Microcosm

by Carl Zimmer

  On-off switches are everywhere in nature. Prophages remain dormant inside E. coli thanks to repressors that keep their genes shut down. Stress causes the repressors to fall off and the prophages to make new viruses. Operons can be found in other bacteria as well. In animals like ourselves, operons appear to be much less common. But even genes that do not sit next to each other on our genome can be switched on by the same master-control protein.

  It is only through the switching on and off of genes that our cells can behave differently from one another, despite carrying an identical genome. They can form liver cells or spit out bone, catch light or feel heat. By learning how E. coli drinks milk, Jacob and his colleagues opened the way to understanding why we humans are more than just amoebas.

  LIVING CIRCUITS

  To an engineer, a circuit is an arrangement of wires, resistors, and other parts, all laid out to produce an output from an input. Circuits in a Geiger counter create a crackle when they detect radioactivity. A room is cast in darkness when a light switch is turned off. Genes operate according to a similar logic. A genetic circuit has its own inputs and outputs. The lac operon works only if it receives two inputs: a signal that E. coli has run out of glucose and another signal that there’s lactose to eat. Its output is the proteins E. coli needs to break down the lactose.

  E. coli has no wires that scientists can pull apart to learn how its circuits work. Instead, they must do experiments of the sort Jacob and Monod carried out. They observe how quickly the bacteria respond to their environment, how quickly they make a certain protein or clear another one away. Scientists combine the results of experiment after experiment into models, which they use to make predictions about how future experiments will turn out. The fundamental discoveries that Jacob, Monod, and others made about E. coli have led other scientists to pick apart the circuitry of other species, including us. But in the fifty years since Jacob squirmed in a cinema seat, scientists have continued to pay close attention to E. coli. They discovered intriguing patterns in E. coli’s circuitry, which they mapped out in more detail than in of any other species, and they’ve discovered that E. coli’s circuitry mimics the sort of circuitry you might find in digital cameras or satellite radios.

  To prove that I’m not dabbling in idle metaphor, I want to probe the wiring of one of E. coli’s many circuits. This particular circuit controls the construction of E. coli’s flagella. It has taken the work of many scientists over many years to discover most of the genes that belong to this circuit. But in 2005, Uri Alon and his colleagues at the Weizmann Institute of Science in Rehovot, Israel, figured out what the circuit does. It acts as a noise filter.

  Engineers use noise filters to block static in phone lines, blurring in images, and any other input that obscures a true signal. In the case of E. coli, the noise is made up of misleading cues about its environment. With the help of a noise filter it can pay attention only to the cues that matter. It’s important for E. coli to ignore noise when it builds a flagellum because the process is a lot like building a cathedral.

  The microbe must switch on about fifty genes, which make tens of thousands of proteins. Those proteins must come together in a tightly choreographed assembly. First the motor must insert itself in the membranes. A syringe has to slide through the center of the motor, which then injects thousands of proteins into the growing tail. The proteins squirm through the hollow shaft and emerge to form its new tip. The process takes an hour or two, which for E. coli can mean several generations. A new microbe inherits a partially built tail and passes it on, still unfinished, to its descendants.

  By the time E. coli has finished building these flagella, the crisis may be long over. All that energy will have gone to waste. So E. coli keeps tabs on its surroundings, and if life does seem to be getting better, it stops building its flagella. The only problem with this strategy is that a sign of better times may actually be a fleeting mirage. If E. coli abandons its flagella when a single oxygen molecule drifts by, it may end up stranded in a very dangerous place. To E. coli these false signs are noise it must filter out of its circuits.

  To explain how E. coli filters out noise, I will draw a wiring diagram. An arrow with a plus sign means that a signal or a gene boosts the activity of another gene. A minus sign means that the supply of protein is reduced. The first link in this circuit is from the outside world to the inner world of E. coli. When the microbe senses danger, it sometimes responds by producing a protein called FlhDC.

  FlhDC is one of E. coli’s master switches. It can latch on to many spots along E. coli’s chromosome, where it can switch on a number of genes. These genes make many of the proteins that combine to make flagella.

  In this simple form, E. coli’s flagella-building circuit has a major flaw. It can turn on flagella-building genes in response to stress, but it also has to shut them down as soon as the stress goes away. Once the microbe stops making new FlhDC, the old copies of FlhDC gradually disappear. As they do, the genes FlhDC controls can no longer make their proteins. The complex assembly of flagella comes screeching to a halt in response to the slightest improvement. When conditions turn bad again, this circuit has to fire up its flagella machine from scratch. In a crisis, those delays could be fatal.

  E. coli does not fall victim to false alarms, however, because it has extra loops in its genetic circuit. In addition to switching on flagella genes, FlhDC switches on a backup gene called FliA.

  FliA can switch on the flagella genes as well.

  But FliA is also controlled by another protein, called FlgM. It grabs new copies of FliA as soon as E. coli makes them, preventing them from switching on the flagella genes. Here is the circuit with FlgM added:

  FlgM cannot keep FliA repressed for long, however, because E. coli can expel it through the same syringe it uses to build its flagella. As the number of FlgM proteins dwindles, more FliA genes become free to switch on the flagella-building genes.

  Here, at last, is the full noise filter as reconstructed by Alon and his colleagues:

  This elegant network gives E. coli the best of all worlds. When it starts building flagella, it remains very sensitive to any sign that stress is going away. That’s because FlhDC alone is keeping the flagella-building genes switched on. But once E. coli has built a syringe and begins to pump out FlgM, the noise filters kick in. If the stress drops, so does the level of FlhDC. But E. coli has created enough free FliA genes to keep its flagella-building genes switched on for more than an hour. If the respite is temporary, E. coli will start making new copies of FlhDC, and its construction of flagella will go on smoothly.

  E. coli can filter out noise, but it’s not deaf. If conditions get significantly better, E. coli can stop making flagella. Its extra supply of FliA cannot last forever. The proteins become damaged and are destroyed by E. coli’s molecular garbage crews. If the stress does not return in time, the microbe will run out of FliA, and the circuit will shut down. The good times have truly returned.

  Scientists are now starting to map the circuitry of genes in other species as carefully as Alon and his colleagues have in E. coli. But it will take time. Scientists don’t yet know enough about how the genes and proteins in those circuits build good models. In many cases, scientists know only that gene A turns on gene B and gene C, without knowing what causes it to flip the switch or what happens when it does.

  But Alon has discovered a remarkable lesson even in that tiny scrap of knowledge. He and his colleagues have surveyed the genes in E. coli and a few other well-studied organisms—yeast, vinegar worms, flies, mice, and humans. The arrows that link them tend to form certain patterns far more often than you’d expect if they were the result of chance. E. coli’s noise filter, for example, belongs to a class of circuits that engineers call feed-forward loops. (The loop in the noise filter goes from FlhDC to FliA to the flagella-building genes.) Feed-forward loops are unusually common in nature, Alon and his colleagues have shown. Nature has a preference for a few other patterns as well, which also seem to allow
life to take advantage of engineering tricks like the noise filter. E. coli and the elephant, it seems, are built not only with the same genetic code. They’re also wired in much the same way.

  LIFE ON AUTOPILOT

  An orange winter dusk has settled in. Out my window I can see the webs of bare maple branches. Photons stream through the window and patter on the photoreceptors lining my retina. The photoreceptors produce electric signals, which they trade among themselves and then fire down the fibers of my optic nerves into the back of my brain. Signals move on through my brain, following a network made of billions of neurons linked by trillions of branches. An image emerges. I get up from my desk to turn on the lights. At first I can see nothing outside, but after a moment my eyes adjust. I can still see the trees, down to their twigs.

  I must remind myself how remarkable it is that I can still see them. A moment earlier my vision was exquisitely tuned to perceiving the world at dusk. If it had stayed that way after I turned on the light, I would have been practically blinded. Fortunately my eyes and brain can retune themselves for the noonday sun or a crescent moon. If the light increases, my brain quickly tightens my irises to reduce the light coming in. When the lights go out, my pupils expand, and my retinal neurons boost the contrast between light and dark in my field of vision. An engineer would call my vision robust. In other words, it works steadily in an unsteady world.

  Our bodies are robust in all sorts of ways. Our brains need a steady supply of glucose, but we don’t black out if we skip dinner. Instead, our bodies unload reserves of glucose as needed. A clump of cells develops into an embryo by trading a flurry of signals to coordinate their divisions. The signals are easily disrupted, but most embryos can still turn into perfectly healthy babies. Again and again life avoids catastrophic failure and remains on course.

  Until recently, scientists had no solid evidence for where life’s robustness comes from. To trace robustness to its source, they needed to know living things with a deep intimacy—the same intimacy an engineer may have with an autopilot system, using its plans to carry out experiments. But the blueprints of most living things remain classified. Among the few exceptions is E. coli.

  E. coli faces threats to its survival on a regular basis. Set a petri dish on a windowsill on a sunny day and you bring the microbes in it to the brink of disaster. In order to work properly, a protein needs to maintain its intricate origami-like folds. Overheated proteins shake themselves loose. They can no longer do the job on which E. coli’s survival depends.

  Yet E. coli does not die from a few degrees of extra heat. As the temperature rises, the microbe makes molecules known as heat-shock proteins. They defend E. coli in two ways. Some of them embrace E. coli’s jittery proteins and guide them back into their proper shape. Others recognize heat-snarled proteins that have been damaged beyond repair. They slice these hopeless proteins apart, leaving harmless fragments to be recycled.

  Heat-shock proteins are lifesavers, but E. coli can’t keep a supply of them on hand for emergencies. They are among the biggest proteins in its repertoire, and to survive a blast of heat E. coli may need tens of thousands of them. Making heat-shock proteins in ordinary times would be like paying the local fire company to park all its trucks in your driveway just in case your house catches fire. On the other hand, when you need a fire truck, you need it fast. If E. coli takes too long to manufacture heat-shock proteins, it can die while it waits to be rescued.

  This tricky trade-off attracted the attention of John Doyle, an engineer at the California Institute of Technology, and his colleagues. In past years, Doyle had developed a theory for designing control systems for airplanes and space shuttles. In E. coli he recognized a piece of natural engineering just as impressive as anything he had helped to build. He and his colleagues began to analyze its heat-shock proteins and the way E. coli uses them to survive.

  They found that E. coli controls its supply of heat-shock proteins with feedback. For engineers, feedback is what happens when they allow the output of a circuit to become an input. A thermostat uses a simple form of feedback to keep the temperature of a house stable. The thermostat senses the temperature in the house and turns on the heater if it’s too cold. If the temperature gets too high, it shuts the heater down.

  E. coli’s defense against heat works a lot like a thermostat as well. The key protein in its thermostat is called sigma 32. Even when the temperature is cool, E. coli is constantly reading the gene for sigma 32 and making RNA copies. But at normal temperatures the RNA folds in on itself, and so E. coli cannot use it to make a protein. At normal temperatures the microbe is loaded with sigma 32 RNA but no actual sigma 32 protein.

  Only when E. coli heats up can the sigma 32 RNA uncrumple. Now the ribosomes can read it and make huge amounts of sigma 32 protein. Each sigma 32 protein quickly finds some of E. coli’s gene-reading enzymes and leads them to the genes for heat-shock proteins. E. coli thus makes tens of thousands of heat-shock proteins in a matter of minutes.

  Left unchecked, however, a sudden rush of sigma 32 would be too much of a good thing. The microbe would churn out heat-shock proteins far beyond its needs. In fact, E. coli makes just the right number of heat-shock proteins to cope with a particular temperature. It makes more proteins for higher temperatures, fewer for cooler ones. It exerts this fine control with a series of feedback loops.

  E. coli’s heat-shock proteins don’t just protect against heat. They also control the thermometer protein itself, sigma 32. Some of them grab sigma 32 and tuck it away in a pocket. Others cut it to pieces. In the first few moments of dangerous heat, heat-shock proteins are too busy helping unfolded proteins to attack sigma 32. But once they get the crisis under control, more and more heat-shock proteins become free to grab sigma 32. As the level of sigma 32 drops, E. coli makes fewer new heat-shock proteins.

  This feedback helps keep E. coli from exploding with heat-shock proteins. It also controls the level of heat-shock proteins. If E. coli is merely warm rather than scorching, the heat-shock proteins quickly reduce the level of sigma 32. But as the temperature increases, they have to cope with more unfolded proteins, and thus they allow sigma 32 to remain high so that E. coli will produce more heat-shock proteins. And once E. coli cools down to a comfortable temperature, its thermostat shuts down the heat-shock proteins almost completely.

  E. coli’s robust self-control comes from the feedback loops built into its network. To engineers this principle is second nature. The autopilot in a Boeing 777 uses the same kinds of feedback to keep the plane level as it is buffeted by wind shears and downdrafts. In neither case does robustness come from some all-knowing consciousness. It emerges from the network itself.

  THE BIG PICTURE

  Put genes together into circuits and they can do much more than they could on their own. Put circuits together and you create a living thing.

  In the 1940s, Edward Tatum and other scientists got the first hints of what certain genes in E. coli were for. As of 2007, researchers had a pretty good idea of what about 85 percent of its genes do, making E. coli the gold standard of genetic familiarity. Scientists have created online databases for E. coli’s genes, its operons, its metabolic pathways. Mysteries remain—there are forty-one enzymes drifting around inside E. coli for which scientists have yet to find genes, for example—but a rough portrait of E. coli’s entire system is emerging, the closest thing biologists have to a complete solution to any living organism.

  Bernhard Palsson, a biologist at the University of California, San Diego, has overseen the construction of a model of E. coli’s metabolism. As of 2007, he and his colleagues had programmed a computer with information on 1,260 genes and 2,077 reactions. The computer can use this information to calculate how much carbon flows through E. coli’s pathways, depending on the sort of food it eats. Palsson’s model does a good job of predicting how quickly E. coli will grow on a diet of glucose and how much carbon dioxide it will release. If Palsson switches off the oxygen, the model shunts carbon into an o
xygen-free metabolic pathway, just as E. coli does. If Palsson leaves out a particular protein, the model metabolism rearranges itself just as the metabolism of a real mutant E. coli would. It predicts E. coli’s behavior in thousands of conditions. The model and E. coli alike make the best of whatever situation they face, adjusting their metabolism in order to grow as fast as they can.

  How does E. coli’s metabolism manage to stay so supple when it is made up of hundreds of chemical reactions? With thousands of possible pathways it could choose from, why does it choose among the best few? Why doesn’t the whole system simply crash? Part of the solution lies in the shape of the network itself, the very layout of its labyrinth.

  When scientists map the pathways that a carbon atom can take through E. coli’s metabolism, the picture they see looks like a bow tie. On one side of the bow tie are the chemical reactions that draw in food and break it down. These reactions follow each other along simple pathways, a fan of incoming arrows. Eventually the arrows all converge on the bow tie’s knot. There the pathways get much more complicated. The product of a reaction may get pulled into many different reactions, depending on the conditions at that moment. It is there, in the knot, that E. coli creates the building blocks for all its molecules. The building blocks enter the other side of the bow tie—an outgoing fan of pathways. Each pathway produces a very different sort of molecule—this one a membrane molecule, that one a piece of RNA, another one a protein. The pathways on the far side of the bow tie fan out without crossing over. A molecule on its way to becoming a protein does not become a piece of DNA.