Chemotherapy selectivity

Why does a drug aimed at a tumour also attack hair, gut lining, and blood?

A cytotoxic drug is described as aimed at a tumour, and then the predictable things happen: the hair goes, the mouth and gut ulcerate, the white cell count collapses. Three tissues, in three organ systems, none of them the target.

That is not a side effect in the ordinary sense of an unlucky off-target hit. It is the same effect, arriving where it was not wanted — which tells us that whatever the drug recognises, hair follicles, gut lining and bone marrow have it too. What is it?

Ask what a classical cytotoxic actually does. It alkylates DNA, or crosslinks the two strands, or starves the cell of the nucleotides needed to copy them, or jams the spindle so chromosomes cannot be pulled apart. Notice what is absent from that list: anything that identifies a cell as malignant. There is no cancer receptor being bound. The drug is a general insult to the machinery of copying and dividing.

So where does any selectivity come from? Only from a proxy. A cell that is replicating its DNA or building a spindle is exposed at that moment; a cell doing neither is much less so. Cancer, the reasoning went, divides more than the tissue around it, so a drug that punishes division punishes cancer more. Hair matrix, gut crypt and marrow progenitors are collateral precisely because they answer to the same proxy.

Now test the proxy honestly, because this is where the folk version fails. The small intestinal lining is replaced roughly every three to five days, its crypt cells cycling on something like a daily schedule; the marrow turns out neutrophils on the order of a hundred billion a day (both recalled figures). Many solid tumours, meanwhile, double in volume over months. On the plain reading — "chemotherapy kills whatever divides fastest" — the gut should be devastated and the tumour largely spared. Both happen, but the tumour also shrinks, and often shrinks more.

Be careful here rather than glib, because volume doubling time is not cell cycle time: a tumour can have cells cycling briskly while its bulk grows slowly, since much of what it produces dies or is shed. So the arithmetic does not cleanly refute the proxy. What it does show is that division rate cannot be the whole account — the ordering of tissues by proliferation does not predict the ordering by clinical damage.

What is missing is recovery. Consider why the drug is given in cycles — a dose, then two or three weeks of nothing, then another. If kill were everything, continuous exposure would be better. It is not, and the interval is the point. Normal marrow and gut hold a reserve of stem cells that are largely out of cycle at the moment of exposure, survive it, and repopulate the tissue during the gap. The tumour, as a rule, has no comparably organised quiescent reserve, and frequently has lost the checkpoint machinery that would let it pause and repair. The therapeutic gain lives in the difference between two recovery curves, not in a difference between two kill rates.

Which gives us a falsification test with a visible answer. If exposure only harms cells that happen to be in cycle at the time, then a tissue with a resting subpopulation should be damaged incompletely. Around 85 to 90 per cent of scalp follicles are in their growth phase at any moment (recalled), the rest resting — so a single exposure should thin rather than clear the scalp, and hair should return afterwards from the quiescent stem cells in the follicle. Both are what is observed.

Two refinements, and one place where my own account gives way.

The refinement first. "Selectivity by proxy" is not the same as no selectivity. It is a real and useful bias, and it explains the pattern of toxicity exactly — hair, mucosa and marrow are hit because the proxy is honest about them. It is only the second half, the claim that cancer sits at the top of the proliferation ranking, that does not survive contact with the data.

Now the failure. I have leaned on the idea that tumours are vulnerable because they have lost the checkpoints that would let them arrest and repair. Testicular germ cell tumours refute that as a general rule: they are among the most curable of all solid cancers by conventional chemotherapy, and their sensitivity is attributed to intact apoptotic machinery — they respond to damage by dying promptly rather than by tolerating it. So the arrow does not point one way. What covers the gap is that repair capacity and readiness-to-die are two separate axes, and a tissue's fate under a cytotoxic depends on where it sits on both.

I have not quoted a growth fraction or a cell cycle time for any particular tumour. Those vary by orders of magnitude between and within cancers, and a single number would be doing rhetorical work its evidence cannot support. The weakest figures I have used are the recalled renewal intervals and the follicle growth-phase share; the firmest claim is the structural one, that classical cytotoxics carry no cancer-specific target at all.

• Why resistance to a cytotoxic emerges, and whether it is selection of pre-existing cells or something acquired under treatment.
• Two neighbouring pieces set the boundaries here: Dose-response explains the therapeutic window this whole argument lives inside, and Programmed cell death explains the dying itself; this piece asks only how a drug with no target nevertheless discriminates.

Every term the collection defines is gathered in the glossary.