained for breeding based on their high availability and stability [15]. Although more than 200 SSRs have been submitted to public databases [16], the number is inadequate with respect to the 1,500 cultivars of tree peony. Compared with crop plants such as maize, wheat, and soybean, or ornamental plants such as rose, molecular markers, especially SSRs, are still needed for future breeding of tree peony. This is especially true taking into consideration its importance, barely transformed nature, and woody characteristics, and the lack of genomic information. Construction of a dense genetic linkage map and development of genome-wide molecular markers are also essential for marker-assisted selection of new tree peony hybrids. Because relationships among wild species of tree peony and their taxonomic position within Paeoniaceae are still unclear, developed polymorphic SSRs would also aid studies of Paeonia evolution, comparative genetics, and population structure.

              Results

              Sequencing and characterization of reads

              R454 sequencing of the tree peony library generated 675,221 reads averaging 356 bp long, with a maximum length of 590 bp (Figure 1). Among these nucleotides, adenine was the most abundant (30.4%), followed by cytosine (26.7%), thymine (23.5%), and guanine (19.4%). G + C content was 46.1%. Clean reads were deposited in the NCBI public database (Accession number: SRA098186).
              http://static-content.springer.com/image/art%3A10.1186%2F1471-2164-14-886/MediaObjects/12864_2013_5597_Fig1_HTML.jpg
              Figure 1

              Length distribution of 454 sequencing reads. X- and y-axes refer respectively to sequence length in bp and the number of sequences of a given length.

              Identification of SSR loci

              MISA was used to analyze a total of 240,672,018 bp of sequences, from which 237,134 SSRs were identified. Of the examined sequences, 164,043 contained SSRs; 27% harbored more than one SSR, with a high proportion of SSRs (43%) present in compound formation (Table 1). The distribution of identified SSR motifs in the cloned sequences was nearly evenly divided between the 400-bp 5′-terminus region and the remaining region outside the 5′-terminus (Figure 2). SSRs with repeat motifs of 1 to 4 bp (mono-, di-, tri-, and tetra-nucleotides) accounted for 99.8% of the total, with di-nucleotide repeats the most abundant (Table 2; Figure 3). SSRs with mono-nucleotide repeats accounted for only 2% of SSRs in tree peony DNA. Proportions of tri-nucleotide and tetra-nucleotide repeats were almost equal, with the combined number of tetra-, penta-, and hexa-nucleotide repeats accounting for at most 8.2% of SSRs.
              Table 1

              Occurrence of microsatellites in the surveyed tree peony genome

              Category

              Number

              Total number of sequences examined

              675221

              Total size of examined sequences (bp)

              240672018

              Total number of identified SSRs

              237134

              Number of SSR-containing sequences

              164043

              Number of sequences containing more than 1 SSR

              44362

              Number of SSRs present in compound formation

              70570

              http://static-content.springer.com/image/art%3A10.1186%2F1471-2164-14-886/MediaObjects/12864_2013_5597_Fig2_HTML.jpg
              Figure 2

              Distribution of SSR start positions from the 5′-terminus of the cloned library insert. The x-axis indicates the number of bp from the 5′ terminus of a sequence to the SSR start site. The y-axis corresponds to the number of SSRs beginning at that start position.

              Table 2

              Microsatellite motif length distribution

              Plant species

              Number of motif repeats

              Total

              Percent

              Mono

              Di-

              Tri-

              Tetra-

              Penta-

              Hexa-

              B. distachyon

              30,573

              9,407

              10625

              990

              196

              84

              51875

              2.45

              S. bicolor

              55,906

              38,138

              28480

              5368

              946

              726

              129564

              5.4

              O. sativa

              64,734

              37,282

              29189

              2565

              604

              261

              135265

              2.54

              A. thaliana

              34,843

              9386

              5596

              169

              41

              57

              50092

              0.53

              M. truncatula

              1,20,383

              20999

              9647

              1079

              216

              137

              152461

              0.94

              P. trichocarpa

              194,557

              54304

              25130

              3178

              772

              665

              278606

              1.66

              P. suffruticosa

              4560

              185911

              27235

              18953

              99

              376

              237134

              --

              Data for B. distachyon, S. bicolor, O. sativa, A. thaliana, M. truncatula, and P. trichocarpa were obtained from Sonah et al. 2011 (PLoS One, 6: 1–9).

              --: the percent was not calculated due to the Genome of P. suffruticosa was unkown.

              http://static-content.springer.com/image/art%3A10.1186%2F1471-2164-14-886/MediaObjects/12864_2013_5597_Fig3_HTML.jpg
              Figure 3

              Total numbers of each repeat motif. The x-axis indicates different repeat motif. The y-axis indicates the number of SSRs with various repeat motifs. 1: mono- nucleotide repeats; 2: di- nucleotide repeats; 3: tri- nucleotide repeats; 4: tetra- nucleotide repeats; 5: penta- nucleotide repeats; 6: hexa- nucleotide repeats.

              Relative frequency of different SSR repeat motifs

              A summary of SSRs, including repeat motif and total number of different repeat motifs, is shown in Table 3. Of the two possible types of mono-nucleotide repeats, the most abundant was (A/T)n, as in most plants; (G/C)n was much less common in tree peony, accounting for only 0.05% of total SSRs. SSR frequency decreased with increasing motif length (mono- to hexa-nucleotide repeats); most SSRs were composed of mono-, di-, tri-, or tetra- nucleotide repeats, with only a very small share contributed by penta- and hexa-nucleotide repeats. The di-nucleotide repeat (AC)n was more common than (AG)n and (AT)n. With respect to tri-nucleotide repeats, A/T-rich repeats were dominant in tree peony (Table 3), with AAC/GTT the most abundant tri-nucleotide motif (65.7%) followed by AAG/CTT (20%). The repeats CCG and ACG were less frequent or absent. The most frequent penta- and hexa-nucleotide repeat motifs were sequences containing the di-nucleotide CpG: AACGT/ACGTT and AAGGAG/CCTTCT, respectively.
              Table 3

              Frequency of mono-, di-, tri-, and tetra-nucleotide repeat motifs in the tree peony genome

              Repeat type

              Tree peony genome

              Mono-nucleotide

              4560

              A/T

              3956

              C/G

              604

              Di-nucleotide

              185911

              AC/GT

              124208

              AG/CT

              59711

              AT/AT

              1868

              CG/CG

              124

              Tri-nucleotide

              27235

              AAC/GTT

              17890

              AAG/CTT

              5394

              AAT/ATT

              106

              ACC/GGT

              1756

              ACG/CGT

              369

              ACT/ATG

              606

              AGC/CGT

              579

              AGG/CCT

              268

              ATC/AGT

              262

              CCG/CGG

              5

              Relative frequencies of different SSR repeat lengths

              The lengths of SSRs were classified into 11 groups (Figure 4). The shortest SSRs (10 bp) constituted 1% of the total. SSRs with lengths of 21–30 and 101–110 bp accounted for 26% and 29% of SSRs, respectively (Figure 4). Among di-nucleotide SSRs, the most abundant repeated length was 28 bp, followed by 12 bp and then 30 or 14 bp. The most common length of tri-nucleotide SSRs was approximately 15 bp, with smaller numbers of 18- and 21-bp sequences. Repeat lengths of tetra-nucleotide SSRs ranged from about 20–28 bp.
              http://static-content.springer.com/image/art%3A10.1186%2F1471-2164-14-886/MediaObjects/12864_2013_5597_Fig4_HTML.jpg
              Figure 4

              SSR length distribution. The x-axis indicates the length of SSRs (bp). The y-axis indicates the number of SSRs with different length.

              Compound SSR analysis

              About 26% of identified SSRs were compound. Interruption distance ranged from 5–195 bp, with most interruptions 5–20 bp long (Figure 5). Many of the compound SSRs were composite, being made up of various combinations of mono- to hexa-nucleotide repeats, such as (mono-nucleotide repeat)n-(tetra-nucleotide repeat)n, (tetra-nucleotide repeat)n-(tetra-nucleotide repeat)n, (mono-nucleotide repeat)n-(tri-nucleotide repeat)n, (tri-nucleotide repeat)n-(tri-nucleotide repeat)n, or (hexa-nucleotide repeat)n-(tri-nucleotide repeat)n. About 56% of repeat motifs were found in compound SSRs, revealing the complexity of the tree peony genome.
              http://static-content.springer.com/image/art%3A10.1186%2F1471-2164-14-886/MediaObjects/12864_2013_5597_Fig5_HTML.jpg
              Figure 5

              Distribution of compound SSR interruption distances. The X-axis indicates the length of interruption between SSRs (bp). The y-axis indicates the number of SSRs with different interruption length.

              Microsatellite distribution in different genomic regions of tree peony using Arabidopsis, poplar, and grape reference sequences

              The distribution of SSRs from tree peony was analyzed based on Arabidopsis, grape, and poplar coding regions (Figure 6). Many sequences (25.6%) were mapped onto Arabidopsis CDSs, whereas only 0.1% and 0.3% were mapped onto CDSs of grape and poplar, respectively. A large number of sequences, 28.1%, 21.7%, and 22.9%, respectively, could not be mapped onto any Arabidopsis, poplar, or grape genomic region. More tree peony SSRs mapped to 5′ UTRs than to 3′ UTRs in the above three species, while 14,290, 23,133 and 5,982 SSR-containing sequences were mapped to introns of Arabidopsis, grape, and poplar genomes, respectively (Figure 6).
              http://static-content.springer.com/image/art%3A10.1186%2F1471-2164-14-886/MediaObjects/12864_2013_5597_Fig6_HTML.jpg
              Figure 6

              Distribution of SSR reads mapping onto Arabidopsis , poplar, and grape genomes. The x-axis includes three reference plants namely Arabidopsis, Grape species and Poplar; The y-axis indicates the numbers of sequences with SSRs mapped with the reference plants at various positions within genes/genomes. UTR = Untranslated region; CDS = Coding DNA sequence.

              Unlike Arabidopsis, grape and poplar genomes have not been fully annotated; Arabidopsis was consequently used as a reference plant for further study (Figure 7). Most SSRs with mono-nucleotide repeats (94%) could not be mapped, and only 0.3% were observed in CDS regions. Among SSRs with di-nucleotide repeats, 44% mapped within 3′ UTRs, 5′ UTRs, introns, and CDSs, with 27% of these found in introns. Approximately 55% of SSRs with tri-nucleotide repeats mapped within CDSs, while most SSRs with tetra-, penta-, or hexa-nucleotide repeats mapped onto intergenic positions, or could not be mapped onto the Arabidopsis genome. With respect to c-type SSRs (i.e., without an interruption between two motifs), 33% mapped to introns, and 9% in total were mapped onto 3′ UTR, 5′ UTR, or CDS regions. In regard to c*-type SSRs (i.e., with an interruption between two motifs), 27% were mapped onto introns and 26% to intergenic regions (Figure 7). Among tree peony SSRs that were mapped to Arabidopsis CDS regions, those with tri-nucleotide repeats were the most abundant, followed by di-nucleotide repeat SSRs; in contrast, intergenic regions and introns contained more di-nucleotide and compound SSRs than did CDS regions (Table 4).
              http://static-content.springer.com/image/art%3A10.1186%2F1471-2164-14-886/MediaObjects/12864_2013_5597_Fig7_HTML.jpg
              Figure 7

              Distribution of SSR reads with various repeat motifs mapping onto the Arabidopsis genome. The x-axis indicates various position within genes/genome of Arabidopsis; The y-axis indicates the relative numbers with different repeat motifs mapped with various position of genes/genome of Arabidopsis. Abbreviations are: p1 = mono-nucleotide repeats; p2 = di-nucleotide repeats; p3 = tri-nucleotide repeats; p4 = tetra-nucleotide repeats; p5 = penta-nucleotide repeats; p6 = hexa-nucleotide repeats; c* = compound SSR without an interruption between two motifs; c = compound SSR with an interruption between two motifs; UTR = Untranslated region; CDS = Coding DNA sequence.

              Table 4

              Microsatellite distribution in different genomic regions of tree peony using the Arabidopsis genome as a reference

              Microsatellite distribution

                 

              Repeat motif

                 

              Mapping type

              c

              c*

              p1

              p2

              p3

              p4

              p5

              p6

              Intergenic

              8173

              727

              203

              18797

              410

              754

              2

              10

              No-mapped

              10464

              218

              3560

              25867

              5039

              766

              24

              80

              3′ UTR

              915

              180

              0

              1534

              14

              0

              0

              0

              CDS

              1755

              397

              15

              4674

              7440

              0

              0

              9

              Intron

              12760

              1051

              1

              28034

              130

              135

              0

              0

              5′ UTR

              921

              284

              0

              10409

              59

              0

              0

              1

              Multi-mapped

              3963

              1024

              0

              12700

              544

              0

              0

              0

              Compound SSRs are designated as follows: c* = no interruption between two motifs; c = interruption between two motifs; p1–p6 refers to the repeat motif (e.g., mono-nucleotide, di- nucleotide, etc.).

              Functional annotation of SSR-containing coding sequences

              Gene Ontology (GO) analysis was performed on sequences with SSRs mapping onto CDSs. Numbers of genes and GO classifications are displayed in Figure 8. Genes were classified into three major categories based on their sub-cellular function: cellular component, molecular function, and biological processes. Genes with functions related to cell and cell part (GO ID: 0044464), macromolecular complex (GO ID: 0032991), and organelle (GO ID: 0043231) were the most abundant genes in the cellular component category. The molecular function category was rich in genes associated with binding (GO ID: 0005488), catalytic activity (GO ID: 0003824), and structural molecules (GO ID: 0005198). Genes related to cellular process (GO ID: 0019941), metabolic process (GO ID: 0008152), and response to stimulus (GO ID: 0009607) were the most heavily represented in the biological process category.
              http://static-content.springer.com/image/art%3A10.1186%2F1471-2164-14-886/MediaObjects/12864_2013_5597_Fig8_HTML.jpg
              Figure 8

              GO classification of SSRs in coding regions, including the number/percentage of genes putatively involved in different subcellular functions. The x-axis refers to different functional classes within a cell performing various functions. The y-axis indicates the percentage (left) or number (right) of genes within SSRs belonging to various functional classes.

              Validation of SSR assays

              A major advantage of the approach used in this study is the ability to obtain SSRs rapidly, thus greatly reducing the time and expense required to check for polymorphisms. In this study, 100 primer pairs were selected for validation using three tree peony accessions (Additional file 1: Table S1), and 24 primer pairs with high amplification effect were subsequently used for polymorphism analysis among 23 accessions (Table 5). The number of alleles per locus ranged from two to five; expected heterozygosity varied between 0.0850 and 0.7275, whereas observed heterozygosity ranged from 0.0000 to 0.8410 (Table 5). It has been confirmed that P. rockii and P. ostii are the ancestors of all 21 cultivars analyzed in this study. UPGMA analysis of SSR data resulted in cultivars ‘Yao Huang’ (1), ‘Dou Lv’ (2), ‘Shui Jing Bai’ (3), and ‘Liu Li Guan Zhu’ (4) from the Zhongyuan cultivar group clustering together, demonstrating their close genetic relationships to one another. Cultivars from the Japanese cultivar group—‘Taiyoh’ (10), ‘Shima Nisshiki’ (11), and ‘Gun Pou Den’ (12), derived from the Zhongyuan group, clustered with this latter group. All of these cultivars formed a major branch in the UPGMA dendrogram (Figure 9). Cultivars ‘Huai Nian’ (14), ‘Ju Yuan Shao Nv’ (15), and ‘Xin Xing’ (16) from the Xibei cultivar group clustered together, reflecting their close inter-relationships, and formed another branch in the UPGMA dendrogram (Figure 9). The results of cluster analysis of these SSR genotypes, consistent with known genetic relationships, are similar to results obtained using EST-SSR, TRAP, and SRAP markers [17, 18], and confirm the usefulness of the SSR markers developed in this study.
              Table 5

              SSR loci amplified from 23 accessions of tree peony

              Locus

                

              23 accessions

                   
               

              Repeat

              Forward primer (5′-3′)

              Reverse primer (5′-3′)

              Size (bp)

              Ta (ºC)

              Na

              He

              Ho

              2A

              TGG6

              AACTGCGCTAGTCGTCCCCATAAAC

              AAAGCCGCCTACAGAGGATGTTCAT

              268

              57

              3

              0.4647

              0.0435

              19A

              CA16

              TAACATCTCACTACCACTCAGGCGA

              CATAAGGGTGATGATCATGTGGTTG

              164

              54.5

              3

              0.6493

              0.0000

              25A

              TGT10

              CAATCCCTTTTGTAATGCCCCTTTC

              CAGGCTGTACTAGCAAAGGCTTCCA

              215

              54.5

              3

              0.5565

              0.0000

              26A

              TTG7

              TGGGCCCTACAAGTGATGATATTCC

              ATGGAATCCAGGTTTGTGAATGTGA

              245

              54.5

              3

              0.559

              0.0000

              30A

              CA13

              TGTCATACCGACTTCGGCTAGGCTA

              AAGGGTGATCGTGTGGTTGATGTTT

              265

              54.5

              4

              0.7275

              1.0000

              31A

              CT11

              AGCGCGTTTAATTGCTCTTACCTTG

              CTCCCTCCTCTAACTCCATGCTTGC

              303

              54.5

              3

              0.6261

              0.0000

              36A

              (TGG)6gctttggccggttcg(CTT)5

              GACTGTAGTGATGGTGGTGGATTGG

              AGCTTATGAACCCTGATGATGACGC

              261

              57

              3

              0.5797

              0.0000

              48A

              CAG5

              ACAGCGTCAGCAGACAGGAAGTACC

              AAGAGTACCTGTCACCCCATCCAAA

              364

              57

              4

              0.5913

              0.0000

              49A

              TGC5

              TCTGGGTGATAGGTGGAGCTGGTGC

              GGAAGACGCCCACAATGAAATCACA

              314

              57

              4

              0.6696

              0.0000

              50A

              CA13

              CACGGCTTTAAAATGCGTCTCAACT

              AGGCTGGTGATAGTGTTGTTGATGC

              252

              54.5

              4

              0.5295

              0.0000

              53A

              TCC5

              CTCTTGTCAACCCCCACTGCCTCCT

              GAAGGGACTTTCGCTGGAATCTGGC

              353

              59

              4

              0.6802

              0.0000

              54A

              (CT)9(CA)14

              TGTCGGGCGGTAAGTTTAGGGAAGA

              CCACTTGGGTTCTGTTGGAGACTCG

              388

              59

              3

              0.5034

              0.0435

              56A

              AC15

              CAGGTGGCATTTTTGGCTTCTCTCT

              TTGGCCCAATCACATGTAATCCCTC

              388

              57

              3

              0.5217

              0.0000

              58A

              GCA6

              TAGGATGACAAAGTGCAGGAAACCC

              TGCTCAAACTCATCCTCAAGCTGTG

              318

              57

              2

              0.085

              0.0000

              59A

              AC18

              TACAACACTTCTCGCCTAACGCACC

              AGACATGGTGCAAGTATGGGAGACG

              270

              59

              3

              0.4908

              0.0000

              63A

              (TC)9(AC)17

              CACCGCATATCTCCAACCTCACCTC

              TTGGGTAGAGATAGGAGGTTGGGGC

              277

              59

              3

              0.6609

              0.0000

              65A

              TGG5

              CATACCTCCATCATGATGCTGCTGT

              ATGAAGGCTCAGTAAGAACCTCGGA

              355

              57

              3

              0.3053

              0.0000

              73A

              CAG5

              CCATCTCAGGGTCAGGGTTCTCGTA

              TAGAGTGTACCTTCACCCCCATCGG

              375

              59

              4

              0.6928

              0.1379

              78A

              AC16

              TATCAAATGGGGATGGTCTCCTCTT

              AATTCTGCCACTATGAGCTCGATCT

              314

              54.5

              5

              0.6899

              0.1579

              79A

              GCA5

              AGAGGAAGTTTGAGGCCATCAGTCG

              CAACTGTAGCCTTCTGTTCCTGCCC

              367

              57

              2

              0.4638

              0.0000

              80A

              GTG5

              AAGGTTATGGTGGCAGTGAAGATGA

              ACCGTCGTACTACCACTTACAGCCG

              207

              54.5

              4

              0.6773

              0.3043

              87A

              TG15

              TGTAATCGATCGAGTTTCTTGGGTC

              CCTAACACTCCACCACTAAGTCGCT

              188

              56

              3

              0.6261

              0.0000

              91A

              (GT)9ttgta(TG)16

              TCAGCCCCTAGCATAGAAGAATCCA

              TCTCACTACCACCTACGCGATGTTC

              384

              60

              3

              0.6032

              0.0000

              Size = size of cloned allele; Ta = annealing temperature, Na = number of alleles; He = expected heterozygosity; Ho = observed heterozygosity.

              http://static-content.springer.com/image/art%3A10.1186%2F1471-2164-14-886/MediaObjects/12864_2013_5597_Fig9_HTML.jpg
              Figure 9

              UPGMA dendrogram constructed from 23 tree peony accessions using the markers developed in this study. Numbers at leaf tips refer to accession code numbers listed in Additional file 2: Table S2.

              Discussion

              The number of SSRs obtained in this study from tree peony was higher than that generated from other plants, including Arabidopsis, Medicago truncatula, Oryza sativa (rice), and Sorghum bicolor (sorghum) [1, 19]. The frequency of A/T repeats present in tree peony was between dicots and monocots [18]. The percentage of tetra-, penta-, and hexa-nucleotide repeats observed in tree peony (8%) was higher than in Sorghum (5.4%), Populus (1.66%), Medicago (0.94%), rice (2.54%), Brachypodium (2.45%), and Arabidopsis (0.53%) [20].

              The frequency of di-nucleotide repeats in tree peony was not consistent with that observed in Brachypodium by Sonah et al. [19]. Similar to rice, AG/CT repeats were well represented. AG/CT and AT/AT repeats were abundant in tree peony, accounting for 41.9% and 41.0%, respectively, of identified SSRs, while AT/AT repeats were more frequent in Populus (60.5%) and Medicago (59.9%) [19]. CG/CG repeats were relatively uncommon in tree peony, however, similar to Populus, Medicago, and Arabidopsis, suggesting that CG-rich motifs are the least preferred in dicot genomes. In human, Caenorhabditis, and Arabidopsis genomes, the most common di-nucleotide repeats are (AC)n, (AG)n, and (AT)n, respectively, demonstrating that different species have different motif frequency distributions.

              With respect to tri-nucleotide repeats, AGC/CGT, AGG/CCT, and CCG/CGG have been observed more frequently in monocots than in dicots. A/T-rich repeats were the dominant tri-nucleotide SSRs in tree peony, similar to the results of Sonah et al. [19]. In tree peony, the sparseness or absence of CCG and ACG repeats may be due to highly mutable CpG di-nucleotide repeats, as evidenced in rice by the tendency of tri-nucleotide repeats, with few exceptions, to consist of various combinations of C and G. Transcriptional repression by DNA methylation depends on CpG density; CCG repeats may also be selected against by the requirements of the splicing machinery, with maintenance or absence of CCG possibly an active process [20]. The total absence of a particular repeat motif may indicate that the sequence is not preferred by the mechanism generating repeats or that strong selective pressure exists against repeated occurrence of particular sequences [20].

              The characteristically short lengths of SSRs may have functional implications with respect to their evolution or the genes involved in plant physiology and development. In a previous study [4], rice SSRs were divided into two groups based on the length of SSR tracts and their potential as informative genetic markers: Class I microsatellites contained perfect SSRs ≥ 20 bp long and Class II microsatellites contained perfect SSRs 12–20 bp long. Class II microsatellites tended to be less variable because of less possibility of slipped-strand mispairing over the shorter SSR template. In tree peony, 85% of SSRs were categorized as Class I microsatellites and 1% as Class II microsatellites. Longer perfect repeats (Class I) have been determined to be highly polymorphic [21]. In future studies of tree peony SSRs, attention should focus on Class I microsatellites, with an emphasis on evaluation of polymorphism and its implications.

              Length variation of repeated units may be due to differences in generation and fixation mechanisms of simple repetitive DNA. The inherent ability of a sequence to form alternative DNA conformations may be important for SSR generation, but does not explain differences observed among taxa. Enzymes or other proteins responsible for various aspects of DNA processing, such as replication and repair, and for chromatin remodeling, may be involved in the taxon specificity of microsatellite characteristics. It should be emphasized that not only do genomes differ in degree of repetitiveness [22], but also in preferred microsatellite types. In plant genomes, the frequent occurrence of repeat motifs of a particular sequence and length is the result of selection pressure applied on the specific motif during evolution [20]. The molecular mechanism responsible for the origin of microsatellites is still a subject of controversy, with many theories—such as replication slippage and unequal crossing-over—proposed to explain their occurrence [19, 23, 24]. The essential basis for species-specific accumulation of particular motif repeats, repeat lengths, and G/C content, which may influence unique microsatellite distribution patterns and evolution, is also still unclear. Variations in repetition purity and motif length enable site-specific adjustment of mutation rate and mutation effect, evidence indicating that common SSR alleles may offer potential selective advantages [25]. The increasing number of species with sequenced genomes should provide a foundation for the study of microsatellite evolution and even lead to discovery of the genetic/genomic role of microsatellites.

              SSR frequency in monocot CDS regions is twice that of dicots [18]. It has been suggested that SSRs in different gene positions may perform varied functions. In animals, including mammals and other vertebrates, introns contain more poly (A/T) than poly (C/G) repeats. In Caenorhabditis elegans, however, intergenic regions show an interesting preference for poly (C/G) over poly (A/T) repeats [21], indicating that preferences may vary among organisms. In tree peony, the abundance of tri-nucleotide repeats mapping onto CDS regions was consistent with results found for the six species studied by Sonah et al. [19]. Tang et al. [26] examined SSRs in the Arabidopsis genome, and found that SSRs generally were preferentially located in upstream gene regions, especially 5′ UTRs; as in tree peony, tri-nucleotide repeats were the most common repeats found in coding regions. The accumulation of tri-nucleotide repeats in coding regions is primarily due to the triplet-repeat nature of codons [19]. The various numbers of repeats in coding regions are a potential source of quantitative and qualitative phenotypic variation [26]. SSRs in 5′ UTRs and CDSs may modify the expression or function of genes with which they are associated [26].

              In rice, 80% of GC-rich tri-nucleotide repeats occur in predicted exons, while AT-rich tri-nucleotide repeats are distributed evenly across all genomic components. Di-nucleotide and tetra-nucleotide repeats are predominantly situated in noncoding—mainly intergenic—regions. (GA)n repeats usually occur in regions with a balanced (close to 50%) GC content, favoring robust PCR amplification, whereas (CA)n and (AT)n are rare in gene-rich regions [4]. Tri- and hexa-nucleotide repeats have been shown to be the most common repeats in eukaryotic coding regions [20, 27]. In our study, SSR-containing genes encoding for binding, catalytic, and structural molecules were abundant in the GO molecular function category, similar to results found in Brachypodium[19]. While such SSR-containing genes may perform multiple functions in tree peony, the importance of SSRs within genes remains to be further explored.

              The SSR markers identified in this study should be useful for population genetic studies, and are potentially amplifiable across the genus. Plant genomes are complex, and contain large amounts of repetitive DNA, including microsatellites, which has immediate practical implications for the success of SSR marker development. Observed differential patterns of SSR marker distribution may be helpful for studying microsatellite evolution in a monocot-dicot system. SSR markers developed in this study have potential application to genomic research, marker-assisted breeding, DNA fingerprinting of genetic resources, molecular mapping of tree peony and related species, and map-based cloning of candidate genes. Hypervariable microsatellites are a useful source of polymorphic DNA markers for linking genetic maps with genomic sequences, and ultimately with phenotypic variation. They provide an opportunity to use SSR markers to investigate the wide range of genetic diversity that exists in wild relatives outside of the tree peony gene pool. Because SSRs are associated with vital functions and characteristics, such as transcription factor binding, RNA shape, DNA structure and packaging, and DNA length and orientation [24], the SSRs obtained in this study may be important for investigating plant genomic structural evolution and for providing insights into phenotypic variation in species during their evolution.

              Conclusions

              This study represents the first application of next-generation sequencing for high-throughput microsatellite development in tree peony. The large size of the tree peony genome, approximately 16 G (data from private correspondence), hampers its sequencing, and the species is not highly amenable to transformation because of its woody characteristics. Consequently, the 237,134 microsatellites obtained in this study should be useful for marker-assisted breeding and functional characterization of genes related to trait formation. In addition, because the phylogenetic position of Paeoniaceae is still unresolved, the uncovered microsatellites may serve as a data resource for evolutionary studies in the family.

              Methods

              Plant materials

              Leaves of tree peony (Paeonia suffruticosa Andrews) were collected from the Peony Germplasm Garden, Institute of Botany, Chinese Academy of Sciences (Beijing, China). Three cultivars—‘Liu li guan zhu’ , ‘Fu gui hong’ , and ‘Wu cai die’—were used for primer validation. Twenty-three accessions of tree peony were used for marker validation (Additional file 2: Table S2).

              Genomic DNA isolation, library preparation, and R454 sequencing

              Total genomic DNA was extracted using the CTAB method [17]. Genomic DNA (500 μl; 600 μg) was fragmented with nitrogen at 45 psi for 2 min; 500–750-bp fragments were used for further study. Both fragment ends were polished and ligated to adaptors using T4 ligase. After PCR amplification of fragments with adaptor primers, selective hybridization was performed using eight biotin-labeled probes—pGA, pAC, pAAT, pAAC, pAAG, pATGT, pGATA, and pAAAT—and streptavidin-coated beads (Dynabeads; Invitrogen, Grand Island, NY 14072, USA) [2831]. Library quality inspection and sequencing of clones was carried out as described by Yang et al. [31].

              DNA (5 μg per plate) was sequenced on a Roche 454 GS FLX sequencer using Titanium reagents. Processing and analysis of sequencing data was performed with GS-FLX Software v2.0.01 (454 Life Sciences/Roche, Werk Penzberg82372, Penzberg, Germany). Raw sequences in SFF files were base-called using the python script sff_extract.py developed by COMAV (http://​bioinf.​comav.​upv.​es), and then processed to remove low-quality and adaptor sequences using the programs tagdust [32], LUCY [33], and SeqClean [34] with default parameters.

              SSR locus search and mapping

              The program MISA (Microsatellite identification; http://​pgrc.​ipk-gatersleben.​de/​misa/​) was used to identify reads and contigs containing SSRs. Criteria used for selection were a minimum of five repeats for simple motifs and three repeats for complex or imperfect repeats, a motif length of 2–10 bp, and, for compound SSRs, a maximum interruption distance of 100 bp between different SSRs. To facilitate SSR detection, only 1- to 6-nucleotide motifs were considered, and the minimum repeat unit was defined as 10 for mono-, 6 for di-, and 5 for tri-, tetra-, penta-, and hexa-nucleotides. SSR position, number of different repeat types, and length (motif bp × number of motifs) were analyzed using the ‘bespoke’ function in MISA [35] and plotted using Open Office Calc.

              To map coding regions, all reads containing SSRs were compared against Arabidopsis (ftp://​ftp.​arabidopsis.​org/​home/​tair), grape (http://​www.​genoscope.​cns.​fr) and poplar (ftp://​ftp.​jgi-psf.​org) public databases using the program BWA-SW [36, 37]. Map position was categorized as follows: 3′/5′ UTR, CDS, intergenic, intron, non-mapped, or multi-mapped. The repeat unit type (1–6, compound, or compound with interruption) was then determined.

              GO annotation was conducted by searching against the Nr database using Blast2GO (E-value = 10-6) [38]. WEGO [39] and custom scripting were used to assign each GO ID to the related ontology entry.

              Primer acquisition and validation

              Primer pairs for flanking sequences of each unique SSR were designed automatically using Primer3 [40], with target microsatellites containing at least five repeats and yielding PCR products of 80–500 bp. One hundred primer pairs were synthesized and used for validation (Additional file 1: Table S1). Screened primer pairs giving good amplification were subsequently used to characterize genetic diversity among 23 accessions of tree peony (Additional file 2: Table S2). PCR protocols and components were as described in [17], with modifications to annealing temperatures.

              Number of alleles and expected and observed heterozygosities were calculated using POPGEN1.32 [41]. A dendrogram was constructed based on Nei’s unbiased genetic distances [42] using the unweighted pair-group method with arithmetic averages (UPGMA) as implemented in NTSYSpc-2.02 [43].

              Abbreviations

              SSR: 

              Simple sequence repeat

              QTL: 

              Quantitative trait locus

              EST: 

              Expressed sequence tag

              CTAB: 

              Cetyltrimethylammonium bromide

              MISA: 

              Microsatellite identification.

              Declarations

              Acknowledgements

              We appreciated for the technical help by Dr. Dahai Wang from Autolab. LTD. Co. (Beijing, China) and Miss Jing Sun from HeBei United University for the marker development. This study was supported by the National Natural Science Foundation of China (Grant No. 31272201) and the National High Technology Research and Development Program of China (863 Program, Grant No. 2011AA100207).

              Authors’ Affiliations

              (1)
              Beijing Botanical Garden, Key Laboratory of Plant Resources, Institute of Botany, The Chinese Academy of Sciences
              (2)
              International Center for Bamboo and Rattan, Key Laboratory on the Science and Technology of Bamboo and Rattan
              (3)
              University of Chinese Academy of Sciences

              References

              1. Lawson MJ, Zhang LQ: Distinct patterns of SSR distribution in the Arabidopsis thaliana and rice genomes. Genome Biol 2006, 7:R14.PubMed CentralPubMedView Article
              2. Li YC, Korol AB, Fahima T, Nevo E: Microsatellites within genes: Structure, function, and evolution. Mol Biol Evol 2004, 21:991–1007.PubMedView Article
              3. Trifonov EN: The tuning function of the tandemly repeating sequences: molecular device for fast adaptation . In Evolutionary Theory and Processes: Modern Horizons. Edited by: Wasser SP. Amsterdam. The Netherlands: Kluwer Academic Publishers; 2004:115–138.View Article
              4. Temnykh S, DeClerk G, Lukashova A, Lipovich L, Cartinhour S, McCouch S: Computational and experimental analysis of microsatellites in Rice ( Oryza sativa L.): Frequency, length variation, transposon associations, and genetic marker potential. Genome Res 2001, 11:1441–1452.PubMedView Article
              5. Pazos-Navarro M, Dabauza M, Correal E, Hanson K, Teakle N, Real N, Real D, Nelson MN: Next generation DNA sequencing technology delivers valuable genetic markers for the genomic orphan legume species. Bituminaria bituminosa. BMC Genet 2011, 12:104.PubMed CentralPubMedView Article
              6. Malausa T, Gilles A, Meglécz E, Blanquart H, Duthoy S, Costedoat C, Dubut V, Pech N, Castagnone-Sereno P, Délye C, Feau N, Frey P, Gauthier P, Guillemaud T, Hazard L, Le Corre V, Lung-Escarmant B, Malé PJ, Ferreira S, Martin JF: High-throughput microsatellite isolation through 454 GS-FLX Titanium pyrosequencing of enriched DNA libraries. Molecular Ecol Resour 2011, 11:638–644.View Article
              7. Perry JC, Rowe L: Rapid microsatellite development for water striders by next-generation sequencing. J Hered 2011, 102:125–129.PubMedView Article
              8. Castoe TA, Poole AW, Gu W, de Koning APJ, Daza JM, Smith EN, Pollock DD: Rapid identification of thousands of copperhead snake ( Agkistrodon contortrix ) microsatellite loci from modest amounts of 454 shotgun genome sequence. Mol Ecol Resour 2010, 10:341–347.PubMedView Article
              9. Abdelkrim J, Robertson BC, Stanton J-AL, Gemmell NJ: Fast, cost effective development of species-specific microsatellite markers by genomic sequencing. Biotechniques 2009, 46:185–191.PubMedView Article
              10. Santana QC, Coetzee MPA, Steenkamp ET, Mlonyeni OX, Hammond GNA, Wingfield MJ, Wingfield BD: Microsatellite discovery by deep sequencing of enriched genomic libraries. Biotechniques 2009, 46:217–223.PubMedView Article
              11. Rasmussen DA, Noor MAF: What can you do with 0.1× genome coverage? A case study based on a genome survey of the scuttle fly Megaselia scalaris (Phoridae). BMC Genomics 2009, 10:382–390.PubMed CentralPubMedView Article
              12. Ferrucho RL, Zala M, Zhang Z, Cubeta MA, Garcia-Dominguez C, Ceresini PC: Highly polymorphic in silico-derived microsatellite loci in the potato-infecting fungal pathogen Rhizoctonia solani anastomosis group 3 from the Colombian Andes. Mol Ecol Resour 2009, 9:1013–1016.PubMedView Article
              13. Pan L, Xia Q, Quan Z, Liu H, Ke W, Ding Y: Development of novel EST-SSRs from sacred lotus ( Nelumbo nucifera Gaertn) and their utilization for the genetic diversity analysis of N. nucifera . J Hered 2010, 101:71–82.PubMedView Article
              14. Lepais O, Bacles CEF: Comparison of random and SSR-enriched shotgun pyrosequencing for microsatellite discovery and single multiplex PCR optimization in Acacia harpophylla F. Muell. Ex Benth 2011, 11:711–724.
              15. Shu QY, Wischnitzki E, Liu ZA, Ren HX, Han XY, Hao Q, Gao FF, Xu SX, Wang LS: Functional annotation of expressed sequence tags as a tool to understand the molecular mechanism controlling flower bud development in tree peony. Physiol Plant 2009, 135:436–449.PubMedView Article
              16. Yuan JH, Cheng FY, Zhou SL: Genetic structure of the tree peony ( Paeonia rockii ) and the Qinling mountains as a geographic barrier driving the fragmentation of a large population. PLoS One 2012, 7:e34955.PubMed CentralPubMedView Article
              17. Zhang JJ, Shu QY, Liu ZA, Ren HX, Wang LS, DeKeyser E: Two EST-derived makrer systems for cultivar identification in tree peony. Plant Cell Rep 2012, 31:299–310.PubMedView Article
              18. Han XY, Wang LS, Shu QY, Liu ZA, Xu SX, Tetsumura T: Molecular characterization of Tree Peony germplasms using sequence-related amplified polymorphism markers. Biochem Genet 2008, 46:162–179.PubMedView Article
              19. Sonah H, Deshmukh RK, Sharma A, Singh VP, Gupta DK, Gacche RN, Rana JC, Singh NK, Sharma TR: Genome-wide distribution and organization of microsatellites in Plants: An insight into marker development in Brachypodium . PLoS One 2011, 6:e21298.PubMed CentralPubMedView Article
              20. Cho YG, Ishii T, Temnykh S, Chen X, Lipovich L, Park WD, Ayres N, Cartinhour S, McCouch SR: Diversity of microsatellites derived from genomic libraries and GenBank sequences in rice ( Oryza sativa L.). Theor Appl Genet 2000, 100:713–722.View Article
              21. Tóth G, Gáspári Z, Jurka J: Microsatellites in different eukaryotic genomes: survey and analysis. Genome Res 2000, 19:967–981.View Article
              22. Hancock JM: Simple sequences and the expanding genome. Bio Essays 1996, 18:421–425.
              23. Levinson G, Gutman GA: Slipped-strand mispairing: a major mechanism for DNA sequence evolution. Mol Biol Evol 1987, 4:203–221.PubMed
              24. Schlötterer C, Tautz D: Slippage synthesis of simple sequence DNA. Nucleic Acids Res 1992, 20:211–215.PubMed CentralPubMedView Article
              25. Kashi Y, King DG: Simple sequence repeats as advantageous mutators in evolution. Trends Genet 2006, 22:253–259.PubMedView Article
              26. Tang JF, Baldwin SJ, Jeanne JME, Van der Linden CG, Voorrips RE, Leunissen JAM, Van Eck H, Vosman B: Large-scale identification of polymorphic microsatellites using an in silico approach. BMC Bioinformatics 2008, 9:374.PubMed CentralPubMedView Article
              27. Metzgar D, Liu L, Hansen C, Dybvig K, Wills C: Domain-level differences in microsatellite distribution and content result from different relative rates of insertion and deletion mutations. Genome Res 2002, 12:408–413.PubMedView Article
              28. Kandpal RP, Kandpal G, Weissman SM: Construction of libraries enriched for sequence repeats and jumping clones, and hybridization selection for region-specific markers. Proc Natl Acad Sci U S A 1994, 91:88–92.PubMed CentralPubMedView Article
              29. Armour JA, Neumann R, Gobert S, Jeffreys AJ: Isolation of human simple repeat loci by hybridization selection. Hum Mol Genet 1994, 3:599–605.PubMedView Article
              30. Glenn TC, Schable NA: Isolating microsatellite DNA loci. Methods Enzymol 2005, 395:202–222.PubMed
              31. Yang T, Bao SY, Ford R, Jia TJ, Guan JP, He YH, Sun XJ, Jiang JY, Hao JJ, Zhang XY, Zong XX: High-throughtput novel microsatellite marker for faba bean via next generation sequencing. BMC Genomics 2012, 13:602.PubMed CentralPubMedView Article
              32. Lassmann T, Hayashizaki Y, Daub CO: TagDust–a program to eliminate artifacts from next generation sequencing data. Bioinformatics 2009, 25:2839–2840.PubMedView Article
              33. Chou HH, Holmes MH: DNA sequence quality trimming and vector removal. Bioinformatics 2001, 17:1093–1104.PubMedView Article
              34. Chen YA, Lin CC, Wang CD, Wu HB, Hwang PI: An optimized procedure greatly improves EST vector contamination removal. BMC Genomics 2007, 8:416.PubMed CentralPubMedView Article
              35. Thiel T, Michalek W, Varshney R, Graner A: Exploiting EST databases for the development and characterization of gene-derived SSR-markers in barley ( Hordeum vulgare L.). Theor Appl Genet 2003, 106:411–422.PubMed
              36. Li H, Durbin R: Fast and accurate short read alignment with Burrows-Wheeler transform. Bioinformatics 2009, 25:1754–1760.PubMedView Article
              37. Niu B, Zhu Z, Fu L, Wu S, Li W: FR-HIT, a very fast program to recruit metagenomic reads to homologous reference genomes. Bioinformatics 2011, 27:1704–1705.PubMedView Article
              38. Conesa A, Gotz S: Blast2GO: A comprehensive suite for functional analysis in plant genomics. Int J Plant Genomics 2008, 2008:619832.PubMed CentralPubMedView Article
              39. Ye J, Fang L, Zheng H, Zhang Y, Chen J, Zhang Z, Wang J, Li S, Li R, Bolund L, Wang J: WEGO: a web tool for plotting GO annotations. Nucleic Acids Res 2006, 34:W293-W297.PubMed CentralPubMedView Article
              40. Rozen S, Skaletsky H: Primer3 on the WWW for general users and for biologist programmers. Methods Mol Biol 2000, 132:365–386.PubMed
              41. Nei M: Estimation of average heterozygosity and genetic distance from a small number of individuals. Genetics 1978, 89:583–590.PubMed
              42. Yeh FC, Boyle TJB: Population genetic analysis of co-dominant and dominant markers and quantitative traits. BELG J Bot 1997, 129:157.
              43. Rohlf FJ: NTSYS-pc: Numerical Taxonomy and Multivariate Analysis System, Version 2.1. NY, USA: Exeter Software, Applied Biostatistics Inc.; 2000.

              Copyright

              © Gao et al.; licensee BioMed Central Ltd. 2013

              This article is published under license to BioMed Central Ltd. This is an open access article distributed under the terms of the Creative Commons Attribution License (http://​creativecommons.​org/​licenses/​by/​2.​0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

              Advertisement