Molecular Verification of the UK National Collection of Cultivated Liriope and Ophiopogon Plants

A collection of cultivated Liriope and Ophiopogon plants was established in 1996–1998 and subsequently hosted at a horticultural college. Uncertainties about the identification of the accessions, compounded by potential errors in propagation and labelling have led to waning confidence in the identities of the plants in the collection. The potential for using DNA barcoding to determine the species identities of the accessions was investigated. The DNA barcode regions of the plastid ribulose-1,5-bisphosphate carboxylase/oxygenase large subunit gene (rbcL) and nuclear ribosomal internal transcribed spacer (nrITS) were amplified. DNA sequence analysis allowed the sequences of the accessions to be compared to reference sequences in public databases. A simple haplotype map of the characteristic polymorphic positions in the rbcL regions was used to clearly distinguish between the two genera and assign Ophiopogon accessions to individual species or sub-groups of species. The ITS sequence data confirmed these genus and species assignations and provided greater resolution to distinguish between closely related species. The combination of two DNA barcodes allowed most of the accessions to be assigned to individual species. This molecular verification confirmed the identity of about 70% of the accessions, with the remaining 30% demonstrating a range of mistaken identities at the species and genus levels.


Introduction
Over the years, there has been increasing recognition of the value of the genetic diversity captured in cultivated plants and the need to preserve, propagate and document it. The UK National Cultivated Plant Collection (NCCP) scheme makes a valuable contribution to the conservation of genetic diversity in cultivated plants (http://www.nccpg.com/National-Collections.aspx). The open model used in the UK allows any suitable collection to be registered with the NCCP, ranging from those held by internationally prestigious botanic gardens to the personal collections of individual enthusiasts. However, the overall value of the scheme is dependent upon the quality of the collections and the correct identification of cultivars within each collection. The identification of cultivated plants is often more difficult than of plants from the wild, due to their complex and often uncertain histories of selective breeding and hybridisation. This problem is exacerbated in ornamental plants, where the selection of unusual floral morphologies may have modified floral characteristics that are important for species assignment.
To date, there is a little evidence of DNA techniques being applied to the identification of cultivated, ornamental plants, but they could prove to be a useful method for the validation of National Plant Collections and as a tool for the wider horticultural community. Identification tests based on DNA barcoding would be much faster than traditional methods of identification that require growth to the flowering stage, in parallel with control plants. In the present study, the relevance of DNA-based tests to the identification and reclassification of Liriope and Ophiopogon species is highlighted.

DNA Barcoding of the rbcLa Region of Liriope and Ophiopogon Accessions
The ribulose bisphospate carboxylase large subunit is encoded in the plastid by the rbcL gene, which, in Liriope and Ophiopogon, is over 1400 bp. The DNA barcoding of the rbcL region typically analyses a partial sequence from either the 5 rbcLa region or the 3 rbcLb region ( Figure 1). The sequencing of the plastid rbcLa barcode region of 75 of the National Collection Ophiopogon and Liriope specimens was reported previously [26]. Multiple sequence alignment of this barcode dataset along with sequences from the GenBank database revealed very little sequence variation, with just four characteristic single nucleotide polymorphic (SNP) positions observed at Positions 172, 216, 392 and 431 (Figure 2). The G/A SNP at Position 431 was found to be genus-specific, with the guanine present in Ophiopogon substituted by an adenine in Liriope. This is the target of a PCR test to distinguish the two genera ( Figure 1) [26]. Four Ophiopogon samples (BTG_693, BTG_695, BTG_709 and BTG_711) and three Liriope samples (BTG_628, BTG_667 and BTG_677) were found to not fit this SNP pattern. The explanation that these accessions had been misidentified at the genus level is supported by results described in the next sections.
The remaining three SNPs at Positions 172, 216 and 392 were identical in all the confirmed Liriope accessions but varied between clusters of Ophiopogon species (Figure 2).  Haplotype map of the rbcL barcode region in different Ophiopogon and Liriope species. The map is a combination of published data from the rbcLa [24] and rbcLb [26] barcode regions, augmented with unpublished sequences from the GenBank database. Diagnostic SNPs unique to single or related groups of species are highlighted.

DNA Barcoding of the rbcLb Region of Liriope and Ophiopogon
A study by Shiba et al. [28] observed a number of SNPs in the rbcLb region of Liriope and Ophiopogon species and identified corresponding haplotypes that were characteristic of individual species or groups of related species. A full rbcL haplotype map comprising 15 informative SNP positions was compiled by combining the rbcLa and rbcLb maps and retaining the haplotype categories used by Shiba et al. [28] (Figure 2). Using the Type 2 haplotype as the consensus sequence, Type 1 was defined by four SNPs (three unique); Type 3, by two SNPs (one unique); Type 4/5, by two/three unique SNPs; and Type 6 (Liriope), by seven SNPs (five unique) ( Figure 2).
Primers were designed to amplify a region containing all the informative SNPs in the rbcLb region ( Figure 1). This region was sequenced in all the Ophiopogon accessions, along with three Liriope specimens (BTG_628, BTG_667 and BTG_677) previously identified as having an anomalous genus-specific rbcLa SNP. The rbcLb SNP pattern for these three accessions confirmed that they all belonged in the Ophiopogon genus. Conversely, the detection of four misidentified Ophiopogon samples (BTG_693, BTG_695, BTG_709 and BTG_711) from the rbcLa SNP at Position 431 was confirmed by their full rbcL Type 6 Liriope haplotypes.
The Ophiopogon sequences were matched with the haplotype panel, allowing the identities of those Ophiopogon species with a unique haplotype to be determined ( Figure 2). For example, the four unique SNPs of Haplotype 1 are found only in a subgroup of O. japonicus (I). Accessions BTG_688, BTG_691 and BTG_692 were confirmed as O. japonicus (I), whilst three plants (BTG_705, BTG_706 and BTG_708) with an unknown species designation also showed this O. japonicus haplotype. The three samples originally designated as O. intermedius (BTG_682, BTG_683 and BTG_684) were also found to have a Type 1 rather than Type 2 haplotype, indicative of O. japonicus (I). By contrast, the accession (BTG_704) originally labelled as O. wallichianus (an unaccepted synonym for O. intermedius) had the expected Type 2 haplotype. However, the taxon originally labelled as O. chingii (BTG_679) showed the Type 2 rather than the expected Type 3 haplotype.
Two samples (BTG_667 and BTG_685) were found to have the rbcL Haplotype 5 characteristic of O. jaburan; sample BTG_685 was therefore correctly designated as O. jaburan, but BTG_667 was originally misidentified as a Liriope. On the other hand, the original designation of sample BTG_686 as O. jaburan was not supported by the rbcL haplotype. One specimen, originally identified as O. clarkei (BTG_680), could not be identified by the rbcL haplotype as there are no rbcL sequences available in public databases. In other cases, the rbcL haplotype was not definitive at the species level and required supporting evidence from the internal transcribed spacer (ITS) barcoding to confirm species identity.

DNA Barcoding of the nrITS Region of Liriope and Ophiopogon
Whilst rbcL barcoding allowed the identification of certain Ophiopogon species by haplotype matching, there was insufficient variation to discriminate between all species, particularly in the Liriope genus. In order to identify these species and resolve the apparent misidentification of certain accessions, the entire collection was subjected to DNA barcoding of the nuclear ribosomal ITS region. The sequence quality of a number of accessions was poor when sequenced in both directions using ITS1 and ITS4 primers. This was resolved in some cases by the use of the plant-specific primers ITS5P and ITS8P, but in other samples, the ITS sequence traces were of poor quality with both primer pairs, indicating low quality DNA templates. However, it was possible to resolve conflicts between the sequence traces obtained with each primer pair and by reference to other sequences in the collection, allowing these sequences to be used in the subsequent analysis.
The samples' sequenced ITS amplicons were aligned with Liriope and Ophiopogon sequences from the NCBI GenBank database. The sequences were aligned using the Clustal W MegAlign package of DNAStar (DNAStar Inc.). The evolutionary relationships of the members of both genera were inferred with the Maximum Likelihood method based on a Kimura 2-parameter model using the MEGA X software package ( Figure 3). The resulting phylogenetic tree was used to identify accessions from the collection by their location on the tree ( Figure 3). The tree morphology is consistent with that found by Wang et al. [4], for ITS sequences, with both Liriope and Ophiopogon supported as monophyletic genera. Ancestral states were inferred using the Maximum Likelihood method [29] and Kimura 2-parameter model [30]. The bootstrap consensus tree inferred from 1000 replicates is taken to represent the evolutionary history of the taxa analysed. Branches corresponding to partitions reproduced in less than 50% bootstrap replicates are collapsed. The percentages of replicate trees in which the associated taxa clustered together in the bootstrap test (1000 replicates) are shown next to the branches. Evolutionary analyses were conducted in MEGA X [31]. The numbers represent the BTG_ sample numbers (Table S1).
All the taxa confirmed as belonging to the genus Liriope by rbcL haplotype had ITS sequences that showed close similarity to L. muscari, including those accessions originally identified as L. graminifolia, L. minor and L. spicata (Table S1). Of the five samples labelled as L. graminifolia, four (BTG_625, BTG_626, BTG_627 and BTG_630) clustered with L. spicata sequences from the database rather than with the two L. graminifolia sequences available in the GenBank database (KF671304.1 and KF671305.1). The fifth sample (BTG_628), labelled as L. graminifolia, was identified as O. planiscapus based on both rbcL and ITS sequences.
Wang et al. [4] showed two major lineages within the genus Ophiopogon Two O. longifolius sequences (KX231369 and KX231372) from the database formed a distinct branch on the tree, separate from their closest relatives O. japonicus, and two sequences (BTG_667 and BTG_685) from the collection fell on the same branch; one of these (BTG_667) was originally classified as L. muscari. No O. jaburan ITS sequences are available in the database, but the two taxa identified as O. jaburan by their rbcL haplotype had quite distinct ITS sequences that formed a separate branch within Sub-Clade B3.
All of the Ophiopogon ITS sequences obtained from the plant collection fell within Sub-Clade B2 or B3. This indicates the absence of O. intermedius and O. clarkei from the collection, even though some accessions were labelled as such.

Deduction of Individual Taxon Identity
The combination of rbcL haplotype and ITS cluster analysis allowed the identity of all the taxa tested to be deduced at the sub-clade or species level. In most cases, the ITS data supported and provided further resolution to the rbcL results, including the seven misidentified genus taxa. In no cases was there a direct conflict between the two barcodes.

Liriope spp.
All the samples predicted by rbcL haplotype to belong to the genus Liriope had ITS sequences that confirmed this classification. The Liriope ITS sequences fell into two distinct groups with close similarity to either L. muscari or L. spicata. In the majority of cases, this was consistent with the original species designation. Out of 46 accessions in the Liriope collection, 27 original identifications were confirmed as correct. These 27 were all L. muscari. The remaining 19 misidentified taxa comprised all the plants originally designated as L. graminifolia, L. minor and L. spicata as well as eight of the accessions labelled as L. muscari. Their ITS sequences allowed the original designation to be ruled out and identification as either L. spicata or L. muscari to be supported (Table S1).

O. japonicus
The identities of all the accessions showing an O. japonicus rbcL Type 1 or 2 haplotype were confirmed by their corresponding ITS sequences. Six accessions were originally labelled as O. japonicus and three of these (BTG_688, BTG_691 and BTG_692) were confirmed as correct. The other three were shown to be O. planiscapus (BTG_690) or L. muscari (BTG_693, BTG_695). Three of the undetermined Ophiopogon species were shown to be O. japonicus (BTG_705, BTG_706 and BTG_708), while the other two, BTG_709 and BTG_711, were confirmed as Liriope spp. The three accessions labelled as O. intermedius (BTG_682, BTG_683 and BTG_684) all proved to be O. japonicus, whilst the single accession labelled as O. chingii (BTG_679) was found to be the only Type II O. japonicus in the collection.

O. bodinieri and O. planiscapus
The differentiation of O. bodinieri and O. planiscapus proved to be more difficult, since the two species lack resolution within the ITS sequences, so identification relied more heavily on the rbcL haplotype. Within the group of plants with ITS sequences related to O. bodinieri and O. planiscapus, some had the rbcL Type 2 haplotype (typical of O. bodinieri as well as O. japonicus II) and others had the Type 3 haplotype consistent with O. planiscapus. This latter sub-group included three cultivars with the "Black" phenotype (BTG_697, BTG_699 and BTG_701) characteristic of ornamental O. planiscapus cultivars [32] and had ITS sequences that fell within Subclade B3a as defined by Wang et al. [4]. Therefore, a combination of Type 3 rbcL haplotype and ITS Subclade B3a was read as support for O. planiscapus, whilst a Type 2 rbcL haplotype and ITS Subclade B3b combination was interpreted as support for O. bodinieri. Using this identification system, five plants originally labelled as O. planiscapus (BTG_697, BTG_698, BTG_700, BTG_701 and BTG_702) were confirmed as O. planiscapus. Two Liriope accessions (BTG_628 and BTG_677) were also identified as O. planiscapus.
Two accessions, BTG_690 and BTG_699, had anomalous barcode results because their apparent location in the ITS Subclade B3b did not match their Type 3 rbcL haplotype. Conversely, accessions BTG_703 and BTG_704 were located in ITS Subclade B3a, which did not match their Type 2 rbcL haplotype. Inspection of the Subclade B3 ITS sequences revealed three SNP positions (A/G 46, C/T 231 and C/T 545) that distinguish O. planiscapus (KC798477.1) from the most closely related O. bodinieri sequence (KF671232.1). The O. planiscapus SNP pattern was found in all nine of the accessions having a Type 3 rbcL haplotype, including BTG_690 and BTG_699. These were therefore assigned to O. planiscapus.
The O. bodinieri pattern was found in all of the accessions (BTG_678, BTG_680, BTG_686, BTG_703 and BTG_704) with a Type 2 rbcL haplotype. Accession BTG_678 was therefore confirmed as O. bodinieri, whilst the other four were re-assigned from O. clarkeii, O. jaburan, undetermined Ophiopogon species and O. intermedius, respectively.
In total, the species identities of 36 accessions were confirmed, and another four unconfirmed Ophiopogon accessions were identified at the species level, out of a total of 73 taxa tested. The remaining 33 taxa were misidentified, seven at the genus level and the other 26 at the species level.

Discussion
This paper describes the application of DNA barcoding to the confirmation of plant species identity in a horticultural collection of ornamental plants. The value of this molecular verification is demonstrated by the discovery of a significant proportion (46%) of mistaken identities.
There are a number of explanations for these misidentifications. One of the most notable observations is that the collection is less rich in diverse species than the original records indicated. The identification of rarer species by the original collector appears to have been over-optimistic, since all the verified species are those found in other collections in the USA and Europe [1][2][3]7] that are widely traded commercially. One possibility is that the collector was misled by new varieties with unusual features resembling species not normally found in trade. It is also not clear whether accessions were identified by vegetative characters alone, since some of the plants in the collection rarely flower.
One issue that could lead to misclassification is confusion around accepted nomenclature. This was certainly the case for accessions BTG_624 (L. exiliflora) and BTG_704 (O. wallichianus). However, in both cases, the identity inferred from DNA barcoding did not match the accepted synonym (L. muscari and O. intermedius, respectively), so the misidentifications were not just the result of changes in taxonomic nomenclature.
A more likely reason for mislabeled taxa in living collections is the gradual accumulation of errors occurring during many years of propagation and re-labelling. Over some periods, the collection was grown outdoors and prone to damage from rabbits, risking the unnoticed loss of some specimens and overgrowth of others. As mentioned above, the irregular flowering of some taxa means that the curators of the collection had to rely on leaf morphology to monitor the correct labelling of accessions.
Another compounding issue is the hybridisation of plants, either before collection or during many generations of propagation. Hybridisation between species is known to occur in Ophiopogon [5] and is often performed deliberately. One advantage of using a combination of plastid and nuclear barcodes is that hybridisation may be detected if there is a discrepancy between the two barcodes. However, the consistency between the rbcL and ITS identifications suggests that this was not a prevalent issue in this particular collection.
The use of two different barcode regions offered other benefits. The rbcL region allowed unambiguous differentiation between the two genera, and the haplotype panel provided a simple, unambiguous method to discriminate between four groups of Ophiopogon species. This provided the first stage of a "two-tier" approach to species identification [33,34]. The second stage involved determining the similarity of the ITS barcode sequence to reference barcodes using a phylogenetic tree. This gave unequivocal results for most Liriope and Ophiopogon species, where good quality reference sequences were available for comparison. The correct species designation of O. bodinieri and O. planiscapus accessions required a combination of rbcL haplotype and ITS SNP comparison to provide a reliable identification, given the close similarity of the ITS sequences.
The molecular testing of the collection was carried out tabula rasa, without reference to the prior labelling or morphological characteristics of the accessions. The subsequent examination of characteristic morphological features (particularly flowering) (Table S1) showed that these were consistent with the molecular classification.
One of the drivers for conducting this research was a particular interest by the curators in the "black" varieties of Ophiopogon (Table S1) and the fact that a number of different accessions had different cultivar names. This barcoding study was able to confirm that all of these cultivars belong to the same species, O. planiscapus, and there were no significant differences in their rbcL or ITS sequences that might support the contention that these cultivars are genetically distinct. Different "pattern based" approaches to DNA testing (DNA fingerprinting techniques), like RFLP, RAPD, ISSR or AFLP, would be more suited to detecting genetic differences between the cultivars [35,36]. These results underline the importance of not relying solely on the morphological characteristics of specimens for their identification and agree with the idea that DNA barcoding should be routinely integrated in the research of alpha taxonomists [37,38]. It is also proposed that DNA barcoding should be more widely performed as a routine quality control method in biological research and presented in publications to verify the authenticity of the samples used.
Due to the low quality of ITS sequencing results of some Liriope and Ophiopogon samples when using the conventional ITS1/4 primers, which could have been linked to fungal contamination or the poor quality of the DNA purified, some of the samples had been sequenced using the plant-specific primers (ITS5p and ITS8p) [37,39]. The use of plant-specific primers helped in improving the quality of sequencing results, allowing correct identification.
The conclusion of the study is that DNA barcoding revealed a surprising level of misidentification in a national collection of horticultural plants. This indicates that other collections of cultivated ornamental plants may be similarly unreliable as living reference standards of species identity and that DNA barcoding would be an appropriate approach to verification.

Plant Collections
The Liriope and Ophiopogon collections were maintained under standard glasshouse conditions, with plants of each accession cultivated in individual pots. The collection comprised one labelled pot per accession (with a few duplicates) containing single or a small number of plants. The collection records indicated a total of 54 Liriope accessions and 34 Ophiopogon accessions, but only 46 Liriope and 27 Ophiopogon taxa were available to be tested (Table S1). The Liriope collection included five named species (L. exiliflora, L. graminifolia, L. minor, L. muscari and L. spicata) and some taxa only wallichianus accession (BTG_704) was renamed as its accepted synonym, O. japonicus, and L. exiliflora (BTG_624) was renamed as its accepted synonym, L. muscari [2,5,8]. Each available accession was photographed, and these are included in the inventory of the collection shown in Table S1.

Plant Material and Genomic DNA Extraction
Fresh leaves were collected from 74 different accessions of Ophiopogon and Liriope at Brooksby Melton College (Melton Mowbray, Leicestershire, UK) from the UK National Plant Collections for Ophiopogon and Liriope. Details of the genera, species and accession numbers are in Table S1. Samples were stored at −80 • C. DNA was extracted from 100 mg of frozen material, previously ground to a fine powder in liquid nitrogen with a mortar and pestle, using the DNeasy Plant Mini Kit (Qiagen Inc., Germantown, MD, USA) following the manufacturers' guidelines. DNA concentration and quality were assessed using NanoDrop (Thermofisher) technology (Table S2).

PCR Protocols
PCRs were carried out using primers as detailed in Table 1. PCR reaction mixes contained 1X MyTaq Red Mix (Bioline), 0.2 µM of each forward and reverse primer, and 1 µL of gDNA as the template. A G-Storm GS1 Thermal Cycler (G-Storm Ltd., Somerton, UK) was used with the following program: PCR products were run on 2% (w/v) agarose, 1X TBE gels with 1 µL of SYBR ® Safe DNA Gel Stain (Invitrogen, Paisley, United Kingdom) at 100 V for 30 min and analysed in a Gel Doc™ EZ Gel Documentation System (BioRad, Oxford, United Kingdom). PCR products were submitted for sequence analysis to Macrogen (www.dna.macrogen-europe.com) to verify the authenticity of the starting material.

DNA Sequence Analysis
The obtained sequences were used to implement contigs and generate multiple sequence alignments using the CLC Main Workbench 7.5.1 software (Qiagen, Germantown, MD, USA) (Supplementary Data S1). To create the alignment, the following parameters were selected: Gap open cost = 10.0, Gap extension cost = 1.0, and the very accurate progressive alignment was selected.
Published Liriope and Ophiopogon ITS DNA sequences were obtained from the National Center for Biotechnology Information (NCBI) GenBank database (http://www.ncbi.nlm.nih.gov/). Phylogenetic analyses were conducted using the MEGA X [29] software package. The evolutionary history was inferred with the Maximum Likelihood method [30] based on the Kimura 2-parameter model [31].
Author Contributions: J.A. and P.G. curated the national collections of Liriope and Ophipogon, J.A. and A.S. conceived the project; T.S. and A.S. designed and supervised the experimental strategy, and wrote the paper; E.M. performed the experiments, analysed the sequence data, and wrote part of the paper; D.B. photographed and sampled the collections, and wrote part of the paper. All authors have read and agreed to the published version of the manuscript.
Funding: This research received no external funding. It was funded in part by an internal award from the De Montfort University HEIF fund.