Genotypic identification of Panicum spp. in New South Wales, Australia using DNA barcoding

Australia has over 30 Panicum spp. (panic grass) including several non-native species that cause crop and pasture loss and hepatogenous photosensitisation in livestock. It is critical to correctly identify them at the species level to facilitate the development of appropriate management strategies for efficacious control of Panicum grasses in crops, fallows and pastures. Currently, identification of Panicum spp. relies on morphological examination of the reproductive structures, but this approach is only useful for flowering specimens and requires significant taxonomic expertise. To overcome this limitation, we used multi-locus DNA barcoding for the identification of ten selected Panicum spp. found in Australia. With the exception of P. buncei, other native Australian Panicum were genetically separated at the species level and distinguished from non-native species. One nuclear (ITS) and two chloroplast regions (matK and trnL intron-trnF) were identified with varying facility for DNA barcode separation of the Panicum species. Concatenation of sequences from ITS, matK and trnL intron-trnF regions provided clear separation of eight regionally collected species, with a maximum intraspecific distance of 0.22% and minimum interspecific distance of 0.33%. Two of three non-native Panicum species exhibited a smaller genome size compared to native species evaluated, and we speculate that this may be associated with biological advantages impacting invasion of non-native Panicum species in novel locations. We conclude that multi-locus DNA barcoding, in combination with traditional taxonomic identification, provides an accurate and cost-effective adjunctive tool for further distinguishing Panicum spp. at the species level.

www.nature.com/scientificreports/ Traditionally, morphological features were used to differentiate Panicum spp. (Fig. 1) 11 . However, species identification based on morphology is not a trivial task as morphological differences between species can be subtle, even when considering native and non-native species 12,13 . A microscope is frequently needed to observe critical features such as the shape of the abscission scar at the base of the fertile lemma. Morphological keys to species are also heavily biased towards reproductive characters thereby rendering identification of sterile specimens difficult, if not impossible, even for a grass specialist. Although precise identification is possible using morphological keys, especially if reproductive material is available 14 , successful usage of these keys requires a clear understanding of morphological structures and a proficiency in using keys. For example, the taxonomic key to differentiate Panicum effusum R.Br. (native to Australia) and P. hillmanii Chase (introduced to Australia from North America) is based on the shape of the abscission scar of the fertile lemma. The abscission scar of the fertile lemma of P. effusum is entirely basally located and less than 0.5 mm wide while P. hillmanii has a crescentic abscission scar of the fertile lemma, extending upwards from the base, and is more than 0.5 mm wide 15 . The level www.nature.com/scientificreports/ of expertise required to detect minute morphological differences presents a major challenge for the inexperienced and examination by a grass taxonomist may ultimately be required for consistency in identification. Molecular technologies are increasingly used to develop reliable methods for plant and animal species identification 16 . A PCR-based genotyping method, DNA barcoding, has been extensively applied for this purpose 17 . DNA barcoding is a method that uses short but informative standardised DNA regions ("barcodes") to identify or differentiate between species [18][19][20] . It was first proposed in 2003 17 , and was utilised as an important  complementary method to traditional morphological identification 21 , for vegetation and floristic surveys 22 , ecological forensics 23 , regulatory enforcement 24,25 , community phylogenies, comparative biology and phylogenetic diversity 26 . Selection of the "barcode" is critical to establish a successful DNA barcoding platform to identify Panicum species. An ideal barcode should be a short DNA sequence that can be routinely amplified using a standard PCR method. The amplified product should also be easily sequenced with universal primers that are anchored in highly conserved DNA regions, and the sequences should be easily aligned without extensive manual editing 22 . Most importantly, these regions should be able to differentiate between the target species 18 . However, unlike animals where the sequence for cytochrome oxidase 1 (CO1) in mitochondrial DNA was proposed as the universal barcode for species identification 17 , the identification of an universal barcode for many plant species, and Panicum in particular, remains challenging due to inter-species mutation and technical reliability 27 . Unfortunately, CO1 is not suitable for use in plants as the nucleotide substitution rate within mitochondria in plant cells is relatively low 28 . Additionally, there has been difficulty in locating highly heterogeneous regions in plant DNA due to a lack of sequence polymorphism, slow mutation rates 29 , frequent introgression or species hybridisation between related species 30 , and incomplete lineage sorting 22 .
To overcome these issues, a multi-locus approach for plants was demonstrated to improve identification capability and reliability 18 . Multiple barcoding studies have further suggested that a combination of rbcL and matK sequences are suitable for DNA barcode GAP analysis in Panicum spp [31][32][33][34][35] . Moreover, the use of the chloroplast gene ndhF, alone or in combination with rbcL and matK, has been proposed 32,36 . Additionally, the use of the ndhF region may also increase the resolution level when used to discern between grass species 37,38 . Unfortunately, the use of trnH-psbA for differentiating Panicum species has not proven useful, as the existence of inversions or mononucleotide repeats at this locus can result in incorrect alignments or additional difficulties in sequencing 39 . To date, the nuclear ribosomal Internal Transcribed Spacer (ITS) locus has not been used as a species discriminating barcode in Panicum spp., but it has been proposed that ITS is a suitable marker for genetically similar species and could be used as a core or complementary barcode 40 . Currently, the optimal suite of barcoding loci has not yet been fully established for identification of various Panicum species in Australia. Therefore, this study has focused on the use of the nuclear locus ITS as the core barcode for genotypic identification of ten native and non-native Panicum species found in southern New South Wales, together with two plastid loci, matK and trnL intron-trnF as complementary loci 41 .
The establishment of a robust and objective method for identification using both genetic markers and morphological traits is required to address and overcome the challenge of differentiation of Panicum species in both field monitoring and laboratory studies and would enable unambiguous identification of field samples collected at any stage of the plant's growth cycle. To achieve this outcome, we developed and validated a DNA barcoding method for identification and differentiation in ten species of Panicum that are frequently found in south-eastern Australia. This study also tested the hypothesis that Panicum species with a smaller genome size have a greater potential to become invasive in a novel environment 42 , by determination of the genome size of several indigenous and non-native Panicum species in southern New South Wales.

Results
Sampling. Panicum plants (106 individuals) were sampled from geographically dispersed locations within a 200 km radius of Wagga Wagga, New South Wales, Australia (Fig. 2). Morphological examination of these specimens at the Australian National Herbarium (CANB) revealed that five Panicum species were captured by field sampling. To bolster the number of species included in the DNA barcode GAP analysis, sampling of herbarium specimens held by CANB was undertaken. A total of 40 samples (17 field samples and 23 herbarium samples), representing ten indigenous and non-native Panicum species, were included in the analysis (Table 1).
DNA barcode gap analyses. PCR amplification and sequencing were undertaken for all samples for the three selected regions: ITS, matK and trnL intron-trnF. Sequenced loci of these three regions were submitted to GenBank and their accession numbers were listed after the specimen's name in the phylogenetic tree. Alignments of each region were truncated to 641, 730, and 750 bp for ITS, matK and trnL intron-trnF, respectively. Concatenated loci, one nuclear locus with either one or two plastid loci, were calculated for barcoding gaps. Further intraspecific and interspecific distance analyses were performed on eight Panicum species (Table 2). P. buncei (native) and P. coloratum (non-native) were not included in these analyses as they were not genetically separated by any of the three regions).
Phylogenetic tree inferred using Bayesian inference clustered most species into highly supported clades (Fig. 3). All native species (P. effusum, P. queenslandicum, P. decompositum, P. laevinode, P. buncei) were clustered into a large group, although the posterior probability was low (59%). The majority of the non-native species (P. hillmanii, P. capillare, P. miliaceum and P. gilvum) was clustered into clades separated at the species level. Most species (both native and non-native) were classified into monophyletic groups. Exceptions included the nonnative species P. coloratum, which clustered with the native P. buncei. www.nature.com/scientificreports/ novel environments, total genome size of five Panicum species, P. capillare, P. decompositum, P. effusum, P. gilvum and P. hillmanii, was determined using flow cytometric analysis of cells collected from fresh leaf tissue. Determination of genome size was based on coefficient of variation (CV) values below 10% (Fig. 4). The calculated genome size (1C value) of P. capillare, P. decompositum, P. effusum, P. gilvum and P. hillmanii was 1.24 pg, 1.49 pg and 1.52 pg, 0.21 pg and 0.24 pg, respectively, (Table 3). No significant differences in genome size were observed for samples of the same species collected from geographically distant locations.  www.nature.com/scientificreports/

Discussion
Selection of the "barcode" for sequencing is critical when establishing a successful DNA barcoding approach or platform effective in differentiating individual Panicum species. An ideal barcode is typically a short DNA sequence that can be routinely amplified using a standard PCR method. The amplified product should be easily sequenced with universal primers, which are anchored in highly conserved DNA, and the sequence result should also be aligned without extensive manual editing 22 . Additionally, but most importantly, the barcode should strongly differentiate the Panicum species, and ideally, there should be no overlap between intraspecific and interspecific divergence 18,43 . Furthermore, the efficacy of any DNA barcoding methodology depends on the extent of differences between intraspecific and interspecific divergence in a selected locus or combined loci 43 . In our study, the ITS locus showed the highest minimum interspecific distance (0.71%), a distance that was significantly greater than the highest maximum intraspecific distance (P. effusum, 0.34%). This confirmed that ITS may be suitable as a standalone locus for the differentiation of selected Panicum species in Australia. In contrast, we found significant overlap between intraspecific and interspecific distances for both matK and the trnL intron-trnF regions. Therefore, the individual application of either loci alone may be problematic for species differentiation in Panicum due to lack of intraspecific distance observed. However, the use of these loci in combination with ITS presents advantages when attempting to detect hybridization although there is currently no field or herbarium evidence of Panicum species hybridisation in southern New South Wales. Sequence combinations from nuclear and chloroplast genomes could provide additional information for enhanced species identification. For example, trnL intron-trnF shows the greatest prevalence among all noncoding chloroplast DNA sequences in GenBank to date 41 ; and may assist in identification at the genus or species level in ambiguous specimens.
We compared the genotypic identification of indigenous and invasive Panicum species in Australia, and found that the native species P. buncei, P. decompositum, P. effusum, P. laevinode and P. queenslandicum were separated clearly from the non-native species P. capillare, P. gilvum, P. hillmanii, and P. miliaceum. These findings suggest that native Australian Panicum species have maintained a unique genetic fingerprint despite potential for hybridisation with non-native counterparts. Diversity in location-dependent accessions of P. miliaceum has recently been described, suggesting that genetic variation could be inherent at the population level 44 . This has potentially important implications for chemical or bioactive properties associated with this species. Interestingly, we noted that one non-native species, P. coloratum, was genetically more closely aligned with the native P. buncei than with non-native counterparts 45 . Further evolutionary analysis of these species, particularly with respect to correlating the molecular results with voucher specimens located in Australian herbaria, and those more globally, may be required to ensure correct identification.
The genome size of P. gilvum, P. hillmanii, P. decompositum and P. effusum has not previously been reported even though these species are frequently encountered across southern Australia. Genome sizes of P. capillare, P. decompositum and P. effusum were shown to be similar to ploidy size of other previously described Panicum species 42 . The genome sizes of P. hillmanii and P. gilvum were surprisingly smaller than predicted, and therefore we suggest a role for genome size in Panicum species identification and possibly prediction of invasive potential. Certain naturalised plants exhibit smaller genome size in contrast to their non-invasive or indigenous counterparts 42 , with the hypothesis that small genome size may confer biological advantage for adaptation in novel habitats, possibly due to enhanced tolerance of extreme environments or via altered regulatory gene divergence 35,46 . Given the challenging environmental conditions frequently encountered across inland Australia, and the successful establishment of these particular invasive grasses across southern Australia, the smaller genome size of the majority of non-native Panicum species investigated could be considered as supporting evidence for this hypothesis 47,48 .
Our results have shown that the use of the nuclear ITS region (and to a lesser extent the two cpDNA regions, namely matK and trnL intron-trnF) allowed clear identification and differentiation for eight of ten Panicum species evaluated, with only P. buncei and P. coloratum unable to be segregated using this method. We suggest www.nature.com/scientificreports/ www.nature.com/scientificreports/ that additional loci are likely required for further resolution at the species level, assuming the original taxonomic identification was correct. With the exception of P. buncei, discrimination between native and non-native species was achieved. Further studies to evaluate additional Panicum species from diverse habitats across Australia could confirm the utility of this approach. In addition to the techniques presented, other molecular tools, including whole or partial genome sequencing 49 , high resolution melt curve analysis 50 , short tandem repeats (STR) 51 , or some combination of the above, may prove useful for rapid and refined species differentiation through estimation of other genetic parameters.
In conclusion, this study reports the use of a DNA barcoding method for distinguishing field samples of Panicum species regardless of phenological growth stage, in isolation or in combination with traditional morphological identification. Rapid identification of Panicum grasses, including those commonly implicated in crop and pasture incursions 4 or in hepatotoxicity outbreaks in livestock 7 , could assist producers, industry advisors, agronomists and weed scientists to identify invasive grasses accurately and quickly for control or eradication. This knowledge may also provide further insight into changing patterns of species distribution, and facilitate the development of efficacious weed management practices to limit invasive incursions or toxic outbreaks in pastures and croplands in Australia and internationally.   www.nature.com/scientificreports/ small leaf section was collected from each plant and stored at -80 °C with silica gel to maintain tissue integrity before DNA extraction. In addition, fresh leaf tissue samples were also collected and stored at 4 °C for determining genome size using flow cytometry.

Materials and methods
To supplement field-collected plant material, an additional 23 dried leaf samples representing nine previously identified Panicum species (Table 1), were sampled from voucher specimens held within the CANB collection (Acton, ACT, Australia). Dried leaf segments from archived plants of each of targeted species were provided by David E. Albrecht. DNA extraction and barcoding. Genomic DNA extraction was performed as described previously 52 .
Sanger sequencing and DNA barcode GAP analysis. PCR products were bidirectional Sanger sequenced using the same primers by the Australian Genome Research Facility, Brisbane. Sequences were read in Geneious version 11.0.5 57 . Forty-three sequences from each locus were aligned with a cost matrix of 65% similarity (Geneious version 11.0.5). Sequence alignments were analysed using MEGA7.0.26 58 to calculate intraspecific and interspecific genetic distances with the Kimura 2-parameter (K2P) model. Sequences of three loci for each Panicum specimen were further concatenated for DNA barcode GAP analysis. Concatenated sequences of the same regions from Setaria italica, a member the tribe Paniceae, was used as an outgroup to root the tree. Phylogenetic relationships between species were inferred by MrBayes 3.2.6 59 using default settings (four gamma categories, Markov chain Monte Carlo (MCMC) setting include chain length 1 million, subsampling every 1000th generation, burn-in length was first 250,000 iterations) with GTR substitution model for the nuclear DNA locus (ITS) and GTR + R substitution model for two concatenated chloroplast DNA loci (matK and trnL intron-trnF) as suggested by JModelTest 2.1.10 60 .
Flow cytometry. Fresh leaf tissue was stored at 4 °C in moist paper towelling with cytometric analysis performed within 48 h using a Gallios Flow Cytometer (Beckman Coulter, USA). Depending on the species analysed, Raphanus sativus L. (red globe radish, 1C = 0.55 pg), or Solanum lycopersicum L. (tomato, 1C = 1.06 pg) were used as internal reference species for assessment of genome size. R. sativus was also used to calibrate S. lycopersicum within each run to confirm the reliability of each run. A composite leaf tissue sample of each targeted Panicum species and the reference plant, similar in size, were chopped using a clean razor blade in a premixed buffer solution, consisting of 1 ml WPB nuclear isolation buffer (0.2 M Tris. HCl, 4 mM MgCl 2 .6H2O, 2 mM EDTA Na 2 .2H 2 O, 86 mM NaCl, 10 mM sodium metabisulfite, 1% PVP-10, 1% (v/v) Triton X-100, pH 7.5) 61 , 50 μg propidium iodide (PI) (Sigma-Aldrich, Castle Hill, New South Wales, Australia), and 10 μl RNase A solution. At least 10,000 nuclei were analysed each run. Each specimen was analysed in triplicate with three technical replicates within 7 days of leaf collection to ensure reproducibility 62 .