RESEARCH

A gDNA Microarray for Genotyping Salvia Species

Alexandra Olarte • Nitin Mantri • Gregory Nugent •

Hans Wohlmuth • Chun Guang Li •

Charlie Xue • Edwin Pang

Published online: 27 December 2012

� Springer Science+Business Media New York 2012

Abstract Salvia is an important genus from the Lamia-

ceae with approximately 1,000 species. This genus is dis-

tributed globally and cultivated for ornamental, culinary,

and medicinal uses. We report the construction of the first

fingerprinting array for Salvia species enriched with poly-

morphic and divergent DNA sequences and demonstrate

the potential of this array for fingerprinting several eco-

nomically important members of this genus. In order to

generate the Salvia subtracted diversity array (SDA) a

suppression subtractive hybridization (SSH) was per-

formed between a pool of Salvia species and a pool of

angiosperms and non-angiosperms to selectively isolate

Salvia-specific sequences. A total of 285-subtracted geno-

mic DNA (gDNA) fragments were amplified and arrayed.

DNA fingerprints were obtained for fifteen Salvia geno-

types including three that were not part of the original

subtraction pool. Hierarchical cluster analysis indicated

that the Salvia-specific SDA was capable of differentiating

S. officinalis and S. miltiorrhiza from their closely related

species and was also able to reveal genetic relationships

consistent with geographical origins. In addition, this

approach was capable of isolating highly polymorphic

sequences from chloroplast and nuclear DNA without

preliminary sequence information. Therefore, SDA is a

powerful technique for fingerprinting non-model plants and

for identifying new polymorphic loci that may be devel-

oped as potential molecular markers.

Keywords Microarray � Genotyping � Salvia � Diversity

Introduction

Salvia (Lamiaceae) is an important genus with approxi-

mately 1,000 species distributed widely in many regions of

the world including the Mediterranean area, southern

Africa, Central and South America, and Asia [1]. This

genus is largely cultivated for ornamental, culinary, and

medicinal uses [2, 3]. For example, S. officinalis L. is used

to preserve foods and employed as a spice for flavoring [3].

Its essential oil is used in perfumery and cosmetics [4] and

the plant and its derivates are known to have a wide-range

of biological activities, such as antibacterial, antioxidative,

anti-inflammatory, hypoglycemic activity [3, 5]. Another

example is the roots and rhizome of Salvia miltiorrhiza

Bunge, known as danshen, which is officially listed in the

Chinese Pharmacopeia to be used for the treatment of

menstrual disorders, menorrhalgia, insomnia, menostasis,

A. Olarte � N. Mantri (&) � G. Nugent � E. Pang

Health Innovations Research Institute, School of Applied

Sciences, RMIT University, Melbourne, VIC 3000, Australia

e-mail: nitin.mantri@rmit.edu.au

A. Olarte

e-mail: alolarte27@yahoo.com

G. Nugent

e-mail: gregory.nugent@rmit.edu.au

E. Pang

e-mail: eddie.pang@rmit.edu.au

H. Wohlmuth

Southern Cross Plant Science and Medicinal Plant Herbarium,

Southern Cross University, Lismore, NSW 2480, Australia

e-mail: hans.wohlmuth@scu.edu.au

C. G. Li � C. Xue

Health Innovations Research Institute, School of Health

Sciences, RMIT University, Melbourne, VIC 3000, Australia

e-mail: chun.guang.li@rmit.edu.au

C. Xue

e-mail: charlie.xue@rmit.edu.au

123

Mol Biotechnol (2013) 54:770–783

DOI 10.1007/s12033-012-9625-5

blood circulation diseases, and other cardiovascular dis-

eases [6].

One of the fundamental requirements for the commer-

cialization of Salvia is the accurate identification of com-

mercially important species from its close relatives.

Traditionally, the correct species identification was per-

formed by morphological or chemical techniques [6–8].

Nowadays, molecular biology techniques have given rise to

new possibilities for identification based on the genetic

composition of the plant which is unaffected by environ-

mental, developmental, and biological factors [9, 10].

However, DNA-based analyses of important medicinal

herbs such as Salvia has several challenges since they are

not model plants, their genome sizes are usually unknown

and there is a lack of molecular-marker-based approaches

for plant improvement [11]. PCR-based methods such as

random amplified polymorphic DNA (RAPD) and ampli-

fied fragment length polymorphism (AFLP) have been

commonly used to fingerprint different species of Salvia

[2, 4, 12–14]. However, the PCR-based methods are based

on gel electrophoresis, which is time consuming, and cor-

relating bands on gels for the allelic variants is difficult and

can lead to inaccurate interpretations [15]. Sequencing-

based methods have also been used for species identi-

fication of these genera. For instance, chloroplast and

mitochondrial DNA regions have been amplified and

restriction digested (polymerase chain reaction–restriction

fragment length polymorphism; PCR–RFLP) to genotype

Salvia species endemic to the Mediterranean region [16].

Furthermore, sequences of the internal transcribed spacer

(ITS) region have been amplified, sequenced and aligned in

order to genotype different Salvia [17, 18]. However,

chloroplast and ITS regions have not always been found to

be polymorphic among closely related Salvia species. For

example, high sequence similarities were found among

these DNA regions of some Japanese Salvia, which made it

difficult to determine evolutionary relationships among

them [19]. In addition, it is very unlikely that any chloro-

plast loci could be linked to a gene responsible for an

important agronomic trait, since these genes are usually

found in nuclear genome. Therefore, there is a need for

genetic markers in Salvia that could be useful not only for

the accurate identification of closely related species but that

could also be linked to desirable agricultural traits.

DNA microarrays can overcome many of the limitations

of the PCR-based fingerprinting methods. For instance,

they provide fixed data features, removing the positional

variation inherent in gel fingerprints. In addition, this

technique is based on hybridization, ensuring that common

elements are identical and genetically informative [20]. A

technology called diversity array technology (DArT) uses

microarrays to detect DNA polymorphism at several hun-

dred genomic loci with the advantage that no prior

knowledge of sequence information is needed [15]. The

main disadvantage of DArT is the method of genome

complexity reduction used, since a large number of

homologous/monomorphic sequences remain present in the

array, which will affect the level of polymorphism detected

[21].

In 2007, we reported a novel DNA-fingerprinting tech-

nique called subtracted diversity array (SDA) [22]. This

technique uses an alternative subtraction suppression

hybridization (SSH) that removes the highly conserved

genomic DNA from the DNA representations that are used

to develop the array. Previously, an SDA was constructed

using gDNA representations from angiosperms and non-

angiosperms to isolate sequences specific to flowering

plants. The array successfully genotyped different flower-

ing plants to the clade, order and family level [23]. In order

to increase the discriminatory power of the angiosperm

SDA, an Asterid-specific SDA was also constructed. This

SDA was capable of successfully fingerprinting 25 Asterid

species representing 20 families and 9 orders [24]. The

broad subtraction approach (between clades) conducted to

produce the Asterid-specific SDA was able to isolate highly

polymorphic regions which were found to be novel species-

specific loci and chloroplast regions that have been

extensively used in phylogenetic analyses. However, its

discriminatory power was not enough to differentiate

among species of the same genera (data not shown). In this

article, we report the construction of the first genera-

specific SDA for Salvia species employing a narrow sub-

traction approach and demonstrate its fingerprinting

potential. In addition, we report the identification of highly

discriminatory and species-specific sequences that may

possibly be developed as molecular markers.

Materials and Methods

Plant Material

In order to obtain a genomic DNA (gDNA) representation

for the subtraction, a total of 151 species including

angiosperms and non-angiosperms were sourced (Table 1).

Non-angiosperms were collected from Toolangi State Park,

Victoria (Australia) and identified [25]. Angiosperms were

obtained only from verified nursery species; a total of 126

species were sourced to represent all angiosperm clades. In

addition, a total of ten Salvia species (42 plants) were

sourced to represent the different centres of diversity

around the world. Salvia miltiorrhiza and S. sinica Migo

plants were obtained from seeds which were used in a

previous study [26]. The other Salvia species were obtained

from verified specimens from various plant nurseries

(Table 1).

Mol Biotechnol (2013) 54:770–783 771

123

Table 1 Description of the angiosperm and nonangiosperm species used for DNA extraction and development of genome representations for

preparing the Salvia-specific SDA

Representations Species

Nonangiosperms (25 species) Adiantum raddianum Dicksonia antarctica Riccardia eriocaula

Azolla sp. Equisetum hyemale Selaginella sp.

Blechnum chambersii Ginkgo biloba Sphagnum australe

Blechnum fluviatile Grammitis billardieri Sticherus tener

Bryum billardieri Hymenophyton flabellatum Thuidium sp.

Catagonium nitens Marchantia sp. Weymouthia cochlearifolia

Cyathea cooperi Microsorum pustulatum Wollemia nobilis

Cyathophorum sp. Polystichum proliferum

Dawsonia superba Racopilum cuspidigerumvar.convolutaceum

Magnoliids (6 species) Cinnamomum verum Illicium anisatum Nymphaea gigantea

Houttuynia cordata Magnolia denudate Peumus boldus

Monocots (23 species) Acorus calamus Dioscorea polystacha Polygonatum multiflorum

Acorus gramineus Fritillaria thunbergii Ruscus aculeatus

Aloe vera Iris domestica (syn. Belamcanda chinensis) Serenoa repens

Bambusa beecheyana Iris versicolor Trachycarpus fortunei

Bletilla striata Lilium longiflorum Zea mays

Coix lacryma-jobi Lomandra longifolia Zephyranthes sp.

Colocasia esculenta Ophiopogon japonicus Zingiber officinale

Curcuma longa Pinellia cordata

Eudicots not placed in either the Rosids

or Asterids subclades (19 species)

Aconitum carmichaelii Chelidonium majus Eschscholzia californica

Aquilegia sp. Clematis hexapetala Grevillea robusta

Berberis fortunei Clematis montana Gypsophila oldhamiana

(syn. Mahonia fortunei) Clematis serratifolia Hamamelis virginiana

Berberis japonica Clematis songarica Phytolacca acinosa

(syn. Mahonia japonica) Dianthus caryophyllus Ranunculus sp.

Buxus sempervirens Dianthus superbus Rumex crispus

Rosids (33 species) Abutilon theophrasti Citrus aurantium Isatis tinctoria

Agrimonia eupatoria Citrus reticulata Oenothera biennis

Agrimonia pilosa Crataegus monogyna Oenothera odorata

Albizia julibrissin Dichroa febrifuga Oxalis pes-caprae

Alchemilla xanthochlora Filipendula ulmaria Passiflora edulis

Althaea officinalis Firmiana simplex Pelargonium sp.

Armoracia rusticana Glycyrrhiza glabra Poncirus trifoliata

Astragalus membranaceus Glycyrrhiza uralensis Rosa rugosa

Baptisia tinctoria Gynostemma pentaphyllum Ruta graveolens

Allocasuarina sp. Humulus lupulus Sophora flavescens

Catha edulis Hypericum perforatum Urtica dioica

Asterids (45 species) excluding

Lamiaceae

Achillea millefolium Codonopsis thalictrifolia Sambucus nigra

Adenophora potaninii Codonopsis pilosula Scrophularia ningpoensis

Angelica archangelica Coffea arabica Scrophularia nodosa

Angelica dahurica Cynara scolymus Solidago canadensis

Aralia chinensis Digitalis purpurea Symphytum officinale

Artemisia abrotanum Eupatorium perfoliatum Syringa vulgaris

Artemisia absinthium Ilex paraguariensis Tanacetum cinerariifolium

Artemisia lactiflora Impatiens sp. Tanacetum parthenium

Artemisia pontica Inula helenium Taraxacum officinale

Atropa belladonna Lycium barbarum Tetrapanax papyrifer

772 Mol Biotechnol (2013) 54:770–783

123

DNA Extraction and Development of Pools

Total DNA was extracted from fresh leaves using a mod-

ification of the standard CTAB procedure [27]. Approxi-

mately 0.5 g of leaves was ground with liquid nitrogen to a

fine powder. The powder was dissolved in 5 ml of CTAB

Buffer (3 % CTAB, 100 mM Tris–HCL pH 8.0, 20 mM

EDTA pH 8.0, 1.4 M NaCl), 1 ml of 10 % PVP and 1.2 ml

of 10 % CTAB. Subsequently, the mixture was incubated

at 60 8C for 1 h and centrifuged at 13,000 rpm for 10 min.

The supernatant obtained was further purified with a dou-

ble chloroform extraction followed by precipitation with

7.5 M ammonium acetate and 100 % ethanol. The pre-

cipitated DNA was resuspended in sterile water and sub-

sequent cleanup was performed by using the DNeasy�

column of the DNeasy� Plant Mini Kit (Qiagen) following

the protocol in the user manual.

All DNA samples were pooled based on the Angiosperm

Phylogeny Group (2009) classification [28] in order to obtain

representations of the following seven groups: Salvia (Tester

pool), non-angiosperms, Monocots, Magnoliids, Rosids,

Asterids (excluding Lamiaceae), and Eudicots not placed

in either Rosids or Asterids subclades (Eudicots and Core

Eudicots) (Table 1). About 10 lg of DNA was bulked for

each representation, with each pool having equal amounts of

gDNA per species. Subsequently, each pool was separately

concentrated using the DNeasy� column of the DNeasy�

Plant Mini Kit (Qiagen). The concentration and purity of the

DNA pools were evaluated spectrophotometrically.

Subtraction, Library, and SDA Construction

The method used for subtraction, library and microarray

construction was prepared using our previously described

method [22].

Subtraction was performed using the PCR-SelectTM

cDNA Subtraction Kit (Clontech). The Salvia pool (tester)

was represented by equal amounts of DNA extracted from

42 different plants of which 25 were Salvia miltiorrhiza

(five plants from five different lines). The driver pool was

obtained by bulking 700 ng of each representation with the

exception of the Salvia pool (Table 1).

The Salvia-specific DNA obtained after the subtraction

was purified using the QIAquick PCR Purification Kit

(Qiagen). Then, approximately 100 ng of the purified PCR

products were ligated into the pGEM�-T Easy vector

(Promega) and transformed into heat-shock competent

Escherichia coli JM109 (Promega) according to the user

manual. Positive transformation was determined by PCR

amplification of the cloned insert using nested primers 1

and 2R (Clontech). Plasmids containing cloned inserts

Table 1 continued

Representations Species

Asterids Bacopa monnieri Petroselinum crispum Trachelospermum jasminoides

Camellia sinensis Physalis peruviana Tussilago farfara

Centella asiatica Plantago major Valeriana officinalis

Chamaemelum nobile Platycodon grandiflorus ‘Apoyama’ Verbascum thapsus

Clerodendrum trichotomum Platycodon grandiflorus Withania somnifera

Salvia (13 species) Subtraction pool (total of 42

plants)

S. sinicag(five plants from Zhejiang

province)

Not included in the SDA

development

S. przewalskiig S. lanceolataf

S. officinalisa (five plants) S. microphyllac

S. lyratab S. miltiorrhizag S. fruticosad,a,e

S. elegansc (5 plants from 5 different populations)

S. sclaread,a,e Shandong province

S. mexicanac Shanxi province

S. runcinataf Henan province

S. lavandulifoliaa Hebei province

S. miltiorrhiza f. alba (Shandong)

a Native to the Mediterraneanb Native to North Americac Native to Central Americad Native to North Africae Native to Central and western Asiaf Native to South Africag Native to China

Mol Biotechnol (2013) 54:770–783 773

123

which showed a single band were isolated from subcultured

transformed cells using the DirectPrep 96 Miniprep Kit

(Qiagen).

The cloned inserts were subsequently PCR amplified

from the corresponding plasmid using the nested primers.

After that, PCR products were precipitated in 96 % ethanol

and 3 M sodium acetate (pH 5.2). The precipitation was

carried out at -20 8C overnight. The pellets obtained were

washed with 70 % ethanol, air dried and resuspended in

10 ll of 50 % DMSO. Finally, a total of 285 clones were

transferred individually into a 384-well plate (Genetix,

Hampshire, UK).

The 285 clones together with the controls were printed on

aminosilane-coated slides using a BioRobotics� MicroGrid

II Compact arrayer (Genomic Solutions) at RMIT Univer-

sity, Australia. Eight subarrays were gridded on a Corning�

GAPS II coated slide (Corning Incorporated Life Sciences,

Acton, MA). Each subarray was composed of 285 samples

and 15 controls. Among the positive controls were three

housekeeping genes (ribulose-1,5-bisphosphate carboxyl-

ase/oxygenase, ribosomal RNA and chlorophyll a/b-binding

protein) sourced from Cicer arietinum [29]. A single printed

slide was used to perform two hybridization experiments,

where each hybridization reaction was tested with 4

subarrays.

Labeling of Target DNA and SDA Hybridization

Target synthesis and hybridization of the microarray slides

were performed by modifying our previously described

method [22]. Only 30 ng of biotin-labeled sample was

mixed with 17.5 ll of 2 9 hybridization buffer (5 9 SSC,

0.2 % SDS, 50 % formamide); 0.5 ll of 1 mg/ml Human

Cot1 DNA (Invitrogen); 0.5 ll of 5 mg/ml PolyA (Sigma-

Aldrich) and 0.5 ll of 10 mg/ml salmon sperm DNA

(Sigma-Aldrich). After that, the mixture (made up to 35 ll

with sterile water) was denatured at 100 8C for 2 min and

applied to the array under a 22 9 22 mm lifter slip (Grale

Scientific, Australia). Hybridization was performed over-

night in a water bath at 42 8C. All hybridizations were

performed with four technical replicates (subarrays) and

two biological replicates, for a total of eight data points per

array feature.

After hybridization, the coverslips were removed and

the slides were washed twice in 19 SSC, 0.1 % SDS at

40 8C for 8 min, once in 0.19 SSC, 0.1 % SDS at 40 8Cfor 8 min and once in 0.19 SSC at room temperature for

5 min. Then, detection of the biotinylated DNA targets

bound on the array was performed by a protocol modified

from Mirus Label IT� lArray� Biotin Labeling Kit [24].

In brief, the slides were washed once in 69 SSPE-T buffer

(0.9 M NaCl, 0.06 M NaH2PO4�H2O, 0.006 M EDTA,

0.005 % Triton X-100, pH 7.4) at room temperature for

5 min. Subsequently, the detection solution (0.8 ll of

25 lg/ul of BSA, 0.5 ll of 0.8 lg/ll streptavidin-labeled

CyTM3 dye (Amersham Pharmacia, UK), made to 200 ll

with 69 SSPE-T) was applied to the wet surface of the

slide and covered by a 25 9 60 mm lifter coverslip (Grale

Scientific) to evenly distribute the solution. The slides were

incubated at 37 8C for 40 min in a waterproof hybridiza-

tion chamber in the dark. Finally, the slides were washed

three times in 69 SSPE-T buffer for 5 min, rinsed with

deionized water and dried with an air gun.

Data and Statistical Analysis

Slides were scanned at 10 lm resolution and gain of 50

PMT using a Perkin Elmer array scanner. Images were

captured and quantified with the ScanArray Express�

Microarray Analysis System. The program quantified the

signal intensity of each spot using adaptive circle method

and LOWESS function. Local background was subtracted

during quantitation. The signal-to-noise ratio obtained for

each spot by ScanArray Express� was defined as:

Signal-to-noise ratio

¼ Mean foreground�Mean background

Standard deviation of background

This signal value was considered to have the most accurate

background correction since it also accounted for varia-

tions in background intensity over the array. Quantified

data was exported to Microsoft Excel (Microsoft) and

abnormal spots that were not automatically flagged by the

software were flagged manually.

The following series of data transformations and

subsequent statistical analyses that were performed with

the raw signal intensity (signal-to-noise ratio) are sum-

marized in Fig. 1. In brief, a mean of the signal intensity

was obtained for each feature between the four technical

replicates. The data was then normalized across the

slides using the mean signal intensity for all 285 features

in all hybridizations performed. After normalization, the

data of the biological replicates was combined to pro-

duce a single value per feature. This data was transferred

to PASW Statistics 18 to perform a hierarchical cluster

analysis of the 15 Salvia genotypes. The dissimilarity

dendrogram was generated using the average distance

linkage between-groups and Square Euclidian metrics.

In addition, the normalized mean data was used for

principal component analyses (PCA) and correlate bivar-

iate analysis using MINITAB� Release 14.1 and PASW

Statistics 18, respectively. These analyses were able to

distinguish the most discriminatory features useful for

fingerprinting.

774 Mol Biotechnol (2013) 54:770–783

123

Sequencing of Selected most Discriminatory

and Species-Specific Features

The cloned inserts were re-amplified from the corre-

sponding isolated plasmid using SP6/T7 primers. Ampli-

fication products were bi-directionally sequenced by

Macrogen Inc. (Korea). Vector and primer sequences were

removed and nucleic acid and protein homology searches

were performed using blastN and blastX programs through

the National Center of Biotechnology Information [30].

All sequences have been deposited in EMBL Nucleo-

tide Sequence Database (Accession number HE96289 to

HE962398).

Results and Discussion

The first step to generate the Salvia-SDA was to use a

suppression subtractive hybridization (SSH) between a

gDNA pool of ten Salvia species (tester) and a pool of 151

non-Lamiaceae angiosperm and non-angiosperm species

(driver) to eliminate DNA fragments common to these two

pools (Fig. 2). The species belonging to the Lamiaceae

were not included in the driver pool since this would have

produced an over-subtracted library resulting in fewer

polymorphic sequences between the Salvia. Therefore, a

very conservative approach was employed for the sub-

traction. During SSH, the tester (Salvia) DNA was divided

in two parts and different adaptors were ligated to each part

followed by two hybridization steps between the tester and

the driver. In the first hybridization, the homologous

sequences between the driver and the tester were hybrid-

ized, leaving almost all of the Salvia-specific DNA single

stranded. Subsequently, in the second hybridization,

Salvia-specific DNA should have been the only double

stranded DNA with a different adaptor sequence allowing

exponential amplification of these sequences by PCR.

These potential Salvia-specific sequences were then iso-

lated by cloning, amplified and used to construct the

Salvia-specific SDA. Once constructed, the Salvia-specific

SDA was validated and tested for its ability to successfully

differentiate (fingerprint) different Salvia species. Each

fingerprint was obtained by hybridizing a biotin-labeled

genomic DNA of each Salvia species to the SDA (Fig. 3).

Finally, the hybridization patterns (fingerprints) were sub-

sequently compared to identify the most discriminatory

features and species-specific features that may be devel-

oped as potential molecular markers.

Subtraction Efficiency and Validation of the Microarray

The Salvia-specific microarray was first validated by

determining the subtraction efficiency, i.e., if the con-

structed array contained only Salvia-specific sequences. In

order to determine this efficiency, the tester and driver pool

were used as targets and separately hybridized onto the

array. Theoretically, a complete subtraction should result in

the absence of any hybridization with the driver pool as all

driver sequences are supposed to be eliminated. However,

thirty-four (12 %) positive features were observed after

hybridizing the driver target, indicating 88 % efficiency.

This subtraction efficiency was lower than that obtained for

the angiosperm [22] and the Asterids [24] SDAs where 97

and 99 % of efficiency was achieved, respectively.

During this subtraction, we used 30-fold excess of driver

(1:30 tester:driver ratio) as recommended by the kit man-

ufacturer (Clontech, CA). This same ratio was used for the

rather broad subtractions performed to obtain the angio-

sperm and Asterids SDAs which effectively eliminated the

common sequences between tester and driver. However,

the Salvia subtraction was more stringent than that of

previous work since the subtraction needed to be per-

formed to the genus level. It may be possible that a higher

concentration of driver DNA may be needed to remove

highly and partially homologous sequences from the tester

Fig. 1 Flow-chart of data analyses. The input of these analyses was

the signal-to-noise ratio obtained after scanning and quantitation of

spot intensities

Mol Biotechnol (2013) 54:770–783 775

123

pools. Mathematical models for suppression subtractive

hybridization have shown that by increasing the excess

ratios, the likelihood for enriching for highly specific tester

sequences is increased [31]. Therefore, it is possible that

higher subtraction efficiencies are achieved using higher

concentration of driver. However, the results of 88 %

efficiency are satisfactory for a subtraction with this high

level of stringency where genera-specific sequences are

isolated.

Microarray Scoring

A new scoring method was developed using the raw data of

the signal intensity since the binomial data scoring repre-

sented clear limitations in the analysis. Initially, the scoring

of the array was performed by converting the signal-

to-noise ratio of each feature to binomial using the same

analysis as our previously described method [22]. However,

we discovered that several of the features were giving signal

for all Salvia species though with different intensities. This

made us realize that since the species we assayed belonged

to the same genera; they shared highly similar gDNA

sequences. These sequences from closely related species

would potentially differ from each other by less than 10 %

thus generating variable signal. Therefore, due to the limi-

tations represented and possible misleading effects of the

binomial scoring, all the subsequent analyses were per-

formed with the raw signal intensity.

Previous array-based studies have also found that

binomial scoring may present difficulties in the analysis of

out-breeding species. For instance, a study based on DArT

found that the bi-allelic assessment limited the range of

possible analysis in 17 Eucalyptus individuals [32]. Fur-

thermore, another DArT study on Arabidopsis thaliana

successfully identified segregating markers in a F2 popu-

lation based on hybridization signal intensity differences

Fig. 2 The process of

construction of the Salvia SDA.

The genomic representations of

tester and driver were used to

perform a suppression

subtractive hybridization (SSH)

in order to eliminate the DNA

fragments common to these two

pools. Salvia-specific sequences

were then isolated by cloning,

amplified and used to construct

the Salvia-specific SDA

776 Mol Biotechnol (2013) 54:770–783

123

[33]. One common feature between the current study and

the previous Eucalyptus DArT study is that the species

analyzed are mainly cross-pollinated [34, 35]. Therefore,

the cross-pollination in these species could have increased

the level of heterozygosity, which will generate interme-

diate signal intensities that could have complicated the

analysis of binomial data. Consequently, the direct com-

parison between signal intensities is the preferred option

for a cross-pollinated genus such as Salvia, since the

assessment based on the presence and absence (dominant)

could have potentially misleading effects due to the higher

heterozygosity expected in this genus.

Evaluation of the Fingerprinting Potential of the SDA

Fingerprints for fifteen Salvia genotypes were obtained,

which included fingerprints for thirteen species and two

accessions of Salvia officinalis. Out of these thirteen, three

fingerprints were of S. lanceolata Lam., S. microphylla

Kunth, and S. fruticosa Mill., which were species not used

in the construction of the SDA (Table 1). The entire data

set from these experiments has been deposited in Gene

Expression Omnibus (GSE39403).

The relationship between the Salvia genotypes was

examined by a hierarchical cluster analysis constructed

based on the 285 features (Fig. 4). This dendrogram

grouped the 15 genotypes into three distinctive clusters.

Cluster A contained species native to the Mediterranean

region. Cluster B grouped all species native to Africa or the

Americas together with S. sclarea L. and S. fruticosa,

which are native to Europe, North Africa and Asia, and

Cluster C grouped all Salvia species from China.

The hierarchical cluster analysis clearly differentiated

important commercial Salvia species and accessions within

species. For instance, it was possible to differentiate Salvia

officinalis from its closely-related species S. fruticosa. These

Fig. 3 Process of biotin

labeling and hybridization. Each

fingerprint was obtained by

hybridizing a biotin-labeled

genome representation of each

Salvia species to the SDA.

Then, the biotinylated targets

bound to the array were labeled

with streptavidin-CyTM3 dye

Mol Biotechnol (2013) 54:770–783 777

123

two species have a high morphological similarity [7], which

has led to adulteration in commercial products. For example,

most of the commercial dried sage imported into North

America not only consists of S. officinalis but it is often

mixed with S. fruticosa since S. officinalis has a slow growth

rate in winter months [36]. In the cluster analysis, the three

accessions of S. officinalis were found to be closely related to

S. lavandulifolia Vahl, in particular to S. officinalis var.

purple, while S. fruticosa was located in a separate cluster

(Fig. 4). These results are in agreement with a recent taxo-

nomical classification [7], where S. lavandulifolia was found

to be a subspecies of S. officinalis and S. fruticosa was clearly

differentiated from these two taxa. Furthermore, the SDA

could also effectively differentiate S. miltiorrhiza from its

close relatives, S. przewalskii Maxim. and S. sinica. Roots of

S. przewalskii and S. sinica are often sold as Danshen (roots

of S. miltiorrhiza), even though chemical fingerprinting by

high performance liquid chromatography (HPLC) have

found that the levels of total phenolic acids and diterpenoids

in the roots of S. miltiorrhiza are generally higher [37, 38].

The above results indicate that this Salvia-specific SDA

could be a useful tool for fingerprinting closely related spe-

cies and accessions that have been used in the construction of

the array.

The results also demonstrate the ability of the SDA to

fingerprint species that were not used in its construction.

For instance, it was possible to fingerprint S. lanceolata,

S. microphylla, and S. fruticosa, which grouped with other

Salvia species of similar geographical origin in the hier-

archical cluster analysis (Fig. 4), even though they were

not part of the original subtraction pool. Similar results

were found for the angiosperm and Asterid-specific SDAs

which were able to fingerprint species outside the initial

subtraction pool [23, 24]. Therefore, this SDA may have a

wider applicability in fingerprinting other species of Salvia

apart from the ones used to construct the array, with the

advantage of not requiring prior sequence information.

Further, the hierarchical analysis revealed genetic rela-

tionships consistent with geographical origins (Fig. 4). The

dendrogram shows three major clusters. The first cluster

included the European Salvia, the second cluster had the

native American and African species together with S. sclarea

and S. fruticosa, and the third clustered the Chinese Salvia.

However, these results differ from previous genotyping

studies. For instance, recent analyses of 51 worldwide

collected accessions of Salvia with AFLP have grouped the

genotypes into two main clusters [14]. The first one

included all species form Central and South America and

the second one contained species form Europe, Africa and

Asia. This clear differentiation between African and

American Salvia was not found in the present study

(Fig. 4). Interestingly, the AFLP and the SDA-based clas-

sification also presented differences with previous phylo-

genetic studies performed in the tribe Mentheae and in the

genus Salvia using the sequences of the amplified nuclear

rDNA ITS and chloroplast DNA regions [1, 39]. It is

important to note that the main aim of this study was to

develop a microarray for fingerprinting Salvia species and

not to perform a phylogenetic analysis of this genus.

During the construction of the SDA, the Salvia pool was

enriched with S. miltiorrhiza, S. sinica, and S. officinalis.

Therefore, the SDA could be overrepresented with

Fig. 4 Dissimilarity

dendrogram for the SDA

hybridization patterns of fifteen

Salvia genotypes using the 285

features (Squared Euclidian

distance, between groups

linkage). The steps of the

dendrogram show the combined

clusters and the values of the

distance coefficients at each

step; the values have been

rescaled to numbers between 0and 25, preserving the ratio of

the distances between the steps

778 Mol Biotechnol (2013) 54:770–783

123

sequences from these three species, and as a result the

phylogenetic analyses obtained from this array could be

biased on the distances given across the species and major

clusters. Previous SDA studies have shown the utility of

the SDA for inferring genetic relationships consistent with

the Angiosperm Phylogeny Group (2009) classification

[28]. For instance, the angiosperms-specific SDA has

shown to be useful in classifying different families with the

respective clade and species within their correspondent

families [23]. However, in order to apply this technique for

phylogenetic analyses in Salvia, a more comprehensive

array would have to be constructed with equal genomic

representations from all Salvia subgenera. In addition, a

wider range of species would have to be genotyped in order

to obtain a more detailed phylogenetic and evolutionary

analysis of the genus.

Identification of most Discriminatory

and Species-Specific Features

Identification of the most discriminatory and species-

specific features was performed by a series of statistical

analyses (Fig. 1). Firstly, principal components analysis

(PCA) was performed in order to identify the most dis-

criminatory DNA fragments capable of generating a fully

resolved phylogeny of the genotypes analyzed. As it can be

seen in Fig. 5, a high percentage of variation (80.5 %) may

be explained by the first two components. Based on this

analysis, only the four most distant features from zero on

the X axis were chosen as the first component explained

most of the variation (71.1 %). Upon examination of these

four features (A16, I7, N6, P4) it was observed that they

were the only features that had a high variance and a high

mean across the fingerprints; however, none of them

showed any species specificity.

In order to find species-specific features, a second

analysis based on the magnitude of the variance for the

signal intensity of each feature across the 15 genotypes was

performed. Ten species-specific features (E13, F5, G4,

G13, H17, J9, N7, N12, N13, O1) were identified which

were not previously detected by PCA. However, more

than one feature showed specificity to the same species,

implying there were features with the same patterns of

variation. Pearson bivariate correlation performed among

the 14 features (four chosen by PCA and ten species-

specific), indicated that there were positive significant

correlations between H17 and J9 (r = 0.98, P \ 0.01),

H17 and G4 (r = 0.99, P \ 0.01), G13 and N7 (r = 0.99,

P \ 0.01) and between N6 and I7 (r = 0.83, P \ 0.01).

Therefore, J9, G4, N7, and I7 were eliminated and only ten

features (Table 2) were selected as they were the most

discriminatory and each had unique patterns of variation

among the 15 genotypes.

Finally, a second hierarchical cluster was performed using

only the ten features selected above (three chosen by PCA and

seven species-specific) (Fig. 6). It was found that the clus-

tering of the species was consistent with the major clusters

obtained using the full set (Fig. 4). However, five of the spe-

cies in Cluster B were displaced relative to the original den-

drogram. In the second dendrogram, S. microphylla clustered

with S. lanceolata and S. lyrata L., whilst S. mexicana L. and

S. runcinata L.f. appeared to be more distantly related. In

contrast, in the original dendrogram, S. microphylla and

S. mexicana clustered together, whilst S. runcinata,

S. lanceolata and S. lyrata were closely related (Fig. 4). Based

on the above results, it was inferred that these ten features were

among the most useful features for fingerprinting these fifteen

Salvia genotypes. Among this subset, species-specific features

were found for S. officinalis, S. sclarea, S. przewalskii,

S. elegans Vahl, and S. runcinata (Table 2; Fig. 6). Also

specific features which discriminated between species were

found, for instance feature N12 was specific for S. sinica,

S. miltiorrhiza, and S. przewalskii and H17 can differentiate

S. officinalis and S. lavandulifolia from the other species.

Sequence Identity of the most Discriminatory

and Species-Specific Features

The species-specific features together with three features

chosen by PCA were sequenced and analyzed using blastN

and blastX [30] (Table 2). Four out of the ten features

matched with homologous sequences in GenBank. Features

E13 [EMBL: HE962398] and P4 [EMBL: HE962389]

corresponded to known chloroplast loci. Feature N12

[EMBL: HE962396] matched to an uncharacterized geno-

mic DNA, F5 [EMBL: HE962397] corresponded to an

uncharacterized gene and the other six features were not

Fig. 5 Principal component analysis plot for the 285 features. The

first principal component accounts for 71.1 % of variation and the

second component explained only 9.4 % of variation. The squaresrepresent features that account for most of the variability found across

the genotypes

Mol Biotechnol (2013) 54:770–783 779

123

Table 2 Predicted locus/function of the ten sequenced SDA features using blastN program [30]

Feature

ID

Length

(bp)

Specific to target Matching

database

entry

Putative identity E value

A16a 556 – No hits NA

E13b 372 Differentiate

accessions of S.officinalis

DQ673256.1 Forsythia europaea psaA-psbB fragment, chloroplast. 7e-151

GQ996975.1 Antirrhinum majus cpl protease proteolytic subunit protein (cplP) gene,

complete cds, chloroplast

8e-150

F5b 218 S. przewalskii DY322087.1 Ocimum basilicum uncharacterized cDNA sequence 6e-14

G13b 145 S. elegans – No hits NA

H17b 612 S. officinalis – No hits NA

S. lavandulifolia

N12b 423 S. miltiorrhiza FJ148301.1 Daucus carota subsp. sativus uncharacterized sequence 2e-06

S. sinica

S. przewalskii

N13b 556 S. runcinata – No hits NA

N6a 250 – No hits NA

O1b 118 S. sclarea – No hits NA

P4a 526 DQ673256.1 Forsythia europaea psaA-psbB fragment, chloroplast. 2e-77

DQ983917.1 Jacobaea uniflora isolate SGO10 tRNA-Met (trnM) gene, partial

sequence; ATP synthase epsilon subunit (atpE) and ATP synthase beta

subunit (atpB) genes, complete sequence; and ribulose 1,5-bisphosphate

carboxylase/oxygenase large subunit (rbcL) gene, partial sequence;

chloroplast

2e-65

The best match is shown as the putative identity for each sequence. E value regarded as significant if \1e–10. NA indicates the absence of

significant dataa Features that were chosen by PCAb Features that are part of the seven species-specific features

Fig. 6 Dissimilarity

dendrogram for the 15

genotypes using only the ten

most discriminatory and

species-specific features

(Squared Euclidian distance,

between groups linkage). The

steps of the dendrogram show

the combined clusters and the

values of the distance

coefficients at each step; the

values have been rescaled to

numbers between 0 and 25,

preserving the ratio of the

distances between the steps. The

black shaded circles in the

dendrogram represent species-

specific features which

discriminated across the species

780 Mol Biotechnol (2013) 54:770–783

123

recognized as known DNA sequences or proteins [EMBL:

HE962390, HE962391, HE962392, HE962393, HE962394,

HE962395].

Features E13 [EMBL: HE962398] and P4 [EMBL:

HE962389], had a significant alignment with the same

psaA-psbB chloroplast region [GeneBank:DQ673256.1], as

shown in Table 2. However, these two polymorphic fea-

tures did not overlap each other since they aligned to dif-

ferent parts of the psaA-psbB region which is 32,412 bp

long. Interestingly, this same region has been found to be

polymorphic among members of the Oleaceae family

(Jasminum and Menodora), where a series of inversions

have been identified [40]. This high level of polymorphism

found in the psaA-psbB region among different members of

the Lamiales (Oleaceae and Lamiaceae family), could

make it a source of potential markers for the Lamiales

species. It is important to note that in this study this

sequence was isolated by the subtraction process, indicat-

ing that this technique is capable of identifying highly

polymorphic sequences that are highly specific for the taxa

under study.

The six unknown and the two uncharacterized sequences

are almost certainly part of nuclear DNA, since if they

were part of chloroplast or mitochondrial DNA they would

have matched to GenBank accessions (which has an

extensive amount of chloroplast and mitochondrial data-

base entries). The presence of these polymorphic nuclear

DNA features suggests that the SDA may be a more reli-

able approach for fingerprinting closely related species and

hybrids. The reason for this may be that the detection of

hybridization/introgression is not reliably accomplished by

examination of chloroplast DNA since it is uniparentally

inherited [41]. In addition, dependence on a single nuclear

locus could have misleading results since the hybrid could

be homozygous at many loci [42]. Selection of nuclear loci

have also been performed using sequence databases as a

framework; however, this approach is dependent on the

phylogenetic proximity of the taxa under study to a number

of DNA sequences being available in the databases [43].

Therefore, the narrow subtraction approach conducted to

produce the SDA is a good alternative to identify new

nuclear polymorphic loci for taxa that do not have closely

related species with available genomic libraries.

Furthermore, the potential nuclear sequences identified

in the array could be associated or linked to a gene

responsible for an important agronomic trait, since these

genes are usually found in the nuclear genome. Previous

studies have already found relationships between essential

oil content and genotyping profiles in S. officinalis [12],

S. fruticosa [13], and other Mediterranean Salvia [16].

However, as of date no significant relationship has been

found between important bioactive components and geno-

typing profiles in several other species. For instance,

previous studies on S. miltiorrhiza have been unable to

correlate molecular profiles obtained with ITS sequences

with HPLC profiles of important bioactive components

[17, 18]. Current genotyping studies on ten S. miltiorrhiza

populations using the SDA have found significant corre-

lation between the signal intensity of two SDA features and

previous chemical analyses on the same geographical

populations (data not shown). Preliminary results of this

work can be found in a published progress report available

online [44]. Therefore, it is expected that upcoming reports

would corroborate the importance of this technique to

identify potential molecular markers which could assist in

future breeding programs.

Advantages and Limitations of the SDA

The SDA technique has shown potential advantages over

other fingerprinting techniques.

(i) The narrow subtraction approach allowed us to

selectively enrich the SDA with Salvia-specific DNA

sequences, which at the same time allowed the

fingerprint of Salvia species that were not used to

construct the array. This indicates that a SDA can be

constructed using a representative pool of individuals

instead of all members of the group under study. The

Salvia-specific SDA was constructed with only ten

species and it has the potential to fingerprint other

Salvia species. This is especially advantageous for

fingerprinting a broad genus like Salvia, which has

approximately 1,000 species with wide morphological

and ecological variation.

(ii) The ability of the SDA to fingerprint based on

polymorphic regions from chloroplast and nuclear

DNA increases the possibility of differentiating closely

related species and hybrids. However, sequences from

ribosomal and chloroplast regions have been success-

fully used to genotype Salvia species [16, 45, 46]. In

addition, intergenic spacers between trnH-psbA and

trnL-trnF together with rbcL and the nuclear ITS

region have been used to conduct molecular phyloge-

netic analyses in the tribe Mentheae and in the genus

Salvia [1, 19, 39]. Although, it could be argued that

SDA does not offer a significant advantage over

sequencing-based methods to genotype Salvia, it is

important to note that it is very unlikely that any

chloroplast or ITS region could be linked to a gene

responsible for an important agronomic trait. Conse-

quently, SDA offers a significative advantage since the

polymorphic nuclear sequences identified in this study

could potentially be employed not only for genotyping

but also they could be associated with genes respon-

sible for important agronomical traits.

Mol Biotechnol (2013) 54:770–783 781

123

(iii) The microarray scoring based on the direct compar-

ison between signal intensity is more appropriate

than a binomial scoring to analyze a cross-pollinated

genus. Therefore, SDA offers and advantage over

dominant markers such as RAPD and AFLP analysis

which have been commonly used to genotype Salvia

species. Co-dominant microsatellites could also offer

a good option for genotyping out-breeding species.

For instance, SSR derived from S. miltiorrhiza EST

sequences were able to detect genetic diversity

among test samples of S. miltiorrhiza and distinguish

it from other Salvia plants [47]. However, it is

important to note that microsatellites need previous

sequence knowledge and are low-throughput.

(iv) The SDA has proven to be a powerful tool to identify

species-specific sequences that may be employed for

unequivocal identification of commercial species

such as S. officinalis and S. miltiorrhiza. However,

the application of this technique for authentication of

commercial species would involve an investment in

the equipment necessary to print, hybridize, and scan

the SDAs. Furthermore, it will require good quality

scans of several species that can serve as references,

followed by standardization of data and statistical

analysis. Therefore, it would not be advisable to use

the entire array to identify only one or a couple of

commercial species since it would be more time and

consuming expensive than PCR. Instead, specific

primers could be developed for the amplification of

the species-specific features previously identified.

Then, variations in the amplification products or

sequence polymorphisms identified between the

related species could be employed for the unequiv-

ocal identification of the species of interest. Once

polymorphic PCR-based markers are developed from

these sequences, a fast and easy protocol for routine

quality control could be develop in order to prevent

substitution and misidentification of commercial

species of Salvia.

Conclusions

To the best of authors’ knowledge, this is the first finger-

printing array constructed for Salvia species and the first

genus-specific SDA. Using this array it was possible to fin-

gerprint 15 Salvia genotypes and to construct a hierarchical

cluster which was found to be consistent with the geo-

graphical origin and was able to differentiate S. officinalis

and S. miltiorrhiza from their closely related species.

Based on the above results it is possible to conclude that

the narrow subtraction approach conducted to produce the

SDA is a good alternative to isolate highly variable DNA

sequences specific to a genus as large as Salvia without

preliminary sequence information. Therefore, SDA is a

powerful technique to fingerprint non-model plants where

little genomic sequence information and few markers are

available.

Acknowledgments The authors gratefully acknowledge the support

from the Rural Industries Research and Development Corporation,

RMIT University and the Australian Postgraduate Scholarship awar-

ded to Alexandra Olarte. We acknowledge the technical support from

A/Prof. Reg Lehmann from MediHerb Australia and Claudia Salazar

for their assistance with the graphics in this study.

References

1. Walker, J. B., Sytsma, K. J., Treutlein, J., & Wink, M. (2004).

Salvia (Lamiaceae) is not monophyletic: Implications for the

systematics, radiation, and ecological specializations of Salviaand tribe Mentheae. American Journal of Botany, 91, 1115–1125.

2. Bruna, S., Giovannini, A., Benedetti, L. D., Principato, M. C., &

Ruffoni, B. (2006). Molecular analysis of Salvia Spp. through

RAPD markers. ISHS Acta Horticulturae, 723, 157–160.

3. Topcu, G. (2006). Bioactive triterpenoids from Salvia species.

Journal of Natural Products, 69, 482–487.

4. Echeverrigaray, S., & Agostini, G. (2006). Genetic relationships

between commercial cultivars and Brazilian accessions of Salviaofficinalis L. based on RAPD markers. Revista Brasileira dePlantas Medicinais, 8, 13–17.

5. Longaray-Delamare, A. P., Moschen-Pistorello, I. T., Artico, L.,

Atti-Serafini, L., & Echeverrigaray, S. (2007). Antibacterial

activity of the essential oils of Salvia officinalis L. and Salviatriloba L, cultivated in south Brazil. Food Chemistry, 100,

603–608.

6. Cai, Z., Lee, F. S. C., Wang, X. R., & Yu, W. J. (2002). A capsule

review of recent studies on the application of mass spectrometry

in the analysis of Chinese medicinal herbs. Journal of MassSpectrometry, 37, 1013–1024.

7. Reales, A., Rivera, D., Palazon, J. A., & Obon, C. (2004).

Numerical taxonomy study of Salvia sect. Salvia (Labiatae).

Botanical Journal of the Linnean Society, 145, 353–371.

8. Zhong, G. X., Li, P., Zeng, L. J., Guan, J., Li, D. Q., & Li, S. P.

(2009). Chemical characteristics of Salvia miltiorrhiza (Danshen)

collected from different locations in China. Journal of Agricul-ture and Food Chemistry, 57, 6879–6887.

9. Kiran, U., Khan, S., Mirza, K. J., Ram, M., & Abdin, M. Z.

(2010). SCAR markers: A potential tool for authentication of

herbal drugs. Fitoterapia, 81, 969–976.

10. Joshi, K., Chavan, P., Warude, D., & Patwardhan, B. (2004).

Molecular markers in herbal drug technology. Current ScienceIndia, 87, 159–165.

11. Canter, P. H., Thomas, H., & Ernst, E. (2005). Bringing medic-

inal plants into cultivation: Opportunities and challenges for

biotechnology. Trends in Biotechnology, 23, 180–185.

12. Boszormenyi, A., Hethelyi, E., Farkas, A., Horvath, G., Papp, N.,

Lemberkovics, E., et al. (2009). Chemical and genetic relation-

ships among sage (Salvia officinalis L.) cultivars and Judean sage

(Salvia judaica Boiss.). Journal of Agriculture and Food Chem-istry, 57, 4663–4667.

13. Skoula, M., El Hilali, I., & Makris, A. M. (1999). Evaluation of

the genetic diversity of Salvia fruticosa Mill. clones using RAPD

markers and comparison with the essential oil profiles. Bio-chemical Systematics and Ecology, 27, 559–568.

782 Mol Biotechnol (2013) 54:770–783

123

14. Braglia, L., Casabianca, V., De Benedetti, L., Pecchioni, N.,

Mercuri, A., Cervelli, C., et al. (2011). Amplified fragment length

polymorphism markers for DNA fingerprinting in the genus

Salvia. Plant Biosystems, 145, 274–277.

15. Jaccoud, D., Peng, K., Feinstein, D., & Kilian, A. (2001). Diversity

arrays: A solid state technology for sequence information inde-

pendent genotyping. Nucleic Acids Research, 29, E25.

16. Karaca, M., Ince, A. G., Ay, S. T., Turgut, K., & Onus, A. N.

(2008). PCR-RFLP and DAMD-PCR genotyping for Salviaspecies. Journal of the Science of Food and Agriculture, 88,

2508–2516.

17. Xu, H., Wang, Z. T., Cheng, K. T., Wu, T., Gu, L. H., & Hu, Z. B.

(2009). Comparison of rDNA ITS sequences and tanshinones

between Salvia miltiorrhiza populations and Salvia species.

Botanical Studies, 50, 127–135.

18. Han, J. P., Liu, C., Li, M. H., Shi, L. C., Song, J. Y., Yao, H.,

et al. (2010). Relationship between DNA barcoding and chemical

classification of Salvia medicinal herbs. Chinese Herbal Medi-cines, 2, 16–29.

19. Takano, A., & Okada, H. (2010). Phylogenetic relationships

among subgenera, species, and varieties of Japanese Salvia L.

(Lamiaceae). Journal of Plant Research, 124, 245–252.

20. Kingsley, M. T., Straub, T. M., Call, D. R., Daly, D. S., Wun-

schel, S. C., & Chandler, D. P. (2002). Fingerprinting closely

related Xanthomonas pathovars with random nonamer oligonu-

cleotide microarrays. Applied and Environment Microbiology, 68,

6361–6370.

21. Li, T. -X., Wang, J. -K., Bai, Y. -F., & Lu, Z. -H. (2006).

Diversity suppression-subtractive hybridization array for profiling

genomic DNA polymorphisms. Journal of Integrative PlantBiology, 48, 460–467.

22. Jayasinghe, R., Kong, S., Coram, T. E., Kaganovitch, J., Xue, C.

C., Li, C. G., et al. (2007). Construction and validation of a

prototype microarray for efficient and high-throughput genotyp-

ing of angiosperms. Plant Biotechnology Journal, 5, 282–289.

23. Jayasinghe, R., Hai Niu, L., Coram, T. E., Kong, S., Kaganovitch,

J., Xue, C. C. L., et al. (2009). Effectiveness of an innovative

prototype subtracted diversity array (SDA) for fingerprinting plant

species of medicinal importance. Planta Medica, 75, 1180–1185.

24. Mantri, N., Olarte, A., Li, C. G., Xue, C., & Pang, E. C. K.

(2012). Fingerprinting the Asterid species using subtracted

diversity array reveals novel species-specific sequences. PLoSONE, 7, e34873.

25. Duncan, B. D., & Isaac, G. (1994). Ferns and allied plants of

Victoria, Tasmania and South Australia. Melbourne University

Press in association with Monash University, Carlton.

26. Li, C. G., Sheng, S. J., Pang, E. C., Marriot, P., May, B., Zhou, S.

F., et al. (2009). Cultivar variations of Australian-grown Danshen

(Salvia miltiorrhiza): Bioactive markers and root yields. Chem-istry & Biodiversity, 6, 170–181.

27. Doyle, J. J., & Doyle, J. L. (1987). A rapid DNA isolation pro-

cedure for small quantities of fresh leaf tissue. PhytochemicalBulletin, 19, 1–15.

28. Bremer, B., Bremer, Kr., Chase, M., Fay, M., Reveal, J., Soltis,

D., et al. (2009). An update of the Angiosperm Phylogeny Group

classification for the orders and families of flowering plants: APG

III. Botanical Journal of the Linnean Society, 161, 105–121.

29. Coram, T. E., & Pang, E. C. K. (2005). Isolation and analysis of

candidate ascochyta blight defence genes in chickpea. Part I. Gen-

eration and analysis of an expressed sequence tag (EST) library.

Physiological and Molecular Plant Pathology, 66, 192–200.

30. National Center of Biotechnology Information (www.ncbi.nlm.

nih.gov).

31. Gadgil, C., Rink, A., Beattie, C., & Hu, W. S. (2002). A math-

ematical model for suppression subtractive hybridization. Com-parative and Functional Genomics, 3, 405–422.

32. Lezar, S., Myburg, A. A., Berger, D. K., Wingfield, M. J., &

Wingfield, B. D. (2004). Development and assessment of

microarray-based DNA fingerprinting in Eucalyptus grandis.

Theoretical and Applied Genetics, 109, 1329–1336.

33. Wittenberg, A. H. J., van der Lee, T., Cayla, C., Kilian, A.,

Visser, R. G. F., & Schouten, H. J. (2005). Validation of the high-

throughput marker technology DArT using the model plant

Arabidopsis thaliana. Molecular Genetics and Genomics, 274,

30–39.

34. Claßen-Bockhoff, R. (2007). Floral construction and pollination

biology in the Lamiaceae. Annals of Botany, 100, 359–360.

35. Pound, L. M., Wallwork, M. A. B., Potts, B. M., & Sedgley, M.

(2002). Self-incompatibility in Eucalyptus globulus ssp. globulus(Myrtaceae). Australian Journal of Botany, 50, 365–372.

36. Dudai, N., Lewinsohn, E., Larkov, O., Katzir, I., Ravid, U.,

Chaimovitsh, D., et al. (1999). Dynamics of yield components

and essential oil production in a commercial hybrid sage (Salviaofficinalis x Salvia fruticosa cv. Newe Ya’ar no. 4). Journal ofAgriculture and Food Chemistry, 47, 4341–4345.

37. Cao, J., Wei, Y. J., Qi, L. W., Li, P., Qian, Z. M., Luo, H. W.,

et al. (2008). Determination of fifteen bioactive components in

Radix et Rhizoma Salviae Miltiorrhizae by high-performance

liquid chromatography with ultraviolet and mass spectrometric

detection. Biomedical Chromatography, 22, 164–172.

38. Li, M. H. (2008). Investigation of Danshen and related medicinal

plants in China. Journal of Ethnopharmacology, 120, 419–426.

39. Walker, J. B., & Sytsma, K. J. (2007). Staminal evolution in the

genus Salvia (Lamiaceae): Molecular phylogenetic evidence for

multiple origins of the staminal lever. Annals of Botany, 100,

375–391.

40. Lee, H. L., Jansen, R. K., Chumley, T. W., & Kim, K. J. (2007).

Gene relocations within chloroplast genomes of Jasminum and

Menodora (Oleaceae) are due to multiple, overlapping inversions.

Molecular Biology and Evolution, 24, 1161–1180.

41. Fazekas, A. J., Kesanakurti, P. R., Burgess, K. S., Percy, D. M.,

Graham, S. W., Barrett, S. C. H., et al. (2009). Are plant species

inherently harder to discriminate than animal species using DNA

barcoding markers? Molecular Ecology Resources, 9, 130–139.

42. Chase, M. W., Salamin, N., Wilkinson, M., Dunwell, J. M.,

Kesanakurthi, R. P., Haidar, N., et al. (2005). Land plants and

DNA barcodes: Short-term and long-term goals. PhilosophicalTransactions of the Royal Society of London, Series B: BiologicalSciences, 360, 1889–1895.

43. Alvarez, I., Costa, A., & Feliner, G. N. (2008). Selecting single-

copy nuclear genes for plant phylogenetics: A preliminary anal-

ysis for the Senecioneae (Asteraceae). Journal of MolecularEvolution, 66, 276–291.

44. Pang, E. (2011). Using frontier technologies for the qualityassurance of medicinal herbs (pp. 1–21). RIRDC publications.

RMIT University, Melbourne.

45. Bertea, C. M., Luciano, P., Bossi, S., Leoni, F., Baiocchi, C.,

Medana, C., et al. (2006). PCR and PCR-RFLP of the 5S-rRNA-

NTS region and salvinorin A analyses for the rapid and

unequivocal determination of Salvia divinorum. Phytochemistry,67, 371–378.

46. De Mattia, F., Bruni, I., Galimberti, A., Cattaneo, F., Casiraghi,

M., & Labra, M. (2011). A comparative study of different DNA

barcoding markers for the identification of some members of

Lamiaceae. Food Research International, 44, 693–702.

47. Deng, K. J., Zhang, Y., Xiong, B. Q., Peng, J. H., Zhang, T.,Zhao, X. N., et al. (2009). Identification, characterization and

utilization of simple sequence repeat markers derived from Salviamiltiorrhiza expressed sequence tags. Yao Xue Xue Bao [ActaPharmaceutica Sinica], 44, 1165–1172.

Mol Biotechnol (2013) 54:770–783 783

123