Methods sections in dozens of cellular senescence studies may have listed antibodies against a bacterial protein instead of the mammalian protein

Methods sections in dozens of cellular senescence studies may have listed antibodies against a bacterial protein instead of the mammalian protein

On August 21, Nature reported on an investigation by independent molecular biologist Sholto David. The methods sections of at least 54 papers state that the researchers tested for mammalian β-galactosidase, but list antibodies against the version of this enzyme found in the bacterium E. coli. The editors of Cell are investigating one case, and Springer Nature said that it would assess the concerns.

Senescence is a state in which a cell stops dividing but remains alive. Researchers often detect it by measuring the activity of β-galactosidase, an enzyme that accelerates a chemical reaction. The X-Gal stain produces a blue signal after reacting with either mammalian β-galactosidase or its bacterial counterpart.

An antibody serves a different purpose. It must bind to a specific protein. Although X-Gal reacts with both versions of the enzyme, an antibody recognizes the mammalian and bacterial proteins as different targets. The reagent catalog number can be used to verify which protein the antibody is intended to recognize.

In a June analysis of confusion between p16, a marker of cell cycle arrest, and a cytoskeletal protein with a similar name, catalog numbers helped distinguish one molecule from the other. In the current investigation, the error arises from the enzymes sharing a name and from the use of the same X-Gal assay for both.

The Abcam product page for ab9361 explicitly states that the antibody is reactive with E. coli. David compared the reagent numbers reported in the methods sections with manufacturers' catalogs and compiled 54 cases in his analysis. A catalog number makes it possible to trace the chain from a measurement to the protein that the antibody was meant to target.

One case concerns a Cell paper published in 2016 on partial reprogramming in mice used as a model of premature aging. Its reagent list includes Millipore AB986, while figure 5E reports a reduction in the liver β-galactosidase signal after cyclic activation of reprogramming factors. The editors of Cell are investigating David's concern about this specific case.

“If papers list the wrong antibodies, such predictions may follow the wrong lead,” biochemist Aled Edwards told Nature.

Edwards was referring to AI models that read papers and propose protein targets for treatment. If a study's methods associate an experimental result with the wrong protein, that error can enter the data from which these models select targets.