Transgenic Res (2015) 24:587–603

DOI 10.1007/s11248-015-9867-7

REVIEW

Genetic basis and detection of unintended effects

in genetically modified crop plants
Gregory S. Ladics • Andrew Bartholomaeus • Phil Bregitzer • Nancy G. Doerrer •
Alan Gray • Thomas Holzhauser • Mark Jordan • Paul Keese • Esther Kok • Phil Macdonald •

Wayne Parrott • Laura Privalle • Alan Raybould • Seung Yon Rhee • Elena Rice •
Jörg Romeis • Justin Vaughn • Jean-Michel Wal • Kevin Glenn

Received: 18 January 2015 / Accepted: 14 February 2015 / Published online: 26 February 2015
 The Author(s) 2015. This article is published with open access at Springerlink.com

Abstract In January 2014, an international meeting to explore current knowledge and identify areas re-
sponsored by the International Life Sciences Institute/ quiring further study on unintended effects in plants
Health and Environmental Sciences Institute and the and to discuss how this information can inform and
Canadian Food Inspection Agency titled ‘‘Genetic improve genetically modified (GM) crop risk assess-
Basis of Unintended Effects in Modified Plants’’ was ments. The meeting featured presentations on the
held in Ottawa, Canada, bringing together over 75 molecular basis of plant genome variability in general,
scientists from academia, government, and the agro- unintended changes at the molecular and phenotypic
biotech industry. The objectives of the meeting were levels, and the development and use of hypothesis-
driven evaluations of unintended effects in assessing
Electronic supplementary material The online version of conventional and GM crops. The development and
this article (doi:10.1007/s11248-015-9867-7) contains supple- role of emerging ‘‘omics’’ technologies in the assess-
mentary material, which is available to authorized users. ment of unintended effects was also discussed. Several
G. S. Ladics A. Gray
DuPont Pioneer Agricultural Biotechnology, DuPont Centre for Ecology and Hydrology, CEH Wallingford,
Experimental Station, 200 Powder Mill Road, Crowmarsh Gifford, Wallingford,
Wilmington, DE 19803, USA Oxfordshire OX10 8BB, UK

A. Bartholomaeus T. Holzhauser
Therapeutics Research Centre, School of Medicine, Division of Allergology, Paul-Ehrlich-Institut, Paul-
Queensland University, Brisbane, QLD 4072, Australia Ehrlich-Strasse 51-59, 63225 Langen, Germany

A. Bartholomaeus M. Jordan
Faculty of Health, School of Pharmacy, University of Cereal Research Centre, Agriculture and Agri-Food
Canberra, Locked Bag 1, Canberra, ACT 2601, Canada, 101 Route 100, Morden, MB R6M 1Y5, Canada
Australia
P. Keese
P. Bregitzer Office of the Gene Technology Regulator, Australian
National Small Grains Germplasm Research Facility, Government, MDP54, GPO Box 9848, Canberra,
US Department of Agriculture – Agricultural Research ACT 2601, Australia
Service, 1691 S. 2700 W., Aberdeen, ID 83210, USA
E. Kok
N. G. Doerrer (&) RIKILT Wageningen UR, P.O. Box 230,
ILSI Health and Environmental Sciences Institute, 1156 6700 AE Wageningen, The Netherlands
15th St., NW, Suite 200, Washington, DC 20005, USA
e-mail: ndoerrer@hesiglobal.org

123
588 Transgenic Res (2015) 24:587–603

themes recurred in a number of talks; for example, a can be developed that directly assess the protein for
common observation was that no system for genetic toxicity and allergenicity and measure levels of
modification, including conventional methods of plant metabolites that may be associated with the protein’s
breeding, is without unintended effects. Another function. Unintended changes, on the other hand,
common observation was that ‘‘unintended’’ does not could materialize as a consequence of gene insertion,
necessarily mean ‘‘harmful’’. This paper summarizes from random mutations that take place during the
key points from the information presented at the transformation and tissue culture process, or from
meeting to provide readers with current viewpoints on pleiotropic effects of the introduced protein, and there
these topics. is no single direct test for them.
In January 2014, an international meeting spon-
Keywords Unintended effects
GM crop plants
sored by the International Life Sciences Institute/
Environmental risk assessment
Allergenicity
Health and Environmental Sciences Institute and the
Toxicity Canadian Food Inspection Agency titled ‘‘Genetic
Basis of Unintended Effects in Modified Plants’’ was
held in Ottawa, Canada, bringing together over 75
Introduction scientists from academia, government, and the agro-
biotech industry. The objectives of the meeting were
As genetically modified (GM) crops worthy of com- to explore current knowledge and identify areas re-
mercialization became available, procedures were in- quiring further study on unintended effects in plants
stituted to ensure that these plants were as safe for and to discuss how this information can inform and
food, feed, and environmental release as their con- improve GM crop risk assessments.
ventional (non-GM) counterparts. These procedures The potential for an unintended effect to present a food
addressed two broad categories of changes that could or feed hazard is currently assessed through composi-
be considered in a GM crop safety assessment: in- tional analyses and agronomic studies to compare the GM
tended and unintended. The intended change in a new crop with a genetically similar conventional counterpart.
GM product is the desired phenotype brought about by Some regulatory authorities, such as those in the Euro-
the introduced transgene. Because many transgenes pean Union (EC 2013), may also require animal feeding
express a known and characterized protein, procedures tests. Some aspects of testing for unintended effects seem

P. Macdonald E. Rice
Canadian Food Inspection Agency, 1400 Merivale Rd, Monsanto Company, 700 Chesterfield Pkwy W., CC5A,
Ottawa, ON K1A 0Y9, Canada Chesterfield, MO 63017, USA

W. Parrott J. Romeis
Center for Applied Genetic Technologies, University of Agroscope, Institute for Sustainability Sciences ISS,
Georgia, 111 Riverbend Road, Athens, GA 30602, USA Reckenholzstr. 191, 8046 Zurich, Switzerland

L. Privalle J. Vaughn
Bayer CropScience, 407 Davis Drive, Morrisville, University of Georgia, 111 Riverbend Road, Athens,
NC 27560, USA GA 30602, USA

A. Raybould J.-M. Wal

Syngenta Ltd, Jealott’s Hill International Research Centre, Dept. SVS, AgroParisTech, 16 rue Claude Bernard,
Bracknell RG42 6EY, UK 75231 Paris, France

Present Address: K. Glenn

A. Raybould Monsanto Company, 800 N. Lindbergh Blvd, U4NA,
Syngenta Crop Protection AG, Schwarzwaldallee 215, St. Louis, MO 63167, USA
4058 Basel, Switzerland

S. Y. Rhee
Department of Plant Biology, Carnegie Institution for
Science, 260 Panama St., Stanford, CA 94305, USA

123
Transgenic Res (2015) 24:587–603 589

to be generally accepted, such as the use of related con- assembled in novel combinations (reviewed in
ventional comparators. Nevertheless, many questions are Weber et al. 2012).
still being discussed, for example, whether it is sufficient • There are genes that are present in different
to limit testing for unintended effects to those subject to numbers or absent altogether in different indi-
testable hypotheses, or whether (and when) the precau- viduals within a crop (e.g., Lai et al. 2010; Lam
tionary principle requires a broader look for genomic et al. 2010; Potato Genome Consortium 2011;
changes via profiling methods. As noted during the McHale et al. 2012).
opening presentation and by several other speakers, • Horizontal gene transfer is not uncommon, with
‘‘unintended’’ is not synonymous with ‘‘harmful’’ (e.g., pararetroviral (double-stranded DNA virus) se-
NRC 2004). quences being particularly abundant in the gen-
This manuscript summarizes four broad areas dis- omes of crop plants (e.g., Liu et al. 2012; Staginnus
cussed at the meeting: the molecular changes associ- et al. 2007).
ated with plant genetic variability, the types of
The effects of naturally occurring insertions are of
unintended genomic changes in GM plants, the use of
particular interest because plant genetic engineering is
hypothesis-driven evaluations of unintended effects,
typically mediated by the insertion of a modified
and the use of emerging technologies in the assess-
T-DNA sequence from Agrobacterium tumefaciens or
ment of unintended effects. This paper is based on the
other vector DNA sequences into the genome. This
meeting presentations, with new and updated infor-
insertion may potentially disrupt the function of native
mation added where appropriate. For each section, the
genes and can create rearrangements at the site of in-
primary contributors are noted, but comments and
sertion. Indeed, roughly half of T-DNA insertions
edits from other authors have been included. The au-
exhibit less than 8 bp of ‘‘filler’’ DNA at the junction
thors’ individual papers in their entirety are available
site, while the other half contain larger additions,
as Online Resource 1. Presentations and other infor-
generally between 8 and 100 bp (Forsbach et al.
mation from the meeting can be found at http://www.
2003). Short insertions, comparable to those seen at
hesiglobal.org/i4a/pages/index.cfm?pageID=3654.
T-DNA junctions, have been observed in induced
double-strand break experiments (Lloyd et al. 2012;
Molecular basis of plant genetic variability1 Vu et al. 2014). Such insertion variation is common,
even in closely related rice varieties, and reflects the
High-throughput sequencing and other genomic fact that errors in double-strand break repair are fre-
technologies have made it possible to evaluate the quent in natural and breeding populations (Vaughn
nature and extent of naturally occurring genomic and Bennetzen 2014). Thus, the ‘‘filler’’ DNA ob-
changes in plants. These were extensively reviewed by served in T-DNA insertions has a clear counterpart in
Weber et al. (2012), who noted the following: naturally occurring DNA insertions.
In summary, the view that only transgenes result in
• Single-nucleotide changes are common, with a
insertions, and that these have a unique ability to
background rate of seven new mutations per
disrupt gene expression (e.g., Fagan et al. 2014) is not
billion bp of DNA, or roughly seven new muta-
supported by the available data. Instead, plant gen-
tions for every soybean (Glycine max) plant in
omes are very dynamic and plastic, as predicted by
every field (Ossowski et al. 2010).
Barbara McClintock (1984) in her Nobel address, and
• Insertions from transposons can be very common as
undergo frequent insertions and other rearrangements.
well, with rates as high as 50 novel insertions per plant
per generation reported in a variety of rice (Oryza
sativa; Naito et al. 2006). Transposon insertions are
also very common in soybean (Tian et al. 2012). Molecular basis of unintended effects in GM plants
• Plants create novel genes through transposon
capture, whereby pieces of different genes are In addition to the potential for insertional effects of
transgenes, other mechanisms such as somaclonal
1
Section based on presentations from Wayne Parrott and Justin variation and pleiotropy can contribute to unintended
Vaughn. effects.

123
590 Transgenic Res (2015) 24:587–603

Somaclonal variation2 also causes premature senescence of the flag leaf (leaf
tip necrosis) that can reduce potential yield in the
Semi-differentiated plant tissues cultured in vitro are absence of disease. If the wheat Lr34 gene is moved
critical for most plant transformation methods. In vitro into barley (Hordeum vulgare), the negative effect
culture induces genetic and epigenetic changes, ter- becomes stronger and the plants exhibit stunted
med somaclonal variation (SCV; Larkin and Scow- growth and sterility (Risk et al. 2013). One possible
croft 1981), that are another possible source of explanation is that wheat has regulatory mechanisms
unintended variation. Although SCV is potentially that control the expression of the gene in a manner that
useful as a source of novel mutations, it is contrary to minimizes its negative effects; on the other hand, the
the objective of making limited, predictable changes barley lines tested were primary transgenic lines and
as a result of transgene introduction. For example, had not undergone selection against SCV (described in
significant and negative changes have been noted in the previous section). The amount of leaf tip necrosis
the agronomic performance and malting quality of varies among wheat genotypes, but plant breeders
tissue-culture-derived barley. Yield losses of 15–84 % have selected lines that maximize the benefit while
have been observed in non-transgenic derivatives of reducing the negative effect of the gene. The same
transgenic plants, i.e., non-transgenic segregants may be possible in barley if the detrimental effects
derived from heterozygous transgenic plants (Bregit- seen are due to somaclonal variation rather than to
zer et al. 1998). Although certain adjustments to the pleiotropy.
in vitro environment can reduce the severity of SCV, Prediction of whether pleiotropy (and therefore the
the most effective way to eliminate it in barley has possibility of unintended effects) is likely to occur as
been to backcross transgenic plants to plants without the result of a transgene depends on knowledge of the
any SCV (such as the wild-type parent used in biochemical mechanism of the encoded protein. Genes
making the original transgenic plant), with selection affecting basic cellular functions that are needed by
at each generation based only on the presence of the many traits (such as ABC transporters) are more likely
transgene. For example, a single backcross to barley to be pleiotropic. Similarly, genes in which alternative
cultivar Conlon recovered the majority of yield loss splicing occurs in the pre-mRNA or that encode a
caused by SCV in a group of transgenic Conlon- protein affecting multiple pathways (e.g., transcription
derived lines. On average, the yield loss in the factors or other regulatory proteins or molecules)
primary transgenic lines was 31 %, compared with could potentially be pleiotropic.
6 % in the backcross-derived lines (Bregitzer and A gene’s origin may also be an indicator of whether
Dahleen 2008). pleiotropy is likely to occur, but this is harder to pre-
dict. There could be more pleiotropy if the gene ori-
Pleiotropic effects of transgenes3 ginates from another species due to lack of regulatory
controls (e.g., expression of wheat Lr34 in barley) or
Some unintended effects might be caused by less pleiotropy due to lack of a pathway or function in
pleiotropy, the effect of a single gene (whether a native the recipient species compared with the donor (origi-
gene or a transgene) on multiple traits (Fagan et al. nal) species. An example of this is the lack of antho-
2014). When both positive and negative effects are cyanin production in transgenic tomato (Lycopersicon
caused by the same gene, it is referred to as an- esculentum) fruit after the introduction of genes
tagonistic pleiotropy. An example of antagonistic regulating flavonol and anthocyanin production from
pleiotropy is the wheat (Triticum aestivum) gene Lr34, maize (Zea mays) (Bovy et al. 2002). The tomato
which encodes an ABC transporter, a molecule in- plants had increases in some flavonols but no antho-
volved in the transport of metabolites across mem- cyanin production because tomato lacks sufficient
branes (Krattinger et al. 2009). Lr34 provides durable expression of another gene required for anthocyanin
resistance to a number of wheat diseases; however, it production. Variation in regulatory mechanisms is not
only observed at the species level—different geno-
types of the same species can have variation in
2
Section based on presentation by Phil Bregitzer. regulatory mechanisms (e.g., Schiessl et al. 2014),
3
Section based on presentation by Mark Jordan. which is one reason there is so much phenotypic and

123
Transgenic Res (2015) 24:587–603 591

phenologic diversity in crop species. Without this type extension of conventional breeding techniques and
of diversity, selection during plant breeding would be that the focus of assessment should be on the novel
less effective. trait rather than on how the trait was obtained.

Similarity between unintended effects The GM product development process5

in conventional plant breeding and biotechnology4
The extensive vetting involved in the generation and
Plant breeders have successfully improved crop yields selection of one or a few ‘‘elite’’ (i.e., top-performing)
despite having little or no information on the genes and events minimizes the likelihood of unsafe unintended
gene networks that impact yield. In the case of maize, it effects associated with the GM crop products that are
is clear that improvements in grain yield have been taken to commercialization. Many ideas, traits, and
associated with significant changes in many other traits events are evaluated to identify an event for which it is
(Tollenaar and Lee 2010). Whereas in conventional worth seeking approval (e.g., Phillips McDougall
plant breeding, the exact functions of the combined 2011; Privalle et al. 2012). As noted earlier, this is not
genes are mainly unknown, biotechnology enables the unlike the approach taken by breeders seeking to de-
introduction of specific genes with expected effects on velop a new and improved variety.
endogenous pathways and phenotypes. For GM crops, as for those derived from conven-
Although crop plants derived from conventional tional breeding, the most important selection criterion
plant breeding and GM differ in the level of molecular is efficacy/performance (i.e., does the trait impart the
data required for commercialization, both conven- desired phenotype, meeting product specifications). In
tional breeding products and GM products undergo a the case of GM crops, the next most important criteria
similar process of selection for intended characteris- applied in identifying the lead event are those related to
tics and elimination of undesirable phenotypes (Pri- the molecular characteristics of the event. The inserted
valle et al. 2012). While it is clear that unintended DNA ideally should be present at a single locus and as a
effects occur in any type of breeding program, in- single copy. There should be no vector backbone pre-
cluding conventional crossing (Cellini et al. 2004; Kok sent in the event and the insertion should not have
et al. 2008; Schnell et al. 2015), the discussion on the disrupted an endogenous gene or created a chimeric
potential for unintended events tends to be focused on novel fusion protein. There should be minimal locus
GM organisms. rearrangement and the integrity of the gene cassette
This similarity between the changes caused by should have been preserved. Importantly, while none
biotechnology and conventional plant breeding is re- of these parameters have been demonstrated to impact
flected in the approach to regulation used in Canada. the safety of the crop, the consideration of these pa-
The Canadian regulatory scope covers plants devel- rameters is based on hypothetical, minimal-probability
oped to possess characteristics sufficiently different possibilities. Since most GM products require multiple
from those of the same or similar species, regardless of approvals, the requirements from the strictest juris-
the method used. As a consequence of the product- dictions dominate the event selection criteria.
based regulatory approach, in Canada the regulated Once the elite event is identified, an extensive safety
plant is referred to as a plant with a novel trait (PNT). assessment is conducted that includes studies on the
PNTs include GM crops as well as some produced by safety of the newly expressed protein, molecular
more conventional breeding techniques. The approach characterization of the insert, impact of the insert on
is designed to take advantage of the knowledge, ex- plant performance and composition, environmental
pertise, and regulatory framework that are already impact, and wholesomeness of the crop (SCBD 2000;
present in regulatory departments and agencies but Codex 2003). The assessment includes phenotypic and
applied to conventional products. It is also an ac- agronomic comparison between the new plant variety
knowledgment that the Canadian Government policy and a genetically similar comparator that is already on
considers that PNT crops should be considered as an the market and considered as safe. GM foods are among

4
Section based on presentations by Elena Rice, Esther Kok,
5
and Phil Macdonald. Section based on presentation by Laura Privalle.

123
592 Transgenic Res (2015) 24:587–603

the most highly studied foods consumed, and the regis- they were to occur, based on relevant legislation or
tration dossiers are scrutinized by regulatory agencies regulations (Evans et al. 2006), and builds scenarios
around the world. To date, approval has not been withheld comprising a series of events leading from the pro-
for any event based on an unintended effect. posed use of the particular GM crop to the identified
harmful effects. These scenarios, or ‘‘pathways to
harm’’, allow the risk assessor to devise testable hy-
Hypothesis-driven evaluation of unintended effects potheses about the likelihood, frequency, or magni-
tude of the events in the pathway. Data are collected to
Approaches to risk assessment6 test these hypotheses and thereby characterize risk
(Raybould 2011).
Risk is a combination of the seriousness and likelihood A concern raised about the latter approach is that it
of a harmful effect following a course of action. Risk represents a biased approach to assessment. Effective
assessment characterizes the amount of risk associated risk assessment does involve bias in that representative
with an activity. It contributes to making decisions protection goals must be selected from among all the
about whether to undertake an activity, such as the possible effects of using a GM crop. Limited resources
import, field testing, or cultivation of a specific GM are then targeted to test hypotheses about the prob-
crop. Some authors have raised concern (e.g., Craig ability and consequences of those effects. These will be
et al. 2008) that the amount of data required for risk strong tests of clear hypotheses, which, if corroborated,
assessments of GM crops is increasing and becoming provide high confidence in conclusions of low risk.
detrimental to decision-making in many countries.
Two possible approaches to risk assessment have Problem formulation in environmental risk
been described. In the ‘‘bucket’’ approach (Raybould assessment7
2011), data on the properties of the GM crop are col-
lected in an untargeted manner, often termed profiling. Environmental risk assessment (ERA) for GM crops
Profiling could comprise measurements of the crop’s deals almost exclusively with the phenotype and
gross phenotype, composition of key tissues, tran- considers all plant traits that may have been altered by
scriptome, proteome, metabolome, and so on. By the transformation, whether intended or unintended.
comparing these profiles with those of a suitable Of particular interest are any unintended changes in
conventional crop, the risk assessor is supposed to be traits that may make the GM plant more persistent or
able to identify changes, which in turn indicate that invasive (‘‘weedy’’) in either agricultural or natural
there may be changes in the GM crop that are poten- environments. These include changes in the properties
tially harmful (see ‘‘Omics technologies’’ below). of the seeds (such as developmental rates, number,
There are several limitations to this approach. First, release from the plant [shattering], dormancy, and
what to regard as harmful is defined by policy; it is not germination rates) that are important in the ‘‘regen-
discovered in data (Sarevitz 2004; Sanvido et al. eration niche’’ of the plant’s establishment and spread,
2012). Second, even if harm is defined, profiling will and in those traits that affect the plant’s competitive-
collect data that do not predict the seriousness or ness (such as seedling vigor, plant height, growth
probability of harm following use of the GM crop. rates, and resistance to pests and disease).
These data are thus irrelevant for risk assessment and The first stage in problem formulation in an ERA is
may impair decision-making because they distract to identify a set of environmental protection goals
from data that are relevant. Furthermore, it is difficult derived from local, national, or international policy.
to interpret the effect of an altered metabolite in the These may be broadly stated (e.g., the Cartagena
absence of a hypothesis. Protocol) or more specific laws, statutes, or even
The second approach regards risk assessment as a guidelines, but collectively they enable a risk assessor
hypothesis-testing exercise. The risk assessor identi- to identify those aspects of the environment that must
fies those effects that would be regarded as harmful if be protected. These can sometimes be formally

6 7
Section based on presentation by Alan Raybould. Section based on presentation by Alan Gray.

123
Transgenic Res (2015) 24:587–603 593

defined in terms of assessment endpoints (e.g., ‘‘insect It is unlikely that systemically toxic proteins that
pollinator abundance’’), which are whatever will be are unrelated to the parent plant variety or to the
measured to ascertain whether protection has been function of the transgene will be produced de novo
achieved as intended (Paes de Andrade et al. 2014). in a GM plant (Weber et al. 2012). The reason is
The second stage of problem formulation is to seek a that systemic toxicity of an ingested protein requires
link between the cultivation of the GM crop and the at least three highly specific structural characteris-
assessment endpoint that may result in harm (i.e., the tics: (1) resistance to digestion, (2) ligand specificity
pathway to harm). For example, insect pollinators are for the gut uptake transporters, and (3) ligand/re-
likely to be harmed if the plant presents a hazard (e.g., an ceptor specificity for site- and species-specific re-
insecticidal protein that negatively affects the insect) to ceptor-mediated toxicity (Hammond et al. 2013). A
which the insect may be exposed. Steps along the path- change in any one of these three characteristics is an
way to harm can be recast as risk hypotheses that can be implausible outcome from either conventional plant
validated or rejected from existing data, or by designing breeding or gene transfer; thus, the probability of
new experiments or trials where appropriate. For ex- having all three occur in the same plant is vanish-
ample, validation of the hypothesis ‘‘the insect is not ingly small. Similarly, the potential for random
harmed by the protein’’ or ‘‘the protein is not expressed genome effects to modify existing non-toxic proteins
in pollen’’ allows a confident risk assessment without to create a toxic protein is also essentially zero
further experimentation. Wolt et al. (2010) describe the (Weber et al. 2012). This prediction is evidenced by
process in detail and Raybould (2011) and earlier papers the current knowledge of conventional corn and
referred to therein give specific examples of formulating other food crop varieties that have millions of sin-
and testing risk hypotheses. Gray (2012) and Tepfer et al. gle-nucleotide polymorphisms (SNPs) across geno-
(2013) give practical examples of the use of problem types (Tenaillon et al. 2001), but have never resulted
formulation in ERA for GM crops. in a novel toxic protein in a food crop (Steiner et al.
2013).
Evaluation of food and feed safety The de novo generation of the machinery necessary
to produce a toxic secondary metabolite is also very
As mentioned earlier, the approach to food and feed unlikely. Such an event has not been observed in the
safety described by the Codex Alimentarius Commis- extensive range of varieties produced by genetic ma-
sion (Codex 2003) is based on comparison of the GM nipulation in conventional and GM crop breeding over
crop to a conventional counterpart employing a weight- the past century. The reactivation of dormant path-
of-evidence approach. In this way, the assessment is ways (i.e., pathways present in an ancestor of the crop
focused on identification of potential new hazards or that are inactive in the modern variety) has also never
changes in hazard levels in the GM product. been observed and is implausible due to the accumu-
lation of mutations in non-functional DNA, progres-
Focus on plausible outcomes8 sively degrading any residual potential functionality
(Steiner et al. 2013).
For food and feed safety risk assessment of GM crops, As in conventional plant breeding, there are natural
it is necessary to exclude hypothetical, extreme, or variations in the levels of compounds, including those
scientifically implausible circumstances. Instead, the of toxicological relevance, in crops developed through
purpose is to (1) identify practical, biologically plau- modern biotechnology. Examples of plausible
sible outcomes based on the extensive data now mechanisms include the up (and down) regulation of
available, and (2) to define the nature of the at-risk pre-existing endogenous plant toxins, increased/de-
group(s) to be addressed (population or individual creased uptake of heavy metals from the soil or water
health risks), the type of hazard(s) of concern (e.g., Cd, As, Se), altered levels of nutrients or
(toxicological, dietary, immunological), and the risk antinutrients associated with population health out-
time metric (acute, sub-chronic, or chronic). comes, altered production of pesticide metabolites,
altered levels of toxic substrates (precursors) due to
blocking of an enzyme pathway, and altered release or
8
Section based on presentation by Andrew Bartholomaeus. availability of endogenous toxins.

123
594 Transgenic Res (2015) 24:587–603

Assessment for changes in levels of existing allergens9 proven to be efficient (alternative) tools for the iden-
tification and quantification of known allergens in
According to the European Union perspective on the plants. However, such tests may sometimes be con-
assessment of the overall allergenicity of whole GM sidered as complex and insufficiently standardized and
plants, in cases where the recipient species of a genetic needing further developments and validation before
modification is a known allergen, (e.g., soybean) the they can be routinely used for safety assessment
qualitative and quantitative composition of endoge- (Fernandez et al. 2013).
nous allergens in the GM crop and its conventional
counterpart should be compared (Metcalfe et al. Examples of natural variation in allergens10
1996). The concentration of endogenous allergens
within a plant species is highly variable, and the Allergies to fruits and vegetables affect up to *4 % of
comparative analysis should consider the influence of the population in Europe (Zuidmeer et al. 2008).
the cultivars and of the conditions of cultivation, Carrot (Daucus carota) and apple (Malus domestica)
harvest, storage, and processing on the expression of are among the most prevalent elicitors of allergic re-
allergens (see ‘‘Examples of natural variation in al- actions to foods in northern and central Europe. In
lergens’’ below). The European Food Safety Authority apple, variation in patient reaction and/or allergen
(EFSA) guidance and the European Commission (EC) levels has been observed among cultivars (Bolhaar
regulation have recommended including ‘‘key’’ en- et al. 2005), stored vs. unstored fruit (Sancho et al.
dogenous allergens (i.e., such as those listed in the 2006a, b), and patient geographical areas. Similarly,
OECD consensus documents) in the comparative the carrot isoallergens (related allergens from the same
compositional analysis of plant materials collected species) Dau c 1.01 and Dau c 1.02 were quantified
from controlled field trials. This aims to assess whe- using ELISA (Foetisch et al. 2011) in two cultivars,
ther the GM plant is more allergenic than its conven- ‘Rodelica’ and ‘Nerac’, in a two-year study. Initial
tional counterpart (EFSA 2011; EC 2013). As evaluation of the field data suggests a large influence
described by the EC, ‘‘key allergens’’ are well-char- of the year of cultivation and an apparent difference
acterized allergens that are relevant for public health between the two cultivars (unpublished data, re-
because of their allergenic potency and abundance. search project BÖL 03OE349 granted by the Ger-
They are generally well-conserved proteins with im- man Federal Ministry of Food, Agriculture and
portant metabolic and physiological functions in the Consumer Protection). Furthermore, some isoaller-
plant, such as enzymes, defense proteins, or storage gens might be more relevant than others for clinical
proteins. Any significant change in key allergen levels reactivity, and the level of allergens can increase or
could thus be directly related to the specific allergy decrease depending on genetic and environmental
risk and could also indicate the possible occurrence of factors. Studies on the allergenicity of apple and
other types of unintended effects. carrot have focused on non-transgenic cultivars;
Non-targeted analyses, such as proteomic ap- however, they can be considered model foods to
proaches using mass spectrometry (e.g., matrix-as- study the influence of genetic and environmental
sisted laser desorption/ionization (MALDI) or factors on the composition of panallergenic struc-
electrospray ionization time-of-flight mass spec- tures (functionally related allergenic molecules
trometry (ESI-TOF MS) in combination with different found in different species) and the isoallergen dis-
separation methods such as 2-dimensional gel elec- tribution in fruits and vegetables. Understanding the
trophoresis or liquid chromatography), have been biochemical pathways of allergen synthesis and the
rapidly developed (Goodman et al. 2013) and can help range of natural variability may support hypothesis-
to identify significant changes in endogenous allergen driven studies on unintended effects in GM plants
expression. They may not require human sera and have intended for human consumption.

9 10
Section based on presentation by Jean-Michel Wal. Section based on presentation by Thomas Holzhauser.

123
Transgenic Res (2015) 24:587–603 595

Environmental risk assessment of GM crop plants season-long water consumption or root growth and
development were observed. These studies did not
Example of hypothesis-driven testing: drought- reveal any potential for adverse environmental
tolerant maize11 impacts.

The evaluation of Monsanto’s recently introduced Example of hypothesis-driven testing: assessment

DroughtGard maize hybrids (event MON 87460) of potential for weediness12
was used as a case study to illustrate how hypothesis-
driven testing can be used for safety assessment. This In Australia, previous experience on assessment of
product expresses a bacterial cold shock domain pro- weediness has been used to assess transgenic crops.
tein B (Bacillus subtilis CSPB), which imparts re- Identification of characteristics that are relevant to
duced yield loss under water-limited conditions weediness/invasiveness has been based on practical
compared with conventional corn (Castiglioni et al. experience with more than 1200 major environmental
2008). CSPB is a member of the cold shock domain- and agricultural weeds in diverse landscapes (Randall
containing (CSD-containing) protein family. Under 2012). Weed scientists have produced a robust and
environmental stress, CSD-containing proteins mod- simple weed risk assessment protocol that can be
erate stress responses in bacteria and plants, primarily readily applied to any plant (Keese et al. 2014). In
through stabilization of RNA and improved cellular addition, the large datasets available from weed risk
function (Cristofari and Darlix 2002; Chaikam and assessments include plants across the whole risk
Karlson 2008). Like endogenous CSD proteins found spectrum and allow rigorous validation tests to be
in bacteria and plants, the CSPB protein in MON conducted (Virtue et al. 2008; Stone and Byrne 2011).
87460 interacts with RNA and accumulates and lo- The most advanced method for weed risk assess-
calizes to rapidly growing tissues and in developing ment is based on the post-border weed risk manage-
reproductive organs, thereby helping to maintain cel- ment protocol (Auld 2012), which was developed as a
lular function in those tissues during stress (Nemali means of prioritizing existing weeds for control. It can
et al. 2014). Under water-limited conditions, there is a be adapted to risk assessment of GM plants (Keese
trend toward improved ear growth rate for MON et al. 2014) by comparing the weed risk of the GM
87460 compared with the control plants, while the plant to that of the conventional counterpart. This
common mechanisms of plant response to drought approach is used to identify significant changes based
stress are not altered in transgenic CSPB-expressing on three factors: the risk context (i.e., the environment
maize plants (Castiglioni et al. 2008; Nemali et al. where the GM crop might be present), the ability of the
2014). When plants were grown under well-watered GM plant to spread and persist, and the potential
conditions, no appreciable differences between CSPB- negative impacts on biodiversity, non-target organ-
expressing lines and the control were detected (Cas- isms, soil nutrients, etc.
tiglioni et al. 2008). The weed risk approach specifies relevant charac-
Based on the understanding of the CSPB mode of teristics of GM plants that affect spread and persis-
action, the ERA for MON 87460 included six hy- tence (invasiveness) and those that potentially give
pothesis-driven studies that answered specific ques- rise to negative impacts on human or animal health, or
tions relevant to the nature of the trait, in addition to the environment. These characteristics capture chan-
the standard phenotypic and agronomic field trials in ges due to either intended or unintended effects.
the presence and absence of the trait (Sammons et al. Changes that have no or negligible effect on weed risk
2014). The studies included assessments for persis- need not be explored. The post-border weed risk
tence outside of cultivation; root growth and devel- assessment approach therefore provides guidance on
opment; and drought, cold, heat, and salt tolerance the data requirements, for both intended and unin-
(Sammons et al. 2014). No additional abiotic stress tended traits, that are considered relevant for the ERA
tolerances were identified and no differences in of a GM plant.

11 12
Section based on presentation by Elena Rice. Section based on presentation by Paul Keese.

123
596 Transgenic Res (2015) 24:587–603

Unintended effects on non-target organisms13 (Garcia-Alonso and Raybould 2014). If differences

are detected, their likely biological relevance will be
A common concern associated with the growing of assessed by considering the range of values known for
GM crops is over their potential to have adverse im- the conventional crop varieties that have a history of
pacts on non-target organisms. Arthropods in par- safe use. The aim of this assessment is to identify
ticular form a major part of the biodiversity in potentially harmful unintended changes that, if found,
agricultural landscapes, and many are valued because would trigger a more detailed assessment (Romeis
they provide important ecosystem services, including et al. 2008). This approach is considered sufficiently
biological control, pollination, and decomposition, or conservative given the fact that more than 99 % of all
cultural services, including human enjoyment and transformation events are eliminated during prior
education (Sanvido et al. 2012; Garcia-Alonso and agronomic and phenotypic analyses (e.g., Phillips
Raybould 2014). Therefore, potential impacts that GM McDougall 2011).
plants may have on valued non-target arthropods It is sometimes argued that risk assessments should
(NTAs) are addressed in ERA. include experiments to study the impact of unintended,
Both the intended change (e.g., production of a Bt transformation-related effects on non-target organ-
Cry [crystal] protein for target insect control) and any isms. For example, the EFSA requests non-target
unintended changes could cause unintended effects on studies using GM plant material as a test substance to
valued non-target organisms. Since consequences of ‘‘… give indications on possible interactions between
the intended change can be anticipated, it is possible to plant compounds and reflect realistic exposure con-
construct conceptual models (pathways to harm) of ditions through bioavailability’’ (EFSA 2010). The
how growing of the GM plant could harm valued justification for these additional data is that the com-
NTAs and to formulate risk hypotheses that can sub- positional analyses do not necessarily target specific
sequently be tested (Raybould 2011). A common hy- metabolites known to be involved in non-target-or-
pothesis is that the stressor (i.e., the Cry toxin) does ganism–plant relationships. This approach has many
not reduce the abundance and ecological functions of limitations. For example, it is usually unknown which
NTAs under field conditions. This hypothesis is metabolites are involved in these interactions, and
typically tested within a tiered framework that moves different metabolites are likely to affect different non-
from laboratory or early-tier tests using species that target species differently. Furthermore it is more likely
are readily available, amenable to testing, and able to to detect differences in plant tissue composition
detect potential hazards, to more complex (higher-tier) among different plant varieties or even plant batches
experiments that evaluate the risks under more real- than between the tissue from GM plants and their non-
istic exposure conditions such as field studies (Romeis transformed control (e.g., Meissle et al. 2014). Such
et al. 2008). Laboratory studies (termed tier 1 tests) are experiments and their results may thus add confusion
particularly powerful for testing the risk hypothesis; in rather than certainty to the ERA. The published lit-
cases where no adverse effects are detected under erature on the non-target impact of Bt maize, for ex-
these highly controlled laboratory and worst-case ex- ample, provides a number of examples where studies
posure conditions, a ‘‘no effect’’ conclusion can be using GM plant tissue as a test substance have resulted
drawn with high confidence (Raybould et al. 2007; in inconclusive results (Romeis et al. 2013).
Romeis et al. 2011). As a solution, the unintended, transformation-re-
In the case of unintended, plant-transformation-re- lated effects that might adversely affect NTAs should
lated effects, the assessment typically follows a be identified during the problem formulation phase of
weight-of-evidence approach taking into account in- the ERA taking into account the results from the
formation from the molecular characterization of the molecular, phenotypic, and agronomic characteriza-
particular GM event and from comparisons of com- tion and the compositional analyses. If such charac-
position and agronomic and phenotypic characteristics teristics are identified as stressors of concern,
of the GM plant with its conventional counterpart pathways to harm can be constructed and testable risk
hypotheses can be formulated. This is a precondition
to design and execute meaningful studies that provide
13
Section based on presentation by Jörg Romeis. data to support the ERA.

123
Transgenic Res (2015) 24:587–603 597

Emerging technologies and application the new plant variety would be required to confirm its
to assessment of unintended effects safety. This approach complements current approaches
for targeted compositional analyses, but the information
Omics technologies14 content of omics technologies will be much greater.
Statistical and chemometric methods are available to
In contrast to the rationale described above suggesting rapidly screen profiles of new plant varieties with ref-
that a hypothesis-driven approach is sufficiently in- erence to profiles from plant varieties that we consider as
formative for risk assessments, there are calls for the safe (Van Dijk et al. 2014). To effectively implement
use of global profiling technologies to survey the plant omics technologies to improve current risk assessment
more broadly than is currently feasible using a targeted procedures, it is necessary to establish simple, common
approach. As previously noted, unintended effects can protocols for omics analyses and related data analyses
be found in both conventional and GM-based breeding with the aim to (1) compare the new plant inbred or
programs (Kok et al. 2008; Weber et al. 2012; Flint- variety to an appropriate comparator (e.g., in the case of a
Garcia 2013). New plant breeding programs are GM crop, a conventional counterpart) and (2) compare
moving toward a broader range of molecular biologi- the new plant inbred or variety to a larger set of geno-
cal breeding techniques to achieve more complex types of the same species that are considered safe. This
genetic alterations within the framework of shorter approach will often relate to data that the plant breeder
breeding programs (Lusser et al. 2011). In the future, will already have available, thus considerably reducing
other strategies that aim for even more profound the regulatory burden for plant breeders while safe-
changes, such as strategies based on synthetic biology, guarding the food supply in the future.
may be applied to plant breeding.
For these reasons, it is useful to have more infor- Genome-scale metabolic networks
mative and cost-efficient analytical methods to screen and metabolomics15
for unintended, potentially adverse, effects. Omics
technologies may meet these criteria. Omics technolo- As suggested above, omics technologies have the po-
gies, whether transcriptomics (mRNA profiling), pro- tential for enabling comprehensive and quantitative
teomics (protein profiling) or metabolomics (metabolite assessment of metabolic changes in response to ge-
profiling), have already shown their added value in netic modification. While non-targeted metabolite
different areas (Tanaka 2010; de Ligt et al. 2012; Rauch profiling can detect thousands of compounds, it is not
et al. 2012). In plant materials, these methodologies can easy to understand the significance of the changed
be applied in a reproducible and informative way metabolites in the biochemical and biological context
(Fernie and Schauer 2009; Van Dijk et al. 2010, 2012; of the organism. To derive biochemical explanations
Oms-Oliu et al. 2013). Such studies have confirmed that or hypotheses for the observed metabolite changes
differences between a single-trait transgenic plant va- from non-targeted metabolomics studies, it is impor-
riety and its conventional counterpart will typically be tant to examine the changed metabolites in the context
smaller than differences between comparable conven- of a genome-scale metabolic network of the organism.
tional varieties for the analytes and time points examined. Much progress has been made in the last few dec-
New plant varieties could be screened for any aberrant ades to represent metabolism at a genome scale
omics profile, i.e., a profile that is different from the (Thiele and Palsson 2010). The advances in genome
profiles of plant varieties considered as safe. New plant sequencing and emerging fields such as biocuration
varieties for which compositional profiles fall within the (biological database management) and bioinformatics
range of profiles derived from plant varieties already enabled the reconstruction of genome-scale metabolic
considered safe would not require further assessment. networks for model organisms (Bassel et al. 2012).
This would not mean that a plant with an altered profile is Genome-scale metabolic networks have been pre-
not safe, but in the case of a profile outside the range of dicted from reference metabolic pathway databases
varieties considered as safe, a further detailed analysis of such as MetaCyc (Caspi et al. 2012), PlantCyc (Zhang

14 15
Section based on presentation by Esther Kok. Section based on presentation by Seung Yon Rhee.

123
598 Transgenic Res (2015) 24:587–603

et al. 2010), and KEGG (Kanehisa et al. 2012). These lines. Several algorithms that have the goal of
genome-scale metabolic networks are now available minimizing the change in the metabolic network upon
for several plant species such as Arabidopsis thaliana perturbation have been developed, and appear to per-
(Mueller et al. 2003; Zhang et al. 2010), poplar form better than FBA in explaining fluxes of mutants
(Populus trichocarpa; Zhang et al. 2010), Chlamy- (Segrè et al. 2002; Shlomi et al. 2005; Herrgård et al.
domonas reinhardtii (May et al. 2009), medic (Med- 2006). This type of modeling could point to bio-
icago truncatula; Urbanczyk-Wochniak and Sumner chemical explanations for unintended or unexpected
2007), grasses (Youens-Clark et al. 2011), and night- metabolite changes, which could help devise hy-
shade plants (Bombarely et al. 2011). pothesis-driven assessment strategies.
The genome-scale metabolic networks can be used Using the genome-scale metabolic network of Ara-
to create predictive metabolic models (Sweetlove et al. bidopsis, Rhee and colleagues tested the effect of single
2008). Metabolic models can be used to predict genetic perturbations of 136 genes (129 knock-out and 7
metabolic fluxes under a variety of scenarios such as overexpression lines) by comprehensively profiling the
genetic perturbations (Feist and Palsson 2008). Two metabolites using 11 analytic platforms including GC–
types of metabolic models have been used: kinetic and MS, LC–MS, and ESI (Quanbeck et al. 2012). Com-
stoichiometric. Kinetic modeling uses enzyme kinet- parison of the metabolite profiles across the mutants
ics to numerically simulate and test metabolic fluxes showed that metabolic networks were robust to pertur-
and can explain the mechanism of flux changes, but bations of single metabolic genes and the genetic per-
the difficulty of determining in vivo enzyme kinetics turbations changed the network more locally than
has limited this modeling to a small number of path- globally (Kim and Rhee, unpublished results). This study
ways. The other modeling approach uses the stoi- revealed relationships between characteristics of the
chiometry of reactants and products in reactions to perturbed genes and metabolic changes. More analyses
solve for the most likely fluxes with added constraints of this type would help in identifying the relationships
based on thermodynamics, directionality, and flux between changed metabolites and their potential impact
capacity of reactions (Thiele and Palsson 2010). This on the metabolic system and biology of the organism,
approach has been used to build genome-scale which in turn would inform if altered composition could
metabolic models of several plant species such as have toxic or other harmful effects for food and feed
Arabidopsis (Poolman et al. 2009; de Oliveira Dal’- safety in any given crop.
Molin et al. 2010), maize (Saha et al. 2011), and C.
reinhardtii (Chang et al. 2011). Most of these models
have not been validated extensively using flux mea- Conclusions
surements, though advances in metabolic flux analysis
using 13C-labeling and metabolomics approaches hold The meeting summarized here was intended to present
promise (Schwender 2008; Sweetlove et al. 2008; and discuss a broad range of viewpoints, rather than to
Allen et al. 2009). arrive at a consensus on particular points. Neverthe-
These metabolic models have been applied in a less, several themes recurred in a number of presen-
variety of studies ranging from metabolic engineering, tations. For example, there seemed to be little
drug discovery, drug target discovery, identification of disagreement with the observation that no system for
novel gene function, evolutionary processes, network genetic modification, including conventional methods
behaviors, and interpretations of mutant phenotypes of plant breeding, is without unintended effects. It was
(Feist and Palsson 2008). The most common algorithm also commonly observed that ‘‘unintended’’ is not
used in these studies has been flux balance analysis synonymous with ‘‘harmful’’. The testing methods
(FBA), which attempts to balance the stoichiometry of used to identify unintended effects from transgene
the metabolites within the metabolic network with a introduction are based on analyzing the agronomic
goal (objective function) of maximal growth rate or performance of the crop and composition of the har-
maximal biomass accumulation. While flux predic- vested parts (e.g., grain, fruit, or forage). This testing,
tions from FBA match experimental data reasonably in combination with hypothesis-based testing of the
well (Burgard and Maranas 2003), its assumptions effects and potential safety issues associated with
may not always hold true, especially for GM or mutant transgene expression minimizes the likelihood of

123
Transgenic Res (2015) 24:587–603 599

unsafe unintended effects associated with the GM crop Auld B (2012) An overview of pre-border weed risk assessment
products that are taken to commercialization. and post-border weed risk management protocols. Plant
Prot Q 7:105–111
The types of changes, such as genomic disruption Bassel GW, Gaudinier A, Brady SM, Hennig L, Rhee SY, Smet
caused by (trans)gene insertion, that were once seen as ID (2012) Systems analysis of plant functional, transcrip-
capable of leading to the production of unintended ef- tional, physical interaction, and metabolic networks. Plant
fects (e.g., in the form of novel toxins [Kessler et al. Cell 24:3859–3875
Bolhaar ST, van de Weg WE, van Ree R, Gonzalez-Mancebo E,
1992]) turn out to be routine occurrences during all types Zuidmeer L, Bruijnzeel-Koomen CA, Fernandez-Rivas M,
of plant breeding and are ubiquitous in crop plants. De- Jansen J, Hoffmann-Sommergruber K, Knulst AC, Gilissen
spite the ongoing presence of these changes during plant LJ (2005) In vivo assessment with prick-to-prick testing
breeding and selection, there is not a single documented and double-blind, placebo-controlled food challenge of
allergenicity of apple cultivars. J Allergy Clin Immunol
example whereby these changes have led to the pro- 116:1080–1086
duction of previously unknown toxins. All reported cases Bombarely A, Menda N, Tecle IY, Buels RM, Strickler S,
of crop toxicity have been associated with the inadver- Fischer-York T, Pujar A, Leto J, Gosselin J, Mueller LA
tent elevation of known toxins, such that testing for their (2011) The Sol Genomics Network (solgenomics.net):
growing tomatoes using Perl. Nucleic Acids Res 39(Data-
presence has become a part of the breeding process in base issue):D1149–D1155
some crops to prevent inadvertent increases in toxin Bovy A, de Vos R, Kemper M, Schijlen E, Pertejo MA, Muir S,
levels (Steiner et al. 2013). Collins G, Robinson S, Verhoeyen M, Hughes S, Santos-
Emerging tools and resources such as genome-scale Buelga C, van Tunen A (2002) High-flavonol tomatoes
resulting from the heterologous expression of the maize
metabolic networks, quantitative network modeling, and transcription factor genes LC and C1. Plant Cell
metabolomics may help assess the effects of genetic 14:2509–2526
modification on metabolism and may facilitate rational Bregitzer P, Dahleen LS (2008) A single backcross effectively
assessment of potential unintended effects of genetic eliminates agronomic and quality alterations caused by
somaclonal variation in barley. Crop Sci 48:471–479
modification on metabolism. The uses of networks and Bregitzer P, Halbert SE, Lemaux PG (1998) Somaclonal
other omics-based technologies are still under assess- variation in the progeny of transgenic barley. Theor Appl
ment and are viewed as possible means of identifying Genet 96:421–425
potential unintended effects not tested by targeted ap- Burgard AP, Maranas CD (2003) Optimization-based frame-
work for inferring and testing hypothesized metabolic
proaches. As discussed during the meeting, there is not objective functions. Biotechnol Bioeng 82:670–677
yet agreement as to whether the analytical and statistical Caspi R, Altman T, Dreher K, Fulcher CA, Subhraveti P,
methods currently available are sufficient to determine Keseler IM, Kothari A, Krummenacker M, Latendresse M,
whether the profile of a new variety is meaningfully Mueller LA, Ong Q, Paley S, Pujar A, Shearer AG, Travers
M, Weerasinghe D, Zhang P, Karp PD (2012) The
different (either globally or in the behavior of individual MetaCyc database of metabolic pathways and enzymes
data points) from those of existing varieties. and the BioCyc collection of pathway/genome databases.
Nucleic Acids Res 40(Database issue):D742–D753
Acknowledgments The authors gratefully acknowledge all Castiglioni P, Warner D, Bensen RJ, Anstrom DC, Harrison J,
speakers and participants for their contributions to the January Stoecker M, Abad M, Kumar G, Salvador S, D’Ordine R,
2014 meeting in Ottawa, Canada, on the Genetic Basis of Unin- Navarro S, Back S, Fernandes M, Targolli J, Dasgupta S,
tended Effects in Modified Plants. The authors also acknowledge Bonin C, Luethy MH, Heard JE (2008) Bacterial RNA
the expert technical assistance of Virginia M. Peschke (Oakside chaperones confer abiotic stress tolerance in plants and
Editorial Services) in the preparation of this paper. improved grain yield in maize under water-limited condi-
tions. Plant Physiol 147:446–455
Open Access This article is distributed under the terms of the Cellini F, Chesson A, Colquhoun I, Constable A, Davies HV,
Creative Commons Attribution License which permits any use, Engel KH, Gatehouse AM, Kärenlampi S, Kok EJ, Leguay
distribution, and reproduction in any medium, provided the JJ, Lehesranta S, Noteborn HP, Pedersen J, Smith M (2004)
original author(s) and the source are credited. Unintended effects and their detection in genetically
modified crops. Food Chem Toxicol 42:1089–1125
Chaikam V, Karlson D (2008) Functional characterization of
two cold shock domain proteins from Oryza sativa. Plant
Cell Environ 31:995–1006
References Chang RL, Ghamsari L, Manichaikul A, Hom EF, Balaji S, Fu
W, Shen Y, Hao T, Palsson BØ, Salehi-Ashtiani K, Papin
Allen DK, Libourel IG, Shachar-Hill Y (2009) Metabolic flux JA (2011) Metabolic network reconstruction of Chlamy-
analysis in plants: coping with complexity. Plant Cell domonas offers insight into light-driven algal metabolism.
Environ 32:1241–1257 Mol Syst Biol 7:518

123
600 Transgenic Res (2015) 24:587–603

Codex Alimentarius (2003) Guideline for the conduct of food Goodman RE, Panda R, Ariyarathna H (2013) Evaluation of
safety assessment of foods derived from recombinant- endogenous allergens for the safety evaluation of ge-
DNA plants. CAC/GL 45-2003 netically engineered food crops: review of potential risks,
Craig W, Tepfer M, Degrassi G, Ripandelli D (2008) An test methods, examples and relevance. J Agric Food Chem
overview of general features of risk assessments of 61:8317–8332
genetically modified crops. Euphytica 164:853–860 Gray A (2012) Problem formulation in environmental risk
Cristofari G, Darlix JL (2002) The ubiquitous nature of RNA assessment for genetically modified crops: a practitioner’s
chaperone proteins. Prog Nucleic Acid Res Mol Biol approach. Colln Biosaf Rev 6:10–65
72:223–268 Hammond B, Kough J, Herouet-Guicheney C, Jez JM (2013)
de Ligt J, Willemsen MH, van Bon BW, Kleefstra T, Yntema Toxicological evaluation of proteins introduced into food
HG, Kroes T, Vulto-van Silfhout AT, Koolen DA, de Vries crops. Crit Rev Toxicol 43(Suppl 2):25–42
P, Gilissen C, del Rosario M, Hoischen A, Scheffer H, de Herrgård MJ, Fong SS, Palsson BØ (2006) Identification of
Vries BB, Brunner HG, Veltman JA, Vissers LE (2012) genome-scale metabolic network models using experimen-
Diagnostic exome sequencing in persons with severe tally measured flux profiles. PLoS Comput Biol 2(7):e72
intellectual disability. N Engl J Med 367:1921–1929 Kanehisa M, Goto S, Sato Y, Furumichi M, Tanabe M (2012)
de Oliveira Dal’Molin CG, Quek LE, Palfreyman RW, KEGG for integration and interpretation of large-scale
Brumbley SM, Nielsen LK (2010) AraGEM, a genome- molecular data sets. Nucleic Acids Res 40(Database
scale reconstruction of the primary metabolic network in issue):D109–D114
Arabidopsis. Plant Physiol 152:579–589 Keese PK, Robold AV, Myers RC, Weisman S, Smith J (2014)
EC (2013) Commission Implementing Regulation (EU) No Applying a weed risk assessment approach to GM crops.
503/2013 of 3 April 2013 on applications for authorisation Transgenic Res 23:957–969
of genetically modified food and feed in accordance with Kessler DA, Taylor MR, Maryanski JH, Flamm EL, Kahl LS
Regulation (EC) No 1829/2003 of the European Parliament (1992) The safety of foods developed by biotechnology.
and of the Council and amending Commission Regulations Science 256:1747–1749
(EC) No 641/2004 and (EC) No 1981/2006. Off J Eur Kok EJ, Keijer J, Kleter GA, Kuiper HA (2008) Comparative
Union L157:1–48 safety assessment of plant-derived foods. Regul Toxicol
EFSA (2010) Guidance on the environmental risk assessment of Pharmacol 50:98–113
genetically modified plants. EFSA J 8:1879 Krattinger SG, Lagudah ES, Spielmeyer W, Singh RP, Huerta-
EFSA (2011) EFSA Panel on Genetically Modified Organisms Espino J, McFadden H, Bossolini E, Selter LL, Keller B
(GMO). Guidance for risk assessment of food and feed (2009) A putative ABC transporter confers durable resis-
from genetically modified plants. EFSA J 9:2150 tance to multiple fungal pathogens in wheat. Science
Evans J, Wood G, Miller M (2006) The risk assessment–policy 323:1360–1363
gap: an example from the UK contaminated land regime. Lai J, Li R, Xu X, Jin W, Xu M, Zhao H, Xiang Z, Song W, Ying
Environ Int 32:1066–1071 K, Zhang M, Jiao Y, Ni P, Zhang J, Li D, Guo X, Ye K, Jian
Fagan J, Antoniou M, Robinson C (2014) GMO Myths and Truths, M, Wang B, Zheng H, Liang H, Zhang X, Wang S, Chen S,
2nd ed. Earth Open Source, London. http://www.earth Li J, Fu Y, Springer NM, Yang H, Wang J, Dai J, Schnable
opensource.org/index.php/reports/gmo-myths-and-truths PS, Wang J (2010) Genome-wide patterns of genetic
Feist AM, Palsson BØ (2008) The growing scope of applications variation among elite maize inbred lines. Nat Genet
of genome-scale metabolic reconstructions using Escher- 42:1027–1030
ichia coli. Nat Biotechnol 26:659–667 Lam HM, Xu X, Liu X, Chen W, Yang G, Wong FL, Li MW, He
Fernandez A, Mills ENC, Lovik M, Spoek A, Germini A, W, Qin N, Wang B, Li J, Jian M, Wang J, Shao G, Wang J,
Mikalsen A, Wal JM (2013) Endogenous allergens and Sun SS, Zhang G (2010) Resequencing of 31 wild and
compositional analysis in the allergenicity assessment of cultivated soybean genomes identifies patterns of genetic
genetically modified plants. Food Chem Toxicol 62:1–6 diversity and selection. Nat Genet 42:1053–1059
Fernie AR, Schauer N (2009) Metabolomics-assisted breeding: a Larkin PJ, Scowcroft WR (1981) Somaclonal variation: a novel
viable option for crop improvement? Trends Genet 25:39–48 source of variability from cell cultures for plant improve-
Flint-Garcia SA (2013) Genetics and consequences of crop ment. Theor Appl Genet 60:197–214
domestication. J Agric Food Chem 61:8267–8276 Liu R, Koyanagi KO, Chen S, Kishima Y (2012) Evolutionary
Foetisch K, Dahl L, Jansen B, Becker WM, Lidholm J, van Ree force of AT-rich repeats to trap genomic and episomal
R, Broll H, Kaul S, Vieths S, Holzhauser T (2011) DNAs into the rice genome: lessons from endogenous
Development and in-house validation of allergen-specific pararetrovirus. Plant J 72:817–828
ELISA tests for the quantification of Dau c 1.01, Dau c 1.02 Lloyd AH, Wang D, Timmis JN (2012) Single molecule PCR
and Dau c 4 in carrot extracts (Daucus carota). Anal reveals similar patterns of non-homologous DSB repair in
Bioanal Chem 399:935–943 tobacco and Arabidopsis. PLoS One 7:e32255
Forsbach A, Schubert D, Lechtenberg B, Gils M, Schmidt R Lusser M, Parisi C, Plan D, Rodrı́guez-Cerezo E (2011) New
(2003) A comprehensive characterization of single-copy plant breeding techniques: state-of-the-art and prospects
T-DNA insertions in the Arabidopsis thaliana genome. for commercial development. IPTS report, JRC 63971,
Plant Mol Biol 52:161–176 EUR 24760 EN. doi:10.2791/54761. Accessed 20 Oct 2014
Garcia-Alonso M, Raybould A (2014) Protection goals in May P, Christian JO, Kempa S, Walther D (2009) ChlamyCyc:
environmental risk assessment: a practical approach. an integrative systems biology database and web-portal for
Transgenic Res 23:945–956 Chlamydomonas reinhardtii. BMC Genomics 10:209

123
Transgenic Res (2015) 24:587–603 601

McClintock B (1984) The significance of responses of the Privalle LS, Chen J, Clapper G, Hunst P, Spiegelhalter F, Zhong
genome to challenge. Science 226:792–801 CX (2012) Development of an agricultural biotechnology
McHale LK, Haun WJ, Xu WW, Bhaskar PB, Anderson JE, crop product: testing from discovery to commercialization.
Hyten DL, Gerhardt DJ, Jeddeloh JA, Stupar RM (2012) J Agric Food Chem 60:10179–10187
Structural variants in the soybean genome localize to Quanbeck SM, Brachova L, Campbell AA, Guan X, Perera A,
clusters of biotic stress-response genes. Plant Physiol He K, Rhee SY, Preeti Bais P, Dickerson JA, Dixon P,
159:1295–1308 Wohlgemuth G, Fiehn O, Barkan L, Lange I, Lange BM,
Meissle M, Zünd J, Waldburger M, Romeis J (2014) Develop- Lee I, Cortes D, Salazar C, Shuman J, Shulaev V, Huhman
ment of Chrysoperla carnea (Stephens) (Neuroptera: DV, Sumner LW, Roth MR, Welti R, Ilarslan H, Wurtele
Chrysopidae) on pollen from Bt-transgenic and conven- ES, Nikolau BJ (2012) Metabolomics as a hypothesis-
tional maize. Sci Rep 4:5900 generating functional genomics tool for the annotation of
Metcalfe DD, Astwood JD, Townsend R, Sampson HA, Taylor Arabidopsis thaliana genes of ‘‘unknown function’’. Front
SL, Fuchs RL (1996) Assessment of the allergenic potential Plant Sci 3:1–12
of foods derived from genetically engineered crop plants. Crit Randall RP (2012) A global compendium of weeds, 2nd edn.
Rev Food Sci Nutr 36:S165–S186 Department of Agriculture and Food, Western Australia,
Mueller LA, Zhang P, Rhee SY (2003) AraCyc: a biochemical Perth
pathway database for Arabidopsis. Plant Physiol 132: Rauch A, Wieczorek D, Graf E, Wieland T, Endele S,
453–460 Schwarzmayr T, Albrecht B, Bartholdi D, Beygo J, Di
Naito K, Cho E, Yang G, Campbell MA, Yano K, Okumoto Y, Donato N, Dufke A, Cremer K, Hempel M, Horn D, Hoyer
Tanisaka T, Wessler SR (2006) Dramatic amplification of a J, Joset P, Röpke A, Moog U, Riess A, Thiel CT, Tzschach
rice transposable element during recent domestication. A, Wiesener A, Wohlleber E, Zweier C, Ekici AB, Zink
Proc Natl Acad Sci USA 47:17620–17625 AM, Rump A, Meisinger C, Grallert H, Sticht H, Schenck
Nemali KS, Bonin C, Dohleman FG, Stephens M, Reeves WR, A, Engels H, Rappold G, Schröck E, Wieacker P, Riess O,
Nelson DE, Whitsel JE, Sammons B, Silady RA, Anstrom Meitinger T, Reis A, Strom TM (2012) Range of genetic
D, Sharp RE, Patharkar OR, Clay D, Coffin M, Nemeth mutations associated with severe non-syndromic sporadic
MA, Leibman ME, Luethy M, Lawson M (2014) Physio- intellectual disability: an exome sequencing study. Lancet
logical responses related to increased grain yield under 380:1674–1682
drought in the first biotechnology-derived drought tolerant Raybould A (2011) The bucket and the searchlight: formulating
maize. Plant Cell Environ. doi:10.1111/pce.12446 and testing risk hypotheses about the weediness and
NRC (2004) Safety of genetically engineered foods: approaches invasiveness potential of transgenic crops. Environ Biosaf
to assessing unintended health effects. Chapter 3. Unin- Res 9:123–133
tended effects from breeding, pp 39–71. Committee on Raybould A, Stacey D, Vlachos D, Graser G, Li X, Joseph R
Identifying and Assessing Unintended Effects of Ge- (2007) Non-target organism risk assessment of MIR604
netically Engineered Foods on Human Health, National maize expressing mCry3A for control of corn rootworm.
Research Council. National Academies Press, Washington J Appl Entomol 131:391–399
Oms-Oliu G, Odriozola-Serrano I, Martı́n-Belloso O (2013) Risk JM, Selter LL, Chauhan H, Krattinger SG, Kumlehn J,
Metabolomics for assessing safety and quality of plant- Hensel G, Viccars LA, Richardson TM, Buesing G, Troller
derived food. Food Res Int 54:1172–1183 A, Lagudah ES, Keller B (2013) The wheat Lr34 gene
Ossowski S, Schneeberger K, Lucas-Lledó JI, Warthmann N, provides resistance against multiple fungal pathogens in
Clark RM, Shaw RG, Weigel D, Lynch M (2010) The rate barley. Plant Biotechnol J 11:847–854
and molecular spectrum of spontaneous mutations in Romeis J, Bartsch D, Bigler F, Candolfi MP, Gielkens MMC,
Arabidopsis thaliana. Science 327:92–94 Hartley SE, Hellmich RL, Huesing JE, Jepson PC, Layton
Paes de Andrade P, Parrott W, Mercedes Roca M (2014) R, Quemada H, Raybould A, Rose RI, Schiemann J, Sears
Genetically modified organisms—environmental risk MK, Shelton AM, Sweet J, Vaituzis Z, Wolt JD (2008)
assessment guide, 2nd edn. ILSI Brasil. http://www.ilsi. Assessment of risk of insect-resistant transgenic crops to
org/Brasil/Pages/ViewItemDetails.aspx?WebId=C34AB3 nontarget arthropods. Nat Biotechnol 26:203–208
F5-C89B-49B3-9740-31F407A2A6FD&ListId=91D4243D- Romeis J, Hellmich RL, Candolfi MP, Carstens K, De Schrijver
A11D-4CB9-B694-551373D9E8C5&ItemID=73. Accessed A, Gatehouse AMR, Herman RA, Huesing JE, McLean
14 Aug 2014 MA, Raybould A, Shelton AM, Waggoner A (2011)
Phillips McDougall (2011) The cost and time involved in the Recommendations for the design of laboratory studies on
discovery, development and authorisation of a new plant non-target arthropods for risk assessment of genetically
biotechnology derived trait. http://croplife.org/plant-bio engineered plants. Transgenic Res 20:1–22
technology/regulatory-2/cost-of-bringing-a-biotech-crop- Romeis J, McLean MA, Shelton AM (2013) When bad science
to-market/. Accessed 14 Aug 2014 makes good headlines: Bt maize and regulatory bans. Nat
Poolman MG, Miguet L, Sweetlove LJ, Fell DA (2009) A Biotechnol 31:386–387
genome-scale metabolic model of Arabidopsis and some of Saha R, Suthers PF, Maranas CD (2011) Zea mays iRS1563: a
its properties. Plant Physiol 151:1570–1581 comprehensive genome-scale metabolic reconstruction of
Potato Genome Sequencing Consortium (2011) Genome maize metabolism. PLoS One 6(7):e21784
sequence and analysis of the tuber crop potato. Nature Sammons B, Whitsel J, Stork LAG, Reeves W, Horak M (2014)
475:189–195 Characterization of drought-tolerant maize MON 87460

123
602 Transgenic Res (2015) 24:587–603

for use in environmental risk assessment. Crop Sci 54: along chromosome 1 of maize (Zea mays ssp. mays L.).
719–729 Proc Natl Acad Sci USA 98:9161–9166
Sancho AI, Foxall R, Browne T, Dey R, Zuidmeer L, Marzban Tepfer M, Racovita M, Craig W (2013) Putting problem
G, Waldron KW, van Ree R, Hoffmann-Sommergruber K, formulation at the forefront of GMO risk analysis. GM
Laimer M, Mills EN (2006a) Effect of postharvest storage Crops Food 4:10–15
on the expression of the apple allergen Mal d 1. J Agric Thiele I, Palsson BØ (2010) A protocol for generating a high-
Food Chem 54:5917–5923 quality genome-scale metabolic reconstruction. Nat Protoc
Sancho AI, Foxall R, Rigby NM, Browne T, Zuidmeer L, van 5:93–121
Ree R, Waldron KW, Mills EN (2006b) Maturity and Tian Z, Zhao M, She M, Du J, Cannon SB, Liu X, Xu X, Qi X, Li
storage influence on the apple (Malus domestica) allergen MW, Lam HM, Ma J (2012) Genome-wide characteriza-
Mal d 3, a nonspecific lipid transfer protein. J Agric Food tion of nonreference transposons reveals evolutionary
Chem 54:5098–5104 propensities of transposons in soybean. Plant Cell 24:
Sanvido O, Romeis J, Gathmann A, Gielkens M, Raybould A, 4422–4436
Bigler F (2012) Evaluating environmental risks of ge- Tollenaar M, Lee EA (2010) Strategies for enhancing grain
netically modified crops—ecological harm criteria for yield in maize. In: Janick J (ed) Plant breeding reviews, vol
regulatory decision-making. Environ Sci Policy 15:82–91 34. Wiley, Hoboken
Sarevitz D (2004) How science makes environmental contro- Urbanczyk-Wochniak E, Sumner LW (2007) MedicCyc: a
versies worse. Environ Sci Policy 7:385–403 biochemical pathway database for Medicago truncatula.
SCBD (2000) Cartagena Protocol on Biosafety to the Conven- Bioinformatics 23:1418–1423
tion on Biological Diversity: annex III, risk assessment. van Dijk JP, Leifert C, Barros E, Kok EJ (2010) Gene expression
Montreal, Canada. http://bch.cbd.int/protocol/text/. Ac- profiling for food safety assessment: examples in potato
cessed 14 Jan 2015 and maize. Regul Toxicol Pharmacol 58:S21–S25
Schiessl S, Samans B, Hüttel B, Reinhardt R, Snowdon RJ van Dijk JP, Cankar K, Hendriksen PJM, Beenen HG, Zhu M,
(2014) Capturing sequence variation among flowering- Scheffer S, Shepherd LVT, Stewart D, Davies HV, Leifert
time regulatory gene homologs in the allopolyploid crop C, Wilkockson SJ, Gruden K, Kok EJ (2012) The
species Brassica napus. Front Plant Sci. doi:10.3389/fpls. identification and interpretation of differences in the
2014.00404 transcriptomes of organically and conventionally grown
Schnell J, Steele M, Bean J, Neuspiel M, Girard C, Dormann N, potato tubers. J Agric Food Chem 60:2090–2101
Pearson C, Savoie A, Bourbonnière L, Macdonald P (2015) van Dijk JP, de Mello CS, Voorhuijzen MM, Hutten RCB, Arisi
A comparative analysis of insertional effects in genetically ACM, Jansen JJ, Buydens LMC, van der Voet H, Kok EJ
engineered plants: considerations for pre-market assess- (2014) Safety assessment of plant varieties using tran-
ments. Transgenic Res 24:1–17. doi:10.1007/s11248-014- scriptomics profiling and a one-class classifier. Regul
9843-7 Toxicol Pharmacol 70:297–303
Schwender J (2008) Metabolic flux analysis as a tool in Vaughn JN, Bennetzen JL (2014) Natural insertions in rice
metabolic engineering of plants. Curr Opin Biotechnol 19: commonly form tandem duplications indicative of patch-
131–137 mediated double-strand break induction and repair. Proc
Segrè D, Vitkup D, Church GM (2002) Analysis of optimality in Natl Acad Sci USA 111:6684–6689. doi:10.1073/pnas.
natural and perturbed metabolic networks. Proc Natl Acad 1321854111
Sci USA 99:15112–15117 Virtue JG, Spencer JE, Weiss JE, Reichard SE (2008)
Shlomi T, Berkman O, Ruppin E (2005) Regulatory on/off Australia’s Botanic Gardens weed risk assessment proce-
minimization of metabolic flux changes after genetic dure. Plant Prot Q 23:166–178
perturbations. Proc Natl Acad Sci USA 102:7695–7700 Vu GT, Cao HX, Watanabe K, Hensel G, Blattner FR, Kumlehn
Staginnus C, Gregor W, Mette MF, Teo CH, Borroto-Fernández J, Schubert I (2014) Repair of site-specific DNA double-
EG, Laimer da Câmara Machado M, Matzke M, Sch- strand breaks in barley occurs via diverse pathways
warzacher T (2007) Endogenous pararetroviral sequences primarily involving the sister chromatid. Plant Cell 26:
in tomato (Solanum lycopersicum) and related species. 2156–2167
BMC Plant Biol 7:24 Weber N, Halpin C, Hannah LC, Jez J, Kough J, Parrott W
Steiner HY, Halpin C, Jez JM, Kough J, Parrott W, Underhill L, (2012) Crop genome plasticity and its relevance to food
Weber N, Hannah LC (2013) Evaluating the potential for and feed safety of genetically engineered breeding stacks.
adverse interactions within genetically engineered breed- Plant Physiol 160:1842–1853
ing stacks. Plant Physiol 161:1587–1594 Wolt JD, Keese P, Raybould A, Fitzpatrick JW, Burachik M,
Stone LM, Byrne M (2011) Comparing the outputs of five weed Gray A, Olin SS, Schiemann J, Sears M, Wu F (2010)
risk assessment models implemented in Australia: are there Problem formulation in the environmental risk assessment
consistencies across models? Plant Prot Q 26:29–35 for genetically modified plants. Transgenic Res 19:425–
Sweetlove LJ, Fell D, Fernie AR (2008) Getting to grips with the 436
plant metabolic network. Biochem J 409:27–41 Youens-Clark K, Buckler E, Casstevens T, Chen C, Declerck G,
Tanaka H (2010) Omics-based medicine and systems pathology: Derwent P, Dharmawardhana P, Jaiswal P, Kersey P,
a new perspective for personalized and predictive medi- Karthikeyan AS, Lu J, McCouch SR, Ren L, Spooner W,
cine. Methods Inf Med 49:173–185 Stein JC, Thomason J, Wei S, Ware D (2011) Gramene
Tenaillon MI, Sawkins MC, Long AD, Gaut RL, Doebley JF, database in 2010: updates and extensions. Nucleic Acids
Gaut BS (2001) Patterns of DNA sequence polymorphism Res 39(Database issue):D1085–D1094

123
Transgenic Res (2015) 24:587–603 603

Zhang P, Dreher K, Karthikeyan A, Chi A, Pujar A, Caspi R, Karp Zuidmeer L, Goldhahn K, Rona RJ, Gislason D, Madsen C,
P, Kirkup V, Latendresse M, Lee C, Mueller LA, Muller R, Summers C, Sodergren E, Dahlstrom J, Lindner T,
Rhee SY (2010) Creation of a genome-wide metabolic Sigurdardottir ST, McBride D, Keil T (2008) The preva-
pathway database for Populus trichocarpa using a new lence of plant food allergies: a systematic review. J Allergy
approach for reconstruction and curation of metabolic Clin Immunol 121:1210–1218
pathways for plants. Plant Physiol 153:1479–1491

123