Ki67, EdU, or BrdU: Choosing a Proliferation Readout for Flow Cytometry

The grant figure needs a proliferation number by Friday, the cells were fixed three weeks ago, and nobody pre-labelled anything. That single fact — whether the sample was labelled before the culture started — decides more about which proliferation readout you can use than any comparison table will. This post walks the three antibody- and nucleoside-based readouts (Ki67, EdU, BrdU) in the order a practitioner actually eliminates them, and names the panel and cell-cycle consequences of each choice.

Ki67, EdU or BrdU: the three questions that decide it

Ki67, EdU and BrdU answer different questions about proliferation, and the differences are not a matter of sensitivity or convenience. Work through these in order.

• Could you label the cells at the start of the culture? If no — archived, fixed, or clinical samples — only Ki67 is available. EdU and BrdU both require the nucleoside to be present while cells are in S phase.
• Do you need to know when cells divided, or only whether they are cycling now? A nucleoside pulse marks the cells that were replicating DNA during the window you defined. Ki67 gives a snapshot of current cycling status with no temporal resolution.
• What else is in the panel? This is the question that gets skipped and then costs a repeat experiment. The detection chemistry of each readout constrains which fluorochromes survive to acquisition.

Ki67 — a snapshot, and a graded one

Ki67 is present through G1, S, G2 and M and falls away in cells that have left the cycle, which is what makes it the standard readout for the growth fraction of a population. It is also the only one of the three that separates quiescent G0 cells from cycling G1 cells — DNA content alone cannot, because both are 2C. If your question is “what fraction of this population is out of G0”, Ki67 answers it in a fixed sample with a single intracellular stain.

Ki67 also requires nuclear access, so the fix/perm choice is part of the decision. The same trade-off that governs intracellular cytokine staining and its gating controls applies here: the permeabilization that gets antibody to the target is also what degrades some surface epitopes and some fluorochromes.

EdU — temporal precision without DNA denaturation

EdU and BrdU both report S phase during a defined pulse. The practical difference is detection. BrdU is detected by antibody, and the antibody cannot reach the base inside intact double-stranded DNA, so the protocol requires denaturation — acid, heat, or DNase digestion. EdU is detected by a copper-catalysed reaction between the alkyne on the nucleoside and an azide-conjugated dye, which is small enough to reach the incorporated base without opening the duplex. Thermo Fisher states the point directly: EdU assays “are not antibody-based and therefore do not require DNA denaturation for detection of the incorporated nucleoside”, whereas BrdU detection requires denaturation with HCl, heat, or DNase.

That has a downstream consequence people discover late: acid denaturation destroys a fraction of protein epitopes and quenches fluorescent proteins, so a BrdU protocol constrains the surface panel and any reporter line you run alongside it.

EdU is not free of biology, either. In mouse embryonic stem cells, EdU incorporation drove accumulation in G2/M followed by apoptosis, and a DNA damage response was triggered in every cell type the authors analysed (Kohlmeier et al., Chromosome Research, 2013). Worth noting where the dose sits: those experiments used 10 µM EdU, a pulse or a labelling period spanning S phase — below the 20 µM in the protocol above. A short pulse on a robust cell line is generally tolerable. Continuous labelling is where it bites, and the threshold is lower than most people assume — the damage response in that paper came from labelling through a single S phase, not from days of exposure. Stem cells are the documented worst case.

BrdU — when it is still the right call

BrdU keeps one real advantage: decades of published protocols and reference data, particularly for in vivo labelling, and a validated SOP that already exists in many labs. If your panel has no acid-sensitive markers and no fluorescent-protein reporter, switching to EdU buys convenience rather than accuracy.

Where each one goes wrong

All three of these readouts sit downstream of a clean parent population, and all three inherit any gating error above them. Doublets are the specific hazard when the readout is paired with DNA content: two G1 cells scored as one event look like a G2/M cell, so FSC-A versus FSC-H singlet discrimination is not optional before a bivariate EdU/DNA or BrdU/DNA plot. The same applies to the model-fitting side of the analysis covered in Dean-Jett-Fox cell cycle fitting.

One more boundary worth stating plainly: none of these three resolve division history. If the question is how many times a cell has divided, the readout is dye dilution. The same applies when you need to identify non-dividing bystanders inside a responding population, and the metrics are the ones described in CFSE division index and proliferation index. Ki67, EdU and BrdU can each tell you a cell is cycling; none of them tells you it is on generation four.

The short version

Whichever you pick, the panel decision comes before the protocol decision, because the detection chemistry is what removes fluorochromes from your options. If you are deciding where a proliferation channel can sit without spilling into the markers you care about, the fluorophore spectrum viewer shows spillover into each detector for a given instrument configuration before anything is stained.