EVALUAS KEAMANAN MAKANAN YANG DIMODIFIKASI SECARA GENETIK

The Safety Evaluation of Genetically Modified Foods

Introduction

The use of gene technology in food produc-tion is the cause of significant controversy that has been
fuelled by the activities of various pressure groups and by the nature of some media reporting. This
heightened con-cern has led to a complex situation where the discussion of genetically modified (GM)
food raises a whole series of social issues that go well beyond the strictly scientific assessment of
safety. This chapter is intended to high-light those safety issues that are addressed when an objective
scientific evaluation of GM foods is undertaken. Discussion is focused on health risks to the
consumer, and issues that are associated with agricultural practice are not considered. In addition, the
frequently raised question of benefit versus risk and the subtleties of risk perception are not covered.
In this regard, the Agriculture and Environ-ment Biotechnology Commission Report Crops on Trial
(AECB, 2001) provides a useful per-spective on the ethical and social impact of and public attitudes
towards GM technology.

The safety assessment of GM food addresses a series of well established and internationally accepted
questions. However, there are differences in implementation, most notably between Europe and the
USA. In Europe, GM food legislation is dominated by EC Regulation 258/97 on Novel foods and
ingredients. This legislation (EC, 1997a) demands a formal process of pre-market approval that
draws on the opinions of independent scientific committees in each Member State. In practice, the
competent authority of the country where a product is first intended to be marketed undertakes the
initial safety assessment. All other European Member States then have an opportunity to comment on
the initial opinion. Where safety concerns are raised, scientific evaluation is passed on centrally to the
Scientific Commit-tee for Food. Some novel foodstuffs, includ-ing GM soybean derivatives, were
already marketed in Europe before this legislation came into force, but they were subject to ear-lier
evaluation by national safety committees, notably by the Advisory Committee on Novel Foods and
Processes (ACNFP) in the UK (ACNFP, 1994).

In the USA, the Food and Drug Adminis-tration (FDA) holds the prime responsibility for GM food
safety under the Federal Food, Drug and Cosmetic Act. The FDA policy on foods developed by
biotechnology is outlined in a policy statement made in 1992 (FDA, 1992). While US regulatory
involvement is less hands-on, the scientific principles used for safety evaluation have much in
common with those of the EU. A recent review of the regulatory process in the USA has led to a
requirement for data submission 120 days prior to the marketing of bioengineered foods (FDA,
2001a), and updated extended guide-lines have been announced (FDA, 2001b).

Application of GM Technology in Food

GM technology is applied to food in a wide variety of different ways. Food production involves the
use of many ingredients, additives and enzyme preparations that are derived from a diversity of
sources. These include microorganisms that are often exploited as cell factories for the production of
food ingredients and processing aids. Microorganisms used in this way are often subject to genetic
improvement, but they are used to produce purified food components that are often nature identical.
Calf chymosin, the active ingredient of rennet, is used exten-sively by the dairy industry in cheese
making and provides a good example of a GM cell factory application. A gene for the chymosin
enzyme has been introduced into the yeast Kluveromyces lactis, the filamentous fungus Aspergillus
niger and the bacterium Escherichia coli, and its expression in these heterologous hosts provides an
alternative source to tradi-tional rennet extracted from the stomachs of slaughtered calves. This
represents one of the earliest commercial applications of GM food technology and it has been subject
to detailed safety evaluation both in the USA and in Europe.
Another example of a microbial cell factory that currently is awaiting EU safety approval is the use of
a GM strain of Bacillus subtilis for the manufacture of the vitamin riboflavin (ACNFP, 1997). This is
distinct in that the vitamin is a metabolite rather than the direct protein product of the introduced gene,
and it is representative of another generic application of GM technology in metabolic engineering.

Traditional biotechnology includes the exploitation of microorganisms for food fer-mentation, and
this has evolved into a major sector of the food industry. This covers the use of yeasts in brewing and
bread making, and the use of lactic acid bacteria in dairying and the production of fermented
vegetables and meats. GM technology has been used for the development of improved strains of these
food fermentation microorganisms. Examples include a bread-making strain of Sac-charomyces
cerevisiae with improved maltose metabolism and a brewing strain of the same species that expresses
an amylase gene derived from another closely related Saccharo-myces species. In both cases, safety
evaluation has been undertaken by the ACNFP (ACNFP, 1994), but food products based on the use of
these GM yeast strains have never been marketed in Europe. Because of the manu-facturing methods
used for bread and beer, the derived food products would not contain viable GM yeast cells or, in the
case of a live beer, only very low numbers. In contrast, some dairy products, such as yoghurt, contain
high numbers of viable lactic acid bacteria and, as a consequence, viable GM derivatives of these
microorganisms would also be pres-ent. These distinctions are important for safety evaluation, and the
particular issues raised by viable GM microorganisms in food have been the subject of detailed
consideration by a variety of organizations including ILSI (1999) and, most recently, FAO/WHO
(2001a).

Crop plants account for the vast majority of current GM food applications. GM plant material might
be marketed as an intact food, as in the case of fresh fruits, or it might be subject to processing, as in
the case of canned tomato paste. The safety evaluations of both fresh GM tomato fruit and canned
GM tomato paste have been undertaken (ACNFP, 1995, 1996). Genetic modification involved the
introduction of an antisense or sense gene for the natural tomato pectinase enzyme. Expression of
the engineered gene causes reduced pectinase activity during the fruit ripening process with
consequent reduced texture change. This lessens mechanical dam-age to the ripe fruit and reduces the
associated spoilage. In addition, altered processing char-acteristics improve the yield of tomato paste,
with associated economic benefits.

Commodity crops, such as soybean, are used for the production of a range of purified or semi-purified
derivatives that are added to processed foods. The extraction of oils, flours and other food ingredients
from commodity crops has implications for safety evaluation in that transgene DNA and expressed
proteins are subject to varying degrees of inactivation and removal. In addition, it is pertinent that
these derivatives are used extensively in pro-cessed foods, making consumer exposure very
widespread.

In many examples of GM food, the tech-nology has been used to improve agricultural performance.
The most widespread applica-tions are the introduction of genes that confer tolerance to otherwise
non-specific herbicides and the gene for a Bacillus thuringiensis insecti-cidal protein that provides in-
built protection from pest damage. While GM food technol-ogy currently is dominated by these first-
generation agronomic applications, there is a considerable effort devoted to second-generation traits
that are intended to benefit the food manufacturer or the consumer. Examples of second-generation
traits include improvements to seed storage proteins, oil content and starch, the removal of allergens
and the fortification of micronutrients and antioxidants. A good example is the develop-ment of GM
yellow rice in which three genes for the biosynthesis of -carotene were intro-duced into a
conventional rice cultivar. As -carotene is the provitamin for vitamin A, this has the potential to
address a serious nutrient deficiency that causes blindness in many children in developing countries.
This very brief overview of GM food applications inevitably will be incomplete, but critically it
serves to emphasize the diversity of GM technology uses in food. It establishes the need for case-by-
case consideration of safety, and this is already well established as a key guiding principle in safety
assessment. It is particularly relevant that purification and processing, such as heat treatment, may
inactivate or remove introduced foreign genes and their expressed proteins, with obvious implications
for safety assessment. Some examples of the diversity of GM food applications that have been the
subject of safety evaluation are listed in Box 15.1, and a more complete position is readily available
from the EC release, Facts on GMOs in the EU (EC, 2000).

GM Technology

The in vitro manipulation of DNA is common to all applications of GM technology. For each target
material, there is a need to use a specific technology for DNA delivery and to ensure its subsequent
maintenance. This can vary depending on the target species that is the subject of genetic modification
and it has implications for safety evaluation. One of the central safety assessment criteria is an analy-
sis of the DNA that has been introduced as well as the methods that have been employed during
transgene delivery. In bacteria, gene technology is generally more advanced and controllable than is
the case for plants. The relatively small genome size and the avail-ability of an increasing number of
whole bacterial genome sequences provide a valuable pool of detailed information. An important
advantage is the ease with which homologous recombination can be used to facilitate the directed
delivery of transgenes to specific genome sites. In general, it is pos-sible to devise food-compatible
genetic sys-tems for the exploitation of GM technology, using both plasmid-based and chromosome
integration systems.

In contrast, plants suffer from the fact that the exploitation of homologous recombi-nation remains a
challenging objective. Gen-erally, DNA delivery does not involve control over the genome site into
which a transgene is integrated and, in many cases, the delivered DNA becomes rearranged during
genetic modification. Transgene delivery in plants involves three technologies: protoplast trans-
formation; microparticle bombardment or biolistics; and Agrobacterium binary vectors. The latter two
processes are the most widely used. As its name suggests, microparticle bombardment involves the
penetration of plant tissue by tungsten particles coated with transgene DNA. Agrobacterium
technology exploits the disease features of the pathogen Agrobacterium tumefaciens, which has
evolved a natural process for the delivery of bacterial genes to the plant genome. Using a process
related to bacterial conjugation, a specific tract of Agrobacterium DNA is transferred to the plant
genome, where its expression leads to the development of a gall. Substrates that support the growth of
Agrobacterium are produced within the gall. By disarming this disease process, it is possible to
engineer Agrobacterium vectors that will carry trait genes into GM plants. In addition to the
integration of trait genes within the plant chromosomes, it is possible to integrate transgenes within
the genomes of plant chloroplasts. Because these organelles have genetic features that are related to
those of the prokaryotic bacteria, this involves the use of homologous recombination to place
transgenes in a predetermined location. A diagrammatic overview of the various techniques used in
the construction of GM plants is included in Fig. 15.1.

An important part of the safety evalua-tion process is the provision of molecular data that demonstrate
the nature of foreign DNA that has been inserted as a result of genetic modification. The method used
to transform a GM plant has a bearing on this. It is well established that biolistic delivery often leads
to extensive structural rearrangement of the integrated DNA. In some cases, this is so extensive that it
is extremely difficult to unravel its DNA sequence. This has led to a preference for the use of
Agrobacterium delivery, which can also cause the integration of multiple DNA copies but is less
prone to causing structural rearrangement of the transgeneic DNA.
One limitation of Agrobacterium had been its restriction to dicotyledonous plants, but recent
developments have extended its use to also include monocotyledonous plants such as rice and maize
(Ishida et al., 1996).

Another distinct issue that has been of concern to regulatory authorities is the realization that small
fragments of transgene DNA can sometimes be integrated at second-ary genome sites following
biolistic delivery. Recently, this was found to have taken place in the GM soybean line that is widely
used commercially. Careful analysis of this specific case eliminated any safety concerns, but this
observation does serve to emphasize the need to investigate unintended secondary integration of
potentially small fragments of transgenic DNA.

Further options that influence safety and containment are:

The use of a vector designed to deliver transgenic DNA into the plant chloroplast genome by
homologous
recombination provides controlled integration at a known site. In addition, the lack of chloroplasts
in pollen pro-vides environmental containment of transgenes.

The separation of trait and plant selec-tion genes on distinct DNA fragments facilitates their
unlinked integration in the plant genome. Alternatively, the plant selection gene can be flanked by
sequences that are recognized by site-specific recombinases (e.g. cre/lox). In both cases,
elimination of the selection gene is possible using conventional plant crosses.

The elimination of unnecessary DNA that was used during bacterial stages is very
straightforward, but this has not always been undertaken, leading to problems with antibiotic
resistance genes and more complex rearrangement of trait DNA.

The marker elimination process des-cribed in Fig. 15.1(b) is realized by conventional crosses that
facilitate the segregation of trait and selection genes. For site-specific recombination, the
recombinase gene is introduced from a separate GM plant to effect marker deletion, and this gene
can be removed by a subsequent conventional cross.

The Use of Selectable Marker Genes

In order to effect the introduction and expres-sion of transgenic DNA, there is a universal need to use
some form of selection to dif-ferentiate the transformed cells. This usually involves the use of a
selectable marker gene that may be physically linked to the chosen trait genes. For the primary
transformation of GM plants, antibiotic resistance genes have often been exploited as convenient
selection markers. The general importance of antibiot-ics for human and veterinary medicine and the
high profile of microbial drug resistance have made this practice controversial. The nature of the
selection system used in GMO construction is a particular focus of safety assessment.

The nptII gene, originally derived from E. coli transposon Tn5, is used frequently as a plant selection
marker, and to this end it has been equipped with a plant-specific promoter to facilitate its expression
in plant cells. A com-prehensive argument about the safety of nptII used as a plant-selectable marker
has been developed, and this was first formally pre-sented by Calgene (1990) and accepted by the US
FDA and other regulatory bodies. A sig-nificant factor is the limited importance of kanamycin and
neomycin in the treatment of bacterial infections in humans mainly as a consequence of their relative
toxicity. In addi-tion, it is recognized that antibiotic resistance is already widespread in bacteria, and
rare gene transfer from a GM food source is unlikely to be of practical consequence (Nap et al.,
1992).

Alternative selection markers that avoid the use of antibiotic resistance genes are becoming available,
and mechanisms, such as the cre/lox system (Dale and Ow, 1991), have been developed to facilitate
the removal of selection markers after GM plant construc-tion. Recently, Novartis Seeds announced
their intention to phase out the use of antibiotic resistance genes. Their Positech marker system uses
a selection for growth on mannose and relies on a gene for phosphomannose isomerase (Anon., 2000).
Other approaches include the use of co-transformation of trait and selection genes followed by
segregation of the latter. This has been effective in both Agrobacterium trans-formation (Komari et
al., 1996) and biolistic transformation (ACNFP, 1996). Figure 15.1 includes a schematic
representation of these alternative selection methods and strategies to eliminate selection markers
from the final GM plant constructs. Despite these developments, GM plant material carrying
antibiotic resis-tance genes continues to be put forward for consideration by regulatory authorities
who need to assess safety on the basis of objective scientific criteria.

In contrast to nptII, several GM plants have been developed in which other antibiotic resistance genes
have been introduced. In these cases, the antibiotic resistance genes have not been engineered as plant
selection markers and they retain their original bacte-rial promoter. Genes in this category include bla,
aad and nptIII, which confer resistance to ampicillin, streptomycin/spectinomycin and amikacin,
respectively. All of these antibiotics have greater use in clinical medicine than kanamycin and
neomycin. The most common reason for the presence of these genes in GM plants is that the trait gene
was first engi-neered using a bacterial vector and E. coli clon-ing techniques where these bacterial
selection markers would be of value. Subsequently, the complete bacterial vector has been delivered
to the GM plant without first removing these now redundant DNA sequences. The bacterial marker
genes are not directly selectable in plants, and there is no good reason for their presence in GM
material destined for use as food. Despite this, there are scientifically valid arguments that they do not
pose a safety hazard. Most persuasive is the fact that drug resistance is already widespread in bacteria
as a direct consequence of the very extensive use of antibiotics in human and veterinary medi-cine
and as animal growth promoters. Also, the process of gene transfer from a GM plant to a
microorganism is likely to be a very rare event, and this is discussed in more detail below. This issue
has been controversial for regulatory authorities, with somewhat differ-ent views emerging on the use
of antibiotic resistance genes in GM plants. The ACNFP has adopted a cautious position (ACNFP,
1995) and, in general, the inclusion of anti-biotic resistance genes in GM plants is widely
discouraged.

The Safety Assessment Process

Despite differences in safety administration, notably between the USA and Europe, the scientific
issues that are addressed during the safety assessment of GM food are very con-sistent. Details have
been published by both the FDA (1992) and the EC (1997b), and an overview of the main features is
presented in Box 15.2. A key principle is that an integrated, stepwise and case-by-case approach is
required. Safety assessment is aided by the use of decision trees that give guidance on the specific
points that need to be addressed for an individual case. Substantial equivalence plays a key role in
identifying differences between a GM food and its conventional counterpart, and these become a
focus for further consideration. The role of substantial equivalence in safety assessment is discussed
in detail below. The information that is required for the safety assessment includes: details of the
genetic modification; the stabil-ity of the modification and potential for its transfer; protein expression
and its effect on function, allergy and toxicity; potential secondary effects; composition; intended uses
and effects of cooking and processing; and potential intake and dietary impact.

The Role of Substantial Equivalence

The concept of substantial equivalence plays an important role in the safety evaluation of GM food.
First, it is important to stress that this concept is not in itself a safety assessment process and, in
particular, it is not intended to identify hazard. Rather, it acts to identify key issues that require
detailed safety evaluation. Substantial equivalence was developed under the guidance of the World
Health Organization (WHO), the Organization for Economic Cooperation and Development (OECD)
and the Food and Agriculture Orga-nization of the United Nations (FAO) follow-ing on from the
realization that conventional approaches to safety assessment, as used for pharmaceutical products,
have serious limitations (FAO/WHO, 1991, 1996; OECD, 1993). These limitations also explain why,
in contrast to pharmaceuticals, animal feeding experiments are less prominent in the safety evaluation
of novel foods. The general approach to safety testing using animal feed-ing involves the consumption
of increased amounts of the test substance until an adverse effect is detected. The sheer bulk of many
whole foods prevents this increased exposure, and data interpretation is com-promised by the fact that
foods are complex mixtures of many different chemicals. Another critical point is that food
contributes to nutrition. Toxicology testing using animals depends on the establishment of optimal
nutrition to provide a controlled background against which to evaluate any effect from the fed test
substance. In the case of whole GM food material, it is self-evident that any observed negative effect
is as likely to arise from disturbed nutrition as it is to be caused by novel technology. These
limitations of conventional animal testing approaches for the safety evaluation of biotechnology prod-
ucts were emphasized during the safety assessment of other novel technologies: food irradiation and
mycoprotein. Thus there is a well-established scientific basis for seeking new approaches to the safety
evaluation of novel foods.

The substantial equivalence concept recognizes that many people have eaten conventional foods over
a very long period of time and this establishes an accepted level of safety. Genetic modification
involves the introduction of only small changes, and thus a comparative approach can be used to
reveal any intended and unintended differences between GM material and its conventional
counterpart. The OECD (1993) described substantial equivalence as embodying the idea that existing
organisms used as food, or as a source of food, can be used as the basis for comparison when
assessing the safety of human consumption of a food or food component that has been modified or is
new. By concentrating on the safety assessment of the differences between a GM derivative and its
conventional counterpart, it can be concluded that the established and accepted safety of the
conventional food has not been compromised. The comparative approach involves the evaluation of a
large body of phenotypic data that include agronomic traits and details of chemical composition.
Typically, the latter includes fats, proteins, solvent-extracted hydrophilic matter, fatty acids, amino
acids, micro-nutrients, antinutrients, crude fibre, ash and moisture content. Particular attention is
given to any known toxins or allergens.

Substantial equivalence has been debated widely in recent years, and it has attracted some vigorous
criticism (Millstone et al., 1998; RSC, 2000) and equally robust defence (Burke, 1999; Kearns and
Mayers, 1999; Trewavas and Leaver, 1999). Recently, the FAO/WHO (2000) report, Safety Aspects of
Genetically Mod-ified Foods of Plant Origin, addressed criticism of the application of the concept of
substantial equivalence and reaffirmed its usefulness. In particular, it emphasized that the determina-
tion of substantial equivalence is not in itself an end point but rather the starting point for safety
evaluation. The substantial equiva-lence concept is likely to be challenged further as genetic
manipulation technology advances. The stacking of multiple traits and the engineering of some
second-generation traits that may deliberately alter metabolic flux are examples of some areas where
the application of substantial equivalence may prove complex.

One major concern is the capacity of substantial equivalence to reveal unintended consequences of
genetic modification. In this regard, it is especially pertinent that conven-tional breeding is as likely as
GM technology to generate unintended effects. Conventional breeding can exploit techniques, such as
mutagenesis and induced polyploidy through the use of colchicines, that might be expected to lead to
unpredictable genetic changes. It is curious that there is relatively little concern with respect to this
very similar risk issue. Overall, the controversy surrounding sub-stantial equivalence serves to
highlight the fact that its role in safety evaluation often is misunderstood. Also, it is clear that substan-
tial equivalence does indeed have limitations.

A current area of research activity concerns the possibility of exploiting new methodologies such as
molecular profiling techniques to provide a more detailed analytical comparison. Currently,
substantial equivalence involves an analysis of com-position and phenotypic parameters that is
undertaken with a targeted approach. In contrast, molecular profiling is non-targeted and more
holistic. Molecular profiling encom-passes three distinct technologies: metabolic profiling;
proteomics; and DNA microarrays. These technologies interrogate sequential steps in the expression
of genes through mRNA, proteins and metabolism, as is illustrated in Fig. 15.2.
Metabolite profiling exploits a variety of chemometric technologies to gather gross data on the
distribution of individual meta-bolites. Critically this does not involve the prior selection of individual
target molecules for quantification. In particular, gas chromatographymass spectroscopy (GCMS),
nuclear magnetic resonance (NMR) and high performance liquid chromato-graphy (HPLC)
techniques are capable of detecting, resolving and quantifying a wide range of compounds in a single
sample.

Proteomics involves the use of two-dimensional gel analysis to separate individ-ual proteins that are
present in a particular tissue. An example of this technique applied to four different conventional
varieties of tomato fruit is given in Fig. 15.3. While these profiling techniques are intended to reveal
differences using a holistic analysis, there is a need to follow up with an assessment of the safety
implications for any differences that are found. Proteomics has a valuable dimension in this regard in
that it is possible to identify individual protein spots by relating them to the genes by which they were
encoded. This relies on the availability of genome sequence data and thus the ease of application
varies for individual species. The identification process involves the use of MALDI-TOF mass spec-
trometry to determine the mass of protein fragments generated by specific proteolytic cleavage.
Because the cleavage pattern is predictable from the amino acid sequence of a given protein,
identification is possible. Alternatively Q-TOF mass spectrometry can be used to gain sequence data
directly from a protein spot, and this can be matched to information available in sequence databases.

In our own laboratory, this latter approach has been used successfully to iden-tify an unknown protein
that appeared in a GM tobacco plant. In this case, an experimen-tal transgenic line with a very severe
morpho-logical abnormality was being studied. The gross proteome was remarkably similar to that of
a normal non-GM syngenic plant but two new protein spots were detected in the GM plants. These
were both identified as the same plant lectin, suggesting that a plant defence mechanism had been
stimulated in response to the genetic modification. The fact that one protein generated two protein
spots is important as it emphasizes the complexity that can arise where post-translational modifi-
cation of a protein takes place. Proteomics is a powerful technique that couples gel electro-phoresis
with mass spectrometry and permits the visualization and possible identification of several thousand
proteins representing the detectable part of the cells total complement of proteins (the proteome). The
technical limi-tation is associated with the relatively slow and demanding gel electrophoresis process,
but alternatives to this are currently being developed.

DNA microarrays with many thousands of gene sequences arrayed on nylon or glass substrates permit
simultaneous examination of the steady-state levels of the cells mRNAs (the transcriptome).
Differential labelling of mRNA extracted from a GM plant and its conventional counterpart with
fluorescent dyes, and their hybridization to a DNA microarray generates data on the relative
abundance of individual mRNA molecules for each individual gene. This approach depends on the
availability of sequence data with which to design the probes used in microarrays and it gives instant
data on the identity of any differences that are revealed.

Molecular profiling techniques generate a very large volume of data, but the point of interest is
detecting biological components that are changed in the GM food. Determining the significance of the
changes and their rele-vance to safety assessment is difficult without a background context based on
the natural variation in the levels of the various biological components in conventional varieties.
While the analytical power of molecular profiling approaches is without question, it remains to be
seen how useful they will be in practice for safety evaluation. The new techniques will need to be
validated and, in addition, the significance of any observed differences will need to be determined.
Clearly, gene expres-sion under normal circumstances is dynamic, and any differences seen in a GM
derivative will need to be considered against this back-ground. The potential of molecular profiling
was highlighted by the Royal Society of Canada (RSC, 2000) and has been debated by FAO/WHO
(2000). In Europe, a significant programme of research supported by the EC and the UK Food
Standards Agency is under way with the intention of investigating the potential of molecular profiling
and the rela-tive merits of the different techniques in the context of safety evaluation
The Impact of the Introduced Trait

Clearly, in most cases, genetic modification involves the introduction of a new intended trait to a
conventional food material. In the context of substantial equivalence, it is thus likely that an intended
difference will be present at least at the level of the primary food material. This last point is
significant because the processing of a food material may, to a greater or lesser extent, eliminate the
difference. A good example is purified oils prepared from GM plants such as oilseed rape or maize
where it is difficult to differen-tiate these products from their conventional counterparts.

The introduced trait is often readily amenable to safety assessment using conven-tional toxicology
approaches. For example, the insecticidal protein produced by Bacillus thuringiensis has been
evaluated extensively in isolation from GM plants that have been transformed with its gene. One
frequently expressed concern is that expression of the same gene in bacteria and a GM plant may
generate different protein products and this needs to be addressed. An obvious possibility is that
glycosylation may occur in plants when it is not found in bacteria.

In addition to potential toxicity or other physiological effect, there is a need to assess the potential that
a newly expressed protein might be allergenic. This involves an assess-ment of the stability of the
protein and data-base searching to determine homology with known allergens. Known allergens share
cer-tain common properties that include resis-tance to heat, stomach acid and degradation by
gastrointestinal tract enzymes, all of which can be assessed.

FAO/WHO have undertaken several recent reviews of GM food safety assessment that have
addressed the question of allergenic potential, and this has culminated in the report of an expert
evaluation that took place very recently in Rome during 2001 (FAO/WHO, 2001b). The FAO/WHO
report from the Rome meeting updated a widely used decision tree that was designed to guide the
assessment of allergenic potential, and this new version is reproduced as Fig. 15.4. This most recent
report drew a distinction between expressed proteins that were derived from sources with known
problems of allergenicity and sources with no known allergenicity. In the former case, initial
investigation is recommended to be based on an analysis of sequence homology to known allergens
pres-ent in the source material. A negative result is followed up by immunoassays to investigate
possible immunoglobulin E (IgE) binding and possibly in vivo studies using patients allergic to the
source food. It was recognized that the use of human in vivo methods would raise ethical issues and
that their use would need to be considered on a case-by-case basis. Where the expressed protein is
derived from a source with no known allergenicity, the initial inves-tigation is also focused on
database searches. A negative result is followed by targeted serum screening using samples containing
high levels of IgE antibodies broadly related to the gene source. A positive result would suggest that
the protein was potentially aller-genic. Following a negative result, further studies of pepsin resistance
and the use of suitable animal models are recommended. The FAO/WHO consultation emphasized the
importance of maintaining and constantly updating an allergen database. Also it recog-nized that
animal models had not been evaluated for all food allergens but sufficient scientific evidence was
available to suggest that animal models will contribute valuable information regarding the
allergenicity of foods derived from biotechnology.

Fate of Consumed DNA

Following the consumption of GM food, the fate of any novel introduced genetic material is pertinent.
Concern extends to the possible uptake by host cells and by microorganisms that inhabit the
gastrointestinal tract. The general sensitivity of consumed DNA to inactivation and degradation
provides one of the best established barriers to the transfer of transgenes. The enzyme
deoxyribonuclease I produced by the salivary glands, pancreas and small intestine is a potent
degradative enzyme, and the low pH of the stomach removes adenine and guanine residues from
DNA, thereby eliminating its biological activity (Beever and Kemp, 2000).
Several recent studies add to our under-standing of DNA survival following its con-sumption. An in
vivo study by Mercer et al. (1999) investigated the effect of human saliva on DNA survival using
competitive polymer-ase chain reaction (PCR) and tested biological activity by measuring
transformation into the naturally competent oral bacterium Strepto-coccus gordonii. Despite evidence
of DNA degradation, sufficient biologically active DNA survived exposure to saliva to generate
transformants, albeit at a reduced frequency. Duggan et al. (2000) investigated DNA degra-dation by
ovine saliva and ovine rumen fluid, measuring biological activity using E. Coli transformation. PCR
amplification of DNA was possible for 30 min after exposure to rumen fluid, but transforming ability
was lost within 1 min. In contrast, the ability to transform E. coli was retained even after 24 h
exposure to saliva.

These studies suggest that, although degraded, DNA remains available for trans-formation in the oral
cavity. In contrast, it is inactivated rapidly and loses biological activity further down the
gastrointestinal tract. Chambers et al. (2000) investigated the fate of the pUC18 ampicillin resistance
gene in vivo using chicken feeding experiments, and their findings reinforce this view. Both bacteria
carrying pUC18 and transgenic maize carrying the bla gene were studied. PCR restriction fragment
length polymorphism (RFLP) was used to differentiate the test bla gene from naturally occuring bla
genes that may have been present in the gastrointestinal tract microflora. The pUC18 gene lacks a PstI
site that is present in the natural gene. While the maize-derived gene could be detected in the crops of
experimental chickens, it did not persist further down the intestinal tract. In contrast, bacteria carrying
pUC18 provided protection, allowing bla gene detection throughout the intestinal tract.

A somewhat different picture has emerged from the work of Schubbert et al. (1994, 1997). These
authors fed mice with bacteriophage M13mp18 DNA, and the fate of this foreign DNA in the animals
was followed by several methods. Fragments of M13mp18 DNA were detected in the contents of the
small intestine, the caecum, the large intestine, the faeces and in blood. It was calculated that 24% of
orally administered DNA was detected in the gastrointestinal tract and 0.10.01% was retrieved from
blood. M13mp18 DNA fragments were traced by PCR to peripheral leucocytes and located by
fluorescence in situ hybridization (FISH) in about 1 of 1000 white cells between 2 and 8 h after
feeding and in spleen or liver cells up to 24 h after feeding. M13mp18 DNA could be traced by FISH
in the columnar epithelial cells, in the leucocytes, in Peyers patches of the caecum wall, in liver cells,
and in B cells, T cells and macrophages from spleen. These findings suggest transport of foreign DNA
through the intestinal wall and Peyers patches to peripheral blood leucocytes and into several organs.
Upon extended feeding, M13mp18 DNA could be cloned from total spleen DNA into a vector.
Schubbert et al. (1998) extended this study and obtained simi-lar results using a plasmid expressing
the gene for green fluorescent protein. The broad con-clusion from this work is that consumed DNA
may survive, cross the gut epithelium, enter the bloodstream and interact with mamma-lian cells.
However, the significance of these results has been questioned. The experiments were artificial in the
sense that mice were exposed to large amounts of pure prokaryotic DNA. Beever and Kemp (2000)
questioned the fact that the DNA used in these experiments was unmethylated and contained
sequences likely to cause up-regulation of inflammatory cell activity and to stimulate a significant
immune response. It is possible that this induced response contributed to the detection of DNA in
white blood cells.

Overall, there is substantial evidence to suggest that DNA degradation is very effi-cient in the
gastrointestinal tract but some may remain biologically active in the mouth for sufficient time to be
taken up by naturally competent oral bacteria. The implications of the observations made by Doeffler
and colleagues are less clear, but suggest that, despite its hostile environment, some DNA may
survive in the gastrointestinal tract.

DNA Transfer from GM Plant Material to Bacteria

Concern about gene transfer from GM plant material into microorganisms needs to be related to the
nature of the genes that are involved. As already discussed, the use of antibiotic resistance genes has
been a par-ticular concern. The most likely mechanism by which microorganisms might acquire
transgene DNA is by its release from GM food and its subsequent uptake by naturally competent
bacteria. This is a real possibility in the oral cavity, as has been demonstrated by the work of Mercer
et al. (1999) that is described above. In general, the status of bac-terial gene transfer by natural
genetic trans-formation has been reviewed extensively by Lorenz and Wackernagel (1994). A few
studies have investigated DNA transfer from GM plant material to microorganisms, and these studies
suggest that such an event would be extremely rare.

Schluter et al. (1995) used a model system based on the plant pathogenic species Erwinia
chrysanthemi. A transgenic potato car-rying an integrated pBR322 plasmid was used, and the plant
pathogenic property of Erwinia provided an intimate association between the plant material and the
potential bacterial recipient. Erwinia causes soft rot by lysing plant tissues, and it is able to support
the replication of the pBR322 plasmid and the expression of its antibiotic resistance genes. Evidence
for plant to bacterium transfer was not detected. Related in vitro experiments were undertaken to
provide quantitative data on gene transfer. This was estimated to have a maximum probability of 5.8
1014 for an experiment using 0.9 g of potato tuber and 6.4 108 bacteria.

DeVries and Wackernagel (1998) used naturally competent Acinetobacter and focused on the plant
selection marker derived from the nptII kanamycin resistance gene. In these experiments, the recipient
Acinetobacter strain carried an inactive homologue of the nptII gene that was under the control of a
bac-terial promoter, and this strategy increased the likelihood of detecting a positive result. In this
case, transforming DNA did not need to be capable of autonomous replication or illegitimate
recombination to be detected. Homologous recombination between the plant-derived nptII gene and
the mutant resi-dent gene would repair the defect in the latter gene, leading to the recovery of
kanamycin-resistant transformants by marker rescue. This event was detected readily at a frequency
of 0.9 104 per nptII gene. However, when artificial homology between the transgene and the
recipient genome was absent, trans-formation frequency fell below the 1.3 1013 limit of detection.
The marker rescue data sug-gest an efficient mechanism for DNA transfer in which as few as 2.5
103 transgenic potato cells could generate a transformant, and res-cue of the kanamycin resistance
marker was effective even in the presence of a more than a 6 106-fold excess of plant DNA. It is
impor-tant to emphasize that this process depends on the provision of artificial homology and
represents marker rescue rather than the recovery of unique DNA from the transgenic plant. However,
it clearly demonstrates that naturally competent bacteria are very effec-tive at taking up transgene
DNA and, in the event of DNA homology being available, transformation is likely to take place.
Gebhard and Smalla (1998) obtained similar results using marker rescue by Acinetobacter and DNA
from GM sugarbeet.

As has been emphasized by Nielsen et al. (1998), selective pressure is of critical impor-tance in
assessing the consequences of gene transfer. The occurrence of a gene transfer event in itself is
unlikely to be of any great sig-nificance unless it led to selective advantage in the recipient.
Conversely, the existence of selective advantage could make even a very rare genetic event important.

Post-market Monitoring

Post-market monitoring of GM foods is considered by many to be an important component of safety

assurance. While this a desirable objective, it is far from easy to realize. It is difficult to gather data on
which members of the population were eating par-ticular GM foods. A proposal in the UK to use
supermarket loyalty cards was rejected on grounds of the invasion of privacy. Also, GM foods, if
they are present in the diet of any population, will be derived from such a range of crops and
processes and present in such a large range of foods that it will be extremely difficult, if not
impossible, to correlate con-sumption of any particular GM food to any recognizable syndrome.
Experience in the USA, where a large number of people have been eating products containing GM
soya and GM maize for a number of years without any ill effects, suggests that any effects will be
very small. Despite the difficulties, post-market monitoring continues to be evaluated as a potentially
valuable approach. Indeed the recent FDA/WHO report on allergenicity recommended:
Post-market surveillance is a valuable tool in the monitoring of adverse effects and long-term
sequelae of foods derived from biotechnology and the Consultation recognized that the feasibility of
certain aspects of its implementation would need further investigation. (FAO/WHO, 2001b)

Conclusions

GM food includes a range of distinct appli-cations of modern biotechnology, ranging from cell
factories for the production of nature-identical ingredients through to the provision of GM fruits to be
eaten fresh and unprocessed. The safety evaluation of GM food involves well-established procedures
that have been developed extensively over the past decade, and these procedures are subject to
ongoing review and updating. There is broad consensus on the relevant risk issues, and structured
approaches have been designed to focus on those issues that are relevant to individual cases. To date,
these processes have proved effective in dealing with GM food innovations but, as more complex trait
development is undertaken, it will be important that safety assessment continues to meet the
challenge.