IEEE Spectrum June, 2013 - 52

The age of plenty

ters of whirring, chugging, and occasionally banging machines
do their work.
What goes on here is akin to what happens in one of the server
farms that run Google's search engine. The lab is an information
factory whose primary purpose is to answer a single question more
than a 100 million times per year: Does this seed contain variants
of genes that are associated with these traits? The difference is that
the database is, quite literally, the DNA of various crop strains and
the search is executed using biochemistry. Armed with the answers
to their queries, breeders can know whether it's worth growing
that seed and which of thousands of other seeds to cross it with.
The basic procedure is this: First, DNA is
extracted from a bit of a plant and unzipped
down the middle, so that a complementa-
ry sequence can bind to it. (In the search
engine analogy, that DNA is the database.)
Then short sequences of DNA, called genetic
markers, are added. (These are analogous
100 million
to the search terms.) Usually not the com-
number of genetic
plete genes themselves, these markers code
experiments
for distinct patches of DNA in the plant's
duPont Pioneer's
genome that are of interest to breeders.
new roboticized
genotyping lab can
For instance, a particular sequence might
perform per year
be found in all maize that can survive in
near-desert conditions but not in any that
can't. The marker, which is chemically la-
6 TerabyTes
beled so it can be identified later, will stick
maximum total
daily data output of
to the plant's unzipped DNA if it's a match
bgi's sequencing
or wash away if it isn't.
machines
Decades ago, it was a tricky procedure.
Now, what was once months of work that
us $500 000
would earn you an advanced degree is
cost of each of
done tens of thousands of times each day.
128 sequencing
At Pioneer's lab, robotic arms swing to a
machines bgi
hypnotic beat, moving cassettes of plant
purchased in 2010
material from one step to another, adding
just the right amount of just the right chemi-
cals and markers to each of thousands of
DNA samples. All the while technicians and scientists keep track of
the results on giant screens.
What's all this automation good for? The most easily measurable
metric is that "running a query" now costs mere pennies instead of
hundreds of dollars. "We couldn't do hundreds of millions of data
points a year if it cost what it did five years ago," says John Arbuckle,
senior research director for Pioneer's production genotyping
facilities. That means more varieties can be screened and breeders
can make a better selection of what plants to cross-pollinate with
which. "Out of the thousands of possible plants to put out into the
field, we can choose the fraction that is likely to succeed," he says.
52

|

jun 2013

|

north american

|

SPectrum.ieee.orG

That means getting not just a better crop-"increasing the rate of
genetic gain," he calls it-but also a faster one. "What used to take
seven or eight generations to get now takes three to five."
To fully appreciate the change, you have to understand what an
insanely frustrating experience breeding was before industrial-
scale genotyping. If you wanted to see if any of your soybean
strains were vulnerable to a particular disease, you couldn't just
query their genomes to see if any had the protective genes. You
had to find a farm where the disease had occurred in the previ-
ous season, rent out the field, and "then pray that the weather
would be right for that disease to occur," says John Soper, Pioneer's
vice president of crop genetics R&D, who started his career as a
soybean breeder in the 1980s. "And 9 times out of 10 it didn't."
On the economic side, the technology can be used to improve the
yield of crops for markets that were heretofore too small to catch the
attention of the big seed firms. It makes "the technology more ac-
cessible to the crops and regions that in the past haven't been able
to afford it," says Arbuckle. Those markets might be characterized
by a peculiar mix of conditions-soil, weather, pests, temperature-
that don't occur elsewhere, and can't be served particularly well
by what companies already have on hand. For instance, Pioneer is
part of an effort led by the United States Agency for International
Development to come up with maize that can thrive in the million
hectares of South Asia that experience excruciating heat during the
crop's crucial flowering season. Genotyping can enable a breeder to
predict which maize has genes that might make it flourish in such
conditions and breed those into a high-yield strain.

ofTen you can geT everyThing you wanT from a planT's own
genome, but sometimes you just can't. There are categories of prob-
lems for which no good solutions exist in a plant's genome. Insect
resistance is generally such a problem, says Soper. Or if you want
to add a nutrient to a crop that just doesn't make it, you're going to
have to add some new blueprints to that crop's genome.
The tried-and-true method of genetic modification involves a par-
ticularly useful pest called Agrobacterium tumefaciens (or Rhizobium
radiobacter if you're feeling formal), a soil bacterium that causes
tumors in plants. How it forms those tumors is the interesting part:
It injects a small segment of DNA into the infected plant's cells,
and that DNA gets incorporated into the cell's genome. Starting
in the late 1970s, scientists realized that they could engineer the
agrobacterium's delivery system to inject useful genes such as the
one for the toxin made by the microbe Bacillus thuringiensis (thus
insect-resistant corn and cotton) or for an enzyme unaffected by
the herbicide glyphosate (thus Roundup Ready soybeans).
It's not quite that simple, of course, and the method has some
hang-ups. The agrobacteria inject their DNA payloads into indi-
vidual cells, so to get a whole genetically modified plant, you often
http://SPectrum.ieee.orG
Table of Contents for the Digital Edition of IEEE Spectrum June, 2013
