Among companies pioneering protein structure determination, Astex stands out for its anti-establishment approach
By Aaron J. Sender
In 1982 Harren Jhoti dropped out of University of London’s Birkbeck College. “I was going to be a rock star,” he says. He gave his punk band a year to sign a record deal and then went back to the books.
In 1999 he dropped out again — this time from a position as head of structural biology and bioinformatics at Glaxo Wellcome — to start his own company. Today the 39-year-old CSO of the Cambridge, UK-based structural biology startup Astex Technology, dressed in a smart black suit and freshly pressed blue shirt, says he prefers classical music to the Clash. “I’m not sure whether I still have some of that anti-establishment punk mentality,” he says, “but I still like to challenge mainstream views.”
As he and renowned University of Cambridge crystallographer Tom Blundell began thinking of starting a structural biology company, Celera and Incyte were betting on getting millions of dollars per subscription to their genomic databases. Structural Genomix of San Diego was gearing up to offer the protein structure equivalent. And Syrrx, also in San Diego, put its efforts into ground-up development of robots and automation to generate thousands of proteins more quickly than anybody else could.
But Jhoti, whose college days’ band was called Emperor’s Clothes, believed that these platform-and-data business models weren’t wearing any.
“The economics just didn’t make sense,” he says. “The value was not in solving the protein crystal structures themselves. It’s how you use that information.”
With the US, UK, and German governments each supporting large-scale protein structure projects, “we expected, when we set up Astex, that most proteins will have their structures solved within the next five to 10 years,” he says. “We believe it’s pre-competitive.” So instead of promising to churn out thousands of structures a year, as Structural Genomix and Syrrx did, Astex from the beginning turned its sights on crystal structures simply as a means to design drugs.
True, Structural Genomix has now shifted its plans from targets to leads as well. And both that company and Syrrx, each with its own synchrotron beamline, are better equipped to determine far more new structures each year.
“We’re not going to be seduced by numbers,” says Jhoti. “Unless you’re going to solve hundreds and hundreds of novel protein structures, I just don’t understand the economics of building your own beamline. And this comes back to the original business model of Syrrx and Structural Genomix: Simply put, if they were going to go after 5,000 novel structures in five years they have to build a beamline,” he says. “The question is, do you think that’s the right thing to do? Do you think that’s where the value is?”
Astex has a more modest ambition of looking at 15 to 20 proteins a year. The company prefers to leave most new structures to others, and instead focus on using x-ray crystallography on proteins bound with small molecules. Astex immerses the protein crystal into a cocktail of small molecules and then shoots x-rays at it to see whether any of the compounds have bound.
“The bottom line is we just use the x-ray crystallography as a method of getting to the leads,” says Jhoti. “We always were focused on small lead compounds and that put us in a good position, because fortuitously it’s all about product now, it’s all about small molecules.”
With its pitch as a drug discovery company, Astex has raised more than $40 million, a huge amount for a British company. That’s enough, according to Jhoti, to get the company through the next two-and-a-half years. Astex is getting ready to move to a new 35,000 square-foot building to house the more than 80 employees that now crowd its current 15,000 square feet of space. With plans to add another 50 employees by next year, Astex also bought another 35,000 square feet adjacent to the new building to allow room to grow.
Deals from day one
Fortunately for Astex, Jhoti has had more luck signing deals as a biotech entrepreneur than as a punk rocker. From day one, the company had already lured a customer: Johnson & Johnson’s Janssen Pharmaceutica. “As a first time doing a biotech startup, I thought this is not as hard as people make it out,” says Jhoti. “It was fascinating, to be honest. They called us up and said they’d heard that I’d left Glaxo and that I was going to set a company up with Tom Blundell and they wanted to know whether we would do a collaboration.”
At first Astex was screening proteins with known structures against compounds that Janssen was trying to tweak for a snugger fit into the target’s active site — a fee-for-service arrangement. Satisfied with Astex’s work, Janssen expanded the deal to solve a novel structure, to which Astex retains the rights. “So we kept that structure in-house and are running our lead discovery approaches against it,” says Jhoti.
But a big pharmaceutical company like Janssen is not one for monogamy. “When we started to look for x-ray crystallography they were one of the only companies,” says Jan Hoflack, Janssen VP of medicinal chemistry and enabling technology. Now he is looking at Structural Genomix, Syrrx, and others as well. “It’s good to have diversity,” he says.
From birth to boom
Astex headquarters sits just a few miles from where crystallography was born at the University of Cambridge. It took decades for Lawrence Bragg, Max Perutz, and John Kendrew to solve the first structures there. For his PhD at the University of London, Jhoti worked on determining the structure of transferrin, a 60,000 molecular weight protein that soaks up iron from storage sites and carries it to different cells. “My supervisor had started working on that project in 1972. That’s when they actually got the first crystals. And then they spent all the other time just trying to solve the structure,” he says.
Developments over the past decade have led to a boom in structure determination: High-energy synchrotron x-ray sources, faster digital detectors, better hardware and software for managing and analyzing the gigabytes of x-ray diffraction data, and even something as simple as methods for freezing crystals so that they are not damaged by the powerful radiation, have made structure determination easier and faster. In fact, during the last 10 years the number of deposits in the Protein Data Bank jumped from fewer than 2,000 to nearly 16,000 structures. The number of publicly available structures will continue to mushroom as government-funded projects take off.
“It’s great for us that lots of novel protein structures are being solved,” says Jhoti. “A lot of them will be solved out in the public domain. And the challenge will be to use these structures to come up with lead compounds.”
Automation ain’t all that
Astex is one of some 60 companies, the majority of them biotechs, occupying Cambridge Science Park. It shares its building with Toshiba R&D, and the European offices of Incyte and Amgen are directly across the road. The ground floor houses Astex’s two laboratory x-ray sources. Glass doors partition each from the rest of the room to protect those in the room from the high-powered x-rays currently bouncing off a crystal inside. A computer screen displays the diffraction patterns recorded by the Jupiter CCD camera aimed at the crystal, a similar model used to collect data at the Spring 8 synchrotron in Japan. “It is between three to five times faster than conventional image-plate detectors for x-ray data collection,” says Jhoti.
The x-ray source on the left has just finished collecting structure data on an Alzheimer’s associated target for Janssen. A video monitor above magnifies the crystal still mounted on a nylon loop on the x-ray source.
After the data collection, a proprietary software called AutoSolve turns the diffraction patterns emitted by the protein-ligand complex into a three-dimensional structure within minutes, a process that usually takes days.
Wet labs, however, dominate most of the building. In one room scientists experiment with various DNA constructs to find the one that expresses the best protein for crystallization. In another, researchers transfer purified proteins into wells for crystallization trials. What’s glaringly lacking is the type of automation boasted by Structural Genomix and Syrrx to speed these trials.
Proteins by definition are idiosyncratic. That’s why they have different biochemical functions. That’s why we are alive. Often thousands or millions of conditions — such as various pH levels, salt content, polarity, and temperatures — must be tried before the appropriate environment for crystallization is found. In the end, many proteins never crystallize.
“We’ve invested a considerable amount of human effort as well as finances in establishing high-throughput crystallization platforms,” says Structural Genomix CSO Stephen Burley. The company can perform 2 million crystallization experiments simultaneously. And Syrrx’s crystallization robot, Agincourt, uses minuscule amounts of protein and processes hundreds of thousands of crystallization trials a day.
Jhoti, however, doesn’t believe automation is the answer to find the right conditions to crystallize a protein. “We haven’t embraced automation in the same way Syrrx has. We’ve crystallized very difficult proteins without the need for extensive automation,” he says. “Crystallization robots have been around for 15 to 20 years. There is actually nothing new in automation in crystallization. There has been a significant amount of hype in terms of saying robots are going to solve all the problems in crystallization.”
The P450 Club
Instead of automation, Astex prefers to invest in high-quality, experienced protein experts. But unlike in the US where many top scientists commonly join small startups, “in the UK most of the best scientists would not consider going into biotech as a viable option for their careers,” says Jhoti. So he has had to be aggressive. After signing Janssen he poached its head of protein expression, Jeff Yon, to fill Astex’s director of protein technology position. “He was actually one of the key scientists we were working with, so it was kind of tricky recruiting him,” says Jhoti. “It’s our benefit and their loss.”
When he heard that researchers at the Scripps Institute had cracked the structural biology of a rabbit P450, a critical class of enzymes involved in drug metabolism and toxicity, Jhoti hopped on a plane to California to recruit two of the key lab members, Jose Cosme and Pamela Williams.
Many groups worldwide have been trying to get hold of a human P450 for more than 10 years. P450s interact with almost every medication and their activity is a major component of many drug failures in development or withdrawal from the market. Bayer, for example, had to withdraw its statin, Baycol, last year because of suspected P450-associated side effects. Just before Christmas, Astex announced it had it solved the world’s first human P450. “The more difficult structural biology problems will be solved with highly experienced expertise, rather than high-level automation,” says Jhoti.
The assertiveness has paid off; Astex’s P450 program has landed it two more pharma collaborators — AstraZeneca and another yet to be announced. “They metabolize everything that you put in the body,” says Nigel Gensmantel, director of physical and metabolic science at AstraZeneca Charnwood in Loughborough, UK. “So being able to understand how to control P450 activity is crucial to drug design.”
AstraZeneca was the first member of what Astex calls its P450 consortium. Members pay an up-front fee to join the club and then pay a milestone for each new P450 delivered. They will also have access to Astex’s structure-based screening of small molecules. AstraZeneca has several potential drug compounds that would probably get canned because of their unwanted interaction with P450s, says Gensmantel. “If we could get rid of that P450 activity then probably we could have something that could be a candidate for development.”
Although Jhoti believes his company’s steady eye on drug development gives it an advantage over competitors such as Structural Genomix, the head start is far from insurmountable. Although, Astex was the first to sign a drug development deal, Structural Genomix has recently announced three collaborations of its own: Millennium, Aventis, and Graffinity. And Syrrx has signed on Pharmacia, Celera, and Cubist Pharmaceuticals.
But in order to achieve the drug development royalties these structural biology companies covet, they still have to convince pharmaceutical companies that structure is valuable enough on its own for drug discovery to merit giving up a chunk of drug profits. “We’re under no illusion that structure-based design is going to provide all the answers. Designing drugs is just too complex,” Jhoti readily admits. “But it’s clearly going to have a big impact.”
Although the dreams of his youth did not crystallize, he says, his experience chasing a record deal helped him see how to make his current dream a success. “You need to convince people of your vision.” But this time he won’t give up so easily.