A kept thing lost on schedule at every moult — a lower termite’s hindgut flagellates, dead days before the skin is shed, restored only by drinking hindgut fluid from a nestmate who has not moulted, again and again, or the colony starves on wood
R893 closed on a gift given once: a koala mother serves her joey a purpose-made first meal, and the installed microbes last the eater for life. Its closing turn asked for the opposite tenure — a partnership that cannot be given once, because its keeper keeps losing it on schedule. It exists underground and inside dead wood, and the first thing to say about it is a correction to the picture that suggests itself.
Stocked for wood. The lower termites — the older termite lineages, as distinct from the higher termites, which manage without this arrangement — eat wood and other plant fibre they cannot fully process alone. In an enlarged chamber of the hindgut, the paunch, they house single-celled flagellates: parabasalids and oxymonads, anaerobic eukaryotes found in this lineage and essentially nowhere free-living. The flagellates are central to breaking down cellulose, and they do not travel alone — most of the prokaryotes in the gut are associated with the protists themselves, attached to their surfaces or living inside them, so what is kept is a community stacked inside a community. A young termite acquires it early, by about the third instar, first from its parents and then, in an established colony, from the circulation among nestmates. All of a colony’s flagellates ultimately trace back to the founding parents; no new kinds arrive from outside.
Lost before the shed, not with it. It is natural to picture the loss as mechanical: the hindgut lining is insect skin, skin is shed at a moult, so the tenants are cast out with their wall. The review literature calls that the most common error in the whole story. In the termites studied, the large flagellates are already dead and gone from the gut days before the host moults — about seven days before in Kalotermes flavicollis, about six days before in Coptotermes formosanus. The lining is indeed shed, but it is retained inside the body for two or three days and then breaks up and is voided in pieces; the flagellates are not riding it out. What kills them is genuinely unsettled. The familiar explanation blames the gauntlet of the moulting gut — the host stops feeding, the gut fluid thickens, oxygen may intrude — conditions these protists are known to be exquisitely sensitive to. An older line of work, revisited in the 2017 review this page leans on, points instead at the host’s own moulting hormone: in the termites’ sister group, the wood-feeding cockroach Cryptocercus, rising ecdysone is exactly the signal that triggers the flagellates’ protective sexual-encystment cycles, long before feeding changes. Why a comparable signal should arm the cockroach’s symbionts and disintegrate the termite’s is the open question, framed there as a possible cost of the changed development that came with eusociality. This page reports the dispute; it does not settle it.
The contrast that proves the loss is a choice of history. Cryptocercus keeps what the termite loses. When its moult approaches, its flagellates encyst or otherwise harden, are held inside the shed hindgut lining while the new one forms, then escape back into the gut and repopulate it. After establishment, a Cryptocercus nymph is nutritionally autonomous — it could, in principle, live alone. A developing termite cannot. It emerges from each moult in an aposymbiotic phase with respect to the large flagellates: stocked gut gone, wood still indigestible, and no reservoir of its own to draw on.
So the colony is the reservoir. There is exactly one road back, and the review is blunt about the exclusions: not eating faeces — termite faeces never contain active flagellates, cysts or other resistant stages — and not feeding on a shed lining, its own or anyone else’s. The sole mechanism of refaunation is repeated proctodeal trophallaxis with an intermolt donor: the newly moulted termite puts its mouth to a nestmate’s anus and drinks hindgut fluid, live flagellates included, and does so repeatedly until its own paunch is stocked again. Note the timing hidden in the word intermolt: the donor must be a nestmate whose own moult is not imminent, because a colony where everyone moulted at once would be a colony of empty guts begging from empty guts. The keeping therefore lives in the colony’s asynchrony as much as in any individual’s gut. The review’s framing, offered as interpretation and labelled as such here, is that this periodic death was a tipping point: once your symbionts die at every moult and only a relative can replace them, independent living is off the table, and cooperation stops being a strategy and becomes the substrate.
Why this keeper is different from the others in the run. Buchnera was inherited inside the egg and never risked a journey. The squid sieved its partner from the sea, once, at birth. The koala was served its community once, in a first meal, for life. The termite is served again and again, by companions rather than a parent, not at one threshold age but at every moult of its development — a standing obligation written into the one event, growing, that it cannot skip. Its partnership is not a birthright and not a meal. It is a subscription the colony pays on its behalf, in the only currency that carries it: another body, still stocked, close enough to ask.
One moult, in order: the stocked paunch, the dying days before the shed, the empty gut, the drinking, the stocked paunch again. Schematic, compressed from the published record.
Walk it in order.
Schematic only: the stages compress work across species, studies and decades into one developing termite’s moult. Labelled cautions: the six- and seven-day figures are for the two named species, not a termite-wide timetable; what kills the flagellates is disputed — gut conditions around the moult and the host’s hormonal environment are rival accounts, and the hormonal reading is a hypothesis revisited from Cleveland’s older work, not a settled result; the claim is about the large flagellates — bacteria and archaea in the same gut are a wider community and this page does not say all of them vanish on the same schedule; higher termites lack these flagellates entirely and are not a counterexample but a different arrangement; and the tipping-point reading of eusocial origins is the 2017 review’s framing, reported here as interpretation, not as consensus.