Silicon, silicone and Alaskan salmon

Rubber, or more technically silicone, can be scaled down. That is, you can make a structure out of rubber and then radically continue to shrink it. Silicon is crystalline and silicone is amorphous, but it still works.

"It turns out that it has exquisite properties for making very small things," said Gajus Worthington, CEO of Fluidigm, which has devised a lab-on-a-chip array with approximately 30,000 hoses along with a computer to control and analyze biological samples piped into the chip. The chip is about 1.5-inch on either side and is enclosed in a 4x6-inch housing.

"We're increasing density at 4 times per year," Worthington said.

Approximately 17 life science companies have installed Fluidigm's BioMark system, which started shipping in November 2006. Effectively, the BioMark system can replace more traditional, and more costly, lab equipment used by drug researchers or hospitals. But one of the more interesting examples of how the technology works has been taking place in Alaska for the past two years.

There Jim Seeb, who until recently was the chief scientist for the Alaska Fish and Game department, set up a lab at Port Moller in the Aleutian Islands. Port Moller is about a week away--as the salmon swims--from the main fishing grounds.

During the 2006 and 2007 summer fishing seasons, Seeb performed genetic analysis on fish coming back home to spawn. With the data, the Fish and Game department was able to determine what fishing grounds to open. If the data showed, for instance, that the number of fish returning to the Lake Clark area was small, the department could shut that off and let breeding recover in the area. Meanwhile, if it showed that large populations were coming toward Ugashik, the department could open it up and tell the fishermen in advance where to be.

Salmon return to where they were born to breed, and the populations from various locations are slightly genetically different, so a gene scan can tell a scientist where a fish was born, and where it's headed during breeding time.

"It was dead on in 2006 and 2007," Seeb said. The predictive model created through the genetic data was more than 90 percent accurate. Approximately 40 million fish came back to spawn in the 2007 season and fisherman caught around 30 million, he said.

In 2005, Seeb used microsatellites to analyze where the salmon would go. That worked well too, but it cost twice as much, and he couldn't cover as many drainages.

Seeb is now at the University of Washington and experimenting--with a grant from the Gordon and Betty Moore Foundation--on ways to expand the genetic analysis program. One of the big questions is whether the samples caught and genetically tested accurately reflect the fish population swimming upstream.

How it works
A robot pipes fluids--a solution holding genetic markers or a medicine--into the small repositories on the right and left of the central, black chip. The fluids then travel through channels and then into the chip.

The chip itself contains roughly 30,000 hoses, 7,000 valves (which control the flow of fluids in the tube), and 5,000 reactive chambers (where a protein might be mixed with another molecule and then studied).

The valves are the key part, Worthington said. Microfluidic chips, designed to control the flow of liquids on a small scale, have been around for years. Miniaturizing valves, though, has been tricky. Most microfluidic chips consist of channels grooved out of silicon or glass and the valves themselves are mechanical devices.

For the valves, Fluidigm exploits the flexible nature of rubber. The hoses are arranged in a perpendicular array with hoses running left and right sitting atop of hoses running up and down. Increasing the pressure on a right-left hose causes it to bulge downward. The upper walls of these hoses are thicker than the bottom walls, so they bulge in only one direction. The bulge pushes into the downward hose and pinches it off.

"How do you control fluids? You do it with valves. Nobody had figured out how to make a very small value," Worthington asserted. The computer system, of course, has to precisely control the pressure in the hoses so that the appropriate reactions will take place in the reactive chambers.

The system is effectively a shrunken version of the equipment drug companies use today. In these systems, robots pipe fluids into small chambers capable of holding about 10 microliters, and then analyze the contents of the individual chambers. They might be looking for genetic material, or studying the chemical composition of a sample. The BioMark system uses the same robots, but the fluid chambers and all of the other plumbing is much smaller. The reactive chambers, for instance, are only 10 nanoliters, or 1,000 times less.

Ultimately, this reduces the cost and time of conducting experiments, as well as the amount of biohazard that has to be disposed of. Fluidigm started in 1999 with a process for preparing protein samples, but is now concentrating on the BioMark system.

"We have a 6 times increase in throughput," Worthington said.