Forensic Science International 152 (2005) 121–131 www.elsevier.com/locate/forsciint

A molecular identification system for grasses: a novel technology for forensic botany J. Warda,*, R. Peakalla, S.R. Gilmorea,b, J. Robertsonc

aSchool of Botany and Zoology, Australian National University, Canberra, ACT 0200, Australia bCentre for Forensic Science, Canberra Institute of Technology, GPO Box 826, Canberra, ACT 2601, Australia cNational Manager, Forensic and Technical Services Division, Australian Federal Police, GPO Box 401, Canberra, ACT 2601, Australia

Received 29 April 2004; received in revised form 2 July 2004; accepted 7 July 2004 Available online 21 September 2004

Abstract

Our present inability to rapidly, accurately and cost-effectively identify trace botanical evidence remains the major impediment to the routine application of forensic botany. Grasses are amongst the most likely plant species encountered as forensic trace evidence and have the potential to provide links between crime scenes and individuals or other vital crime scene information. We are designing a molecular DNA-based identification system for grasses consisting of several PCR assays that, like a traditional morphological taxonomic key, provide criteria that progressively identify an unknown grass sample to a given taxonomic rank. In a prior study of DNA sequences across 20 phylogenetically representative grass species, we identified a series of potentially informative indels in the grass mitochondrial genome. In this study we designed and tested five PCR assays spanning these indels and assessed the feasibility of these assays to aid identification of unknown grass samples. We confirmed that for our control set of 20 samples, on which the design of the PCR assays was based, the five primer combinations produced the expected results. Using these PCR assays in a ‘blind test’, we were able to identify 25 unknown grass samples with some restrictions. Species belonging to genera represented in our control set were all correctly identified to genus with one exception. Similarly, genera belonging to tribes in the control set were correctly identified to the tribal level. Finally, for those samples for which neither the tribal or genus specific PCR assays were designed, we could confidently exclude these samples from belonging to certain tribes and genera. The results confirmed the utility of the PCR assays and the feasibility of developing a robust full- scale usable grass identification system for forensic purposes. # 2004 Elsevier Ireland Ltd. All rights reserved.

Keywords: Grasses; Molecular identification system; Forensic botany; Indels; PCR assay; Mitochondrial genome

1. Introduction either awareness or botanical knowledge among evidence- collection teams and prosecutors, and the difficulty in rou- Forensic botany is the study of plants and plant matter as tinely identifying trace material by traditional morphologi- they pertain to criminal investigations [1]. Botanical evi- cal methods, using whole-plant identification or botanical dence remains under-utilised in forensics due to the lack of experts [1]. Grasses are amongst the most likely plant species * Corresponding author. Tel.: +61 2 6125 8161; encountered in trace evidence searches, and have the poten- fax: +61 2 6125 5573. tial to provide links between crime scenes and individuals, E-mail address: [email protected] (J. Ward). relate an item to a suspect, support or disprove alibis or

0379-0738/$ – see front matter # 2004 Elsevier Ireland Ltd. All rights reserved. doi:10.1016/j.forsciint.2004.07.015 122 J. Ward et al. / Forensic Science International 152 (2005) 121–131 provide vital crime scene information. They provide con- sampled set of 20 grass taxa representing the Pooideae siderable potential as contact DNA evidence because of their and the Panicoideae subfamilies [13]. Within each sub- ubiquitous nature in both urban and rural environments, and family, two major tribes were represented and within each their morphological adaptations for seed dispersal [2]. Many tribe, two major genera were represented by each of two grass seeds have hooks, barbs, spines, hairs or sticky cover- species. Within the Poeae tribe, sampling was extended to ings that allow them to be transported easily by attaching to examine variation within species, with two cultivars of the bodies of animals and thus humans [3–5]. each of the four species represented (Table 1). The phy- Molecular technology is now being used routinely in logenetic framework for the sample design was based forensic investigations involving humans [6–8], but applica- on Watson and Dallwitz, Clark et al., GPWG and GPWG tion of these techniques to plant evidence is still novel [14–17]. [9–12]. Our present inability to routinely, rapidly and accu- The 20 samples, which were sequenced and used for rately identify trace botanical evidence remains the major primer design [13], provide our ‘control’ set of samples for impediment to the routine application of forensic botany. testing the PCR assays developed in this study. A further set Our objective is to design a molecular DNA-based of 26 species from within the Pooideae or Panicoideae identification system for grasses consisting of several subfamilies represent our ‘experimental’ set of samples PCR assays that, like a traditional morphological taxonomic (Table 1). All samples were sourced as seeds that were key, provide criteria that progressively identify an unknown obtained from seed, grain and turf companies within grass sample to a given taxonomic rank depending on the Australia and subsequently germinated in sterile soil in outcome of the previous step. Ultimately for those grasses of glasshouse conditions. Shortly after germination, leaf tissue potential widespread forensic interest, such as turf grasses samples were collected for DNA extractions. that are frequently encountered in the urban environment, identification to cultivar level will be essential. This might be 2.2. DNA extraction followed by further population level analyses in an attempt to match samples to a point of origin. Total genomic DNA was isolated from 0.5–1.0 g of As the first step towards the development of this grass seedling tissue that was ground in liquid nitrogen using a molecular identification system, we examined DNA sequence modified CTAB protocol of Doyle and Doyle [18]. The variation among a phylogenetically representative and hier- crude DNA was resuspended in 200 mL of TE Buffer (Tris archically sampled set of 20 grass species at chloroplast, 100 mM, EDTA 1 mM, pH 8.0) and excess polysaccharides mitochondrial and nuclear DNA loci [13]. The sampling removed by adding 1/15th volume of 20% (w/v) SDS and strategy was based on a sound phylogenetic framework to one-third volume of 5 M K-acetate (pH unadjusted) and maximise the chance of finding diagnostic markers, at dif- vortexing. The samples were then incubated at 4 8C for ferent taxonomic levels, that will characterise both known 30 min and the polysaccharides removed by centrifugation and unknown samples. Based on this analysis we identified a followed by a phenol: chloroform (1:1) extraction. The DNA set of putatively informative indels of variable size that were was then washed in 70% ethanol before final re-suspension diagnostic of subfamily, tribe and genus ranks within our set in 50 mL of TE buffer. DNA yield was estimated by elec- of 20 samples. The objective of this study was to develop trophoresis and visual comparison with lambda DNA PCR assays for these putatively diagnostic indels and to test (BioLabs) of known concentration. their utility for identifying unknown grass samples. Our specific objectives were: 2.3. Primer design of taxa-specific molecular 1. to develop PCR assays for five putatively informative markers indels located in the grass mitochondrial genome; 2. to develop a PCR method for the reliable assay of smaller DNA sequences spanning putatively informative indels indels on agarose gels; in the mitochondrial genome were selected as the starting 3. to test the utility of these molecular markers for grass point for primer design in this study. Primers were designed identification of twenty six unknown grass taxa; in conserved flanking regions on either side of the indels 4. to evaluate the future research needs for our continuing using the online primer design program Primer3 [19]. Our development of a robust grass identification system for goal was to obtain primers that would amplify all grasses forensic purposes. under standard PCR conditions without the need for species- specific optimisation, while providing diagnostic informa- tion for target groups. This is an important consideration for 2. Material and methods unknown samples. Primer sequences and PCR product sizes are summarised in Table 2. Below we briefly summarise 2.1. Sampling information about the target indels for these PCR assays and our predictions in terms of their utility for identification in Our initial study of the patterns, extent and location of unknown grass samples based on our knowledge of DNA DNA sequence variation was based on a hierarchically sequence variation [13]. J. Ward et al. / Forensic Science International 152 (2005) 121–131 123

Table 1 The sample design and outcomes of the PCR assays for the 20 control samples and 26 experimental samples Subfamily Tribe Genus Species POO PAN POE TRI PANI AND PAC PAS SOR ZEA ID- ID- ID- SF Tribe Genera Control Pooideae Poeae Lolium perenne Pooideae Poeae Lolium perenne Pooideae Poeae Lolium multiflorum Pooideae Poeae Lolium multiflorum Pooideae Poeae Festuca longifolia Pooideae Poeae Festuca longifolia Pooideae Poeae Festuca arundinacea Pooideae Poeae Festuca arundinacea Pooideae Triticeae Triticum aestivum Pooideae Triticeae Triticum durum Pooideae Triticeae Hordeum vulgare Pooideae Triticeae Hordeum sp. Panicoideae Paniceae Panicum maximum Panicoideae Paniceae Panicum coloratum Panicoideae Paniceae Paspalum notaturn Panicoideae Paniceae Paspalum dilatatum Panicoideae Andropogoneae Sorghum bicolor Panicoideae Andropogoneae Sorghum halepense Panicoideae Andropogoneae Zea mays Panicoideae Andropogoneae Zea diploperennis Experimental-group1 Pooideae Poeae Lolium rigidum Pooideae Poeae Festuca rubra Pooideae Triticeae Triticum monococcum Pooideae Triticeae Hordeum bulbosum Panicoideae Paniceae Panicum laxum X Panicoideae Paniceae Paspalum wettsteinii Panicoideae Andropogoneae Sorghum angustum Panicoideae Andropogoneae Zea luxurians Experimental-group2 Pooideae Poeae Puccinellia ciliata Pooideae Poeae Poa labillardieri Pooideae Poeae Dactylis glomerata Pooideae Triticeae Elymus scaber Pooideae Triticeae Agropyron elongatum Pooideae Triticeae Secale cereale Panicoideae Paniceae Cenchrus ciliaris X Panicoideae Paniceae Pennisetum clandestinum X Panicoideae Paniceae Digitaria milanjiana X Panicoideae Andropogoneae Themeda australis X Panicoideae Andropogoneae Andropogon gayanus X Panicoideae Andropogoneae Bothriochloa bladhii X Experimental-group3 Pooideae Bromeae Bromus stamineus X Pooideae Stipeae Austrostipa verticilliata X Pooideae Aveneae Avena sativa X Panicoideae Arundinelleae Arundinella setosa XX Panicoideae Neurachneae Thyridolepis mitchelliana XX Panicoideae Isachneae Coelachne perpusilla ****** ******* Key:*, subfamily, tribe or genus identification as indicated by the PCR assay; H, correct identification; X, incorrect identification; *, excluded sample (no PCR product amplified); POO, Pooideae subfamily; PAN, Panicoideae subfamily; POE, Poeae tribe; TRI, Triticeae tribe; PANI, Paniceae tribe; AND, Andropogoneae tribe; PAC, Panicum genus; PAS, Paspalum genus; SOR, Sorghum genus; ZEA, Zea genus. 124 J. Ward et al. / Forensic Science International 152 (2005) 121–131

Table 2 Primer sequence, primer lengths, annealing temperatures and PCR product size for PCR amplification of the taxa-specific molecular markers 0 0 Primer name Primer sequence 5 –3 Tm (8C) Length Product size Subfamily-specific PCR primers POO/PAN-nad7F AGCGRGCCAATGTATCAAT 60 19 215–150 POO/PAN-nad7R GCTCACGCCATTTGAGGT 60 18 – Tribe-specific PCR primers And/Pan-nad7F GAACGGAGAAGTGGTGGAAC 60 20 250–150 And/Pan-nad7R GGTCCTCCGACCAGATGTGT 62 20 – And/Pan-nad7F1 CCGCTAGCACGGTGGAGGT 66 19 – Poe/Tri-nad7F TACTGGCAAAGACCGTCTGG 61 20 320–200 Poe/Tri-nad7R GCTCACGCCATTTGAGGT 60 18 – Poe/Tri-nad7F1 GACGGAAATGGAAGGTCC 58 18 – Genus-specific PCR primers Zea/Sor-nad7F CAAAGGGACGGTTGAGCA 61 18 350–250 Zea/Sor-nad7R ACGGTCTTTGCCAGTAGTGC 60 20 – Zea/Sor-nad7F1 GAGGACCGACCTGGGTTT 60 18 – Pas/Pac-nad5F TAGGGAGCGCAATTGCTAGA 61 20 220–170 Pas/Pac-nad5R CCACTCACTGCTTCCCCTAA 60 20 – Pas/Pac-nad5F1 CAGCCGGACGGACTACTATA 58 20 –

2.4. Subfamily PCR assay 2.5. Tribe and genus PCR assays

The basis of our subfamily PCR assay is a large indel Several small indels of six or seven base pairs in the nad 7 with a maximum size of 64 bps in intron 1 of nad 7, which and nad 5 introns, respectively, distinguished among our distinguished the subfamily Pooideae from the Panicoi- control samples at the tribe and genus levels (Fig. 1). We deae, with the Pooideae characterised by the large deletion predict that these indels will assist in the identification of (Fig. 1). Given the wide representation of tribes and other grasses not included in our control samples. Within the genera within these two subfamilies in our control set Pooideae subfamily, our control species within the Poeae of samples, we predict that this indel will be phylogen- tribe were distinguished by a 6-bp deletion in intron 1 of nad etically informative well beyond the tribes and genera 7 from the Triticeae tribe and tribes within the Panicoideae. represented in our initial study. Note that some minor Therefore, a PCR assay spanning this indel is predicted to variation in the size of the PCR fragment is predicted identify Poeae but not other tribes. Within the Panicoideae within the Panicoideae subfamily due to smaller indels that subfamily, the control species within the Andropogoneae are known from the DNA sequences to occur in the same tribe were distinguished by a different 6-bp deletion in intron region but for which it was unavoidable to exclude in the 1ofnad 7 from the Paniceae and the tribes within the PCR assay. Pooideae. Similar to above, a PCR assay spanning this

Fig. 1. The informative indels located in the 20 control samples used to design PCR assays for the identification pathway, illustrating how the assays can progressively identify a grass sample to a given taxonomic level. J. Ward et al. / Forensic Science International 152 (2005) 121–131 125 indel is predicted to identify Andropogoneae but not other agarose gel electrophoresis. These PCR assays were also tribes. designed to be suitable for trace and degraded samples, by With the control samples we also identified indels within virtue of their small PCR product size. This is important the Panicoideae subfamily that enable some identification of because with many forensic samples, the quality, quantity genera. For example, within the Paniceae tribe, the Paspa- and purity of DNA are poor, or they often contains frag- lum samples exhibited a 7-bp deletion in intron 4 of nad 5 mentary and degraded DNA. that was not found in the related Panicum or other genera. Therefore, a PCR assay spanning this indel is predicted to 2.7. PCR conditions distinguish Paspalum from Panicum and all other genera. A similar case was found within the Andropogoneae tribe. The Approximately 20 ng of template DNA was used in a Zea samples were characterised by a 6-bp deletion in intron 20 mL PCR reaction consisting of 2 mLof10 reaction 1ofnad 7 which distinguished them from Sorghum and all buffer (Perkin-Elmer-100 mM Tris–HCl, pH 8.3, 500 mM other genera in our sample set. A PCR assay spanning this KCl, 15 mM MgCl2, 0.01% (w/v) gelatin), 0.2 mM dNTPs in region is predicted to identify Zea, but not other genera. an equimolar ratio and 0.75 units of TaqGold DNA poly- merase (5U/mL) (Perkin-Elmer). For subfamily PCR reac- 2.6. Development of an informative assay for detecting tions containing two primers, 0.2 mM each of forward and small indels reverse primers was used. For tribe/genus PCR reactions containing three primers, 0.15 mM of forward primer F, Standard agarose gel electrophoresis conditions do not 0.15 mM of internal forward primer F1, and 0.2 mMof provide sufficient resolution to distinguish fragments differ- reverse primer R were used per reaction. Amplification ing in size by only a few base pairs. Ordinarily, such analysis was performed in a Corbett Research Thermal Cycler with is achieved by the more time consuming acrylamide gel an initial cycle at 94 8C for 10 min, denaturation of 94 8C for electrophoresis. To overcome this limitation we used a 30 s, annealing of 55 8C for 30 s, extension of 72 8C for 30 s simple method that ensures reliable detection of these for thirty cycles and a final cycle of 72 8C for 5 min to smaller indels. This approach uses three primers in a single complete extension of amplified products. PCR reaction (Fig. 2). Two primers (F and R) were designed PCR products were separated on 2% agarose gels (Fisher in the conserved flanking regions on either side of the indel. Biotec), containing 0.5 mg/mL of ethidium bromide in TAE An internal left primer (F1) was also designed so that the last buffer (0.04 M Tris–acetate, 0.001 M EDTA) at 100 V for three base pairs of the 30 end of the primer corresponded with 90–120 min. PCR products were visualised using a UV the first three base pairs of the insertion. Hence, the taxa transilluminator. without the deletion produce two PCR products, while the taxa with the deletion amplifies only one PCR product. This 2.8. Tests of the PCR assays procedure provides a positive control by amplifying from all Pooideae and Panicoideae samples; and generates fragment To assess the effectiveness of our primer design and size differences between the taxon-specific PCR fragments PCR conditions we assayed the 20 control samples and the positive control, that can be readily resolved by from which the primer design was based [13]. In these cases

Fig. 2. An example of a tribe PCR identification assay, showing the positions of three primers (F, R, F1) designed to distinguish between samples containing small indels and the resulting PCR products. 126 J. Ward et al. / Forensic Science International 152 (2005) 121–131 the expected number of PCR fragments and their size were of approximately 64 bp in size in the Pooideae subfamily but known. not the Panicoideae subfamily (Fig. 3). As anticipated we did To determine whether our series of PCR assays could be detect some variation in the size of the PCR fragment within used to identify unknown samples we assayed 26 additional the Panicoideae. These differences may eventually be useful species representing a wide range of tribes and genera from for aiding the identification of samples at lower taxonomic within the Pooideae and Panicoideae subfamilies. Our levels. experimental samples were further divided into three groups. The Andropogoneae/Paniceae assay (And/Pan-nad7F, Group 1 consisted of eight samples representing different And/Pan-nad7R, And/Pan-nad7F1) successfully distin- species, but belonging to the same genera as represented in guished all Andropogoneae samples from the Paniceae our control set. Group 2 consisted of 12 samples representing samples. Andropogoneae samples produced only one PCR new genera, belonging to the four tribes represented in the fragment as a consequence of their diagnostic 6-bp deletion, control set. Group 3 consisted of six samples belonging to while all Paniceae samples produced two PCR fragments genera and tribes not represented by any of the control (Fig. 4). samples. The Poeae/Triticeae tribe assay (Poe/Tri-nad7F, Poe/Tri- To avoid any bias in the scoring of our results, this second nad7R, Poe/Tri-nad7F1) successfully distinguished all phase of testing was performed ‘blind’, in that knowledge of Poeae samples from the Triticeae samples. Poeae samples the species identity was not known to the first author (JW) produced only one PCR fragment as a consequence of their during the laboratory and scoring phases. A set of control diagnostic 6-bp deletion while all Triticeae samples pro- samples were also amplified along with the unknown sam- duced two PCR fragments (Fig. 5). ples for each PCR assay to act as positive controls and to The Zea/Sorghum genus assay (Zea/Sor-nad7F, Zea/Sor- provide size standards. nad7R, Zea/Sor-nad7F1) successfully distinguished all Zea samples from the Sorghum samples. Zea samples produced only one PCR fragment as a consequence of their diagnostic 3. Results 6-bp deletion while all Sorghum samples produced two PCR fragments (Fig. 4). 3.1. Outcomes of PCR trials for the control samples The Paspalum/Panicum genus assay (Pas/Pac-nad5F, Pas/Pac-nad5R, Pas/Pac-nad5F1) distinguished all Pas- For all 20 control samples the appropriate PCR assays palum samples from the Panicum samples. Paspalum produced the correct number and size of fragments, con- samples produced only one PCR fragment as a con- firming that our primer design and PCR conditions were sequence of their diagnostic 7-bp deletion, while all appropriate (Table 1). The subfamily PCR assay (POO/PAN- Panicum samples produced two PCR fragments nad7F, POO/PAN-nad7R) reliably detected a large deletion (Fig. 5).

Fig. 3. A subfamily identification PCR assay. The first 12 samples are identified as belonging to the Pooideae and the last eight samples are identified as belonging to the Panicoideae. C is the negative control lane, and a ladder marker was used as a size standard and loaded on either side of the samples. J. Ward et al. / Forensic Science International 152 (2005) 121–131 127

Fig. 4. A tribe and genus identification PCR assay. The first half of the gel is an Andropogoneae/Paniceae tribe assay. The first four samples are identified as belonging to the Paniceae and the last four samples are identified as belonging to the Andropogoneae. The second half of the gel is a Zea/Sorghum genus assay. The first two samples are identified as belonging to Sorghum and the last two samples are identified as belonging to Zea. C is the negative control lane, and a ladder marker was used as a size standard and loaded in the first lane of each of the assays.

3.2. Outcomes of PCR trials for the experimental samples markers, however they were different species. Seven out of the eight samples were correctly identified at all taxonomic 3.2.1. Group 1 samples—new species, but represented levels (Table 1). The exception was Panicum laxum that genera produced a single fragment similar to that in the case of This group contained eight samples representative of the Paspalum control samples, rather than the double fragment tribes and genera used to identify and design the molecular indicating the insertion as expected. Therefore, our predic-

Fig. 5. A tribe and genus identification PCR assay. The first half of the gel is a Paspalum/Panicum genus assay. The first two samples are identified as belonging to Paspalum and the last two samples are identified as belonging to Panicum. The second half of the gel is a Poeae/ Triticeae tribe assay. The first eight samples are identified as belonging to the Poeae and the last four samples are identified as belonging to the Triticeae. C is the negative control lane and a ladder marker was used as a size standard and loaded in the first lane of each of the assays. 128 J. Ward et al. / Forensic Science International 152 (2005) 121–131 tion that this 7-bp deletion is diagnostic of Paspalum does level. Our group 3 exemplified a set of samples for which the not hold. tribal- or genus-specific PCR assays were not designed; nonetheless, we could confidently exclude these samples 3.2.2. Group 2 samples—new genera, but represented from belonging to certain specific tribes and genera. These tribes results confirm the potential utility of our molecular identi- This group contained samples representing 12 additional fication system, and the feasibility of developing a full-scale genera. However, these genera all belonged to four tribes usable system for forensic or other needs. represented in the control samples. All 12 samples were As we anticipated would happen, in this early phase of correctly assigned to their subfamily and tribes by the our development of a molecular identification system most respective assays (Table 1). At the generic level, all samples of our existing indel-based assays only provide a partial were correctly excluded from belonging to either Paspalum identification. Based on our control set of 20 samples only, or Zea because they exhibited the two PCR products known a common feature of our tribal- and genus-level assays was to characterise other genera within those tribes. that the deletion (i.e. the shorter sequence) appeared to be diagnostic of a specific tribe or genus. However, when we 3.2.3. Group 3 samples—new tribes expanded our sampling to include samples representing This group contained six species belonging to tribes not additional tribes or genera we sometimes discovered the represented in the control samples. The DNA sample of deletion was also found in related samples. For example, Coelachne perpusilla originating from an old herbarium the tribal assay within the Pooideae revealed that the specimen did not produce a PCR fragment at this subfamily deletion initially detected in the Poeae, is also common level and was therefore excluded from subsequent analyses. to its sister tribe Aveneae. The lack of a deletion was The remaining five samples were correctly placed into the common to the Triticeae as well as the relatively closely appropriate subfamily (Table 1). At the tribe and generic related Stipeae and its sister taxa Bromeae. Hence, this levels where no specific assay existed for these samples, deletion actually appears to distinguish a deeper taxo- PCR fragments were still produced. For the representatives nomic group than tribes, but still along accepted phylo- of the Bromeae and Stipeae two PCR fragments of similar genetic lines (see [15,17,20,21] for current phylogenetic sizes to the bands for the Triticeae were produced; therefore; views of grass relationships). these samples could not belong to the Poeae. For Aveneae a We encountered only one unexpected finding within our single band of similar size to the Poeae was produced, experimental group of samples: the misplacement of Pani- excluding it from the Triticeae. For the Neurachneae and cum laxum with Paspalum species. The taxonomy and Arundinelleae two PCR fragments of similar sizes to the systematics of the tribe Paniceae with almost half the genera Paniceae were produced for the tribal assay, confirming and 60% of the Panicoid grasses, has been recognised as the these samples could not belong to the Andropogoneae, while central phylogenetic problem of the subfamily [22].In the generic level assay indicated these samples did not particular, molecular phylogenetic studies indicate the genus belong to Paspalum. Panicum is polyphyletic [21–25]. This suggests that the These findings indicate the potential to misidentify sam- current taxonomy of the genus does not reflect natural ples below the subfamily level for those tribes not repre- evolutionary groupings. There is also some uncertainty sented in our control samples with our existing set of PCR about the phylogenetic relationship between Paspalum assays. Some explanations for our findings and solutions to and Panicum species [22,23,26]. Our finding that Panicum this problem are discussed below. Nevertheless, although laxum contains the same 7-bp deletion that we found in the positive identification was not achieved in these cases, we other species of Paspalum may indicate that Panicum laxum were able to confidently exclude some genera and tribes is more closely related to Paspalum or other genera than to from consideration. other species of Panicum and is perhaps misplaced in this genus within the current taxonomy. Alternatively, this 7-bp deletion may not be diagnostic of the genus Paspalum as a 4. Discussion whole, but rather a subgeneric grouping instead. Phylogenetic uncertainty is not restricted to the grass We have confirmed that for our control set of 20 grass genera Panicum and Paspalum. Indeed the grass literature samples, our molecular identification system of five primer offers many other examples [17], indicating that our combinations can identify samples to a given taxonomic existing understanding of the phylogeny of the Poaceae rank. Furthermore, we successfully extended this approach remains incomplete and some questions remain to be to correctly identify unknown samples. Our group 1 samples resolved regarding the generic- and species-level classifi- being different species, but belonging to genera represented cations. As long as these ambiguities are recognised and in our control set, were all correctly identified to the genus taken into consideration when designing taxa-specific level with one exception. Similarly, our group 2 samples genetic markers, they should not pose problems for the being different genera, but belonging to tribes represented in design and implementation of a molecular identification our control set, were also correctly identified to the tribal system. J. Ward et al. / Forensic Science International 152 (2005) 121–131 129

4.1. Features of the molecular identification system species in the sequence databases likely limit the application of this approach in botanical forensics. The outcomes of this study, demonstrate the ‘in principle’ feasibility of developing a molecular identification system 4.2. Future development and research directions for grasses that operates as a molecular key (Fig. 1). For the first PCR assay, one can determine to which subfamily a One limitation of using a diagnostic indel for identifica- grass sample belongs. Depending on the outcome of this tion is that while the presence of the deletion can confirm a initial assay, one then chooses the appropriate PCR assay sample belongs in a specific group, the absence of the that identifies to which tribe the sample in hand belongs and deletion may characterise the remaining groups; and there- so on, until an identification is made. The taxonomic rank to fore cannot be diagnostic. However we can exclude a sample which identification can be made will vary among taxa from belonging to a taxon with the deletion by such PCR depending on the taxon in question, but certain taxa can assays. Additional diagnostic markers will need to be devel- also be excluded, thereby providing increasingly precise oped to enable identification of the remaining taxa at the taxonomic information. different taxonomic levels. Again this process of identifica- An important consideration in the development of a tion can proceed as a molecular key, with the choice of which molecular system for species identification is that despite PCR assay to pursue depending on the outcome at previous the availability of a wide and growing array of DNA profil- steps. ing techniques (see reviews of [27–32]), not all techniques Other strategies available for improving this identifica- will meet forensics standards or be transferable among tion system include incorporating negative checkpoints and forensic laboratories. This molecular identification system multiple assays at strategic steps. Firstly, in a well-con- reduces species identification to a set of simple PCR tests structed traditional morphological taxonomic key, if one targeting phylogenetically informative indel sites. The five takes the wrong turn the subsequent options should imme- indels tested here are all from introns in the mitochondrial diately indicate this is the wrong pathway. In a similar way, genome, and four of them are from intron 1 in the nad 7 gene. with our proposed molecular identification system, it would Mitochondrial DNA sequences have the advantage of being be ideal to ensure that if one takes a wrong pathway this can in generally high copy number in plant and animal cells, be readily detected at the next step. For example, if one making the recovery of a useful amount of DNA less misinterprets the results of a tribal-level assay and therefore problematic with forensic samples. By targeting just the chooses the wrong generic assay when proceeding to the mitochondrial DNA and not a combination of organelle next level of identification, this error should be obvious genomes (e.g. nuclear and/or chloroplast DNA), the possi- because either no product is formed, or products of very bility of differential copy number presenting problems in the different size to those expected are generated. In order to amplification reactions is reduced. achieve this, primer design considerations will need to take In this system the PCR targets specific single locus DNA into account both the need to span a diagnostic region (such sequence and avoids the problems associated with reprodu- as an indel) and to be positioned in those flanking regions cibility and profile complexity encountered with other sys- that provide informative differences between the right and tems such as amplified fragment length polymorphisms wrong pathways. Whether this is feasible or not will depend (AFLPs) and randomly amplified polymorphic DNA on the degree of phylogenetically informative sequence (RAPDs) [27,29]. Even if identification to species or genus variation in the relevant taxa. The second way to enhance is not possible, identification with confidence to higher the reliability of this system for identification is to incorpo- taxonomic rank may be informative to the direction of an rate more than one PCR assay at important junctions in the investigation. This method could also enable the identifica- identification pathway,thus providing a double-checking tion of the multi-components of botanical mixtures without system. cloning, which would normally pose difficulties for AFLPs, There is a decreasing availability of indels at lower RAPDs and sequence-based approaches. taxonomic levels [13]. Therefore, for genera-, species- The simplicity of these assays will, in theory, allow easy and cultivar-level identification alternative diagnostic mar- conversion to kit form and will minimise the expertise, kers will be necessary. Single nucleotide polymorphisms expense and equipment required in contrast to sequencing (SNPs) are single-base changes at a specific position in the methods. Nonetheless, in some cases it may remain desirable genome, in most cases with two alleles (reviewed in [31,36]) to sequence data or even proceed with a morphology-based that promises to be particularly useful at these taxonomic approach for identification, but the molecular identification levels. SNP abundance combined with relatively low muta- system can rapidly narrow the search for useful compar- tion rates [37,38] facilitates SNP identification and subse- isons. Several other identification systems combine DNA quent utilisation as forensic markers. SNPs are readily sequencing and GenBank searches to make either an exact detected in grasses [39,40] and we have begun to catalogue match or to retrieve sequences on which to compare to crime their presence in our grass DNA sequence analysis. For cases scene samples through phylogenetic analyses [33–35]. The requiring species or cultivar identification with SNPs, agar- vast numbers of plant species and the poor coverage of ose based assays such as allele-specific PCR could be used 130 J. Ward et al. / Forensic Science International 152 (2005) 121–131

(e.g. [41]), although in some cases more technically involved and discrimination of grass species on the basis of their seed SNP assays may be necessary. proteins, J. For. Sci. Socy. 23 (1983) 297–305. In order to identify informative DNA sequence we need a [3] C.A. Lamp, S.J. Forbes, J.W. Cade, Grasses of Temperate hierarchically sampled and phylogenetically representative Australia: a Field Guide, Inkata Press, Melbourne, 1990. [4] W.S. Judd, C.S. Campbell, E.A. Kellogg and P.F. Stevens, Plant set of samples to maximise the prospects of identifying Systematics: A Phylogenetic Approach, Sinauer Associates related samples not included in the design of the identification Inc., U.S.A., 1999. system. The next phase of our research is to obtain sequences [5] P.H. Raven, R.F. Evert, S.E. Eichhorn, Biology of Plants, W.H. from several suitable genes for a set of samples of the most Freeman and Company, U.S.A, 1999. important grass taxa, including representatives of all major [6] E. Hagelberg, I.C. Gray, A.J. Jeffreys, Identification of the grass subfamilies. These sequences will be subjected to skeletal remains of a murder victim by DNA analysis, Nature phylogenetic analysis to identify indels or SNPs that can 352 (1991) 427–429. be targeted by simple and reliable PCR tests to provide useful [7] J.M. Butler, B.C. Levin, Forensic applications of mitochon- taxonomic information to assist identification of trace bota- drial DNA, Foc. 16 (1998) 158–162. nical evidence. The molecular identification system we are [8] M. Bataille, K. Crainic, M. Leterreux, M. Durigon, P. de Mazancourt, Multiplex amplification of mitochondrial DNA developing for grasses may also serve as a model system for for human and species identification in forensic evaluation., the development of similar methods to verify the species Forensic Sci. Int. 99 (1999) 165–170. composition of a Chinese medicine or herbal tea or to identify [9] A. Linacre, J. Thorpe, Detection and identification of cannabis dangerous or endangered botanical substances. by DNA, Forensic Sci. Int. 91 (1998) 71–76. [10] L. Congiu, M. Chicca, R. Cella, R. Rossi, G. Bernacchia, The use of random amplified polymorphic DNA (RAPD) markers 5. Conclusion to identify strawberry varieties: a forensic application, Mol. Ecol. 9 (2000) 229–232. We have demonstrated the feasibility of developing a [11] L.D. Kumar, M. Kathirvel, G.V. Rao, J. Nagaraju, DNA full-scale molecular identification system that will ulti- profiling of disputed chilli samples (Capsicum annum) using ISSR-PCR and FISSR-PCR marker assays, Forensic Sci. Int. mately provide the forensic community with a rapid, reli- 116 (2001) 63–68. able, objective and informative identification method in [12] S. Gilmore, R. Peakall, J. Robertson, Short tandem repeat grasses. Importantly, because the method is PCR based it (STR) DNA markers are hypervariable and informative in will be applicable to trace and degraded samples that are Cannabis sativa: implications for forensic investigations, For- frequently encountered in forensic investigations, but for ensic Sci. Int. 131 (2003) 65–74. which identification by traditional morphological methods [13] J. Ward, A molecular forensic identification system for trace fail. Beyond applications in forensic botany, this method botanical evidence, Honours Thesis, The Australian National also offers a tool of widespread utility where identification of University, 2002. trace botanical material can provide vital information that in [14] L. Watson and M.J. Dallwitz, The Grass Genera of the World, the past has been ignored or unattainable. C.A.B International, UK, 1992. [15] L.G. Clark, W. Zhang, J.F. Wendel, A phylogeny of the grass family (Poaceae) based on ndhF sequence data, Syst. Bot. 20 (1995) 436–460. Acknowledgements [16] Grass Phylogeny Working Group, A phylogeny of the grass family (Poaceae) as inferred from eight character sets, in: S. We thank the Forensic and Technical Services Division Jacobs, J. Everett (Eds.), Grasses: systematics and evolution, of the Australian Federal Police for funding this project and CSIRO Publishing, Melbourne, 2000, pp. 3–7. Dan Ebert for his technical assistance. We also acknowledge [17] Grass Phylogeny Working Group, Phylogeny and subfamilial the seed, grain and turf suppliers, the Australian Tropical classification of the grasses, (Poaceae), Ann. MO. Bot. Gard. Crops and Forages Genetic Resource Centre, the Australian 88 (2001) 373–457. Winter Cereals Collection, the Australian National [18] J.J. Doyle, J.L. Doyle, A rapid DNA isolation procedure for small quantities of fresh leaf tissue, Phytochem. Bull. 19 Botanic Gardens, CSIRO Plant Industry, and the Australian (1987) 11–15. National Herbarium for providing grass seed samples or [19] S. Rozen, H.J. Skaletsky, Primer3 on the WWW for general grass material. users and for biologist programmers, in: S. Krawetz, S. Mis- ener (Eds.), Bioinformatics methods and protocols: Methods in molecular biology, Humana Press, NJ, 2000, pp. 365–386. References [20] R.J. Soreng, J.I. Davis, Phylogenetic structure in Poaceae subfamily Pooideae as inferred from molecular and morpho- [1] H. Miller-Coyle, C. Ladd, T. Palmbach, H.C. Lee, The green logical characters: misclassification versus reticulation, in: S. revolution; Botanical contributions to forensics and drug Jacobs, J. Everett (Eds.), Grasses: systematics and evolution, enforcement, Croat. Med. J. (2001) 42. CSIRO Publishing, Melbourne, 2000, pp. 61–74. [2] D.J. Carroll, M. Cambell, J.W. Thorpe, J. Robertson, The [21] E.A. Kellogg, Classification of the grass family, Flora of potential use of isoelectric focusing for the identification Australia 43 (2002) 19–36. J. Ward et al. / Forensic Science International 152 (2005) 121–131 131

[22] L.M. Giussani, J. Hugo Cota-Sanchez, F.O. Zuloaga, E.A. [33] R. Bever, M. Cimino, Forensic Molecular Biology: identifica- Kellogg, A molecular phylogeny of the grass subfamily Pani- tion of plants from trace evidence., in: Twelfth International coideae (Poaceae) shows multiple origins of C4 photosynth- Symposium on Human Identification Biloxi, Mississippi, esis, Am. J. Bot. 88 (2001) 1993–2012. 2001. [23] R. Gomez-Martinez, A. Culham, Phylogeny of the subfamily [34] H. Hsieh, H. Chiang, L. Tsai, S. Lai, N. Huang, A. Linacre, J.C. Panicoideae with emphasis on the tribe Paniceae: evidence Lee, Cytochrome b gene for species identification of from the trnL-F cpDNA region, in: S. Jacobs, J. Everett (Eds.), the conservation animals, Forensic Sci. Int. 122 (2001) 7– Grasses: systematics and evolution, CSIRO Publishing, Mel- 18. bourne, 2000, pp. 136–140. [35] S.K. Verma, L. Singh, Novel universal primers establish [24] F.O. Zuloaga, O. Morrone, L.M. Giussani, A cladistic analysis identity of an enormous number of animal species for forensic of the Paniceae: a preliminary approach, in: S. Jacobs, J. application, Mol. Ecol. Notes 3 (2003) 28–31. Everett (Eds.), Grasses: systematics and evolution, CSIRO [36] A.J. Brookes, The essence of SNPs, Gene 234 (1999) 177–186. Publishing, Melbourne, 2000, pp. 123–135. [37] S. Gilmore, R. Peakall and J. Robertson, Organelle DNA [25] M.R. Duvall, J.D. Noll, A.H. Minn, Phylogenetics of Paniceae haplotyping reveals information about crop-type and prove- (Poaceae), Am. J. Bot. (2001) 88. nance of Cannabis samples. XIX International Congress of [26] S. Mathews, R.E. Spangler, R.J. Mason-Gamer, E.A. Kellogg, Genetics, Melbourne, 2003. Phylogeny of Andropogoneae inferred from phytochrome B, [38] R. Thomson, J.K. Pritchard, P. Shen, P.J. Oefner, M.W. Feld- GBSSI, and ndhF, Int. J. Plant Sci. 163 (2002) 441–450. man, Recent common ancestry of human Y chromosomes: [27] M.K. Morell, R. Peakall, R. Appels, L.R. Preston, H.L. Lloyd, Evidence from DNA sequence data, Proc. Natl. Acad. Sci. DNA profiling techniques for plant variety identification, Aust. U.S.A. 97 (2000) 7360–7365. J. Exp. Agric. 35 (1995) 807–819. [39] J.A. Rafalski, Novel genetic mapping tools in plants: [28] K. Weising, H. Nybom, K. Wolff, W. Meyer, DNA Finger- SNPs and LD-based approaches, Plant Sci. 162 (2002) 329– printing in Plants and Fungi, CRC Press, USA, 1995. 333. [29] R. Peakall, PCR-based genetic markers and their applications [40] V. Kanazin, H. Talbert, D. See, P. DeCamp, E. Nevo, T. Blake, to Turfgrass breeding, ITS Res. J. 8 (1997) 243–259. Discovery and assay of single-nucleotide polymorphisms in [30] G.K. Chambers, E.S. MacAvoy, Microsatellites: consensus and barley (Hordeum vulgare), Plant Mol. Biol. 48 (2002) 529– controversy, Comp. Biochem. Physiol. B Biochem. Mol. Biol. 537. 126 (2000) 455–476. [41] C.R. Newton, A. Graham, L.E. Heptinstall, S.J. Powell, C. [31] I. Gut, Automation in genotyping of single nucleotide poly- Summers, N. Kalsheker, J.C. Smith, A.F. Markham, Analysis morphisms, Hum. Mutat. 17 (2001) 475–492. of any point mutation in DNA The amplification refractory [32] D. Klein, Quantification using real-time PCR technology: appli- mutation system (ARMS), Nucl. Acids Res. 17 (1989) 2503– cations and limitations, Trends Mol. Med. 8 (2002) 257–260. 2516.