Corn (maize) is arguably man's first, and perhaps his greatest, feat of genetic engineering. Its huge ears--each packed with firmly attached kernels filled with starch, protein, and oil--make it a food staple. Contemporary corn, unlike its wild grassy ancestor teosinte, can't survive without people because it can't disperse its own seeds. The origins of maize have long intrigued geneticists, but only recently have new molecular methods enabled evolutionary sleuths to pinpoint its origins and identify the genetic modifications (GMs) that enabled the radical transformation of teosinte into contemporary maize. On page 1206  of this issue, Jaenicke-Després, Doebley, and their colleagues ( 1) provide the latest chapter in this detective story and suggest that prehistoric people were quick to adopt GM corn.
Teosinte and corn ( Zea mays) don't look much alike, but they are interfertile. Teosinte-corn hybrids arise in the wild but look so different from either parent that they were originally classified as a different species ( Zea canina). In the 1920s, Beadle examined chromosomes in teosinte-corn hybrids and concluded that the two plants belonged to the same species, and even shared the same chromosomal order of genes. That should have resolved the question of corn's origins, but it didn't.
In 1938, the eminent maize geneticist Mangelsdorf proposed that maize evolved from an extinct South American maize species and that teosinte originated from a cross between another grass, Tripsacum, and maize ( 2). Although cumbersome, this hypothesis was widely accepted, and Mangelsdorf and Beadle sparred publicly for years. Upon retirement, Beadle organized an expedition to Mexico to look for more wild maize relatives, returning with seeds that proved invaluable to the next generation of molecular archaeologists. The Tripsacum hypothesis was briefly resurrected in the mid-1990s, but by then molecular evidence overwhelmingly favored the notion that teosinte was the ancestor of modern maize ( 3).
So how, when, and where was teosinte transformed into maize? Beadle gave his mentor, Emerson, credit for the idea that just a few mutations changed teosinte into maize ( 4). Analyzing backcrossed maize-teosinte hybrids with molecular probes, Doebley's group came to a startlingly similar conclusion: The differences between maize and teosinte could be traced to just five genomic regions ( 5). In two of these regions, the differences were attributable to alternative alleles of just one gene: teosinte glume architecture ( tga1) and teosinte branched ( tb1), which affect kernel structure and plant architecture.
The tga1 gene controls glume hardness, size, and curvature ( 6). Teosinte kernels are surrounded by a stone-like fruitcase, assuring their unscathed passage through an animal's digestive tract, which is required for seed dispersal. But the plant's reproductive success is the consumer's nutritional failure. Not surprisingly, one of the major differences between maize and teosinte kernels lies in the structures (cupule and outer glume) enclosing the kernel. Maize kernels don't develop a fruitcase because the glume is thinner and shorter and the cupule is collapsed. The hardness of teosinte kernels comes from silica deposits in the glume's epidermal cells and from impregnation of glume cells with the polymer lignin. The maize tga1 allele supports slower glume growth and less silica deposition and lignification than does the teosinte tga1allele.
The tb1 locus is largely responsible for the different architecture of the two plants. Teosinte produces many long side branches, each topped by a male flower (tassel), and its female flowers (ears) are produced by secondary branches growing off the main branches. Modern corn has one main stalk with a tassel at the top. Its lateral branches are short and bear its large ears. Much of the difference is attributable to the tb1 gene, originally identified in a teosinte-like maize mutant. Mutations generally abrogate gene function, indicating that the maize allele acts by suppressing lateral shoot development, converting grassy teosinte into slim, single-stalked modern corn and male into female reproductive structures ( 7).
Knowing that this cluster of traits is controlled by just two genes makes it less surprising that genetic differences in these genes could render teosinte a much better food plant. Yet however useful to people, a tga1 mutation would have been detrimental to teosinte, making it more vulnerable to destruction in the digestive tract of the consumer and so less able to disperse its seeds. Thus, the only way this mutation could have persisted is if our ancestors propagated the seeds themselves. This implies that people were not only harvesting--and likely grinding and cooking--teosinte seeds before these mutations came along, but also were selecting for favorable features such as kernel quality and cob size. In turn, this suggests a "bottleneck" in corn evolution: Several useful GMs were brought together in a single plant and then the seeds from this plant were propagated, giving rise to all contemporary maize varieties. Such a prediction can be tested by calculating the number of generations and individuals it would take to account for the molecular variability present in contemporary maize. The results of such a test suggest a bottleneck for maize domestication of just 10 generations and a founding population of only 20 individuals ( 8). Did this happen once or many times? Because genetic differences arise at a fairly constant rate, this question can be answered by constructing family trees using similar sequences from different varieties of teosinte and contemporary maize. The results are unequivocal: All contemporary maize varieties belong to a single family, pointing to a single domestication event.
Knowing how quickly differences arise, how many there are today, and where the family of origin survives, it is possible to determine when--and where--it all started. The answer is that maize most probably arose from teosinte of the subspecies parviglumis in the Balsas River basin of southern Mexico roughly 9000 years ago ( 9). Recent redating of cobs from the Guilá Naquitz cave (about 500 km from the Balsas River basin) demonstrated that they were more than 6200 years old, providing archaeological support for the molecular findings ( 10, 11). These earliest corn cobs don't look much like those of modern corn, but they look even less like teosinte cobs (see the figure). They are tough and have several rows of tightly attached kernels, implying that the plants wouldn't have survived without people to detach and plant the seeds. By contrast, teosinte's reproductive structure, the rachis, falls apart when mature to release its hard seeds. Thus, even 6000 years ago, ancient maize cobs were already corn-like.
Primitive popcorn. Teosinte ( left) and primitive maize ( right). Primitive maize was "reconstructed" by crossing teosinte with Argentinian corn.
CREDIT: JOHN DOEBLEY
The GM corn spread far--and fast. Maize appears in the archaeological record of the southwestern United States more than 3000 years ago ( 12), and it is evident that cob size had already increased under selection. The Jaenicke-Després et al. study ( 1) examines the selection of traits that can't be observed in fossilized cobs. Taking tiny samples of fossil cobs from the Ocampo Caves in northeastern Mexico (2300 to 4400 years old) and the Tularosa Cave in the Mogollon highlands in New Mexico (650 to 1900 years old), the authors extracted DNA and amplified, cloned, and sequenced small DNA fragments of the tb1 gene, the pbf gene that controls the amount of storage protein, and the su1 gene encoding a starch-debranching enzyme whose activity affects the texture of corn tortillas. They compared their ancient DNA sequences with those of 66 maize landraces (the corn grown by indigenous farmers) from South, Central, and North America and 23 lines of teosinte parviglumis.
They report that alleles of these genes typical of modern corn were already present more than 4000 years ago, implying that plant architecture and kernel nutritive properties were selected early, long before corn reached North America. All 11 ancient cobs carried the tb1 allele present in modern corn, but fewer than half of the 23 teosinte varieties carried this allele. Similarly, all ancient samples contained a pbf allele that is common in corn but rare in teosinte. The predominant modern su1 allele was found in all of the older Mexican cobs, but the younger New Mexican cobs had several different alleles, suggesting that this gene was still undergoing selection when maize reached North America.
The authors conclude that "... by 4400 years ago, early farmers had already had a substantial homogenizing effect on allelic diversity at three genes associated with maize morphology and biochemical properties of the corn cob." This suggests that once this special combination of GMs was assembled, the plants proved so superior as a food crop that they were carefully propagated and widely adopted, perhaps causing something of a prehistoric Green Revolution. It also implies that the apparent loss of genetic diversity following the introduction of high-yielding Green Revolution wheat and rice varieties in the 1960s and 1970s, and attending the rapid adoption of superior GM crops today, is far from a new phenomenon.
V. Jaenicke-Després et al., Science 302, 1206  (2003). J. Doebley, Trends Genet. 8, 302 (1992) [Medline] . S. White, J. Doebley, Trends Genet. 14, 327 (1998) [Medline] . J. Doebley et al., Nature 386, 485 (1997) [Medline] . A. Eyre-Walker et al., Proc. Natl. Acad. Sci. U.S.A. 95, 4441 (1998) [Medline] . Y. Matsuoka et al., Proc. Natl. Acad. Sci. U.S.A. 99, 6080 (2002) [Medline] . B. F. Benz, Proc. Natl. Acad. Sci. U.S.A. 98, 2104 (2001) [Medline] . D. R. Piperno, K. V. Flannery, Proc. Natl. Acad. Sci. U.S.A. 98, 2101 (2001) [Medline] . B. B. Huckell, J. World Prehist. 10, 305 (1996).
V. Jaenicke-Després et al., Science 302, 1206  (2003).
J. Doebley, Trends Genet. 8, 302 (1992) [Medline] .
S. White, J. Doebley, Trends Genet. 14, 327 (1998) [Medline] .
J. Doebley et al., Nature 386, 485 (1997) [Medline] .
A. Eyre-Walker et al., Proc. Natl. Acad. Sci. U.S.A. 95, 4441 (1998) [Medline] .
Y. Matsuoka et al., Proc. Natl. Acad. Sci. U.S.A. 99, 6080 (2002) [Medline] .
B. F. Benz, Proc. Natl. Acad. Sci. U.S.A. 98, 2104 (2001) [Medline] .
D. R. Piperno, K. V. Flannery, Proc. Natl. Acad. Sci. U.S.A. 98, 2101 (2001) [Medline] .
B. B. Huckell, J. World Prehist. 10, 305 (1996).
The author is at the Huck Institute for Life Sciences, Pennsylvania State University, University Park, PA 16802, USA. E-mail: firstname.lastname@example.org 
Reprinted with permission from Science, Vol 302, Issue 5648, 1158-1159, 14 November 2003