For more than
a decade, researchers have been trying
to create accurate computer models of
Escherichia coli (E. coli), a
bacterium that makes headlines for its
varied roles in food poisoning, drug manufacture
and biological research.
By combining laboratory data with recently
completed genetic databases, researchers
can craft digital colonies of organisms
that mimic, and even predict, some behaviors
of living cells to an accuracy of about
Now, NSF-supported researchers at the University
of California at San Diego have created
a computer model that accurately predicts
how E. coli metabolic systems
adapt and evolve when the bacteria are
placed under environmental constraints.
Bernhard Palsson, Rafael Ibarra (now at
GenVault Corporation in Carlsbad, California)
and Jeremy Edwards (now at the University
of Delaware at Newark) report their findings
in the November 14 issue of Nature.
"Ours is the only existing genome-scale
model of E. coli," says Palsson.
In addition, while many approaches to
genetics experiments "knock out" individual
genes and track the results, the new model
takes a whole-system approach. Changing
one aspect of a genetic code could be
irrelevant if an organism adapts and evolves,
says Palsson. The constraints-based models
allow the E. coli to evolve more
naturally along several possible paths.
Scientists may use the approach to design
new bacterial strains on the computer
by controlling environmental parameters
and predicting how microorganisms adapt
over time. Then, by recreating the environment
in a laboratory, researchers may be able
to coax living bacteria into evolving
into the new strain.
The resulting strains may be more efficient
at producing insulin or cancer-fighting
drugs than existing bacterial colonies
engineered by researchers using standard
"Now we have a better tool to predict how
bacteria evolve and adapt to changes,"
says National Science Foundation program
director Fred Heineken. "As a result,
this constraints-based approach could
lead to better custom-built organisms,"
The researchers based their digital bacteria
on earlier laboratory studies and E.
coli genome sequences, detailed genetic
codes that have been augmented with experimental
information about the function of every
Such digital models are known as "in
silico" experiments -- a play on
words referring to biological studies
conducted on a computer.
In the first days of testing on living
organisms, the bacteria did not adapt
into the strain predicted by the simulation.
Yet, with more time (40 days, or 500-1000
generations), the E. coli growing
in the laboratory flasks adapted and evolved
into a strain like the one the in
silico model predicted.
"The novelty of the constraints-based approach
is that it accounts for changes in cellular
properties over time," says Palsson. "Fortunately,"
he adds, "the other advantage is that
it actually works surprisingly often."
For many years, drug manufacturers have
manipulated the genetic code in E.
coli strains, creating species that
can produce important substances, such
as the hormone insulin for use by people
with diabetes or the experimental cancer
Using the new constraints-based techniques
Palsson and his colleagues developed,
drug manufacturers and bioprocessing companies
could use computers to determine the genetic
code that could yield the most efficient
and productive versions of E. coli,
and then use adaptive evolution to create
bacterial strains that have the desired
Says Palsson, "This development potentially
opens up a revolutionary new direction
in the design of new production strains."
In addition, says Palsson, "now that we
have gained a greater understanding of
this process in E. coli, developing
similar simulations of other organisms
should take less time."
Photomicrograph of E. coli.
Photo Credit: Image courtesy of National
Institute of Allergy and Infectious Diseases,
National Institutes of Health
The two images show how E. coli bacteria
in a laboratory evolved over time to metabolize
and grow at a rate predicted by a computer
simulation. The bottom left axis represents
the amount of oxygen that the bacteria
consumed, the bottom right axis represents
the amount of glycerol (the food) the
bacteria consumed, and the vertical axis
represents the rate at which the bacteria
The colored surfaces are called "phenotype
phase planes." They graphically represent
the researchers' in silico (computer)
prediction for the possible ways in which
the bacteria could grow under specific
environmental conditions. The red shading
represents faster growth conditions, the
green represents slower growth.
The red line is the "line of optimality,"
the optimal growth rate predicted by the
researchers' computer model. Region 1
represents a phase where growth is not
optimal, Region 2 represents a phase where
the bacteria consume too much food and
therefore have to secrete some as byproducts
The white dots are measurements of how
several E. coli specimens were metabolizing
food, and how fast they were growing when
the researchers tested them. After 40
days (700 generations), the bacteria evolved
to metabolize as predicted by the researchers'
in silico model (all specimens cluster
along the line of optimality).
Select image for a larger version.
Photo Credit: Images courtesy
of the Genetic Circuits Research Group
(Rafael U. Ibarra, Jeremy S. Edwards,
and Bernhard O. Palsson).
Select images for larger versions
versions (Total Size: 128KB) of images
from this document