Rice Science, 2009, 16(2): 119­123
Copyright  2009, China National Rice Research Institute.
Published by Elsevier BV. All rights reserved
DOI: 10.1016/S1672-6308(08)60067-0
Metabolome Comparison of Transgenic and Non-transgenic Rice by
Statistical Analysis of FTIR and NMR Spectra
Keykhosrow KEYMANESH, Mohammad Hassan DARVISHI, Soroush SARDARI
(Biotechnology Department, Pasteur Institute of Iran, #69 Pasteur Ave., Tehran, Iran)
Abstract: Modern biotechnology, based on recombinant DNA techniques, has made it possible to introduce new traits
with great potential for crop improvement. However, concerns about unintended effects of gene transformation that
possibly threaten environment or consumer health have persuaded scientists to set up pre-release tests on genetically
modified organisms. Assessment of `substantial equivalence' concept that established by comparison of genetically
modified organism with a comparator with a history of safe use could be the first step of a comprehensive risk assessment.
Metabolite level is the richest in performance of changes which stem from genetic or environmental factors. Since
assessment of all metabolites in detail is very costly and practically impossible, statistical evaluation of processed data of
grain spectroscopic values could be a time and cost effective substitution for complex chemical analysis. To investigate
the ability of multivariate statistical techniques in comparison of metabolomes as well as testing a method for such
comparisons with available tools, a transgenic rice in combination with its traditionally bred parent were used as test
material, and the discriminant analysis were applied as supervised method and principal component analysis as
unsupervised classification method on the processed data which were extracted from Fourier transform infrared
spectroscopy and nuclear magnetic resonance spectral data of powdered rice and rice extraction and barley grain
samples, of which the latter was considered as control. The results confirmed the capability of statistics, even with initial
data processing applications in metabolome studies. Meanwhile, this study confirms that the supervised method results in
more distinctive results.
Key words: rice; principal component analysis; discriminant analysis; nuclear magnetic resonance; Fourier transform
infrared spectroscopy; transgene; safety assessment; metabolome analysis
The genetically modified (GM) plants are made
through introduction of new genes possibly from distant
phylogenic species. These plants have shown improved
performance in facing abiotic stress [1]
and pest and
pathogen attacks [2-3]
, or production of more qualified
or nutritionally valuable products [4]
. As the area under
cultivation of GM plants and their products' portion in
the market is increasing around the world, more
consumers are exposed to the outputs of recombinant
DNA techniques [5]
. This tendency has led to debates
about the possible unintended effects of new products
on environment as well as the consumer health. As a result,
scientific bodies suggest safety assessment procedures
through which some of them have acquired international
acceptance [6]
. The `substantial equivalence' concept
which is based on the comparison of a GM product
with a traditionally bred parent with a long record of
safe usage is usually the first step in safety supervision[7]
.
Fingerprinting approaches at metabolite level could be
very informative and reflect the changes which stem
from genetic manipulation, but investigation of all
chemical components of the target organism is
technically very difficult. On the other hand, statistical
analysis of spectra results from spectroscopy techniques
could be a time and cost effective and accurate way
for metabolome comparison studies [8]
.
There are many reports about metabolome
studies based on nuclear magnetic resonance (NMR)
and Fourier transform infrared spectroscopy (FTIR).
These studies cover a vast range from medical
investigations to safety-intended studies of GM
products [9-11]
. Reaching a convincing result by these
methods depends on efficient exploitation of data
from the spectra and then analysis of the data through
a plausible classification method. Generally, there are
two main classification methods. First, supervised
method, in which the predefined classes are the
starting point for the analysis process that leads to
models based on the input clusters, and the second one
Received: 23 September 2008; Accepted: 2 January 2009
Corresponding author: Soroush SARDARI (ssardari@hotmail.com;
sardari@pasteur.ac.ir)
120 Rice Science, Vol. 16, No. 2, 2009
is unsupervised method that in a multivariate manner
assesses how similar a set of samples are and the
samples with more similarities are included to a group.
Eventually, every statistical procedure for classification
belongs to one of these two major methods [12]
.
Usually, experiments with similar aims as ours are
done by use of a few specialized computer applications.
However, in this study, in addition to assessment of
the transgenic rice in comparison with its parental
variety, we intended to examine the power of less
specialized and more available softwares for such
experiments. We used transgenic rice in combination
with its traditionally bred parent as test materials and
exerted linear discriminant analysis as supervised and
principal component analysis as unsupervised classification
methods on the processed data extracted from NMR
and FTIR spectra of transgenic and non-transgenic
rice and barley grain samples which were used as
control.
MATERIALS AND METHODS
Plant materials
The seed samples were derived from the rice
variety Tarom mulaiee and its counterpart variety
which is transformed with cry1Ab gene from Bacillus
thuringinesis (Bt). The samples have been provided
from transgenic plants and their parental variety which
was grown similarly as control. For further processing
of collected data and spectral analysis, the different
samples taken from the same non-transgenic and
transgenic rice seeds samples numbered 1­10 and 11­
20, respectively. A barley sample of an unknown
variety was used merely as control. The treatment
procedures on this sample were like the rice samples.
DNA was extracted from rice samples and the
existence of the transgene was confirmed by use of
relevant primers and observation of concerned band
(Fig. 1). The DNA was extracted as described by
Wang et al [13]
, and subjected to one cycle of 94C for
5 min, 35 cycles of three steps each (94C, 1 min;
60C, 1 min; and 72C, 3 min) in 25 L of PCR buffer
(10 mmol/L Tris-HCl, pH 8.4, 50 mmol/L KCl and
1.5 mmol/L MgCl2) containing 0.2 mmol/L of each
dNTP, 20 ng each of RG100 primers, and 40 ng of
each of hpt or cry1Ab Bt primers and 1 U of Taq
polymerase. PCR products were then analyzed by
agarose gel electrophoresis [14]
.
H-NMR sample preparation
For each sample, 2.5 g of complete rice grain was
powdered. Sample extraction was carried out in three
steps. In the first and second steps, 10 mL methanol
was added to the rice powder in a flask. The third step
was addition of 15 mL of 2:1 mixture of methanol and
dichloromethane. In each step, the powdered rice grain
and solvent were stirred for 45 min and the resulting
solution was transmitted to a glass vessel through
filter paper. After solvent volatilization, a 13 mg per
700 L solution was prepared as sample for H-NMR
spectroscopy by addition of d6-DMSO to the residue.
The NMR instrument was a 400 MHz Brucker.
FTIR sample preparation
For each sample, 2.5 g of complete rice grain was
fine powdered. The thin tablets used as samples in
FTIR were prepared by mixing the fine powdered rice
with KBr with 2% ratio. The FTIR instrument was a
Brucker Tensor 27.
Statistical analysis
Linear discriminant analysis (LDA) is a frequently
used supervised classification method. Discriminant
analysis finds a set of prediction equations based on
independent variables that are used to classify
individuals into groups [15]
. The main target of principal
Fig. 1. PCR analysis of seeds from putative transgenic rice.
PCR primers for rice locus RG100 (1.0 kb product) and cry1Ab
gene (1.2 kb product) as described by Ghareyazie et al [14]
. Sample
DNA: Lane 1, Non-transgenic rice; Lane 2, Transgenic rice; Lane 3,
Water control (no DNA); Lane 4, Molecular weight markers. The
primers for locus RG100 are 5 -GCTGGACGTGCCAAAGAGAG-3
(forward) and 5 -CGAACCACAGCCACAGCATG-3 (reverse). The
primers for the cry1Ab gene are 5 -GGCGGCGAGAGGATCGAGAC-
3 (forward) and 5 -GGCGGGACGTTGTTGTTC-3 (reverse).
Keykhosrow KEYMANESH, et al. Metabolome Comparison of Transgenic and Non-transgenic Rice by FTIR and NMR Spectra 121
component analysis (PCA) is to reduce the number of
variables in such a way that most of the variation
would be expressed by new uncorrelated variables or
principal components. Based on principal components,
classification of samples is possible in a way that
between-class variance is maximized and individuals
in the same group have the least difference [16]
.
There are many commercial computer applications
that are used successfully in similar experiences. In
this study, we used NCSS (2006 release) and MESTREC
(version 4.9.9, 2006) to implement the analysis process
from the first phase which is transformation of
spectroscopy diagram in a manner that can be used as
an input material for the future multivariate analysis.
Preparing data for analysis of H-NMR and FTIR
spectra
Using the MESTREC software, H-NMR spectra
of samples were transformed in such a way, in which
every spectrum was redefined as a point in a multidimensional
space. In fact, it was the key operation in
which its results were used as input for the oncoming
analysis phase. For the FTIR spectra, according to the
above-mentioned process, for first stage operation, we
used area under the curve option of NCSS application
to calculate the area under the curve for 187 limited
areas which were defined for all spectra. In this way,
the results for each spectrum mirrored the shape of it;
therefore, these data were used as raw inputs for the
statistical analysis stage leading to classification. Fig.
2 shows the way of data preparation for H-NMR and
FTIR experiments in a graphical manner.
RESULTS
As shown in Fig. 3, principal component analysis
indicated no definite clusters. The points that represent
the samples are dispersed in a non-distinctive manner
in the PCA diagram of FTIR. The first and second
scores represent 69.89% and 17.41% of total variations,
respectively. The PCA diagram of H-NMR is better
clustered though the accumulation of samples has no
rational meaning. In this case, the first and second
scores cover 23.33% and 19.97% of variations,
respectively. The difference between diagrams could
be due to the fact that in the latter case only 43.3% of
the total variations has been involved in contrast to
87.3% of FTIR case, so the less variation among
individuals has led to less disseminated points.
Classification of samples by LDA was carried out
Fig. 2. The way that the H-NMR and FTIR spectra were processed for use in statistical analysis.
The images show graphically how the spectroscopic graphs were prepared for data extraction. The images are merely figurative and for better
understanding of the process.
122 Rice Science, Vol. 16, No. 2, 2009
through stepwise selection in which retention or
removal of variables depended on PIN and POUT values
in NCSS application, respectively. For PIN, 0.5 and
for POUT, 0.99 are known as suitable values since
these amounts would permit the most number of
efficient variables. As indicated in Fig. 4, for FTIR
spectra, discriminant functions enabled the classification
with 95.23% accuracy. H-NMR spectra of samples
were classified with 100% accuracy.
DISCUSSION
The results showed the advantage of supervised
methods over unsupervised one in case of classification.
Similar conclusion has been acquired in other studies[17]
.
However, we should have in mind that unsupervised
methods would be more reliable in the case that there
is no record of the samples in hand.
At this level, statistical comparison of extracts
from different plant materials only informed us
whether the compared materials have any significant
differences and based on which the materials could be
distinguished and be classified in different groups.
From the statistical view point, abundance of
variables which are calculated integral of limited
counterpart areas of spectra obliged us to use data
reduction procedures such as PCA and LDA that not
only classify the samples but also implement the
classification by use of the variables which represent a
major portion of the variation.
The differences between transgenic and nontransgenic
materials could have different reasons.
These differences may be resulted from metabolites
that the genetic manipulation is done for their
production. On the other hand, the differences may be
the result of increase or decrease in production of
metabolites other than the targeted one. These changes
may include some unpredicted events which probably
Fig. 3. Diagrams resulted from principal component analysis of FTIR and H-NMR spectra.
Numbers 1 to 10 are common rice samples, and numbers 11 to 20 are transgenic rice samples and number 21 is a barley sample.
Fig. 4. Diagrams resulted from linear discriminant analysis of FTIR and H-NMR spectra.
Keykhosrow KEYMANESH, et al. Metabolome Comparison of Transgenic and Non-transgenic Rice by FTIR and NMR Spectra 123
lead to unwanted effects on consumer health. Therefore,
identifying the reason of differences between transgenic
and their non-transgenic counterparts could be the
next step in such investigations. In addition, the
statistical comparison of plant materials has further
advantages than a primitive stage in regulations
related to transgenic or genetically modified (GM)
products; in fact, it can be used as a quick method for
identification of transgenic products in cases that there
is no reliable information about the plant material, e.g.
in an imported consignment. Due to increasing share
of GM products, the improvement and facilitation of
identification methods play a vital role in accurate
implementation of regulations which are approved in
many countries for control and assessment of GM
products [18]
.
REFERENCES
1 Thomsona J A, Mundreea S G, Farrant J M. The development
of genetically modified maize for abiotic stress tolerance. S
Afr J Bot, 2007, 73: 494­495.
2 Rafiq M, Fatima T, Husnain T, Bashir K, Khan M A,
Riazuddin S. Regeneration and transformation of an elite
inbred line of maize (Zea mays L.), with a gene from Bacillus
thuringiensis. S Afr J Bot, 2006, 72: 460­466.
3 Sharma A, Sharma R, Imamura M, Yamakawa M, Machii H.
Transgenic expression of cecropin B, an antibacterial peptide
from Bombyx mori, confers enhanced resistance to bacterial
leaf blight in rice. FEBS Lett, 2000, 484: 7­11.
4 Alexander J S, Sachdev H P S, Matin Q. Genetic engineering
for the poor: Golden rice and public health in India. World
Dev, 2008, 36: 144­158.
5 Thomson J. Genetically modified food crops for improving
agricultural practice and their effects on human health. Trends
Food Sci Tech, 2003, 14: 210­228.
6 Miraglia M, Berdal K G, Brera C, Corbisier P, Holst-Jensen
A, Kok E J, Marvin H J P, Schimmel H, Rentsch J, van Rie J
P P F, Zagon J. Detection and traceability of genetically
modified organisms in the food production chain. Food Chem
Toxicol, 2004, 42: 1157­1180.
7 Konig A, Cockburn A, Crevel R W R, Debruyne E,
Grafstroem R, Hammerling U, Kimber I, Knudsen I, Kuiper
H A, Peijnenburg A A C M, Penninks A H, Poulsen M,
Schauzu M, Wal J M. Assessment of the safety of foods
derived from genetically modified (GM) crops. Food Chem
Toxicol, 2004, 42: 1047­1088.
8 Rischer H, Oksman-Caldentey K M. Unintended effects in
genetically modified crops: Revealed by metabolomics?
Trends Biotechnol, 2006, 24: 102­104.
9 Charlton A, Allnutt T, Holmes S, Chisholm J, Bean S, Ellis N,
Mullineaux P, Oehlschlager S. NMR profiling of transgenic
peas. Plant Biotechnol J, 2004, 2: 27­35.
10 Serkova N J, Spratlin J L, Eckhardt S G. NMR-based
metabolomics: Translational application and treatment of
cancer. Curr Opin Mol Ther, 2007, 9: 572­585.
11 Emura K, Yamanaka S, Isoda H, Watanabe K N. Estimation
for different genotypes of plants based on DNA analysis using
near-infrared (NIR) and Fourier-transform infrared (FT-IR)
spectroscopy. Breeding Sci, 2006, 56: 399­403.
12 Colquhoun I J, Le Gall G, Elliott K A, Mellon F A, Michael
A J. Shall I compare thee to a GM potato? Trends Genet,
2006, 22: 525­528.
13 Wang K H, Cho Y G, Yoon U H, Eun M Y. A rapid DNA
extraction methods for RFLP and PCR analysis from a single
dry seed. Plant Mol Biol Rep, 1998, 16: 1­9.
14 Gharayazie B, Alinia F, Menguito C A, Rubia L G, de Palma
J M, Liwanag E A, Cohen M B, Khush G S, Bennett J.
Enhance resistance to two stem borers in an aromatic rice
containing a synthetic cry1A(b) gene. Mol Breeding, 1997, 3:
401­414.
15 Park C H, Park H. A comparison of generalized linear
discriminant analysis algorithms. Pattern Recogn, 2008, 41:
1083­1097.
16 Shih F Y, Zhang K. A distance-based separator representation
for pattern classification. Image Vis Comput, 2008, 26: 667­
672.
17 Rezzi S, Axelson D E, Héberger K, Reniero F, Mariani C,
Guillou C. Classification of olive oils using high throughput
flow 1
H NMR fingerprinting with principal component
analysis, linear discriminant analysis and probabilistic neural
networks. Anal Chim Acta, 2005, 552: 13­24.
18 Fontes E M G. Legal and regulatory concerns about
transgenic plants in Brazil. J Invertebr Pathol, 2003, 83: 100­
103.