CRISPR and gene editing

What CRISPR actually does

CRISPR is borrowed technology. Bacteria store short fragments of viral DNA in repeating arrays — clustered regularly interspaced short palindromic repeats — as a memory of past infections, then transcribe them into RNA that guides an enzyme to destroy the same virus if it returns. Emmanuelle Charpentier and Jennifer Doudna shared the 2020 Nobel Prize in Chemistry for showing that this bacterial system could be reprogrammed to cut any chosen sequence of DNA.

The tool has two parts. A guide RNA of roughly twenty bases carries the address, base-pairing with the matching stretch of the genome. The Cas9 protein is the scissors: once the guide has found its match, and provided a short motif called a PAM sits immediately alongside, Cas9 cuts both strands. Changing the target means changing twenty bases of RNA, which is why the technique spread so fast. It is programmable and it is cheap.

What happens after the cut is done by the cell, not by you, and that is the half candidates skip. A double-strand break is dangerous, so the cell repairs it at once. Non-homologous end joining pushes the ends back together and often loses or gains a few bases, scrambling the reading frame and switching the gene off — useful when the aim is to disable something. Homology-directed repair copies from a supplied template, so it can rewrite a gene precisely, but it is far less efficient and works mainly in dividing cells.

1. Design a guide RNA complementary to the target, and screen it computationally against the rest of the genome for near matches.
2. Deliver the guide and the Cas protein into the cell — by viral vector, by lipid nanoparticle, or by electroporating cells that have been taken out of the body.
3. Cas9 binds, checks the match against the guide, and cuts.
4. The cell repairs the break: imprecisely by end joining, or precisely from a template if one has been supplied and the pathway is available.
5. Sequence the edited cells to confirm what actually happened — the intended change, any unintended changes elsewhere, and whether every cell was edited or only some.

Cutting both strands is the crude part, and the two newer refinements avoid it. Base editors attach a chemical-modifying enzyme to a Cas protein disabled so that it nicks one strand rather than cutting through, converting one DNA letter directly into another — C to T, or A to G — without breaking the double helix. Prime editors attach a reverse transcriptase and use an extended guide carrying the new sequence with it, writing in any substitution plus small insertions and deletions. Both sit further behind conventional CRISPR in the clinic, and both address what makes cutting risky.

Somatic and germline: the line everything turns on

Every serious question in this topic sorts into one of two piles, and an answer that fails to sort them will drift. Somatic editing changes cells in a person who already exists — blood stem cells, T cells, liver cells — and the change lives and dies with that patient. Germline editing changes an embryo, an egg or a sperm, so the alteration sits in every cell of the resulting person, including the cells that will make their own eggs or sperm. It descends to their children and to everyone after them.

Somatic editing is ethically continuous with the rest of medicine, and saying so is not a dodge. A competent adult consents, the risks and benefits fall on the person who agreed to them, a regulator assesses the evidence as it would for any medicine, and if it goes wrong the harm is contained. It raises real questions about cost and long-term unknowns, but no new question of principle: it is beneficence and non-maleficence weighed by the patient, exactly the machinery in the four pillars.

Germline editing breaks that structure at the point that matters most. The person who bears the consequences does not exist yet and cannot be asked, which is not something consent and capacity can solve, because there is nobody there to have capacity. The effects are not contained: an error propagates down a family line indefinitely and, at scale, into the population. And there is no way to withdraw a treatment from a person who has been made out of it.

The first approved therapy is already in the NHS

This stopped being hypothetical in November 2023, when the UK regulator, the MHRA, authorised the first CRISPR-based therapy anywhere in the world: Casgevy, or exagamglogene autotemcel, for sickle cell disease and for transfusion-dependent beta-thalassaemia in patients aged twelve and over. Regulators in the United States and the European Union followed within months. Britain was first, which is worth knowing precisely because most candidates cannot say it.

The mechanism is elegant and you should be able to describe it. Blood stem cells are collected from the patient, edited outside the body to disable a control element that normally switches off fetal haemoglobin production shortly after birth, and infused back. The edited cells make fetal haemoglobin again, which does not sickle and which substitutes for the faulty adult haemoglobin. Notice what that means: the therapy does not repair the sickle cell mutation at all. It reactivates a healthy backup the body already owns.

It is not a simple injection, and candidates who describe it as one get pushed. Before the edited cells go back, the patient needs conditioning chemotherapy to clear the existing bone marrow, which brings infection risk, weeks in hospital and a real risk of infertility, so fertility preservation forms part of the conversation. It is given at a small number of specialist centres, and follow-up so far runs to years rather than decades, which matters when the claim being made is durability.

Cost is the live issue. Reported international list prices run to millions per patient, paid once for what may be a lifetime of benefit — a very different shape from a drug taken daily. NHS access in England was agreed through managed access arrangements with confidential pricing: NICE recommended it for transfusion-dependent beta-thalassaemia in 2024, and for severe sickle cell disease in 2025. Check the current position before quoting anything, and notice the justice question underneath: sickle cell disease in the UK predominantly affects Black African and Caribbean communities and has a documented history of under-recognition, which puts this inside health inequalities.

Somatic editing is wider than one product. Edited immune cells have been used in leukaemia, including a UK case in 2022 in which a child received base-edited donor T cells after other options had run out, and in 2025 a US team reported treating an infant who had a rare urea cycle disorder with a base-editing therapy built for that child’s individual variant. That last case previews a new regulatory problem: a trial population of one cannot be studied the way medicines usually are.

He Jiankui: the case every interviewer knows

In November 2018 the Chinese scientist He Jiankui announced that twin girls had been born from embryos he had edited, days before an international summit on human genome editing opened in Hong Kong, where he then defended the work in person. He had targeted CCR5, which encodes a receptor HIV uses to enter cells, intending to make the children resistant to infection. The reaction from the scientific community was close to unanimous condemnation, including from researchers who support heritable editing in principle. He was convicted in China in 2019 of practising medicine illegally and served a three-year prison sentence.

Know why it was condemned, because the reasons are the syllabus.

• There was no unmet medical need. The father was HIV positive and the mother was not, and established techniques already prevent transmission to a child, so the editing addressed a problem with a safe existing solution.
• The consent process was inadequate. The documentation given to the couples was dense and framed around an AIDS vaccine programme, and it is doubtful the parents understood they were enrolling children in a first-in-human genetic experiment.
• The edits were not what he intended. Rather than reproducing the naturally occurring protective variant, the procedure produced novel changes of unknown consequence, and there was evidence that not every cell carried the same edit.
• Disabling CCR5 is not free of risk. The receptor has roles in immune defence, and there is evidence that losing it may worsen the course of some other infections.
• Oversight was evaded rather than obtained. The work was concealed from his institution and from clinical staff involved in the pregnancies, which is why no regulator stopped it.

The case did two useful things. It made an abstract argument concrete, and it produced a broad international consensus that heritable human genome editing should not proceed at this time — a position restated at the third international summit on human genome editing, held in London in 2023 — while laboratory research continues under regulation. What it did not produce is enforceable global governance, because there is none. The constraints are national law, professional norms, and the certainty that a scientist who tries it will lose their career.

The ethics that actually earn marks

Start with consent, because it has the most structure. Every other intervention in medicine is justified either by the patient’s own agreement or, where they cannot agree, by a best-interests judgement made for a particular existing person. A germline edit has neither, because the person is partly constituted by the decision itself. Some philosophers argue this makes the consent objection incoherent rather than decisive, since nobody consents to their own conception. The honest position is that the objection is really about the scale and irreversibility of what is decided for someone, not about consent in its ordinary clinical form.

Safety sits underneath all of it. In a germline context you cannot pilot the intervention, because the first evidence of harm is a person. That asymmetry is why even scientists who expect heritable editing to become technically safe argue that it is not yet, and why the strongest arguments for it are conditional on data that does not exist.

The therapy and enhancement boundary is where an answer either sounds thought through or sounds rehearsed. The extremes are easy: correcting a variant that causes a fatal childhood disease is treatment, editing for height is not. The middle is genuinely unclear. A variant that halves lifetime cholesterol prevents disease that has not happened, and vaccination enhances an immunity nobody was born with without anybody finding it sinister. A defensible reformulation is that the line falls between restoring a function most people have and reaching beyond it — then admit that this still does not settle where inherited resistance to infection belongs.

The argument candidates most often miss is the disability rights critique, and interviewers notice when you can voice it properly. Framing a condition as a defect to be edited out expresses a judgement about the value of lives currently lived with it — the expressivist objection — and many disabled people report receiving exactly that message from screening and selection programmes. The counter-argument is that treating a condition is not a verdict on the people who have it, and that many people living with painful or progressive disease want it prevented. The strongest answer holds both: a condition can be worth preventing while the people living with it are owed full respect and full accommodation, and a society funding the first while neglecting the second is what the critique is really attacking.

Equity is the objection that scales fastest. A one-off treatment priced in the millions does not distribute itself evenly, and globally the countries carrying the heaviest burden of sickle cell disease are the least able to buy the therapy that treats it. If heritable editing ever arrived, it would arrive first through private fertility clinics in wealthy countries, which is a mechanism for converting social inequality into biological inequality.

Finally, hold on to the fact that some of this already happens by another route. Preimplantation genetic testing creates embryos by IVF, tests them for a condition known in the family, and transfers one that is unaffected — the same end point germline editing would deliver, reached by selection rather than modification, and licensed in the UK for a published list of serious conditions. That does not make editing permissible. It does weaken the claim that the goal is unnatural or unprecedented, and it presses the objector to say what modification adds that selection has not already conceded.

What UK law permits, and how to say it in the room

The UK position is unusually clear, and quoting it accurately is an easy way to sound informed. The Human Fertilisation and Embryology Act 1990, as amended in 2008, allows only a permitted embryo to be placed in a woman, and an embryo whose nuclear or mitochondrial DNA has been altered is not a permitted embryo. Germline modification for reproduction is therefore prohibited outright — not merely unfunded or discouraged, prohibited — subject only to the narrow exception described below.

Research is a separate matter. The Human Fertilisation and Embryology Authority licenses research on human embryos, editing included, for defined purposes, and no embryo may be kept beyond fourteen days of development or the appearance of the primitive streak, whichever comes first. A UK laboratory received the first licence to use CRISPR on human embryos for research in 2016, studying the genes that govern the earliest days of development. Those embryos are never transferred to a woman. Some scientific bodies have argued for reviewing the fourteen-day limit case by case; UK law has not changed at the time of writing.

One qualification stops your answer being too neat. The UK legalised mitochondrial donation in 2015, through regulations made under the same Act, and that is a heritable genetic change: nuclear DNA comes from the intending mother while the mitochondria, which carry their own small genome, come from a donor, and daughters born this way pass the donated mitochondria on. It is tightly regulated, licensed case by case, restricted to a significant risk of serious mitochondrial disease, and in 2025 the Newcastle team reported the first published series of births under the UK programme. Say that the UK prohibits heritable modification of nuclear DNA and permits one narrow mitochondrial exception, and you will be right where most candidates overstate.

None of this arrives as a lecture request. You will be asked something short — should we edit human embryos, is gene editing playing God, what did you make of the Chinese twins — and marked on whether the answer has a spine. The spine is the somatic and germline distinction, with an approved NHS therapy on one side and an international prohibition and one notorious case on the other. Practise naming the strongest objection to your own position before the interviewer does, the habit answering a dilemma live is built around, and if you reached this article without the sequencing and screening material, read genomic medicine and screening first.