We analysed 48,961 virtual Physcomitrella transcripts with the MISA software 2526. In 3,108 (6.3 %) of the EST sequences one or more microsatellites were found, 3,723 microsatellite motifs were detected in total. This corresponds to a frequency of 98.8 SSR counts per mega base pairs (counts/Mbp) and one SSR per 10.1 kilo base pairs (kbp). Interestingly, the majority of the SSRs were dimer repeats with 2,095 individual SSRs (56.3 %), whereas only 1,315 (35.3 %) were trinucleotide repeats. This is an unexpected result because trimer repeats are reported to be the most frequent SSRs in plant ESTs 17. Other SSRs were 273 (7.3 %) tetranucleotides, 39 (1.1 %) pentanucleotides and one hexanucleotide SSR (Table 1). This corresponds to SSR counts/Mbp of 55.6 for dimers, 34.9 for trimers, 7.2 for tetramers, 1.0 for pentamers and 0.03 for hexamers, respectively. Among each of the SSR classes the different possible repeat motifs were not evenly distributed (Table 1). The average motif length including compound SSRs was 21.3 base pairs (bp) with 26 % of the motifs being larger than 20 bp.

472 EST sequences contained more than one SSR. 381 ESTs contained two SSRs (80.7 %), 59 contained three (12.5 %), 23 contained four (4.9 %), six contained five (1.3 %) and one EST each contained six, seven respectively 13 SSRs (each 0.2 %). For the generation of PCR markers, multiple SSRs in one EST being separated by less than 100 bp were defined as in 'compound formation' and subsequently handled as one single potential marker. Multiple SSRs being separated by more than 100 bp were further treated as separate single marker loci. In total, 3,171 SSRs were available as potential marker loci, comprising 2,924 SSRs in single and 247 SSRs in compound formation.

Oligonucleotide primer sequences were successfully derived for 2,951 (93 %) of the 3,171 single and compound microsatellites using the Primer3 27 software in batch mode with MISA. Primers could not be designed for SSR motifs comprising too short or inappropriately composed flanking sequences.

For 376 SSRs, primer pairs were synthesized and PCR was performed using the standard Physcomitrella lab strain 'Gransden'. PCR led to successful product amplifications in 329 of 376 SSRs (87.5 %), of which 27 SSR amplicons (7.2 %) yielded longer sequences than expected (size difference from 0.1 up to 2 kbp), most likely due to presence of introns. The frequency of successfully amplified SSRs was more than 20 % higher than described for Hordeum vulgare, where SSR markers were also derived with MISA/Primer3 25. This may be due to the high quality of the clustered Physcomitrella EST database and our specific PCR conditions with a touchdown PCR and a decreased final primer annealing temperature. Moreover, differing intron frequencies with 22 % in Hordeum 25 and 8.4 % in Physcomitrella, as detected in SSR amplicons, may also contribute to the higher rate of successful PCRs in Physcomitrella.

The applicability of the EST-derived SSRs as molecular markers could be proven with side by side amplification of the 376 SSRs in the lab strain 'Gransden' and the french accession 'Villersexel-K3', both of which we are using as parental lines in a genetic mapping approach. 110 markers (33.1 %) were polymorphic between the two accessions (Figure 1).

The informative properties of the EST-derived microsatellites were further evaluated in 64 SSR markers selected from the 376 markers and a collection of 21 worldwide Physcomitrella accessions (Table 2, Figure 2). In total, 238 alleles, including 30 null alleles (13.2%), were detected with a maximum of 7 alleles and an average of 3.7 alleles per SSR. Only one SSR turned out to be monomorphic in the analysed accessions. The polymorphism information content (PIC) based on the 21 Physcomitrella accessions was calculated for all 64 SSR markers. PIC values ranged from 0.0 – 0.78 with a mean PIC value of 0.47 ± 0.17. This PIC is comparably high for EST-derived SSRs, which are generally known to be less informative than SSRs derived from genomic sequences 2829. An explanation for the high amount of SSR polymorphism, including numerous null alleles, could be a high degree of genetic diversity in the studied accessions.

Physcomitrella patens belongs to the family Funariaceae. To gain insight into the interspecies transferability, the 64 Physcomitrella SSR marker were analysed in two further species of the Funariaceae, Physcomitrium sphaericum and Funaria hygrometrica (Table 1, Figure 2). Whereas 51 SSR PCRs (79.7 %) performed well in the more closely related Physcomitrium sphaericum, only 22 (34 %) did so in the more distantly related Funaria hygrometrica (Figure 2). Given this data, we estimate that for Physcomitrium sphaericum about 2,350 Physcomitrella SSRs with the appendant primer pairs can be transferred directly and still about 1,000 Physcomitrella SSRs are applicable in Funaria hygrometrica. Hence, researchers working on closely related moss species may benefit from the Physcomitrella SSRs.

Our results are in accordance with prior reports about interspecies transferability of EST-derived SSRs for numerous seed plants including Triticum aestivum, Hordeum vulgare, Festuca arundinacea, Oryza sativa, Medicago truncatula and Picea taeda, where the transferability decreased with increasing phylogenetic distance and transfer success rates differed from 96 % to 40 % 1725303132.

We aimed to achieve a clear-cut comparison of our results for Physcomitrella with those of other plant species. Unfortunately, to classify surveys on EST-derived SSRs was rather complicated due to large differences in the applied SSR search strategies and by the limited number of analysed species. As a consequence, we identified microsatellites in a comparative side by side search using the MISA software upon the gene indices of 24 plants species. Accounting for the present availability of EST sequences, the species selection encompasses as many major clades of the plant kingdom as possible. In total, two green algae, two true mosses, a fern, a fern palm, the ginkgo tree, two conifers, ten dicots and five monocots were analysed (see Additional file 5). However, for some classes among the plants no appropriate EST datasets were available, e. g. liverworts, hornworts and magnoliidae.

The chosen plant sequence datasets consisted of assembled and non redundant EST sequences ('tentative consensus sequences') and of single non redundant ESTs ('singletons'). While the gene indices were build using the same principal method, there are slight variations in the chosen assembly parameters for each of the four databases they originate from, namely The Institute for Genomic Research (TIGR) 3334, Plant Genome Database (PlantGDB) 3536, COSMOSS 937 and New York Plant Genomics Consortium (NYPG) 383940. Methodological variations in generation of the underlying cDNA sequences, e. g. 5'-, 3'- or full length cDNA enrichment, may have led to additional bias in the datasets. Since EST databases are available for only a limited number of species yet, and because they may be affected by a certain bias, care has to be taken in relating deviating observations to certain plant classes or biological characteristics.

The sequence databases ranged in size from 2 megabases up to over 93 megabases of nucleotide information. The overall average EST sequence length was 733 bp with a standard deviation of 180 bp (see Additional file 5). Significantly deviating sequence lengths were detected for Aquilegia, Arabidopsis and Oryza with long average sequence lengths of 1129, 1184 and 1053 bp and Adiantum, Cycas and Helianthus with short average sequence lengths of 483, 476 and 478 bp, respectively. The long sequence averages indicate a better coverage, whereas the short ones most likely are due to a more fragmentary coverage of the full length cDNAs. The dependency of the long average sequence length on coverage of the individual transcript could be confirmed for the Arabidopdsis GI which contains more than 5,000 full length cDNA sequences 33. Because of the extreme EST sequence length variablity in the different datasets, we believe it to be better to compare EST-derived SSR frequencies by calculating the SSR counts per megabase (counts/Mbp) rather than by SSRs per EST as has sometimes been done previously 192030.

The average GC-content in the 24 datasets was 45.4 % with a standard deviation of 5.3 %. Significantly increased GC-contents were detected for the green algae Chlamydomonas and Mesostigma with 58.3 % and 51.7 %, respectively, the moss Tortula with 53.0 % as well as the grasses (51.1% for Triticum, 52.5 % for Hordeum, 51.5 % for Saccharum and 53.7 % for Oryza). Ginkgo, Aquilegia and Medicago deviated from the average with significantly reduced GC-contents of 40.0 %, 40.0 % and 39.4 % respectively.

The MISA search statistics for all 24 analysed plant species have been made available at COSMOSS 41. The overall frequency of SSRs varied 6.9 fold among the 24 gene indices (Figure 3, see also Additional file 1). The average SSR frequency was 114.7 counts/Mbp with a standard deviation of +/- 60.1 counts/Mbp. The green alga Mesostigma, the fern palm Cycas, both conifers (Pinus and Picea) as well as the monocot Allium cepa were significantly deviating from the average with an extremely reduced SSR frequency of 37.3, 52.9, 41.5, 47.3 and 38.9 counts/Mbp, respectively. The monocot Oryza, however, as well as the dicots Aquilegia and Mesembryanthemum revealed a significantly increased frequency with 258, 240 and 239 counts/Mbp, respectively. The high SSR frequency of rice has been reported in earlier results 18. The hitherto unnoted significantly biased frequencies of the further seven species indicate a much higher variability of EST-derived SSR frequencies among plants than previously reported. Our data for Cycas, Ginkgo, Picea and Pinus seem to indicate that a low SSR frequency is an intrinsic characteristic for gymnosperm gene indices. Among the monocots we ascertained the highest variability in the SSR frequencies with Allium (38.9 counts/Mbp) and Oryza (258.0 counts/Mbp).

Pearson's correlation coefficients (R) of the SSR frequencies with the average GC-content of the analysed species were not significant with -0.12 for dimer SSRs, 0.23 for trimer, 0.24 for other SSRs and 0.04 for the total SSRs. Thus correlations of SSR frequencies with the average GC-content of the analysed species could not be concluded based on our data.

In all datasets, more than 91 % of the detected SSRs were dimer and trimer repeats with the exception of the green alga Mesostigma where they accounted for only 58 % of the total SSRs. Among tetra-, penta- and hexamers, the tetramers are the most abundant in 20 of 24 plants, whereas in the remaining four plant species hexamer SSRs are most abundant (Figure 3, see also Additional file 1).

The ratio of the two most common SSR classes, the dimer and trimer repeats, varied among the datasets. In 15 plants trimer SSRs were more frequent than dimer SSRs. Interestingly, the higher dimer than trimer SSR frequency of Physcomitrella was also true for eight more species among the analysed sample. This dominance of dimer SSRs in nine of 24 plants contradicts the general assumption that trimer SSRs are most frequent in plant ESTs 17.

The average frequency of trimer SSRs is 61.5 counts/Mbp in the 24 datasets with a standard deviation of +/- 42.8 counts/Mbp. Mesostigma, Ginkgo, and Pinus deviate significantly with only 15.7, 15.8 and 18.6 counts/Mbp, respectively. Aquilegia and Oryza deviate significantly with 151.8 and 201.8 counts/Mbp. The extreme dominance of trimer SSRs is the main contribution to the observed overall SSR increase in Aquilegia and even more evidently in Oryza.

The average frequency of dimer SSRs is 46.7 counts/Mbp with a standard deviation of +/- 28.4 counts/Mbp. Mesostigma, Allium and Mesembryanthemum deviate significantly from the average with only 4.3, 12.6 and remarkable 152.4 counts/Mbp, respectively. The observed significant overall SSR increase in Mesembryanthemum is due to its extremely high dimer SSR frequency. In Mesostigma and Allium, significantly reduced trimer as well as dimer SSR frequencies were the reason for the low overall SSR frequency.

In Mesostigma, the observed frequencies were especially remarkable (see Additional file 1 and the Mesostigma-MISA statistics on COSMOSS 41): Although it showed the lowest overall SSR frequency of all analysed plants it revealed by far the highest tetramer SSR frequency with 14.1 counts/Mbp (37.9 % of the total SSRs) and furthermore the second highest pentamer SSR frequency with 2.9 counts/Mbp (7.9 % of the total SSRs). The increased tetramer frequency was mainly due to an increased AATT/TTAA motif frequency (9.9 counts/Mbp) and the increased pentamer frequency due to an increased AAATT/AATTT frequency (2.4 counts/Mbp).

Dimer SSRs are much more variable in length throughout the analysed species than trimer SSRs with the exception of Mesostigma (Figure 4, see also Additional file 2). The average standard deviation of the dimer SSR length with +/-11.9 bp is more than two times larger than that of trimer SSRs with only +/- 5.2 bp. On average, dimer SSRs (18.7 bp +/- 2.9) were significantly longer (p = 0.03) than trimer SSRs (17.5 bp +/- 0.8). The reduced average dimer length in Mesostigma of only 13.0 bp with a small standard deviation (+/- 1.4 bp) might be correlated to the corresponding reduced dimer counts/Mbp. The average dimer length was longest in the two conifers with 24.5 bp, respectively 23.4 bp. Further significant deviations from the average SSR length (see Additional file 2) could neither be correlated with SSR counts/Mbp nor with certain taxonomic clades.

The abundancies of the four canonical non-redundant SSR dimer motifs varied strongly among the analysed plants (Figure 5, see also Additional file 3). The average counts/Mbp for the SSRs motifs AG/CT, AC/GT, AT/TA and CG/GC were 27.4 (+/- 27.6), 7.8 (+/- 7.5), 9.8 (+/- 6.8) and 1.3 (+/- 1.9). The large standard deviations reflect the extremely varying motif abundancies in the analysed species. As an amendment to earlier studies, where AG/CT was found to be the most abundant dimer motif in several plants 12, in our analysis this was persistent for only 16 of the 24 datasets. In Chlamydomonas, AC/GT and in Mesostigma, Ginkgo, Picea, Pinus, Gossypium, Solanum and Allium AT/TA was the most abundant dimer SSR motif. For Picea, the high abundance of AT/TA repeats has been noted in earlier reports 42. In Mesembryanthemum an exceptionally high frequency of the AG/CT motif with 140 counts/Mbp is the main contribution to the species increased overall SSR frequency.

The abundancies of the ten canonical non-redundant SSR trimer motifs varied as well (Figure 6, see also Additional file 4). The average counts/Mbp for the SSRs motifs were 10.8 (+/- 23.3) for CCG/CGG, 4.0 (+/- 3.6) for AGT/ACT, 6.0 (+/- 6.2) for AGG/CCT, 7.1 (+/- 7.2) for AGC/GCT, 4.0 (+/- 3.3) for ACT/ATG, 4.0 (+/- 4.0) for ACG/CGT, 5.8 (+/- 4.6) for ACC/GGT, 3.5 (+/- 2.8) for AAT/ATT, 12.5 (+/- 15.0) for AAG/CTT and 4.4 (+/- 4.8) for AAC/GTT. Like for the dimer motifs, the large standard deviations reflect the extremely varying trimer motif abundancies in the analysed species. In the four grasses, CCG/CGG was the most abundant trimer motif and the same was due for Chlamydomonas. The increased CCG/CGG frequency has been described earlier for grasses and has been related to a high general GC-content 12. In this context the CCG/CGG increase in Chlamydomonas, which has the the highest GC-content (58.3 %) of all 24 gene indices, was consistent, whereas the low CCG/CGG frequency in the alga Mesostigma and the moss Tortula, which also had significantly increased GC-contents of 51.7 and 53.0 %, respectively, did not follow the earlier assumed rule.

The detected low CCG/CGG frequency in Allium contradicts the earlier reported generalisation that an increased CCG/CGG frequency is specific for monocots 121843. An exeptionally high frequency of the SSR motif AAG/CTT contributes to the increased overall SSR frequency in Aquilegia. The canonical motifs encompassing the three stop codons AAT, ATG and AGT were not exceptionally reduced most likely due to the presence of the untranslated regions in the EST databases.

A collection of 21 worldwide Physcomitrella accessions including the standard laboratory accession 'Gransden' plus two related Funariaceae, Physcomitrium sphaericum and Funaria hygrometrica, were selected for analysis (Table 1). The collection includes japanese, australian, african, european and american accessions 44. Details of the collection will be presented elsewhere (Mark von Stackelberg Gabriele Schween, Stefan A. Rensing, Ralf Reski, manuscript in preparation). Plants were grown axenically on solid media and in liquid culture according to the small scale cultivation protocol as described 45. Prior to DNA extraction, plant material was harvested, shock frozen in liquid nitrogen and stored at -80°C.

The DNA extraction protocol is based on Doyle and Doyle 46. Approximately 0.3 g plant material was ground in liquid nitrogen to a fine powder and incubated in 5 mL prewarmed CTAB buffer (2% CTAB, 1.4 M NaCl, 20 mM EDTA, 0.5 % PVP 40, 100 mM Tris [pH 8.0], 0.2 % [v/v] beta-mercaptoethanol) at 65°C for one hour with occasional agitation in a 15 mL polypropylene tube. Afterwards, 5 mL chloroform : isoamylalcohol (24 : 1) was added. Organic and aqeous phase were vigorously mixed followed by separation at 2,500 g for 10 minutes. The aqeous phase was then transferred into a fresh 15 mL polypropylene tube. RNAse A was added to a final concentration of 100 μg/mL and the solution incubated for 30 minutes at 37°C. One tenth volume sodium acetate (3 M, pH 5.2) was added and DNA precipitated overnight with one volume isopropanol at -20°C. DNA was pelleted for 30 min at 2,500 g/4°C. The supernatant was decanted and the pellet incubated in 10 mL washing solution (76 % ethanol, 10 mM ammoniumacetate) for 20 minutes at room temperature (RT). After a short centrifugation, washing buffer was decanted and 70 % ethanol added for a 5 minute incubation at room temperature. After a short spin the supernatant was removed completely and the pellet air dried at room temperature. DNA was dissolved in 200 μL TE buffer overnight at 4°C. The solution was warmed to 65°C for 10 minutes to promote dissolving. DNA quality and concentration was examined by electrophoresis in 0.8 % agarose gels.

For a comparative microsatellite analysis, clustered EST sequence datasets of 24 plant species were chosen (see Additional file 5). For 18 species we downloaded the assembled non-redundant EST-based nucleotide sequence datasets (tentative contigs of the gene indices) from TIGR 3334 and for two species each from PlantGDB 3536 and from NYPG 383940. For the moss Physcomitrella patens we used a non-redundant gene index of the standard lab strain 'Gransden' consisting of high quality, vector-clipped, clustered, assembled and annotated ESTs available via COSMOSS 937. The same filtering and assembly procedure as for Physcomitrella was applied to derive a gene index for all available EST sequences of the moss Tortula ruralis, also available via COSMOSS. All applied databases are summarized in See Additional file 5. The percentage of the nucleotides guanin and cytosin (GC-content) was calculated for the gene indices with the EMBOSS 47 program GEECEE.

SSR motifs were detected using the Perl script MISA 2526. The MISA definition of microsatellites was by unit size (x) and minimum number of repeats (y): 1/100, 2/6, 3/5, 4/5, 5/5, 6/5 (x/y). Mononucleotides were excluded from further analysis. The maximal number of interrupting basepairs in a compound microsatellite was set to 100. MISA detects perfect SSRs only.

For Physcomitrella patens, MISA was used in direct conjunction with the Primer3 software 27 to derive appropriate PCR oligonucleotides. The results of the microsatellite search for Physcomitrella patens are available at COSMOSS 41. A maximum of three different primer pairs per SSR have been provided. Furthermore PCR results for 376 microsatellites, polymorphic SSRs between 'Gransden' and 'Villersexel-K3' and PIC values of 64 SSRs for 21 Physcomitrella accessions have been provided at COSMOSS 41.

The length of the SSR PCR amplicons was set to 100 – 280 bp. Oligonucleotide parameters for Primer3 were set to a length of 18 – 27 bp with an optimum of 20 bp, a GC content of 20 – 80 % with an optimum of 50 %, a melting temperature (Tm) of 57 – 63°C with an optimum of 60°C, and a primer Tm maximum difference of 1°C. SSR marker were amplified in a 20 μL PCR mix containing 2 μL of 10 × RED-Taq-PCR buffer, 0.1 mM dATP, dCTP, dGTP and dTTP, 5 pmol each of two primers, 0.5 Units RED-Taq-Polymerase (SIGMA-Aldrich) and 4 ng plant DNA. Cycling was carried out in a Biometra thermal cycler T1 starting with an initial DNA denaturation at 95°C for 2 min. The first cycle consisted of 30 sec denaturation at 92°C, primer annealing for 30 sec at 60°C and elongation for 30 sec at 72°C. In each of the 10 subsequent cycles the annealing temperature was decreased by 0.7°C. The final 25 cycles consisted of 15 sec denaturation at 92°C, 15 sec primer annealing at 52°C and 30 sec elongation at 72°C.

Electrophoretic size separation of SSR PCR products was performed in 3 % MetaPhor (Cambrex) high resolution agarose gels in 0.5 fold TBE (45 mM Tris-borate, 1 mM EDTA, pH 8.0). We thereby circumvented the lengthy and elaborate use of denaturing polyacrylamide gels. As described in 48 MetaPhor agarose is capable of resolving microsatellite size differences of only two basepairs. The comparably high costs of this agarose can be reduced by reusing it several times.

SSRs were scored visually in the gel according to their amplified fragment size. Different sizes of one marker were scored as different alleles. Same sizes of one SSR and different sizes which could not be visually distinguished were scored as the same allele. Repeated abscence of PCR products in 'Gransden'were scored as PCR failures. Repeated absence of PCR products in the 20 additional accessions were scored as null alleles if PCR worked out in 'Gransden'. The polymorphism information content (PIC value) was calculated according to Botstein et al. 49 as follows:

PIC = 1 − ∑ j = 1 n P i j 2 MathType@MTEF@5@5@+=feaafiart1ev1aaatCvAUfKttLearuWrP9MDH5MBPbIqV92AaeXatLxBI9gBaebbnrfifHhDYfgasaacH8akY=wiFfYdH8Gipec8Eeeu0xXdbba9frFj0=OqFfea0dXdd9vqai=hGuQ8kuc9pgc9s8qqaq=dirpe0xb9q8qiLsFr0=vr0=vr0dc8meaabaqaciaacaGaaeqabaqabeGadaaakeaacqqGqbaucqqGjbqscqqGdbWqcqGH9aqpcqaIXaqmcqGHsisldaaeWbqaaiabdcfaqjabdMgaPjabdQgaQjabikdaYaWcbaGaemOAaOMaeyypa0JaeGymaedabaGaemOBa4ganiabggHiLdaaaa@3EA9@

where Pij is the frequency of the jth allele for marker i and summation extends over n alleles.