Berta Strulovici was faced with a choice. When she joined Merck five years ago from Tularik to head its high-throughput screening operations, the pharmaceutical company had no centralized screening facility. "That was my mandate," she says.
She knew what she wanted. "People at Merck used to do a lot more biochemical rather than intact cell assays," she says. "And from the beginning I was looking for technologies that allow miniaturization, particularly of intact cells."
So she formed a group of four people to evaluate the options. But no such technology existed. Meanwhile, a small startup in San Diego, Aurora Biosciences (now PanVera, a wholly owned Vertex subsidiary), was getting started on plans for a product that seemed to be the perfect solution. Its Ultra High Throughput Screening System promised it all: 3,456 well plates in the same footprint of the industry standard 96-well plate, assay for intact cells with volumes scaled down 100-fold, and a throughput of 100,000 compounds a day.
"The idea was very noble," says Strulovici. "But it was just an idea. They had a lot of things on paper at the time, but they had no products, nothing to even show."
She needed to make a decision: "Should we wait four to five years and buy something off the shelf? Or should we get in early and participate in the development?"
Although committing time and resources on vaporware presented some risks, she decided it was worth it. "Others were barely talking about 384-well plates, while Aurora was presenting assays in two-microliter volumes in extremely high density 3,456-well microtiter plates," she says. "That not only decreases consumable costs, but also increases throughput because you just have one plate instead of 36."
As it turns out, Merck was able to push the screening system's throughput even farther than advertised — up to 200,000 compounds a day. And although it became fully operational just a year ago, it has already saved Merck millions of dollars. "A single assay that could have cost $250,000 now costs just $14,000," says Strulovici.
Merck's UHTSS occupies a 16,000-square-foot, single-story building in North Wales, Pa., about 35 miles northwest of Philadelphia. Inside, microtiter plates cruise along a four-lane conveyor belt, two lanes in each direction, that acts as the main thoroughfare linking the various parts of the screen machine: a compound store that holds as many as 2 million potential drugs; incubators that mimic physiological conditions; liquid-handling stations specially designed for minute amounts of reagents.
Smaller conveyor belts intersect the main one, and gates pop up to direct traffic, sending the bar-coded plates to their appropriate exits. A half-dozen robots lift and stack plates and move them on and off the conveyor. "The system knows at all times which plate is where and what it is in each well," says Oleg Kornienko, a Merck senior research chemist who manages the UHTSS system.
A database collects data on the fly and automatically identifies hits, measuring response versus concentration of any given compound. A "cherry picking" robot then "picks hits in the primary screen from individual wells on separate plates and either dilutes or consolidates those potential actives to a new plate and registers it with the database," says Kornienko.
There's also a mini-me version of the system, used to develop assays. "It has an identical reader. And it has an identical reagent dispenser. But it's just lower throughput," says Kornienko. "So I don't need to waste a lot of the big system's time," he says.
Assay development is the current bottleneck, says Strulovici. Not every assay can be scaled down to work in such tiny wells. Factors that are often ignored in larger wells must be considered. Evaporation, for example, becomes an issue. So each new assay must first be modeled on the smaller system. "We are now trying to see how we can enhance our efficiency of developing the assays so they are ready and waiting for the system."
The building that houses the UHTSS system had to be carefully designed. "It was new, outfitted from scratch for this operation," says Strulovici. For example, the screening system requires a completely dust-free environment. "When you work with such tiny wells and low volumes, lint could be a problem," she explains. "And this is particularly important for the technology that Aurora had, because lint actually fluoresces in one of the same wavelengths as the chemistry."
One advantage of working with Aurora early on was the ability to help direct the UHTSS's development, say the Merck researchers. Aurora designed the instrumentation around its own fluorescence-based assays. "The reader that came with the UHTSS could only read fluorescence. And if you want to do luminescence, for example, it wouldn't do that," says Kornienko.
So at Merck's insistence, Aurora added a 1,536-well plate capability to the system so that it would be compatible with third-party readers.
Strulovici also wanted the ability to link the Aurora-supplied database with Merck's. "But each of them had a different schema. Each had its own design. And it wasn't so easy to create the handshake between the two," she says. So she got Aurora to add an open-source API.
"I have to say we were pretty aggressive," she says of the push to direct the final product's design. "We knew that it was not going to be a design that fits only Merck. But for our investment we wanted to make sure we have more flexibility," she says
Even though Merck's collaboration with Aurora has been scaled down since biopharmaceutical company Vertex acquired Aurora more than a year ago, Strulovici is confident she made the right decision. "We knew it was not going to live forever as a tiny company," she says. Without taking some risks, "we would never work with startup companies and they have a lot to offer."
"If Aurora would have stayed Aurora for many years to come, we could have collaborated a lot more with them for new development," she says. "There are some very talented engineers there who are very eager to continue to work with us. The door appears open at the moment." In fact, PanVera just delivered Merck an updated, more sensitive reagent dispenser. "But we'll see how it evolves," says Strulovici.
Merck is not taking any chances. "We are purchasing this big multi-year service and maintenance agreement from PanVera. We want to make sure that this expensive system keeps running," says Kornienko.
Regardless of its developer's future, "the system has completely lived up to its expectations," he says. It seems to have paid off for Merck to get in early with Aurora: Although Vertex says it will continue to maintain PanVera's current customers, it would not say whether it plans to pursue any new UHTSS accounts. So with the only other installations at Bristol-Myers Squibb, Pfizer, and Vertex, "We are one of the only four companies in the world to have this system," says Kornienko — leaving all others fluorescing green with envy.