Leafcutter ants (Atta cephalotes) maintain four worker castes — defensive patrollers, leaf harvesters, brood carers, and fungal gardeners — from the same genome. The castes differ in body size, but what determines which job a worker performs is not fully explained by size alone.
Gilbert et al. (Cell, 2025) identified two neuropeptides that specify caste behavior. CCAP (crustacean cardioactive peptide) is elevated in leaf harvesters and promotes leaf-cutting behavior. Neuroparsin-A (NPA) is abundant in defensive patrollers and suppresses brood care. When researchers injected CCAP into defensive ants and brood carers, the ants approached leaves and attempted to cut them. When they knocked down NPA expression in defensive ants using inhibitory RNA, the ants began ferrying brood to the fungal pile — the signature behavior of a different caste. Two molecules, adjusted up or down, reprogram which social role an ant performs.
The reprogramming is incomplete. Defensive ants induced to cut leaves were not very good at it. The neuropeptide changes the behavioral target but not the skill. The caste switch is motivational, not developmental — the ant wants to do a different job but hasn't practiced.
The structural finding is the cross-species comparison. Naked mole-rats — eusocial mammals with division of labor between workers and breeders — show substantial overlap in the gene-expression patterns that distinguish worker types. Ants and naked mole-rats last shared a common ancestor approximately 600 million years ago. That ancestor was not eusocial. The division-of-labor systems evolved independently. Yet the molecular machinery governing those systems converges on the same neuropeptide families.
The convergence goes beyond parallel function. When researchers treated naked mole-rat astrocytes with the ant version of neuroparsin-A — a peptide that naked mole-rats do not naturally produce — gene expression changed in a pattern consistent with receptor activation. The ant molecule could interact with mammalian signaling machinery, likely through the insulin receptor. The neuropeptide from a 600-million-year-distant lineage was still legible to the mammalian brain.
This is not the familiar convergence of form — wings in birds and bats, eyes in vertebrates and cephalopods — where similar selection pressures produce similar shapes from different materials. Here the materials themselves converge. The specific molecular signals that assign social roles in an insect colony are close enough to the signals in a mammalian colony to activate each other's receptors. The convergence is not at the level of anatomy or behavior but at the level of individual signaling molecules.
The implication: the problem of assigning social roles within a colony constrains the molecular solution so tightly that evolution arrives at the same answer independently. The solution space for “make this individual a worker instead of a carer” is not large. The neuropeptide families that can modulate behavioral motivation while leaving morphology intact are limited. Two lineages, separated by more evolutionary time than separates humans from sponges, converge not just on the same social structure but on the same chemical vocabulary for implementing it. The command is shared because the language has few words.