Although consensus has now been reached on a general two-locus DNA barcode for land plants, the selected combination of markers (rbcL + matK) is not applicable for ferns at the moment. Yet especially for ferns, DNA barcoding is potentially of great value since fern gametophytes—while playing an essential role in fern colonization and reproduction—generally lack the morphological complexity for morphology-based identification and have therefore been underappreciated in ecological studies. We evaluated the potential of a combination of rbcL with a noncoding plastid marker, trnL-F, to obtain DNA-identifications for fern species. A regional approach was adopted, by creating a reference database of trusted rbcL and trnL-F sequences for the wild-occurring homosporous ferns of NW-Europe. A combination of parsimony analyses and distance-based analyses was performed to evaluate the discriminatory power of the two-region barcode. DNA was successfully extracted from 86 tiny fern gametophytes and was used as a test case for the performance of DNA-based identification. Primer universality proved high for both markers. Based on the combined rbcL + trnL-F dataset, all genera as well as all species with non-equal chloroplast genomes formed their own well supported monophyletic clade, indicating a high discriminatory power. Interspecific distances were larger than intraspecific distances for all tested taxa. Identification tests on gametophytes showed a comparable result. All test samples could be identified to genus level, species identification was well possible unless they belonged to a pair of Dryopteris species with completely identical chloroplast genomes. Our results suggest a high potential of the combined use of rbcL and trnL-F as a two-locus cpDNA barcode for identification of fern species. A regional approach may be preferred for ecological tests. We here offer such a ready-to-use barcoding approach for ferns, which opens the way for answering a whole range of questions previously unaddressed in fern gametophyte ecology.
Citation: de Groot GA, During HJ, Maas JW, Schneider H, Vogel JC, Erkens RHJ (2011) Use of rbcL and trnL-F as a Two-Locus DNA Barcode for Identification of NW-European Ferns: An Ecological Perspective. PLoS ONE 6(1): e16371. doi:10.1371/journal.pone.0016371
Editor: Juergen Kroymann, French National Centre for Scientific Research, Université Paris-Sud, France
Received: September 20, 2010; Accepted: December 20, 2010; Published: January 26, 2011
Copyright: © 2011 de Groot et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Funding: This work was supported by grants from the Schure-Beijerink-Popping Fund (KNAW, SBP/JK/2009-21) and Miquel Fund. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors have declared that no competing interests exist.
The development of universal DNA barcoding markers for land plants is challenging and the exact choice of loci has been heavily debated –. Recently, the Plant Working Group of the Consortium for Barcoding of Life decided on a standard two-locus barcode for all land plants, consisting of portions of the rbcL and matK plastid genes . It was immediately emphasized that this core barcode might have to be augmented with supplementary loci in some groups due to lack of discriminatory power and/or primer universality. Ferns form one such group. While rbcL routinely has been used for studies on fern phylogeny (e.g., , ), species discrimination is sometimes insufficient , . The generation of matK sequences for ferns is currently problematic, because this part of the chloroplast genome underwent a strong restructuring during the evolution of the fern clade . None of the currently existing primer sets are likely suitable for all lineages of land plants ,  and efforts are now focusing on the development of complex primer assays to achieve reliable amplification and sequencing of matK among land plants.
Nonetheless, while the feasibility and necessity of DNA barcoding for general taxonomic purposes has sometimes been criticized , , fern ecology is a typical example of a field for which its main goal, sample identification , would be of high practical and scientific value. Homosporous ferns have two free-living generations. The small and short-lived gametophytic generation plays an important role in the reproduction and dispersal ecology of ferns . Many aspects of its ecology (e.g. tolerance to light and drought stress) differ radically from those of the sporophyte . Studies on wild gametophyte populations and spore banks are therefore as essential for a proper ecological understanding of fern species as those on their sporophytes. Gametophytes however typically show very limited morphological complexity, making identification based on morphological features to species or even genus level often impossible –. Pteridologists are therefore in strong need of a proper alternative method for gametophyte identification.
Schneider & Schuettpelz  tested the principle of DNA barcoding of fern gametophytes using rbcL and successfully identified a cultivated gametophyte as Osmunda regalis. However, whether rbcL shows sufficient variation to allow general identification below genus level remains uncertain , . Moreover, nowadays it is widely accepted that any valid plant barcode will be multi-locus, preferably existing of a conservative coding region like rbcL, in combination with a more rapidly evolving region, which is most likely non-coding . The non-coding trnL intron and trnL-F intergenic spacer (IGS) have been repeatedly suggested for this purpose , ,  and were successfully used by Li et al.  for identification of a mysterious aquatic gametophyte. Besides the technical issues of primer universality and sequence quality and complexity, Schneider & Schuettpelz  mentioned three potential difficulties for any tested marker to overcome: incomplete sampling of the online records to be used as a reference for identification (GenBank/EBI), the occurrence of misidentified and erroneous sequences in these online databases, and the potential inability of the marker to discriminate among species. An additional practical problem for fern gametophyte barcoding is the acquirement of sufficient DNA for multi-locus sequencing from minuscule samples.
We overcame the problems described above and created a reference database of trusted rbcL and trnL-F sequences, in this case for the ferns of NW-Europe. We then tested its potential in the context of primer universality and species discrimination power on a set of previously unidentified gametophyte samples originating from Dutch spore banks. This is the first test of the question whether the combination of rbcL and trnL-F (intron and IGS) possesses all the necessary qualities for standard DNA barcoding across a wide, complex variety of fern taxa. We herewith offer a ready-to-use approach for DNA barcoding of ferns, to be used in any fern ecological study, and show its strength when applied to a regional species pool.
A total of 77 rbcL sequences, belonging to 52 taxa, were selected and incorporated into the reference database (Table S1). For trnL-F, 74 sequences were selected, belonging to 47 taxa. New sequences were produced for a total of 26 species, for 13 species we had to rely completely on GenBank accessions. For a few species, only rbcL accessions were available from GenBank, which explains the difference in total number of included sequences between the two markers. Primer universality proved high for both rbcL and trnL-F. PCR amplification was successful for all tested taxa using the standard protocol, except for a herbarium sample of Matteucia struthiopteris, most likely due to poor DNA quality. PCR-protocols were robust. Sequencing success was >95% for rbcL, and equally high for trnL-F when using internal primers. The slightly higher effort needed for trnL-F in some taxa (mainly belonging to the families Blechnaceae and Thelypteridaceae) resulted from the presence of a homopolymer C-repeat in the intergenic spacer.
Maximum parsimony analysis of the total dataset yielded the phylogram shown in Figure 1 (50% majority rule bootstrap consensus). Topology is mostly consistent with currently accepted pteridophyte phylogenies based on the same markers , , . More importantly, 100% of the genera and 100% of the included species form their own well-supported monophyletic clade (bootstrap support >70%). Mean minimal interspecific P-distance per species (distance to nearest neighbour) was 0.031. Distances were however very skewed towards lower values, which is in line with general results for plastid loci in land plants . Within genera, mean minimal interspecific P-distances were smallest for Polystichum (0.006) and Dryopteris (0.008). These genera had relatively short internal branches but still generally showed high bootstrap support (Figure 1).
Unrooted phylogram of the 50% majority rule bootstrap consensus tree from an analysis of the combined rbcL and trnL-F sequence data for 46 fern taxa occurring in NW-Europe. Bootstrap support values are given with each node. In case of multiple accessions per taxon, a sample number was added behind the taxon name.
Figure 2 presents an overview of sequence divergences based on a comparison between maximal intraspecific and minimal interspecific distances for all species for which multiple samples were included in the analysis. Exact values are given in Table S2. For all 24 taxa tested, minimal interspecific distances are clearly higher than maximal intraspecific distances. Not surprisingly, the smallest ratio was found for the subspecies of A. trichomanes, but even there minimal interspecific distance was 1.3 times bigger than maximal intraspecific distance.
Sequence divergence across all 24 taxa for which sequences of multiple individuals were available. Divergence is given as the relation between the uncorrected maximal intraspecific and minimal interspecific P- distances. Along the black line both distance values equal each other.
To test the applicability of the generated database for fern species identification in ecological field studies, we used material of a total of 88 gametophytes which resulted from two different soil spore bank analyses [G.A. de Groot, unpublished data]. In one experiment soil samples were taken from different depths in four forests in the Dutch IJsselmeer polders. More than 25 fern species have been recorded for these forests, which were planted on the bare land of a former sea bottom and over the last decades acted as giant diaspore traps . Samples were spread out in the greenhouse of the botanical gardens of Utrecht University and 86 of the resulting gametophytes were selected for analysis. Two additional gametophytes resulted from two different soil cores taken from the IJsselmeer lake bottom in spring 2008. Extraction of DNA from tiny amounts of gametophyte material proved well possible using a normal commercial extraction kit. Only two out of 88 attempts failed to produce a workable DNA solution. Table 1 gives an overview of the performance of sequencing and all necessary preceding steps. rbcL amplification was unproblematic and successful for all samples, in four random cases trnL-F amplification failed completely for unknown reasons. In some cases however, amplified bands were weak even after a nested PCR was performed and sequencing produced only a short fragment. Still, usable rbcL and trnL-F sequences were eventually produced for 97% and 99% of the samples respectively. Sequence coverage was around 84% for rbcL and 89% for trnL-F. The two tiny gametophyte samples extracted from lake-bottom sediments yielded usable DNA and were successfully sequenced.
Identification using various methods classified the tested samples into six taxonomic groups. Identification results are given in Table 2. Identification at genus level succeeded for all samples using maximum parsimony analysis. 44 out of 79 samples with usable sequences of rbcL and trnL-F (56%) could easily be assigned to a single species. Maximum parsimony analyses using the rbcL + trnL-F reference database compiled in this study successfully identified samples as Dryopteris dilatata, Athyrium filix-femina or Thelypteris palustris. For D. dilatata and A. filix-femina, online BLASTn searches using trnL-F already successfully yielded the same identification, while searches using rbcL resulted in multiple options. BLAST analysis failed on samples identified as T. palustris in the phylogenetic analysis, either because of lack of resolution (rbcL) or because the species was not present in the online records (trnL-F). Two samples that were identified as Polystichum setiferum s.l. using the plastid markers could be identified as P. aculeatum using the additional PgiC marker. Parsimony analyses classified the remaining 35 samples (44%) as either D. oreades s.l. or D. carthusiana s.l. (see materials and methods for an explanation on used species groups); further identification to species name remained impossible using this method since the species have identical cpDNA. In a limited number of cases, bootstrap support turned out to be insufficient (bootstrap values 50–70%) to distinguish D. dilatata and D. expansa from D. carthusiana s.l. in the parsimony analyses (Table 2). PgiC was not able to resolve these complexes either: sequences of D. carthusiana, D. dilatata and D. cristata differ from D. filix-mas and D. affinis, but are identical among species in both combinations. The fact that BLAST results for rbcL or trnL-F do sometimes give only one or two options is simply due to lack of online accessions for the other species.
Selecting an Appropriate Barcode for Ferns
In line with previous results of for land plants in general  and for bryophytes , it proved possible to amplify and sequence trnL-F in ferns using a limited set of universal and very reliable primers, even when using suboptimal starting material. Primer universality of rbcL was already shown by various authors (e.g., , ) and is confirmed by our results. As in many plastid spacer regions in plants (e.g., –), a mononucleotide repeat of varying length is also present in fern trnL-F sequences. Such microsatellites potentially lead to reduced sequence quality and the use of any spacer region for plant DNA barcoding has therefore been criticized . The use of a single pair of internal primers however successfully solved this problem in our case. Fazekas et al. ,  showed that the use of proofreading enzymes fused to nonspecific dsDNA binding domains can also greatly improve sequence quality in case of mononucleotide repeats.
While Taberlet et al.  concluded that the trnL intron generally shows a low barcoding resolution among land plants, variation of the combined trnL and trnL-F spacer region appears to be surprisingly high for ferns. When combined with rbcL, both distance- and character-based tests showed a 100% resolving power at both genus and species level (Figure 1 and 2) for all included taxa with different chloroplast genomes. Although based on limited sample sizes, the calculated sequence divergences indicate the presence of a clear ‘barcoding gap’ across European fern taxa: enough difference between inter- and intraspecific distances to discriminate a species from its nearest neighbours . Maximum parsimony analysis yielded the same well-supported differentiation, even though testing for it in a fundamentally different way. When adding this to its highly universal primers and robust PCR conditions, we conclude that trnL-F might be a valuable substitute for the problematic matK spacer , , and in combination with rbcL possesses al the necessary qualities to form a powerful barcode for species identification of pteridophytes, at least in NW-Europe.
The fern chloroplast genome is maternally inherited and hence non-recombining . Recent allotetraploids therefore are likely to still share an identical chloroplast genome with the diploid parent that functioned as chloroplast donor or with a related polyploid sharing the same donor. This makes it often impossible to distinguish such related species based on the cpDNA, as sequences will be identical for any chloroplast marker. The same problem can be expected for apomictic aggregates. We explicitly chose to ignore this issue in our performance tests (see materials and methods), since plant DNA barcoding efforts are currently restricted to the single-copy cpDNA , ,  and similar problems are thus to be expected for any candidate locus. The uniparental inheritance of the chloroplast is however an important issue to address, since this focus on the cpDNA can clearly reduce identification success in plant groups like temperate ferns, which are known for a high frequency of hybridization among closely related species  and a relatively frequent formation of apomictic lineages . Application of a nuclear marker may often be useful for valid species discrimination in case of allopolyploid species, as well as for proper identification of hybrid individuals. We therefore suggest a further search for nuclear markers which are cheap and easy to apply. An example is the nuclear PgiC gene encoding cytosolic phosphoglucose isomerase . PgiC has discriminatory power for European polystichoids, as shown by Bremer & De Groot . They reported a simple identification method based on band lengths on an agarose gel, which discards the time and money consuming cloning step often discouraging barcoders from using nuclear loci . However, for some other genera PgiC proved less useful, as discriminatory power was insufficient a. Clearly, more effort is needed to find alternative, more universal nuclear markers with sufficient discriminatory power.
Methods for Species Assignment
We assessed both barcoding resolution of the tested markers and identity of the test samples by use of two different methods of comparison: a character-based parsimony analysis (checking for well-supported monophyly of species, with or without the presence of a query sequence) and a distance-based analysis (checking for (pair-wise) sequence similarity based on the number of identical nucleotide positions). Distance-based comparison methods are fast and can provide a rank list of nearest neighbours accompanied by a simple score of similarity . This is useful when comparing samples with a limited reference database from which the correct species might be missing. Parsimony analysis might then yield a monophyletic clustering with the wrong species if other species are sufficiently different, while distance values will most likely still indicate a small percentage of dissimilarity. The other way around however, number of similar positions (similarity) is not always an indication of relatedness : species with equal similarity scores might still be different at specific positions. Such character differences are taken into account when testing for well-supported monophyly based on parsimony. Additionally, the use of bootstrapping provides a measure of the expected mis-assignment due to local homoplasy . Maximum parsimony analysis and bootstrapping are however computationally demanding, which might be a problem for barcoding applications. In our view however, both methods are complementary and we therefore advocate a joint use both in case of assessing the optimal set of barcodes and in case of actual species assignment tests by barcoding.
Proper sequence alignment is essential for any assignment method based on phylogenetic analysis, either distance- or parsimony-based . Creating an acceptable global alignment above genus level is often impossible for loci with relatively high variation , , thus limiting the use of this type of methods for species assignment. An easy option is to keep problematic (highly diverging) sequences out of the alignment. In our case, sequences of five (evolutionary older) species could not be aligned with the rest, but the resolving power of the less-variable rbcL locus was already high enough to form well-supported monophyletic species. This once more indicates the strength of using a hierarchical multi-locus strategy with rbcL as a backbone locus, as proposed by the CBOL Plant Working Group . In practice, a preliminary species assignment for most NW-European ferns will be already possible based on rbcL and might be verified or specified by comparison with trnL-F sequences. This can either be done manually, by use of pair-wise distance analysis, or by performing a second parsimony analysis using data of a single genus only. At the same time, current advances in comparative genomics induce innovations in bioinformatic techniques for automated analysis of large and diverse sequence databases . Newly developed tools for (multiple) sequence alignment are more dynamic, using adjustable local algorithms and instant global replotting, and as such are better suited for alignment of distantly related sequences , .
Application of DNA Barcoding in Fern Ecology
In part of the BLAST searches performed for test sample identification, MPI values were equally high for some Asian and American species as for the (more likely) European species. Such species were specifically ignored here, but these results indicate that DNA barcoding of ferns might be more problematic on a global scale. Testing the rbcL + trnL-F barcode on a wider scale and/or a further search for (additional) fern barcodes cannot be avoided. Finding universal primers for the exhaustive variation in closely related fern species will be extremely difficult. For ecological applications a regional barcoding approach might therefore be the best choice, as its goal is simple: finding the most likely identification for a specimen encountered, given its local environment. For this purpose, a regional approach is most efficient, as it enables the use of a restricted reference database of trusted sequences for all species of a specific region or ecological community , . Our study approached such a database for the native ferns of North-western Europe. We acknowledge that restricting the reference set to a certain region neglects the possibility of invasions and garden escapes (although typical examples might be included in the database), but the same is true for common practice in morphological identification.
Even though 44% of the tested samples could not be resolved to a single species because they belonged to one of the polyploidy complexes described above, all samples could be resolved to genus level. The fact that even difficult samples with low quantities of degraded DNA, like gametophytes resulting from spores of the long-term lake-bottom spore bank, could successfully be sequenced and identified opens the way for answering a whole range of previously ignored questions in the field of fern gametophyte ecology. Gametophytes derived from samples of the deeper soil layers typically were small and unhealthy. Such individuals would never have produced sporophytes of identifiable size, thus making them unidentifiable by conventional morphological methods. However, these gametophytes represent the few long-term surviving spores in the soil and potentially yield valuable information about past vegetation composition and diversity. Likewise, morphological identification is inapplicable when studying population biology or reproductive success of gametophytes in the field. In such cases barcoding is a very efficient and valuable technique. Already, some ecologists used a barcoding approach to identify a specific unknown plant sample for practical purposes , , . We now offer a complete and ready-to-use approach for wider application of DNA barcoding in ecological studies on ferns in north-western Europe.
Materials and Methods
Sequence Database for DNA Barcoding
Taxon sampling and origin of sequences.
52 taxa, representing 23 genera, were included in our reference database and together cover the diversity of terrestrial ferns occurring and sporulating in the wild in North-western Europe. This region harbours the taxa most likely to reach the area from which the samples of our case study (see below) originated, and as such is a typical example of a regional approach to be used by ecologists. We defined “North-western Europe” as comprising the British Isles, The Netherlands, Belgium, Luxemburg, Northern France and Western Germany (thus excluding the Nordic countries), and a species list for this area was compiled from Stace , Lambinon et al.  and Van der Meijden . Fresh water ferns and horsetails were excluded since they fell outside the scope of the spore bank studies for which the database was originally developed. In order to obtain a set of trusted reference sequences, we used self-produced sequences for each taxon whenever possible. Sequences were either obtained from previous studies of the authors or sequenced de novo from freshly collected leaf material or herbarium specimens. Identifications were checked by experienced fern taxonomists. GenBank accessions were used for a small number of taxa for which no properly identified material was available. Additional GenBank accessions were added to the database when available in order to represent multiple individuals per species. A full list of sequences present in the database and their origin is given in Table S1.
DNA extraction, amplification and sequencing.
Freshly collected specimens were stored on silica prior to extraction. DNA was extracted using the GenElute™ Plant Genomic DNA Miniprep Kit (Sigma-Aldrich, St. Louis, USA) following the manufacturer's protocol. We amplified two plastid regions, a c.1300 bp. long fragment of the rbcL gene (using 1FN and 1361R; Table 3) and a c. 900 bp. long combination of the trnL gene and trnL-trnF intergenic spacer together referred to as the trnL-F region (using FERN-1 and ‘f’; Table 3). For samples with problematic amplification we additionally used internal primers (Table 3). For rbcL, this concerned some seemingly random samples, for trnL-F a few taxa were inherently difficult because a homopolymer C-repeat was present in the intergenic spacer.
DNA amplification was performed in 25 µl reactions containing 1× buffer, 3.5 µM MgCl2, 0.1 µM primers, 200 µM dNTP, 1% BSA and DMSO, 0.25 U RedTaq™ polymerase (Sigma-Aldrich, St. Louis, USA) and 1.5 µl of DNA template. Thermal cycling conditions for rbcL were: 50 s at 96°C, 30 cycles of 50 s at 96°C, 50 s at 53°C and 90 s at 72°C, and a final extension of 7 min at 72°C. For trnL-F the following protocol was used: 5 min at 95°C. 30 cycles of 5 s at 95°C, 30 s at 53°C and 90 s at 72°C, and a final step of 10 min at 72°C.
PCR products were purified on a 96-wells plate (Thermo Fisher Scientific, Waltham, USA) using gel filtration with Sephadex™ G-50 (GE Healthcare, Uppsala, Sweden). Sequencing was performed by Macrogen (Seoul, Korea and Amsterdam, The Netherlands) using the amplification primers (except for samples amplified with internal primers, which were short enough to sequence with forward primers only). All obtained sequences are available in GenBank (accession numbers listed in Table S1).
Sequence alignment and data analysis.
Sequences were edited and assembled in SeqMan 4.0 (DNAStar Inc., Madison, USA) and manually aligned using PAUP* 4.0b10 . Large indels were coded using simple indel coding as described by Simmons & Ochoterena .
To estimate species discrimination based on the chosen markers, we used a combination of two complementary measures: a comparison of interspecific and intraspecific distance values and a tree-based analysis. The tree-based strategy was performed using maximum parsimony rather than using a neighbour joining approach, since besides pure distance values also specific shared characters can prove useful in discriminating different lineages for barcoding purposes , . Most-parsimonious trees were generated in PAUP* 4.0  using random taxon additions, TBR swapping and equal weights. Heuristic bootstrap analysis  was performed with 1000 bootstrap replicates, 100 random addition cycles per bootstrap replicate, TBR swapping and equal weights. Species discrimination success was then based on monophyly: a species was successfully resolved when forming a monophyletic group with sufficient bootstrap support (bootstrap value >70%). Uncorrected minimal interspecific P-distances were calculated for all included species. Species discrimination based on distance values was tested for all species with multiple sequences in the database, according to CBOL guidelines . Discrimination was considered successful if the minimal interspecific distance involving a species was larger than its intraspecific distance .
As previous studies showed that the resolving power of multi-locus barcodes is almost always higher than that of single-locus barcodes , , , , all testing was based on the combined dataset of rbcL and trnL-F. Four taxa (Botrychium lunaria, Botrychium matricariifolium, Ophioglossum vulgatum and Osmunda regalis) could not be properly included in the trnL-F alignment due to large sequence divergence (differences at so many positions that any alignment would be unreliable). Evidence for monophyly of these species thus was based on rbcL sequences only, but P-distances could be calculated from the total dataset.
As explained in the discussion, recent polyploids may share their complete chloroplast genome with their diploid chloroplast donor or with a related polyploid originating from the same donor, which makes them undistinguishable by any cpDNA marker. The same problem is seen in apomicts. Our database includes several examples of such complexes. Dryopteris cristata (L.) A. Gray and D. carthusiana (Vill.) H.P. Fuchs belong to a complex of allotetraploids sharing ancestral diploid genomes . Stein et al.  showed that both species have identical cpDNA, which they most likely inherited from a shared extinct parent called D. semicristata . The tetraploid Dryopteris filix-mas (L.) Schott is closely entangled with the apomictic D. affinis (Lowe) Fraser-Jenk  aggregate and the species presumably share parents, one of which being D. oreades Fomin , . Polystichum aculeatum (L.) Roth is an allotetraploid derivative from the cross between P. setiferum (Forssk.) Woynar, which donated the cpDNA, and P. lonchitis (L.) Roth . Asplenium adiantum-nigrum L. is interpreted as a segmental allotetraploid (at least in Europe) that involved A. onopteris L as the maternal parent . Finally, A. fontanum is most likely donor of the cpDNA to A. foreziense . For a proper analysis of the discriminatory power of the chosen markers, we chose to collapse the taxonomy in each of these cases and name each member of the group after the (also included) diploid cpDNA donor . For D. cristata and D. carthusiana the donor could not be included and their sequences were instead named D. carthusiana s.l.. An overview of the performed merges can be found in Table S3.
A Case Study: Identification of Fern Gametophytes from Spore Bank Analyses
We used material of a total of 88 gametophytes which resulted from two different soil spore bank analyses [G.A. de Groot, unpublished data]. Gametophytes ranged in size from c. 1 mm2 to 1.5 cm2. The smallest gametophytes resulted from the deeper forest soil layers (>15 cm) and lake bottom samples, and were sampled for DNA extraction at small size because they didn't grow any further and started to look unhealthy. Selection was based on an optimal sampling scheme for the spore bank analysis, but we made sure that the observed morphological variation (mainly presence of glandular hairs) was covered. Selected gametophytes were rinsed with water to avoid contamination and dried on silica in a 1.5 ml Eppendorf tube. Because of the very small size of the gametophytes, it was impossible to use only part of the material and use the rest for further culturing. Instead, complete individuals were grinded with mortar and pestle and a tiny amount of sand, after which DNA was extracted using the GenElute™ Plant Genomic DNA Miniprep Kit (Sigma-Aldrich, St. Louis, USA) following the manufacturer's protocol, but using 10% of all volumes and a final elution in 30 µl water . For 86 out of 88 gametophytes this resulted in dissolved DNA of adequate concentration to be used in multiple PCR reactions. We prefer this method above tissue-direct PCR  since in case of tiny samples the latter leaves no option for retrial or application of multiple markers. rbcL and trnL-F were amplified in respectively three and two parts using the primers listed in Table 3 (with the earlier described reaction mixture and cycling protocol) and sequenced by Macrogen (Seoul, Korea) using the forward primers. Acquired sequences were edited and assembled in SeqMan 4.0 (DNAStar Inc., Madison, USA).
BLASTn searches were applied to all produced sequences using the available online databases (i.e. GenBank, EMBL). Since a considerable part of the existing online accessions involved partial sequences, BLAST results were ranked by maximal percent identity (MPI) instead of maximal bit core. Since here we specifically test a regional approach for ecological purposes and non-European species were considered unlikely to be the correct identification, such species were ignored in all output. Identification at genus level was considered successful when all hits with MPI scores >95% involved a single genus. Species identification was considered successful only when the highest MPI included a single European species and scored above 95%.
Additionally, all sequences were manually aligned with the above described reference database using PAUP* 4.0b10 . Heuristic parsimony analyses using the combined rbcL and trnL-F database were performed separately for every unknown gametophyte by including one individual at a time in the analysis. Bootstrapping  was performed with 20 bootstrap replicates, 100 random addition cycles per bootstrap replicate, TBR swapping and equal weights. Genus identification was considered successful when the unknown gametophyte formed a monophyletic group together with all members of a single genus, with a bootstrap support >70%. An equal strategy was applied for identification at species level.
Two samples could only be identified as Polystichum setiferum s.l. using above described methods since P. setiferum and P. aculeatum (both occurring in the polder forests) share the same cpDNA. The nuclear PgiC gene for cytosolic phosphoglucose isomerase  was applied to discriminate up to species level, following a method based on band size differences described by Bremer & De Groot . The length of the amplified fragment for PgiC differs between Polystichum lonchitis and Polystichum setiferum, resulting in a difference in band lengths on agarose gel between the two species, and a double banding pattern in their hybrid derivative, Polystichum aculeatum. These patterns easily discriminate the three species and can be used for identification of unknown samples. New primers were developed for this purpose (Table 3). Thermal cycling protocols were copied from Ishikawa et al. .
rbcL and trnL-F sequence information. Sequence origin (O = sequenced by authors in previous study, G = downloaded from Genbank, N = sequenced for this study), voucher information (herbarium, collection number. EB = Ecology & Biodiversity Group, Utrecht University voucher collection), collector, publication information and Genbank accession numbers of the rbcL and trnL-F sequences utilized for this study.
Maximal intraspecific genetic distances, minimal genetic distances towards the nearest neighbour and their ratio. Data are presented for all taxa with multiple individuals in the dataset. Ratios are calculated as maximal interspec. distance/minimal intraspec. distance. All distances are uncorrected P-distances based on the combined rbcL and trnL-F sequence data. N = number of individuals in the dataset.
Used species complexes. Overview of species merged into a single complex because of identical chloroplast genomes. All sequences were named after the diploid cpDNA donor, unless this donor could not be included (Dryopteris “semicristata”, see text).
We thank Sander van Hal, Jasper Wubs, Yunfan Liu, Mariska Meijer, Irene Blotenburg and Ellen Jochems for their help in the laboratory, and Piet Bremer, Maarten Christenhusz, Patrick Acock and Bas van de Riet for their help with sample collection and identification. Thanks to Staatsbosbeheer, Natuurmonumenten and Het Flevolandschap for allowing us on their properties. Special thanks go to Koos Swart of the Netherlands Institute of Ecology (NIOO-KNAW) for drilling soil cores from the IJsselmeer lake bottom for use in spore bank analyses.
Conceived and designed the experiments: GAG RHJE HS. Performed the experiments: GAG HS JCV JWM. Analyzed the data: GAG. Wrote the paper: GAG RHJE HJD HS.
- 1. Chase MW, Salamin N, Wilkinson M, Dunwell JM, Kesanakurthi RP, et al. (2005) Land plants and DNA barcodes: short-term and long-term goals. Philos Trans R Soc London [Biol] 360: 1889–1895.
- 2. Kress WJ, Wurdack K, Zimmer EA, Weigt L, Janzen DH (2005) Use of DNA barcodes to identify flowering plants. Proc Natl Acad Sci U S A 102: 8369–8374.
- 3. Fazekas AJ, Burgess KS, Kesanakurti PR, Graham SW, Newmaster SG, et al. (2008) Multiple multilocus DNA barcodes from the plastid genome discriminate plant species equally well. PLoS One 3: e2802.
- 4. Hollingsworth PM, Forrest LL, Spouge JL, Hajibabaei M, Ratnasingham S, et al. (2009) A DNA barcode for land plants. Proc Natl Acad Sci U S A 106: 12794–12797.
- 5. Pryer KM, Schuettpelz E, Wolf PG, Schneider H, Smith AR, Cranfill R (2004) Phylogeny and evolution of ferns (Monilophytes) with a focus on the early leptosporangiate divergences. Am J Bot 91: 1582–1598.
- 6. Schneider H, Russell SJ, Cox CJ, Bakker F, Henderson S, et al. (2004) Chloroplast phylogeny of asplenioid ferns based on rbcL and trnL-F spacer sequences (Polypodiidae, Aspleniaceae) and its implications for biogeography. Syst Bot 29: 260–274.
- 7. Jansen T, Schneider H (2005) Exploring the evolution of humus collecting leaves in drynarioid ferns (Polypodiaceae, Polypodiidae). Plant Syst Evol 252: 175–197.
- 8. Schneider H, Ranker TA, Russell SJ, Cranfill R, Geiger JM, et al. (2005) Origin of the endemic fern genus Diellia coincides with the renewal of Hawaiian terrestrial life in the Miocene. Proc R Soc Lond [Biol] 272: 455–460.
- 9. Duffy AM, Kelchner SA, Wolf PG (2009) Conservation of selection on matK following an ancient loss of its flanking intron. Gene 438: 17–25.
- 10. Hollingsworth ML, Clark AA, Forrest LL, Richardson J, Pennington RT, et al. (2009) Selecting barcoding loci for plants: evaluation of seven candidate loci with species-level sampling in three divergent groups of land plants. Mol Ecol Res 9: 439–457.
- 11. Li F-W, Tan BC, Buchbender V, Moran RC, Rouhan G, et al. (2009) Identifying a mysterious aquatic fern gametophyte. Plant Syst Evol 281: 77–86.
- 12. Will KP, Mishler BD, Wheeler QD (2005) The perils of DNA barcoding and the need for integrative taxonomy. Syst Biol 54: 844–851.
- 13. Spooner DM (2009) DNA barcoding will frequently fail in complicated groups: an example in wild potatoes. Am J Bot 96: 1177–1189.
- 14. Hebert PDN, Gregory TR (2005) The promise of DNA barcoding for taxonomy. Syst Biol 54: 852–859.
- 15. Farrar DR, Dassler C, Watkins JE Jr, Skelton C (2008) Gametophyte Ecology. In: Ranker TA, Haufler CH, editors. Biology and Evolution of Ferns and Lycophytes. Cambridge: Cambridge University Press. pp. 222–256.
- 16. Watkins JE Jr, Mack MK, Mulkey SS (2007) Gametophyte ecology and demography of epiphytic and terrestrial tropical ferns. Am J Bot 94: 701–708.
- 17. Schneller JJ (1975) Untersuchungen an einheimischen Farnen, insbesondere der Dryopteris filix-mas-Gruppe, 3. Teil: Ökologische Untersuchungen. Ber Schweiz Bot Ges 85: 110–159.
- 18. Cousens MI (1981) Blechnum spicant, habitat vigor of optimal, marginal, and disjunct populations, and field observations of gametophytes. Bot Gaz 142: 251–258.
- 19. Dyer AF (1994) Natural soil spore banks – can they be used to retrieve lost ferns? Biodiversity Conserv 3: 160–175.
- 20. Schneider H, Schuettpelz E (2006) Identifying fern gametophytes using DNA sequences. Mol Ecol Notes 6: 989–991.
- 21. Kress WJ, Erickson DL, Jones FA, Swenson NG, Perez R, et al. (2009) Plant DNA barcodes and a community phylogeny of a tropical forest dynamics plot in Panama. Proc Natl Acad Sci U S A 106: 18627–18632.
- 22. Taberlet P, Gielly L, Pautou G, Bouvet J (1991) Universal primers for amplification of three non-coding regions of chloroplast DNA. Plant Mol Biol 17: 1105–1109.
- 23. Taberlet P, Coissac E, Pompanon F, Gielly L, Miquel C, et al. (2006) Power and limitations of the chloroplast trnL (UAA) intron for plant DNA barcoding. Nucleic Acids Res 35: e14.
- 24. Schuettpelz E, Pryer KM (2007) Fern phylogeny inferred from 400 leptosporangiate species and three plastid genes. Taxon 56: 1037–1050.
- 25. Geiger JMO, Ranker TA (2005) Molecular phylogenetics and historical biogeography of Hawaiian Dryopteris (Dryopteridaceae). Mol Phylogenet Evol 34: 392–407.
- 26. De Groot GA, Erkens RHJ, During HJ (2008) WODAN: de invloed van verspreidings- en vestigingsmogelijkheden op het ontstaan van biodiversiteit. Gorteria 33: 59–60.
- 27. Liu Y, Yan H-F, Cao T, Ge X-J (2010) Evaluation of 10 plant barcodes in Bryophyta (Mosses). J Syst Evol 48: 36–46.
- 28. Shaw J, Lickey EB, Beck JT, Farmer SB, Liu E, et al. (2005) The tortoise and the hare II: relative utility of 21 noncoding chloroplast DNA sequences for phylogenetic analysis. Am J Bot 92: 142–166.
- 29. Shaw J, Lickey EB, Schilling EE, Small RL (2007) Comparison of whole chloroplast genome sequences to choose noncoding regions for phylogenetic studies in angiosperms: the tortoise and the hare III. Am J Bot 94: 275–288.
- 30. Erkens RHJ, Maas JW, Couvreur TLP (2009) From Africa via Europe to South America: migrational route of a species-rich genus of Neotropical lowland rain forest trees (Guatteria, Annonaceae). J Biogeography 36: 2338–2352.
- 31. Devey DS, Chase MW, Clarkson JJ (2009) A stuttering start to plant DNA barcoding: microsatellites present a previously overlooked problem in non-coding plastid regions. Taxon 58: 7–15.
- 32. Fazekas AJ, Steeves R, Newmaster SG (2010) Improving sequencing quality from PCR products containing long mononucleotide repeats. Biotechniques 48: 277–285.
- 33. Fazekas AJ, Steeves R, Newmaster SG, Hollingsworth PM (2010) Stopping the stutter: Improvements in sequence quality from regions with mononucleotide repeats can increase the usefulness of non-coding regions for DNA barcoding. Taxon 59: 694–697.
- 34. Gastony GT, Yatskievych G (1992) Maternal inheritance of the chloroplast and mitochondrial genomes in cheilanthoid ferns. Am J Bot 79: 716–722.
- 35. Consortium for the Barcoding Of Life (2007) Non-COI Barcode Regions - Guidelines for CBOL Approval. URL http://www.barcoding.si.edu [accessed 6 may 2010].
- 36. Barrington DS, Haufler CH, Werth CR (1989) Hybridization, reticulation and species concepts in the ferns. Am Fern J 79: 55–64.
- 37. Lovis JD (1978) Evolutionary patterns and processes in ferns. Adv Bot Res 4: 229–415.
- 38. Ishikawa H, Watano Y, Kano K, Ito M, Kurita S (2002) Development of primers sets for PCR amplification of the PgiC gene in ferns. J Plant Res 115: 65–70.
- 39. Bremer P, De Groot GA (2010) Polystichum ×bicknellii (Christ) Hahne, nieuw in Nederland. Gorteria 34: 146–152.
- 40. Altschul SF, Gish W, Miller W, Myers EW, Lipman DJ (1990) Basic local allignment search tool. J Mol Biol 215: 403–410.
- 41. Koski LB, Golding GB (2001) The closest BLAST hit is often not the nearest neigbor. J Mol Evol 52: 540–542.
- 42. DeSalle R, Egan MG, Siddall M (2005) The unholy trinity: taxonomy, species delimitation and DNA barcoding. Philos Trans R Soc London [Biol] 360: 1905–1916.
- 43. Ptacek T, Sell SM (2005) A tiered approach to comparative genomics. Brief Funct Genomic Proteomic 4: 178–185.
- 44. Ovcharenko I, Loots GG (2003) Finding the needle in the haystack: computational strategies for discovering regulatory sequences in genomes. Curr Genomics 4: 557–568.
- 45. Ovcharenko I, Loots GG, Hardison RC, Miller W, Stubbs L (2004) zPicture: Dynamic alignment and visualization tool for analyzing conservation profiles. Genome Res 14: 472–477.
- 46. Stace CA (1997) New Flora of the British Isles. 2nd edn. Cambridge: Cambridge University Press. 1232p p.
- 47. Le Clerc-Blain J, Starr JR, Bull RD, Saarela JM (2010) A regional approach to plant DNA barcoding provides high species resolution of sedges (Carex and Kobresia, Cyperaceae) in the Canadian Arctic Archipelago. Mol Ecol Res 10: 69–91.
- 48. Van de Wiel CCM, Van der Schoot J, Van Valkenburg JLCH, Duistermaat H, Smulders MJM (2009) DNA barcoding discriminates the noxious invasive plant species, floating pennywort (Hydrocotyle ranunculoides L.f.), from non-invasive relatives. Mol Ecol Res 9: 1086–1091.
- 49. Pryer KM, Schuettpelz E, Huiet L, Grusz AL, Rothfels CJ, et al. (2010) DNA barcoding exposes a case of mistaken identity in the fern horticultural trade. Mol Ecol Res. doi:10.1111/j.1755-0998.2010.02858.x.
- 50. Lambinon J, De Langhe J-E, Delvosalle L, Duvigneaud J (1998) Flora van België, het Groothertogdom Luxemburg, Noord-Frankrijk en de aangrenzende gebieden (Pteridofyten en Spermatofyten). 3rd edn. Meise: Nationale Plantentuin van België. 970p p.
- 51. Van der Meijden R (2005) Heukels' Flora van Nederland. 22nd edn. Groningen: Wolters-Noordhoff. 685p p.
- 52. Little DP, Barrington DS (2003) Major Evolutionary Events in the origin and diversification of the fern genus Polystichum (Dryopteridaceae). Am J Bot 90: 508–514.
- 53. Wolf PG, Soltis PS, Soltis DE (1994) Phylogenetic relationships of Dennstaedtioid ferns: evidence from rbcL sequences. Mol Phylogenet Evol 3: 383–392.
- 54. Schuettpelz E, Schneider H, Huiet L, Windham MD, Pryer KM (2007) A molecular phylogeny of the fern family Pteridaceae: assessing overall relationships and the affinities of previously unsampled genera. Mol Phylogenet Evol 44: 1172–1185.
- 55. Trewick SA, Morgan-Richards M, Russell SJ, Henderson S, Rumsey FJ, et al. (2002) Polyploidy, phylogeography and Pleistocene refugia of the rockfern Asplenium ceterach: evidence from chloroplast DNA. Mol Ecol 11: 2003–2012.
- 56. Swofford DL (2003) PAUP*. Phylogenetic Analysis Using Parsimony (* and Other Methods). Version 4.0b10. Sunderland: Sinauer Associates.
- 57. Simmons MP, Ochoterena H (2000) Gaps as characters in sequence-based phylogenetic analyses. Syst Biol 49: 369–381.
- 58. Sarkar IN, Planet PJ, DeSalle R (2008) CAOS software for use in character-based DNA barcoding. Mol Ecol Res 8: 1256–1259.
- 59. Felsenstein J (1985) Confidence limits on phylogenies: an approach using the bootstrap. Evolution 39: 783–791.
- 60. Newmaster SG, Fazekas AJ, Ragupathy S (2006) DNA barcoding in land plants: evaluation of rbcL in a multigene tiered approach. Can J Bot 84: 335–341.
- 61. Kress WJ, Erickson DL (2007) A two-locus global DNA barcode for land plants: the coding rbcL gene complements the non-coding trnH-psbA spacer region. Plos One 2: e508.
- 62. Landergott U, Holderegger R, Kozlowski G, Schneller JJ (2001) Historic bottlenecks decrease genetic diversity in natural populations of Dryopteris cristata. Heredity 87: 344–355.
- 63. Stein DB, Hutton C, Conant DS, Haufler CH, Werth CR (2010) Reconstructing Dryopteris “semicristata” (Dryopteridaceae): molecular profiles of tetraploids verify their undiscovered diploid ancestor. Am J Bot 97: 998–1004.
- 64. Werth CR (1989) Isozyme evidence on the origin of Dryopteris cristata and D. carthusiana. Am J Bot 76: 208.
- 65. Fraser-Jenkins CR (2007) The species and subspecies in the Dryopteris affinis group. Fern Gaz 18: 1–26.
- 66. Quintanilla LG, Escudero A (2006) Spore fitness components do not differ between diploid and allotetraploid species of Dryopteris (Dryopteridaceae). Ann Bot 98: 609–618.
- 67. Ekrt L, Trávníček P, Jarolímová V, Vít P, Urfus T (2009) Genome size and morphology of the Dryopteris affinis group in Central Europe. Preslia 81: 261–280.
- 68. Sleep A (1971) Polystichum hybrids in Britain. Britt Fern Gaz 10: 208–209.
- 69. Vogel JC, Barrett JA, Rumsey J, Gibby M (1999) Identifying multiple origins in polyploidy homosporous pteridophytes. In: Hollingsworth RM, Bateman PM, Gornall RJ, editors. Molecular systematics and plant evolution. London: Taylor & Francis. pp. 101–117.
- 70. Sleep A (1983) On the genus Asplenium in the Iberian peninsula. Acta Bot Malac 8: 11–46.
- 71. Mota de Oliviera S (2010) Diversity of epiphytic bryophytes across the Amazon. PhD thesis. Utrecht: Utrecht University. 152p p.
- 72. Li F-W, Kuo L-Y, Huang Y-M, Chiou W-L, Wang C-N (2010) Tissue-direct PCR, a rapid and extraction-free method for barcoding of ferns. Mol Ecol Res 10: 92–95.