Parasite host manipulation

Why does an infected ant climb a grass stem each evening and clamp its jaws where a grazing sheep will swallow it?

An ant carrying the lancet liver fluke does something no healthy ant does. As the evening cools it leaves the trail, climbs a grass blade, bites down on it, and stays there until morning warms it — at which point it lets go and forages normally. The obvious reading is that the parasite has taken control. But before accepting that, ask the harder question: control is expensive and specific, so what exactly would have to be true for a fluke to be selected to do it — and how would we tell that from an ant that is simply ill?

Start with the constraint the parasite faces. This fluke matures in the bile ducts of a grazing mammal. Its eggs leave in dung, are eaten by a land snail, multiply there, and are shed in balls of slime that ants drink. And now it is stuck. An ant will not, in the ordinary course of its life, be eaten by a sheep. Ants forage on the ground by day and are back in the nest by night, which is precisely when and where sheep are not eating them. The parasite's problem is not survival; it is a transport problem with a very low background rate of success.

So what would improve that rate? Being somewhere a grazer's mouth will pass, at a time a grazer is grazing. That is a specific request, and the observed behaviour answers it with uncomfortable precision: climb, attach at the top of a blade, hold through the cool hours, release when it gets hot. Notice the release. An ant clamped to a stem in the midday sun would desiccate and die, and dead ants are eaten by nobody — so the behaviour is conditional on temperature, and its conditionality is the strongest single argument that it is not simply malfunction. Sickness is not usually reversible twice a day on a thermal cue.

Now the part that makes this more than a curiosity. Dissect an infected ant and you find dozens of cysts in the abdomen, and one — occasionally two — lodged near the nerve cluster in the head. The one in the head does not become infective. Whatever it is doing, it is not going to be a fluke in a sheep. So we have a trait that benefits others and costs the individual performing it, which is exactly the pattern that ought to be selected away.

What rescues it? Ask which entity is being selected. Trematodes multiply asexually inside the snail, so the cercariae arriving in a single slime ball are expected to be a clone. If the fluke in the head shares its genotype with the fluke in the abdomen, then an allele causing "lodge in the head and manipulate" is copied forward through the ones that get swallowed, even though every individual carrying out the act is a dead end. The behaviour is not altruism in any interesting sense; it is a genotype spending one copy of itself to move the others. How reliably clonal a given ant's infection actually is remains an empirical question — ants can drink from more than one slime ball — and the strength of the argument scales with that relatedness.

This is the domain's most dangerous kind of story, because a manipulation account is always available and usually wrong. Most behavioural changes in infected animals are pathology, not strategy: fever, lethargy, loss of appetite, seeking warmth. Some are the host's defence rather than the parasite's doing. So a claim of manipulation needs to earn itself, and the working criteria are reasonable ones — the change should be specific and elaborate rather than general debilitation, it should plausibly and ideally demonstrably raise transmission, and it should appear in a parasite whose life cycle actually requires that step.

The lancet fluke meets the first criterion strongly and the second by inference: a clamped ant at the top of a blade at grazing time is in the right place, but directly measuring how much that raises the odds of ending up in a sheep is hard in the field, and I would not pretend the quantitative case is closed.

Two further honesties. The proximate mechanism is not established. What the head-lodged fluke secretes, and which circuits it acts on, is not something I can state — "it controls the brain" is a description of the outcome, not an account of the cause. And the assumption that manipulation must work through the brain has already been embarrassed once: in the fungus that makes carpenter ants bite onto leaf veins, careful reconstruction found the fungus filling the mandible muscles and body cavity while leaving the brain uninvaded. The lesson generalises — the route to a behaviour may be peripheral.

What makes the whole account falsifiable is that each link predicts an absence. If uninfected ants clamped at the same rate when chilled, temperature rather than infection explains it. If ants carrying only abdominal cysts, with no fluke in the head, clamped just as readily, the head-lodged individual is not doing the work. And the load-bearing claim — that this raises transmission — fails outright on one observation: if clamped ants were swallowed no more often than foraging ants, there is nothing for selection to have built, and the behaviour would have to be reclassified as a side effect of having a fluke near one's nerves.

• How relatedness within a single infected host is actually measured, and what mixed infections do to the argument.
• Why some manipulations work through peripheral tissue rather than the nervous system, and how that was discovered.

Every term the collection defines is gathered in the glossary.