Introduction

The human diet includes a lot of domesticated plant species (for instance Hordeum vulgare L., Zea mays L., Oryza sativa L., etc.) derived from an evolutionary continuum of people–plants interactions, leading to plants showing higher yield, easier cultivation and detoxification from natural poisonous metabolites [37]. Since the beginning of the domestication, around 10,000 years ago, plants and derived were also used in medicine, and in the last years, the nutritional therapies and phytotherapy have emerged as new concepts and healing systems. However, several spontaneous plants are potentially toxic for human beings [27, 32, 34, 36]. The accidental ingestion of toxic plant portions (for instance seed, fruit, root, etc.) can cause severe poisoning or even death [24, 34]. In the last years, the plant exposures are among the most frequent poisoning cases reported by poison control centres [25]. Exposures to plants containing oxalate crystals, such as Philodendron and Dieffenbachia, are one of the most common toxic plant exposures reported in the USA [1]. The same toxic compounds were also present in several African species such as Dieffenbachia amoena, Alocasia macrorrhiza and Monstera deliciosa [8] and in spontaneous or ornamental plant from other countries. To further complicate the picture, the amount of poison in a plant tends to vary with location, age of the plant, season and weather conditions [10, 18, 35].

The clinical diagnosis of intoxicated patients is typically based on the morphological analysis of plant fragments in the stomach contents [20]. This method is very tedious to perform, requires a considerable amount of training and usually a variable proportion of plant fragments remains unidentifiable. In addition, the plant species identifications can be difficult without residuals showing distinctive taxonomic elements. Recently, a new technique for plant identification based on the analysis of a short, standardised DNA region known as “DNA barcoding” ([13]; www.barcoding.si.edu) has been proposed. The basic idea of the discrimination system is simple: the sequence variation of the DNA barcode must be high enough between species, so that they can be discriminated from one another. In practice, a DNA sequence from such a standardised DNA region can be generated from a small tissue sample taken from an unidentified organism and can be compared to a library of reference sequences from described species. A match of the sequence from the unknown organism to one of the reference sequences can provide a rapid and reproducible identification [2, 13].

However, the selection of a barcode locus is complicated by the trade-off that arises between the need for universal application and maximal rates of sequence divergence [17]. A portion of the mitochondrial coxI (cytochrome c oxidase subunit 1) gene was largely applied as DNA barcode sequence for animal identification [13], but in land plants, this sequence is highly invariant and therefore unsuitable for use as DNA barcodes. Several alternative regions have been proposed, such as the plastidial conserved rpoB, rpoC1, rbcL and matK genes, several intergenic plastidial spacers (trnH-psbA, atpF-atpH and psbK-psbI) and the nuclear internal transcribed spacers [3, 9, 17]. A lot of research groups have proposed different combinations of these loci to increase the efficiency in species identification, especially for the critical groups [30]. However, both the identity and the number of the most appropriate regions for plant DNA barcoding remain contentious [14, 15].

The objective of the current study is to evaluate the universal application of DNA barcoding approach to univocally identify toxic plants starting from different plant portions. In this study, we provide an evaluation of the performance of five candidate barcoding loci on a selected group of poisonous plant species. The final objective is to develop a rapid molecular tool for toxic plants identification starting from a small portion of plant tissue from different biological samples such as stomach contents and/or faeces of patients. This tool could be useful for poison centres to identify the ingested plants univocally and to define the suitable medical treatments rapidly.

Materials and methods

Taxonomic sampling

A total of 50 land-plant species showing different poisonous substances was selected to test the performance of different candidate genomic regions for DNA barcoding analyses empirically. A complete list of investigated species is given in Table 1. Our sampling strategy was designed to have sufficient density of sampling within angiosperm, such that sets of closely related species are included as well as plants of divergent genera or families. Our taxon set was divided in three groups.

Table 1 List of analysed plants divided in different groups (Gr) (see “Materials and methods”)
Full size table

Group I: Hornamental plants as well as spontaneous plants containing different toxic substances (Table 1)

Several of these show attractive poisonous portions such as fruits or flowers, and these are accidentally eaten, mainly by children. If the amount of ingested poisonous substance reaches significant levels in the body, clear symptoms of poisoning were highlighted. These samples were selected, because they represent a diverse set of species across the angiosperms (including monocots and eudicots) with various levels of phylogenetic distance. This provides a sound assessment of universality of the tested DNA barcoding markers.

Considering that the goal of our work is to set up a practical DNA barcoding approach suitable to identify poisonous plants starting from a small portion accidentally eaten, in this group, DNA was extracted from standard materials, such as leaves, and also from fruits-seeds that represented the accidentally eaten poisonous structures (Table 1).

Group II: Phylogenetically related plants showing different poisonous portions and degrees of toxicity were selected to evaluate the discriminating power of different DNA barcoding markers among closely related taxa

In this group, species of the genera Aconitum (IIa) and Sambucus (IIb) were included.

Group III: Congeneric edible and poisonous plant species

In the clinical diagnosis of intoxicated patients, it is very important to distinguish fragments of edible plants from the poisonous species elements. For this reason, some species of Solanum (group IIIa) and Prunus (group IIIb), two large angiosperm genera constituted by edible and poisonous plants, were analysed by different DNA barcoding markers. The final goal was to test the power of DNA barcoding to distinguish common edible fruits from fragments of toxic plants accidentally ingested.

DNA isolation and amplification

Young leaves or fruits-seeds (100 mg) of the 50 selected plants were used for DNA analysis. DNA was isolated using the DNeasy Isolation and Purification kit (Qiagen, Milan, Italy) to obtain high-quality DNA, free of polysaccharides or other metabolites that might interfere with DNA amplification. Purified DNA concentration of each sample was estimated both fluorometrically and by comparison of ethidium bromide-stained band intensities with λ DNA standard.

To assess the comparative performance of different DNA markers, each DNA sample (Table 1) was analysed with all the five candidate DNA barcoding regions. These included two coding (rpoB and matK) and one non-coding (trnH-psbA intergenic spacer) plastidial DNA region, that feature in the preferred barcode solutions of different research groups [14, 15]. Two additional nuclear DNA regions were also analysed to evaluate the performance of nuclear genome in comparison with plastidial one. These regions are the sqd1 (UDP sulfoquinovose synthase) and At103 (Mg-protoporphyrin IX monomethyl ester cyclase) developed by Li et al. [22] as universally amplifiable markers for phylogenetic reconstructions.

PCR for each candidate marker was performed using puReTaq Ready-To-Go PCR beads (Amersham Bioscience, Italy) in a 25-μL reaction according to the manufacturer’s instructions. PCR cycles consisted of an initial denaturation for 7 min at 94°C, 35 cycles of denaturation (45 s at 94°C), annealing (30 s at different temperature; see Table 2) and extension (1 min at 72°C), and a final extension at 72°C for 7 min. Details of primers used for amplification are given in Table 2. Products were submitted for sequence analysis to Macrogen (http://www.macrogen.com). The heavy DNA strands were bidirectionally sequenced using an ABI 155 3730XL automated sequencer at Macrogen Inc., Korea. Manual editing of raw traces and subsequent alignments of forward and reverse sequences enabled us to assign edited sequences for most species. The 3′ and 5′ terminals were clipped to generate consensus sequences for each taxon.

Table 2 Details of primers used for amplification of the five candidate DNA barcoding markers
Full size table

Data analysis

Our criterion for assessing universality of the five DNA candidate markers involves the establishment of which regions could be routinely amplified and sequenced in the maximum number of analysed plants. We have tested only the most universal primer combinations (Table 2) for each candidate DNA marker to facilitate the interpretation of successes and failures. For all taxa and loci, we conducted PCR amplification in a two stage trial. In the first stage, we used the standard PCR conditions described above, starting from 10 ng of DNA template. The second stage included only samples that did not amplify or that produced multiple PCR products. Samples of both types of failure were re-amplified using 1 and 25 ng of DNA template. The samples that failed to amplify were repeated at lower stringency (with a reduction of 5°C in the annealing temperature as described in Table 2) and 40 PCR cycles. Only in case of negative amplification in all conditions the PCR reaction was considered a failure.

Each marker was also evaluated, including the sequence length and the alignment success in all the analysed plant groups. Evaluation of comparative levels of variation and discrimination of different five markers were undertaken using MEGA 4.0 [31] to generate Kimura 2-parameter (K2P) distance matrices for each locus. K2P distances were used following guidelines from the Consortium for the Barcoding of Life to evaluate performance among barcoding loci (http://www.barcoding.si.edu/protocols.html). The discrimination values (K2P sequences divergence converted into percent) were computed for each locus in all the analysed groups.

Results

Amplification and sequencing success

A total of 50 poisonous plants (Table 1) was used for DNA extraction. High DNA quality was obtained from all the analysed samples. For nine taxa (Table 1), the DNA was extracted both from leaves and fruits-seeds; however, no difference was observed in the PCR success, sequencing quality and sequences alignments for the template DNA obtained from different plant portions.

Results of our DNA barcoding analysis across the three selected plant groups showed a conspicuous difference among the five tested loci with respect to amplification success, PCR products lengths and sequences quality (Table 3). All the three plastidial markers exhibited high PCR success with standard primers with 10 ng of DNA as template for DNA amplification; non-relevant differences were detected among the three plant groups (Table 3). In the case of nuclear markers, only the At103 was successfully amplified in the standard PCR condition (10 ng of DNA as template) for a large number of plants of our dataset, while sqd1 showed some amplification problems for several samples of group I and IIb (Table 3) both starting from 1, 10 and 25 ng of DNA.

Table 3 Summary of the proportion of individuals successfully amplified and sequenced and divergence values from five candidate barcoding regions in three analysed groups of poisonous plants
Full size table

All the PCR products corresponding to the five DNA markers were easily sequenced, with exclusion of four samples (Table 3). The evaluation of the sequence quality from the tested loci demonstrated that high-quality bidirectional sequences were routinely obtained from almost all loci. Success rates for bidirectional sequences were highest for plastidial markers. The greatest problems in obtaining bidirectional sequences with few ambiguous bases were encountered with the At103 sequences in part attributable to a high frequency of mononucleotide repeats disrupting individual sequencing reads. This required more manual editing. In order to properly assess the performance of this marker, only the first 250 bp of sequence was used for statistical analyses.

The accession numbers of the considered sequences are listed in Table 5 of the Electronic supplementary materials.

Evaluation of the discrimination value of the five candidate loci

The sequence divergence values for each tested locus were computed to define which is the best DNA barcode marker for poisonous plants identification (Tables 3 and 4). The trnH-psbA ranked first in divergence value in the analysed groups compared with the other plastid and nuclear regions. However, the high variability of this DNA spacer did not allow to align properly the sequences related to species of the group I and to define the genetic diversity values (Table 3).

Table 4 Vale of sequence divergences (Kimura 2-parameter d value) computed between individuals of each of the congeneric group (IIa, IIb, IIIa and IIIb)
Full size table

The matK gene showed consistent genetic variability among samples of group I and IIa; however, the divergence among the plants of group IIb was very low: Sambucus ebulus L. and Sambucus nigra L. showed the same matK sequence (Table 4). This marker is also able to distinguish edible from poisonous plants of group III (Table 4).

The rpoB sequences showed the lower sequence divergences than the other four tested markers. The efficacy of this marker is very limited in several groups such as IIa and IIb (Tables 3 and 4); however, edible plants of group IIIb were clearly distinguished from toxic species (i.e. Prunus laurocerasus L.).

Nuclear markers showed consistent genetic variability among all analysed groups; however, we underline that sqd1 was amplified in a limited number of species, and in group IIa, sequences of this gene showed a large conserved trait. On the contrary the At103 showed a satisfactory level of amplification and a wide genetic variability in all the analysed groups. This marker seems also to allow a good distinction among congeneric species, including toxic and edible species included in groups IIIa and IIIb.

The results from data-mining sequences in GenBank by using BLAST analysis notwithstanding their relatively crude nature search engine indicated that matK and trnH-psbA were successful at returning a correct match for 24 and 22 samples, respectively (data not shown). This suggests that generic and public databases such as NCBI or EMBL have already catalogued numerous matK and trnH-psbA sequences referable to poisonous plants species. The other three loci had a few too many sequences in GenBank to conduct a robust test (both rpoB and sqd1 showed three correct matches, while no-corresponding sequence for the At103 was found in GenBank).

Discussion

The main purpose of this work is to provide a rapid and accurate identification system of unidentifiable poisonous plants based on DNA analysis. DNA barcoding is promising in providing a practical, standardised, species-level identification tool that can be used for different study including forensic analyses [7] and animal diet determination when the food is not identifiable by morphological criteria [7, 33]. Similar situations are found in the poisonous centres, where toxic plants should be identified starting from different portions of plant material (fruits, seed, leaves, etc.). In most cases, portions of the most attractive fruits and seeds, accidentally ingested from children, represent the only mean for species identification [20]. For these reasons, our research was conducted starting from different plant materials, including attractive fruits and seeds. DNA was extracted from plant samples collected in field or house and stored at room temperature without specific methods to preserve the DNA quality. DNA amplification and sequence analysis suggest that plant materials used for DNA extraction do not affect the results of DNA barcoding analysis; this supports the application of the DNA barcoding approach as a practical identification tool in the poisonous centres.

In our work, we tested the effectiveness of five different DNA barcode regions. The three plastidial markers provided a good performance; whereas, concerning the nuclear regions, At103 only showed a wide applicability in the analysed groups. Since an ideal DNA barcode should be applied to a large number of species with standard PCR conditions [4, 13], according to our results, the sqd1 gene was excluded among the candidate regions.

Among the three plastidial markers, all worked across the analysed groups with a single set of PCR conditions. This is an impressive performance given by the range of taxonomic diversity encompassed. However, the trade-off in the universality of the rpoB is its relatively low level of species discrimination power. For this reason, we consider this marker unsuitable for the poisonous plants identification. On the contrary, the non-coding trnH-psbA intergenic spacer is the most viable candidate in the analysed poisonous plants. PCR priming sites within highly conserved flanking sequences, combined with a non-coding region that exhibits high numbers of substitutions, made the trnH-psbA spacer highly suitable as a plant barcode. Previous researches suggested that this intergenic spacer has alignment problems, especially in some monocots [4] and conifers [14]. Our results confirm these problems (Table 3), even if the preliminary investigations carried out in GenBank showed good results for this marker, and overall, we concur with Kress and Erickson [16] that trnH-psbA is one of the most suitable plant DNA barcode.

The matK gene showed easy amplification and alignment in the analysed species and high level of discrimination values in several analysed groups (Tables 3 and 4). Lahaye et al. [19] identified this gene as a universal DNA barcode for flower plants. Recently, the CBOL Plant Working Group (http://www.barcoding.si.edu/plant_working_group.html) confirmed this prompt and suggested the combination of matK with rbcL as a universal plant DNA barcode. Although we did not test the rbcL because several publications underline the low discriminating power of this gene [9, 14, 26], our data are in agreement with the CBOL statement on the role of matK gene as a universal DNA barcode.

Concerning the nuclear markers, our analysis focused on the Conserved Ortologous Sequences genes [22]. Among these, At103 was amplified and sequenced in a large part of the tested samples and showed a consistent genetic diversity, supporting its application as a DNA barcode. We underline that nuclear markers can provide a more reliable assessment of hybridization than uniparentally inherited plastid DNA regions [3, 11], and this is very important in the plant kingdom where hybridization events are largely diffused. For example, Sambucus and Prunus are two important genera considered in our work showing different natural and commercial hybrids [23, 29]. Breeding events between poisonous and non-poisonous plants could result in individuals showing different degrees of toxic substances [12, 21, 28]. Only by the combination of the plastidial marker with the nuclear one is it possible to distinguish hybrids from parental species.

In conclusion, our data confirm the utility of DNA barcoding for plant poisonous species identification. Our results support the application of matK as the best plastidial marker. Among the nuclear sequences, we suggest At103 as the most suitable candidate; however, future analysis will be required to evaluate the performance of this marker in a large plant dataset.

The next step of this research is toward the establishment of a dedicated poisonous plants database where all species are described either by morphological and molecular (DNA barcoding) approaches. Our preliminary in silico tests suggested that the availability of deposited reference sequences of matK and trnH-psbA relative to toxic plants could enable a fast identification of poisonous species with high efficiency. The dedicated database supported by proper bioinformatics technologies [2] will lead to the development of an innovative tool suitable for a rapid identification of toxic plant by poisonous centres.