stanford medicine

 ARCHIVESnext issue

Microbe computers - Built from the stuff of life


Microbe computers

Built from the stuff of life

Drew Endy is an engineer who uses biological molecules to build microscopic computers. One of his goals: Make it easy to compute inside living cells.

Bacteroides thetaiotaomicron is one of the good guys. Known simply as Bt, it is an abundant member of the estimated 500 to 1,000 species of bacteria that live in the human intestine and one of the lead players in a remarkable symbiosis between man and microbe. Bt does what our bodies can’t. It ingests complex sugars and breaks them down into usable nutrients that can then be absorbed by our digestive system. We may be stardust, but we are also microbe excrement.

Inflammatory bowel disease, in contrast, is about as unpleasant as the name implies. The complex condition has baffled science for decades. Among its peculiarities is the fact that the bacteria in the guts of sufferers vary significantly in type and number from those of healthy humans, Bt chief among them.

“Unfortunately, we don’t know whether these variations in the bacterial populations are a root cause or a resulting symptom of inflammatory bowel disease, and, until recently, we were very limited in the tools available to answer that question,” says Justin Sonnenburg, PhD, a Stanford assistant professor of microbiology and immunology who studies relationships between gut microbes and health. He’s in the vanguard of scientists applying a new research tool that provides unprecedented control of microbes.

That tool is a computer, but not a computer made of silicon, metal and plastic. It’s a computer made of DNA, RNA and enzymes, residing within the confines of a single cell.

The build

This biological microcomputer sprang from the mind of Drew Endy, PhD, an assistant professor of bioengineering at Stanford. In three scientific papers released over a 13-month span in 2012 and 2013, Endy and a team of researchers from his lab showed how they used ordinary genetic engineering techniques to turn the bacterium E. coli — that stalwart of the Petri dish — into a machine capable of the basic functions of a computer: logic, data storage and data transmission. They also showed that their techniques will work in any type of living cell, not just bacteria.

And while others have accomplished similar feats, Endy’s system has the singular advantage of being able to amplify the information flow.

“Amplification is what makes this system the best,” says Endy. “It’s the equivalent of the transistor in an electronic device. It’s what makes our computer really useful.”

So it’s a cool bit of engineering, but it’s more than that.

The work “… clearly demonstrates the power of synthetic biology and could revolutionize how we compute in the future,” said fellow biological engineer Jay Keasling, PhD, at the University of California-Berkeley, quoted in the San Jose Mercury News.

Other researchers echoed this opinion, as did the Journal of Biological Engineering, which recognized one of the three articles — “Engineered cell-cell communication via DNA messaging” — as its Publication of the Year.

Speed-wise at least, IBM won’t feel threatened by the biocomputer. “The microbial processor operates in the millihertz time frame — about one cycle every 1,000 seconds, or about four times per hour,” Endy says, “But in biology it doesn’t always matter; slow can be beautiful.”

The biological computer opens up a host of research avenues never before imagined, much less pursued. Microbes could be engineered to detect cancer, for instance, and then tag malignant cells with fluorescent markers for easy identification. Other cells might be programmed to detect those markers and deliver with pinpoint accuracy pharmaceuticals they themselves manufacture on the spot, much as Bt produces and excretes nutrients. Biological computers might even someday be able to reprogram cancer cells to shut off their own growth.

Unfortunately, at first biologists often have a hard time understanding his work, Endy says. “When I talk about this to groups of biologists, the initial response is usually, ‘Harrumph. That’s not how biology does it,’” he says. “They’re not used to thinking like engineers.” Once he explains that he’s using biology to build something simpler and easy to control, something useful, they begin to warm up to the idea.

Endy knows of several scientists starting projects using his system and he hopes many more will take it up. An advocate of open-source technology (which, as with open-source software, makes its discoveries and technologies free to the public), he has made the instructions available free online. A video primer is also on YouTube (; it’s been viewed nearly 30,000 times.

It starts with a memory

The biological computer has been a quest of five years for Endy and the researchers in his lab. The first step was creating the data-storage component. To do that, Endy, postdoctoral researcher Jerome Bonnet, PhD, and graduate student Pakpoom Subsoontorn worked to master the precise interaction of two enzymes that, when working together, can flip a DNA sequence end for end and flip it back again. The key to biocomputing is that the microbe produces the enzymes, so it controls which direction the sequence points. With this bit of biotechnology, the researchers have created an equivalent to the 1s and 0s of binary data storage that are at the heart of most every computer today.

“If you are reading along a particular section of DNA and it reads one way, we can arbitrarily label that section a zero. If it reads the opposite way, we can call that orientation a one,” Endy explains.

‘The microbial processor operates in the millihertz time frame — about one cycle every 1,000 seconds, or about four times per hour. But in biology ... slow can be beautiful.’
Drew Endy

Such a biologic memory device is the equivalent of a lone binary digit — a “bit” in computer shorthand — a quantity that is infinitesimal in storage terms. It takes eight bits to form just a single “byte” of data. A byte, in turn, is enough to store just one typed character of information. It would take some 15,000 bytes — or 120,000 of Endy’s bits — to store this article. But, like the biocomputer’s slow clock speed, in biology such small amounts of data amount to a lot of computing power in the hands of the right people.

Endy’s team then went a step beyond, formulating a clever way to retrieve the data that doesn’t require DNA sequencing. By engineering the microbe to glow different colors under ultraviolet light depending upon the direction the memory bit points, reading the data becomes as easy as shining a UV light on the microbes. If the section of DNA points one way, the microbe glows red. If the section points the other, it glows green.

To grasp how this might work, imagine a microbe that is programmed to detect the telltale chemical signature of cancer in the intestine. To start, one would swallow a million or so replicas of this specially programmed microbe. Once in the intestine, any of the microbes that encountered the signature of cancer would kick into gear and produce the enzymes necessary to flip its memory bit. Then, after the microbes exit the body (in a bowel movement) the researcher could illuminate them with UV light and know immediately whether cancer was present in the patient.

“Of the three core components of the biological computer, digital data storage was by far the hardest to create,” Bonnet says. “It took us three years and hundreds of tries to get just a single bit working right every time.”

The biological Internet

After conquering the data-storage challenge, Endy and team created a way to transmit data between cells. Their technique enables the data to literally go viral, using an innocuous virus known as M13 that makes itself at home in bacteria, living off nutrients cadged from its host. Like a freeloading houseguest texting its friends across town, M13 broadcasts its own genome to other cells. This is the infection stage of M13’s life cycle, and Endy and his team have repurposed it to fashion a biological Wi-Fi able to transmit virtually any DNA sequence between cells.

The technique requires a bit of genetic subterfuge. Normally, M13 works by sealing its own DNA within an additional brief genetic sequence, a sort of genetic package. To transmit a message, all the microbe has to do is add this packaging sequence and M13 will send it off to other nearby microbes, oblivious that the message inside the package is not its own DNA. Endy has parasitized the parasite.

“We can send any genetic messages we want and we can send them to specific cells within a complex microbial community,” says Monica Ortiz, PhD, a former graduate student working in Endy’s lab and now a postdoctoral researcher at Harvard. M13 can send genetic codes measuring in the tens of thousands of characters. While that’s a modest bandwidth compared with gigabit Ethernet, it’s plenty for biocomputing.

Researchers now have the means to control the behavior of not just a single microbe, but an entire community of cells. Cells engineered with M13-based communications might be orchestrated to start growing or stop growing, to cluster together or swim away. A group of microbes could turn on the production of insulin en masse when they detect sugar, morphine in the presence of pain, or anti-inflammatories at the site of ulceration. Contemplate for a moment the concerted effort that even a small portion of the trillion bacteria in the intestine might produce if working in a coordinated, pre-programmed fashion.

A logical conclusion

The last function Endy tackled was logic. Most computers are built to perform Boolean logic — named after George Boole, the mathematician who proposed a system of logic in the 19th century. Boolean logic in an electronic computer typically takes the form of 1s and 0s. One is true; zero is false. Answer true, gate opens, electrons flow. Answer false, gate closes, electrons don’t flow. With just these two yes-or-no states, binary computers are able to accomplish all the astounding things they do today.

“In a biological setting, the possibilities for logic are as limitless as in electronics,” says Bonnet.

In the biocomputer, the silicon gates are replaced by genetic gates that open and close to similarly control the flow, but instead of electrons the flow in this case is an enzyme that travels along a DNA strand. Answer true, enzyme flows. Answer false, enzyme doesn’t flow.

‘These tools will be useful immediately at the basic science level, helping us to better understand relationships of microbes to human health.’
Justin Sonnenburg

For example, one of the most basic logic gates is the “AND” gate, which gives an output of true when its two inputs are true — when “a” and “b” are both true. An “OR” gate, on the other hand, is true when either a or b, and possibly both, are true. Testing whether a or b, but not both, is true requires an “XOR” gate, known as an “exclusive or” gate. And so it goes, until there is a gate for every possible logical combination. Endy’s team has demonstrated biological equivalents for all the major logic gates known to electronic computing.


It is in the combination of logic and data transmission, however, where the biocomputer really begins to reveal its greater potential. The biological transistor is able to turn a tiny amount of information into a very large flow of data. In electronics, this is known as signal amplification.

With electronic signal amplification, a very small change in electrical flow is sufficient to open and close gates that control massive rivers of electrons. “The biological transistor, what we call a ‘transcriptor,’ does the same thing. A small change in gene expression can produce a very large change in cell behavior,” Endy says.

The transistor, often described as the greatest technological advance of the 20th century, was conceived precisely with amplification in mind as a way to replace unreliable vacuum tubes in relaying telephone calls across the continent. Electrical signals attenuate, weakening as they travel. By boosting the fading signal with transistors — by amplifying signals — it is possible to rebroadcast them across great distances.

Biological systems are no different. Genetic signals can now be amplified as they move through a community of cells, enabling the orchestration of large numbers of cells.

Calculating possibilities

Sonnenburg and postdoctoral scholar Weston Whitaker, PhD, were familiar with Endy’s work from the published papers, but they had been thinking about using biological parts to build machines for a long time. Whitaker learned many of the techniques of the field at Berkeley. Folding in his own ideas and adding the genetic parts from Endy’s lab, they were off.

“Conceptually, the biocomputer is fairly straightforward. Technically, however, there are many issues, as you might expect when trying to transfer genetic parts between E. coli, where Endy worked, into Bt — two organisms separated by more than 2 billion years of evolution,” says Sonnenburg. “Weston is clever, but such work often includes screening hundreds of variants of a genetic part to find just the right ones for Bt, so things are proceeding, but slowly.”

When Sonnenburg imagines the trillions of bacteria in every human colon, however, he sees only potential. “On a sheer numbers basis, there are 10 times more single-celled microbes on and in our body than all the remaining cells of our body combined,” he notes. “We’re just beginning to explore the implications of computing within these cells.”

The biological computer means that Sonnenburg could record each time one of his programmed Bt cells encounters certain environmental factors, such as inflammation in an intestine riddled with Crohn’s disease or ulcerative colitis.

“These tools will be useful immediately at the basic science level, helping us to better understand relationships of microbes to human health. In the long term, they open the possibility of altering the microbes’ community structure and function to prevent and treat diseases,” Sonnenburg says.

A research proposal drafted by Whitaker describes a three-pronged strategy to put the biocomputer to work. First, he plans to program microbes to detect inflammation. Already he has designed a genetic logic switch to detect inflammation in a mouse colon afflicted by colitis. Wiring this switch to the production of a fluorescent protein would allow programmed Bt microbes to tag afflicted areas for easy identification.

Second, using Endy’s logic and data storage tools, Whitaker and Sonnenburg plan to program microbes to gather and record information about what’s happening in the intestine that will help them understand the factors contributing to inflammation.

Third, they plan to program therapeutic bacteria and coordinate groups of them to produce and deliver immune-suppressants directly to the site of inflammation. The programmed bacteria would be research probe and drug factory in one. All three research directions were impossible before the biological computer.

More biocomputing?

Endy says the long-term goal for his work is to make biology easier to engineer — and the more people working on that goal the better. Toward that end, he created a public benefit charity, the BioBricks Foundation, and developed the BioBrick Public Agreement to make it easier for people to freely develop uses of genetic computers.

He has formally donated the transcriptor and biological logic gates to the public domain via the BioBrick Public Agreement. That means anyone is free to use them. A similar declaration for the biological Internet is in process.

The only piece of biocomputer technology Stanford and Endy have patented is the biological digital memory.

“Some other groups have patented technologies claiming to accomplish a similar goal,” explains Endy. “If we have a patent, we can assure the technology is free and available to all simply by not pursuing our patent rights. But if we don’t have a patent, someone else could claim the technology and restrict its use.”

Meanwhile, Sonnenburg, Whitaker and graduate student Liz Stanley have begun working with programmed microbes in laboratory mice, and Bonnet has returned to his native France to apply the biological computer to studying diabetes.

All of these researchers say they recognize that years of clinical testing and regulatory review lie ahead before the biocomputer can be used in humans, but the fact that they can contemplate the possibility for the first time is profound. Asked to predict possible directions the biological computer might venture, Endy hesitates for a moment, wary of overstatement.

“I’m not entirely certain where the biological computer might lead from a clinical or therapeutic perspective,” he says. “But, I do know that very modest amounts of computing inside living cells will be incredibly useful.”


E-mail Andrew Myers






©2013 Stanford University  |  Terms of Use  |  About Us