Content uploaded by Ping Song
Author content
All content in this area was uploaded by Ping Song on Apr 01, 2015
Content may be subject to copyright.
FOOD COMPOSITION AND ADDITIVES
Polymerase Chain Reaction Technology as Analytical Tool in
Agricultural Biotechnology
MARKUS LIPP
Monsanto Co., 800 N. Lindbergh Blvd, St. Louis, MO 63167
RAYMOND SHILLITO
Bayer CropScience, PO Box 12014, 2 T.W. Alexander Dr, Research Triangle Park, NC 27709
RANDAL GIROUX
Cargill Inc., 2301 Crosby Rd, Wayzata, MN 55391
FRANK SPIEGELHALTER
GeneScan USA, 2315 N. Causeway Blvd, Suite 200, Metairie, LA 70001
STACY CHARLTON
Syngenta Biotechnology, PO Box 12257, 3054 Cornwallis Rd, Research Triangle Park, NC 27709-2257
DAV I D P INERO
Eurofins Scientific, 3507 Delaware, Des Moines, IA 50313
PING SONG
Dow AgroSciences, 9330 Zionsville Rd, Indianapolis, IN 46268-1053
The agricultural biotechnology industry applies
polymerase chain reaction (PCR) technology at
numerous points in product development.
Commodity and food companies as well as
third-party diagnostic testing companies also rely
on PCR technology for a number of purposes. The
primary use of the technology is to verify the
presence or absence of genetically modified (GM)
material in a product or to quantify the amount of
GM material present in a product. This article
describes the fundamental elements of PCR
analysis and its application to the testing of grains.
The document highlights the many areas to which
attention must be paid in order to produce reliable
test results. These include sample preparation,
method validation, choice of appropriate reference
materials, and biological and instrumental sources
of error. The article also discusses issues related
to the analysis of different matrixes and the effect
they may have on the accuracy of the PCR
analytical results.
T
he class of plant products developed through modern
biotechnology has been described as genetically
modified (GM), genetically engineered (GE),
genetically modified organism (GMO), transgenic, biotech,
and recombinant. For the present discussion, the term
“genetically modified” (GM) will be used for its simplicity
and broad recognition. This discussion will address
polymerase chain reaction (PCR) technology as it applies to
food biotechnology in the soybean and corn (maize) industries
only, though the principles are applicable to other crops as
well. PCR is only one of the techniques that are used for the
detection of GM material in a product. Although
protein-based test technology is available and applied to
testing (1), especially in the seed and grain industry, the
remainder of the article will focus exclusively on PCR
technology.
A number of countries have adopted, or are in the process
of developing, legislation related to the approval of GM
products. Authorities in many countries require that DNA
sequence information be provided as part of the registration
package. In addition, a PCR detection method that is specific
to the event may also be requested. The term “event” is used to
describe a plant and its offspring that contain a specific
insertion of DNA. Such an event is distinguishable from other
events by its unique site of integration of the introduced DNA.
A PCR method that can distinguish such an event from all
other events is described as being “event-specific” and
generally is based on the detection of a junction fragment
between the original plant DNA and the introduced DNA.
Uses of PCR Technology
The agricultural biotechnology industry applies PCR
technology at numerous steps throughout product
development, much as it does with immunoassays (1). The
major uses of PCR technology during product development
include gene discovery and cloning, vector construction,
transformant identification, screening and characterization,
and seed quality control. Commodity and food companies, as
well as third-party diagnostic testing companies, rely on PCR
136 LIPP ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 88, NO. 1, 2005
Received February 27, 2004. Accepted by AH July 28, 2004.
Corresponding author's e-mail: markus.lipp@monsanto.com.
technology to verify the presence or absence of GM material
in a product or to quantify the amount of GM material present
in a product. Quantitative PCR technology also has been used
to estimate GM copy number and zygosity in seeds and
plants (2–4).
The grain handling and grain processing industry uses
PCR to certify compliance with contracts between buyer and
seller. PCR testing is used for 4 specific purposes in the grain
handling/processing industry:
PCR testing for unapproved events.—In countries that
have a defined approval process for GM crops, an event may
be approved for use in the country of production but not yet
approved for use in an importing country. In these instances,
the importing country often requires that the grain shipment
be tested for the presence of specific GM events to ensure that
the grain shipment does not contain these unapproved events.
Such testing often relies on qualitative PCR because the
detection of these events, in most cases, is at a zero-tolerance
threshold.
PCR testing for GM content.—Most countries that have
adopted mandatory labeling rules for food or feed have set
tolerances for the adventitious presence of GM material in
grain products or the final foods based on a percent GM
(weight-to-weight) content. In these countries, food and feed
manufacturers and retailers often choose to originate/obtain
grain and grain products below the defined regulatory
threshold to avoid labeling their products. In this case, grain
must come from a non-GM identity preservation program and
be certified to contain GM grains at a level below the
threshold specified in the contract. To meet this need for
testing, several laboratories currently are adopting
quantitative PCR for percent GM determinations.
PCR testing for non-GM labeling.—In some cases, food
manufacturers and retailers wish to use positive labeling for
their non-GM products. These companies hope to gain market
share among consumers who wish to avoid GM products. In
most cases, the use of positive labeling requires that the grain
and grain products originate from a non-GM identity
preservation program and test negative or at least below a
certain threshold for GM DNA. Qualitative PCR testing is
most often used to certify compliance with a non-GM
contract.
PCR testing for presence of a high-value commodity.—In
certain cases, it is desirable to show that a commodity is made
up of a specific crop commodity (e.g., low phytate maize,
soybean with altered oil profile). PCR could be used for this
purpose by testing for the GM trait that conveys the
characteristic, although the grain may also be tested by
quantifying the improved quality of the commodity.
The PCR Process
The PCR process mimics in vitro the natural process of
DNA replication occurring in all cellular organisms in which
the DNA molecules of a cell are duplicated prior to cell
division. In contrast to natural DNA replication, the DNA
reproduction during PCR does not cover the entire sequence
of the original DNA molecules but is restricted and targets a
specific, relatively short, region of the template DNA
molecules. Short, single-stranded, synthetic DNA molecules
called the primers give the specificity of the reaction. They are
designed to be complementary to their intended binding site.
Most commonly, 2 primers are involved and the DNA section
in between the distal ends of their binding sites are replicated
during the reaction.
A single cycle of the PCR and the corresponding
temperature profile are typically divided into the 3 phases:
denaturation, annealing, and elongation. At the end of this
procedure, the targeted DNA region has been replicated into
2 copies of the original double-helix molecule. This process of
selective duplication is repeated multiple times in a cyclic
reaction. The repetitive DNA duplication is driven solely
through quick and precise shifts in the reaction temperature,
facilitated by the thermocycler instrument. DNA replication is
catalyzed by heat-stable DNA polymerases, previously
isolated and characterized from Thermus aquaticus (5). The
kinetics of the DNA reproduction resemble an exponential
amplification in which the replicas of distinct length
(amplicons) accumulate quickly and outnumber the original
template molecules. The distinct size of the amplified copies
allows them to be detected by gel electrophoresis in the
background of nonamplified DNA.
Sample Preparation
Sampling
If applying PCR to test for GM material, one must carefully
conduct the sampling in a manner that avoids erroneous
results. When PCR-based diagnostic assays are used to test for
the presence of GM material in seed or grain, a number of
sampling steps occur (Figure 1): (1) Sampling the
consignment of seed or grain to obtain the bulk sample.
(2) Sampling the bulk sample to obtain the laboratory sample.
(3) Subsampling the laboratory sample to obtain the test
sample. (4) Sampling the meal that results from grinding the
test sample to obtain the analytical sample. (5) Sampling the
DNA solution that results from extraction of the meal sample
to obtain the test portion.
Typically, sampling considerations are limited to the
laboratory and test sample with regard to the desired threshold
of detection. The subsequent sampling step (meal) is typically
driven by the traditional procedures of a particular laboratory
or the capacity of the equipment used. It is incumbent on the
analytical laboratory to have characterized and standardized
the production of the meal sample as part of the method
development. The final sampling step (DNA to be used in a
single PCR) is typically constrained to <200 ng DNA by the
limitations of the technique and the equipment used.
Each of the sampling steps has the potential to introduce
error that may impact the detection of GM material at the
desired threshold. As such, designing a sampling strategy that
will be suitably representative requires knowledge of the
particle size characteristics of the test sample and the meal, the
genome size of the species in question, and the limit of
LIPP ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 88, NO. 1, 2005 137
detection (LOD) or range of quantification (ROQ) of the
analytical technique. Sampling considerations have been
addressed in a number of recent papers (6, 7).
Although it is often overlooked, the overall sampling
method must be carefully designed and characterized.
Particular attention should be given to understanding the
limitations of the analytical technique as it relates to the
sampling plan when testing for GM material at low thresholds
of detection.
The following is an example of a theoretical exercise that
demonstrates the impact of sampling on the ability of a
qualitative detection method to detect small quantities. To
frame the exercise, we must first make some basic
assumptions about the biological material to be analyzed: (1)
the objective of the analysis is to detect trace amounts of GM
material in maize; (2) the GM material is in the hemizygous
state (and no information is available on the genetic
background/variety); (3) the PCR sample size is
approximately 40 000 maize genomes, which equals
~109 ng (8); (4) to compensate for sampling error at the DNA
sampling stage, the minimum nominal number of target
sequences must be 20 per reaction as suggested by Kay and
Van den Eede (9); (5) the target gene is present as a single
copy per haploid genome and, thus, with the same relative
abundance as the endogenous control gene.
The first calculation estimates the minimum level of
percent GM (on a weight/weight basis assuming all kernels
have equal weight) that would be detected in the sample and is
called the relative LOD. To calculate this value, we need to
divide the minimum number of genome copies that can be
reliably detected in a qualitative analysis (20 copies) by the
total number of copies in the reaction tube (40 000 copies).
The result of this must be multiplied by 2 to account for the
fact that in the hemizygous state, the GM gene is present in
only 1 of the 2 parental genomes. The result of the calculation:
20 ¸ 40 000 (ratio of GM genomes to total) ´ 2 (hemizygosity
correction factor) ´ 100 = 0.1%. This represents the lowest
concentration of GM that can be reliably detected using
qualitative PCR.
In this example, the meal sample from which the DNA
preparation is extracted must be at least 0.1% GM in order to
achieve reliable detection. This fact imposes constraints on
the choice of pool size and meal sample size. In this example,
a seed pool containing a single GM seed must result in a meal
sample that is at least 0.1% GM with high confidence.
If one accepts the limitations of the analytical technique
outlined above (i.e., LOD equal to 20 target sequences per
reaction, equivalent to 0.1% GM), the absolute maximum
pool size for which one can expect reliable detection of the
presence of a single positive kernel in the pool is 1000 seeds, if
all the seeds in the pool are ground to meal and the entire
resulting meal preparation is used for DNA extraction
(thereby eliminating sampling error at the meal sample stage).
Using a pool size >1000 will result in situations in which some
positive pools will have GM levels below the LOD of the
analytical technique. In order to test to a very low threshold,
multiple sample approaches must be adopted to increase the
overall sample size. For example, assuming a binomial
distribution and testing a single pool of 998 seeds with a
negative result provides 95% confidence that GM content is
<0.3%. In order to have 95% confidence that GM content is
<0.1%, one must test 2995 seeds, whether in a single pool or
as multiple pools, with no pools testing positive
(http://www.seedtest.org/upload/cms/user/seedcalc6.zip).
DNA Extraction and Matrix Effects
The performance of an analytical method will vary with the
nature of the sample under study. Typically, a method will be
developed and validated for only 1 sample type or a very
restricted set of different matrixes. Modifications to the
method may be required to accommodate other matrixes,
thereby creating a different method/procedure.
This section outlines aspects that apply to effects of the
sample matrix on the performance of the method. Although
the considerations focus on effects from evidently different
matrixes, it must be noted that the term “sample matrix” may
not be clearly defined. There may be ambiguity, and the
analyst must decide if the unknown sample falls in the same
138 LIPP ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 88, NO. 1, 2005
Figure 1. Sampling steps for PCR-based diagnostic
assays for GM detection.
category as the samples used during validation of the method.
A detailed scope of the method will minimize such ambiguity
but cannot entirely prevent it.
PCR methods for detection of GM traits are commonly
developed and validated with samples of ground seed because
validation studies, proficiency schemes, and check sample
programs with large samples of whole seed are not practical.
However, initiating validation and proficiency programs with
ground seeds ignores the variability in the preparation
process. Some organizations, notably the International Seed
Testing Association (ISTA), have carried out proficiency
programs using whole seed.
In routine applications, the analyst will frequently use PCR
to analyze grain. However, even a matrix that appears simple
from the method developer’s point of view, i.e., maize kernels,
may be challenging in a laboratory situation. For example, a
seed sample may have been treated with chemicals such as
fungicides, which were not present in the samples used for
method validation. These compounds may interfere with or
inhibit the PCR, particularly if the DNA extraction procedure
is not tailored for their effective removal. If the method does
not include appropriate controls, this inhibition can lead to
false-negative results. In addition, any 2 maize samples are
unlikely to be exactly the same with regard to such
characteristics as moisture, fiber, starch, and residues of
chemicals, and at some point the assumption must be made
that they belong to the same matrix, although the boundaries
of this category cannot be exactly defined.
A prudent approach to this challenge is not to rely solely on
method validation for a particular matrix or to assume all
samples that are considered the same matrix will behave in
exactly the same way. Controls that monitor performance
should be developed as part of the method to detect potential
effects originating from the individual unknown sample under
study. The effectiveness of these controls should be
demonstrated during method development and, ideally, also
during validation.
(a) PCR inhibitors.—Various compounds in plant
material and food products can be co-extracted with genomic
DNA and inhibit the PCR. This may lead to false-negative
results because of failure of the PCR (10). Likewise, partial
inhibition is very likely to bias results from a relative
quantitative PCR assay using 2 PCR systems in parallel,
because the 2 PCR systems will rarely be affected to exactly
the same extent.
PCR inhibitors from plants and processed foods include
polysaccharides, proteins, phenolic compounds, and other
uncharacterized plant secondary metabolites (11–13).
Moreover, covalent cross-linking of proteins to DNA through
carbohydrates can render the DNA unsuitable as a PCR
template (14, 15). The inhibitors may vary in the extent to
which they affect individual PCRs using different primer pairs
in separate or multiplex PCRs (16, 17).
Although it is necessary to use or develop a DNA
extraction method that sufficiently removes PCR inhibitors
from the matrixes that fall within the scope of the method, it is
not practical to identify and characterize all potential
inhibitors. A more feasible approach is to test DNA extracts
for the presence of inhibitors and then modify the DNA
isolation protocol or the PCR conditions to reduce the effects
of PCR inhibitors, if necessary (16–19). Generally, a positive
control DNA (spike) is added to the PCR. Specific
amplification of the control DNA is tested in the presence and
absence of DNA extracts to see if the extract inhibits
amplification. This should preferably be done with DNA
extracts from known negative samples so that the spike DNA
is the only source of a constant amount of target DNA in the
PCR. There are several examples of DNA extraction methods
that target removal of specific inhibitors (12, 13, 18, 20, 21).
An inhibition control in the method is a requisite to monitor
potential inhibition arising from an individual sample in
routine application of the method. Such a control will
considerably lower the risk of false-negative results caused by
PCR inhibition. During method development and validation,
the fitness of an inhibition control can be assessed by
analyzing samples with small amounts of analyte close to the
anticipated or intended detection limit. In PCRs that contain
purposely added inhibitory compounds or crude DNA
extracts that are known to contain inhibitors, the failure of the
PCR (false negative) should be indicated by the malfunction
of the inhibition control. Holden et al. (10) have shown that
the amplification of an endogenous sequence does not always
fulfill this requirement. The sensitivity of this type of reaction
as an inhibition control is restricted, probably because the
presence of the endogenous target is greater than the presence
of the GM target; therefore, it still can be amplified
successfully in partially inhibited reactions where
amplification of the GM target fails. Ideally, accept/reject
criteria for negative results for an individual sample consider
the outcome of the corresponding inhibition control and are
described under the appropriate sections of method validation.
(b) DNA degradation, fragmentation, and
extractability.—Processing of raw agricultural commodities
to food ingredients and finished food products usually
comprises steps that extract, fragment, or otherwise
compromise the DNA molecules. Complex sample matrixes
may also require multiple consecutive DNA extraction and
purification steps that lower the DNA yield of the overall
procedure. Reduced size of the DNA molecules that can be
extracted from a processed matrix is of concern if a
considerable portion of the fragments cannot function as PCR
templates because their insufficient size does not span the
entire target sequence. Also, certain types of DNA damage
during food processing may interfere with the DNA’s ability
to serve as a PCR template. Whereas it may not be required or
feasible to address the exact nature of these limitations,
validation studies will reasonably show their impact on the
applicability of a method to a processed matrix.
These effects imply that a detection or quantification limit
established for a given method is restricted to the matrix used
during validation (most commonly, raw material, i.e, ground
seeds) and cannot be projected to any other matrix outside the
scope of the method. Decreasing the amount or otherwise
compromising the integrity of DNA that can be extracted from
LIPP ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 88, NO. 1, 2005 139
the sample and amplified in PCR reciprocally increases the
detection and quantification limits expressed in terms of
percent GM DNA relative to plant species DNA. In extreme
cases, only minute amounts of DNA can be extracted from
certain matrixes (e.g., refined oil, modified starch, and
soybean lecithin). If the DNA yield is extremely low, or only
an insignificant portion of the extracted DNA is present in
amplifiable fragments larger than the PCR target sequence,
attempts to establish the LOD of a PCR method by validation
likely will fail. With significant lack of extractable and
amplifiable DNA, relative quantification will fail as well, with
poor reproducibility and very high or entirely unknown
quantification limits. Modifications to DNA extraction
procedures may compensate for extremely low DNA yield
from such matrixes to some extent and may result in an
improvement of the detection limit, but are still not likely to
result in methods that are practicable in the real world.
In contrast to the decrease of DNA content by most food
processing, it is possible that some procedures such as
freeze-drying may concentrate the DNA in the matrix and,
thereby, decrease the lower limits of the method’s
applicability. The detection limit will remain unknown for the
processed matrix unless established by appropriate validation.
Because the concept of relative quantitative PCR involves
the measurement of 2 analytes in parallel, removal,
degradation/fragmentation, or restricted extractability of
DNA have the potential to shift the ratio between these
2 analytes if they are not equally subject to the adverse effects.
Variation in nuclease susceptibility of different regions of the
genome is a well-established phenomenon particularly with
respect to Dnase I sensitivity of actively transcribed chromatin
regions (22–25). Differences in nuclease susceptibility may
arise as a result of a number of factors. Some of these factors
may be relatively global in scope, such as higher-order
chromatin structure, whereas others may be quite local, such
as methylation state of specific sequences, nucleosome
placement, or level of transcriptional activity. Whatever the
root cause, differential nuclease susceptibility between the
endogenous control gene and the target sequence can result in
biased results particularly when dealing with processed
materials. For example, processing that preferentially
eliminates the DNA sequence used to measure the amount of
plant species DNA would increase the result given in percent
GM DNA relative to plant species DNA in the processed
matrix. Although results that are expressed in terms of GM
DNA relative to plant species DNA are, strictly speaking, still
correct, the restriction that may arise to their interpretation is
evident. They do not necessarily reflect the relative
composition of the original material.
(c) Mixed products with DNA from multiple
sources.—The specificity of a PCR assay for its intended
target analyte is determined by the likelihood that sequences
that are similar to, but not the same as the target sequence are
not amplified and therefore do not lead to false-positive
results. Besides the ability of the primers and PCR conditions
to discriminate between the intended and unintended
detection of closely related sequences, the composition of the
DNA in the sample matrix plays a major role. The minimum
requirement for the specificity of a PCR assay for GM
detection is that it does not cross-react with any of the DNA
sequences that are present in the non-GM plant genome.
Furthermore, it is critical that sequences from other GM
events in the same species do not interfere with the assay in a
manner that could lead to positive signals that do not originate
from the intended GM analyte. However, PCR analyses that
are targeted to a sequence present in many different events
(e.g., the 35S promoter) will react with those multiple events.
Compliance with the minimum specificity requirements
can be established during development and validation of a
method. However, the scope of such method would be fairly
limited and not necessarily suitable for routine application in
daily laboratory practice. For example, a method for detection
of a GM maize event may show a satisfactory specificity on
maize samples prepared for a validation study; however, this
does not demonstrate the method’s suitability for analysis of
commercial maize samples that contain foreign materials,
such as soybean or wheat. Therefore, it is advisable to design
and validate a method so that it is specific for the intended GM
event, even in the presence of DNA from other plant species
that are likely to be present as adventitious material in routine
grain or processed samples.
The same argument applies to methods that are intended
for use in more complex food matrixes that may have even
more sources of DNA from different organisms than a grain or
flour sample. Experimental assessment of the method’s
specificity on DNA from various sources will define the scope
of the methods that must be considered during method
validation.
PCR methods that target artificial transitions between
DNA sequences that naturally do not occur in juxtaposition
are believed to be less problematic with regard to their
specificity than PCR methods that target a sequence located
within only 1 genetic element that has been used in genetic
engineering of plants. For example, the detection of
cauliflower mosaic virus (CaMV) 35S promoter sequences is
not necessarily conclusive evidence that the DNA detected
originated from a GM plant. The virus itself may be present
(the CaMV is a common plant virus affecting e.g., Brassica
species) and lead to a positive result, and the likelihood of its
presence could vary considerably with the sample matrix, i.e.,
the ingredients of a finished food product. PCR assays that
detect CaMV itself by targeting a viral nucleotide sequence
other than 35S promoter can be useful to investigate the
probability of 35S promoter positive results originating from
CaMV (26). In other words, the presence of the 35S promoter
can be caused by a naturally occurring plant virus and its
presence per se cannot be linked to the presence of GM plants
without further additional information. Similarly, the
detection of a gene derived from Bacillus thuringiensis may
be the consequence of traces of DNA from soil bacteria
present in a grain sample. If possible, these limitations should
be considered in the scope of the method and instructions
given for interpretation of results.
140 LIPP ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 88, NO. 1, 2005
The presence of CaMV in considerable amounts can also
affect results from quantitative PCR that measure the ratio
between 35S promoter DNA and a species-specific DNA
sequence. Although the result would probably still reflect the
ratio of the 2 types of DNA correctly, it may defy reasonable
interpretation because the contributions from CaMV
sequences and GM DNA cannot be distinguished.
A further major restriction to the interpretation of 35S
promoter quantification in complex matrixes is GM DNA
from different crop plants that could contribute to the value
obtained for the 35S promoter DNA; most commonly, only
1 plant species at a time is targeted in the species-specific PCR
in parallel. Multiple significant contributions to the numerator
of the percent GM calculation, without the contribution of all
respective plant species to the denominator, would
consequently yield an overestimate that does not reflect
percent GM.
Applications of PCR
PCR can be used in 2 primary ways in the detection of GM
DNA in plants. These are termed quantitative PCR, which
yields an estimate of the amount of the specific analyte
present, and qualitative PCR, which yields a yes/no answer as
to the presence of GM material.
Quantification of DNA
The analyst must be aware of the measurement uncertainty
in the determination of the amount or concentration of DNA
used in an experiment. The following list contains a number of
factors that contribute to this uncertainty (27). This list is not
considered exhaustive:
The following factors are known to influence the accuracy
and precision of DNA quantification by UV spectrometry:
(1) presence of other components absorbing at 260 nm, e.g.,
proteins, RNA; (2) ratio of single-stranded vs double-stranded
DNA; their absorption coefficients differ; (3) size distribution
of DNA in solution.
The following factors are known to influence the accuracy
and precision of DNA quantification by
fluorescence-spectrometry: (1) size distribution; (2) in cases
of dyes that bind exclusively to double-stranded DNA,
single-stranded DNA, if present, will not be determined at all.
Currently, all DNA concentration quantification
techniques have limitations in their use and application.
Spectrophotometric analyses (i.e., A260-A280) require a
relatively large amount (2.5–5.0 mg) of DNA of almost pure
quality. DNA extracted from certain food matrixes is unlikely
to meet this requirement. Spectrophotometric assays are also
unable to differentiate between single- and double-stranded
DNA or between DNA and RNA. Fluorometric assays require
that a DNA standard of a comparable size, and in the case of
Hoescht assays, adenine and cytosine content be used (28, 29).
Thus, all DNA quantification methods have their strengths
and weaknesses, although the spectroscopic determination
with absorbance at 260 nm is commonly used. Regardless of
method choice, the analyst must recognize that the uncertainty
of the DNA determination will be an inherent part of the total
uncertainty of the method and, thus, be reflected in each
analytical result. No method will be more precise and accurate
than the estimates of the concentrations of its calibrators.
For some applications, DNA is diluted exhaustively
(through sequential dilution the number of PCR target
molecules is decreased) to determine the copy number of the
limit of a method, or to quantify the amount of DNA by
measurement of an endogenous gene. For very low numbers
of molecules, stochastic effects will predominate.
The best estimate of the absolute quantity of DNA in a
given reaction volume can be determined using one of the
techniques described above. This value is multiplied by the
genome size (8) as given in literature to express the quantity as
copy number or genome equivalent. The uncertainty of this
value is determined by the uncertainty of the DNA
measurement technique and the genome size, as well as
dilution and any absorption by the apparatus. However, we are
not aware if any good estimates of uncertainty for the
biological variation of the genome size as expressed in
equivalent/mole or equivalent/g are available. Thus, the
analyst needs to be cautious about using copy number
calculations in reporting results.
Qualitative PCR
In a qualitative analytical setup, the PCR components are
combined with DNA extracted from the unknown sample. If
the DNA sample contains the target DNA sequence in
question, this sample DNA will function as the template DNA
that can be amplified successfully. Together with appropriate
negative controls, detection of the correct PCR product
indicates the presence of the targeted DNA sequence in the
original sample. Absence of PCR product in conjunction with
suitable positive controls implies the DNA target was not
present in detectable amounts.
Qualitative PCR products are commonly analyzed by
agarose or polyacrylamide gel electrophoresis (30). Applying
a voltage will cause the negatively charged DNA to migrate
and will separate DNA fragments according to their length.
The very large numbers of identical DNA molecules that are
the product of the PCR form a distinct band that can be
visualized as UV fluorescence using the fluorophor ethidium
bromide or other means.
At the end of the PCR, the intensity of the signals may vary
between samples. However, the signal does not necessarily
correlate in a linear way to the amount of target DNA that was
present in the beginning of the reaction, primarily because
PCR of this type stall and enter a plateau phase after large
amounts of PCR product have been made, usually due to
exhaustion of one or more substrate(s). Analyzing PCR
products after the reaction is finished restricts the results to
merely detected or not detected (positive or negative).
Qualitative PCR assays are used in 2 main ways. The first
way is a simple test to determine whether the sequence in
question is present in a bulk sample (usually flour or other
processed material). The second way is semiquantitative. If
the sample is made up of seeds or grain, or other discrete units,
LIPP ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 88, NO. 1, 2005 141
a number of test samples can be used to estimate the number of
particles (seeds or grains) that contain the target analyte. As an
example, instead of testing a single bulk of 1000 seeds or
grains, the analyst can make 10 pools of 100 seeds or grains.
The number of pools that test positive for the analyte gives an
estimate of the number of positive seeds in the lot. This
method works only when the percentage of positive seeds in
the sample is low (typically below 5%). For example, if 5 out
of 10 pools each containing 100 seeds test positive, then the
calculated level of the adventitious presence is 0.69% and the
95% confidence limits are 0.21 to 1.66% (31). This is much
more information than can be gained by testing a single pool,
in which the answer would be “positive.”
Another advantage of the semiquantitative approach is that
the method can be applied in a range that is well above the
LOD; thus, the likelihood of false-positive or -negative results
will be significantly lower. Nevertheless, care has to be taken
that contamination with fragments of seed, grain, or dust does
not cause false-positive results.
Quantitative PCR
There are various approaches to quantification of GM
material in a sample using PCR. In all cases, quantification by
PCR determines the amount of GM DNA vs a reference DNA
target (e.g., maize or soy DNA). This is not a direct
weight-to-weight measurement. The following is a discussion
of various real-time PCR chemistries, as well as different
approaches for standard curve generation and data analysis.
Real-time PCR technology allows for the monitoring of
fluorescence associated with amplification products
throughout the PCR process. This technology is available
with different types of fluorescent chemistries. Examples are
fluorogenic probes (i.e., TaqMan
®
, FRET), Scorpion
™
primers, and SYBR
®
Green.
With TaqMan fluorogenic probes, an additional
oligonucleotide, located between the 2 primers required for
amplification, is added. This probe is labeled with a
fluorescent reporter dye and a corresponding quencher dye.
The quencher dye absorbs the fluorescence from the reporter
dye, and when they are in close proximity, no fluorescent
signal is emitted. During the extension phase of the PCR, the
polymerase breaks down the probe, thus physically separating
the quencher dye from the reporter dye. This results in the
reporter dye emitting a fluorescent signal upon excitation. The
amount of fluorescence can be measured in real-time and used
for quantification purposes.
With Scorpion primers, no additional probe is added, but 1
of the 2 PCR primers is specially configured and labeled with
a fluorescent dye and a corresponding quencher, as with the
aforementioned fluorescent probe. The primer is arranged to
form a hairpin loop structure, enabling the quencher to be in
close proximity with the reporter dye. During the extension
phase, the Scorpion primer unfolds to hybridize with the new
daughter strand, and the hairpin loop within the Scorpion
primer disassociates, separating the fluorescent reporter dye
from the quencher.
SYBR Green quantification is a completely different
approach. It uses a dye (SYBR Green) that binds to
double-stranded DNA and quantifies the amount of
double-stranded DNA produced. SYBR Green will quantify
both specific and nonspecific PCR products. In contrast,
Scorpion primers and fluorogenic probes will only hybridize
with specific PCR products downstream in the PCR, resulting
in higher specificity and lower background noise than SYBR
Green assays.
Using real-time PCR equipment, such as the ABI Prism
®
7700 or 7900 (Applied Biosystems, Foster City, CA),
LightCycler
®
(Roche Diagnostics, Indianapolis, IN),
I-Cycler
®
(Bio-Rad, Richmond, CA), or the Mx3000P™
(Stratagene, La Jolla, CA), fluorescence can be monitored and
analyzed through this process via computer interface. As
previously mentioned, fluorescent signals will eventually
plateau after a number of PCR cycles; therefore, end point
fluorescence is not suitable for quantification. However, by
plotting the fluorescence vs cycle number (Figure 2), and
assigning a threshold within the exponential phase of the
fluorescence amplification, the corresponding cycle number
at which fluorescence crosses this threshold will be inversely
proportional to the amount of DNA target in the sample.
Therefore, this cycle number, referred to as cycle threshold
(C
T
), can be used to quantify target DNA amounts.
With the resulting quantitative data, real-time PCR
techniques can be used to determine the percentage of a GM
DNA sequence in a sample. Values can be determined for a
GM target DNA and compared to total target species DNA,
which is determined by use of a species-specific, preferably
single-copy reference gene (an endogenous control gene).
Quantification units will, therefore, be the amount of the GM
sequence expressed as a percentage of the reference DNA
(example, 1% RR soy DNA vs total soy DNA). PCR for GM
and reference sequences can be performed either in separate
tubes, or in the same reaction using reporter dyes that
fluoresce at different wavelengths (multiplex). With multiplex
reactions, individual component concentrations (i.e., primers,
probes, deoxynucleotides) have to be carefully titrated and
validated to protect against competitive effects between the
2 target amplifications. Multiplex PCR also introduces some
errors into the final results because of limitations in
fluorescence multicomponent analysis (32).
Quantification can be achieved with 2 approaches: One
approach is to construct a standard curve and interpolate
values into the standard curve to extract data. The second
approach is the comparative C
T
method (32).
There are 2 common ways of constructing a standard curve
for quantitative analysis. One common way is the use of a
serial dilution of DNA of known concentration and GM
content. GM DNA from certified reference materials,
plasmids, or hybrid amplicons (33) can be used for the
standard curve. In any case, the target DNA copy number must
be precisely quantified beforehand. Examples of DNA
concentration quantification techniques are fluorometry (i.e.,
using Hoechst or Picogreen
®
dyes), melting curve analysis
142 LIPP ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 88, NO. 1, 2005
(i.e., using SYBR Green), or spectrophotometric analysis.
These were discussed previously (see above).
Plant GM DNA concentration can be converted to copy
number equivalents by using conversion factors as reported
by Arumuganathan and Earle (8) or by referring to the Web
site “C Value” (34). However, genome sizes may vary
depending on the variety within the species, so varietal
information, if available, should be considered. If using
certified reference materials (CRMs) of a certain percent GM,
that percentage must be taken into consideration when
calculating GM copy number equivalents as well.
With this technique, reference and GM C
T
values from the
sample can be interpolated against the corresponding standard
curve to determine the copy number equivalents: Dividing the
calculated GM copy numbers by the reference copy numbers
will yield the result in percent GM target DNA versus total
species DNA. Factors such as ploidy and copy numbers per
genome of GM and reference gene have to be taken into
account in order to relate DNA fraction to mass fraction.
A second common approach in construction of standard
curves is the use of CRMs of different percent GM, measuring
the difference in C
T
values from the GM and reference target
genes for each standard. This approach is referred as the D C
T
standard curve method. D C
T
values will be inversely
proportional to percent GM, and the D C
T
of the sample can be
interpolated against the D C
T
of the standards to approximate
the sample’s percent GM. Certified reference standards of
some GM varieties are commercially available at 0, 0.1, 0.5, 1,
2, and 5% concentrations.
An alternative to standard curves is the comparative C
T
method. This method is also referred to as the DD C
T
method (32). In this method, the amount of target, normalized
to an endogenous reference and relative to a calibrator, is
given by the following formula:
Amount of target DNA = 2
– DD
C
T
in which
DDC
T
= DC
T
, sample – DC
T
,calibrator
For this calculation to be valid, the efficiency of the
reference and target amplifications must be equivalent.
Each of the 3 quantitative approaches discussed has its own
advantages and limitations. The serially diluted standard
curve method requires the smallest amount of validation and
optimization. It is also less sensitive to variations in PCR
efficiency and more rugged in its application with various
sample matrixes. In contrast, it may be affected by dilution
errors and may be biased against the particular standard used.
The DC
T
standard curve method uses more standards, and
therefore reduces bias, but is more sensitive to variations in
PCR efficiency and requires a greater amount of validation. It
is very useful when DNA is consistently analyzed from the
same sample matrix. Of all the quantification methods, the
comparative C
T
method requires the greatest amount of
validation and optimization and is most prone to errors arising
from changes in PCR efficiency. Nevertheless, in the event of
a successful validation, this could be the most cost-effective
and highest-throughput method, as it eliminates the need for a
standard curve and also eliminates the error arising from
standard curve dilutions. However, the inherent problems
associated with the comparative C
T
method usually restrict its
application to 1 or very few sample matrixes and extraction
methods, because it is unlikely that PCR efficiencies are
constant.
Method Validation
There is sometimes confusion about the terms used to
describe interlaboratory studies for the purpose of method
performance, material performance, and laboratory
performance studies. The reader is reminded that these are
3 distinct activities. The International Union of Pure and
Applied Chemistry (IUPAC) classifies (35) interlaboratory
studies into the following 3 categories: (1) Method
performance.—Determines the bias and precision of an
analytical method. (2) Material performance.—Assigns a
value and an uncertainty (or reliability) to a characteristic
(usually concentration) of a material. (3) Laboratory
performance.—Permits the evaluation of each participant
against preset criteria or criteria estimated from the study
itself.
Although the procedures regarding statistical data
evaluation from these 3 types of interlaboratory studies may
be identical, the use and interpretation of the resulting
statistical estimates will be determined by the primary purpose
of the study.
The validation of methods consists of 2 phases. The first is
an in-house validation of all of the parameters except
LIPP ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 88, NO. 1, 2005 143
Figure 2. Amplification plot from Applied Biosystems
ABI Prism 7700 showing serial dilution of sample. Each
dilution was run in duplicate. The y-axis represents
normalized, logarithmically converted fluorescence
intensity data. The x-axis represents the cycle number.
(A) Set threshold across samples’ exponential phase;
(B) sample with highest concentration has lowest C
T
(point at which amplification curve crosses threshold);
(C) sample with lowest concentration has highest C
T
;
(D) plateau phase; all samples have very similar
fluorescence; quantification is not possible here.
reproducibility. The second is a collaborative trial, the main
outcome of which is a measure of the repeatability and
reproducibility in order to estimate the transferability of
methods between laboratories. It is our experience that a
small-scale collaborative trial should be performed to test the
ruggedness of a particular method before the expense of
organizing a large-scale trial is incurred. In case any
improvement of the method or the method description is
needed, only limited costs are incurred through the pretrial,
whereas a failure of a full interlaboratory method validation
because of an ambiguous method description is a very costly
failure. Implementation of an already validated method in a
laboratory needs to include the confirmation that the
implemented method performs as well under local conditions
as it did in the interlaboratory method validation.
A method must be validated by using the protocols and
reaction conditions under which it will be performed. For
example, the protocol should not be changed using higher or
lower numbers of amplification cycles. These and other
changes, as well as the application to a different matrix, are
likely to affect method characteristics such as the specificity or
sensitivity. PCRs may have a tendency of unspecific
background amplification at low rates, which can be tolerated
if the specified conditions and number of cycles do not result
in analytical artifacts as demonstrated by validation. However,
they can be expected to result in artifacts if operated with more
cycles and/or under nonoptimal conditions.
This document deals primarily with the use of
interlaboratory studies for the assessment of method
performance. However, before use at a particular location, any
method must be subjected to an in-house validation
procedure.
The results of a determination are often expressed in terms
of percent of a sample that contains a particular
biotechnology-derived sequence. In a quantitative test, this
measurement actually involves 2 PCR-based determinations:
that of the primary analyte (e.g., an inserted gene sequence)
and that of the endogenous or comparator sequence (e.g., an
endogenous maize gene). Each of these determinations has its
own uncertainties, and the 2 are likely to have different
measurement characteristics. In most applications, the
primary analyte will be present at low concentrations and the
comparator will be present at concentrations 10–1000 times
higher. Thus, it is important that both measurements are
properly validated. In cases in which the measurement is
expressed directly as a percentage (as in the use of DC
T
), these
factors must be considered when validating the method.
Validation Parameters
The method performance study or method validation
establishes the performance characteristic for a specific
method application, i.e., a specific analytical procedure for a
well-defined scope. In the following text, the spirit of the most
relevant terms has been captured by a simple definition for use
in this document (Table 1). For a more detailed discussion and
explanation of the definitions, refer to the Procedural Manual
of Codex Alimentarius (36).
The concepts of a LOD, limit of quantitation (LOQ), and
ROQ are not yet explicitly defined by Codex Alimentarius.
These parameters can be considered to pertain to the
applicability of the method. However, as they are useful for
some applications, definitions are also given in Table 1.
The LOD is the amount of analyte at which the analytical
method detects the presence of the analyte at least 95% of the
time (<5% false-negative results). This and the false-positive
rate are the only parameters required for a qualitative method,
other than specificity.
Determination of an LOD is not necessarily needed to
establish the validity of a method for a given qualitative
application if it can be shown that the false-negative rate is
<1% in the range of the application. For example, the precise
determination of the LOD to be 1 ng/kg does not add much
value when the scope of the method validation extends only
for concentrations ranging in g/kg. Similar considerations
apply for the LOQ. However, the range over which a method
is applicable (ROQ) should always be established and
included in the validation study.
Many quantitative methods are applied and have their most
linear response near the LOQ. In such cases, it is important to
know the LOD and LOQ in order to determine whether an
observed result is, in fact, significantly different from the
background, and can be satisfactorily quantified. In a
quantitative method, it is common practice to assume that
LOD is the signal strength of a blank increased by 3 times the
standard deviation of the blank. However, this method gives at
best an estimate, relies on normal Gaussian distribution of the
blank measurements around zero, and may give a lower value
than the actual LOD. Its use is not valid in methods such as
quantitative PCR, in which the distribution of measurement
values for blanks is typically truncated at zero and thus is not
normally distributed. Thus, the LOD needs to be
experimentally determined unless the targeted concentrations
are well above the LOD, and the LOD, therefore, becomes
irrelevant.
For a quantitative method, it is also important to know
whether the LOQ for a particular matrix is close to the values
to be measured. Again using a traditional approach, the LOQ
has to be expressed as the signal strength of a blank increased
by 6–10 times the standard deviation of the blank. These data
must be experimentally determined, as discussed previously,
unless it is known from other sources that the measured values
range so high above the LOQ that this information becomes
irrelevant. However, this method to determine the LOQ leads
only to an estimate of the LOQ that may be an artificially high
or low approximation.
In practice, 2 procedures have been used to determine the
LOQ. The first approach is to assay a number of negative
samples that have been supplemented (spiked) with known
amounts of analyte. The LOQ is then the level at which the
variability of the result and percent recovery of the analyte
meet certain pre-set criteria. For small molecules, these
criteria have typically been a coefficient of variation of 20 and
70–110% recovery (37). DNA recovery, however, may be
difficult from some matrixes, e.g., starches or ketchup, and
144 LIPP ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 88, NO. 1, 2005
lower recovery efficiencies may have to be accepted. When
recovery efficiencies are low, this must be stated in the
validation data and in the analytical report. The second, more
complete approach is to test the method using a number of
samples that contain known amounts of the GM material. This
is more complicated, as it requires access to significant
quantities of reference materials that contain a known range of
concentrations of the GM event of interest. Procedures for
assessing LOD and LOQ during the validation of qualitative
and quantitative PCR methods are described in the following
text.
Parameters Common to Qualitative and Quantitative
Methods
A number of the parameters measured in a validation are
common to both qualitative and quantitative methods; some
are more applicable to the quantitative methods. Those
common to both types of methods will be discussed first.
(a) Specificity.—Specificity is the starting point for a
method and needs to be considered during primer design.
Primers should be checked against the known sequence of the
event insert and pertinent databases for possible matches.
Experimental confirmation of the specificity must be
performed. The following suggests a reasonable approach; the
experiments should be performed during development or
in-house validation of an assay before a larger validation is
conducted.
For event-specific assays: (1) analyze a total of at least
10 sources, including nontarget GM events and any non-GM
plants that may commonly be found as contaminants in the
commodity; (2) test 1 sample from each source (total of at
least 10 DNA samples); (3) analyze 2 replicates for each DNA
sample.
Test results shall clearly indicate that no significant
positive signal is observed.
For assays on plant endogenous genes: (1) analyze a total
of at least 10 different plant samples that comprise different
varieties of the same plant species as well as other plant
species important for food production (such as wheat, rice,
maize, potato, and soybean) and that may commonly be found
as contaminants in the commodity; (2) test 1 sample from each
source (total of at least 10 DNA samples); (3) analyze
2 replicates for each DNA sample.
Test results shall clearly indicate that no significant
positive signal is observed.
(b) Applicability.—It is not feasible to provide reference
materials for every one of the thousands of food matrixes
available, so the use of a seed-derived or other such matrix
reference will usually be necessary. The use of the method in a
new matrix will need to be validated at a minimum via
LIPP ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 88, NO. 1, 2005 145
Table 1. Parameters for method validation
Accuracy The closeness of agreement between the reported result and the accepted
reference value. Accuracy describes how close the measured value is to the
actual value of the known reference sample.
Precision The closeness of agreement between independent test results obtained under
stipulated conditions. Precision describes how well the results agree between repeated
analyses of the same material. Less precision is reflected by a larger standard
deviation of the combined results than the individual results.
Sensitivity Change in the response divided by the corresponding change in the concentration of a
standard (calibration) curve, i.e., the slope of the analytical calibration curve.
Specificity The ability of a method to respond exclusively to the characteristic or analyte.
Specificity describes how often the analyte is not detected.
Ruggedness (robustness) The ability of a method to resist changes in results when subjected to minor changes in
environmental and procedural variables, laboratories, personnel, etc.
Applicability The analytes, matrixes, and concentrations for which a method of analysis
is validated that may be used satisfactorily.
Repeatability Precision under repeatability conditions. These are conditions in which
independent test results are obtained with the same method on identical
test items in the same laboratory by the same operator using the same
equipment within short intervals of time.
Reproducibility Precision under reproducibility conditions. These are conditions in which test
results are obtained with the same method on identical test items in different
laboratories with different equipment and operators.
Limit of detection (LOD) The lowest amount of analyte in a sample that can be detected with
suitable confidence but not necessarily quantified as an exact value
as determined by method validation.
Limit of quantitation (LOQ) The lowest amount of analyte in a sample that can be quantified with suitable accuracy
and precision as determined by method validation.
Range of quantification (ROQ) The range within which the amount of analyte in a sample can be quantified with
suitable accuracy as determined by method validation.
in-house validation, which may be done using spike and
recovery experiments, and the reference material used should
be described on the report to the customer.
Validation of a Qualitative PCR Method
The following section concentrates on issues surrounding
validation of qualitative methods. However, many of the
principles also apply to quantitative methods, especially when
they are used in a qualitative way. The sensitivity of a
qualitative method must be shown to be such that it can
reliably detect 1 positive particle (e.g., seed) in a pool and does
not give rise to a significant number of false positives. A
concept of using false-positive and -negative rates to describe
the accuracy and precision of a qualitative assay has been
developed for microbial assays (38). This concept can be
applied to qualitative PCR assays. A critical issue in the
validation of this type of method is the availability of test
materials that are known to be positive and negative. The
provision of negative reference materials is particularly
important and critical in the case of a qualitative method. Any
impurities present must be only at levels so low that they
become negligible. Development of reference materials is
covered later in this article.
(a) False-positive and -negative results.—By their very
nature, qualitative tests result only in yes/no answers. The
measures of precision and accuracy are the frequencies of
false-negative and/or false-positive results. False-negative
results indicate the absence of a given analyte when in fact the
analyte is present in the sample; false-positive results indicate
the presence of an analyte that is not present in the sample.
Because of the inherent nature of the analytical technique, an
increase in false-negative results will be observed when the
amount of analyte approaches the LOD of the method. Like the
LOD for quantitative methods, the LOD for a qualitative method
can be defined as the concentration at which a positive sample
yields a positive result at least 95% of the time. This results in a
rate of false-negative results of 5% or less. During validation of a
qualitative PCR assay, it is also important to determine the
frequency of false-positive results (a positive result obtained
using a sample that is known to be negative). Both false-positive
and -negative results are expressed as rates (38).
(1) False-positive rate.—This is the probability that a
known negative test sample has been classified as positive by
the method. The false-positive rate is the number of
misclassified known negatives divided by the total number of
negative test samples (misclassified positives plus the number
of correctly classified known negatives) obtained with the
method.
For convenience, this rate can be expressed as percentage:
% false - positive results =
number of misclassified known negative samples
total number of negative test results [incl. misclassified]
100´
(2) False-negative rate.—This is the probability that a
known positive test sample has been classified as negative by
the method. The false-negative rate is the number of
misclassified known positives divided by the total number of
positive test samples (misclassified positives plus the number
of correctly classified known positives) obtained with the
method.
For convenience this rate can be expressed as percentage:
false - negative results, % =
number of misclassified known positive samples
total number of positive test results [incl. misclassified]
´ 100
In order to demonstrate the false-negative rate for a
qualitative assay, a series of samples (e.g., grain/seed pools)
with a constant, known concentration of positive material in a
pool of negative material (e.g., 1 positive kernel in
199 conventional maize kernels) has to be analyzed and the
results evaluated. The concept of confidence intervals and
statistical uncertainty needs to be applied to the risk of
false-positive and/or false-negative results as well. The
desired level of confidence determines the size and number of
pools that need to be tested. For example, 100 positive test
results obtained from 100 independent measurements on truly
positive samples lead to the conclusion that the level of
false-negative results is below 4.5% at a confidence level of
99% for the tested concentration of positive kernels
(expressed as the number of positive kernels in a pool of
negative kernels).
(b) Ruggedness.—As with any validated method,
reasonable efforts must be made to demonstrate the
ruggedness of the assay. This involves careful optimization
and investigation of the impact of small modifications that
could occur for technical reasons.
(c) Acceptance criteria and interpretation of results.—A
validated method includes criteria from which an observed
measurement result can be accepted as valid. It is important to
follow these criteria and to observe the rules for data
interpretation. It is, therefore, important to ensure that the
result of the positive DNA target control is positive. Similarly,
the amplification reagent control (a control containing all the
reagents, except extracted test sample template DNA, which is
replaced by a corresponding volume of nucleic acid-free
water or buffer) must be negative.
In addition to these controls, it is desirable to conduct a
parallel reaction on the same DNA sample using a primer set
that detects an endogenous single copy sequence. This
reaction is performed on every DNA sample and can either be
in the same reaction (multiplexed) or as a separate reaction. In
the case of multiplexed reactions, it is important for the
endogenous reaction not to out-compete the event-specific
reaction for reagents, as the endogenous sequence is likely to
be present at a rate up to 1000-fold the amount of the target
sequence. The control reaction with the endogenous sequence
can give an indication of the quality of the DNA as a template
for the PCR.
The fact that qualitative PCRs are typically performed in
duplicate introduces a further complication: The duplicates
may not agree. It is common practice to repeat PCRs once on
DNA samples that are rejected because of conflicting
(indeterminate) results. A repeated indeterminate result
indicates that the analyte cannot be reliably detected (Table 2)
and that the assay is operating below the LOD as, by
definition, a 95% or better detection rate would be achieved at
146 LIPP ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 88, NO. 1, 2005
the LOD. The sample is, therefore, scored negative. Similar
criteria apply if more replicates are performed on each DNA
sample.
Validation of a Quantitative PCR Method
A harmonized ISO/IUPAC/AOAC protocol was
developed for chemical analytical methods. This defines the
procedures necessary to validate a method (39). All the
principles and rules of the harmonized protocol are applicable
to quantitative PCR methods. The parameters involved in
validation of the performance of a quantitative PCR assay
include LOD, LOQ, ROQ, accuracy, precision, sensitivity,
and ruggedness. Other important factors are acceptance
criteria and interpretation of results, and the units in which the
results are expressed.
A quantitative PCR assay typically consists of 2 assays:
one determines the amount of DNA specific for the GM
product; the other is specific for the amount of plant-specific
DNA. Each of these assays is considered separately, because
they are independent analytical procedures. Thus, all
parameters listed below, including specificity and sensitivity,
have to be assessed individually for each of the assays
involved. A method validation for the whole assay cannot be
appropriately performed unless both assays are validated
individually.
(a) Limits of detection and quantitation (LOD and
LOQ).—If validation of the quantitative PCR assay shows
that the assay can measure GM plant DNA at the required
concentration with acceptable trueness and precision, then it is
often not necessary to determine the LOD and LOQ, as the
method is only being applied above the range where these are
relevant. However, if the method is being used at
concentrations close to the LOD and LOQ (typically
0.1–0.5%), then the assessment of the LOD and LOQ will
become part of the validation procedure.
(b) Range of quantification (ROQ).—The ROQ of the
method defines the concentration range over which the
analyte will be determined. Typically, the range for a GM
product will run from a tenth of a percent up to a few percent;
the endogenous control range will be close to 100%, unless
the testing of complex mixtures is envisioned. This desired
concentration range defines the standard curves, and a
sufficient number of standards must be used to adequately
define the relationship between concentration and response.
The relationship between concentration and response should
be demonstrated to be continuous and reproducible and
should be linear after suitable transformation.
The quantitative method is designed to operate in the range
of 0.1–100% (DNA %, w/w). However, it is common to
validate a method for a range of concentrations that is relevant
to the ROQ of the application. If a method is validated for a
given range of values, the range may not be extended without
validation. For certain applications (e.g., seed or grain
analysis), the use of genomic DNA for the preparation of the
standard curve (see discussion on the use of plasmid DNA
below) may be considered. Although it is easy to establish a
nominal 100% standard (limited only by the purity of the
materials used), it is difficult to reliably produce standard
solutions <0.1%. This is due to the uncertainties involved in
measuring small volumes and the error propagation if serial
dilution steps are applied. Additionally, the number of target
sites (DNA sequence to be amplified) becomes so small that
stochastic errors will begin to dominate and no reliable
analysis is possible (9, 40). If genomic DNA is chosen to be
used as calibrator, it is important that this calibrator be traced
back (in its metrological meaning) to a reference of highest
metrological order, e.g., a CRM. The range will be established
by confirming that the PCR procedure provides an acceptable
degree of linearity and accuracy when applied to samples
containing amounts of analyte within or at the extremes of the
specified range of the procedure.
(c) Accuracy and precision.—The accuracy of a method
should be compared to known values derived from reference
materials. This may be a challenge in this field because of the
limited availability of such materials. However, the accuracy
must be compared to the best available reference material.
Precision will be determined in the usual way from
single-laboratory (repeatability) and multilaboratory
(reproducibility) studies.
(d) Sensitivity.—A linear relationship of the C
T
as a
function of the logarithm of the concentration of the target
should be obtained across the range of the method. The
correlation coefficient, y-intercept, slope of the regression
line, and percent of residual should be reported. The percent
of residual for each of the calibrators should preferably be
£30%.
In order to obtain a standard curve for event-specific
quantitative assays, standard DNA mixtures can be prepared
by combining purified genomic DNA from GM and non-GM
plant material such as seed or leaves. The content of GM plant
DNA in the mixtures might be 100, 50, 10, 5, 1, 0.5, 0.1, and
0% or as appropriate for a smaller concentration range. Three
replicates should be analyzed for each point on the standard
curve.
For quantitative assays on plant endogenous genes,
standard DNA mixtures can be prepared by combining
purified genomic DNA from the target plant species and that
of a nontarget plant species. For example, for validation of a
maize ADH1 quantitative assay, the target plant species is
maize and the nontarget plant species could be soybean or
another species. The content of DNA of the target plant
species in the mixtures is typically 100, 90, 80, etc., and 0% or
LIPP ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 88, NO. 1, 2005 147
Table 2. Criteria for scoring duplicate qualitative PCR
analyses
Lane 1 Lane 2 Scoring of test
+ + Positive
– + Repeat/indeterminate
+ – Repeat/indeterminate
– – Negative
as appropriate. Three replicates should be analyzed for each
point on the standard curve. When the DC
T
method is used, it
will be the responsibility of the analyst to ensure that the
overall amount of DNA is within the range for which the assay
was validated.
(e) Ruggedness.—The evaluation of ruggedness
(robustness) demonstrates the reliability of a method with
respect to inadvertent variations in assay parameters.
Variations that may be included are reaction volumes (e.g., 25
vs 30 mL), annealing temperature (e.g., plus and minus 1°C),
and/or other relevant variations. The experiments need to be
performed at least in triplicate, and the recovery needs to be
calculated. The response of an assay with respect to these
small changes should not deviate more than ±30% from the
response obtained under the original conditions.
(f) Acceptance criteria and interpretation of results.—A
validated method also includes criteria on which the observed
measurement result can be accepted as valid. It is important to
follow these criteria and to observe the rules for data
interpretation. If a case calls for the deviation from said
criteria and rules, a new method validation study would be
needed to demonstrate the validity of the new rules and
procedures.
At a minimum, the following acceptance criteria are
common to all quantitative PCR methods and applicable to
each PCR run: (1) the result of the positive DNA target control
with, for example 1% GM DNA, the mean of the replicates
deviates <3 standard deviations from the assigned value. A
target DNA control is defined as reference DNA or DNA
extracted from a certified reference material or known
positive sample representative of the sequence or organism
under study. The control is intended to demonstrate the result
of analyses of test samples containing the target sequence.
(2) The amplification reagent control is negative. The
amplification control is defined as a control containing all the
reagents, except extracted test sample template DNA. Instead
of the template DNA, a corresponding volume of nucleic
acid-free water is added to the reaction. (3) The percent of
residual for each of the standards should be <30%.
To accept the result of an unknown sample, the relative
standard deviation of the sample replicates should be <30%.
Reference Materials
General Considerations
Reference materials play a number of roles in
development, validation, and troubleshooting of PCR-based
diagnostics, as well as in the routine conduct of such assays. In
the context of assay validation, positive reference materials
are used to establish the accuracy, precision, sensitivity, LOD,
and false-negative rate in quantitative assays. Negative
reference materials are very important in determining
false-positive rates and specificity.
Reference materials can be of several levels of metrologic
quality: (1) A certified or standard reference material (CRM
or SRM
®
) is accompanied by a specific certificate. This
certificate states that one or more of the property values of the
reference material is certified by a procedure that establishes
the value’s traceability to an accurate realization of the unit in
which the property value is expressed; in addition, the
certificate states a level of confidence of uncertainty (41).
Such reference materials are usually issued by National
Metrology Institutes such as the Institute of Reference
Materials and Measurement (IRMM) of the Joint Research
Center of the European Union and the National Institute of
Standards and Technology (NIST) in the United States. (2) A
reference material (RM) is a reference material or substance
one or more of whose properties are sufficiently
homogeneous and well established to be used for the
calibration of an apparatus, the assessment of a measurement
method, or for assigning values to materials (ISO definition).
(3) A working standard (WS) is a secondary standard in
regular use. This working standard is equivalent to RM if it is
quantified/characterized by comparison with the CRM/SRM.
Quality Standards for Reference Materials
A certificate of analysis will accompany each reference
material. The certificate will describe the characteristics of the
material, both as to the presence of the target material and the
absence of other possibly interfering materials. In addition, a
reference material may even be restricted as to the method for
which it can be used or is validated.
A certificate of analysis for a GM reference material will
address the following factors: (1) target event or sequence;
(2) adventitious presence of other events; (3) strength and
purity; (4) genetic background.
Target event or sequence.—The first consideration is
whether the material is a reference for a particular event. In
most detection processes, this will be the case. However, it is
foreseeable that many reference materials or working
standards used in screening methods will be for a certain
sequence, such as the 35S promoter. Event reference materials
can be used for this purpose, providing it is known how many
copies of the sequence are present in the event.
Adventitious presence of other events.—An important
factor that will influence the true concentration of the samples
prepared for the validation experiments is, of course, the level
of impurity in the reference materials. The starting material
used for the preparation of the reference materials needs to be
characterized for purity with respect to the desired analyte. To
do so, a representative subsample must be analyzed for the
absence or presence of the analyte in both negative and
positive pools. The sample size and measurement error will
determine the confidence level of the results.
Positive reference materials are considered to be those that
contain the event or sequence of interest (e.g., MON810, T25,
35S, NOS). Negative reference materials are materials that do
not contain the stated event or sequence, at least not at a
detectable level. A negative reference material will not have a
value of 0% assigned. The known presence or absence of
other events that may interfere with the analysis should be
stated. As with all analytical blanks or negative and positive
reference materials, an uncertainty about the assigned values
needs to be expressed. Ideally, this uncertainty includes 0%
148 LIPP ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 88, NO. 1, 2005
for the negative materials. For example, regarding Roundup
Ready
®
soybeans, currently available negative reference
materials (Joint Research Center, IRMM, Geel, Belgium) are
assigned to contain <0.03% Roundup Ready soybeans (42).
This value reflects the uncertainty of the methods and
sampling procedures used for the certification exercise of the
materials. This is particularly the case for any material
prepared from grain, as it is not feasible to test every grain
kernel before using it in a reference material. Thus, a reference
material certificate will typically state that it contains less than
a particular amount (e.g., <0.03%) of other events, which
should be listed explicitly, with a known certainty (e.g., 99%).
Similar restrictions or considerations are applicable for all
reference materials used in analytical sciences.
Strength and purity.—Strength and purity are
2 measurements that normally are determined for a reference
material. However, how the strength and purity of a GM
reference material is defined is not yet clear.
(1) Strength.—In some cases, it can be clearly stated that a
sample is of 100% strength. For example, if DNA is prepared
from a single homozygous plant, this DNA sample can be
considered to have a strength of 100%, if a procedure has been
used that produces essentially uncontaminated material. If the
DNA was produced from a heterozygous plant, then it can be
argued that the strength is 50%. This, however, ignores the
effect of chloroplast DNA, which may dilute the sample to
below 100 or 50%, depending on the definition. Special
considerations may apply for hemizygous maize grain
material in which the strength may depend on whether the
applicable gene was introduced via the male or female parent,
resulting from variations in zygosity of the tissues in the
kernel.
(2) Purity.—Classically, purity is the percentage of the
sample that consists of the material. For example, if a sample
contains >99% DNA, then its purity would be >99%.
Strength and purity also may depend on loss of particulate
material during any sieving process, unless such material is
reground and added back to the sample. These factors may be
expressed in alternative terms, such as the mass fraction (w/w)
of the stated material.
Genetic background.—Many reference materials are
prepared by mixing GM with non-GM materials. Comparison
of the 2 sources as to whether they consist of near-isogenic
lines may be included, if desired.
Choice of Reference Material
There are a number of matrixes that can be used to develop
reference materials or working standards for methods of
detection of GM products. Each has its own advantages and
disadvantages for particular purposes. The 3 reference
materials discussed here include: (1) grain or seed or
seed-derived powders; (2) GM DNA of plant origin;
(3) plasmid DNA or amplicons containing the target sequence.
For purposes of this discussion, the term “seed” will
indicate both grain and seed. Seed is the preferred reference
material for testing of grain and oilseed commodities. When
dealing with hybrid crops, seed is preferred, as grain will be
segregating for the trait. Both hybrid and inbred seed may be
acceptable, provided one is cognizant of potential bias that can
be introduced by both approaches. In the case of hybrid seed,
the presence of a triploid endosperm in maize and other
cereals will make a variable contribution to the number of
PCR targets, depending on whether the GM parental line was
the male or female parent. Use of inbred seed as a positive
reference standard for a hybrid crop (e.g., maize) will result in
underestimation of GM content that is present as hemizygous
material if not corrected appropriately.
In addition to these reference materials, there may be
reference materials produced that consist of other plant
organs, or refined and finished products. Examples of such
materials may be leaves, starch, and lecithin. It is beyond the
scope of this paper to discuss these types of materials.
(a) Grain or seed or seed-derived powders.—Given the
present commercial and trade situation and the present target
crops for genetic modification, seed is the most commonly
used reference material. Unfortunately, seed may be
heterogeneous and the uncertainty factor caused by sampling
is significant at all levels. This is particularly true when the
reference materials contain <1% of the event in question,
unless very large samples are supplied. For example, testing
for the presence of any GM seed in a sample of maize or
soybeans at a level of 0.1% would require testing
10 000 kernels (or roughly 3 kg) to achieve a 99% confidence
in the analytical result. Even in this case, there will be
uncertainty in the actual number of positive kernels in any
sample. To use so much material for each reference point is
not practical, and thus, seed powders are used widely.
However, seed mixtures are particularly useful for
validation of qualitative methods. These validations can be
made using pools of seeds. When preparing the pooled
samples for a validation study, the following errors must be
taken into consideration: (1) the positive kernel(s) may
contain a small amount of negative kernels with respect to the
desired analyte; (2) the negative kernels may contain a small
amount of positive kernels.
Consequently, the pools that are prepared for use in the
validation study could be impacted in the following ways:
(1) the number of spiked kernels is smaller than calculated,
i.e., 1 or more of the spiked positive kernels was actually
negative; (2) the number of spiked kernels is larger than
calculated, i.e., the negative kernels contained 1 or more
positive kernels.
Table 3 summarizes the probability that a given negative
seed bulk contains at least 1 positive (GM trait) kernel. Table 3
shows that, for large seed pools and high impurity levels in the
negative material, the chance of observing at least 1 positive
kernel in the pool is essentially 100%. Consequently, when
operating with large pools, the purity of the negative materials
needs to be extremely high. In order to assess all variations of
the true value for a spiked pool, the impurity of the positive
material (the spike) must therefore be taken into
consideration.
Table 4 gives examples of the effect of purity in the
situation in which seed pools are spiked with GM material to
LIPP ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 88, NO. 1, 2005 149
achieve a target of 1% level of GM seeds in non-GM
seeds (6). It illustrates the high level of purity of the negative
material that is needed if large seed pools with a small number
of kernels carrying the GM trait need to be prepared, as may
be the case when very sensitive qualitative methods are tested.
The purity of the negative material is much more important
than the purity of the positive material because much more
negative material than positive material is used to build the
pools.
For example, pools of 300 kernels targeted to have 1% GM
kernels are composed of 3 positive and 297 negative kernels,
pools of 600 kernels are composed of 6 positive and
594 negative, and pools of 1000 kernels contain 10 positive
kernels and 990 negative kernels. In constructing the table, a
variation of approximately 20% was assumed. Thus, the table
shows the frequency of pools of 300 kernels containing at
least 2, but at most 4 positive kernels. (For a pool size of 600,
5–7 positive and for a pool size of 1000, 8–12 positive kernels
were tolerated.)
Seed powders are a compromise that best mimics the
genuine test material and will be processed in a manner similar
to the test material, while avoiding the need to prepare large
amounts of mixtures of positive seeds in pure negative seeds.
Thus, matrix effects and extraction-related artifacts should be
similar between control and test samples. They can be
prepared with a known particle size and can be tested for
homogeneity and accuracy with respect to the expected value.
Seed powders do have considerable preparation and storage
costs and must be tested for stability. Some seed-based
reference materials of this type are currently commercially
available (IRMM, Geel, Belgium).
Preparation of reference materials from seed is
complicated by the particulate nature of such material, which
necessitates statistical considerations as described above. The
starting material can be a sample that contains a certain
number of negative kernels and a certain number of positive
kernels. The purity of the sample of negative kernels is subject
to limitations because of the inability to completely test the
sample; the estimate of purity is based on a subsample of a
larger sample from which the negative material also is
derived. Thus, it is never possible to be 100% sure that a
supposed negative reference material is negative. It can only
be established that the amount of positive material is less than
the LOD of the method used to establish that the sample is
negative and will be additionally restricted by statistical
150 LIPP ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 88, NO. 1, 2005
Table 3. Probability that a given negative seed bulk
contains at least 1 GM trait seed
Conventional reference material impurity level
Seed builk size 0.01% 0.10% 1.00%
100 1 10 63
200 2 18 87
250 2 22 92
300 3 26 95
500 5 39 99
600 6 45 100
1000 10 63 100
Table 4. Number of pools (out of 100) that contain the given number of GM seeds when pools are prepared at a 1.0%
target concentration
Pool size
300 600 1000
GM trait, reference material
impurity, %
Non-GM, reference material
impurity, % 2–4 GM seeds 5–7 GM seeds 8–12 GM seeds
4.0 0.01 100
a
(98, 100) 98 (95, 100) 100 (98, 100)
4.0 0.10 97 (93, 99) 89 (84, 94) 94 (90, 98)
4.0 1.00 23 (16, 30) 3 (1, 6) 1 (0, 2)
2.0 0.01 100 (99, 100) 100 (98, 100) 100 (99, 100)
2.0 0.10 97 (93, 99) 89 (83, 94) 93 (89, 97)
2.0 1.00 21 (15, 29) 2 (0, 5) 0 (0, 2)
1.0 0.01 100 (99, 100) 100 (99, 100) 100 (100, 100)
1.0 0.10 97 (93, 99) 89 (83, 94) 93 (88, 97)
1.0 1.00 21 (14, 28) 2 (0, 5) 0 (0, 2)
0.5 0.01 100 (100, 100) 100 (99, 100) 100 (100, 100)
0.5 0.10 97 (93, 99) 88 (83, 93) 93 (88, 96)
0.5 1.00 20 (14, 27) 2 (0, 4) 0 (0, 1)
a
Median and 5th and 95th percentiles (in parentheses) from distribution of number of seed bulks with specified number of GM trait seeds.
sampling considerations, most notably by the size of
subsample(s) analyzed.
A further complication in comparing a sample to a
reference material is present when material is analyzed for the
presence of maize or other monocotyledonous material. The
various tissues of the maize kernel are of maternal origin, or
receive different contributions from the male and female
parent. This will give rise to a different result when the
proportion of GM DNA is measured in a sample, depending
on whether the GM component came from the male or female
parent. This is particularly relevant when testing for
adventitious presence of small amounts of GM material, in
which the adventitious material may have arisen by incoming
pollen, or by seed mixing. The analyst should be aware that
these 2 sources would give rise to different results, even
though the analyst can do little or nothing to solve the
problem. In addition, processed samples arising from starchy
endosperm, the germ, or the pericarp can be subject to similar
issues. The issue of parental contribution to the DNA content
of different tissue must be addressed when preparing
reference material, and ideally, the type of material (inbred,
hybrid material with description of the cross direction) should
be included on the certificate of analysis.
Preparing reference material often involves the preparation
of materials containing a small percentage of GM in a
background of non-GM material. Some authorities prefer to
use mixtures of the GM seed with what they define as
near-isogenic seed. Near-isogenic lines are ideally lines that
differ only in the gene of interest, although this is impossible
to achieve in practice because of the nature of plant breeding
and the seed multiplication steps that are necessary to produce
enough material for reference seed production. This approach
could lead to obtaining seed that is alike in size and
composition. However, seed for use in the preparation of
reference material must also be of very high purity, and
therefore the GM and non-GM plants must be grown in
isolation from each other, either physically, or in time. To do
so is particularly important for an outcrossing crop such as
maize. Thus, environmental effects will likely overshadow
most possible benefits that could be obtained by using a
near-isogenic line. In a real-life situation, however, the
presence of any GM-seed in a conventional seed lot can be the
result of mechanical mixing or cross-pollination. Both effects
will invariably lead to GM kernels that are not isogenic to the
bulk of the material with which they are commingled.
(b) GM DNA of plant origin.—Using GM DNA derived
from plant materials other than seed provides somewhat more
flexibility than using seed or powders as a reference material.
The storage of DNA solutions may also be easier.
One approach to using GM DNA is to use spiked mixtures
of seed as a starting point for a positive control. Seed that has
been tested and shown to contain below a threshold of the
accepted level of adventitious material is used as a starting
point for the negative control. This approach has some of the
same sampling issues that are associated with using seed, at
least concerning the initial material. However, the actual
reference material can be tested for homogeneity, and
relatively small amounts are required.
Another approach to using GM DNA is to use fresh plant
(leaf) tissue as the source of the DNA. Negative controls can
be derived from a known non-GM plant; positive controls are
derived from mixtures of positive and negative GM DNA to
simulate various percentages of GM presence in grain. In this
case, it is possible to establish that a particular plant is in fact
negative, so that a true negative control can be obtained by use
of plant DNA.
The disadvantage of directly using DNA as reference
material is that it cannot be used to control for extraction and
matrix-related artifacts. Additional biological factors may
also play a role; as discussed elsewhere in this article, different
tissues may exhibit differences in their genetic makeup. For
most agriculturally relevant crops, leaves would not be a
relevant commodity, as typically grain is used for food (e.g.,
maize, soy, canola, wheat, barley). And, as with any reference
material, stability may be an issue; the stability of a
DNA-based reference material should be tested and
documented in the certificate of analysis.
Plasmid DNA or Amplicons Containing the Target
Sequence
Using a plasmid or amplicon (or cloned fragment)
containing the cloned target sequences may be attractive for
certain purposes, such as protocol optimization and
troubleshooting, as well as for an additional positive control.
Those using plasmid or amplicon DNA as reference material
usually validate the assay using the matrix of interest and only
then use the plasmid as reference. As with GM DNA, use of
plasmid or amplicon DNA ignores matrix effects compared to
the test material. The reference DNA is compared to the
reference material and is therefore traceable back to the
physical standard. However, this type of DNA has some
special characteristics.
It is possible that plasmid DNA may behave differently in a
reaction if presented as a closed circular, relaxed, or linear
form. In addition, measuring the actual copy number of
plasmid or amplicon DNA added to a reaction poses special
challenges. There is no accurate method to measure such
small amounts of DNA, so the amount added must be inferred
from a dilution series, which may have to take into account the
absorption of low concentrations of DNA onto surfaces. In
addition, the presence of concentrated plasmid solutions in
laboratories poses a potent contamination hazard.
Summary of Reference Material Considerations
In summary, the analyst conducting PCR to detect the
presence of a GM material in seed or processed materials must
make a number of decisions. A key decision is the type of
reference material to use. This decision will be influenced by
the availability of reference materials and any consideration of
matrix effects. In any case, each method should be validated in
the laboratory using a reference material of the highest
metrological standard available (SRM or CRM if possible).
LIPP ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 88, NO. 1, 2005 151
The laboratory may then use a reference material or working
standard that has been calibrated back to the CRM/SRM.
Sources of Errors
Biological Sources of Errors
In determining the percent GM value for an unknown
sample, the laboratory must convert the analytical result
(copies of the GM gene/copies of the endogenous gene) into a
percent GM value (weight to weight). This conversion
assumes there is a direct 1:1 relationship between the
endogenous control gene and the GM gene. However, there
are many biological factors that can affect this 1:1 relationship
and, as such, this basic assumption is not valid in many
circumstances. Of most significance is the effect of biological
factors on the 1:1 relationship. This effect is most pronounced
in maize and wheat grains and grain products, but soybeans
and cotton are not exempt from the basic physiological issues
discussed below. In this discussion, we will focus on the major
factors that impact the 1:1 ratio assumptions. A number of
these impacts appear easy to account for or to develop an
adjustment factor for. However, it is important to remember
the test portion used for testing likely contains a mixture of
GM events, and there is no understanding of the relative
contributions of these events in the test portion. As such, the
use of conversion factors to account for differences in copy
numbers is not readily possible. In addition, there are a
number of issues that arise from plant breeding and
physiology that can impact the conversion factors. In this
discussion, maize and maize products will be used as the basis
for discussion.
(a) Hybrid status.—A large proportion of maize grown in
the world is produced by using hybrid seed. However, the
grain produced for this seed and used for the production of
food and feed does not maintain the homogeneous genotype
and segregates the traits based on simple Mendelian
inheritance patterns. The relationship between the GM gene
and the endogenous gene can be significantly affected,
depending on whether the GM gene comes from the maternal
or paternal parent, or both. The most significant effects of
hybrid status appear in the endosperm fraction of maize
products, based on its triploid condition, and are discussed
under tissue type effects below.
(b) GM gene copy number.—Several laboratories use the
35S and NOS screen to quantify the presence of GM in grain
and grain products. With this approach, the testing laboratory
assumes that the endogenous gene and the GM gene are
present in a 1:1 relationship in the GM grains present in the
sample. This relationship is correct in some of the GM events
currently on the market. However, it does not hold true for all
of the GM events that could be present in a grain or grain
products. For example, the maize event Bt11 contains 2 copies
of the 35S promoter for every 1 copy of the maize
(endogenous) gene in the DNA extract. In this specific case,
the percent GM level in this sample could be overestimated by
a factor of 2´. In general, the assumption that there is a 1:1
ratio generally leads to the overestimation of the amount of
GM present in the sample (43).
(c) DNA degradation.—There is an assumption in the
testing of grain and grain products that the DNA present in the
grain is of good quality and present in all cells. The goal of the
analytical laboratory is to remove DNA in as intact a form as
possible and subject it to the PCR analysis. All DNA testing
laboratories acknowledge that DNA quality is a critical
element in performing a rugged analytical test. The presence
of high-quality DNA in mature grain, however, may not be an
accurate assumption. There is evidence that the cells of the
endosperm undergo apoptosis, or programmed cell death,
during the development of the maize kernel (44, 45). One
hallmark of apoptosis is the degradation of the nuclear DNA
into small fragments. These studies show that most of the
DNA in mature kernels is degraded to some degree.
(d) DNA endoreduplication.—Endoreduplication results
when replication is not coordinated to the cell cycle.
Endosperm development in maize is characterized by a period
of intense mitotic activity followed by a period in which
mitosis is essentially eliminated and the cell cycle becomes
one of alternating S and G phases, leading to
endoreduplication of the nuclear DNA. This leads to the
polyploidization, in which a single cell can contain several
copies of the genome. This process is initiated with the onset
of starch and storage protein synthesis and results in
polyploidy values ranging from 6C to 96C. (The DNA amount
in the unreplicated gametic nucleus of an organism is referred
to as its C-value, irrespective of the ploidy level of the
taxon; 46, 47). This biological process combined with
apoptosis challenges the assumption that maize kernels are a
good source of genetic material for quantification of percent
GM, as it impacts the relative contribution of germ and
endosperm representation in ground maize kernels. This
complexity in genetic contribution and copy
representativeness must be taken into account when
quantitative PCR is used.
(e) Outcrossing vs inbreeding.—In the production of
maize grain in the field, there are 2 sources of pollination.
Within any field, a majority of the kernels will be fertilized
with pollen by plants within the same field. In this case, the
plant population will maintain the same average genetic
constitution as original seed materials. In the production of
non-GM maize, external pollination from neighboring fields
of GM maize contributes to the production of maize with
adventitious GM material. In this case, the resulting maize
grain does not contain the same ratio of GM copies to the
endogenous control.
(f) Variability in the genome.—One major assumption
during the development of the PCR testing metrics for GM is
the calculation of the number of copies in the PCR. This value
is calculated to determine the lowest GM concentration that
can be estimated from the test portion subjected to PCR
analysis. Generally, the size of a genome is estimated for each
species and this value is used to calculate such terms as the
theoretical LOD. The assumption that maize varieties are
consistent in genome size (pg/2C) is not supported by the
152 LIPP ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 88, NO. 1, 2005
published literature. For example, the maize genome size is
shown to vary by up to 40% (8, 48). This is not restricted to
widely disparate maize accessions, and significant variation
can exist within a single seed company’s breeding program
(49). A high degree of variability is not restricted to maize, and
similar variability can be found in soybeans (50, 51).
(g) Effects of grain processing.—One of the first steps in
processing maize into a food ingredient involves the
separation of embryo (germ) and endosperm. The germ is
further processed for the extraction of maize oil and
production of maize gluten feed. The endosperm is processed
to produce starches, syrups, and maltodextrins.
The embryo and endosperm are significantly different at
the genetic level. During the development of a seed, one of the
2 pollen nuclei fuses with the egg cell. This diploid cell type
continues to develop into the embryo. The second pollen
nucleus fuses with 2 polar bodies, producing a triploid cell
type; this forms the endosperm of the developing seed. As
such, the embryo maintains a diploid status (1:1 relationship),
but the endosperm does not, and the proportion of endogenous
gene to GM gene can be either 2:1 or 1:2, depending on hybrid
status. Based on this fact, it may be inappropriate to assume a
1:1 ratio of the GM gene:endogenous gene when testing food
ingredients or food products for the percent GM level.
Analytical/Instrumental Sources of Errors
Total analytical error (or measurement error) refers to
assay errors from all sources derived from a data collection
experiment. The accuracy and precision of a PCR method for
GM detection or quantification are subject to influences of
total analytical error. Total analytical error is of paramount
importance in judging the acceptability of PCR-based GM
detection or quantification methods. Errors in PCR assays can
be classified as follows: (1) random (indeterminate);
(2) systematic (determinate).
(a) Random error.—Because the intrinsically uncertain
nature of the measurement technique is the source of random
error, this kind of error occurs in every analysis and is not
predictable. The amount of random error can be greatly
increased or decreased by a variety of factors in PCR-based
GM detection or quantification methods. Such factors include
the number and complexity of steps in the method, the number
of replicates, the skill of the analysts who perform the assay,
and the laboratory conditions. For example, 2 analysts with
different skills in PCR technique in the same laboratory or
2 analysts with similar skill in 2 different laboratories could
produce different results using the same sample and same
assay procedure. Therefore, a rugged PCR assay is a very
important factor in reducing random errors. Random errors
can be reduced by increasing the number of data points and
calculating the mean of these data. The average of a large
number of data points that are affected only by random error is
always accurate.
In a single qualitative PCR, a proper number of replicates
per sample is essential to reduce random errors. In the
quantitative PCR method, 3 repeats per sample are the
minimum acceptable for the collection of each data point.
Repeatability and reproducibility standard deviations are
usually used to estimate random errors. Good statistical
practices, such as the measurement of coefficient of variation
(%CV) or the repeatability standard deviation (RSD
r
), need to
be implemented in a quantitative PCR method and will aid the
analyst in the evaluation of detection methods and the
performance of such a method under local conditions.
Generally, a CV value of replicates should not exceed 30%.
(b) Systematic error.—Systematic errors cause results to
deviate from the expected or true values in a constant manner.
Sources include improper instrument calibration procedures,
insufficient purity of reagents, and improper operation of the
measurement instrument. Generally, systematic errors cannot
be reduced by the application of statistical methods, such as
taking the average of replicate measurements. This kind of
error may often be identified by careful validation and data
analysis and subsequently minimized by modifying the
analytical procedure.
The quality of reagents used in DNA extraction, PCR
amplification, and labeled probes in a quantitative PCR assay
can affect the test results. DNA quantification is one of the key
steps in a PCR-based assay. Instrument errors caused by a
malfunctioning UV-visible spectrophotometer or
fluorescence spectrophotometer (for instance, an expired bulb
for emission/excitation light sources), or improper operation
of the instruments will affect the accuracy of DNA
measurements, resulting in errors in downstream applications.
Routine use of pipet devices in PCR-based assays is another
source of instrumental errors. All pipet devices require
calibration on a regular basis.
The most critical instrument in PCR-based methods is a
thermal cycler. Because temperature change, especially in the
annealing step, can alter PCR amplification efficiencies,
temperature verification of a thermal cycler is recommended.
Temperature changes in the heat and cooling block of a
thermal cycler can be checked by using a verification system,
and many manufacturers will offer this as a service.
Real-time PCR technology has been extensively developed
in recent years. Several different types of instrument are
commercially available. They differ in design and many
specifications, including the heating and cooling system,
source of excitation/emission light, detection range of
fluorescence, and calculation algorithm (software). For this
reason, a quantitative method using real-time PCR technology
should be specific to certain types of instruments (e.g., block
cycler). Application of an assay across instruments of
different platforms without calibration or validation will result
in errors.
(c) Quality control.—In order to reduce the total analytical
error, quality control steps such as training of analysts,
standard operation protocols (SOP), and regular instrument
maintenance should be implemented throughout the assay
process. Good laboratory practice (GLP), quality assurance
systems according to ISO17025 (52), or other equivalent
quality assurance management systems are highly
recommended. Quality assurance schemes may be required in
laboratories or facilities where PCR-based methods are used
LIPP ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 88, NO. 1, 2005 153
for GM detection and quantification in order for the results to
be accepted in some countries. Tools from statistical process
control (SPC) can be used to objectively evaluate the nature
(random error vs systematic error) and amount of
measurement uncertainty.
Conclusions
PCR technology is often used for the detection of products
of agricultural biotechnology. It is critical that such methods
are reliable and give the same results in laboratories across the
world. This can be achieved by proper validation of the
methods. The choice of the appropriate reference material will
impact the reliability and accuracy of the analytical results. It
is important that analysts pay proper attention to the effect of
specific matrixes on the methods. In addition, numerous
biological and analytical factors need to be taken into account
when reporting results. This is particularly important when
interpreting quantitative data.
Acknowledgments
We thank Croplife International and Analytical
Environmental Immunochemical Consortium for the financial
support to prepare this manuscript. We also thank Pat Feeney
for her invaluable help in preparing the manuscript.
References
(1) Lipton, C.R., Dautlick, J.X., Grothaus, C.D., Hunst, P.L.,
Magin, K.M., Mihaliak, C.A., Rubio, F.M., & Stave, J.W.
(2000) Food Agric. Immunol. 12, 153–164
(2) Ingham, D.J., Beer, S., Money, S., & Hansen, G. (2001)
Biotechniques 31, 132–134
(3) Schmidt, M.A., & Parrott, W.A. (2001) Plant Cell Rep. 20,
422–428
(4) Song, P., Cai, C.Q., Skokut, M., Kosegi, B.D., & Petolino,
J.F. (2002) Plant Cell Rep. 20, 948–954
(5) Chien, A., Edgar, D.B., & Trela J.M. (1976) J. Bacteriol.
127, 1550–1557
(6) Remund, K., Dixon, D.A., Wright, D.L., & Holden, L.R.
(2001) Seed Sci. Res. 11, 101–119
(7) Whitaker, T.B., Freese, L., Giesbrecht, F.G., & Slate, A.
(2000) Paper 2000-10, Journal Series of the Department of
Biological and Agricultural Engineering, North Carolina
State University, Raleigh, NC
(8) Arumuganathan, K., & Earle, E.D. (1991) Plant Mol. Biol.
Rep. 9, 208–218
(9) Kay, S., & Van den Eede, G. (2001) Nature Biotech., 19, 405
(10) Holden, M.J., Blasic, J.R., Bussjaeger, L., Kao, C., Shokere,
L.A., Kendal, D.C., Freese, L., & Jenkins, G.R. (2003) J.
Agric. Food Chem. 51, 2468–2474
(11) Rossen, L., Norskov, P., Holmstrom, K., & Rasmussen, O.F.
(1992) Int. J. Food Microb. 17, 37–45
(12) Séjalon-Delmas, N., Roux, C., Martins, M., Kulifaj, M.,
Becard, G., & Dargent, R. (2000) J. Agric. Food Chem. 48,
2608–2613
(13) Tengel, C., Schussler, P., Setzke, E., Balles, J., &
Sprenger-Haussels, M. (2001) Biotechniques 31, 426–429
(14) Laws, G.M., Skopek, T.R., Reddy, M.V., Storer, R.D., &
Glaab, W.E. (2001) Mutat. Res. 484, 3–18; Erratum in:
Mutat. Res (2002) 505, 93
(15) Tichopad, A., Didier, A., & Pfaffl, M.W. (2004) Mol. Cell.
Probes 18, 45–50
(16) Weissensteiner, T., & Lanchbury, E. (1996) Biotechniques
21, 1102–1108
(17) Permingeat, H.R., Reggiardo, M.I., & Vallejos, R.H. (2002)
J. Agric. Food Chem. 50, 4431–4436
(18) Koonjul, P.K., Brandt, W.F., Farrant, J.M., & Lindsey, G.G.
(1999) Nucleic Acids Res. 27, 915–916
(19) Reiss, R.A., & Rutz, B. (1999) Biotechniques 27, 920–922,
924–926
(20) Seeger, C., Batz, H.G., & Orum, H. (1997) Biotechniques 23,
512–517
(21) Lipp, M., Anklam, E., Brodmann, P., Pietsch, K., & Pauwels,
J. (1999) Food Control 10, 379–383
(22) Tikoo, K., Gupta, S., Hamid, Q.A., Shah, V., Chatterjee, B.,
& Ali, Z. (1997) Biochem. J. 322, 273–279
(23) Watson, J.D., Hopkins, N.H., Roberts, J.W., Steitz, J.A., &
Weiner, A.M. (1987) Molecular Biology of the Gene, 4th Ed.,
Benjamin Cummings Publishing Co. Inc., San Francisco,
CA, pp 690–696
(24) Goldmit, M., Schlissel, M., Cedar, H., & Bergman, Y. (2002)
EMBO J. 21, 5255–5261
(25) Xiao, L., & Casero, R.A. (1996) Biochem. J. 313, 691–696
(26) Wolf, C., Scherzinger, M., Wurz, A., Pauli, U., Hubner, P., &
Luthy, J. (2000) Eur. Food Res. Technol. 210, 367–372
(27) Ausubel, F.M., Brent, R., Kinsdton, R.E., Moore, D.D., &
Siedman, J.G. (1999) in Short Protocols in Molecular
Biology, John Wiley & Sons, Inc., New York, NY, pp
A3–10, 11
(28) Labarca, C., & Paigen, K. (1980) Anal. Biochem. 102,
344–352
(29) Singer, V.L, Jones, L.J., Yue, S.T., & Haugland, R.P. (1997)
Anal. Biochem. 249, 228–238
(30) Permingeat, H.R., Reggiardo, M.I., & Vallejos, R.H. (2002)
J. Agric. Food Chem. 50, 4431–4436
(31) Seedcalc 6, ISTA Web site, http://www.seedtest.org/
en/content—1—1143.html, to download directly,
http://www.seedtest.org/upload/cms/user/seedcalc6.zip
(32) Applied Biosystems (1997) User Bulletin #2: ABI Prism
7700 Sequence Detection System, Applied Biosystems,
Foster City, CA
(33) Pardigol, A., Guillet, S., & Popping, B. (2003) Eur. Food
Res. Technol. 216, 412–420
(34) C-Value Web site: http://www.rbgkew.org.uk/
cval/introduction.html
(35) Horwitz, W., Britton, P., & Chirtel, S. (1998) J. AOAC Int.
81, 1257–1265
(36) Codex Alimentarius Commission (2001) Procedural Manual,
12th Ed., FAO, Rome, Italy, p. 64
(37) EPA Pesticide Regulation (PR) Notice 96-1 (1996) Notice to
Manufacturers, Formulators, Producers and Registrants of
Pesticides Products, http://www.epa.gov/PR_Notices/pr96-1.html
(38) AOAC Official Methods Program Manual (2002) AOAC
INTERNATIONAL, Gaithersburg, MD, Appendix X, pp
14–15
154 LIPP ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 88, NO. 1, 2005
(39) Horwitz, W. (1995) Pure Appl. Chem. 67, 331–343
(40) Huebner, P., Waiblinger, H.U., Pietsch, K., & Bordmann, P.
(2001) J. AOAC Int. 84, 1855–1864
(41) ISO Guide 30:1992, ISO Guide 31:2000, ISO Guide
32:1997, ISO Guide 33:2000, ISO Guide 34:2000, ISO Guide
35:1989, International Standardization Organization, Geneva,
Switzerland
(42) Institute of Reference Material and Measurement of the Joint
Research Center of the European Union (2004) Geel,
Belgium, http://www.irmm.jrc.be/rm/food.pdf
(43) AgBios (2004) Transgene Copy Number, http://www.agbios.com
(44) Young, T.E., Gallie, D.R., & DeMason, D. (1997) Plant
Physiol. 115, 737–751
(45) Gallie, D.R., & Young, T. (2000) Plant Mol. Biol. 42,
397–414
(46) Dilkes, B.P., Dante, R.A., Coelho, C., & Larkins, B.A. (2002)
Genetics 160, 1163–1177
(47) Schweizer, L., Yerk-Davis, G.L., Phillips, R.L., Srienc, F., &
Jones, R.J. (1995) Proc. Natl. Acad. Sci. USA 92, 7070–7074
(48) Poggio, L., Rosato, M., Chiavarino, A.M., & Naranjo, C.A.
(1998) Ann. Bot. 82, 107–115
(49) McMurphy, L.M., & Rayburn, A.L. (1991) Crop Sci. 31,
63–67
(50) Grielhuber, J., & Obermayer, R. (1997) Heredity 78,
547–551
(51) Graham, M.J., Nickell, C.D., & Rayburn, A.L. (1994) Theor.
Appl. Genet. 88, 429–432
(52) ISO 17025 (1999) International Standardization
Organization, Geneva, Switzerland
LIPP ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 88, NO. 1, 2005 155
