A label nobody checks
© BMC Microbiology, 2026
A new paper from the lab has just appeared in BMC Microbiology, and like most of the work we end up writing about here, it did not begin as a paper at all. It began as a control – one of those experiments you set up expecting nothing, precisely so you can point at it later and say "see, nothing." Instead it filled the microscope with something we did not expect, so it was worth writing it up.
The project that spawned this one belongs to Professor Ronan McCarthy, now at the University of Southampton, who drives our work on artificial sweeteners and bacterial physiology – the genuinely medical-microbiological end of things. And I should admit something else: I have never actually run a replication run-out experiment myself. My interest in it is that of a reader – someone who works on how bacteria copy their DNA, and for whom cephalexin arrives as a name in other people's methods sections, described, almost invariably, as a "cell-division inhibitor." I took that three-word description on trust for years. This paper is, in part, about what that description was quietly hiding.
An experiment that was only ever meant to be a control
The starting point was saccharin. In a recent study led by Ronan's group, we had found that the sweetener causes E. coli cells to stop dividing and grow into filaments. A reasonable question follows: is saccharin doing something specific, or does anything that stops cells dividing produce the same picture? To find out, you need a known cell-division inhibitor to compare against.
E. coli cells filament rapidly after treatment with the antibiotic cephalexin and show a large number of active replication complexes, shown here in the YFP fluorescence channel.
© RudolphLAB, 2026
Cephalexin is the obvious choice. It has been used for decades to block bacterial cell division, and Emma Dunbar – at the time an international summer student in my lab – added it as a quick comparison. I did what one does under time pressure: I took the first paper I could find that used it, saw a concentration of 36 µg/ml, rounded it up to a tidy 40, and we ran the experiment.
The cells filamented violently – far more dramatically than anything we had seen with saccharin. That was the first sign that "just a cell-division inhibitor" was not going to be the whole story.
What I had failed to appreciate
Here is the naïve picture I had been carrying around, more or less without examining it: cephalexin stops cells dividing. Full stop. A clean switch.
That picture is not wrong so much as dangerously incomplete. Cephalexin is a β-lactam antibiotic – a relative of penicillin and ampicillin. Its main job is, after all, to kill bacteria. It does this by jamming the machinery that builds the cell wall. One consequence of that jamming is that cells cannot form the septum needed to divide, so they filament – which is the effect everyone reaches for it for. But the filamentation is a side effect of a drug whose main event is cell death. Treat cells with cephalexin on its own and, given a little time, they bulge and, eventually, burst.
I had been thinking of it as a "stop dividing" button. It is really a slow-acting poison that happens to stop division on its way to killing the cell. That distinction turned out to matter a great deal, because of what cephalexin is used for in my part of biology.
And here is the thing I keep coming back to. The reason I never examined any of this is that I never had to: the shorthand did the thinking for me. "Cell-division inhibitor" is how cephalexin is introduced in many papers, and a three-word label that everyone uses is a label nobody checks. I inherited it and passed it along without ever looking underneath. I doubt I am the only one. When a description becomes standard enough, it stops being a description and starts being a substitute for looking – and that is precisely the condition under which a method's quiet assumptions go unquestioned for decades.
A short detour: what "replication run-out" is
To measure how bacteria manage DNA replication, we often want to count how many rounds of copying a cell has going at once. Fast-growing bacteria are extravagant: they start new rounds of DNA replication before the previous ones have finished, so a single cell can be carrying several overlapping copies of its chromosome at the same time.
Counting these is the trick. The "replication run-out" method, established in the 1980s, is an elegant way to do it. You add two drugs. The first, rifampicin, stops cells from starting any new rounds of replication, while letting the rounds already underway run to completion – hence "run-out." The second, cephalexin, stops the cells dividing, so that all those finished chromosomes stay conveniently packaged inside one long filament where you can count them.
The whole method rests on a quiet assumption: that cephalexin's only relevant contribution is to stop division. It is supposed to be an inert bystander that simply holds the cell still while rifampicin does the interesting work. Our accidental result was starting to suggest that the bystander was not so inert.
What cephalexin was actually doing
When we looked properly – and this is where Iren Grigoryan took the project on and, over a long stretch of careful work, largely built the paper – a consistent picture emerged. Iren repeated Emma's original experiments to make sure we had the necessary repeats confirming the results and then did the great bulk of everything that followed: the lower drug concentrations that most labs actually use, the crucial rifampicin experiments, and the semi-automated measurements of cell length that let us put numbers on what we were seeing. She did not do it in isolation – Dominika Krawiel, a shared PhD student in the lab at the time, contributed experiments and kept things running at the bench when I could not be there – but the shape of the paper is, in large part, Iren's.
Cephalexin on its own did not merely stop cells dividing. It triggered a marked increase in DNA replication activity. Using cells in which the replication machinery is tagged with a fluorescent marker, we could see the number of active replication sites climb well beyond what simply halting division could explain. Other, independent ways of measuring the same thing agreed. The cells were not just failing to divide; they were copying their DNA more than they should.
We chased the cause, and the answer was reassuringly specific: the extra activity depends on DnaA, the protein that fires the "start" signal for a new round of replication. In other words, cephalexin was somehow provoking cells into initiating replication more often. Our best guess is that it is an indirect consequence of the cells becoming so enormous – size and the decision to start replication are known to be linked – but we are honest in the paper that we have not pinned down the mechanism, and did not set out to.
The part that matters for everyone else
At this point we had what looked like an uncomfortable conclusion. If cephalexin, on its own, distorts the very thing run-out experiments are trying to measure, then a method the field has relied on for forty years might be in trouble.
So we tested it directly, the way the method is actually used – cephalexin and rifampicin together, rather than cephalexin alone. And here the story deflates, in the best way. With rifampicin present, the excess replication is almost entirely suppressed. The cells also filament less and burst less. The two drugs, it turns out, partly antagonise one another, and rifampicin happens to shut down exactly the runaway initiation that cephalexin provokes.
That is genuinely good news for the decades of published run-out data: the standard method largely holds. But "largely" is carrying weight in that sentence. A small but real fraction of cells – roughly one in ten in our hands – still behaves oddly even under the full run-out conditions. Not enough to overturn old bulk measurements, where the misbehaving minority is swamped by the well-behaved majority. But potentially quite enough to matter in the kinds of experiments the field is increasingly moving towards: single-cell measurements, sensitive mutant backgrounds, high-resolution microscopy, where a strange 10% does not average away.
The practical message is not "throw out run-out." It is more measured than that: know what your tools are actually made of. Cephalexin is not a neutral spacer; it is an antibiotic with opinions, and in the right circumstances those opinions leave a mark on your data.
Where the paper came from
A word on how this came to exist, because it is a slightly odd origin story.
We had, in truth, already suspected cephalexin was a poor control long before we wrote any of this down – the phenotype was simply too strong to trust. The push to turn it into a paper came, fittingly, from peer review: a reviewer of our earlier replication termination work suggested we use cephalexin as a control for a different question. We already knew it would be a bad control – and decided that this, precisely, was the point worth making. The data that disqualified it from one job qualified it for a small paper of its own.
It also served, along the way, as a test case for some of the AI-assisted reviewing tools I have written about elsewhere.
This paper is the work of three people in particular. Iren Grigoryan, an outstanding undergraduate who joined the lab in her first year and has been with us ever since, is its first author, and deservedly so: she did the overwhelming majority of the work that turned an odd observation into a finished study – in the lab, and in the fiddly, unglamorous image analysis that so much of modern cell biology quietly depends on. Dominika Krawiel, a shared PhD student who has since moved with Ronan to Southampton, contributed important experiments of her own and supervised Iren at the bench during the stretches when I was unavailable – exactly the sort of steadying, day-to-day mentorship that keeps a project moving and rarely gets named in print. And Emma Dunbar, now a PhD student in California, lit the fuse in the first place, as a summer student, with the quick test that refused to behave. It is the kind of contribution – across all three of them – that seldom shows on the polished surface of a published paper, and it is worth naming plainly.
The paper's journey to print was, let us say, eventful – a story for a separate post. It was rejected once, not unreasonably, for lacking the mechanistic depth we had never intended to provide, before finding its current home. Research papers present clean, finished stories, but the story of their creation is rarely as clean.
At any rate, the genuinely good news is that run-out survives the scrutiny largely intact – but "largely" is a word worth taking seriously.
The full paper can be read here.
Similar and related posts: