EST and EST-SSR marker resources for Iris

Background Limited DNA sequence and DNA marker resources have been developed for Iris (Iridaceae), a monocot genus of 200–300 species in the Asparagales, several of which are horticulturally important. We mined an I. brevicaulis-I. fulva EST database for simple sequence repeats (SSRs) and developed ortholog-specific EST-SSR markers for genetic mapping and other genotyping applications in Iris. Here, we describe the abundance and other characteristics of SSRs identified in the transcript assembly (EST database) and the cross-species utility and polymorphisms of I. brevicaulis-I. fulva EST-SSR markers among wild collected ecotypes and horticulturally important cultivars. Results Collectively, 6,530 ESTs were produced from normalized leaf and root cDNA libraries of I. brevicaulis (IB72) and I. fulva (IF174), and assembled into 4,917 unigenes (1,066 contigs and 3,851 singletons). We identified 1,447 SSRs in 1,162 unigenes and developed 526 EST-SSR markers, each tracing a different unigene. Three-fourths of the EST-SSR markers (399/526) amplified alleles from IB72 and IF174 and 84% (335/399) were polymorphic between IB25 and IF174, the parents of I. brevicaulis × I. fulva mapping populations. Forty EST-SSR markers were screened for polymorphisms among 39 ecotypes or cultivars of seven species – 100% amplified alleles from wild collected ecotypes of Louisiana Iris (I.brevicaulis, I.fulva, I. nelsonii, and I. hexagona), whereas 42–52% amplified alleles from cultivars of three horticulturally important species (I. pseudacorus, I. germanica, and I. sibirica). Ecotypes and cultivars were genetically diverse – the number of alleles/locus ranged from two to 18 and mean heterozygosity was 0.76. Conclusion Nearly 400 ortholog-specific EST-SSR markers were developed for comparative genetic mapping and other genotyping applications in Iris, were highly polymorphic among ecotypes and cultivars, and have broad utility for genotyping applications within the genus.


Background
Iris, a genus of 200-300 species in the Iridaceae (Asparagales), is one of the most widely admired and earliest cultivated garden flowers, having appeared in ancient Eygptian artifacts as early as 1950 B.C. [1]. The most widely cultivated, hybridized, and horticulturally important species are I.germanica (tall-bearded Iris), I.pseudacorus (yellow-flag Iris), and I.sibirica (Siberian Iris), each with numerous commercially important cultivars. Iris species are found in diverse habitats on every continent in the Northern Hemisphere and have been important models for the study of plant evolution, ecology, and hybrid speciation [2][3][4][5][6][7][8]. Chromosome numbers and ploidy are highly variable among and within species in the genus, ranging from 2n = 16 in I. attica to 2n = 108 in I. versicolor [3,4]. Similarly, haploid genome lengths are generally large and highly variable in the genus, ranging from 2,000 to 30,000 Mbp [9].
Minimal genomic resources have been developed for Iris, a genus where forward genetic approaches have previously been applied to the study of life history and other traits by genotyping generic DNA markers, e.g., random amplified polymorphic DNA (RAPD) or retrotransposon display (IRRE) markers, in segregating populations developed from interspecific hybrids [7,[9][10][11][12][13][14]. While such markers have facilitated linkage and quantitative trait locus (QTL) mapping in Iris, the uncertain orthology of RAPD and IRRE bands has precluded cross-referencing loci across populations and species. Simple sequence repeat (SSR), restriction fragment length polymorphism (RFLP), and single nucleotide polymorphism (SNP) markers are typically ortholog-specific and, consequently, have been widely used as DNA landmarks for synteny analysis and cross-referencing loci across populations [15][16][17]. Thus far, a limited number of ortholog-specific DNA marker have been described for Iris [18]. The primary goal of the present study was to develop a sufficient number of ortholog-specific DNA markers for genome-wide comparative genetic mapping and other genotyping applications in I. brevicaulis (x = 20), I. fulva (x = 20), and other species in the genus by developing a small EST database and targeting SSRs in ESTs.
SSRs are ubiquitous in transcribed sequences, typically locus-specific and co-dominant, and often multi-allelic, highly polymorphic, and transferrable among species within genera [19][20][21][22]. EST databases have been a rich source of SSRs for the development of ortholog-specific EST-SSR markers for genotyping applications in numerous species of flowering plants [21][22][23][24][25][26][27][28]. When our study was initiated, a limited number of ESTs (201) had been deposited in GenBank http://www.ncbi.nlm.nih.gov/ Genbank/ for a single species in the genus, I. hollandica [29], and were insufficient for EST-SSR marker develop-ment. We developed a small EST database from cDNA sequences produced from normalized cDNA libraries of two species of Louisiana Iris (I. brevicaulis and I. fulva), partly to support the development of several hundred EST-SSR markers for comparative mapping and other genotyping applications in Louisiana Irises and partly to create DNA sequence and ortholog-specific DNA markers resources for the genus as a whole. Previous forward genetic analyses in I. brevicaulis and I. fulva identified QTL for several morphological, life history, and ecological traits [12,13,[30][31][32]. Because ortholog-specific DNA markers were previously lacking for these genera, linkage groups and QTL identified in earlier analyses could not be cross-referenced and comparative genetic mapping was infeasible. Here, we describe the I. brevicaulis-I. fulva EST database and the development, cross-species utility, and polymorphisms of I. brevicaulis-I. fulva EST-SSR markers among wild collected ecotypes of four species of Louisiana Iris (I. brevicaulis, I. fulva, I. hexagona, and I. nelsonii) and horticulturally important cultivars of tall-bearded (I.germanica), yellow-flag (I.pseudacorus), and Siberian (I.sibirica) Iris.

Development of a Leaf and Root EST Database for Iris
Normalized leaf and root cDNA libraries were developed from I.brevicaulis (IB72) and I.fulva (IF174) ecotypes (root and leaf RNAs were pooled and a single cDNA library was constructed for each species). Library quality was checked by sequencing colony-PCR amplified inserts from 295 randomly selected cDNA clones split between the IB72 and IF174 libraries. Of the 295 clones, 251 (85.1%) harbored inserts 800 bp or longer, three lacked inserts (1.1%), and 290 (98.3%) harbored unique inserts. Subsequently, 12,199 cDNA clones were single-passed sequenced and yielded 6,530 ESTs surpassing quality standards, 2,947 from the IB72 and 3,583 from the IF174 library. Less than 1% of the clones lacked cDNA inserts (85/12,199). The vector-and quality-trimmed ESTs were deposited in GenBank (Acc. No. EX949962-EX956238 and FD387191-FD387443), annotated by BLASTX analyses against NCBI databases, assembled, and deposited in a database http://www.genome.uga.edu/IrisDB developed by modifying a previously described EST processing pipeline and database [33]. The mean length of vectorand quality-trimmed ESTs was 578.0 bp. cDNA normalization minimized redundancy in the Iris EST database and yielded a wealth of unique cDNA sequences (unigenes) for EST-SSR marker development and other applications in Iris biology, breeding, and floriculture http://www.genome.uga.edu/IrisDB. The 6,530 ESTs assembled into 4,917 unigenes (3,851 singletons and 1,066 contigs); hence, 75.3% of the ESTs were unique and 78.3% of the unigenes were singletons. cDNAs were normalized using a protocol which has been applied in numerous plant and animal species and minimized abundant transcript resequencing [34][35][36]. cDNA populations in leaves are dominated by abundant transcripts, e.g., chlorophyl A/B binding proteins and rubisco, neither of which were abundant among transcripts isolated by sequencing normalized leaf cDNA libraries. The deepest contigs contained seven ESTs.
Unigenes ranged in length from 100 to 1,673 bp with a mean length of 603.1 bp. Less than one-tenth of the unigenes (433/4,917) were sequenced through the polyA tail. Mean GC contents, which were identical for I.brevicaulis (45.4%) and I.fulva (45.3%), were slightly greater than mean GC contents reported for onion (Allium cepa L.; 41.9%) and Arabidopsis (42.7%) transcripts [37,38]. Unigenes were annotated by BLASTX http:// www.ncbi.nlm.nih.gov/BLAST analyses against the NCBI Non-Redundant Protein http://www.ncbi.nlm.nih.gov/ RefSeq/ and UniProtKB Swiss-Prot and TrEMBL http:// www.expasy.ch/sprot/ databases. Using a BLASTX threshold of <E = 1 10 , significant similarities were found and putative functions were identified for 2,390 Iris unigenes (48.6%). Thirty-two (0.6%) additional unigenes were similar to genes of unknown function. Significant similarities were not found for the other 2,495 Iris unigenes (50.8%). The fraction of unigenes homologous to cDNAs encoding known function genes was similar to onion, an economically important species in the Asparagales [38].
The Louisiana Iris ESTs developed in the present study have moderately increased DNA sequence resources for Iris, which were previously minimal http:// www.ncbi.nlm.nih.gov/, and supplied ESTs for an important basal species in the Asparagales, a family where DNA sequence information has primarily been produced for onion, asparagus (Asparagus officinalis L.), and model species [38,39]. van Doorn et al. [29] previously described 201 I. hollandica ESTs from a tepal cDNA library. Other than the latter, 607 nucleotide sequences for 104 species of Iris had previously been deposited in public databases, the bulk of which were for a limited number of DNA sequence motifs commonly targeted in phylogenetic analyses, e.g., matK. The Sanger ESTs described here were produced before the emergence of next-generation DNA sequencing technologies, which have dramatically increased DNA sequencing throughput and are facilitating deeper and broader DNA sequencing than was previously practical in species with limited DNA sequence resources [40][41][42]. The Sanger ESTs we produced, while limited in number, build the foundation for deeper transcriptome sequencing in Iris using next-generation technologies.

Abundance, Characteristics, and Distribution of SSRs in Louisiana Iris ESTs
SSRs were highly frequent in the Louisiana Iris EST database (Figures 1, 2; Additional File 1; http:// www.genome.uga.edu/IrisDB). We identified 1,447 perfect SSRs (n ≥ 5) in 1,162 unigenes. One-fourth of the 4,917 unigenes in the transcript assembly harbored at least one SSR, a frequency which was much greater than the frequency range (2-12%) in many other flowering plants [20,21,24,43]. The mean SSR density was one per 2,048 bp, which was much higher than the density found in onion (1/25 kb; [38]), another species in the Asparagales, and Arabidopsis (1/14 kb; [44]). When the transcript assembly was mined for perfect and imperfect repeats, 3,487 SSRs (n ≥ 5) were identified in 2,037 unigenes (41.4%) with a mean density of approximately one SSR per 850 bp; imperfect repeats are interrupted short tandem repeats.
SSR repeat numbers ranged from 5 to 30 and lengths ranged from 10 to 69 bp ( Figure 1; Additional File 1). Of the 1,447 perfect SSRs, 1,077 (72.9%) were 14 bp or longer and 694 (48.0%) were 18 bp or longer. The mean repeat number was 9 and the mean repeat length was 23 bp. Of the 1,447 perfect repeats, 807 were dinucleotides (55.8%) and 569 were trinucleotides (39.3%). The most common repeat motifs were AG/CT (50.1%), AAG/CTT (18.9%), and AGG/CCT (7.6%) ( Figure 2). Slightly more than two-thirds of the SSRs were located in UTRs (61.4% in 5'-UTRs, 8.3% in 3'-UTRs, and 30.3% in exons). Of repeats identified in UTRs, 62.3% were dinucleotides and 31.4% were trinucleotides. Conversely, of repeats identified in CDSs, 17.4% were dinucleotides and 82.6% were trinucleotides (Additional File 1). The low frequency of SSRs identified in 3'-UTRs was primarily a function of 5'-Distribution of repeat counts for simple sequence repeats (SSRs) identified in 1,162 unigenes in the I. brevicaulis-I. fulva EST database  The most common dinucleotide repeat motif was AG/CT, which constituted 89.8% of the dinucleotide repeats identifed in Iris ESTs and has been the most common dinucleotide repeat identified in other plant EST databases [20,21,24,43]. AG/CT repeats have been widely targeted for EST-SSR marker development in plants because, in addition to being highly abundant, they are often highly polymorphic, more abundant in UTRs than CDSs, seldom associated with transposons, and consistently amplify and yield robust SSR markers [20,24]. The frequencies of trinucleotide repeats in CDSs and dinucleotide repeats in UTRs appear to be similar in wheat and Iris ( [26]; Additional File 1).

Louisiana Iris EST-SSR Marker Development, Screening, Allele Length Polymorphisms, and Cross-Species Utility
The nine EST-SSR markers were highly polymorphic among tall-bearded, yellow-flag, and Siberian Iris cultivars; heterozygosities ranged from 0.77 to 0.91 (Table 2). Even though I. pseudacorus and I. sibirica belong to the same section (Limniri) as Louisiana Iris [5], a significant decrease in allele amplification was observed in these species, and was comparable to the decrease observed in I. germanica, a species from section Iris. Nevertheless, many of the I.brevicaulis-I.fulva EST-SSR markers developed in the present study amplify alleles from other species and should have broad utility in the genus [1,5].
The 40 EST-SSR markers were highly polymorphic among and consistently amplified alleles from Louisiana Iris ecotypes; the null allele frequency was 0.5% (Table 1; Figure  3; Additional File 4). The number of alleles/locus (n) ranged from two to 18, the mean number of alleles/locus (n) was 8.9, heterozygosities of individual SSR markers ranged from 0.36 to 0.90, and the mean heterozygosity (h) was 0.76. Eighty to 100% of the EST-SSR markers were polymorphic, n ranged from 2.9 to 5.2, and h ranged from 0.41 to 0.65 among Louisiana Iris ecotypes (Table 1; Additional File 4). The number of species-specific alleles

Genetic Diversity Among Wild Collected Ecotypes and Horticulturally Important Cultivars
Because only nine of the 40 I. brevicaulis-I.fulva EST-SSR markers amplified alleles from horticultural cultivars of I.germanica, I.pseudacorus, and I.sibirica, genetic distances and dendrograms were separately estimated from genotypes of the nine EST-SSR markers among accessions of all seven species and of the 40 EST-SSR markers among ecotypes of the four Louisiana Iris species (Figure 4; Additional File 4). Genetic distances (G) ranged from 0.25 to 0.93 among Louisiana Iris ecotypes. The longest genetic distances were interspecific (G = 0.93 between IB70 and IF10, IH32 and IN33, and IB25 and IF17), whilst the shortest genetic distances were intraspecific (G = 0.25 between IF14 and IF17 and IH10 and IH16). Ecotypes assembled into species-specific clusters which were separated by greater genetic distances than ecotypes within species-specific clusters (Figure 4; Additional File 5). Genetic diversity was significant and diffuse among ecotypes or cultivars within species. Only a few EST-SSR markers were needed to identify (distinguish) ecotypes and cultivars.      Teal Velvet-IS4   IP1   IP2   IP3   IP4   IH10   IH15  IH259 IH16  IH32 IH253   IB72   IB70  IB38  IB29   IB48   IB49   IB25 and quality-trimmed ESTs longer than 100 bp were assembled using MEGABLAST and CAP3 TGI Clustering Tools http://compbio.dfci.harvard.edu/tgi/software/. BLASTX http://www.ncbi.nlm.nih.gov/BLAST analyses were performed against the NCBI Non-Redundant Protein Database, UniprotSprot, and UniprotTrembl to identify putative functions of and annotate unique transcripts (unigenes).

EST-SSR Discovery, Marker Development, and Length Polymorphism Screening
Unigenes in the transcript assembly were screened for perfect repeat motifs using SSR-IT (http://www.gramene.org/ db/searches/ssrtool; [24]) and imperfect repeat motifs using FastPCR http://www.biocenter.helsinki.fi/bi/Pro grams/fastpcr.htm. SSRs with a minimum repeat count (n) threshold of n ≥ 5 were selected for further analysis and EST-SSR marker development (Additional File 1). Flanking forward and reverse primers were designed for SSRs in 526 unigenes using Primer 3 (http:// frodo.wi.mit.edu/cgi-bin/primer3/primer3_www.cgi; Additional File 2). To facilitate multiplex genotyping on an ABI3730 XL Capillary DNA Sequencer (Applied Biosystems, Foster City, CA), SSR primers were designed by uniformly varying target amplicon lengths from 100 to 450 bp and end-labeling forward primers with one of three fluorophores, 6FAM, HEX, or TAMRA (Additional File 2). The 526 SSR markers were screened for amplification and length polymorphisms among two I.brevicaulis ecotypes (IB72 and IB25) and an I.fulva ecotype (IF174) on agarose [47].
To assess cross-species amplification, transferability, and allele length polymorphisms, 40 of the 526 I. brevicaulis-I. fulva EST-SSR markers were screened among 26 ecotypes sampled from four Louisiana Iris species (I.brevicaulis, I.fulva, I. nelsonii, and I. hexagona), four yellow-flag Iris cultivars (I.pseudacorus), four Siberian Iris cultivars (I.sibirica), and five tall-bearded Iris cultivars (I.germanica) (Additional File 3). Louisiana Iris ecotypes were collected from Terrebonne Parish and St. Martinville Parish, Louisiana and I.pseudacorus ecotypes were collected from Spring Lake, San Marcos, TX. Tall-bearded and Siberian Iris cultivars were purchased from Schreiner Iris Gardens, Salem, Oregon. The 40 EST-SSR markers were previously mapped and distributed among 21 I.brevicaulis × I.fulva linkage groups (unpublished data). Genomic DNA was isolated from leaves of the 39 ecotypes or cultivars using a modified cetyltrimethylammonium bromide (CTAB) method [48]. SSR markers were genotyped on an ABI 3700 XL Capillary DNA Sequencer as previously described [47,49] and SSR allele lengths were ascertained using GeneMapper (Applied Biosystems, Foster City, CA). Heterozygosities (H) of individual EST-SSR markers were estimated as described by Ott [50]. Genetic distances (G) were esti-mated using the proportion of shared alleles estimator in Microsat, where G = (1 -p) and p is the proportion of shared alleles http://hpgl.stanford.edu/projects/microsat/ . Neighbor-joining (NJ) trees were constructed using the NEIGHBOR program in PHYLIP http://evolution.genet ics.washington.edu/phylip.html and were drawn with TreeView http://taxonomy.zoology.gla.ac.uk/rod/ treeview.html.