Freshwater shrimp kept for twenty hours at a caffeine concentration detected in highly polluted rivers spent less time in the dark than shrimp kept at ordinary background levels, and that shift was about the same size as the one caused by a parasite that manipulates them into being eaten by water birds

A tube half wrapped in black tape

How much caffeine does it take to change the way a river animal handles the risk of being eaten? In a laboratory in Bonn the figure on the label was 10,000 nanograms per litre, a level that has been detected in highly polluted rivers. After twenty hours at that concentration the animals spent less of a five-minute test in the dark than animals held at a hundredth of it.

The animals are Gammarus pulex, the freshwater shrimp of European streams. They are amphipods rather than true shrimp, they shred dead leaves for a living, and they are the standing food supply for a long list of fish and birds. The work is reported in Biology Letters by Elias Fritzsche, Hannah Otto, Rudram Bhagwat, Nicole Bosse, Carola Mauel, Joachim Frommen and Timo Thünken, six of them at the University of Bonn and Frommen at Manchester Metropolitan University.

The apparatus was plain. A clear plastic tube twenty-five and a half centimetres long and four across was wrapped in black tape over half its length, tied to a green styrofoam plate, and filled to a depth of two centimetres. One gammarid went in through a hole in the middle by pipette, got two minutes to settle, and was then watched for five.

The water in that tube is worth following, because the paper describes caffeine going only into the twelve exposure tanks. It came from a separate tank of brook water and was changed before every trial, with its temperature measured each time, ranging from 16.0 to 16.6 degrees Celsius. So the animals spent twenty hours in caffeine and then took the test in water from home, carrying whatever the exposure had done to them out of the tank and into the tube.

An observer with a stopwatch recorded how long each animal spent on the lit side, and the count of crossings between the two sides stood in for activity. Between trials the tube was turned end for end, so that a preference for one corner of the room could not masquerade as a preference for the dark.

The observer was blinded to the treatment conditions, the paper says. Infection is harder to hide: infected animals are recognised by the bright orange parasite visible through the cuticle, and the paper does not say whether that was masked during the trials or raise it as a limitation.

A hundred and twenty animals were put into twelve aerated tanks, ten to a tank, with a single dead leaf each for food and cover. Six tanks were dosed to 10,000 nanograms per litre and six to 100, arranged alternately to keep external factors from lining up with a treatment. A hundred animals were tested and twenty held in reserve.

The parasite that makes its host careless

Half the animals in every tank were already carrying something. Polymorphus minutus is an acanthocephalan, a thorny-headed worm that has to get from a gammarid into a water bird in order to reproduce. From outside it shows as a bright orange spot through the shrimp’s cuticle.

Parasites of this kind are the textbook case of an animal’s behaviour being rewritten from inside. This one inverts its host’s response to gravity and alters how it responds to light, and it increases the shrimp’s chances of ending up inside one of the water birds the worm needs. Running caffeine against it gave the study a yardstick: a behavioural change of known ecological consequence, produced by something that has been under selection to produce it.

On the light measurement, the two came out comparable. Relative to the low-caffeine group, the high-caffeine animals were less photophobic; relative to uninfected animals, the infected ones were too. The two acted independently of one another, and the effect sizes were comparable, a partial eta squared of 0.05 apiece in the statistical model.

They were not alike in kind. The high-caffeine animals also crossed between light and dark more often than the low-caffeine ones, and infection made no statistically detectable difference to that. Infected animals crossed at a rate about 30 per cent above uninfected ones, but the confidence interval on that ratio runs from just below one to just under 1.8, which is why the paper does not claim the difference.

Nor did the two combine. The interaction between them was not detectable either, so caffeine neither sharpened the parasite’s manipulation nor blunted it.

A comparison between two doses

The experiment had no caffeine-free tank, and the reason is in the design. Both doses were chosen to be real: 100 nanograms per litre stands for ordinary background contamination and 10,000 for a highly polluted river. What the study therefore measures is a difference between two concentrations. The results, its authors write, should be read as evidence of concentration-dependent effects instead of the effects of caffeine exposure as such.

The two numbers are also quantities added. The tanks were filled with water from the animals’ own brook and then dosed, and the caffeine solution was made up in tap water, so 100 and 10,000 nanograms per litre are the concentrations that result if the starting water held none. The paper does not report measuring what was already in it.

No predator was anywhere in the experiment. Predation is discussed at length, in the introduction and again in the discussion, but the word appears nowhere in the methods. The increase in predation risk is a step the paper reasons to, not one it observed, and it is careful about the tense: the expectation is that reduced photophobia and higher activity will lower gammarid population sizes, not that they were seen doing so. A shrimp spending more of five minutes on the bright side of a taped tube is not a shrimp that was eaten.

The team also did not record which of the twelve tanks each tested animal came from, so tank could not be entered into the statistics as a random effect. They argue that six independent tanks per treatment and haphazard sampling make it unlikely the effects were driven solely by tank conditions, then add that they cannot completely exclude some degree of pseudoreplication.

The exposure ran twenty hours with the tank water unchanged, under a light regime the authors describe as a natural cycle typical of mid-July. The paper gives no year or calendar date for the experiment and no temperature for the exposure tanks. Two animals were dropped from the analysis for inactivity throughout the trial, and both were infected animals from the low-caffeine tanks.

Small effects in an animal that turns up in numbers

Both effects were rather small, the paper says, before making the argument for why small might still matter. Gammarids often occur at high population densities and do a great deal of the work in a stream. They shred leaf litter, recycle nutrients, feed vertebrate and invertebrate predators alike, and prey on other invertebrates themselves.

That is the argument for why a difference this small might become an ecological question, and the paper’s own phrasing is conditional: the effects could nevertheless be ecologically important. It is also, on its own account, untested. The ecological impact of caffeine pollution needs to be analysed in future studies, the discussion says, and the population decline it anticipates carries a stated reasoning rather than a measurement.

Caffeine lasts in water for months

The reason caffeine turns up in streams at all is that people drink it and wastewater treatment often does not fully eliminate it. It is biodegradable under ideal conditions and rather stubborn outside them. The paper puts its half-life at 100 to 240 days, which it says results in caffeine persisting in aquatic systems. Separately, and on different evidence, it records that caffeine accumulation has been documented in various marine and freshwater organisms.

The compound is stable enough that hydrologists have used it to quantify untreated wastewater in karst systems. Ten thousand nanograms per litre is not an ordinary river, though. It is the polluted end of a range that has been documented worldwide, from urban rivers to remote streams, and the high treatment here was set to match it.

What a concentration does depends heavily on the animal and on how long it sits there. The paper’s own survey of the field has caffeine exposure associated with increased locomotor activity and altered swimming patterns in copepods. Daphnia magna, a water flea, responds more variably, and at higher concentrations or over longer exposures the response can turn severe, running to impaired swimming, feeding inhibition and death.

Fish show the same split between the short term and the long. Brief exposure may increase activity, while longer exposure can lead to developmental abnormalities and lethality. The authors’ summary of all of it is that caffeine can disrupt neural and behavioural processes in aquatic organisms, and that the magnitude and direction of the effects are context-dependent.

The brook has the last word here. The Dransdorfer Bach supplied the gammarids and, by way of a separate holding tank, the water they were tested in, so the animals answered the question in water from home while carrying twenty hours of a dosed tank inside them. What the five minutes showed is that the gap between a hundred nanograms per litre and ten thousand is wide enough to change how a shrimp uses the dark.