Open wounds pose a major infection and mortality risk in animals. To reduce these risks, many animal species apply antimicrobial compounds on their wounds. Ant societies use antimicrobial secretions from the metapleural gland to combat pathogens but this gland has been lost over evolutionary time in several genera including Camponotus. Using behavioral and microbiological experiments, we studied how Camponotus floridanus handles infected wounds without the use of antimicrobial secretions. When we experimentally injured a worker's leg at the femur, nestmates amputated the injured limb by biting the base (trochanter) of the leg until it was severed, thereby significantly increasing survival compared to ants that did not receive amputations. However, when the experimental injury was more distal (at the tibia), nestmates did not amputate the leg and instead directed more wound care to the injury site. Experimental amputations also failed to improve survival in ants with infected tibia injuries unless the leg was amputated immediately after pathogen exposure. Micro CT-scans revealed that the muscles likely responsible for leg hemolymph circulation are predominantly in the femur. Thus, it is likely that femur injuries, by attenuating hemolymph flow, provide sufficient time for workers to perform amputations before pathogen spread. Overall, this study provides the first example of the use of amputations to treat infected individuals in a non-human animal and demonstrates that ants can adapt their type of treatment depending on the location of wounds.

Funding provided by: Deutsche ForschungsgemeinschaftCrossref Funder Registry ID: https://ror.org/018mejw64Award Number: 511474012 Funding provided by: Swiss National Science FoundationCrossref Funder Registry ID: https://ror.org/00yjd3n13Award Number: 310030_156732 Funding provided by: European Research CouncilCrossref Funder Registry ID: https://ror.org/0472cxd90Award Number: 741419

Data includes: 1) Behavioral ethograms 2) Survival data 3) Bacterial qPCR 4) R-Code for analyses of datasets 1-3 5) PDF with R-code on the Bayesian Model in Fig. S1D 6) PDF with R-code for Figures 1-3