Like clockwork

Timing is everything for Tal Danino and his colleagues. Using computer simulation and genetic engineering, they’re synchronizing bacteria into something like a biological metronome, producing regular, adjustable pulses of activity.

Similar genetic clocks are ubiquitous in nature and govern processes like cell generation, heart function and the circadian rhythms that regulate sleep cycles. But Danino and University of California, San Diego (UCSD) colleagues Octavio Mondragón-Palomino, Lev Tsimring and Jeff Hasty are the first to synthesize a coordinated circuit.

Danino, a fourth-year Department of Energy Computational Science Graduate Fellow (DOE CSGF) and doctoral student at UCSD, is lead author of a paper describing the research, published this week in Nature.

Synchronized oscillations are important to harnessing bacteria for specific purposes, says Hasty, who also is Danino’s doctoral adviser. “If you’re going to have a sensor or other biotechnology application, cells oscillating out of phase will not give you much of a signal. But if they’re all in phase they will have a bigger signal.”

Danino says the project also illustrates computers’ ability to portray and predict gene activity and biological processes.

“We’re getting closer and closer to learning how to spatially and temporally model how synthetic circuits and natural circuits behave in bacteria,” he says. The goal in synthetic biology is “to reprogram bacteria to do things we want them to do and to be able to model those ahead of time” before building them in the lab.

Computer modeling can guide researchers through that process, saving time and money, Danino says. “It allows you to do things outside the reach of experimental procedures.”

Computation is even more important, Hasty says, when modeling coupled systems like the one the group described in the Nature paper. “Most of these circuits people are building in cells are noisy or variable. Once you start worrying about variability and noise, the computation goes up,” requiring more time on more powerful computers.

Biologists are beginning to think of genes and cells the same way engineers regard electrical or computational circuits. Genes carry the blueprints for proteins, cellular workhorses that comprise many body tissues but also act as enzymes and antibodies. Manipulating genes can make them into switches and logic gates capable of directing cell activity.

In a previous paper, UCSD researchers demonstrated oscillation in a single bacterium. In effect, they engineered a feedback loop to turn selected genes on and off, producing regularly spaced pulses of activity.

“We’ve basically learned how to build an oscillator in a single cell, but the oscillation wouldn’t be synchronized” across a population of cells, Danino says. “Each cell would do its own thing.”

The researchers want cells to work in concert, and they want to adjust that coordinated behavior the way a metronome’s timing can be adjusted. They did it by harnessing something called quorum sensing.

“In naturally occurring networks, (quorum sensing) is a way to coordinate behavior,” Hasty says. A colony of bacteria encountering a stressor or environmental change may find it advantageous to work together. In quorum sensing, the cells broadcast chemical signals and “agree” on a response.

For the Nature paper, “We built quorum sensing into the architecture of the oscillator,” Hasty says. “The cells broadcast to their neighbors the phase of their oscillation” so the response is synchronized.

The team first placed the luxI gene from the Vibrio fischeri bacterium and the aiiA gene from Bacillus thuringiensis into Escherichia coli, along with another gene that generates a fluorescent protein. All three were placed under the control of identical copies of a luxI promoter gene.

LuxI is a quorum-sensing gene, Danino says: It detects when cells reach a density high enough to generate a chemical signal. AiiA, conversely, is a quorum-disrupting gene that dampens the chemical signal.

Traditional molecular biology techniques would make it difficult to characterize the colony’s behavior.

The protein encoded by the luxI gene acts enzymatically to produce acyl-homoserine lactone (AHL), a molecule that diffuses across membranes to other E. coli, where it activates the luxI promoter. That, in turn, boosts expression of the luxI, aiiA and fluorescent protein-producing genes inserted into the bacteria. AHL levels consequently rise, and it flows out of each successive bacterium and into other bacteria.

But aiiA encodes a protein that degrades AHL. As that protein increases, AHL levels fall, repressing the luxI promoter, which in turns slows synthesis of the luxI, aiiA and fluorescent proteins in a kind of feedback loop: an activator triggers production of a chemical that represses the activator itself.

“That’s how we built this positive and negative feedback loop,” Danino says. “We looked for an organism that would degrade AHL.” The feedback loop generates synchronized waves of gene activity through the E. coli colony. The presence of the fluorescent protein makes the waves visible.

Danino and his colleagues also found knobs that adjust both the amplitude and frequency of the waves. One way they tested relies on regulating cell density to control quorum sensing.

“We built these microfluidic devices that will specifically hold an amount of cells that will activate the signal, but not so many that it will be difficult to study them,” Danino says. The device has a channel to deliver nutrients and gene inducers to E. coli growing in a chamber the width of a hair. Excess cells leak out of the chamber and are swept away through the channel, allowing the bacterial colony to continuously grow while remaining a fixed size.

The experiments showed that in areas of high cell density, like the colony center, and areas of low density, like the colony’s edge, the oscillations were weak and spaced closely together. But like Baby Bear’s bed in “Goldilocks and the Three Bears,” intermediate cell densities are just right – producing stronger, lower-frequency gene response waves than in high- or low-density areas.

“That’s because if you have a small amount of AHL you will get a low signal,” Danino says. “But it also turns out if you have too much AHL it shuts down production,” reducing the signal. Low-density areas produce less AHL; high-density areas produce more.

The researchers used the sensitivity to AHL as another way to control oscillations. Adjusting flow through the microfluidic channel let them change AHL concentrations in the colony, with a corresponding effect on oscillations.

Slowing the flow accelerated oscillations to around once every 55 minutes and decreased the strength of the gene response, the researchers found. Increasing the flow rate gradually slowed the oscillation frequency to around once every 90 minutes and increased amplitude. When flows were faster, gene activity fell to almost zero between waves, while under slow flows gene activity between waves was higher than the baseline.

In the paper, the researchers note that AHL both enables activation of the genes necessary for oscillations and mediates coupling between cells. “By controlling the flow rate you adjust how much of the AHL is washed out and that controls the period and the amplitude of the oscillation,” Danino says. “That’s really a huge benefit of using these microfluidic devices.” Traditional molecular biology techniques would make it difficult to characterize the colony’s behavior.

Computer models agreed well with the experiments, the Nature paper says. They showed waves of gene activity were spaced farther apart and were stronger under fast flows but were of higher frequency and lower amplitude under slow flows. The models showed each oscillation begins as the proteins encoded by aiiA and luxI slowly accumulate, then burst to high levels. The protein bursts suppress AHL and, in turn, the production of the two proteins. The process repeats after the proteins encoded by luxI and aiiA decay enzymatically.

The researchers also used computation to understand how single cells behave. They set the model parameters to artificially block AHL diffusion across cell membranes in the colony. When decoupled from the environment and each other, the individual cells oscillated independently for any cell density. That strengthens the conclusion that coupling through AHL diffusion is one way to synchronize oscillations at intermediate cellular concentrations, the paper says.

The finding about single-cell behavior was possible only through modeling, Danino says. “It’s not something you could shut off in an experiment.”

That demonstrates the increasing value of computation as a tool for understanding gene networks, he adds. “It’s furthering our knowledge about how to design these circuits. There’s typically a lot of fine-tuning between your first design and how you model it in the end. As the years go by we’re narrowing that gap” between model and reality.

Bacteria engineered for synchronized oscillatory behavior could be used in a number of ways, Danino says. He’s working on one possibility at DOE’s Lawrence Berkeley National Laboratory (LBNL), where he’s resumed a collaboration begun during a 2008 DOE CSGF practicum.

Danino is working with Jay Keasling, leader of LBNL’s Joint Bioenergy Institute. Keasling is known for having engineered yeast to produce the powerful antimalaria drug artemisinin.

“We’re looking at increasing production of that and biofuels by using oscillatory behavior,” Danino says. His computer models indicate it’s healthier for cells to produce such substances via oscillations rather than continuously. Now he’s out to test the idea experimentally.

Unlike research described in the Nature paper, the LBNL work will test unsynchronized cell activity, but coordinated oscillations could be next, Danino says.

The DOE CSGF program has given Danino the background support necessary to develop skills in both wet lab work and computation, says Hasty, his doctoral adviser. “What they’re doing is important in getting this kind of multidisciplinary work to go.”

The fellowship is supported by DOE’s Office of Advanced Scientific Computing Research and National Nuclear Security Agency.

Besides the DOE CSGF, the researchers were supported by the National Institutes of Health and General Medicine and the Mexican science agency CONACyT.