What Is CRISPR? A Bacterial Immune System Turned Into an Editing Tool

The acronym stands for clustered regularly interspaced short palindromic repeats, which describes an odd pattern first noticed in bacterial genomes in 1987 and not explained for nearly twenty years: short repeated sequences separated by unique spacers. In 2005 several groups realised the spacers matched viral DNA, and in 2007 a team working on yoghurt bacteria demonstrated experimentally that this was an adaptive immune system. When a virus infects, the bacterium stores a fragment of its DNA as a new spacer; it then transcribes these spacers into short RNA molecules, each of which guides an associated protein to any matching DNA and directs it to cut. The system is inheritable, specific and updatable, which makes it genuinely analogous to immunity, and it was studied for years as an interesting piece of microbiology before anyone proposed using it.

The step that mattered was showing that the guide and the cutter could be separated from their biological context and reprogrammed. The standard system has two components and a repair step:

• •A guide RNA about twenty bases long, synthesised to match whatever sequence is being targeted, which is what makes the system programmable rather than fixed
• •The Cas9 protein, which carries the guide, scans the genome, and cuts both strands of DNA where the guide matches, provided a short adjacent motif is present
• •The cell's own repair machinery then acts, and which pathway it uses determines the result
• •Non-homologous end joining rejoins the cut sloppily, usually inserting or deleting a few bases, which disables the gene and is the easiest edit to achieve
• •Homology-directed repair, if a template DNA is supplied, copies from that template, allowing a specific sequence to be written in, though it works only in dividing cells and at much lower efficiency
• •Newer variants avoid cutting entirely: base editors chemically convert one letter to another, and prime editors write short specified sequences using a reverse transcriptase attached to a disabled Cas9

Targeted gene editing existed before, using zinc finger nucleases and later TALENs, and both worked by engineering a protein to recognise a specific DNA sequence. Designing such a protein was slow, expensive and required specialist expertise, so a single experiment could cost tens of thousands of dollars and take months. The new system changed the targeting element from a protein to a piece of RNA, and RNA of a specified sequence can be ordered for a few tens of dollars and arrives in days. The effect was that a technique confined to well-funded specialist laboratories became available to any molecular biology laboratory in the world, and adoption was correspondingly rapid: within three years of the 2012 paper it was in routine use across biology, and Emmanuelle Charpentier and Jennifer Doudna received the Nobel Prize in Chemistry in 2020, with a long and bitter patent dispute running alongside over contributions from Feng Zhang's laboratory and others.

Most use is in research, where the ability to knock out a gene and see what changes is the basic method of working out what genes do, and the technique has accelerated that work enormously. Medically, the first approved therapy arrived in late 2023: a treatment for sickle cell disease and beta thalassemia that removes a patient's blood stem cells, edits a regulatory gene so that the cells resume making foetal haemoglobin, and returns them, which addresses the disease without correcting the original mutation. Trials are running for inherited blindness, certain cancers via edited immune cells, and a form of amyloidosis treated by editing inside the body rather than in a dish. In agriculture the technique produces crops with edits that could in principle have arisen by breeding, which several jurisdictions regulate more lightly than transgenic organisms. Other applications include gene drives, engineered to spread a trait through a wild population faster than inheritance would allow, proposed for malarial mosquitoes and deeply contentious because releasing one is difficult to reverse.

The technique is not perfectly precise. Off-target cuts at sequences resembling the guide occur, and although improved variants and better guide design have reduced them substantially, they must be checked for in any therapeutic use. Large unintended deletions and rearrangements at the target site have also been reported. Delivery is the harder practical problem, since editing cells removed from the body is manageable and reaching a specific tissue inside a living person is not, and the current therapies work largely because blood stem cells and the liver are accessible. The ethical boundary the field drew is between somatic editing, which affects one patient, and germline editing, which changes eggs, sperm or embryos and is inherited by all descendants. That boundary was crossed in 2018 by He Jiankui, who edited embryos and produced live births, an act condemned across the scientific community for its recklessness and its lack of medical justification, and for which he was imprisoned in China. Germline editing remains prohibited in most countries, and the international consensus is that it should not proceed until safety and governance questions that nobody has answered are settled.

CRISPR is a bacterial immune system that stores fragments of viral DNA and uses matching RNA guides to direct a protein to cut any sequence that matches. Redirecting it with a synthetic guide, demonstrated in 2012, made genome editing programmable by ordering a cheap piece of RNA rather than engineering a custom protein, which put the technique in every laboratory. It underlies an approved sickle cell therapy from 2023, and its limits are off-target cuts and the difficulty of delivering it inside a living body, with inherited germline editing prohibited in most countries.