While starter culture customers prioritize acidification and phage resistance, they also value other qualities including texture and aroma. “You have to acidify quickly but you must also produce texture. You’ve experienced liquid yogurt?” Horvath asks me as we are about to tuck into dinner. “This is due to phages that have killed the texturing strain.” Horvath’s group developed starter cultures for a variety of fermentations including more than 1,000 different French cheeses. The starter culture for pizza cheese is far different to those used to produce Camembert. And because phages can manifest at any time, Horvath’s group has to formulate different starter cultures for the same product with unrelated phage sensitivities.
Much of Horvath’s work involves selecting daughter strains that are immune to the phages that attacked the parent. The process is survival-of-the-fittest straightforward: add a phage to a sensitive strain in the lab, wait patiently and search for survivors—naturally occurring mutants called bacteriophage insensitive mutants (BIM)—perhaps because the phage is unable to attach to the bacterial cell surface.
In September 2002, while attending a symposium in the Netherlands on lactic acid bacteria, Horvath came across a poster presented by Alexander Bolotin. The poster mentioned a repetitive DNA motif called “SPIDR” (spaced interspersed direct repeats), which would later be renamed CRISPR. “We have identified a region with repeats that is very useful for strain identification,” Bolotin stated. Horvath was so intrigued that he snuck a photograph.
Back in the lab, Horvath compared the sequence from one of his own group’s Streptococcus strains (LMD-9) with the strain that Bolotin had studied and other strains. To Horvath’s delight, there was a huge diversity of spacers across the SPIDR repeat regions. Every strain was different, resembling a DNA fingerprint. Horvath noticed something else: some of the sequences alternating with the SPIDR repeats matched the DNA of viruses, suggesting a link between spacers and phages. “By comparing spacer sequences with known [viral] sequences, we saw identities with phage sequences. Yes, in 2003!” At the time, only Mojica and a handful of other investigators were remotely interested in the CRISPR repeats. Horvath tried in vain to get his supervisors interested in a project he dubbed “CRISPy-SPIDRs,” but researching obscure virus biology in the food division of a chemical company wasn’t likely to win many converts. “We were told to stop working on that,” Horvath said. He continued his CRISPR research on the side, running computer searches just like Mojica.
Attitudes changed after Rhodia was acquired by Danisco, catapulting the Danish food ingredient company to second in the starter culture market, trailing only another Danish company, Chrysanthum. In 2004, every other loaf of bread and a third of ice creams contained Danisco ingredients such as color, texturants, or emulsifiers. Suddenly flush with money, Horvath felt reborn. That December, he was finally able to buy a DNA sequencing instrument. “What did we do? CRISPR sequencing! The more we sequenced, the more obvious it became!”
Horvath was on the verge of a crucial discovery in the brief history of CRISPR. It came courtesy of a new colleague, a French expat based, appropriately, in Wisconsin—the cheese state.
“I used to be French,”21 says Rodolphe Barrangou, professor of food science at North Carolina State University (NCSU). We’re speaking in the dramatic setting of the university’s 21st-century Hunt Library, like something out of The Jetsons. Barrangou’s self-confidence borders on a swagger, an impression magnified by his penchant for wearing cowboy boots to alleviate strain on his back following a serious basketball injury. He drives a modest Honda Accord that he has maintained since graduate school, adorned with a personalized CRISPR vanity plate.
Barrangou was born in Paris, but after he moved to North Carolina for his PhD, he fell in love with the Tar Heel state. During his PhD, which was funded by Danisco, Barrangou’s twin interests were developing next-generation probiotics and starter cultures to ferment milk (“Maybe in the spirit of Pasteur” he says, a little cheesily). His first publications were on the bacteria and attendant viruses involved in sauerkraut fermentation, including the SPIDR repeats.
In February 2005, Barrangou and his wife drove their Honda to Madison, Wisconsin, where Barrangou joined Danisco. The reports that year by Mojica and others marked the first signals of a direct connection between CRISPR elements and viruses. Barrangou began characterizing the genomes of starter cultures, checking in with Horvath. He used the CRISPR repeats to fingerprint the S. thermophilus starter cultures (a key ingredient in yogurt manufacturing), using “those peculiar loci to tell which strain is which and where it came from.” The more they sequenced, comparing new strains with older cultures thawed from Horvath’s freezer in Dangé-Saint-Roman, the more it became obvious that the CRISPR repeats could grow and evolve.
A trio of experiments clinched the association. First, Barrangou asked, what happens when a bacterial strain is exposed to a virus? “We saw the [bacterial] immune system activate itself and pick up new pieces of DNA from the viral genome and integrate them into the CRISPR locus in a particular order,” he says. This strongly supported the notion of a link between the spacer content and phage resistance.
Barrangou handled the next experiment, as his lab in Wisconsin was the only lab at Danisco permitted to do this kind of genetic engineering. Horvath was comparing two bacterial strains: DGCC7710 and a daughter (mutant) strain called 7778, which had arisen following a phage challenge performed in 1990. “Suddenly, the black box was opened!” Horvath said. In the resistant daughter strain, Barrangou found two additional spacers in the CRISPR region. “We had the sequence of the phage used in 1990: spacers 1 and 2 were present in the phage. So what did we do? We engineered to prove it was sufficient to provide resistance.” Removing both spacers in the daughter strain resulted in loss of resistance. “Add the two spacers in the parental strain, and without any challenge, it becomes resistant. Bingo!”
Understandably, Barrangou calls this his favorite experiment of all time. “When you swap the two immune systems between two strains, you swap their resistance to some of these sensitivities to viruses. That was essentially the proof that there is a direct link between the CRISPR genotype and the antiviral phenotype.”
The third experiment, Barrangou admits, was a bit lucky. Adjacent to the CRISPR motifs are the CRISPR-associated, or Cas, genes, which encode the nucleases that actually cleave the viral DNA. Inactivating the two biggest Cas genes had a major impact on the CRISPR system: knocking out Cas9 abolished the immune potential, whereas inactivating another gene, Csn2, left the immune potential intact but scrapped the ability of the CRISPR array to acquire new spacers. “This is where sometimes you have to be serendipitous, right?”
The priority for Danisco wasn’t so much to trumpet the results in a major science journal but to patent the CRISPR discovery. Filing the initial patent application on August 26, 2005, gave the company one year to provide additional examples to illustrate the patent. Horvath’s group had to keep quiet and hope nobody scooped them. The inventors were listed as Horvath, Barrangou, their respective bosses, Christophe Fremaux and Dennis Romero, and Patrick Boyaval (the boss of Fremaux and Romero). With the clock ticking, Horvath and Fremaux reached out to Canadian virologist Sylvain Moineau, who was also an expert on S. thermophilus. Moineau was skeptical at first until he reproduced the Danisco team’s results in his own lab. One of Moineau’s postdocs, Hélène Deveau, came to his office one day: “You’re not going to believe this,” she said. DNA analysis showed the CRISPR array had increased in size as the bacteria she was studying gained resistance to phages. “It was just history from there,” Moineau said.22