WHEN Stephen Quake visits his colleagues’ biology labs, he becomes almost claustrophobic. All that glassware – shelves and benches piled high with test tubes, beakers and pipettes – do they really need so much clutter? Not any more, reckons Quake, a physicist at the California Institute of Technology in Pasadena. He is working on a new kind of device that will sweep the benches clear.
And it’s not a moment too soon. Today’s biology experiments, for example those deciphering the complex biochemical pathways of life, often involve several dozen, or even several hundred, different steps. Sequencing genes is a laborious process that involves heating and cooling samples many times. And in chemistry, isolating particular enzymes or possible drugs is a tedious process of trial and error, involving many different experiments.
In many ways, lab work today is not so different to having a roomful of people operating a series of vacuum-tube computers to solve some large mathematical problem. Back in the 1950s, the solution to fast, cheap computers was to etch all the circuitry – thousands of transistors and their connections – onto a single polished silicon chip.
Advertisement
So why not do the same for lab experiments? The idea of reducing the scale of laboratory plumbing to chips just a few millionths of a metre across is called microfluidics, and it is part of the rapidly developing field known as MEMS (microelectromechanical systems). Mainstream MEMS borrow the techniques used for making integrated circuits to build tiny silicon-based devices. Researchers have already made MEMS that can move a few cells across a chip or mix two reservoirs of liquid. But over the past few years Quake’s group, and other researchers around the world, have found that silicon, which is a hard, brittle crystal, is not the best choice of material. Instead they have turned to silicone, a soft, flexible rubber famous for sealing bathtubs and augmenting breasts.
The advantages of using silicone in microfluidics are far from cosmetic. The stiffness of silicon makes it hard to build devices with tiny, moving parts. A silicon valve has to be at least several micrometres across to flex enough to open and close. Also, in order to close completely it needs some sort of flexible cushioning, like the rubber seal on a refrigerator. Quake realised that he could eliminate these problems if the whole of a microfluidic chip was built out of a more flexible material, such as silicone.
Not only that, but silicone turns out to be easier to handle than silicon. Quake’s group uses the most popular silicone, polydimethylsiloxane (PDMS), which is 50 times cheaper than a similar volume of silicon. PDMS does not require clean rooms, vacuum chambers or any other of the expensive and complicated equipment necessary to keep silicon chips smooth and pure. Silicone “is easy enough to work with that you could do it in your own garage,” says Quake.
Despite these differences, the manufacture of silicone chips uses pretty similar methods to those for silicon chips. In 1996, George Whitesides and his colleagues at Harvard University in Cambridge, Massachusetts, developed soft lithography – a collection of techniques for moulding and stamping microstructures into elastic materials. Using Whitesides’s techniques, Quake’s group etch a negative mould of pipes, valves, pumps and reservoirs onto a slice of glass. They then cover the mould with soft, warm PDMS, cool it down, and peel it free. The process is so easy, says Quake, that “my students can sit down and within 48 hours have a chip in their hands”.
Soft and simple
Many simple microfluidic devices have now been built using soft lithography. In 1999, for example, Quake’s group made a cell sorter. This is a vital biological tool, typically used when an experiment produces two types of cells, one of which is marked by a fluorescent dye. The cell sorter’s job is to round up all those marked by the dye. In Quake’s microfluidic sorter, the cells travel in single file down a microscopic pipe. Since PDMS is transparent, photomultiplier detectors can see the cells as they flow down the pipe. At a T-junction, a valve controlled by a photomultiplier that detects fluorescence directs the cells into two tiny reservoirs – one for those cells with the fluorescent dye, the other for those without. Quake’s cell sorter costs a few thousand dollars. That might sound a lot, but it’s less than a tenth of the cost of conventional laboratory equipment that does the same job. And the device is less than a hundredth of the size.
Quake’s cell sorter was just the first step in his silicone revolution. After all, it is still a discrete device. To put it in context, the transistor was a huge improvement over the vacuum tube, but a computer made from discrete transistors would still have a complex, bitty structure. So along came integrated circuits. Was there a way to do something similar with individual microfluidic devices? In September, Quake, along with students Todd Thorsen and Sebastian J. Maerkl, described the crucial step towards the biological equivalent of the integrated circuit (Science, vol 298, p 580). They showed how to integrate thousands of individual microfluidic devices onto a single, low-cost, reliable chip. “We’ve been able to make devices that have the equivalent complexity of large-scale integration [of electronic circuitry],” Quake says.
The secret of the microfluidic integrated circuits is that the chips contain several layers (see Graphic). One layer, called the flow layer, consists of the pipes through which the cells and chemicals flow, as well as the reservoirs in which they are stored or mixed. Above this is a second layer, called the control layer, which regulates the flow of fluids. Channels on the control layer run across the fluid-carrying channels on the flow layer below. A valve is formed where these two channels overlap. A thin membrane separates the layers, and when air is pumped into the control channel it pushes down on the membrane, pinching off the fluid flow. “It’s sort of like stepping on a garden hose,” Quake says. The valves in these layers open and close in milliseconds; a series of valves along a single channel can open and close in sequence, pumping fluid through the channel. And a third layer can be added to control the flow of signals in the control layer.
To demonstrate the power of microfluidic integration, Quake’s group constructed two large-scale devices. The first is a microfluidic analogue of a memory chip. This consists of 1000 individual reservoirs arranged in a rectangular grid. Each reservoir is capable of holding 250 picolitres – about 250 times the volume of a droplet of ink sprayed by an ink-jet printer. The control layer consists of some 3574 valves that control the flow of fluids, rather as the circuitry on a RAM chip controls the flow of bits. It takes just 20 control lines, thin aluminium tubes like hypodermic needles that inject air into the control channels, to direct fluid to any of the individual reservoirs. To demonstrate their device, the researchers injected blue dye into selected chambers to make a low-resolution display. Not surprisingly, they chose to emblazon “CIT” across the first microfluidic integrated circuit, the initials of their place of work, the California Institute of Technology.
But any technology needs a practical application. Enter chemist Frances H. Arnold, a colleague of Quake’s at Caltech. She finds new biological molecules by allowing bacteria to mutate and then selecting for a desired property, for example, the ability to produce a particular molecule that will bind to a receptor in the body, or a useful enzyme for industry. “I have to make large numbers of mutant molecules and screen them for interesting or new biological functions,” Arnold says.
With this kind of time-consuming work in mind, Quake developed a device that duplicates on a chip a procedure that would usually take up the entire resources of a small laboratory. The device has 256 reservoirs that contain a reagent that identifies mutated E. coli bacteria. First, the device is flooded with a solution of bacteria, diluted so that each reservoir on the chip will end up with just one bacterial cell (or none). Mutant E. coli glow red in the presence of the reagent, so photomultiplier detectors can be used to find the reservoirs containing the mutant cells, which can then be plucked off the chip.
Needle in a haystack
Arnold is impressed. “I can look for the proverbial needle in a haystack – a rare molecule or cell that has acquired a feature or behaviour,” she says. This technology will enable Arnold to collect just the cells she wants. “Some day we will be able to automate the whole process,” she says.
Such high-throughput screening is not the only application that Arnold envisages for Quake’s devices. Because cells can be isolated in a reservoir whose contents are controlled by microfluidic flows, researchers can study how a single cell responds to its environment or to changes in its DNA. Previously, any experiments to do this involved hundreds of steps because you couldn’t easily isolate single cells. “By looking only at a population of many cells, we could only see the average behaviour of the population and not all the fun and crazy things the individuals do,” says Arnold.
Quake has now set up a company called Fluidigm, based in San Francisco, to commercialise his devices. Its first product is a layered silicone chip to help chemists with the laborious task of protein crystallisation. In order to design drugs, you need to understand a protein’s biological activity. One way of doing this is to make proteins into crystals and use X-ray diffraction to determine their three-dimensional structure. But making large proteins form crystals can be difficult; it is a trial-and-error process in which scientists have to mix the protein with hundreds of possible reagents to see which combinations work. The problem is compounded by the fact that often only a tiny sample of the protein in question is available, isolated at great expense.
Fluidigm’s approach is to perform these hundreds of experiments simultaneously on a single, disposable chip. The chip can be programmed to mix precise quantities of protein and reagent in its 144 reservoirs. The device uses less than a hundredth of the volume of the often-expensive chemicals needed in standard techniques. And there’s another advantage too. For reasons that Quake does not yet entirely understand, the actual physical kinetics of how the crystals nucleate and grow is also superior. A larger percentage of the experiments yield crystals, and these grow more rapidly.
Such successes, Quake believes, are an indication of what is yet to come. Being able to run forensic tests on a tiny rubber chip without waiting for lab reports could revolutionise law enforcement, for example. In hospitals, microfluidic chips could quickly sequence patients’ genomes or those of the bugs they are coming down with. In your home, they could detect and analyse chemicals in the environment that you might be allergic to, find lead in your water or even pick up the anthrax spores on your post.
If you thought the silicon chip was everywhere, you ain’t seen nothing yet. Just wait till the silicone chip gets out of the lab.