DNA barcoding: a new tool for palm taxonomists?

Abstract

Background and Aims

In the last decade, a new tool – DNA barcoding – was proposed to identify species. The technique of DNA barcoding is still being developed. The Consortium for the Barcode of Life's Plant Working Group (CBOL-PWG) selected two core markers (matK and rbcL) that now must be tested in as many taxa as possible. Although the taxonomy of palms (Arecaceae/Palmae) has been greatly improved in the past decades, taxonomic problems remain. Species complexes, for example, could significantly benefit from DNA barcoding. Palms have never before been subjected to a DNA barcoding test.

Methods

For this study, 40 out of the 48 species of the southeast Asian tribe Caryoteae (subfamily Coryphoideae) were included. In total, four DNA markers – three plastid encoded (matK, rbcL and psbA-trnH) and one nuclear encoded (nrITS2) – were analysed to determine if adequate variation exists to discriminate among species.

Key Results

The combination of three markers – matK, rbcL and nrITS2 – results in 92 % species discrimination. This rate is high for a barcoding experiment. The two core markers suggested by the CBOL-PWG, rbcL and matK, have a low species discrimination rate and need to be supplemented by another marker. In Caryoteae, nrITS2 should be chosen over psbA-trnH to supplement the two ‘core’ markers.

Conclusions

For the first time a test of DNA barcoding was conducted in Arecaceae. Considering that palms have highly variable mutation rates compared with other angiosperms, the results presented here are encouraging for developing DNA barcoding as a useful tool to identify species within this ecologically important tropical plant family.

INTRODUCTION

Recent estimates state that as few as 4 % to approx. 40 % of the species on earth have been described in the scientific literature (Systematics Agenda, 1994; Pennisi, 2003). The intertropical zone shelters high levels of species diversity (Reaka-Kudla et al., 1996; Kress and Erickson, 2008) but faces the strongest threats of extinction (Laurance and Peres, 2006), inducing an ever increasing rate of species loss (Pimm and Raven, 2000; Thomas et al., 2004).

A huge amount of research is still needed to produce a satisfactory inventory of biodiversity. This, combined with the range and intensity of threats to this diversity, has served to emphasize the role of systematics in modern biology (Wilson, 1989; May, 1990). A new tool for identifying species called ‘DNA barcoding’ was proposed earlier this decade by Hebert and colleagues (Hebert et al., 2003, 2004). They advocated the use of short DNA sequences for biological identification. This innovation gave rise to many controversial questions about the nature and purpose of systematics and appropriate funding levels for various sub-disciplines (Lipscomb et al., 2003; DeSalle et al., 2005; Rubinoff, 2005; DeSalle, 2006; Rubinoff et al., 2006).

For the last several years, the plant barcoding community and especially the Plant Working Group of the Consortium for the Barcode of Life (CBOL-PWG) have focused on the identification of a universally informative plant barcode (CBOL Plant Working Group, 2009). In the end, matK and rbcL were selected by the CBOL-PWG as ‘core’ markers for plant identification (http://www.barcoding.si.edu/PDF/PlantWG/CBOL%20Decision%20-%20Plant%20Barcode%20Regions.pdf). It is imperative that these markers be tested in a broad range of laboratories across a diverse set of land plants. Although these markers have been widely used to construct phylogenies in Arecaceae, they have not been evaluated in a barcoding context.

Almost completely restricted to tropical and sub-tropical areas, palms (Arecaceae/Palmae) contain about 2400 species (Govaerts and Dransfield, 2005) and are the sixth largest monocot family. Although the taxonomy of this family has greatly improved at the generic level (Dransfield et al., 2008), problems remain (Baker et al., 2011). This is especially true at the species level. Several species complexes have been identified that are hard to resolve using only morphological characters [e.g. Calamus or Geonoma (Henderson and Martins, 2002; Henderson, 2011)]. Palms are considered study models in the field of tropical forest ecology (Dransfield, 1988; Kahn and De Granville, 1992; Dransfield et al., 2008), but remain poorly represented in herbaria, leading to a severe lack of taxonomic advancement (Henderson et al., 1995). One of the reasons for the relative scarcity of palms in herbaria is that they are more difficult to collect than most other plants due to their (often) large size. This lack of collection and taxonomic knowledge is even more problematic considering that many species in Arecaceae are known to be under major threat (Johnson and IUCN/SSC, 1996). In some palm groups, plastid DNA seems to evolve more slowly than in other monocots (Gaut et al., 1992, 1996; Asmussen, 1999a, b; Hahn, 2002; Lewis and Doyle, 2002; Gunn, 2004; Roncal et al., 2005; Thomas et al., 2006), whereas in others the rate of evolution seems unexpectedly high (Bayton, 2005; Cuenca and Asmussen-Lange, 2007). This makes the analysis and the understanding of inter- and intraspecies relationships more complex. Thus, DNA barcoding in palms could be challenging, but the development of such a tool would be very helpful to taxonomists and ecologists (Valentini et al., 2009). No DNA barcoding test in palms has ever been published.

Tribe Caryoteae (subfamily Coryphoideae) includes three genera, Caryota, Arenga and Wallichia, distributed from mainland Asia to the western Pacific and Australia. This group is being monographed by the first author. This tribe was selected for this experiment for different reasons. Amongst palms Caryota is one of the most difficult palm genera to collect (Dransfield, 1974), leading to a scarcity of material and thus the traditional alphataxonomic approach is difficult. Another reason is that Arenga caudata and Arenga hookeriana form a species complex for which the traditional morphological approach is uninformative (Hodel and Vatcharakorn, 1998; Henderson, 2009). Here, the usefulness of four DNA markers to discriminate among the species traditionally recognized within tribe Caryoteae were tested.

MATERIALS AND METHODS

Markers used

Four DNA regions were amplified. Following the requirements of the CBOL Plant Working Group (2009) parts of two plastid genes – matK and rbcL – were included. Two markers, psbA-trnH, a plastid intergenic spacer (Kress et al., 2005), and the nuclear ribosomal internal transcribed spacer (nrITS2) (Kress et al., 2005; Sass et al., 2007) were added.

Taxon sampling

Species delimitation used here was based on previous work: for Caryota the taxonomic reference was Hahn's revision (Hahn, 1992), for Arenga we followed the checklist published by Govaerts and Dransfield (2005), and for Wallichia, Henderson's revision was our reference (Henderson, 2007). A total of 40 species out of the approx. 48 found in Caryoteae were included in this study (Table 1). For some species more than one accession was examined, resulting in 88 samples in total (Table 1; Supplementary Data Table S1, available online). Material for DNA extraction was obtained from field trips and living collections (FTG, K and Montgomery Botanical Center).

Table 1.

List of the species included in the Tribe Caryoteae sorted by genus

Genus	Species	No. of individuals available for study	No. of samples for which all four markers were amplified	Total no. of sequences
Caryota	bacsonensis	2	0	5
	cumingii	0
	gigas	2	1	7
	kiriwongensis	2	1	7
	maxima	5	3	16
	mitis	4	0	13
	monostachya	4	1	11
	no	2	2	8
	obtusa	4	3	15
	ophiopellis	1	1	4
	rumphiana	1	1	4
	sympetala	3	0	2
	urens	1	1	4
	zebrina	1	0	3
Arenga	australasica	1	0	3
	brevipes	2	1	6
	caudata	10	6	36
	distincta	1	1	4
	engleri	5	2	15
	hastata	2	2	5
	hookeriana	4	1	3
	listeri	4	1	4
	longicarpa	1	0	4
Arenga	longipes	0
	micrantha	0
	microcarpa	1	0	3
	mindorensis	0
	obtusifolia	3	1	10
	pinnata	1	0	3
	plicata	0
	porphyrocarpa	2	1	7
	retroflorescens	1	0	2
	ryukyuensis	2	1	7
	talamauensis	0
	tremula	1	1	4
	undulatifolia	2	2	8
	westerhoutii	4	1	11
	wightii	1	1	4
Wallichia	caryotoides	1	0	3
	disticha	2	1	7
	gracilis	3	1	9
	lidiae	0
	marianneae	1	0	3
	nana	0
	oblongifolia	1	1	4
	triandra	0

For each species, the number of individuals (samples) included in the study (0 indicates that the species was not sampled), the number of individuals (samples) that provided a sequence for all four markers (0 indicates that the species was not included in the analysis because all four markers could not be sequenced) and the total number of sequences obtained for each species (submitted to GenBank) are given.

DNA extraction, amplification and sequencing

DNA was extracted from silica gel-dried (0·01–0·05 g) leaves using a modified cetyl trimethyl ammonium bromide (CTAB) extraction method as described in Kang et al. (1998) or using DNeasy^® Plant MiniKits (Qiagen Inc., Valencia, CA, USA). PCRs were conducted in a total reaction volume of 15 µL containing 8·13 µL of autoclaved ion-exchanged water, 1·5 µL of dNTP mixture (stock of 2·5 mm of each dNTP), 1·5 µL of bovine serum albumin (BSA; 0·25 µg μL⁻¹ stock), 1·5 µL of buffer [200 mm Tris pH 8·8, 100 mm KCl, 100 mm (NH₄)₂SO₄, 20 mm MgSO₄·7H₂O, 1 % (v/v) Triton X-100, 50 % (w/v) sucrose, 0·25 % (w/v) cresol red], 1 µL of each primer (0·67 µm final concentration), 0·3 µL of Taq polymerase and 0·25 µL of DNA. The amplicon size of matK ranges from 846 to 852 bp; primers used (5'–3') were CGTACAGTACTTTTGTGTTTACGAG and ACCCAGTCCATCTGGAAATCTTGGTTC (K. J. Kim, pers. com.), The amplicon size of rbcL is 654 bp; primers used (5'–3') were ATGTCACCACAAACAGAGACTAAAGC and GAAACGGTCTCTCCAACGCAT (Kress and Erickson, 2007; Fazekas et al., 2008). The amplicon size of psbA-trnH ranges from 318 to 820 bp. This length variation is mostly the result of small scattered insertions/deletions without an apparent taxonomic pattern (Supplementary Data Alignment File, available online). In addition there is an approx. 40 bp insertion, composed mostly of A/T, present in Caryota as well as in A. hastata and A. distincta. The primers used (5'–3') were GTTATGCATGAACGTAATGCTC and CGCGCATGGTGGATTCACAATCC (Sang et al., 1997; Tate and Simpson, 2003). The amplicon size of nrITS2 ranges from 538 to 559 bp; primers used (5'–3') were ATGCGATACTTGGTGTGAAT and GACGCTTCTCCAGACTACAAT (Shu-Jiau et al., 2007). PCR for matK and rbcL was performed with an initial denaturation of 2 min 30 s at 94 °C followed by ten cycles under the following conditions: 94 °C for 30 s, 54 °C for 30 s, 72 °C for 30 s, and 25 cycles under the following conditions: 88 °C for 30 s, 54 °C for 30 s and 72 °C for 30 s, terminated by an extension of 72 °C for 10 min. PCR for psbA-trnH was performed with an initial denaturation of 2 min and 30 s at 95 °C followed by 35 cycles under the following conditions: 95 °C for 30 s, 58 °C for 30 s, 64 °C for 1 min, terminated by an extension of 72 °C for 7 min. PCR for nrITS2 was performed with an initial denaturation of 3 min at 94 °C followed by 35 cycles under the following conditions: 95 °C for 30 s, 56 °C for 30 s, 72 °C for 30 s, terminated by an extension of 72 °C for 7 min.

The PCRs were run in a Gene Amp^® PCR system 9700 or on an Eppendorf vapo Protect Mastercycler pro S. PCR products were purified with ExoSAP-IT (exonuclease I and shrimp alkaline phosphatase). Sequencing was performed using the Big Dye terminator sequencing kit version 3·1 (Applied Biosystems, Carlsbad, CA, USA) and an ABI 3700 sequencer (sequencing conducted in the Department of Genome Sciences, University of Washington). The complete sequences were manually edited using Staden (http://staden.sourceforge.net/). Sequences and voucher information are archived in GenBank (accessions JF344793–JF345078; Supplementary Data Table S1)

Analysis of the data

Evaluation of the relative discriminatory power of the different markers was conducted with a suite of PERL scripts (CBOL Plant Working Group, 2009) available at http://www.nybg.org/files/scientists/dlittle/PWG.html. Statistical analyses used the R package (R v2·11·1; MASS v7·3-6; agricolae v1·0-9; http://cran.r-project.org/).

To make meaningful comparisons between markers, samples with missing sequence data were eliminated from the analyses, resulting in a matrix of 26 species and 39 sequences per marker. Micro- and macroinversions found in psbA-trnH were manually re-inverted to see if this could make a difference in the species discrimination. The original psbA-trnH sequences were analysed as one marker and the ‘corrected’ sequences were analysed as a separate marker.

For each marker, all possible pairwise global alignments were calculated using MUSCLE v3·6 (Edgar, 2004) and unambiguous nucleotide changes between sequences were tabulated (pairwise p-distances). A given species was considered distinct if all samples, for that species, could be unambiguously differentiated (i.e. one or more nucleotide changes resulting in an intraspecific distance greater than zero) from all other samples. Discrimination was calculated for marker combinations by summing the number of differences for each sample. An index of sequence depth (coverage) and quality (B) was calculated from manually edited contigs using QV30 as the minimum acceptable value (Little, 2010). Sequence quality was calculated for marker combinations using the mean value for each sample. Thus, estimates of average sequence quality for individual markers and all possible combinations of markers could be compared. Statistical differences in PCR success, species discrimination and sequence quality among markers were examined using the test of Scheffé (1953) at P = 0·05. The binomial distribution was used for the Scheffé tests of PCR success and species discrimination, while the Gaussian distribution was used for the Scheffé test of sequence quality.

RESULTS

The amplification success was of 91·7 % for psbA-trnH (abbreviated as p), 90·58 % for rbcL (abbreviated as r), 88·1 % for nrITS2 (abbreviated as i) and 73 % for matK (abbreviated as m). PCR success was not statistically different among markers (Scheffé test, P = 0·05) (Scheffé, 1953).

The best sequence quality (B₃₀; Little, 2010) was obtained from the ‘core’ markers with almost 0·9 for matK (m) and 0·93 for rbcL (r; Fig. 1). The B₃₀ of these two markers, individually as well as combined (mr), is statistically different from that of all other markers and their combinations (Scheffé test, P = 0·05) (Scheffé, 1953). There were no other unambiguously statistically distinct groupings based on sequence quality (with and without the inclusion of marker combinations). PsbA-trnH (p) has the lowest sequence quality value (B₃₀ = 0·69). The ‘corrected’ psbA-trnH sequences (d; 11 reverse complemented microinversions in positions 83–95 of the alignment; Supplementary Data Alignment File; Fig. 2) were plotted with sequence quality identical to the uncorrected sequences. Between the rbcL/matK (r/m) cluster and the psbA-trnH/‘corrected’ psbA-trnH (p/d) cluster is a cloud of different combinations of matK, psbA-trnH and rbcL (mpr, mp and pr). The sequence quality values are the average of psbA-trnH (p) and matK/rbcL (m/r), ranging from 0·79 to 0·82 (Fig. 1). For the nrITS2 (i) and combinations [nrITS2/matK/psbA-trnH (imp), nrITS2/psbA-trnH/rbcL (ipr), nrITS2/matK (im), nrITS2/matK/rbcL (imr), nrITS2/rbcL (ir)] sequence quality ranges between 0·72 and 0·83.

Fig. 1.

Percentage species discrimination in relation to sequence quality. m = matK; r = rbcL; p = psbA-trnH; i = nrITS2; d= ‘corrected’ psbA-trnH. Squares represent single barcodes. Diamonds represent various combinations of markers. Errors bars are 95 % confidence intervals. Combinations with ‘d’ are not shown for clarity (they exactly mirror the relationship between ‘d’ and ‘p’).

Fig. 2.

An example of microinversions within psbA-trnH sequences. ‘Inverted’ sequences can be ‘corrected’ to the ‘standard’ (most frequently retrieved) sequence type by replacing the bracketed segment with its reverse complement. Nucleotide positions correspond to the full MUSCLE alignment (Supplementary Data Alignment File, available online).

The percentage species discrimination with rbcL (r) alone, matK (m) alone or their combination (mr) is either low [26 % for rbcL (r) alone, the lowest value] or average [48 and 51·8 % for matK (m) and matK + rbcL (mr), respectively]. Sequence quality (B₃₀) for rbcL (r) and matK (m) is uniformity high. PsbA-trnH (p) combines a low B₃₀ with poor species discrimination (37 %) – the lowest species discrimination (26 %) is obtained with ‘corrected’ psbA-trnH sequences (d) and rbcL (r). Although not included in the final data set due to missing data for other markers, the psbA-trnH (p) sequences of Caryota mitis were almost entirely inverted. Different 5' and 3' ends in the two individuals sampled suggest that the inversions have independent origins. Within the cloud of various combinations of matK (m), psbA-trnH (p) and rbcL (r; Fig. 1), the percentage species discrimination ranges between 66 and 81 % – higher than the values obtained with the individual component markers. The highest species discrimination (92 %) is reached with nrITS2 (i) and its various combinations – although at the cost of lower sequence quality (B₃₀). For species discrimination, there were no unambiguously statistically distinct groupings (with, and without, the inclusion of marker combinations).

If identification fails, the size of the resulting ambiguously identified group is measured, for each marker, by the maximum number of indistinguishable species. The minimum value is 0 (meaning identification success) and the maximum value is the size of the study set (meaning an invariant marker). In Caryoteae, the maximum number of indistinguishable species ranges from two for nrITS2 (i) and its combinations (meaning that, at most, one will not be able to distinguish between a set of two species, using this marker) to 19 for rbcL (r; Fig. 3); meaning that if only this marker is used, in some cases, 19 different species are indistinguishable from another (Fig. 3). In contrast, psbA-trnH (p) alone has a maximum number of indistinguishable species of nine and, for some combinations in which it is involved, three [matK/rbcL/psbA-trnH (mrp) and matK/psbA-trnH (mp)]. In combination with psbA-trnH (pr), rbcL still presents a high maximum number of indistinguishable species (eight). If used alone, ‘corrected’ psbA-trnH (d) has the next highest maximum number of indistinguishable species – 16 species are potentially unrecognizable. The combination of ‘corrected’ psbA-trnH with other markers improves the situation [nrITS2/‘corrected’ psbA-trnH (id) = 2; rbcL/‘corrected’ psbA-trnH (rd) = 3] except if the other marker is rbcL [rbcL/‘corrected’ psbA-trnH (rd) = 12]. MatK (m) has an intermediate value if used alone as a barcode – six species are potentially indistinguishable.

Fig. 3.

Maximum number of indistinguishable species for each marker and combination of markers; d = psbA-trnH corrected, i = nrITS2, m = matK, p = psbA-trnH, r = rbcL (ir, for example, represents the combination of nrITS2 and rbcL).

DISCUSSION

Palms are well known in the systematics community for highly variable rates of plastid DNA evolution (Baker et al., 1999; Asmussen et al., 2000; Baker et al., 2000b; Dransfield et al., 2008). We therefore expected that DNA barcoding would not be very serviceable for species discrimination, but, interestingly, our results indicate that DNA barcoding may be a very useful tool – at least within Caryoteae. Sequence quality (B₃₀) ranged between 0·69 and 0·92 (Fig. 1). Both the statistically higher B₃₀ of matK, rbcL and their combination and the low B₃₀ for psbA-trnH conform to prior expectations (CBOL Plant Working Group, 2009). For the animal CO1 sequence, the minimum B₂₀ value, defined in the BARCODE data standard, is 0·75 (http://www.barcoding.si.edu/PDF/DWG_data_standards-Final.pdf). The CBOL has not yet defined sequence quality criteria for plant barcode markers. The criteria used by the CBOL Plant Working Group (2009) mandated a minimum B₃₀ of 0·375 (Little, 2010). To date, most plant barcoding studies have focused on species discrimination, and to a lesser extent marker universality. However, sequence quality must be a criterion in selecting a DNA barcode because the BARCODE data standard requires high-quality sequences (CBOL Plant Working Group, 2009). There is only one published analysis of plant barcode sequence quality: B₃₀ values ranged between 0·40 and 0·85 (CBOL Plant Working Group, 2009; Little, 2010). In comparison, our results are slightly better (0·69–0·93). The discrepancy is likely to be the result of focused vs. diverse sampling (Caryoteae vs. angiosperms, gymnosperms and cryptogams). Given the CBOL Plant Working Group (2009) criteria, all the markers and their combinations tested here provide a reliable enough sequence quality (Fig. 1).

A good DNA barcode should have acceptable sequence quality with species discrimination as high as possible. The combination of matK and rbcL does not provide good species discrimination in the current study (Fig. 1), and a broad range of levels of discrimination success has been obtained in other studies published to date (Hollingsworth et al., 2011). The low discrimination success of ‘core’ markers observed in Caryoteae is similar to what has been observed in Geonomateae and Archontophoenicinae (Conny B. Asmussen-Lange, pers. com.). However, the two ‘core’ regions selected by the CBOL should always be used for DNA barcoding, with supplemental markers, such as psbA-trnH or nrITS2, added as required. A uniform database of markers is required to maximize the utility of DNA barcoding. Therefore, the two ‘core’ makers (matK and rbcL) must be sequenced for all barcoded plants regardless of their utility within a given clade. Failure to create a uniform database of core marker sequences would require a barcoder first to identify a sample using morphology and then sequence the markers best represented in the database for that clade – completely negating the purpose of barcoding.

Taking into account sequence quality and species discrimination, three good options emerge from our data set: nrITS2 combined with matK (im), nrITS2 combined with matK and rbcL (imr) or nrITS2 combined with rbcL (ir; Fig. 1). Some combinations of markers resulted in species discrimination as high as 92 % [nrITS2/matK/psbA-trnH (imp), nrITS2/psbA-trnH/rbcL (ipr), nrITS2/matK (im), nrITS2/matK/rbcL (imr), nrITS2/rbcL (ir); Fig. 1], which is high for plant DNA barcoding. The results presented here show that the addition of a third marker to the two ‘core’ markers greatly improves species discrimination power. Our results also indicate that for Caryoteae nrITS2 is more informative than psbA-trnH and should be combined with rbcL and matK. The maximum number of indistinguishable species also justifies the choice of nrITS2 over psbA-trnH as a supplemental marker. Among the species included in this study, at most two cannot be distinguished (maximum number of indistinguishable species) if matK, rbcL and nrITS2 are used together (imr; Fig. 3) – making these markers quite accurate (although not 100 % reliable).

Multiple divergent nrITS copies are reported from Calamoid palms (Baker et al., 2000a). In a DNA barcode context, this potentially leads to misidentification due to differential sampling of divergent paralogues. We did not observe more than one band for any marker (visualized on agarose gels) nor did any of the electropherograms show evidence of overlapping signals that would be suggestive of the presence of divergent paralogues. These observations lead us to believe that, in tribe Caryoteae, nrITS2 can be reliably used. Compared with other studies that included nrITS2 as a barcode marker in monocots (Yao et al., 2010), our results are rather encouraging (74·2 % species discrimination in the monocots vs. 92 % in our study).

Other markers, or parts of markers, could be informative for barcoding in the palm family, such as ndhF (Hahn, 2002; Cuenca and Asmussen-Lange, 2007), rpb2 or PRK (Thomas et al., 2006; Trenel et al., 2007). These potentially informative markers were not selected by the CBOL for different reasons: for example, ndhF is not found in gymnosperms and some pteridophytes so it cannot be used as a universal marker.

The complex molecular evolution of psbA-trnH makes it difficult to use as a barcode (Lahaye et al., 2005; Whitlock et al., 2010). Microinversions in psbA-trnH (p; Fig.2) have been reported from a variety of angiosperms (reviewed in Whitlock et al., 2010). Whitlock et al. (2010) demonstrated that manually correcting inverted sequences has a positive impact on understanding of the relationship among sequences (i.e. phylogenies). In our case, ‘correcting’ the inversions resulted in lower species discrimination because the inversions are species specific in two-thirds of our sample (p vs. d in Fig. 1). Therefore, ‘correcting’ the microinversion harmonizes one-third of the species and destroys a diagnostic set of bases for two-thirds of the species.

To test if these promising results can be extended to the whole family, this experiment should be replicated with other groups of Arecaceae. If these results were to be applicable to the rest of the family, DNA barcoding could become an ideal supplementary tool for palm systematics and greatly enhance taxonomic knowledge especially within species complexes.

SUPPLEMENTARY DATA

Supplementary data are available online at www.aob.oxfordjournals.org and consist of the following. Table S1: samples, voucher information and GenBank accessions used in this study. Alignment File: MUSCLE alignment of psbA-trnH (.txt file). Positions identical to that of the first sequence are reported as points.