January, students at M.I.T. are let off the leash to follow their fancies. The annual monthlong Independent Activities Period is a playground for the mind, offering courses, seminars, and special events devoted to everything from energy-dispersive x-ray spectroscopy to poetry reading. There’s glassblowing, building spacecraft for mice, and the all-important coolest-stuff-made-of-duct-tape competition. “I wish I didn’t teach an IAP,” says Drew Endy, an assistant professor in biological engineering. “I’d take a whole bunch of the courses.”
But Endy does teach an IAP. This year his class is devoted to building counters – devices that count from, say, 1 to 32. That may not sound like much of a challenge for students at the world’s most prestigious engineering school; in fact, it’s the sort of thing a nerdy middle school kid would solder together. But here’s the rub: The counters his students design won’t be electronic, but biological. They won’t be made of transistors, but DNA. And they won’t be inserted into breadboards, but living bacteria.
While Endy is keen on counters at the moment (they might have practical uses; for example, indicating how many times a given cell has divided since the counter was last reset), they’re just stepping-stones to a new era in biology. Last year, his students programmed bacteria to form polka-dotted colonies. The year before, they designed microorganisms that blinked like Christmas lights. But the real purpose of the course isn’t making a particular biological circuit; it’s figuring out what it takes to make any biological circuit.
Endy is the newest recruit to a cabal of MIT engineers gathered around one of the university’s computer science gurus, Tom Knight. Their aim is to create a field of engineering that will do for biological molecules what electronics has done for electrons. They call it synthetic biology.
“I think this will likely be the most important thing I’ve done,” says Knight, whose track record already includes designing some of the earliest network interfaces, bitmapped displays, and workstations. “We’re at the cusp of some dramatic changes.”
If the notion of hacking DNA sounds like genetic engineering, think again. Genetic engineering generally involves moving a preexisting gene from one organism to another, an activity Endy calls DNA bashing. For all its impressive and profitable results, DNA bashing is hardly creative. Proper engineering, by contrast, means designing what you want to make, analyzing the design to be sure it will work, and then building it from the ground up. And that’s what synthetic biology is about: specifying every bit of DNA that goes into an organism to determine its form and function in a controlled, predictable way, like etching a microprocessor or building a bridge. The goal, as Endy puts it, is nothing less than to “reimplement life in a manner of our choosing.”
And what might the practitioners of this emerging science do with such godlike capability? Within a decade, some hope to create bacteria able to mass-produce drugs that currently have to be painstakingly harvested from rare plants. Others talk about making viruses encased in protein sheaths that can be used to produce fabric with molecular circuitry woven into its warp and weft. In the more distant future, synthetic biologists envision building more complex organisms, like supercoral that sucks carbon out of the biosphere and puts it into building materials, or an acorn programmed to grow into an oak tree – complete with a nifty tree house. And there’s the opportunity to add new chromosomes to the human genome, ushering in a panoply of human augmentations and enhancements.
Synthetic biology has a long way to go before such wonders become possible, but each year’s IAP course brings them closer as the MIT team learns by trial and error. As the course enters its third year, Endy and his students are closing in on an approach that will let them design systems considerably more elaborate than the simple projects they’ve attempted so far. And if getting there doesn’t go smoothly, that’s OK. “Engineers work best by flailing about,” Endy says, “and we’ve been doing as much of that as anybody.”
Drew Endy has a rapid-fire delivery and a high-intensity gaze. As an undergraduate at Pennsylvania’s Lehigh University in the early ’90s, he studied civil engineering. “I like to build stuff,” he says. “I’m a kid in that regard.” But he was also fascinated by biology, which led him to environmental engineering and molecular biology.
For his PhD project at Dartmouth, Endy developed a computer model of T7, a virus that infects the bacterium E. coli. His model’s description of what happens as T7 attacks its prey – for example, which genes are turned on, and when – “wasn’t complete bullshit,” he says. In the uncertain business of simulating biological systems, that counts as success.
But the real test of a model is how well it predicts the outcome of circumstances never seen before. So Endy rearranged T7’s DNA to make mutant strains in which the virus synthesized its proteins in a different order. Then he used his model to predict what the new guys would do when presented with some E. coli to infect. The results weren’t good. “For all the interesting predictions, whenever I went into the lab the opposite thing would happen,” he recalls. “It was really disappointing.”
In the late ’90s, Endy joined the Molecular Sciences Institute, an independent research outfit in Berkeley, California. There he realized there were two different ways forward. He could “go back and understand a whole bunch more about the science of the organism in order to model it better,” which, Endy concedes, is “a fine and valid traditional path,” not to mention the one MSI was devoted to. Or he could take a more radical approach: tear apart nature’s work and reconstitute it in a more logical, malleable form. “I thought, Screw it,” he says. “Let’s build new biological systems – systems that are easier to understand because we made them that way.'”
At any other time, Endy’s idea would have been deeply impractical. Custom-building biological systems meant writing DNA sequences from scratch, and the ability to write sequences trailed far behind the ability to read them. In the 1990s, though, the rate at which DNA could be read revved into high gear, and in 2000 it was becoming clear that DNA synthesis – stringing together pairs of nucleotide bases, the letters of the genetic code – would follow.
One of Endy’s friends at MSI, Rob Carlson, charted the rates at which various biotechnologies were improving. The DNA-reading machines used by the Human Genome Project were doubling in efficiency every 18 months. DNA synthesis was accelerating even more quickly. If reality kept up with these “Carlson curves,” then by 2010 a single lab worker would be able to synthesize a couple of human genomes from scratch every day. No more need for DNA bashing – just write out the sequence you want and synthesize it straightaway.
The Carlson curves also showed that the price of DNA synthesis was falling rapidly. That trend has only continued. In 2000, the cost of assembling sequences to order was roughly $10 to $12 per base pair. Today, it’s down to $2. Some scientists foresee DNA synthesis dropping to 1 cent per base pair within a couple of years. That’s a gene for 10 bucks, a bacterial genome for the price of a car.
Working next to Carlson at MSI, Endy could see the cheap future of DNA synthesis. The era in which you could learn about biology only by studying natural systems was quickly drawing to a close. The era in which the full discipline of engineering could be brought to bear on biological systems was set to begin.
To bring his vision to life – in the most literal sense – Endy knew he needed to go somewhere rich in engineering lore and wisdom, and where his ideas wouldn’t be seen as too weird. The second requirement rather narrowed the field. “When I started broadcasting my idea – ‘I think we should rebuild the natural biological systems we most care about and domesticate their genomes’ – the only place that returned a coherent signal was Tom Knight’s lab at MIT.”
Tom Knight is an MIT institution. As a high school student in the ’60s, he got a summer job working for Marvin Minsky in the artificial intelligence lab and has been thereabouts ever since. A solid man with a purposeful set to his bearded jaw, he exudes a carpenter-like trustworthiness, along with just the right amounts of guru mystique and pocket-protector practicality.
In the early ’90s, Knight was looking for a new challenge. He came across the work of Harold Morowitz, a physicist turned biologist at Yale who specialized in stripped-down bacteria called mollicutes, creatures so small they contain only a billion atoms. In some contexts, a billion is a lot; in others, not so much. To a computer guy, it’s less than the number of transistors you can cram on a chip. “Morowitz’s work laid out, in words I could understand as an engineer, an agenda that seemed so exciting I had to go do it,” Knight says. “Here’s a class of organisms so simple that maybe we can understand everything there is to know about them.”
Knight attacked the project with unbridled energy. From 1993 to 1995, he went back to school, this time to study biology, while continuing to teach as a professor of electrical engineering. Knight and Gerry Sussman, another veteran of the AI lab, started talking to Darpa about biological information-processing systems, networks of genes and proteins turning each other on and off. The Pentagon’s thinking-outside-the-box boys listened.
Knight set about building his own biology lab in the computer science department. “They weren’t used to having biohazard signs around and worrying about chemical disposal,” Knight recalls, “but a lot of people were supportive.” Once the lab was running, he found himself swamped by “the remarkable lack of standards in the biological world.” So naturally he started bringing an engineering-friendly order to it, looking for ways that DNA sequences could be designed to fit together in useful, practical ways. A lifelong Lego fan, he called these sequences biobricks and released the first batch in 2001.
Endy’s arrival at the beginning of 2002 kicked the effort to a new level of ambition and visibility. “He energized the campus in a way I wouldn’t have been able to,” Knight says. With new blood and fresh biobricks, it was time to test the limits. Knight and Sussman suggested setting up a course, bringing in some students, and making stuff. This may seem counterintuitive: How can you teach a subject that hasn’t been invented, and how does teaching it help get it invented, anyway? But the graybeards of synthetic biology were confident that a course could catalyze a revolution. After all, they’d seen it done before, when Lynn Conway taught her pioneering class in the emerging technology of very large-scale integration.
First taught in 1978, Conway’s course is the stuff of geek legend. In the ’70s, large semiconductor companies kept a tight grip on the technology for making integrated circuits and used it for only a limited range of products, mostly industrial and military. Conway, a computer architect from Xerox PARC, and her colleague, Caltech professor Carver Mead, developed a new, more flexible chipmaking technique that decoupled design from fabrication: VLSI. The method allowed engineers to conceive circuitry without having to deal with the details of manufacturing, and Conway brought it to the brightest minds in academia.
What dazzled Conway’s students was that they didn’t just learn about VLSI; they got to use it. They designed circuits in a format that could be transmitted over the newfangled Arpanet to Mead’s prototype chip foundry in California. And thanks to a grant from Darpa, they got their work back one month later in the form of honest-to-goodness chips. That class, and its successors, revolutionized electronics. From PCs and mobile phones to game consoles and PDAs, much of the next two decades of cool computing sprang from Conway’s course.
Hoping to spark a similar transformation in molecular biology, Endy and Knight modeled their Independent Activities Period course on the Conway approach. Students would design DNA circuitry – specifications for genetic sequences, some of them genes describing proteins, some of them sites for proteins to bind to and thus switch genes on and off. The specs would be sent via the Internet to a Seattle-based gene-synthesis shop called Blue Heron, one of the outfits pushing the price of synthesis down Carlson’s curves. The resulting DNA would be shipped back to MIT and slipped into E. coli specimens. Darpa ended up covering the cost of the synthesis, echoing the agency’s role in the VLSI course more than 20 years earlier.
For the first year’s project, Endy drew on work by Michael Elowitz, now an assistant professor at Caltech. In 2000, Elowitz had described a small circuit he’d engineered into E. coli that consisted of three so-called repressor genes rhythmically turning one another on and off. Tie a fluorescent protein to one of those genes and you get bacteria that flash like fireflies. Elowitz called his system a repressilator – an oscillator made of repressors. The IAP students studied it, admired its simplicity, and convinced themselves that, as MIT’s crack studbunnies of geekdom, they’d be able to do something cooler yet. They designed several circuits that were more sophisticated and, in principle, more stable. Then they sent their designs to Blue Heron to be synthesized.
At that point, things started to go wrong. Blue Heron was unable to reproduce half of the student sequences. When the company injected the synthetic DNA into the cells that were supposed to replicate it, the cells refused to cooperate. After the course was over, it took Endy six months to find an effective way to install the artificial sequences so the cells would reproduce them.
Then, when he received the sequences from Blue Heron and inserted them into the bacteria in his lab, they didn’t work. Or so it appeared. He has since realized that many of them probably functioned as intended; he wasn’t able to see flickering lights because the display wasn’t strong enough. “These systems run inside cells that are only 10,000 atoms long,” he points out. “We’ve only just been able to see oscillations in cells for the first time, a year after we started trying. I suspect that many of those systems are working.”
Part of the reason it has been so hard to figure out whether, and how well, the circuits work is compatibility. The initial generation of biobricks fit together physically – the Legos plugged into one another – but they didn’t necessarily fit together functionally. Their parts had been scavenged from various biological systems. “It would have been dumb luck if they all worked together out of the box,” Endy now realizes.
It wasn’t long before a solution arose – thanks as much to budgetary realities as to scientific insights. The 2003 IAP cadre was divided into four teams, each working on a different design. There was enough money to synthesize 20,000 base pairs, giving each team a budget of 5,000. But the designs were all longer than that. So they had to be revised so that basic functions could be carried out by identical parts – building blocks that would be useful in a wide variety of applications. The growing parts list, in turn, forced Endy and Knight to think through exactly how biobricks could be made to work together as well as fit together. They knew they needed a consistent interface between components, like current in electronics. After much head-scratching, they settled on the rate at which RNA polymerase, the molecule that transcribes DNA, travels along the sequence that makes up the component – a measure they call PoPS (polymerase per second).
The result is a library of standardized parts, much like the transistors, capacitors, and resistors of electronics, with a standard kind of signal to run through them. That is, synthetic biology is approaching the sophistication of a child’s first electronics kit.
This infrastructure is still a work in progress. The parts list continues to grow, and some basic standards, such as the optimal signal level – that is, how many PoPS events are required for a serviceable connection – remain to be worked out. But the nature of standardization means that once a few components work together, it should be easy to get more and more to join in. Endy’s colleague Randy Rettberg, a former CTO at Sun Microsystems and now a research affiliate at MIT’s Computer Science and Artificial Intelligence Lab, says simply, “We’re ahead of ourselves. That’s a good place to be.”
When the great physicist Richard Feynman closed the door of his Caltech office for the last time in 1988, he left a striking epigram scrawled on the blackboard: “What I cannot create I do not understand.” That, in a nutshell, is the scientific case for synthetic biology.
To many scientists, the field’s real appeal is that it provides a new way to unlock the mysteries of biology. Trying to do the things that nature does – say, orchestrating the interactions of genes and proteins triggered by some external event – is a way to discover fundamental principles that govern living systems. At Caltech, Elowitz is using his repressilator to learn about the different sorts of noise that interfere with a cell’s ability to send and receive signals internally, and how signal-to-noise issues might affect a cell’s behavior.
But creation is not just a route to understanding; it’s also way to get things done. At the first synthetic biology conference, held last summer at MIT, the most compelling presentation was given by Jay Keasling, a professor of chemical and biological engineering at UC Berkeley. Keasling is modifying bacteria to help make the malaria drug artemisinin. This treatment is far more effective than its current competitors but also far more expensive to produce. Keasling’s approach would be impossibly complicated without synthetic techniques; he’s assembling 10 genes from three different organisms and forging a new metabolic pathway. He has already improved the artificial pathway’s productivity a millionfold. If he can make it a hundred times better still, he’ll churn out artemisinin cheaply enough to save a great many more lives.
Once synthetic biology becomes sufficiently advanced, Knight thinks the big application will be construction. Biological systems are great at producing large-scale structures from small beginnings: Theoretically a seed could be programmed to grow into a house. But for now, what he’s most excited about is not a biological product at all – it’s a reprise of Lynn Conway’s success with the VLSI course in 1978: a generation of students freed from past limits, ready to do things never imagined before.
There are dangers in such power. The first dramatic demonstration of DNA synthesis was the creation of a polio virus, assembled in 2002 by researchers at SUNY Stony Brook. The event itself wasn’t especially worrisome; if a terrorist wanted to obtain polio, a few easily purchased doses of live vaccine would be a much easier starting point. It was, however, a vivid illustration of challenges to come. Blue Heron and its competitors check every sequence they’re asked to make against the genomes of all known pathogens. But what if people find ways to create nasty life-forms unlike anything in nature, whose capacity for hurt looks like nothing seen before?
best defense against rogues and evildoers, Endy says, is a community of technologists committed to finding ways to thwart them, bigger and better-resourced than the Dark Side could ever be. He believes in an open source approach to the task of programming life, modeled explicitly on the open source approach to programming computers, long popular at MIT. This ethos might also defend against implementations that are simply inept, rather than malicious, but nonetheless capable of great harm. As Endy points out, do you really want staple crops fitted with proprietary programming as secure as, say, Windows 95?
How society will respond to these new technologies is as hard to predict as the behavior of a mutant T7. That we will have to respond, though, is beyond question. MIT’s synthetic-biology project could fail to develop the genetic equivalent of resistors and capacitors, transceivers and power supplies; funding agencies may not see the possibilities; industry may decide it’s all too blue-sky. But the ability to design genomes as freely as circuits in silicon will not go away. And that’s a fundamental change. It means that rather than just listening to what nature chooses to tell us, we can now ask questions in nature’s native tongue and compel it to reply. If we don’t phrase the questions correctly, the answers may not mean much. But in that case, we can simply try again. Questions are cheap. Ten bucks a gene in a year. A grand for a genome soon after.
Life will never be the same. Oliver Morton, Wired Magazine