Three CRISPR papers from the last fortnight, in Cell Stem Cell, Science and Nature Biotechnology. The theme, which nobody planned, is that two of the three do not cut any DNA at all.
One is four patients with inherited blood disorders who have now been off transfusions for over a year. One uses CRISPR as a volume knob on the genes that build ribosomes, and finds that turning it up changes what a stem cell decides to do. And one is a short, slightly embarrassing paper explaining that the standard way of measuring RNA knockdown has been flattering everyone's results.
1. Four patients, one edited letter, no more transfusions
Beta-thalassemia and sickle cell disease are both caused by problems with adult haemoglobin, the protein red blood cells use to carry oxygen. In severe thalassemia there is not enough of it and patients survive on regular blood transfusions, often every few weeks for life. In sickle cell disease the haemoglobin is misshapen, which deforms red cells and causes the blockages that produce the disease's characteristic bouts of severe pain.
The trick this group uses does not touch the broken adult gene. Before birth you make a different version of haemoglobin, called fetal haemoglobin, and your body switches it off in infancy. Switch it back on and it does the job the adult protein cannot. So the edit targets the off switch rather than the gene itself.
They do it with a base editor, which is worth pausing on. Classic CRISPR cuts both strands of DNA and lets the cell repair the break, which is effective and a little uncontrolled. A base editor instead chemically converts one letter of DNA into another without cutting through, so there is no break to repair. Theirs is called a transformer base editor, or tBE.
The team had already reported five transfusion-dependent thalassemia patients in China reaching transfusion independence. The question this paper asks is whether that travels. Different mutations cause these diseases in different populations, and an editor tuned on one genetic background is not guaranteed to work on another.
This is a descriptive report on four more patients from three trials: one African patient with sickle cell disease, and three thalassemia patients carrying mutations common in South and Southeast Asia. After more than 12 months of follow-up:
- All four recovered their blood-forming system and stopped red blood cell transfusions entirely.
- The edit persisted, and fetal haemoglobin stayed high across essentially all of their red cells rather than in a lucky subset.
- The sickle cell patient had no vaso-occlusive episodes, the painful blockages that define the disease.
- No off-target mutations, no cancers and no deaths were observed.
What to watch: four patients, described rather than compared against anything. There is no control group and no randomisation, because that is not what this kind of report is. The authors call the results initial, say they support feasibility, and explicitly ask for broader evaluation across diverse populations. Twelve months is also a good sign and not a lifetime; these are people who are meant to live decades with the edit in them.
2. Turn up a cell's ribosome supply and it behaves differently
Ribosomes are the machines that read messenger RNA and build protein, and a large part of each one is made of ribosomal RNA. How fast a cell transcribes its ribosomal RNA genes goes up and down through development, and when it goes wrong you get cancers and a family of conditions called ribosomopathies.
Whether that rate causes anything has been hard to establish, for a dull structural reason: those genes sit in the genome in many near-identical copies, and they are transcribed so heavily that normal genetic tricks do not give you a clean handle on the level. You can observe the rate changing. You could not set it.
TAPIR is the handle. It uses CRISPR not to cut but to activate, a now-standard variant where the Cas protein is disabled as a nuclease and carries machinery that switches a target gene up instead. Pointed at the ribosomal DNA, it raises transcription of the 47S precursor, which is the long RNA the mature ribosomal RNAs get cut out of. What follows:
- Nucleoli, the dense structures inside the nucleus where ribosomes get assembled, grew larger.
- Protein synthesis went up, including in cells that were already dividing fast and might have been expected to be maxed out.
- In neural stem cells, that extra translation pushed the cells towards self-renewal and proliferation, in culture and in a living animal.
- The same tool was used to model disease phenotypes and partially rescue them.
The claim the authors draw is the interesting part: protein synthesis capacity is not just a consequence of what a stem cell is doing, it is one of the things that decides.
What to watch: the rescue is described as partial, in disease models, and this is a research tool rather than anything near a therapy. Pushing stem cells towards self-renewal and proliferation is also exactly the direction cancer goes, which the paper is clear-eyed about given dysregulated ribosomal RNA transcription is a cancer hallmark to begin with.
3. Your RNA knockdown numbers are better than they should be
Some CRISPR systems target RNA rather than DNA. You are not editing the genome, you are chewing up a transcript to stop it being made into protein, which is reversible and does not leave a permanent change behind. The standard way to check it worked is RT-qPCR: convert the RNA to DNA, amplify it, and see how much less of your target there is than in an untreated sample.
This paper found that the guide RNA, the short sequence that tells the CRISPR protein what to grab, does not wash away during RNA extraction. It comes along into the sample, and once there it inhibits the conversion step for any region that spans or sits upstream of where the guide binds. The measurement is depressed for reasons that have nothing to do with the target being gone.
Which means the knockdown looks better than it was. The authors report this as an overestimation across every CRISPR system they tested, and it is worth noticing how cheap this failure is to run into: the artifact lands in the direction that confirms what you were hoping to see, and nothing about the result looks wrong.
Their fix is practical. Use a processive reverse transcriptase with strong strand-displacing activity, meaning an enzyme that pushes through obstacles on the template rather than stalling at them, and confirm with a second, independent method rather than trusting RT-qPCR alone.
What to watch: this is a paper about a measurement, not about whether RNA-targeting CRISPR works. It says the standard assay overstates knockdown; it does not say any particular published result is wrong, and the abstract does not put a number on how large the inflation typically is. If you run these experiments, the useful response is to check your own reagents and amplicon placement.
The thread
CRISPR spent its first decade being described as scissors, and the two most consequential papers here do not cut anything. The base editor rewrites a single letter and leaves the strands intact. TAPIR takes the cutting machinery, breaks it on purpose, and keeps only the targeting, using the most precise address system in biology to deliver a volume knob. The scissors were never the interesting part. Knowing exactly where to stand was.
The third paper is a reminder of the other thing that decides whether any of this works, which is whether you can trust what your assay tells you. A correction that arrives in a short paper in Nature Biotechnology is a correction that arrived cheaply.