The Solexa Story

September 24, 2010


13 years ago, a beer summit in an English pub led to the birth of Solexa and—for now at least—the world’s most popular second-generation sequencing technology.

September 30, 2010 | On a warm August night in 1997, four thirsty Cambridge University chemists met for a couple of pints at their local pub, the Panton Arms. Two young faculty—Shankar Balasubramanian and David Klenerman—were joined by two postdocs, Mark Osborne and Colin Barnes. With a straight face, Balasubramanian insisted such beer summits were essential because the chemistry department didn’t have any conference rooms and his office was too cramped.

While the Panton Arms isn’t yet enshrined in scientific folklore like The Eagle—the pub a mile away where Jim Watson and Francis Crick exclaimed, “We’ve discovered the secret of life” to the bemused lunchtime crowd in 1953—that could change. That summer evening, the four chemists’ ideas for a new approach to DNA sequencing began to ferment. “I remember going home feeling pretty excited, as I often did after a discussion at the Panton Arms,” Balasubramanian said. “The acid test, of course, is how you feel when you wake up and sobered up.” They were still enthused the next morning. “I told the landlord that I’ll make him very famous one day. If I do, free beers for life! But I probably need to help him understand the importance first.”

The significance of that beer summit was finally spelled out 11 years later when, in November 2008, the Panton Arms patrons were among 100 authors on a paper in Nature describing the sequencing of the first African genome using their sequencing technology. “It’s very exciting,” he said in his Liverpudlian accent, slurping a pint a couple of weeks later. Solexa had kept their progress quiet for many years, fearing that as a small British company, it might get scooped by an Applied Biosystems (ABI) with deeper pockets. That same issue of Nature included articles describing the first Asian genome and the first cancer genome, also validating the technology of Solexa sequencing.

Born in India, Balasubramanian grew up in northwest England dreaming of becoming a professional football (soccer) player for his beloved Liverpool. Cambridge University ended up being his safety school. After earning his Ph.D., he spent a few years in the United States, studying the molecular biology of nucleic acids, before returning to Cambridge in 1994 on a Royal Society Fellowship. His research focused on DNA polymerase, the enzyme that replicates DNA, using fluorescence energy transfer (FRET) to study DNA-enzyme complexes.

A pivotal moment came in 1997, when he submitted a manuscript to the journal Biochemistry. The editors requested he perform an additional experiment that would require a laser. Balasubramanian looked up a new faculty member, spectroscopist David Klenerman. “We had a cup of tea, he had a laser, he helped me address the reviewer comments, the paper got published, and then we started talking about things we might do together.”

Those ideas centered on tracking the movement of DNA polymerase, by visualizing single molecules as they incorporated nucleotides step-by-step on a solid surface. One day, perhaps, they could draw inspiration from microarray companies like Affymetrix and parallelize the process. “We were just playing around, to be honest,” said Balasubramanian.

Nevertheless, in November 1997, the pair met with some venture capitalists at Abingworth, Balasubramanian armed with just four acetates. “We have an idea that could increase the rate and lower the cost of gene sequencing by [a factor of] 104 or 105,” he told them. The implications of a 100,000-fold improvement in DNA sequencing technology were not lost on the Abingworth executives or on the chemists performing due diligence, including Ron Cook, an organic chemist in San Francisco who had worked with PCR inventor Kari Mullis.

In the summer of 1998, Abingworth funded the Cambridge chemists, on paper at least, with Osborne and Barnes the only bench scientists. Balasubramanian named the start-up company Solexa. He chose the term ‘sol’ because it signifies light and was a contraction of ‘solo,’ as in single molecules of DNA. “We went from Sol to Sol Molecular, somehow an X appeared in the middle,” he said.

Solexa had the basics of a reversible sequencing chemistry—add one base then pause—a way of imaging single molecules, and a massively parallel array, with spots just 1 micron in diameter. Abingworth’s Hugh Rienhoff (see p. 34) told Balasubramanian that sample preparation could be a nightmare. So for his next presentation, Balasubramanian prepared a new acetate, showing a mixture of lots of colored DNA fragments (represented by squiggles) and an arrow pointing to a surface where they would be dispersed. The projected sequencing capacity was a gaudy 1 billion bases per run. The VCs said they’d be impressed if he could achieve a tenth of that figure.

In 1999, Abingworth gave Solexa a proper launch by putting up $3 million, and later raising a further $20 million. Balasubramanian and Klenerman visited the Sanger Centre to meet David Bentley, Jane Rodgers and Richard Durbin, leaders in the British arm of the HGP. Those conversations planted the seeds for personal genome resequencing, Balasubramanian recalled.

Solexa employee number three was Harold Swerdlow, a lanky, slow-talking New Yorker. Swerdlow set up Solexa’s first physical lab near the Sanger, which opened in May 2001. Nick McCooke became CEO while two key recruits arrived from Glaxo. Medicinal chemist John Milton joined after bumping into Balasubramanian, predictably, in a pub. Milton’s forte was more about blocking DNA synthesis using antiviral drugs, but Balasubramanian reckoned he was the ideal man to head up chemistry. Six months later, bioinformatician Clive Brown arrived (see “What Can Brown Do For Oxford Nanopore?Bio•IT World, Sept 2009), eager to escape the stifling bureaucracy of big pharma.

McCooke was increasingly eager to tout the medical implications of Solexa’s fledgling technology. As early as 2002, he was telling the BBC that patients might soon be getting a complete map of their genetic code from their doctor. “Thanks to Nick, it was all about ‘the $1,000 genome,’” said Brown. “Nick was always banging on me to get the cost down to $1,000,” said Swerdlow. He also wrote a patent on short-read sequencing—shotgun sequencing, realigning the fragments, and counting the differences.

Milton set about inspecting and reinventing the chemical process piece by piece, redesigning chips, surfaces, enzymes, fluorescent labels, optics, and engineering. “There wasn’t a single atom left in the sequence-by-synthesis chemistry at Solexa that was there originally in the academic founders’ labs,” he said. “They built a model T Ford, but we had to build a Ferrari.”

In late 2002, chief science officer Tony Smith (previously with Amersham) traveled to Boston to outline Solexa sequencing at a symposium on the $1,000 genome organized by Craig Venter, and size up some of the competition across the Atlantic. Smith shared the stage with speakers from 454, VisiGen, and Eugene Chan, the upstart founder of U.S. Genomics. “Every board meeting, the investors wanted to know: how are we doing relative to the claims of U.S. Genomics?” said Balasubramanian (see “Wanted: the $1000 Genome,” Bio•IT World, Nov 2002).

As that threat receded, another emerged in spring 2003, when Steve Quake published a rudimentary method for single-molecule DNA sequencing. Quake’s paper was seized by Stan Lapidus, who effortlessly raised $35 million to found Helicos Biosciences. “Quake was the first to publish [and] will say he invented single-molecule sequencing, but we had the first patent on it,” noted Swerdlow. Balasubramanian felt he could have submitted a similar story five years earlier.

But Solexa found that detecting fluorescent tags on single DNA molecules was no picnic. Milton conceded that Solexa’s PowerPoint presentations, which depicted DNA molecules as rigid as Redwoods, were idealistic at best. Klenerman tried putting a loudspeaker under the chip blasting high-frequency sound waves to make the DNA stand on end. When that didn’t work, McCooke faced two stark alternatives: build an enormous sequencer with a huge imaging capacity (as Helicos did) or abandon single molecules altogether.

The decision was made when McCooke heard about a struggling Swiss sequencing firm called Manteia, which had a clever method for amplifying DNA strands into clusters of about 1,000 identical molecules. McCooke bought the cluster chemistry rights for about $3 million and later picked up some of Manteia’s mothballed rigs in an Internet auction—machines that produced Solexa’s first sequence-by-synthesis data. “Buying Manteia instantly advanced us two years overnight,” said Milton.

While Manteia’s cluster chemistry was vitally important, the rest of the technology—the sample prep, surface chemistry, sequencing chemistry and much of the bioinformatics—remained the same. Read lengths steadily improved, first to 12, then to 25 bases. Smith recalls Vicki Harris generating a beautiful piece of data, “a real eureka moment.” Suddenly, investors “could see a set of stairs from the basement up to floor 20. It basically said, ‘It works!’”

At last, Brown could begin building an IT infrastructure to manage and assess the quality of the data. Among the key players were Klaus Maisinger and Tony Cox, who developed ELAND, one of the first short-read alignment program.

In early 2005, Solexa decided to try sequencing its first real genome, the famous ΦX174 virus that Fred Sanger first sequenced 25 years earlier. Brown lured Cox and Maisinger to work over a February weekend with free pizza to run the analysis software. On a Sunday afternoon, he tapped out an email to a dozen colleagues with the unequivocal header:

“WE’VE DONE IT !!!!”

The viral genome had been resequenced with more than 99.9 percent accuracy. An hour later, Smith replied: “This is amazing news!” and pleaded with everyone to keep the news quiet.

Interestingly, Brown and Milton had no interest in writing up their success for a journal article. “We didn’t publish as policy. It was a distraction. It was all hands to the pumps,” said Brown. “My interest is patents,” said Milton. “I’m well past the point that another publication does me any good.”

By the time of Brown’s jubilant email, Solexa had a new CEO, in the form of a former ABI executive, John West. Based in California, West had been in charge of the 3730, ABI’s most powerful sequencer, but as far as developing next-generation sequencing technologies went, he said, “The giant was asleep.”

West had serious doubts about Solexa’s long-term prospects if it remained a purely British company. “We had great scientists, but no one for marketing,” he said. McCooke had started looking at a Bay area biotech called Lynx Therapeutics, which had launched the first commercial massively parallel sequencing business (see, “Just Bead It,” Bio•IT World, Feb 2004), but was in financial straits. West wasted no time negotiating with Lynx CEO Kevin Corcoran, crafting a deal for a reverse merger. He then flew to the U.K. and presented the deal to the Solexa staff. He’d been on the job a week.

By many accounts, the merger was a near catastrophe. No one consulted Milton, Swerdlow, or Brown, and they weren’t happy. Brown called what happened next “a massive merger bun fight” over which company’s technology would survive. Some Solexa board members lobbied to retain the Lynx technology and phase in Solexa’s chemistry as it matured. Their reasoning was that Lynx allegedly had the better instruments, usable chemistry, a viable service business, a NASDAQ stock market listing and an American base. Brown vehemently disagreed: “If you got through what actually was correct and true, it was they had a stock market listing. The chemistry was dreadful, the instruments were useless, and the service business was dying or dead. They were three months away from bankruptcy.” West decided the best course was to make the Solexa chemistry work.

It wasn’t just the Brits who were upset. Lynx scientists had spent years refining their sequencing system, which while hardly perfect, had paying customers. Many of the engineers left, as did Corcoran. Ultimately it was the British company’s surface chemistry, hardware design, and software that was used in the first machines, although Swerdlow admitted they were “pretty flaky.”

By March 2005, Solexa was a publicly traded company on NASDAQ. “It was a stroke of genius economically,” said Swerdlow. “Lynx was worth $12 million, we were worth £20 million. Put it together, it very quickly became a $200-million company.” Solexa started shipping its first machines in mid-2006 to some of the major genome centers. Although later than 454 to market, Smith credited their competition for “showing that something other than Sanger sequencing could work. They whetted the appetite without satisfying it.” West priced the 1G Genetic Analyzer (the 1G stood for 1 gigabase, the target output of 1 billion bases/run) at about $400,000, to be competitive with ABI’s flagship sequencer. By his estimation (see, “Solexa Readies 1G Genetic Analyzer,” Bio•IT World, Feb 2006), the 1G could sequence a personal genome for about $100,000 in three months. But with read lengths of just 30 bases, producing a full genome assembly would be a massive computational challenge. West’s pledge of the $100,000 genome did not materialize in 2006—or 2007 for that matter. “John said the end of the year, just not which year,” said Swerdlow.

Brown’s team was determined to sequence the human genome as West had promised. In mid-2006, Solexa did ten runs of the human genome, producing a low coverage assembly, but priorities changed and the project was shelved. Back then, the flood of data was 10-15 gigabases per week, more than the Sanger Institute. “For about one year, we were the world’s biggest genome center,” Brown recalled. Meanwhile, his open-source software—dubbed “the pipeline” because he never bothered to give it a name—ended up going to every customer, not just the early adopters, even though it was originally only intended for genome centers (and in a rare misstep, the intended instrument software was shelved). It has since evolved into CASAVA.

In November 2006, Illumina CEO Jay Flatley (see, “Jay Talking Personal Genomes”) placed a $650 million offer for Solexa, which instantly provided the third leg to complement Illumina’s genotyping and gene expression platforms. The acquisition culminated what the Cambridge University press office called “one of the greatest commercialisation (sic) success stories to emerge from the University of Cambridge.” Flatley was bullish: “This acquisition... may prove to be one of the most successful acquisitions and new technology introductions in the history of the life science industry.”

But with scientific and commercial priorities changing, Brown and others felt it was time to move on. “I wouldn’t have left had we not done a human genome shotgun,” said Brown.

By February 2007, Illumina had sold and placed 12 machines in the field, and dozens of new orders soon followed. Although the read length was dwarfed by 454’s, the throughput and cost per gigabase were favorable. Flatley anticipated finally reaching the $100,000 genome by year’s end. “A grad student will be able to sequence the genome of an organism as a Ph.D. thesis,” he said. David Bentley’s team did indeed set about Illumina’s first human genome around Christmas 2007.

In 2007, Illumina’s revenues doubled to $360 million, with more than 200 GA instruments installed by year’s end, and doubled again in 2008. BGI and other genome centers expanded their fleets, but it was smaller labs driving much of the demand. In spring 2008, Illumina unveiled the GAII, with enhanced hardware, software and biochemistry upping the read lengths to 50 bases and throughput to 3 gigabases per run.

Using an Illumina fleet, the Sanger Institute’s output in 2008 was astonishing. In July, the institute sequenced its 1 trillionth base. (The Sanger press office helpfully pointed out that if printed in 12-point Courier type, the DNA sequence would span the earth 63 times.) “Approaching 90% of all the DNA ever sequenced has been done with the chemistry I built and Clive’s bioinformatics,” claimed Milton. “It’s been phenomenally successful,” said Brown. The year closed with the Illumina African genome paper published in Nature, with Balasubramanian, Bentley, Brown, Milton, and Swerdlow among the lead authors. “We had a corporate goal in 2008 to make human genome sequencing routine,” said Flatley. “We think we accomplished that.”

Reaching the market ahead of ABI’s SOLiD (see, “The Next Generation in Sequencing Is SOLiD,” Bio•IT World, July 2006) was crucial. “That’s one of the few reasons Illumina dominates the SOLiD system—they got there first,” said Milton. “Once you’re a genome center and everybody’s trained, you stick with [the technology].” That was a major reason why the Sanger elected to stick with Illumina platform—the pipelines and staff were already optimized for it.

“It was a race,” said Smith. “When you have a competitor as ABI, you better not only have the very best technology but you also better commercialize it with ruthless efficiency. If we’d been two years later to market, we’d have been head to head with Helicos, as opposed to a year ahead of ABI. That made an enormous difference.”

Solexa “could so easily have been a complete disaster,” said Milton. “The Brits are the great inventors, and the Yanks are the guys who are great at putting it in a box and selling it.” Brown said: “Enough of the boxes were ticked in terms of efficiency and performance and it’s won. We’re all very proud of it looking back but, God, it was a struggle.” Milton and Brown have since joined Oxford Nanopore, Britain’s third-generation sequencing hopeful, and ironically are linked with Illumina—Oxford’s sales and marketing partner—once again.

Back at the Panton Arms, Balasubramanian said the performance of his technology had exceeded the specs he had scrawled on his acetates a decade earlier. It came down to putting quality first, even when competitors were ahead. “There were times when we could have rushed ahead, could have made claims ahead of performance. But we were very English, actually, and quite understated in how we put this out.” As for the future, “If someone can come out with something that’s 1,000 times better than what [Illumina] can do, wonderful, because it’ll force the revolution even faster. That’s the scientist in me.” The goal of a $1,000 genome was within reach. “If it’s not Solexa sequencing, it’s going to be something else. It’s going to happen.”

Balasubramanian downed his pint and headed back to the department. His students aren’t allowed to run experiments if they’d been drinking at lunch time, but the professor had no such worriers. “I don’t do experiments anymore,” he said.

This article also appeared in the September-October 2010 issue of Bio-IT World Magazine. Subscriptions are free for qualifying individuals. Apply today.