ALIMENTOS MODIFICADOS GENETICAMENTE: SÃO SEGUROS?

Judging the “safety” of food is hardly an exact science.
Through experimental testing it is possible to
certify that some foods will be dangerous for human
consumption, but certifying a complete absence of
danger is (like any effort to prove a negative) beyond
the capability of experimental science.


Sixty-five percent of the Americans queried for an international sur-
vey on genetically modified foods got the answer to the following ques-
tion wrong: “Do ordinary tomatoes contain genes, or is it only
genetically modified tomatoes that do so?” All of our food contains
genes—all our plant food and all our animal food. As one biologist
explained in a teaching module for middle schools, “A pound of broc-
coli, for example, contains about a tenth of an ounce of DNA.” Accord-
ing to a 1997 study, people eat up to a gram of DNA a day. (One ounce
is equivalent to about 28 grams.) Of that, less than 1/250,000 was in-
troduced by genetic engineering. The rest was always in our diet.
When it is eaten, a gene from any source, whether from an animal,
an insect, a plant, a virus, or a bacterium, is broken down during diges-
tion into the building blocks of DNA, the nucleotides adenine (A),
thymine (T), guanine (G), and cytosine (C). The process begins in the
mouth. Saliva contains an enzyme, called deoxyribonuclease, specifi-
cally designed to dismember deoxyribonucleic acid, DNA. More of this
enzyme is produced in the pancreas and in the small intestine. The
stomach also contributes to the process: Stomach acid attacks DNA at
two of the nucleotides, A and G, inactivating the whole molecule.
The breakdown is not 100 percent efficient. Bits and pieces—short
stretches of DNA—escape, a fact that has been trumpeted by certain
activists. It might be the source of the Europeans’ fears, as documented
by another survey, that eating genes will alter their own genes, a con-
cept scientists call “horizontal” gene transfer (“vertical” gene transfer
is what happens when you pass your genes on to your children).
Studies showing that bits of genes persist for a while during the
digestive process have been done in chickens, sheep, and people, as
well as in laboratory mice. The first two digest their food differently
from the last two. In chickens, digestion starts in the crop, the stone-
filled bag in the bird’s throat. Sheep, like people, use saliva at first, but
unlike us they have two stomachs.
In chickens, researchers at the University of Leeds in England
found an antibiotic-resistance marker gene from genetically modified
corn in the crops of all five birds studied. The gene could be found in
the stomachs of two of the chickens, but not in the lower intestines of
any of them. When the scientists looked for a natural plant gene, they
found it in the same places.
The same researchers fed sheep kernels of corn that had been ge-
netically modified to produce the insecticide Bt, a toxin from the bac-
terium Bacillus thuringiensis. Earlier they had sampled fluid from a
sheep’s first stomach, the rumen, and found it inactivated naked DNA
in a test tube within a minute. Five hours after the sheep ate the corn,
the scientists could detect in the rumen the gene encoding the Bt toxin.
But when the experiment was done using silage, only a small fragment
of the gene could be found in the rumen, and that no longer than three
hours later. Silage is made by letting the corn ferment: by the time the
sheep ate it, the plant cells had begun to decay. The DNA was easier to
digest because the enzymes could reach it more easily.
A group at the University of Newcastle upon Tyne, led by Harry
Gilbert, used human volunteers—though they were not normal,
healthy subjects. Each had undergone an ileostomy, in which the il-
eum, part of the digestive tract, is removed. Each wore a bypass bag,
which collects the waste produced by that drastically shortened gut.
The scientists fed these seven ileostomists a burger and a shake con-
taining genetically modified soy. After waiting a suitable time, they
sampled the contents of the bypass bags, looking for the transgene,
which was a gene that made the soy plant resistant to an herbicide.
“Whilst the amount of transgene that survived passage from the small
bowel was highly variable between subjects, the nucleic acid was de-
tected in all seven subjects,” the researchers write. In one person 3.7
percent of the total amount of transgene he or she had eaten made it to
the bag. In the others the amount was less. Gilbert’s team then repeated
their study with healthy volunteers, sampling their feces. “No transgene
DNA was detected in the faeces,” they write, “indicating that the nucleic
acid did not survive passage through the complete intestine.” The ile-
ostomy bags had given the researchers a peek into the digestive pro-
cess. At midpoint in digestion a small percentage of the gene remained
whole. By the time digestion was complete the DNA had been broken
down.
Walter Doerfler, in the Institute of Genetics at the University of
Cologne in Germany, has been studying the fate of DNA eaten by labo-
ratory mice for 10 years. He and his colleagues have reported that some
DNA can escape being digested and pass through the intestinal wall. It
can enter the bloodstream and interact with the mouse’s own cells.
The DNA they used at first came from a bacteriophage called M13. It
was natural DNA, not a transgene. They detected the M13 DNA in the
small intestine, the large intestine, and the blood, as well as in the feces.
They found it in spleen and liver cells, and in one out of a thousand
white blood cells. They even found it in the cells of a mouse fetus,
meaning that it had traveled through the placenta from mother to un-
born child.
Doerfler’s work has been used as proof that eating genetically
modified food is dangerous. Yet, as a report prepared for the U.K.’s
Secretary of State for the Environment, Food, and Rural Affairs noted
in 2003, “It is important to emphasise that these studies are not fo-
cused on transgenes and they are relevant to the fate of all consumed
DNA.” They apply equally to the tenth of an ounce of DNA in an ordi-
nary pound of broccoli as to the pesticide-coding gene engineered into
transgenic Bt corn.
Also important to keep in mind is how Doerfler’s methods made it
possible to detect even the tiniest bit of foreign DNA. In the M13 ex-
periments the mice were fed 50 micrograms (50 millionths of a gram)
of a single, small DNA molecule, one only about 6,400 base pairs
long—a very big dose of a very small DNA sequence. In 50 micro-
grams of M13 DNA the concentration of any one particular segment is
much, much higher than in 50 micrograms of, say, the soybean DNA
used by Gilbert and his colleagues. The soybean DNA is a billion base
pairs long. To sample mouse blood or spleen cells for foreign DNA,
Doerfler and his colleagues used the PCR technique, which can am-
plify the smallest trace of any DNA. PCR doesn’t understand the word
foreign. You must give it a specific DNA sequence to look for, a list of
A’s, T’s, G’s, and C’s in exactly the right order. Perhaps that sequence is
1,000 base pairs long. If so, it will represent more than 10 percent of
the M13 DNA, but only a millionth of the soybean DNA.
To make it a fair search you’d need to feed a mouse a hundred
thousand times more soybean DNA than you fed it M13 DNA. To
match 50 micrograms of M13 DNA—an amount that would disap-
pear in the bottom of a tiny salt spoon—you’d have to feed the mouse
a couple of teaspoonfuls of pure soybean DNA. Yet people—and
mice—do not eat pure soybean DNA. A soybean is mostly starch and
protein. To eat a teaspoonful of soybean DNA in the usual way would
mean eating pounds and pounds of soybeans—a distinctly unnatural
scenario. So feeding a mouse the pure DNA of a simple bacteriophage
is not at all like feeding it food from a genetically modified plant.
What these experiments show, however, is that some DNA does
escape the process of digestion. A little will incorporate itself into our
blood and liver and perhaps even pass through into our unborn chil-
dren. Does that mean it could alter our genes? Yes, in principle, it could.
But has it? People have been eating DNA-rich foods for all of their
evolutionary history. Humans have a particularly eclectic diet, con-
suming foods derived from or containing bacteria and fungi, as well as
a great variety of different plants and animals. That is, we’ve been ex-
posing our digestive tracts to foreign DNA for untold eons before re-
combinant DNA techniques were invented. Because plant genes,
bacterial genes, and animal genes are different from human genes, we
should be able to identify the foreign genes in the human genome—if
they’re there—using computers to compare the vast DNA sequence
databases available today.
Animal genes, plant genes, human genes, and even bacterial genes
all work by the same rules. How genes are transcribed and translated
into proteins is common to all of life. But that’s like saying all books
contain words. Words belong to different languages. A Japanese word
is easy to tell apart from an English one. Even if the two words mean
the same thing in both languages, they can be distinguished. Bacterial
genes have a different structure from plant genes, and plant genes have
a different structure from animal genes. The genes of bacteria are gen-
erally not interrupted by introns, for instance, while both plant and
animal genes have them. The protein-coding sequences are different as
well. A human hemoglobin gene looks very much like a mouse hemo-
globin gene, but very different from a plant hemoglobin gene.
Proteins change more slowly than the DNA sequences that encode
them, so the correspondence between two genes can often be identi-
fied when the protein sequences are compared. So, for instance, when
scientists analyzed the DNA in the mitochondria of plant and animal
cells and the DNA in the chloroplasts of plant cells, they found that in
both structure and sequence these genes were more similar to those of
certain bacteria than they were to the genes in the nuclei of animals
and plants. This discovery confirmed biologist Lynn Margulis’s theory
that both mitochondria and chloroplasts were descended from bacte-
ria that had invaded and taken up residence in cells, creating a symbi-
otic partnership. Other studies found that in plants, at least, the
symbiosis is not at all static: genes from the chloroplast, for instance,
migrate into the plant’s nucleus at a surprisingly high rate.
That we can read in the genes the history of the partnership be-
tween chloroplasts and plants means that we should also be able to
detect plant, viral, and bacterial genes in the human genome—if they
were there. Viral genes are certainly there, but they’re not from plant
viruses or bacteriophage. They’re from viruses that infect animals and
humans. It is true that a handful of human genes do look more like
bacterial genes than like any other kinds of genes. When the human
genome was first sequenced, it was reported that there were about 200
human genes that might have come directly from bacteria—out of the
more than 30,000 genes in the human genome. After more investiga-
tion that number was reduced by half, then three-quarters. Some of
the candidate bacterial genes turned out to be what they appeared to
be: genes from bacteria that had contaminated the sequence analysis
and were not part of the human genome at all. The experts are still
arguing—and probably will for a while—about how to interpret the
rest. Are they due to horizontal gene transfer from bacteria to humans?
Or do they look out of place because all of their other relatives were
lost over the course of evolution? It seems more and more likely that
the relatives do exist, but just haven’t yet been detected. As more and
more animal genomes are sequenced and searched the number of such
genes becomes smaller and smaller. Relatives of the candidate genes
keep turning up.
So despite all of the opportunities cells have had to exchange genes,
genomes maintain their identities. Rather than being surprised that
there are bacterial genes in our genomes, we should perhaps be sur-
prised that there are so few of them.

One set of bacterial genes is often singled out in reports on the
safety of genetically modified foods—such as the reports by the Soci-
ety of Toxicologists in 2003, the American Medical Association in 2000,
and the World Health Organization and U.N. Food and Agriculture
Organization in both 2000 and 2001. These genes are treated as being
somehow different, somehow more suspicious. These are genes that
make the bacteria resistant to antibiotic drugs.
Antibiotic-resistance genes were, from the beginnings of genetic
engineering, commonly used as markers to identify the plant cells that
had picked up the new DNA. When Stanley Cohen and Herbert Boyer
patented their gene-splicing technique in 1980, they pointed out that a
crucial step was being able to separate the transformed cells carrying
the recombinant DNA from the parent cells. By linking an antibiotic-
resistance gene to the gene to be introduced, and then growing the cells
in petri dishes containing the antibiotic, it was easy to pick out the cells
that got the new genes: they were the only ones that survived on the
antibiotic.
Antibiotic resistance wasn’t the only marker Cohen and Boyer sug-
gested. Resistance to heavy metals, or an ability to manufacture a cer-
tain amino acid or another growth factor, would also mark the
transformed cells as different. The antibiotic-resistance markers, how-
ever, were easy to work with and they became a standard tool. When
introducing new genes into plants, the marker gene used most often
was one that conferred resistance to kanamycin, an antibiotic rarely
used in human medicine.
As crops developed using these new techniques got closer to the
marketplace, alarms were raised, both inside the scientific community
and by such groups as Consumers Union (publisher of Consumer Re-
ports magazine). Could eating antibiotic-resistance genes in food in-
crease the rate at which harmful bacteria became resistant to the
antibiotics that doctors depend on to cure diseases? The issue wasn’t
that eating an antibiotic-resistance marker gene could make a person
resistant to—and therefore unable to be cured by—antibiotics. No one
suggested that. The fear was that the ability to survive a dose of antibi-
otics would be transferred from the person’s food to the bacteria that
normally live inside a person’s gut—again, that hypothetical horizon-
tal gene transfer—and that resistance would then spread from an in-
testinal bacterium to disease-causing bacteria.
Theoretically, such a possibility exists, conceded a 2000 report on
agricultural biotechnology by the Subcommittee on Basic Research of
the U.S. House of Representative’s Committee on Science. “But it is
exceedingly unlikely,” the report continued, “because it demands nu-
merous steps, each of which also is highly unlikely.” The resistance gene
would have to escape the plant cell, yet not be degraded by saliva and
stomach acids. It would have to make contact with a bacterium, avoid
being cut to pieces by the bacterium’s restriction enzymes, and recom-
bine with the bacterial chromosome in just the right place and in just
the right way to be inserted.
But that’s only the first step, because the intestinal bacterium
would then, in a separate horizontal transfer, have to pass that very
gene on to a bacterium with a different lifestyle, one that caused a dis-
ease that a doctor would need to treat with the drug in question. For
the kanamycin-resistance gene, that means tuberculosis, one of the few
diseases occasionally treated with kanamycin (which has substantial
enough side effects in people to be used only as a drug of last resort).
This second horizontal gene transfer, too, is rather unlikely, particu-
larly in countries with modern sanitation systems, where surviving gut
bacteria are eliminated in sewage treatment plants.
Considerable effort has been invested in trying to detect such a
transfer—so far without success. The U.K.’s GM Science Review Panel,
formed by Margaret Beckett, the Secretary of State for the Environ-
ment, Food, and Rural Affairs, published a report in 2003 summariz-
ing nearly 20 scientific studies on horizontal gene transfer. In none of
them was such a heritable horizontal transfer detected. Yet the percep-
tion remains that antibiotic-resistance marker genes can move out of
food and into a gut bacterium and cause a health problem.
The researchers’ response has been quite practical. They have de-
vised different approaches, and identified different markers. For ex-
ample, the gene from jellyfish that codes for a green fluorescent protein
has been used as a marker. Cells that contain the gene coding for the
fluorescent protein (together with the gene being introduced) look
green under ultraviolet light. Cells that failed to take up and insert the
new gene remain dark. Another approach uses the sugar mannose. Or-
dinary plant cells growing in petri dishes are fed the sugars sucrose or
glucose for energy. Transformed cells, by contrast, survive when fed
another sugar called mannose, because they have received a gene that
can convert mannose into glucose. Without that gene, the cells starve.
A third approach is to remove the antibiotic marker after it has served
its purpose and before the plant is propagated.
It is worth noting that some of the most common genetically
modified food plants never did contain an antibiotic marker. In
Roundup Ready and other herbicide-tolerant crops, the herbicide it-
self was used to select the plant cells that had picked up the new genes
and to eliminate the cells that didn’t. The herbicide was simply added
to the petri dish together with the transformed plant cells. The cells
that survived were the ones that had incorporated the herbicide-resis-
tance gene.
-----------
Another DNA sequence that has been singled out as being some-
how different is the CaMV 35S promoter. The idea that crops pro-
duced using the CaMV 35S promoter could cause problems for plants
and people seems to have come from an article published by the jour-
nal Microbial Ecology in Health and Disease in 1999. “Cauliflower Mo-
saic Viral Promoter—A Recipe for Disaster” was written by Mae-Wan
Ho and Angela Ryan of the Department of Biological Sciences in the
U.K.’s Open University, together with Joe Cummins, a retired professor
of genetics at the University of Western Ontario. It is an opinion piece,
which means that it wasn’t subjected to peer review, the process by
which journals assess the logic and scientific validity of what they pub-
lish. Mae-Wan Ho is a founder of the Institute of Science in Society,
whose mission, according to its website, is to “work for social responsi-
bility and sustainable approaches in science.” Ho’s biography on the
website says that she has a Ph.D. in biochemistry from Hong Kong
University and is “a leading exponent of a new science of the organism
which has implications for holistic health and sustainable systems.”
Among her writings is “The Golden Rice—An Exercise in How Not to
Do Science,” in which she called Ingo Potrykus’s Golden Rice “a useless
application.” It would not provide better nutrition for the poor, she
argued; it was “worse than telling them to eat cake.”
The CaMV 35S promoter is widely used. According to Michael
Hansen of Consumers Union, “all GE [genetically engineered] crops
on the market contain it.” The reason is simple: all genes need pro-
moters. Without a promoter, a gene can’t be transcribed into messen-
ger RNA. If it is not transcribed, it can’t be translated into a protein.
That is, a gene simply doesn’t work without a promoter. But the tran-
scription machinery in a plant cell is different from that in a bacterial
cell; the promoters that work in plant cells are different from those
found in bacteria. To add a bacterial gene to a plant requires adding a
promoter that will work in a plant cell. The CaMV 35S promoter does.
To work well, as Beachy found out in his first experiments with viral
coat protein genes, a promoter also has to direct a plant to produce an
effective amount of the transcript. The CaMV 35S promoter works
well in plants, ensuring that the gene is transcribed often. It is also
expressed in many different plant tissues throughout the plant’s life.
For these reasons it quickly became popular among molecular biolo-
gists. It’s reliable, so it has been used extensively.
Like Ho, Hansen believes the CaMV 35S promoter is “unstable”
and will, in addition to the job it was meant to do, activate some “nasty
dormant genetic material.” It can cause “large-scale genomic rearrange-
ments,” say Ho and her colleagues. It could activate dormant viruses,
and perhaps even generate new ones. The “promiscuity” of the CaMV
promoter could cause it to trigger cancer. Ho and her colleagues are
particularly alarmed by the fact that plants have lots of transposons—
jumping genes—in their DNA. They say: “The fact that plants are
‘loaded’ with potentially mobile elements can only make things worse.”
Most transposons in plants, as Barbara McClintock found in corn, are
silent or asleep—Ho calls them “tamed.” She and her colleagues worry
that the 35S promoter might somehow awaken them and therefore
“destabilize the transgenic DNA,” and allow it to “generate more exotic
invasive elements.” They conclude that all transgenic plants—those in
which a gene has been transferred from another organism—should be
immediately withdrawn from commercial use. “The available evi-
dence,” they say, “clearly indicates that there are serious potential haz-
ards associated with the use of the CaMV promoter.” But there is no
such evidence.
The finding that frightened Ho and Hansen came from a study
published in Plant Journal in 1999. The study reported that the CaMV
promoter contained a site that is cut-and-pasted with other DNA more
often than usual. The researchers called that site a recombination
hotspot. The hotspot was very close to the end of the promoter. Break-
ing the DNA at such a hotspot would most likely disrupt the promoter’s
activity. But being easily disrupted does not mean that the CaMV pro-
moter is unstable or that it poses a hazard. If the promoter is dis-
rupted—separated from the gene—the gene does not work. The plant
does not express the trait, such as virus resistance, sought by the plant
breeder. Such a plant, in a breeder’s eye, is a loser, a cull. It is discarded.
No matter which promoter is used, precisely where the DNA is
inserted into a plant’s genome is more or less random. A few insertions
affect the plant’s ability to grow and be productive. These plants, or
their progeny, are discarded as well. For this reason, breeders create
many transformed lines. They cross the transformed plants to varieties
that perform well, then evaluate the progeny in the greenhouse and the
field. They check each generation to make sure that it expresses the
new trait. The best plants in the field trials are saved; the rest are dis-
carded. Just as Luther Burbank burned tens of thousands of rejected
blackberry bushes, the developer of a virus-resistant squash will grow
and select plants for 5 to 10 generations or more, weeding out the
underperformers. Among those rejected, obviously, will be any plant
in which the inserted DNA causes a lethal mutation. Even if a het-
erozygote that contains a good copy of the mutated gene grows well,
those that have two copies of such a gene will not grow at all or will die
before setting seed. Such plant families are discarded. Even small ef-
fects on the plant’s growth rate or its yield are noticed and, if deleteri-
ous, cause the breeder to cull. The new trait must be expressed and it
must continue to be expressed in subsequent generations. In fact, one
of the biggest problems for breeders is not that crops created through
molecular techniques acquire extraordinary new traits—that “nasty
dormant genetic material” is turned on—but just the opposite. The
newly introduced genes tend to turn off after a few generations.
Could the CaMV 35S promoter activate something nasty? If it is
nasty enough to affect the plant’s ability to grow and produce a good
crop, the plant and its siblings and progeny will be discarded. Could it
awaken dormant viruses, as Ho and her colleagues assert? Thousands
of copies of dormant viruses have been identified in plant genomes,
and their gene sequences have been analyzed. The viruses are defective
in ways that make it impossible for a promoter to turn them on. They
contain single base changes, and usually short insertions and deletions,
that make them gibberish to the protein-making machinery—with or
without a strong promoter. Moreover, even if such an “activation”—
however unlikely—were to happen in a plant, the result would be that
this one plant would suffer from a viral infection. If the virus stunted
the plant’s growth or reduced its yield, this plant too—or its prog-
eny—would be discarded.
The charge that the CaMV 35S promoter could cause cancer was
constructed in the following way. Ho and her colleagues call the CaMV
promoter “promiscuous” because it works in plants, yeast, algae, and
bacteria. “It has the possibility of promoting inappropriate over-ex-
pression of genes in all species to which it happens to be transferred,”
they claim, adding, “One consequence of such inappropriate over-ex-
pression of genes may be cancer.” That is quite a leap of logic, and it
rests on no facts.
Because the CaMV promoter works in plants, yeast, algae, and bac-
teria does not mean that it is expressed in human cells. Indeed, it
doesn’t work particularly well other than in plants. Even if it did work
in humans, to cause cancer it would not only have to get into a human
cell, it would have to insert itself into the human genome in just the
right spot to turn on a cancer gene. Lots of cancer-causing genes have
been identified and sequenced, along with their promoters. There are
indeed rearrangements of human chromosomes that bring a promoter
together with a cancer-causing gene. But none of them are plant pro-
moters, despite the fact that humans have, over the centuries, eaten a
lot of plant viruses in their food.
CaMV 35S is naturally found in the cauliflower mosaic virus. A
cauliflower—or a cabbage or broccoli or any other of the brassicas—
infected with cauliflower mosaic virus often doesn’t show any symp-
toms. The plants might be stunted. Sometimes the veins in the leaves
of a cauliflower look clear or are banded in green. On the leaves of
certain turnip varieties, particularly one called Just Right, the virus pro-
duces clear spots making the leaf look mottled. On cabbages and Chi-
nese cabbages, black specks called pepper spot or fly speck can develop
after the vegetable has been picked, although some scientists believe
these spots are caused by the turnip mosaic virus, a related virus that
often infects the plants at the same time. The virus is spread by aphids
from a number of weed hosts, including mustard, penny cress,
shepherd’s purse, charlock, and chickweed. “Management is difficult,”
notes a flyer from the University of California at Davis. If the edges of
fields are not conscientiously weeded, the crops will most likely be in-
fected. If the cauliflower or broccoli head shows no sign of infection
(and generally it doesn’t), it is sold, cooked, and eaten, along with
its viruses.
Ho and her colleagues assert that the viral DNA, and particularly
the CaMV 35S promoter, in a virus-infected cauliflower is somehow
different from the CaMV 35S in a genetically modified plant. The
CaMV 35S promoter in the latter is “naked,” they say, and it is known
that human cells take up naked DNA. But, in fact, the CaMV 35S pro-
moter is much less naked in a plant cell than it is in the original virus.
A virus is generally a piece of DNA (or RNA) wrapped in a protein
coat. By contrast, the DNA in a plant is hidden inside a cellulose box
(the plant cell itself), inside the membrane-bound nucleus, and
wrapped up in proteins inside of chromosomes. The CaMV promoter
is much better dressed in a plant cell than it was in the virus.
What’s more, the CaMV 35S promoter is much more dilute in the
plant than it is in the virus. In the virus, the CaMV 35S promoter con-
stitutes roughly 10 percent of the virus’s whole genome. (The length of
the promoter used varies between 350 and 1,200 base pairs; the whole
CaMV genome is 8,000 base pairs.) Put into corn, the same promoter
is only about 0.004 percent of the 2.5 billion base pairs in the corn
genome. In other words, one cauliflower cell might release hundreds
of viruses, and each of those virus particles will contain a CaMV 35S
promoter. But if that promoter were in a corn cell instead, that cell
would have just two copies: one in the paternal genome and one in the
maternal genome. So on both grounds, eating a virus-infected head of
cauliflower brings us into contact with much more of the CaMV 35S
promoter than eating an ear of genetically modified corn. Eating cauli-
flower isn’t high on the list of cancer-causing behaviors.
-------------
If eating genes isn’t hazardous, then, are genetically modified foods
safe to eat? That depends on the gene or genes added. But more im-
portantly, it depends on the food itself. As political scientist Robert
Paarlberg notes in his book The Politics of Precaution: Genetically Modi-
fied Crops in Developing Countries, “Eating any food can be dangerous.”
Many familiar foods naturally contain toxic chemicals—poisons
the plant uses to defend itself against insects and browsing animals.
Lima beans contain a chemical that breaks down during digestion into
hydrogen cyanide, which is poisonous. Toxic psoralens in celery cause
skin rashes. Moreover, psoralen cross-links the strands of DNA to each
other, which can cause cancer. A chemical in cauliflower can make the
thyroid enlarge. Carrots contain a nerve poison and a hallucinogen.
Peaches and pears promote goiters. Strawberries contain a chemical
that prevents blood from clotting and can lead to uncontrollable
bleeding. Peas, beans, cereals, and potatoes contain lectins, which cause
nausea, vomiting, and diarrhea, as do the glycoalkaloids also found in
potatoes. Mushrooms, squash, cucumbers, chickpeas, mustard, manioc,
olives, coffee, and tea all contain chemicals that are toxic to humans.
Yet people generally know from experience how to prepare and eat
these foods safely. Few even know about the toxins, because they are
present at levels too low to harm someone eating a varied diet. (You
would have to eat 400 carrots at a time to receive a toxic dose of nerve
poison.) But some foods contain enough of a harmful chemical to be
toxic in the amounts ordinarily consumed. By trial and error over mil-
lennia, people have learned to eliminate the toxins while preparing the
food. In the Andes bitter potatoes are detoxified through freeze-dry-
ing. In Mexico and the American Southwest they are dipped in clay.
Both methods keep the glycoalkaloids from causing stomach pains and
vomiting.
Manioc, from which cassava meal is ground, “is more poisonous
than the potato,” noted Jack Harlan in The Living Fields. “It contains
cyanogenic glycosides which when broken down by enzymes release
prussic acid, HCN, one of the most deadly compounds known to man.”
To be safely eaten, a manioc tuber must be peeled, grated, and pressed
to expel the juice. The meal must be sun dried, fermented, or heated
overnight until all possible HCN is formed. Then, when the cassava
cake is cooked, the HCN is destroyed by the heat. To eat olives or olive
oil, the extremely bitter oleuropein must first be pressed out—the ancient
Romans used it as a weed killer and insecticide; it is now considered a
pollutant. Yet olive oil, cassava, and potatoes are not only considered to
be safe foods, they are staple foods on which cultures depend.
Because proving “a complete absence of danger is,” in Paarlberg’s
words, “beyond the capability of experimental science,” the practical
definition of a safe food has long been based on experience. “If a food
has been a familiar component of the human diet for some time with-
out any known adverse effects,” Paarlberg explains, “it comes to be ‘gen-
erally recognized as safe’—or GRAS, to use the terminology of the
United States Food and Drug Administration (FDA).”
A novel food can achieve GRAS status if it can be shown to be
chemically equivalent to a food with which we are already familiar.
Such was the case with canola oil when it was approved for sale in the
United States in 1987. Canola oil was the result of a conventional breed-
ing project to modify rapeseed, Brassica napus. Cooking oils pressed
from the seeds had been popular until nutritional experiments in the
1940s showed that erucic acid, one of the major fatty acids in rapeseed
oil, was toxic. By 1968 Canadian plant breeders had reduced the levels
of erucic acid in rapeseed oil dramatically. In 1974, having coined the
term canola (from Canadian Oil) to describe varieties containing less
than 2 percent erucic acid, they began the process of marketing canola
oil in the United States.
Yet the new canola oil was not equivalent to the familiar—and
harmful—rapeseed oil. Nor did it compare exactly with any other com-
mon vegetable oil, because oils vary considerably depending on the
plant and its growing conditions. Instead the Canadian breeders com-
pared the individual fatty acids in canola oil to the fatty acids in soy,
corn, peanut, safflower, sunflower, and olive oils. They checked toxico-
logical databases and did feeding studies in both animals and humans.
They—and the FDA, which accorded it GRAS status in 1987—con-
cluded that canola oil was safe.
Canola oil was used as a case study when the question of how best
to assess the safety of new genetically modified foods was discussed by
the international working group called together by the Organisation
for Economic Co-Operation and Development (OECD) in the early
1990s. Thirty countries belong to the OECD, including 23 in Europe,
Australia, New Zealand, Japan, Korea, Mexico, Canada, and the United
States. The process they agreed upon depends on the idea of familiar-
ity or, to use the FDA’s term, substantial equivalence. As members of
the Food Directorate of Health Canada recently wrote, “Stated most
simply, substantial equivalence encourages investigators to compare a
product which they have to assess with one with which they are already
familiar.”
Substantial equivalence is often misunderstood. For example,
Michael Pollan in the popular book The Botany of Desire writes, “The
Food and Drug Administration told me that, because it operates on
the assumption that genetically modified plants are ‘substantially
equivalent’ to ordinary plants, the regulation of these foods has been
voluntary since 1992.” Pollan’s errors are common ones. As a 2002
handbook for scientists, Genetically Modified Crops: Assessing Safety,
points out, such criticisms follow from “the mistaken perception that
the determination of substantial equivalence was the end point of a
safety assessment rather than the starting point.” The end point in the
United States is GRAS status—the food is generally recognized as safe.
To reach that point plant breeders compare their new crop with one
with which they are familiar, that is, a popular variety of the same plant.
They look for differences in how well and where the two crops grow, as
well as for changes in levels of nutrients and toxic chemicals in the
food. As a group of botanists and nutritionists writing in Nature Biotechnology
pointed out in 2002, “Only those plants that meet the most
stringent performance and safety criteria will advance to the regula-
tors’ desks where the results of the safety studies will be independently
assessed.”
The first part of this assessment is one that plant breeders have
been performing—voluntarily—for years. (The fact that the assess-
ment is voluntary has nothing to do with the idea of substantial equiva-
lence, as Pollan implied.) When the colorful sweet corn called Indian
Summer—whose red, white, yellow, and purple kernels intensify in
color when cooked—and the gourmet orange-fleshed watermelon
called New Queen were developed in the late 1990s, for example, they
were tested through voluntary field trials overseen by All-America Se-
lections. The judges were horticultural professionals; the site of each
trial was a garden at a university, a seed company, or a horticultural
institution. There Indian Summer and New Queen were grown side by
side with the crops most like them then on the market. The judges
compared such traits as yield, time of harvest, and how easy they were
to grow. They looked at the plants’ resistance to pests and diseases, and
marked down idiosyncrasies about each plant’s shape and growth hab-
its (for instance, New Queen needs 9 feet of garden space, while Indian
Summer has to be isolated from any other corn pollen). They judged
each crop’s taste by eating it and its quality by how it looks and keeps. A
crop that outperforms its closest market rival—as both Indian Sum-
mer and New Queen did—is given the stamp of an All-America Selec-
tions Winner.
The second step in determining substantial equivalence is new. As
the authors of Genetically Modified Crops: Assessing Safety note, “Nor-
mally, new varieties of foods and crops have not been subjected to tra-
ditional toxicological testing.” Determining if the bright colors of
Indian Summer and New Queen marked a rise—or fall—in the level
of a nutrient or a toxic chemical was not part of the All-America Selec-
tions trial or any other official test. As Jim Maryanski of the FDA notes,
“This is in contrast to what is done with engineered varieties, where
there are far more tests being done for nutrients, toxins, vitamins and
minerals, and so forth.”
In the late 1960s Wilford Mills of the Pennsylvania State Univer-
sity crossed the popular potato variety Delta Gold with a wild potato
relative from Peru. He called the new variety Lenape. The wild genes
made Lenape highly resistant to attack by insects and potato blight.
“Lenape was a wonderful potato,” remembers Mills’s colleague Herb
Cole. “It chipped golden.” A Pennsylvania potato chip manufacturer
earmarked it as a favorite. A potato breeder in Ontario thought Lenape
would be a good variety for early potatoes, the kind harvested young
and boiled with peas. “He cooked up a batch of potatoes and peas,”
Cole said, telling the story in 2003, “and he got very nauseous. He fig-
ured it was just an accident, so he cooked up some more the next week.
He got even sicker.” He asked a biochemist at his university to analyze
the potatoes. It turned out that they were exceedingly high in
glycoalkaloids, the potato’s natural toxins. He reported the results to
Mills, who enlisted Cole in analyzing the variety further. “The result
was that Penn State recalled the Lenape potato,” Cole said.
Lenape had been released as a public variety, not patented. Recall-
ing it meant contacting every grower of seed potatoes and requesting
that they not market Lenape seed tubers. Cole and Mills did so. “But
some of Lenape’s heritage has carried forward and been bred into other
varieties used today,” Cole added. A list of potato cultivars compiled by
the crop and soil science department of Michigan State University in-
cludes Lenape in the parentage of 13 varieties. “In making the cross,”
Cole concluded, “Bill did what all the people opposed to biotechnol-
ogy say you ought to do. He went back to the origins of the potato and
brought along genes for insect and disease resistance. He also brought
along genes for glycoalkaloids.” But he didn’t know it.
There were no such surprises when Roundup Ready soybeans were
first eaten. Before any were marketed they were compared with regular
soybeans to see if there were differences either in the raw seeds or in
toasted soybean meal. Monsanto’s researchers checked the fatty acid
composition of the oil and its total quantity. They looked at the
amounts of fiber, ash, and water, and compared the carbohydrates and
proteins. They gave special attention to chemicals typically found in
soybeans that could be toxic at higher levels or act as antinutrients.
They did feeding studies in rats, chickens, catfish, and cattle to show
that there were no nutritional differences between Roundup Ready soy
and its market rivals. They found no substantial differences.
Other scientists reported in 1999 that Roundup Ready soybeans
contained 12 to 14 percent less of a substance, called isoflavone, that
might help prevent heart disease, breast cancer, and osteoporosis. Still
a third group of scientists, though, found that Roundup Ready soy
contained too much isoflavone—enough to cause uterine cancer in
laboratory mice. According to scientists from Colorado State Univer-
sity, the difference among the three results is due to environmental
conditions, “the kind of variation in isoflavone levels that can occur
from year to year. . . . Wine afficionados know that the weather can
influence the quality of grapes, causing ‘good’ years and ‘bad’ years for
wine.” The same is true for soybeans. Monsanto’s experiments are the
most trustworthy, in terms of assessing the safety of the crop, the Colo-
rado State researchers confirmed, because Monsanto grew both kinds
of soybeans, the conventional ones and the genetically engineered ones,
in the same fields. With the two types growing side by side, the effects
of the weather can be discounted. Neither of the other experiments com-
pared crops from the same field grown in the same season. As
Genetically Modified Crops: Assessing Safety concludes, “The choice of
the comparator is therefore crucial to the effective application of sub-
stantial equivalence in establishing the safety of a GMO-derived food.”
Some critics of genetically modified food, such as Michael Hansen
of Consumers Union, want to see plant breeders do more. In a chapter
in the 2001 volume Genetically Modified Organisms in Agriculture: Eco-
nomics and Politics, he argued that the FDA should require plant breed-
ers to identify the total number of copies of each gene they insert into a
plant, the location of each one on a chromosome, the structure of each
insert, including a complete genetic map and the full DNA sequence,
and the sequence of at least 10,000 base pairs of DNA on either side of
it. He further believes that the FDA must demand evidence of both the
structural and functional stability of the insert over multiple plant gen-
erations. In 2003 the FDA included Hansen on a committee to review
their guidelines. Jim Maryanski was asked specifically if there was any
case, out of the 50 or more that the FDA has dealt with so far, in which
knowledge about the insertion site said anything of value about food
safety. Maryanski’s answer was a resounding “No.”
To someone familiar with the genomes of plants, suggesting such
requirements is puzzling. Even two inbred strains of the same crop will
show differences in their gene sequences. Rearrangements and trans-
positions are quite common. As for stability, hybrids—whether cre-
ated through new methods or old—are not stable over time. After a
while they decline in performance—leaving the market looking for the
next new variety. Wide crosses, too, trigger instability. Nor are these
necessarily the work of scientists and plant breeders. Some simply hap-
pen naturally, such as the distant crosses that gave rise to our contem-
porary strains of wheat. Avraham Levy and his collaborators at the
Weizmann Institute in Israel have recreated those crosses and studied
what happens to the genes and the genomes. They found many genetic
changes in these wide hybrids: genes shut down, transposons were ac-
tivated, and many genes were simply deleted.
At the molecular level, instability is a fact of life. Making sure the
DNA around the new gene doesn’t change—as demanded by
Hansen—doesn’t provide any information about the safety of the food.
The FDA advisory committee came to the conclusion that having more
information about the inserted gene and the DNA sequence surround-
ing it wasn’t nearly as important as having more information about
the chemical composition of the food produced from the plant.
Indeed, plant geneticists have been surprised at how few gene in-
sertions have a visible effect on plants. Scientists studying the little
laboratory weed Arabidopsis thaliana have used Agrobacterium to make
thousands upon thousands of T-DNA insertions in and near genes to
see what effect disrupting a particular sequence has on the plant. Most
of the time, inserting a new gene has no effect at all; the plants grow
normally. A few insertions cause mutations, but for every hundred only
one or two make a visible change in how the plant looks, grows, or
reproduces.
Gene-activation libraries have been made in which the T-DNA has
a CaMV 35S promoter at its very end to activate genes. In these experi-
ments scientists are doing quite deliberately what Mae-Wan Ho feared
the 35S promoter might do on its own. To the investigators’ disap-
pointment, most of the insertions don’t have any effect. The same is
true of transposon insertions—most don’t cause much change. The
vast majority of plants with new insertions—and their offspring, too—
are quite normal.
How a plant grows and reproduces is simply not as sensitive to the
disruption or activation of any one gene as we might fear. These ex-
periments tell us—if we haven’t yet learned it from irradiating plants,
culturing them, and treating them with chemical mutagens—that
plants and their information storage systems are quite robust. A gene’s
location, while not unimportant, is less important than what the gene
is and what it codes for. Nor is the loss of a gene necessarily a catastro-
phe. Many genes are represented by several copies—sometimes hun-
dreds of copies. Though each copy might be a bit different, they can
often cover for each another so that the loss of one copy goes unno-
ticed. There can also be more than one way to carry out a needed bio-
chemical function; if one route is blocked, another is often available.
The contemporary molecular plant breeder can be confident that
a corn plant with an extra gene is still a corn plant—a fact that Luther
Burbank simply took for granted. And if an insertion is bad for the
plant, the contemporary breeder does precisely what Burbank did—
throws the plant out. But modern breeders have analytical tools avail-
able to them that Burbank didn’t have. They can analyze the proteins,
nucleic acids, fats, and starches, as well as all of the many molecules
(called secondary metabolites) that the genetically modified plant
makes. They can compare this analysis to the chemical profile of the
variety from which the plant was derived and ask, Is this the same plant,
except for that one protein encoded by the added gene?
Calgene’s FlavrSavr tomato was the first genetically modified whole
food. When Calgene brought it to the FDA in 1992, the tomato was
subjected to $2 million-worth of testing by the FDA on top of the test-
ing done by Calgene. In a public meeting the FDA scientists brought
the results of their extensive and sophisticated chemical analyses to a
panel of external advisors; the panel included representatives of public
interest groups and industry, as well as scientists whose specialties
ranged from nutrition to basic plant science. The concluding slide of
the FDA’s presentation had a simple message: Calgene’s transgenic to-
mato . . . is a tomato.