by Carl Zimmer
Terrifying failures like this one leave scientists hoping that someday they will find new antibiotics that are immune to the evolution of resistance. Like Fleming before them, they find promising new candidates in unexpected places. One particularly promising group of molecules was discovered in 1987 in the skin of a frog.
Michael Zasloff, then a research scientist at the National Institutes of Health, noticed that the frogs he was studying were remarkably resistant to infection. At the time, Zasloff was using frogs’ eggs to study how cells use genes to make proteins. He would cut open African clawed frogs, remove their eggs, stitch them back up, and put them in a tank. Sometimes the water in the tank became murky and putrid, yet his frogs—even with their fresh wounds—did not become infected.
Zasloff suspected the frogs were making some kind of antibiotic. He ground up frog skin for months until he isolated a strange bacteria-killing molecule. It was a short chain of amino acids known as a peptide. He and other researchers discovered that it is fundamentally different from all previously discovered antibiotics. It has a negative charge, which attracts it to the positively charged membranes of bacteria but not to the cells of eukaryotes such as humans. Once the peptide makes contact with the bacteria, it punches a hole in their membranes, allowing their innards to burst out.
Zasloff realized he had stumbled across a huge natural pharmacy. Antimicrobial peptides, it turned out, are made by animals ranging from insects to sharks to humans, and each species may make many kinds. We produce antimicrobial peptides on our skin and in the lining of our guts and lungs. If we lose the ability to make them, we become dangerously vulnerable. Cystic fibrosis may be due in part to mutations that disable genes for antimicrobial peptides produced in the lungs. The lungs become loaded with bacteria and swell with fluid.
Having discovered antimicrobial peptides, Zasloff now tried to turn them into drugs. They might be able to wipe out bacteria that had evolved resistance to conventional antibiotics. Antimicrobial peptides might even be resistance proof. In order to become resistant to antimicrobial peptides, bacteria would have to change the way they build their membranes. It was hard to imagine how bacteria could make so fundamental a change in their biology, and experiments seemed to back up this hunch. Some scientists randomly mutated E. coli to see whether it could produce mutants able to survive a dose of antimicrobial peptides. No luck.
But an evolutionary biologist named Graham Bell at McGill University in Montreal suspected that E. coli—and its evolutionary potential—might be more powerful than others had thought. Michael Zasloff, for one, didn’t think so. But as a good scientist, he was willing to put his hypothesis to the test. He teamed up with Bell and Bell’s student Gabriel Perron to run an experiment. Remarkably, his hypothesis failed.
The researchers began by exposing E. coli to very low levels of an antimicrobial peptide. A few microbes survived, which the scientists used to start a new colony. They then exposed the descendants of the survivors to a slightly higher concentration of the antimicrobial peptide. Again most of the bacteria died, and they repeated the cycle, raising the concentration of the drug even higher. E. coli turned out to have a remarkable capacity to evolve. After only six hundred generations, thirty out of thirty-two colonies had done the impossible: they had become resistant to a full dose of antimicrobial peptides. These results raise some serious concerns about how effective antimicrobial peptides will be when they hit the market. E. coli and other bacteria that are hit by low doses of antimicrobial peptides may evolve resistance. If they do, they will survive stronger and stronger doses until they can withstand the full strength of these drugs.
If E. coli can evolve resistance to antimicrobial peptides so quickly, then how did they protect Zasloff’s dirty frogs? E. coli and other bacteria are locked in an evolutionary race with the animals they colonize. When an animal evolves a new antimicrobial peptide, natural selection favors microbes that can resist it. One common counterstrategy is for a microbe to make an enzyme that can cut the new peptide into pieces before it is able to do any damage.
Now the evolutionary pressure shifts back to the animal. Mutations that allow an animal to block the peptide-cutting enzyme may allow it to survive infections. It will pass down the mutation to its descendants. Animals defend against peptide-slicing enzymes by stiffening the peptides. The peptides are folded over on themselves and linked together with extra bonds. But microbes have evolved counterstrategies of their own. For example, some species secrete proteins that grab the antimicrobial peptides and prevent them from entering the bacteria.
One of the most potent ways for animals to overcome all of these strategies is by making lots of different kinds of antimicrobial peptides. New ones can be produced by gene duplication or by borrowing peptides with other functions. The more antimicrobial peptides an animal makes, the harder it is for bacteria to recognize them all. Thanks to this arms race, the genes for antimicrobial peptides have undergone more evolutionary change than any other group of genes found in all mammals.
Compared with this complex, ever-changing attack on antimicrobial peptides, Bell and Zasloff’s experiment is child’s play. They exposed E. coli to a single kind of antimicrobial peptide and created a strong advantage for mutants that could withstand it. Unfortunately, modern medicine works more like Bell and Zasloff’s experiment than like our own evolution. Doctors have only a few antibiotics to choose from when fighting an infection, and they generally prescribe only a single drug to a patient. In a few years this practice gives us resistant bacteria. We might do a better job of fighting bacteria if new drugs came through the development pipeline faster and if doctors could safely prescribe several of them at once.
There are many lessons to be learned from E. coli’s quick evolution of resistance. The most surprising of all is that our own bodies, and those of our ancestors, are actually drug-development laboratories.
EVOLUTION ON DEMAND
When Salvador Luria discovered the jackpot pattern in E. coli’s resistance to viruses in 1942, he provided some of the first compelling evidence that mutations strike randomly and blindly. A vast number of other experiments on E. coli and many other species confirm the steady rate of mutations. But there are a few experiments on E. coli that raise some intriguing doubts. Perhaps some mutations are not so blind after all.
Floyd Romesberg, a chemist at Scripps Research Institute in La Jolla, California, carried out an experiment to watch E. coli evolve resistance to antibiotics. The drug he chose was ciprofloxacin, or cipro for short. Cipro first emerged in the early 1980s as a promising replacement for older antibiotics that had begun to fail. But within a few years, scattered reports of resistance began to appear. Cipro-resistant bacteria are now very common in some parts of the world. In Germany, 15 percent of E. coli were resistant in 2002. In China that same year, one study put the figure at 59 percent.
To understand how cipro-resistant genes evolved, Romesberg and his colleagues injected a disease-causing strain of E. coli into six-week-old mice. They then treated the mice with cipro, and the infection disappeared. Or at least it seemed to. Three days later the mice were sick with E. coli again. When the scientists tested the bacteria, they discovered that the E. coli had become fifty times more resistant to cipro since the start of the experiment.
Cipro kills E. coli by tricking it into committing suicide. It interferes with an enzyme known as a topoisomerase, which normally helps to untangle DNA by snipping it and then joining it back together. Once the topoisomerase has cut the DNA, cipro prevents it from finishing its job. The free ends attract other enzymes whose job it is to chop up loose pieces of DNA. They end up destroying much of E. coli’s chromosome and thus killing the microbe.
It occurred to Romesberg that cipro might cause E. coli to do something else as well: mutate faster. E. coli repairs damaged DNA with enzymes called polymerases. It makes two kinds of polymerases: one that does high-fidelity repair and one that does low-fidelity work. The hi-fi polymerases usually handle all the
repair work while the genes for lo-fi polymerase are switched off by a protein called LexA. But things change when E. coli is in a crisis. When E. coli becomes burdened with a lot of damaged DNA, LexA falls off the lo-fi polymerase genes. Now the lo-fi polymerases help repair E. coli’s DNA. And because they do a less accurate job, they leave behind more mutations.
Romesberg wondered if these extra mutations helped E. coli evolve resistance to cipro faster. While most of the mutations might harm the bacteria, a few might produce topoisomerases that could keep doing their cut-and-paste job even in the presence of cipro. It was possible that extra mutations would arise only during these sorts of crises. Once E. coli could cut and paste its DNA again, its supply of loose DNA would dwindle. LexA would grab on to the genes for lo-fi polymerases and shut them down. E. coli would return to its more careful DNA repair.
Romesberg and his colleagues tested their hypothesis with an elegant experiment. They engineered a strain of E. coli in which LexA did not fall off the lo-fi polymerase genes. Exposed to cipro, these microbes would go on repairing their DNA with exquisite accuracy. Romesberg and his colleagues injected their engineered strain into mice and gave them cipro. In 2005, they reported their results: unable to mutate more, the E. coli evolved no resistance to cipro at all.
Romesberg’s experiment suggests that E. coli is not just passively accumulating mutations. E. coli may have evolved ways to manipulate mutations to its own advantage.
The first inklings of not-so-blind mutations came in a “water, water, everywhere, but not a drop to drink” experiment in 1988. John Cairns, then at Harvard, and his colleagues engineered a strain of mutant E. coli that was almost completely unable to feed on lactose. Its lac operon was in good working order, but the promoter sequence where it could be switched on was slightly mutated. Cairns and his colleagues then gave the bacteria nothing but lactose to eat. The bacteria stopped growing and began to starve. But they did not die out completely.
Over the course of six days, a hundred colonies emerged. Cairns examined their lac operon and found that a mutation had struck the microbes, allowing them to switch on the operon again. Cairns calculated that if the bacteria had been mutating spontaneously at their normal rate, only a single colony would have formed in that time. Instead, Cairns concluded, these microbes had acquired working genes a hundred times faster than they should have.
“Cells may have mechanisms for choosing which mutations will occur,” Cairns and his co-authors wrote.
These “directed mutations,” as they came to be known, caused an uproar. The idea that E. coli could respond to a crisis by mutating a specific piece of DNA smacked of Lamarck. Critics claimed that Cairns’s hypothesis was practically mystical, requiring E. coli to know that mutating a particular part of its DNA would help it in a particular crisis. A wave of other studies followed as scientists tried to figure out just what was happening.
A consensus emerged that these mysterious mutations were not precisely directed toward any particular goal. Many of the bacteria that regained the ability to feed on lactose also carried new mutations on genes that had nothing to do with lactose. Instead of directed mutations, scientists began to speak of “hypermutation.” And by hyper, they meant that during a crisis E. coli’s mutation rates could soar a hundred-or even a thousandfold. Several studies identified E. coli’s lo-fi polymerases as the enzymes that created these extra mutations.
Some scientists argue that hypermutation is an elegant strategy to ward off extinction. Normally, natural selection favors low mutation rates, since most mutations are harmful. But in times of stress, extra mutations may raise the odds that organisms will hit on a way out of their crisis. To avoid starving, E. coli does not need to know that a small mutation to the switch controlling its lactose-digestion genes will hit the jackpot. It just has to change enough DNA until it changes the right one.
Hypermutation has an obvious risk: along with a beneficial mutation, it can also cause many harmful ones. Susan Rosenberg of Baylor College of Medicine and her colleagues argue that E. coli minimizes this risk by spreading it across an entire colony. When E. coli produces extra mutations under stress, an individual microbe experiences them only in one narrow region of its DNA. From one microbe to the next, that window of mutation is in a different spot. As a result, the bacteria are not crippled by mutations all across their genome. At the same time, though, new versions of almost every gene in the E. coli genome can emerge in a colony. When a few microbes hit on the winning solution, they can start growing quickly.
Hypermutations may be a useful way for E. coli to cope with stress, but they may have evolved for very different reasons. Olivier Tenaillon of France’s National Institute of Health and Medical Research points out that it takes a lot of energy and material to build hi-fi polymerases. In times of stress, E. coli may not be able to afford the luxury of accurate DNA repair. Instead, it turns to the cheaper lo-fi polymerases. While they may do a sloppier job, E. coli comes out ahead on balance. Natural selection, Tenaillon proposes, didn’t favor higher mutation rates—just cheaper repairs.
Even if the changing mutation rate in bacteria arose as a side effect, it may still be useful. Tenaillon and his colleagues have demonstrated that E. coli varies enormously in its mutation rate. Under stress, one microbe may mutate a thousand times faster than another. Hypermutation genes must be responsible for the difference, and they can be passed down from one generation to the next.
In different situations, natural selection may favor some mutation rates over others. Tenaillon and his colleagues have observed the average mutation rate in E. coli as it colonizes a mouse. Early on during the colonization, when the bacteria are experiencing a lot of stress, high-mutation microbes are more common. When the bacteria have established stable colonies in the guts of the mice, low-mutating microbes take over. Antibiotics may also drive the rise of high mutators because they can evolve resistance faster than bacteria that mutate more slowly.
Some critics are skeptical of directed mutation, hypermutation, and their intellectual offspring. John Roth of the University of California, Davis, and Dan Andersson of Uppsala University in Sweden argue that Cairns did not discover anything out of the ordinary in his original experiments. The lac operons in the bacteria he used were not entirely shut down, Roth and Andersson claim. They could still produce a few proteins, allowing the bacteria to avoid starvation. An ordinary, random mutation might have copied the lac operon in a microbe, allowing it to digest more lactose and grow faster. Its descendants might accidentally have made a third copy of the genes, and natural selection might have favored that mutation as well.
Through nothing more than spontaneous mutations and natural selection, Roth and Andersson argue, E. coli can expand its collection of lactose-digestion genes. And as the number of copies grows, it becomes more likely that an ordinary mutation will restore one of the operons to good working order. Any microbes that gain a working operon will suddenly multiply far faster than the other bacteria. Mutations may then remove the defective copies, leaving the microbes with a single good version. This process creates the illusion of directed mutations, Roth and Andersson argue, when nothing of the sort has taken place.
The debate, which continues to rage, matters both to the practice of medicine and to our understanding of how life works. If microbes do depend for their survival on an ability to change their mutation rates, then blocking that change could be a way to kill them. Floyd Romesberg has shown that preventing E. coli from raising their mutation rate prevents them from evolving resistance. He and his colleagues are now trying to turn that discovery into a medical treatment. They hope that someday people who take antibiotics will also be able to take a drug to stop microbes from increasing their mutations.
Some scientists suspect that animals and plants can also manipulate their mutations to cope with stress. Susan Lindquist of the Whitehead Institute for Biomedical Research in Cambridge, Massachusetts, and her colleagues discovered that fruit flies have a buffer t
o protect themselves from the harmful effects of mutations. A harmful mutation might cause a protein to fold incorrectly. But the fruit fly’s heat-shock proteins can fold it into its proper shape. Over many generations, Lindquist argues, the fruit flies can generate a lot of genetic diversity that could not exist without the help of their heat-shock proteins.
Lindquist discovered that stress unmasks these mutations. Raising the temperature, adding toxic chemicals, or otherwise abusing the flies makes even normal proteins go awry. The heat-shock proteins become so overworked that they abandon many of the mutant proteins to assume their true shapes. These proteins can have drastic effects on the flies, altering their eyes, wings, or other body parts.
Lindquist proposes that heat-shock proteins let the flies build up a supply of mutations that help them survive a crisis without having to suffer their ill effects in less stressful times. An unmasked mutation may prove helpful to the flies, and new mutations can allow it to remain unmasked even after the stress has disappeared. Lindquist and her colleagues have found a similar mutation buffer in plants and fungi, suggesting that it may be a common strategy. The process Lindquist has proposed is different in the details from hypermutation in E. coli, but the fundamental benefits seem to be the same: harnessing the creative powers of mutations while minimizing their risks.
Roth and Andersson’s gene amplification, on the other hand, may not be limited to a few lactose-starved E. coli. Making extra copies of genes may help many organisms adapt to new challenges.
