That zombie-ant fungi make their hosts climb and bite fast is hardly disputed any longer. What is harder to demonstrate is how they manage it. Recent molecular work has, for the first time, established a direct link between a secreted fungal protein and changes in olfactory perception and social behaviour in ants.
Around noon, a large ant climbs up the vegetation. She staggers a little, but continues to search until she sets her jaws into a leaf or twig. After that, she never lets go. The ant dies: the fungus Ophiocordyceps camponoti-floridani digests her from the inside out and forms a fruiting body from which spores descend. The location, timing, and posture seem precisely tuned to the reproduction of the fungus. But how does such a spectacular scene evolve into scientific evidence for manipulation – and ultimately into an explanation at the molecular level?
The host in this system is Camponotus floridanus, a large ant species from Florida. It is precisely the unmistakable final behaviour that makes these zombie ants so suitable for research, says microbiologist and expert in parasitic fungi Charissa de Bekker of Utrecht University. ‘We absolutely do not see healthy ants climbing up only to then permanently bite fast. It is almost a binary system: if you see that biting, you know the parasite has changed the behaviour. As a result, you are much less in a grey area than when an animal only becomes slightly less anxious or a bit more active.’
Hyperactive
Yet the change begins long before that final notorious death grip. Healthy specimens of Camponotus floridanus forage mainly at night and coordinate their behaviour with nestmates. Within about ten days after infection, that fixed rhythm disappears: the ants also become active during the day, sometimes moving hyperactively and aimlessly, and communicate less and less with their colony mates. ‘The ant’s biological clock seems to be thrown into disarray,’ says De Bekker. ‘They are increasingly found outside the nest and gradually become detached from the social network.’
That social decoupling is a clear part of the behavioural change and is likely beneficial for the fungus. Ants recognise sick nestmates by smell and can clean, attack, or remove them, which can also break the infection. ‘It is very likely that the fungus thus bypasses the colony’s social defences,’ De Bekker believes. ‘Only we haven’t directly tested that specific function yet.’
To monitor the course of infection in a controlled manner, her research group injects ants in the laboratory with fungal cells. ‘With a very small glass needle, we suck up fungal cells in liquid and inject them into the ant under the binocular microscope.’ Afterwards, they stay in bins with a nest area and a kind of ‘little tree’ in which they can climb and bite fast. After two to three weeks, the climbing and biting behaviour follows there as well.
But that does not yet distinguish every behavioural change from general sickness behaviour. In a study in Animal Behaviour (2021), De Bekker’s group therefore also infected ants with Beauveria bassiana, a pathogen that does not cause climbing and biting behaviour. ‘In the Ophiocordyceps ants, the daily activity rhythm was almost gone after about ten days,’ says De Bekker. ‘The Beauveria ants were also sick and their rhythm also changed, but a daily pattern clearly remained visible.’
The difference between the two infections was also reflected in gene activity. Halfway through the Ophiocordyceps infection, nearly three times fewer ant genes showed an activity pattern of approximately 24 hours than in healthy animals, according to a 2023 bioRxiv preprint. With Beauveria, fewer of those rhythms were lost, but the peak activity shifted more strongly. Both fungi thus disrupt the biological clock in a different way.
In addition, light seems to guide the ant to her final position during the climbing and biting phase. Light intensity and direction are related to the climbing route, biting height, and orientation of infected ants. Light likely leads them to a microclimate where temperature and humidity are suitable for the weeks-long growth of the fruiting body. That the location matters is evident when researchers move ants that have bitten fast: in deviant locations and positions, the fungus less often forms a normal fruiting body (Behavioral Ecology, 2019).
For evolutionary biologist Robert Poulin of the University of Otago in New Zealand, manipulation begins precisely with such an advantage. ‘I define it as a change in the host’s phenotype that is specifically caused by a parasite and promotes its transmission, survival, or reproduction,’ he says. ‘Furthermore, the ability to induce that change must be heritable, thus anchored in the parasite’s genome.’
Abnormal behaviour can also be a side effect of tissue damage or an immune response. ‘If the genes involved primarily help to fight the immune system or extract food from the host, they have not necessarily evolved for behavioural manipulation,’ Poulin explains. ‘A molecule that affects a specific molecular target or a neural signalling pathway makes the case more convincing.’
Evidence
Even without molecular evidence, timing can be telling. Poulin mentions hairworms, which only need to go to water when they are adults. Precisely then, the parasite makes its cricket host enter the water. ‘Should that happen when the worm is still tiny, then it would more likely be a side effect. But the change appears here exactly when the parasite is ready to emerge.’
In the case of Ophiocordyceps, he found the evidence convincing even before researchers could measure gene expression. ‘It’s not about an ant showing 10 percent more or less of normal behaviour. She climbs, bites fast, and dies there just before the fungus grows out of her body and starts spreading spores. That can hardly be the product of chance. Natural selection must have brought those components together step by step.’ According to Poulin, hundreds of systems have been labelled as manipulation, ‘but perhaps only a handful where so many different types of evidence converge.’
That Ophiocordyceps manipulates its host is therefore hardly disputed any longer. The next question is how the fungus manages this physically and molecularly. Research on a related zombie-ant fungus showed with three-dimensional microscopy that the fungus forms extensive networks around the muscles, while no fungal cells were found in the brain during the manipulation phase (PNAS, 2017). During the death grip, the jaw muscles appeared strongly contracted, while motor nerves and neuromuscular connections seemed largely preserved (Journal of Experimental Biology, 2019).
It is possible that Ophiocordyceps actually actively limits some of the damage to the ant brain during the manipulation phase
That pattern does not fit a fungus that simply grows into the brain or destroys the nervous system at random. It is possible that Ophiocordyceps actually actively limits some of the damage to the ant brain during the manipulation phase. In manipulated ants, researchers found strongly increased concentrations of the antioxidant ergothioneine; this increase was absent in a Beauveria infection (Journal of Insect Physiology, 2019). A later multi-omics study also pointed to possible neuroprotective roles for kynurenic acid and carnosine (Scientific Reports, 2023). In this way, the fungus may keep important nerve tissue usable for longer while severely affecting the jaw muscles.
‘If the behavioural change only arose because a fungus randomly damages tissue, you would expect it to happen much less precisely,’ says De Bekker. ‘It is precisely that precision that makes us think that all sorts of things are happening at the molecular level and that it is not just random tissue destruction.’
Using genomics, transcriptomics, and metabolomics, her research group is searching for the molecules with which the fungus achieves this. ‘With those, you first get mainly a long list of suspects,’ De Bekker explains. ‘You see that a gene is strongly activated during manipulation, but not yet whether the behaviour is caused by it.’
Fungal proteins
To refine that long list of candidates, De Bekker and colleagues compared, among other things, two specialised fungus-ant pairs (G3: Genes, Genomes, Genetics, 2020). The two specialised Ophiocordyceps species each manipulate a different ant host and are not each other’s closest relatives. Corresponding patterns in gene activity during the manipulation phase are therefore interesting candidates for a shared mechanism. In this way, possible effectors came into view – secreted fungal proteins and other substances that can influence processes in the host. In the ants, the activity of genes involved in olfactory perception, daily rhythms, nutrition, and neuronal functions also changed.
De Bekker and colleagues have now followed one candidate from that long list from fungal protein to molecular target and changed behaviour in the ant: the small, cysteine-rich scramblase-binding peptide, SBP. The corresponding gene is hardly active when the fungus grows independently, but is strongly activated during the manipulated climbing. The peptide was found to bind to scramblase-1, a membrane protein of which related forms in nerve and sensory tissue are involved in signal transduction. As a result, SBP also became interesting as a possible cause of changes in olfactory perception and social behaviour.
Because genetic experiments in ants and Ophiocordyceps fungi are difficult, the team first used the model worm Caenorhabditis elegans. There, SBP and scramblase-1 were found in the same nerve and sensory tissue, and the peptide caused measurable changes in motor skills. Ants treated with SBP avoided citronella less, made less antenna contact, and clustered less with nestmates.
De Bekker and colleagues also suppressed the gene that codes for scramblase-1 in the ant. Those animals also reacted less strongly to citronella. That both interventions weakened the response to citronella supports the hypothesis that this molecular pathway is involved in the altered olfactory perception. The effects of SBP on antenna contact and clustering also indicate that the peptide can contribute to social decoupling. The results have been on bioRxiv as a not yet peer-reviewed preprint since September 2025.
However, SBP is not yet a zombie button: the peptide does not demonstrably make ants climb or bite fast, and the gene has not yet been knocked out in the fungus to test whether the natural manipulation then disappears. ‘We now have a target on one side and a behavioural phenotype on the other,’ says De Bekker. ‘But what happens in between is still a black box.’
She also does not expect one molecule to control the entire behavioural sequence. ‘The fungus probably uses a cocktail of substances that act at different times on communication, the biological clock, the nervous system, and the muscles.’
The infection trials, timing, and biting site have made the zombie ant one of the strongest cases for adaptive manipulation. The scramblase-binding peptide now opens a first molecular pathway within that complex behaviour. How the fungus subsequently links clock, senses, nerve tissue, and muscles into one deadly sequence is still largely unknown.