Conventional Plant Breeding

Conventional plant production occasionally generates foods with undesirable traits, some of which are potentially hazardous to human health. Most crops naturally produce allergens, toxins, or other antinutritional substances (see Chapter 5). Standard practice among plant breeders and agronomists includes monitoring the levels of potentially hazardous antinutritional substances relevant to the crop. For example, canola breeders monitor levels of glucosinolates in breeding lines under consideration for prospective commercial release, while potato breeders monitor for glycoalkaloid content. If a particular breeding line generates too much of an undesirable substance, that line is eliminated from consideration for commercial release.

In the United States, the plant breeding community is largely self-monitored. Regulatory agencies do not evaluate conventional new crop varieties for health and environmental safety prior to commercial release. Some other countries require government agencies to conduct premarket evaluations for new crop varieties, both conventional and biotechnology-derived. Canada has a “merit system” for the commercial release of new varieties of major field crops, in which candidate varieties are grown in government-administered field trials. Performance data from these trials, related to agronomic factors, disease resistance, and food quality characteristics, are compiled for all candidate and standard commercial varieties. The data are collected from multiple locations over multiple years, as cereal chemists analyze the grain for chemical and nutritional composition and plant pathologists conduct tests to determine reactions to relevant diseases. These data are then evaluated by a team of experts from industry, government, and universities.

The breeder of each candidate variety must convince these experts that it is competitive and worthy relative to other commercially available varieties of that crop, based on the performance data from the trials. Only if the committee agrees is the candidate variety allowed to be registered as a new commercial variety. If the variety does not perform within prescribed parameters for all characteristics, it is not commercially released (CFIA, 2003).

Unintended Effects of Conventional Plant Breeding

Naturally Occurring Toxins

All foods, whether or not they are genetically engineered, carry potentially hazardous substances or pathogenic microbes and must be properly and prudently assessed to ensure a reasonable degree of safety. Furthermore, all crop strains, including organic strains, potentially express traits generated by various forms of induced mutagenesis. (Under organic regulations, radiation breeding and induced mutagenesis are acceptable, but irradiation of the final food itself is not. For more information on organic regulations, see USDA-AMS, 2001.)

History provides examples of traditional breeding that resulted in potentially hazardous foods (see Box 3-1). Solanaceous (tobacco family) crops, such as potato and tomato, naturally produce various steroidal glycoalkaloids. These substances are toxic not only to humans, but also to insects and pathogenic fungi. During the course of ordinary plant breeding assessments, breeding lines with increased levels of glycoalkaloids may be identified by the breeder as showing superior insect or disease resistance and retained for possible commercial release. The elevation of glycoalkaloid levels responsible for the pest tolerance may not be noted until people become ill from consuming the foods.

BOX 3-1

Kiwi: A New Food? Puffer fish, chile peppers, and mustard all are traditional foods that humans have learned to consume in moderation to minimize adverse health effects. But what about foods entirely new to the human digestive tract? Although some have (more...)

Tomatine, a glycoalkaloid naturally present in tomatoes, can be produced in hazardous quantities in certain conventionally bred varieties. Ordinarily, alpha tomatine is present in immature tomato fruit, but is degraded as the fruit matures, so that by the time the fruit ripens to the preferred stage for human consumption, tomatine content is reduced to safe levels. Nevertheless, levels of naturally occurring tomatine in ordinary tomatoes bred using conventional methods can vary considerably, primarily based on maturity, type, and environmental growing conditions (Gilbert and Mohankumaran, 1969; Leonardi et al., 2000). In this respect, environment is more responsible for food hazards than genetic makeup or breeding method.

Another example of a possible effect is the unintended elevation of glycoalkaloid content in potatoes. All potatoes produce the toxic glycoalkaloid solanine, but mature potatoes from most cultivars have amounts so small as to be nonhazardous. However, some varieties produce more than others, and certain environmental stimuli, such as growing or storage conditions, can cause potentially hazardous increases in solanine content, even within a usually safe cultivar (Concon, 1988). For example, potatoes exposed to sunlight turn green, making them particularly prone to high solanine content. Dark-skinned varieties are less likely to turn green than light-skinned varieties, but in either case environment seems to be more important to solanine production than genetics.

Certain potato lines have been found to express greater disease- or pest-resistance, and they have been selected as superior, not always with favorable or intended results. The most notorious such selection was the Lenape potato, which was developed using conventional breeding methods (Akeley et al., 1968). After a successful commercial launch, it was found to have dangerously elevated solanine content in the tubers and was removed from the market (Zitnak and Johnston, 1970). More recently, a similar high-solanine potato variety was detected and withdrawn from the market in Sweden (Hellanäs et al., 1995). In this case, the potato was a heritage variety, developed in the United Kingdom during the nineteenth century, but superseded by another variety due to its susceptibility to disease (Kuiper, 2003). Nevertheless, it became popular in Sweden under the name “Magnum Bonum,” until its predilection for overexpressing solanine resulted in its commercial demise (Hellanäs et al., 1995).

In spite of occasional problems with the consumption of potato glycoalkaloids, conventional breeders continue to increase the glycoalkaloid content in the leaves to take advantage of its pest- and pathogen-deterrent properties. Consequently, the U.S. Department of Agriculture (USDA) recommends, but does not require, a limit for glycoalkaloid content in new potato varieties (Sinden and Webb, 1972).

Interestingly, the Lenape (also known as breeding line no. B5141-6) potato has been found to express some useful attributes, such as high solids content. Thus Lenape continues to be used successfully as a parent in conventional breeding programs, providing its genes to new commercial potato varieties, such as Atlantic and Denali. Progeny with genes providing the high solids content are selected or maintained, and the genes responsible for high solanine content in the tubers are selected against or rejected. Additionally, Lenape has been transformed with a genetic construct containing the solanidine glucose-adenosine diphosphate glucosyltransferase (SGT) transgene in the antisense direction, which is designed to interfere with the solanine biosynthesis (Moehs et al., 1997). In field trials, several transgenic Lenape-derived lines expressed substantially less solanine than the parent Lenape, apparently due to antisense expression, an rDNA method for “turning off” an undesirable gene (McCue et al., 2003).

Several breeding programs are developing potatoes derived from conventional crosses between the ordinary potato Solanum tuberosum and relatives of other species, such as S. acaule (Kozukue et al., 1999) or S. chacoense (Sanford et al., 1998; Zimnoch-Guzowska et al., 2000). These are conventional breeding programs in which genes from two different species are exchanged. The intent of these conventional breeding programs is to generate potato varieties with new beneficial features, while minimizing deleterious traits from the foreign species.

Breeders typically monitor levels of toxins in plants that are known to naturally contain them, even though such monitoring is only voluntary in the United States. However, an unexpected and unintended problem may result when combining different species because thousands of genes would be interacting, not just one or two genetic elements of interest. For example, hybrids of S. tuberosum and S. brevidens produced not only the usual glycoalkaloids, but also the toxin demissidine, which is not produced in either parent (Laurila et al., 1996). This singular result shows that non-genetic engineering breeding methods can have unintended effects and generate potentially hazardous new products.

Any time genes are mutated or combined, as occurs in almost all breeding methods, the possibility of producing a new, potentially hazardous substance exists. Conceivably, similar outcomes could result from using rDNA to transfer specific genes from S. brevidans to S. tuberosum, giving rise to hybrids expressing the novel toxin demissidine. In either case, the hazard lies with the presence of the toxin, and not with the method of breeding. Genetic engineering could also be used to transfer only the beneficial genes from S. brevidans, leaving behind the genes responsible for the novel toxin.

Another example of a toxic compound from traditional crops is psoralens in celery (see Box 3-2). Celery naturally produces these irritant chemicals that deter insects from feeding on the plant and also confer protection from some diseases (Beier and Oertli, 1983). Celery plants with an elevated expression of psoralens will suffer less damage from disease and insect predation and have more aesthetic appeal to consumers, who tend to reject insect- or disease-damaged produce.

BOX 3-2

Unintended Health Effects: A Conventionally-Bred Celery Cultivar. Genetic engineering has raised the question of whether its products create unintended health effects for consumers, but how do conventionally bred products compare along the continuum of (more...)

As a result, breeders may observe healthy, undamaged celery lines and select them for commercial release. Unfortunately, workers who harvest high psoralen-producing celery or pack it in grocery stores have, on occasion, developed severe photodermatitis, especially when they are exposed to bright sunlight and the celery is infected with pathogens (Berkley et al., 1986; Birmingham et al., 1961; Finkelstein et al., 1994). There are differences in psoralen content from one variety to another (Beier, 1990; Diawara et al., 1993), but environment seems to have the greatest influence on psoralen production (e.g., Diawara et al., 1995).

Genetic Engineering in Plants

As for conventional breeding techniques, the genetic engineering (GE) breeding and selection process is largely self-monitored and only new varieties of rDNA crops are subjected to assessment prior to commercialization. Evaluation of GE food products for other countries is described above, in the section Conventional Plant Breeding.

Genetic engineering methods are considered by some to be more precise than conventional breeding methods because only known and precisely characterized genes are transferred. In contrast, conventional breeding involves transferring thousands of unknown genes with unknown function along with the desired genes. Similarly, in mutation breeding hundreds or thousands of random mutations are induced in each mutated line.

Plant breeders take a variety of methods used to introduce desired traits into consideration during selection. The location where DNA expressing the desired traits is inserted into the host genome in rDNA technology may be immaterial—for example, if an insertion is made in an inappropriate place, the transformed plant is eliminated or selected against by the breeder, who will then select another plant with a preferable locus of insertion for continued evaluation and development as a new variety.

The number of different lines developed by breeders varies according to the crop, the desired trait, and the choice of the breeder, but is it not unusual to start with evaluations of 2,000 “sister” lines from a cross of two parents to develop just one new variety. Thus 99 percent of sister lines are eliminated over the several years of evaluation, for various reasons. These include poor expression of the desired trait, poor yield performance, increased disease susceptibility, or even a lack of visual and tactile appeal, which is a largely subjective and arbitrary designation, but an important criterion nevertheless.

Genetic engineering techniques require fewer lines or transformation events because the desired trait is known and identified early. Consequently, the evaluations—which still take several years—focus on eliminating any unstable lines or those with deleterious characteristics.

In contrast to breeders using other techniques, genetic engineering breeders typically prefer to start with a small number of plants and then select only a few. For example, the two currently approved GE plant varieties from public institutions, papaya (Gonsalves, 1998; Swain and Powell, 2001) and flax (McHughen et al., 1997), started with only about 30 sister lines.

Unintended Effects from Genetic Engineering

Unexpected and unintended effects can be seen with all methods of breeding. Traditionally breeders observe such off-types regularly; they methodologically eliminate these individuals during the evaluation process, long before preparations are made for commercial release. An unexpected or unintended effect does not imply a health hazard, although clearly a plant expressing novel and unexpected characteristics warrants closer inspection prior to commercial release.

A large body of transgenic plant material is developed by university and government laboratories investigating aspects of rDNA that are not related to commercial interests. The Organization for Economic Cooperation and Development (OECD) lists more than 10,000 field trials with GE plants that were conducted between 1986 and 1999. Most of these authorized trials were conducted on plants genetically engineered by public universities and government research institutions, not for commercial release but to test the plants for unexpected or unintended results (see OECD, 2000). Consequently, most of the information on unexpected or unintended results comes from these sources.

Because GE crops are regulated to a greater degree than are conventionally bred, non-GE crops, it is more likely that traits with potentially hazardous characteristics will not pass early developmental phases. For the same reason, it is also more likely that unintentional, potentially hazardous changes will be noticed before commercialization either by the breeding institution or by governmental regulatory agencies.

The following are among the examples of unexpected or unintended characteristics that are often cited in the scientific literature.

Conventional Animal Breeding

Genetics is crucial for any livestock enterprise and requires breeders to evaluate production conditions and market objectives; the best breeds or lines of livestock for meeting market goals; selection of a breeding system and breed types; and selection of individual animals within the breed type. Generally, production conditions—so called gene-environment interactions—and market factors have the greatest influence on conventional animal breeding strategies.

Breeders develop a breeding plan that may either be continuous if replacement females are selected from within the herd or flock, or be terminal if all breeding females are selected and brought into the herd or flock. The breeding system may involve straight-breeding using only one breed, or cross-breeding using two or more breeds, with the latter expected to provide hybrid vigor, known as heterosis. In sound breeding programs, sires are chosen carefully as they have a great impact on production traits, and dams are chosen carefully based on expected high fertility and desired maternal traits (see Box 3-3). Selection of individual animals for breeding systems within herds or flocks is important, but most animal breeding plans depend on information called “expected progeny difference,” which is an estimate from across all animals within a breed of the genetic potential for desired traits an individual can transmit to progeny.

BOX 3-3

Desirable Traits in Beef Cattle. Certain variables are used in a selection index to make decisions on which sires and dams to use in a breeding plan. Desirable traits include:

Unintended Effects from Conventional Animal Breeding

Genetic Engineering

As with plant breeding, animal breeding programs and the production of transgenic farm animals can result in unintended effects. Most, but probably not all, gene-based modifications of animals for food production or therapeutic claims fall under the FDA Center for Veterinary Medicine regulations of new animal drugs. The animal drug provisions of the Federal Food, Drug, and Cosmetic Act provide the legal framework for developing science-based guidelines for assessing the commercial value of these products to society. Other products of transgenic methods will no doubt be developed that could be viewed as containing food or color additives and vaccines. Development of site-specific gene insertion techniques and animal genome projects could change the scope of potential genetic modifications to yield a wider variety of products than are currently being investigated. The following are some examples of unintended effects of recombinant technology administered to animals.

Unintended Effects from Genetic Engineering

The report Animal Biotechnology: Science-Based Concerns describes some examples of unintended effects found in transgenic pigs (NRC, 2002). The introduction of DNA into random sites in the genome is a mutagenic event that will affect any gene at or near the site of introduction, potentially resulting in unintended effects in the target animal. Given the remarkable variation in sites of gene insertion, the number of gene copies transferred, and the level of gene expression, every animal produced by microinjection is potentially unique in its phenotype (NRC, 2002). This variability contributes enormously to any effort to assess the unintended effects of genetic modification and to evaluate whether any unintended effect that occurs is possibly a cause for a health concern in humans.

An example of a detrimental effect resulting from overexpression of a transgene is illustrated by studies conducted by USDA with transgenic pigs. These studies were the logical extension of earlier work showing that daily injection of recombinant pST markedly increased growth rate, improved the ratio of muscle to fat, and improved feed efficiency (reviewed in Etherton, 2000; Etherton and Bauman, 1998). In these studies, transgenic pigs were created with a gene for human somatotropin or bST to test the hypothesis that overexpression of human somatotropin would result in effects comparable with those observed in studies in which pigs were injected with recombinant pST. Some of the transgenic pig lines did exhibit increased weight gain and were more efficient in converting feed to body weight gain; however, the pigs that showed high levels of either human somatotropin or bST also had a variety of physical problems, including lameness, lethargy, and gastric ulcers (Pursel et al., 1990).

Cloned Animals

A detailed discussion of animal cloning, including the committee's findings and recommendations related to this area, can be found in the subreport included with this volume. The following is a brief overview of unintended effects associated with cloning technology.

Cloning by nuclear transfer from embryonic blastomeres (Willadsen, 1989; Willadsen and Polge, 1981) or from a differentiated cell of an adult (Kuhholzer and Prather, 2000; Polejaeva et al., 2000; Wilmut et al., 1997) requires that the introduced nucleus be reprogrammed by the cytoplasm of the egg and direct development of a new embryo, which is then transferred to a recipient mother to develop to term. The resulting offspring will be identical to their siblings and to the original donor animal in terms of their nuclear DNA, but will differ in their mitochondrial genes and possibly the manner in which the nuclear genes are expressed. Cloning from blastomeres and somatic cells may result in large calves and lambs, the so-called “large offspring syndrome” (Sinclair et al., 2000; Young et al., 1998). More serious abnormalities may also be associated with cloning from somatic cells (Wilmut et al., 1997).

Marker-Assisted Selection

The use of marker-assisted selection in both plants and animals is expected to increase exponentially as data from genome sequencing projects and the density of useful segregating markers increase for economically important species. Marker-assisted selection is already helping plant breeders to identify desirable individuals from heterogeneous populations and to segregate for disease resistance and other features that would ordinarily require several years and substantial field plot experiments. Animals in breeding programs initially will be screened for genes that control simple traits, such as horns, which are undesirable in cattle, and halothane sensitivity, which segregates with metabolic stress syndrome in pigs. As this technique develops over time, easily identifiable markers will be chosen that accompany multiple genes that control more complex traits, such as meat tenderness and taste, growth, calf size, and disease resistance.

Although marker-assisted selection has enormous potential for improving animal health and production traits, it may decrease genetic diversity (Dekkers and Hospital, 2002). Thus short-term gains in productivity may occur at the expense of longer-term improvements due to the extended time before an unintended effect is identified (Dekkers and Hospital, 2002; Dekkers and Van Arendonk, 1998).

It is possible that marker-assisted selection in animal breeding is beneficial because there is a cumulative effect of expression of multiple genes that may be important, whereas effects of individual genes within the chromosomal region associated with the marker are too minor to be of economic importance. Thus the cumulative effects of multiple genes may be exploited to a greater extent by marker-assisted selection for a single gene with a major effect in an attempt to enhance a desired trait. Additionally, this type of marker-assisted selection may counter inbreeding by encouraging breeding strategies that maintain diversity among major genetic loci through exploitation of genes from rare breeds and wild ancestors to improve traits for disease and parasite resistance and adaptability.

Genetic Instability

A frequently stated concern about the use of rDNA technology is that transgenic organisms may be genetically unstable. That is, due to the transfer of genetic material using rDNA methods, the host organism may have some as yet unknown and unobserved mechanism to identify and eliminate the intruding DNA. However, transgenic plants, animals, and microbes have been studied for long enough and intensively enough to determined that genetic instability is not a routine phenomenon. While it is known that transgenic organisms can, on occasion, silence the expression of inserted genes through mechanisms that are now well studied, this phenomenon occurs on an event-by-event basis—not to all transgenic organisms resulting from a particular method or a particular construct.

Although genetically unstable plant varieties can be produced, their likelihood for commercialization is very low. First, if the new trait is unstable, then the organism loses the new trait and reverts back to the traits inherited from the parent plant. Second, domestic and international plant protection regulations exist to protect the intellectual property rights of plant breeders (for more information, see the U.S. Plant Variety Protection Act [P.L. 91-577, 1970]; and the International Convention for the Protection of New Varieties of Plants [UPOV, 1961]). These regulations provide patent-like protection for new varieties of plants under a certificate program. Since only stable varieties are eligible for certification, their production will be favored over unstable varieties.

Gene Transfer Using Foreign DNA

Genetic engineering can be used to transfer DNA within species or across species. In considering examples of cross-species transfers, plants, animals, and microbes all have been engineered with DNA from distant species to give rise to new GE products with realized or potential commercial application. Most examples of commercial GE crops carry DNA from unrelated species, commonly bacteria.

While the process of crossing plant and bacterial DNA is a challenge for conventional plant breeders, gene transfer from bacteria to plants does occur in nature. An example is Agrobacterium (see Chapter 2), which has the capability of transferring a portion of its own DNA into plants, where it integrates into the plant genome. In most cases bacteria do not exchange DNA with plants or animals. Genetic engineering transfer of foreign DNA into animals, such as inserting a growth hormone gene into salmon to generate a faster-growing fish, is currently under review by FDA.

Many microbes have been engineered with genes from plants or animals. GE microbes are used to produce a wide range of pharmaceuticals. For example, bacteria are engineered with the human gene for insulin to produce recombinant human insulin. GE microbes are also used to produce products used in food processing. For example, GE bacteria produce the enzyme chymosin used in cheese making. This product was one of the first GE foods approved for use in the United States and in the United Kingdom (FDA, 2001; University of Reading, 2004).

Gene Transfer Using Same-Species DNA

GE foods have been identified as potentially hazardous because they carry genes from foreign species. However, combining genes from different species may or may not be hazardous, as all organisms, including humans, carry genes inserted from different species. For example, all humans carry genes that have been incorporated from viral infections (Cohen and Larsson, 1988).

Genetic engineering is one method that can be used to transfer genes from either related or unrelated species into a host genome. For example, the Xa21 gene is derived from the rice relative Oryza longistaminata. This gene confers the valuable agronomic trait of disease resistance, and so various rice varieties have been developed using both conventional crossing and rDNA technology to add the Xa21 gene to commercial rice varieties. Regardless of the method of breeding, from the standpoint of composition, the two new rice varieties are virtually indistinguishable.

Another example includes high oleic acid soybeans, discussed earlier. Both conventional breeding via induced mutagenesis and rDNA transfer of genes from other soybean varieties have been used to generate new soybeans with a higher proportion of oleic acid. However, the conventionally bred high oleic soybeans can be unstable—depending on environmental conditions, their fatty acid content may vary—while the GE high oleic soybean remain more stable (DHHS, 1996).

Other rDNA examples exist where foreign DNA is not necessarily present. The first GE food product approved for human consumption in the United States, the Flavr Savr tomato, was engineered with an inverted tomato gene (CFSAN, 1994), along with bacterial genes transferred to facilitate selection of the modified plant. In the absence of foreign DNA or genes from another species, an argument may be made that a GE plant is not transgenic and not unnatural if only native DNA is transferred. One could readily transfer genetic material from one tomato variety to another, or rice to rice, or soybean to soybean without any other contributing source of DNA.

A recent report of a similar case involved a coffee variety being engineered to produce less caffeine as an alternative to current industrial methods of decaffeination (Ogita et al., 2003), which may use benzene or other organic solvents for extraction. Using rDNA methods, Ogita and colleagues (2003) were able to transform coffee and observe a 70 percent reduction in caffeine in leaves. Because their intent was to interfere with the natural caffeine biosynthetic pathway in the coffee plant, the objective also could have been pursued with an induced mutation breeding program. However, the rDNA method was preferred because it was more precise and predictable, as well as less likely to induce deleterious mutations unrelated to caffeine.

Naturally Occurring rDNA and Human-Mediated rDNA

Genetic recombination occurs in both nature and in human-mediated genetic engineering of plants, animals, and microorganisms (see Box 3-4). This section examines recombination as it occurs in nature and as it occurs with genetic engineering, with particular reference to differences between the two processes. Natural recombination occurs in several ways, typically divided between homologous and nonhomologous recombination events. Genetic recombination events that occur specifically in animals are summarized in Chapter 2 and are discussed in detail in the report Animal Biotechnology: Science-Based Concerns (NRC, 2002).

BOX 3-4

Evolutionary History of Naturally Occurring Hybridization in Plants and Animals. Prior to the use of recombinant deoxyribonucleic acid (rDNA) techniques to achieve recombination of genes from different species, crossing among similar species occurred (more...)

Recombination Hotspots and Genetic Instability

The range of loci in the Arabidopsis genome into which T-DNA has been inserted indicates that the insertion location is random or near random (Ecker, 2003). If insertional hotspots do exist, they clearly are not exclusive, but at most they are mildly preferred points of recombinational insertion. Genetic instability at singular loci in itself is not hazardous unless the unstable nature necessarily gives rise to hazardous outcomes. However, genetic instability, when it occurs, affects only one cell or plant at a time, and the result is typically a loss of activity of the relevant gene. An example is the case where a herbicide tolerance gene is inserted into an unstable locus and then grown in a field of the herbicide-tolerant plants. When an instability incident occurs in this crop, the herbicide tolerance is lost—the plants experiencing this instability event revert to being herbicide susceptible. When the field is then sprayed with the herbicide, only those few individual plants suffering the instability succumb to the spray.

Because the timing of such destabilizing events is rare and does not strike an entire population simultaneously, it is unlikely that the loss of a small number of crop plants would be noticed compared with the loss of weeds or with nonherbicide-tolerant volunteer crop plants. If genetic instability were more common or synchronized, such that it were noticeable, the breeder would not be able to obtain certification for the new variety due to failure to meet the genetic stability requirement.

Genetic instability is observed from time to time, both in conventional breeding and in genetic engineering. Robertson (1978) described a conventional corn line with a high degree of genetic instability due to an endogenous transposon. As discussed earlier, unstable pigmentation in GE petunias provides an example of a systemic instability in particular GE plants (Meyer et al., 1992).

Spontaneous mutations change DNA also, affecting everything from single base changes, known as “point mutations,” to entire genomes. Deletions of entire segments from a genome are predictably severe and deleterious, but point mutations or even more substantial mutations, such as those affecting large tracts of DNA in chromosomes, can be beneficial and adaptive.

Because of the ubiquitous nature of spontaneous mutations, all plant varieties—whether conventional or GE—exhibit a small degree of genetic instability and generate mutants with some degree of frequency. If the frequency of spontaneous genetic instability is low, the typical loss-of-function mutants will probably not be noticed; if it is high, the variety cannot be commercialized.

Natural types of recombination can also result in the same effects. Transposons, even Agrobacterium insertions, all interrupt any DNA sequence where they insert, regardless of whether they insert into homologous or nonhomologous DNA. Depending on the function of the interrupted DNA, there may or may not be phenotypic consequences from the insertion.

Agrobacterium naturally inserts DNA into the host plant-cell genome, typically in a genetically stable manner. The substitution of desirable DNA by genetic engineering for the phyto-oncogenic DNA of the wild strains does not affect the mechanics of transfer. The genes responsible for Agrobacterium genetic transformation and recombination in the plant are physically and functionally separate from the T-DNA actually transferred and integrated. The mechanics of Agrobacterium appear the same whether the event is staged through a natural infection process by a crown gall-producing wild-type strain or by the same strain that has been disarmed by the removal of the gall-producing genes and subsequently replaced with known, desirable DNA.

Overview

Predicting the likelihood of unintended hazards from compositional changes associated with genetic modifications does not fit a simple dichotomy comparing genetic engineering with non-genetic engineering breeding. This is because there are many mechanisms shared in common by both GE and non-GE methods, and also because there are techniques that slightly overlap each other. Furthermore, within the scope of genetic engineering (rDNA) technology, several mechanisms for genetically transforming plants are available as options to scientists, such as Agrobacterium-mediated gene transfer and the use of the particle gun (McHughen, 2000). These two examples of genetic engineering are as different mechanistically as the conventional methods of narrow crossing and wide crossing (discussed in Chapter 2). Consequently, it is unlikely that all methods of either genetic engineering or conventional breeding will have equal probability of resulting in unintended changes. It is the final product of a given modification, rather than the modification method or process itself, that is more likely to result in an unintended adverse health effect. In Figure 3-1 some of the methods used to generate GM plants are shown to illustrate the full range of possibilities that might lead to unintended changes (these methods are described in Chapter 2).

Analysis of the Continuum

As noted earlier in this chapter, unintentional changes are possible with all conventional and biotechnological breeding methods for genetic modification. It is possible to represent the likelihood of such changes as a continuum, albeit only a partially understood one. Placement along this continuum has no bearing on risk of adverse outcomes, but only on the probability of unintended changes, which need not be hazardous. The potential for hazard resides in specific products of the modification regardless of whether the modification was intentional or unintentional.

Again, unintended effects do not necessarily imply hazard. If a particular method were inherently hazardous, all products resulting from its use would be potentially harmful. However, it is known that each method can provide safe products, so the key for breeders and regulatory agencies, in their reviews of specific products, is to identify the relatively rare, potentially hazardous products resulting from any method.

In breeding, the developer can generate literally thousands of breeding lines; each might be progeny of the same parents, thus bringing together the same set of genetic information, and yet possess subtle differences. As noted previously in this chapter, conventional breeders may scrutinize 2,000 of these “sister lines” of a single cross of two parent plants. Meiotic recombination allows sister lines to show a continuum of phenotypic expression. Over several years, the breeder eliminates most, or sometimes all, of those breeding lines because they are unsuitable.

Plant breeding is often said to be a process not of selection, but of elimination. Any off-types, unstable lines, or lines showing characteristics such as significant differences in nutrient content, responses to environmental stresses, diseases, or the presence of other undesirable traits are discarded as soon as they are noticed. This winnowing takes place over several years, so the remaining lines identified for prospective commercial release are unlikely, but not guaranteed, to have any significant compositional changes other than those related to the desired trait. For this reason, regulatory scrutiny focuses most often on the new trait and its metabolic perturbations. Nevertheless, the appearance of subtle or obscure phenotypic changes can go unnoticed by breeders or regulators and may subsequently have to be recalled from the market, as in the case of the Lenape potato— developed using non-genetic engineering methods—discussed previously in this chapter (Laurila et al., 1996).

The products of a narrow cross-pollination, from a plant of a given species to another plant of the same species, or even the same subspecies, are less likely, but not guaranteed, to pose unintended risks. This is due to the expectation that pollination is inherently safer than other methods of gene transfer because the range of resulting products is typically limited to those products already present in the species or subspecies. However, even this likely “safe” method is subject to occasional spontaneous mutations or infelicitous genetic recombination events leading to novel and unexpected products, and these products may carry some degree of risk, such as disease susceptibility, leading to, for example, hazardous levels of fungal mycotoxins. Expanding the range of conventional breeding beyond the same species provides a greater opportunity to introduce novel genetic information and thereby novel sources for unexpected effects.

Triticale is a man-made crop developed by breeding wheat with rye; hordecale is a similar product of barley and rye. Both examples bring together thousands of genes in a way that does not occur in nature and which conceivably could result in the production of undesirable toxicants. As discussed previously in this chapter, the conventional potato breeding program combining Solanum tuberosum and S. brevidens produced not only the expected toxicants, but also a new one, demissidine, which is not produced in either parent (Laurila et al., 1996).

Although some products can be reasonably predicted to be hazardous, this does not imply that all products of rDNA—or other methods of bringing together genes—are necessarily hazardous. Agrobacterium and the particle gun are the only two rDNA methods represented on the continuum in Figure 3-1 because they provide almost all of the GE crops approved for commercial release. They are depicted separately because of the documented predilection of the particle gun to introduce multiple, broken, or rearranged DNA segments, in contrast with the usually singular and high-fidelity transfer typical of Agrobacterium.

The more disorganized particle gun mechanism increases the risk of insertion into crucial endogenous sequences. The degree of increased risk is as yet undetermined and may be minimal. However, in both cases sufficient numbers of events are generated and screened to make it possible to identify those with the most desirable insertion features, the best expression of desired genes, and the least likelihood of deleterious features, in much the same way that conventional breeders evaluate their breeding lines.

Continuum Conclusions

Unintended adverse health effects can be either predictable or unpredictable. All organisms undergo spontaneous mutations, giving rise to novel traits, which may carry some hazard. The incidence of such mutations is relatively rare and the type of hazard associated with them is generally not predictable. On the other hand, introduced mutations, such as by rDNA, theoretically allow a gene from any species to be inserted into and expressed by a food crop. Clearly, such a product has the potential to be hazardous if the inserted gene results in the production of a hazardous substance. Although the process of rDNA is itself not inherently hazardous, the resulting product of the process may be.