After Gene Editing, Do Corrected Cells Gain or Lose Ground Over Time?

Agene therapy trial reports one proud number: the share of cells it corrected. Our research desk argues that number is a snapshot taken once, and that in a living tissue the corrected share keeps moving afterward. The tools to track it already exist, borrowed from the study of small human populations.

Demographers who study small populations do not count heads. They count the variance in who leaves descendants, because a few lineages supply most of the next generation. Two forces drive that spread. One is luck, called drift. The other is a real advantage, called selection. Drift scatters; selection steers.

Gene therapy in living tissue makes a mosaic. A virus or lipid particle edits some liver cells and not others. The field reports the percent of genes corrected, measured early. The desk's claim: in a renewing tissue, that corrected fraction is a lineageshare variable subject to both drift and selection, and should be modeled and measured with the demographer's tools.

Where the correction helps a cell grow or survive, the share should move the same direction across animals. Where it does nothing to survival, individual animals should scatter around the starting value with no shared trend. That contrast is the whole experiment.

The demographic bookkeeping is standard population genetics. Effective population size depends on the spread of reproductive contributions, not the raw count, a result settled since the mid-twentieth century. The two forces have distinct signatures: drift is inconsistent across replicates, selection is consistent.

The biology the desk leans on is also real. Clonal hematopoiesis, where one blood stem-cell lineage expands over decades, is well documented and clinically important. A 2017 New England Journal of Medicine study by Siddhartha Jaiswal and colleagues found that such clones nearly doubled coronary heart disease risk. That is lineage dynamics in a human tissue with health stakes. It also shows that expansion is not automatically good news: clonal dominance has a documented harm side, including insertional-oncogenesis and clonal-dominance events in haematopoietic gene-therapy trials. A rising corrected fraction is not by itself a benefit.

The liver example is strong, and it means the hypothesis is less novel there than it may sound. In hereditary tyrosinemia, cells that regain the FAH enzyme outgrow the uncorrected ones. Work by Hao Yin, Wen Xue and colleagues showed this twice: a 2014 Nature Biotechnology study corrected about 1 in 250 liver cells with CRISPR, yet the corrected cells expanded to roughly a third of hepatocytes over about 30 days and rescued the animals. A 2016 follow-up raised initial correction above 6% with a combined delivery approach. In both, the early number understated the final one, because corrected cells had a survival edge.

Take mice treated with a liver base editor correcting FAH deficiency, where corrected cells are known to win. Restrict to animals whose corrected fraction one month after dosing sits between 5% and 40%. By twelve months, the desk predicts the median corrected fraction will be at least 1.5 times the one-month median. In a matched comparator editing a liver target with no survival effect, the twelve-month median should stay between 0.8 and 1.2 times the start.

The desk gives the paired prediction a 60% chance. That figure is a subjective credence, not a calculation. Tyrosinemia is the positive control, not the discovery, and its growth advantage is extreme, far larger than anything plausible for ordinary metabolic targets. The interesting question is whether a smaller version of the same drift shows up in ordinary metabolic targets nobody examined.

Gene and base editing are moving from rare diseases toward common ones, and in vivo base editing has already reached humans, including lipid-lowering programmes aimed at targets such as PCSK9 and a bespoke treatment for a ureacycle disorder. That is where reporting the corrected fraction over time would be actionable. If the fraction reliably rises when the fix helps a cell survive, a modest early edit could grow into a durable one. If it fades where the fix is neutral, durability becomes a design target, not an afterthought. Either way, trials should measure the trajectory, not just the first reading.

• Liver is special. It regrows readily. A clean result there may not carry to heart, muscle or brain, where cells divide little and the framework has least to say. • Prediction is hard. Knowing in advance whether a correction helps a cell survive is genuinely difficult, which is why target arms must be named before dosing. • Abundance is not benefit. More corrected cells is not the same as a patient getting better. That needs functional endpoints this prediction leaves out.

• Reanalysis of published mouse liver editing time courses for drift in the corrected fraction. • A prespecified two-arm mouse study measuring bulk editing and single-cell clonal structure at 1, 6 and 12 months. • Whether gene-therapy trials begin reporting the corrected fraction over time, not just at baseline.

Losing Experts All at Once May Destroy Skills That Gradual Turnover Would Not Did Selection Start Before Life? A Testable Idea About Chemistry That Remembers