by Gina Shaw
Remember when robotics and automation were a far-off, Jetsons-style future for biological research? Neither do we. It wasn’t really all that long ago that “high throughput” meant a few extraordinarily caffeinated, overworked lab assistants preparing and dispensing samples. But how quickly and completely times have changed. Today, benchtop automated liquid handling and sample-dispensing systems are de rigueur for any lab much above the level of your local high school chemistry class.
What are bioscience labs demanding—and getting—from automation and robotics vendors today? One critical factor: process control. Automation eliminates a lot of “human factor” errors, but it also can eliminate human judgment at key points when it’s particularly useful.
A common source of potential automation error lies in pipetting. When you want to transfer liquid from one place to another, you want to be absolutely sure that the pipette has drawn enough liquid into its tip—from the right place—and that it’s then dispensed the liquid on the other side.
“If you have a human doing this, they see tip, work with the liquid, can tell if it’s a short sample or a bubbly sample, and can make the decision to pipette again if needed,” says Jeff Hurwitz, vice president of robotics for Hamilton Robotics, Reno, NV. “Most machines on the market don’t have that ability: they pipette blind. They use capacitance, a change in electrical signal, in trying to find a liquid level in the tube or container. But that can change from hitting the side of the container or even bubbles in the liquid. It doesn’t assure you where the actual liquid meniscus is.”
That’s a problem in a lot of contexts—in blood bank work, in forensic work (where swabbed samples can clog the pipette tip) and of course, in pharmaceutical and biotech work, where scientists want to be absolutely sure what they’re transferring is all a sample of their hopefully-billion-dollar drug, and not 75% bubbles and 25% sample.
?
That’s why Hamilton’s STAR line has liquid sensor technology built into its independent tips. The Total Aspirate and Dispense Monitoring (TADM) system can differentiate between the side of a container, bubbles, and the actual liquid sample. “It monitors aspiration; it monitors travel over the deck with anti-drip technology; and monitors the dispensation,” Hurwitz says.
STAR also has the advantage of being an air-based system, with few of the fittings and connections that can lead to nightmares in troubleshooting when something goes awry. “Most common liquid-filled systems have well over 50 hand-type fittings,” Hurwitz says. “The fluid goes up into the tips, which are connected to tubing and routed through an arm to the back of the unit, to a valve to a syringe, and from there to a main system, so you have splitters and a wash station involved. We have only one piece: a handheld electronic pipette that picks up a disposable tip with our proprietary technology, CO-RE (compressed o-ring and expansion). There’s a flat surface on the tip and on the adapter that line up and pull themselves together to ensure that the tip is straight.”
In addition to avoiding the obvious hazards of multiple connections, air-based systems also avoid pulling sample into the same channel that contains liquid—even with an air gap, it’s almost impossible to prevent a dilution effect with water-based systems. (Try filling up such a system with blue food coloring and then pipetting clear water to see an example.
)??
Process control
A particular challenge for process control in automation is solid phase extraction, something that’s been part of the business model for Caliper Life Sciences (Hopkinton, MA) for two decades. “The difficulty when you’re doing any filtration assay is the possibility of particulates getting clogged in the wells of your filter plate,” says Kevin Keras, business unit manager for automation, consulting, engineering, and services. “You won’t know until afterward when you go to your assay plate. That can be very detrimental if you’re doing forensic work and have a limited sample.”
Zephyr, Caliper’s newest liquid handler, is designed specifically for solid phase extraction and has interchangeable 96 channel HV or 384 channel LV heads. It incorporates a highly intuitive graphical interface designed specifically for solid phase extraction. “You don’t need to know anything about the machine—you can just set up your samples and press the go button,” says Keras.
?
It also incorporates an ultrasonic transducer that allows the user to interrogate every well of the 96-well plate to assure that it’s been processed. “If it’s clogged, you can retry, or manually clear and assure your process,” Keras explains. “There’s a big market for this, because people recognize that it’s been a real problem with traditional solid phase extraction. With this kind of process control, you guarantee that your assay gets performed as a person would do it. These things are common in industrial applications, but they’re still relatively new to the life sciences.”
After hours
Another key element for many users of automated systems is programmability and scheduling. Torben Bruck, a chemical engineer with Dow Chemical’s biotechnology facility in San Diego, was an early adopter of an exciting new system made by Microreactor Technologies (San Francisco, CA)—a bioreactor that can run 24 individual cellular growth reactions, each one with individual temperature controlled and with individual oxygen feed and pH controls. Fitting comfortably in an 18-by-18 footprint, the reactor replaces large, costly shake flask reactor set-ups.
But it had one big problem: the reactions Bruck and his team follow often extend over a time period of five to seven days, and the reactor as it was designed had no alternative to having someone at the lab 24-7 to take samples, track cell growth and determine if the cells needed feeding. “What that means is having me here at strange hours,” says Bruck. “And no individual can process the things that need to be done with 24 reactions, so we have to do it in groups, which takes away from our ability to handle each one. What we wanted was the ability to have the system operate itself as we would if we were here.”
And that’s what Bruck hopes to get from a new adaptation of the Microreactor, the AutoReact Mini-Bioreactor System, developed by Hudson Control Group (Springfield, NJ). Hudson developed a method for automatically feeding and sampling the reactions going on in the Microreactor, and developed algorithms designed to predict when feeding and sampling needed to take place.
“We gave them the ability to do an unlimited number of algorithms to carry this process through a five- to seven-day period,” says Phil Farrelly, Hudson Control’s founder and president. “Each reaction can have its own set of algorithms, and there are separate algorithms for feeding and sampling, so in total there can be 48 sets of algorithms. You can walk away from the whole system for days at a time and just let it run.”
The scale-up isn’t exact, Farrelly admits. “Cells growing in a small 10-ml tube will probably behave differently than in a large flask or in a process tank. But the idea is to get inferences about what to do to optimize your media and growth conditions so that you have knowledge going forward, and this takes the place of a roomful of equipment. Ordinarily, to do 24 reactions at a time you’d need an enormous lab.”
Production and reproduction
?
As AutoReact demonstrates, automation isn’t about replacing people with robots (unless the person being “replaced” is an exhausted scientist who’d prefer to be sleeping early on a weekend morning). “Most labs are pursuing automation because they want to free-up investigators’ time to do other, more important things in the lab,” says Keith Roby, product manager for strategic marketing applications for Beckman Coulter (Fullerton, CA). “With higher throughput and more data, you need to spend more time analyzing that data—so you want to save time in other areas.”
And time is not the only issue. Another critical issue is reproducibility—ensuring the precision and accuracy of results. For example, says Roby, Beckman Coulter has recently been collaborating with Affymetrix on target preparations for gene chips. “Customers who have sent in samples have noted that they would get varying results from technician to technician. Before going automated, if company A sent in a sample, they always had to have technician A do those samples, because they found that if they mixed up the technicians, the controls were off and they didn’t get the same reproducible results,” says Roby. “Gene expression is so expensive and you have so many reagents to pipette, and with something like a 2-microliter sample, it’s not easy to see if you pipetted in well B6 vs. well B5. With gene expression, if you make one mistake, you don’t find out until you analyze the chips, and then you’ve spent $1,000 or more for that sample.”
Beckman Coulter’s BioRAPTR FRD microfluidics workstations, which now have 1,536 and even 3,456-well plates, allow more of an assay to be done at one time and in one place, enhancing reproducibility and eliminating variability. (And at the same time, allowing for the screening of 100,000 compounds in less than a week.)
“If you’re screening for 50,000 compounds, you want to make sure you can do as much of that on one plate in the same day with the same cells you just grew up,” says Roby. “With more cells, you can eliminate the plate-to-plate and cell culture preparation variability. Now, the caveat is that you’re using the same plate dimensions to go from 96 to 3,456 well plates, so your volume is a lot lower and you’re going to have to worry about evaporation, static electricity, and dispensing very small volumes quickly and accurately.”
The BioRAPTR systems address those issues in part with a “sacrificial” barrier around the plate that the system pipettes to. “We’ve seen ‘edge effects,’ in which outside wells will evaporate faster than the inside wells. So there’s a sacrificial row around the outside of the plate, and those wells without sample will take the evaporation,” says Roby. “There’s also an interlocking plate, so that we can get an exchange of gases but the evaporation is a lot less.” The plates, made out of cyclo-olefin polymer, are much flatter than a polystyrene plate and have an extremely low binding capacity.
?
Leading-edge automated systems aren’t just for big pharma and biotech anymore, Roby adds. “The cost of automation is becoming a lot more affordable for clients like hospitals, who are interested in things like automating immunoassays,” he says. “We’re seeing a lot of dedicated automation, with a more focused workflow per robot, to decrease cost. Some customers still want the flexibility of the open platform, but there are more and more who want that dedicated platform where you just load your samples and push a button.”
? ??