The FLAG tag added so coronin could be seen left the mouse looking like a knockout

The reagent inside the result

The Basel group's first move was unglamorous and decisive: before measuring anything, they made coronin 2 and coronin 3 knockout HeLa lines with CRISPR and tested their antibodies on cells that could not contain the antigen. Most of the antibodies failed, including ones the literature cites for exactly this purpose. Only after finding 2 that distinguished wild-type from knockout cells did they compare where the untagged protein sits against where a GFP fusion sits. Median Pearson correlation coefficients with actin rose for coronin 1, coronin 2 and coronin 3 once the tag was attached, at p below 0.0001 in each case.^[1]

The sharper result is in mice. The authors knocked an 8-amino-acid FLAG sequence into the endogenous coronin 1 locus. Those mice reproduced the T-cell numbers of coronin 1 deletion in lymph nodes, spleen and thymus. Eight amino acids is a small addition, and the lesson the paper draws from it is that a tagged protein's functionality has to be tested case by case rather than assumed.^[1]

What the design can separate

Read as a design question, the problem is familiar. An experiment that only ever images the tagged construct produces one picture, and two different states of the world produce it: coronin genuinely sitting on actin, or the tag putting it there. Nothing inside that experiment separates them. This column put the same argument to a single spacecraft in the radiation belts on 7 August, where diffusion and ordered drift leave the same observational signature and one vehicle cannot tell which it saw. The second measurement is the comparison against the untagged endogenous protein, and once it is run the two explanations come apart. Other readings remain open: the authors say a tag could relocate coronin without exposing any cryptic actin-binding site, listing post-translational modification, oligomerisation state and membrane association, and they note that GFP is known to bind actin sites on myosin.^{[1], [2]}

The paper also counts how the field arrived here, and that section deserves more caution than the experiments. Most coronin publications, the authors report, do not test the actin interaction at all; they cite earlier work that did. Among non-mammalian studies that do conclude an interaction, most used the yeast protein Crn1, which carries domains the non-fungal coronins lack, is produced in E. coli and is mixed with rabbit muscle actin. Crn1 also has a higher reported affinity for rabbit actin than for yeast actin. That is a description of how a literature is built; it narrows the evidence base without overturning any individual result.^[1]

What survives the reassessment

The positive findings are narrower than the negative ones and easier to state. Removing coronin 2 and coronin 3 from HeLa cells left actin distribution and repolymerisation after latrunculin washout unchanged, and proteomic and transcriptomic profiling of the same cells found no shift in the actin-interacting proteins the coronin literature had implicated. Power was at least 80 per cent to catch a change above 2-fold at an adjusted p-value below 0.01. What those cells did lose was forskolin-stimulated cAMP production, with or without intact filaments. The authors state plainly that this does not exclude a role for coronins in actin depolymerisation; their claim is that here the proteins' localisation and the cells' actin dynamics did not depend on each other.^[1]

The result that should be tested next is the tag effect itself, the one finding here with reach beyond coronin. A second laboratory would need to do what this one did: validate the antibody on knockout cells, image the endogenous protein and the tagged construct in the same cell type, and report both. Two conditions would make that a real test: a primary cell rather than a transformed line, and a tag other than GFP or FLAG. If such a comparison is published by 31 December 2027 and the untagged protein again sits further from actin than its tagged version, the reassessment holds and the tag belongs in the methods discussion of every localisation claim. If the gap closes in primary cells, the effect is a property of these lines and this construct.^[1]