How CRISPR works

Chapter 9 of Walter Isaacson’s The Code Breaker, “Clustered Repeats”, tells how CRISPR was found, and I wanted to see the mechanism rather than only read about it. So I built a model of it. Step through it below, or press play and watch, and drag to turn it.

A strange pattern. In 1987 Yoshizumi Ishino found short, identical stretches of DNA repeated at regular intervals in E. coli, with ordinary-looking stretches, the spacers, in between. Francisco Mojica found the same pattern in salt-loving archaea, and in 2001 he and Ruud Jansen named it CRISPR: clustered regularly interspaced short palindromic repeats. Beside the array sit genes for CRISPR-associated (Cas) enzymes. The repeats are drawn in amber, the spacers each in their own colour.
A virus attacks. A bacteriophage, a virus that preys on bacteria, lands on the cell and injects its DNA to take the cell over. In 2003 Mojica found that the spacers in the array match the DNA of viruses that attack that bacterium: the array is a record of past infections.
Keeping a mug shot. A bacterium that survives keeps a copy. The enzymes Cas1 and Cas2 cut a short piece out of the virus's DNA and insert it at the front of the array as a new spacer, together with a fresh repeat. The newest spacer always sits first, and the record is passed on to every descendant.
Reading the archive. The cell copies the whole array into one long RNA and cuts it at the repeats into short CRISPR RNAs, or crRNAs. Each carries one spacer: one guide for each virus the bacterium remembers. In the Cas9 system a second RNA, the tracrRNA, helps to make the guides and to hold them in place.
Search and cut. Cas9 holds a guide and scans DNA. It first looks for a short tag called the PAM, then checks whether the stretch beside it matches its guide. On a match it opens the double helix, the guide pairs with the DNA, and Cas9 cuts both strands. The virus's DNA is destroyed. The spacers in the bacterium's own array have no PAM beside them, so Cas9 does not cut its own record.
A tool for any DNA. In 2012 Jennifer Doudna and Emmanuelle Charpentier showed that Cas9 can be programmed: fuse the two RNAs into a single guide, write the sequence you want into it, and Cas9 cuts DNA wherever that sequence occurs. The cell then repairs the break, and the repair is where editing happens: the gene can be switched off, or, given a template, rewritten.

Drag to turn the model. Not to scale: real repeats and spacers are a few dozen letters long, and here one repeat and one spacer make one turn of the helix. Drawn with three.js.

The history is the best part of the chapter. In 1987 a Japanese doctoral student, Yoshizumi Ishino, noticed an odd pattern while sequencing a gene in E. coli: short identical stretches of DNA repeated at regular intervals, with different stretches in between. He mentioned it in the last lines of his paper, said nobody knew what it was for, and moved on. A few years later Francisco Mojica, a graduate student in Alicante studying microbes that live in salt ponds, found the same pattern, first assumed he had made a mistake, and then spent a decade trying to explain it with little funding and much advice to stop.

The answer came in August 2003, when he ran the spacers through a database and found that they matched the DNA of viruses that attack those bacteria. Bacteria, it turned out, have an immune system that remembers. Journals were slow to believe it; the paper took until 2005 to appear. Two food scientists at the Danish company Danisco, Rodolphe Barrangou and Philippe Horvath, confirmed it in their collection of yogurt bacteria: strains collected after a virus attack carried new spacers that matched that virus. And in 2012 Jennifer Doudna and Emmanuelle Charpentier showed how the cutting works and that it can be programmed, which turned a bacterial defence into a tool for editing any DNA.

What I find most elegant is how little it takes. A library of short snippets, a way to copy them into guides, and one enzyme that compares a guide with whatever DNA it meets. The same pattern of matching a stored sample against what arrives is how a lot of security systems work, and it is a fine example of curiosity paying off: a question about salt-loving microbes that turned, years later, into one of the most useful tools in biology.

References

Walter Isaacson, The Code Breaker (2021), chapter 9, “Clustered Repeats”, chapter 12, “The Yogurt Makers”, and chapter 17, “CRISPR-Cas9”.
Francisco J. M. Mojica et al., “Intervening sequences of regularly spaced prokaryotic repeats derive from foreign genetic elements”, Journal of Molecular Evolution 60, 2005.
Martin Jinek, Krzysztof Chylinski, Ines Fonfara, Michael Hauer, Jennifer A. Doudna, and Emmanuelle Charpentier, A programmable dual-RNA-guided DNA endonuclease in adaptive bacterial immunity, Science 337, 816 to 821, 2012.