Data from: Better safe than sorry: Leg amputations as a prophylactic wound care behaviour in carpenter ants
Abstract
Animals often sustain injuries that are susceptible to lethal infections. In social insects, wound care behaviours have evolved to reduce these risks. But the limits of wound care behaviours remain unclear. Here we investigated the wound care behaviours of the ant Camponotus maculatus. Our findings show that amputation of legs infected with Pseudomonas aeruginosa significantly reduced mortality. However, nestmates do not differentiate between infected and sterile injuries, providing the same treatments regardless of the infection. Even though we show that early amputation correlates with higher survival rates, nestmates amputate indiscriminately on legs with fresh or old wounds. Additionally, cuticular hydrocarbon profiles differed between ants with infected or sterile wounds only 24 hours post-injury, a timepoint when amputations are no longer effective. We propose that C. maculatus workers perform prophylactic amputations regardless of injury state or age. This is in sharp contrast to previous studies, which showed clear capabilities to treat infected wounds differently in ants using antimicrobial compounds. This work, therefore, shows the limits of wound care behaviours in social insects, allowing us to better understand the evolutionary drivers of this unique behaviour.
Dataset DOI: 10.5061/dryad.w6m905r28
Description of the data and file structure
The zip file Analyses.zip contains folders divided into 4 analyses sections (analyses_chc_infection; analyses_efficiency_amputations; analyses_ethograms; analyses_time_amputations). Data analysis was conducted using R and RStudio. The zip file Data.zip contains 4 folders with the respective raw data (data_chc; data_efficiency_amputations; data_ethograms; data_time_amputations). Missing data code: NA
Each analysis folder has a corresponding .Rmd file with the code necessary to recreate all the figures and analyses from the study using the raw data csv files provided in the respective data folder. The .Rmd files are also annotated to explain each line of code.
File: Analyses.zip
Description: includes all the code to reproduce all the results of the study (including figures).
File: Data.zip
Description: includes all the raw data to reproduce all the results of the study.
For data_chc there are 2 csv files:
Mastertable_Camponotus_Final.csv
- Compound_ID: refers to our manually identified compound using Nist library and MS spectra
- Retention_Time: Retention time in minutes of our Gas Chromatographic run
- Retention_Index: The universal index calculated using retention time of the respective alkanes.
- Compound: Compound class the respective compound belongs to
- Chainlength: The number of carbon atoms in the compound
- Subsequent lines refer to the individual samples with G=Healthy; I=Infected; S=Sterile; 0h= zero hours after manipulation; 2h= 2 hours after manipulation; 24h = 24 hours after manipulation. The last number refers to the sample repeat. e.g. the 7th healthy sample at timepoint 2h is called: G2h_7.
Metadata_Camponotus.csv
This file contains the metadata necessary for analysis.
- ID= refers to the identification of the sample. Same code as used above but preceeded by Camponotus_
- Treatment= refers to healthy (uninjured), sterile or infected wounds.
- Time (h) = refers to how many hours after treatment the CHC profile was extracted (0, 2 or 24h)
For data_efficiency_amputations all csv files follow the same nomenclature.
Mortality_Ants_0.1OD_144h.csv refers to an experiment in which wounds were infected with a concentration of 0.1OD of pseudomonas bacteria, while Mortality_Ants_0.01OD_144h.csv refers to an experiment in which a ten times lower pathogen dose was used.
- time= time the individual was last observed (dead or alive)
- event= signals if at that time the individual was dead (2) or alive (1)
- treatment= same meaning as above. Healthy= uninjured, infected (signals either if the sample was infected with a high or low pathogen dose, see csv file name either _0.1OD or _0.01OD); sterile= injured ant without infection; amputated= infected ant whose leg was amputated after set amount of time.
- location= signals if the ants were in their colonies (colony) or kept alone in isolation.
- colony= states the mother colony of the trialed individual
- amputated= states if the individual was amputated inside the colony.
- notes= include observer additional notes. ertrunken= individual drowned in the provided food in the isolation chambers.
For data_ethograms the behavioral ethograms were quantified.
data_ethograms_infected_sterile.csv
- Datum= The date in which the experiment was conducted
- ID= the individual ID of the tested individual
- Versuch= the number of the experiment
- Kolonie = the colony used for the experiment
- Farbcode = the color used to mark the individual (with y=yellow, b=blue, g= green, ry= red and yellow, rb= red and blue, rg= red and green)
- Behandlung = Treatment as outlined above, with infiziert= infected, steril= sterile.
- Wundtyp= if the wound was amputated or not by nestmates. verletzt= injured by us. amputiert= leg amputated by nestmates
- Zeit (absolut) = Time intervall of the ethogram (in steps of 10 minutes)
- Zeit(wunde) = Total time the original wound or amputation was present
- Intervall = Actual time the individual could be observed
- Ant(n)= number of antennations by others
- Ant(t)=duration of antennations by others
- Troph(n) =number of trophalaxis by others
- Troph(t) = duration of trophalaxis by others
- BebbB(n) = number of begging behaviour by others
- BeggB(t) = duration of begging by others
- SelfG(n) = number of selfgrooming by others
- SelfG(t) = duration of selfgrooming by others
- AlloG(n) = number of allogrooming by others
- AlloG(t) = duration of allogrooming by others
- WundB(n) = number of woundcare by others
- WundB(t) = duration of woundcare by others
- AmpV(n) = number of amputation attempts by others
- AmpV(t) = duration of amputation behaviour by others
- Amputation = amputation time
isolation_infected_sterile_144h.csv
- ID= individual ID
- Behandlung= Treatment for that individual I=infected; s=sterile
- Zeit= Time of last observation (dead or alive)
- Überleben= Signaling dead (1) or alive (0)
probability_amputations.csv
- Date= Refers to the date of the experiment
- ID= individual id
- Colony= mother colony of individual
- Treatment= Infected or Sterile wound
- Time_2= Time in hh:mm:ss
- Time= Time in minutes
- Amputation= Time of amputation. 0= no amputation, 1= amputated
time_to_amputation.csv
- Date= Refers to the date of the experiment
- ID= individual id
- Colony= mother colony of individual
- Treatment= Infected or Sterile wound
- Time_to_amputation= Time of amputation in hh:mm:ss
- Seconds= Time to amputation in seconds
- Amputated_during_6h= Signifies if amputation occured within the first 6 hours: 1=yes, 0= no
The last folder data_time_amputations contains 3 csv files
amputations_after_24h.csv
- Colony= Mother colony of individual
- Time = Time after wounding the individual is introduced in the colony
- Treatment= If wound is sterile or infected
- Amputated= how many individuals were amputated
- Total= how many individuals were placed in the colony
- Time_amp_checked= time amputation was first discovered (in hours)
amputations_time_fixed.csv
- Colony= Mother colony of individual
- ID= Individual ID of sample
- Hour= Time of death
- Treatment= Infected or sterile wound
- Leg = Treatment on the leg by us after injury. Injured= not amputated. Amptuated= amputated by us after a set amount of time (see next column)
- Time= Time of artificial amputation (in hours)
- Condition = Dead (1) or Alive (0).
Code/software
For statistical analyses and graphical illustrations, we used the statistical software R v4.3.2 (R Core Team 2025) with the user interface RStudio 2023.06.0+421 (Rstudio Team 2020) and the R package ggplot2 v.3.3.5 (Wickham 2016).
To investigate behavioural differences between workers with infected and sterile wounds, we modelled wound care and allogrooming as binary response variables using Hierarchical Generalised Additive Models (HGAM) with a binomial family. A nested random effect with a smoother was included for each individual within a colony, along with a smoother interaction between time and wound condition. To identify time intervals with significant differences, we conducted post-hoc contrasts on the probability of these behaviours between infected and sterile ants. Given that wound care often occurs in short, intermittent bursts, we considered wound care to be present if it was observed for at least one-quarter of the observation interval (2.5 minutes). In contrast, allogrooming, which is more frequent during ant interactions, was considered present if it occurred for at least half of the observation interval (5.0 minutes). The conclusions remained robust regardless of variations in these criteria. Similarly, to determine when amputations were most likely to occur, we recorded the time at which amputated ants were first observed. Binary records per treatment were analysed using an HGAM model, as applied for allogrooming and wound care. Model assumptions were assessed using the DHARMa v.0.4.7 (Hartig 2024) and mgcv v.1.9.1 (Wood 2017) packages.
To determine whether ants with infected wounds undergo amputation sooner than those with sterile wounds, we used Generalised Linear Models (GLMs) to model the time to amputation as a function of wound condition with the package lme4 v1.1.33 (Bates et al. 2015). Colony ID was included as a random effect to account for colony-level variation.
To evaluate the impact of amputation on mortality at low and high pathogen concentrations, we fitted separate Mixed Effects Cox Models to analyse censored survival data for each experiment. The models included treatment (healthy, sterile, infected, infected + amputated) and environment (colony, isolated) as additive effects, with Colony ID as a random effect to account for colony-level variation. Model fitting was performed using the coxme v.2.2.22 (Therneau 2024).
To assess at which time the amputations are no longer efficient in reducing mortality caused by infections, we fitted a Mixed Effects Cox Model for the censoring data on the survival of amputated ants at different times. Colony ID was included as a random effect. Furthermore, to validate that the mortality of workers was caused by the infections, we compared the survival of workers with an infected and sterile wound by cumulative Kaplan-Meier curves with the packages Surv v.3.5.5 and survminer v.0.5 (Kassambara et al. 2024). To identify differences in the mortality of specific groups, we performed pairwise comparisons on estimated marginal means. We controlled for false discovery rates on multiple comparisons by Tukey adjustments using the package emmeans v.1.11.1 (Lenth 2025).
CHC analyses were performed using the packages ggtext, gplots, GCalignR, ggdist, ggside and analyzeGC. Chromatographic data was aligned in R using the package GcalignR v.1.0.7 (Ottensmann et al. 2018). Subsequently, all compounds with a relative abundance below 0.1% (too small to identify) or less than 50% occurrence within all groups were excluded, so as to focus the CHC analyses to differences across groups of interest as done in previous studies (Alciatore et al. 2021, Yusuf et al. 2020). The final compounds were then identified, and a retention index was calculated based on Carlson et al. (1998).
To visualise dissimilarities between wound types over time, we calculated the Bray-Curtis dissimilarities indices for the relative abundance data between CHC samples. We further visualise these dissimilarities by performing non-metric multidimensional scaling (NMDS). To test whether ants sustaining infected wounds show a distinct CHC profile, we assessed differences between groups over time by performing a multivariate analysis of variance (PERMANOVA) with 9999 permutations. These analyses were performed with the package veganv.2.6-10 (Oksanen et al. 2025).
Experimental designs:
Wound care behavioural experiments
To compare the wound care behaviours towards infected and sterile-injured ants, we created 3 sub-colonies between November 8 and 21, 2022. Six previously marked ants were collected per sub-colony, and all of them were wounded at the centre of the right hind femur using sterile scissors. Afterwards, half were exposed to a 10 μL pathogenic solution by immersing their injury in a solution containing approximately. 105 P. aeruginosa bacteria (0.01 OD) for 2 seconds. In the control group, the wound was exposed to sterile PBS for 2 seconds. All the workers were then returned to their respective sub-colonies for observations (N=9 per group).
The experiment was also repeated with a higher infection dose, infecting ants with a ten times more concentrated P. aeruginosa solution (0.1 OD, approx. 106 P. aeruginosa bacteria). Observations followed the same procedure as above (N=9 per group). While these sample sizes are relatively small (reflecting the substantial effort of continuous manual ethogram annotation per individual), they were sufficient to identify clear overall patterns that support our conclusions. We therefore do not believe that larger sample sizes would have altered the main findings. However, as with any study based on limited replication, effect sizes near the margins should be interpreted with caution, and future work with larger datasets may be able to refine or detect subtler trends with greater statistical power.
Treated ants were observed in the nest for two weeks, recording the number of nearby ants, the treated worker’s condition (alive or dead), and overall colony mortality. Deceased ants were removed and inspected for limb amputation. Data on surviving ants and amputations were recorded every hour for the first 24 hours, every two hours for the remainder of the first week, and twice a day during the second week. Ant behaviour inside the nest was monitored using a Panasonic HC-X1000 4K video camera recorder mounted on top of the nest box. We analysed the first 6 hours of video recordings for each trial using VLC media player 3.0.16 Vetinari. We distinguished the following behaviours: (I) Wound care (injury is cleaned by another worker), (II) allogrooming (cleaning the injured worker), and (III) amputation (biting on the trochanter by a nestmate).
Survival experiments
To confirm the role of amputation in reducing mortality rates associated with injury and wound infection, we compared the survival of injured ants with or without amputation and infection. A total of 80 previously uninjured workers (from 4 different colonies) were divided into 4 treatments. The first group, labelled “Healthy,” served as a negative control and received no treatment (N=20). In the second group, “Sterile,” the right hind leg was severed at the centre of the femur using sterile scissors, and the wound was soaked in sterile PBS for 5 seconds (N=20). Workers in the “Infected” group were treated in the same manner as the “Sterile” group, but their wounds were soaked in a 0.01 OD P. aeruginosa solution for 2 seconds (N=20). The “Infected + Amputated” group underwent the same initial treatment as the “Infected” group, but after one hour, the leg was manually amputated at the trochanter using sterile scissors to simulate amputations performed by nestmates (N=20). All workers were kept in isolation chambers and provided ad libitum honey water and water. Isolation experiments were performed in parallel with a social experiment where 60 workers were collected from the same 4 colonies and received the same treatments as described above: (1) “Healthy”, (2) “Sterile”, and (3) “Infected”. All the workers were then put back into their respective sub-colonies (N=20 per group). The isolation and colony survival experiments were repeated, using a 0.1 OD P. aeruginosa solution (N=20 per group). Workers were monitored once per hour for 6 days (144h), and the time of death was recorded.
To determine the time window during which amputations by nestmates remained effective in reducing infection mortality, we manually amputated the infected leg of injured ants kept in isolation at different time points. 10 workers (N=10) were collected per treatment group from four colonies. After manually injuring the femur and exposing the wound to a P. aeruginosa solution (0.1 OD), we amputated the infected leg at 0-, 6-, 12-, or 24-hours post-exposure. Mortality was quantified for the first 6 days post-exposure.
To investigate whether the duration ants carried an infected or sterile injury influenced the frequency of amputations by nestmates, we introduced injured nestmates at different timepoints after injury and/or infection into sub-colonies. Individual workers were collected from each of the 4 colonies and received one of three treatments: (1) “Healthy”, (2) “Sterile”, or (3) “Infected”. Half of the workers were immediately returned to the nest after manipulation, and the other half were held in isolation chambers with water and honey water provided ad libitum for 24 hours before being reintroduced to the nest (N=10 per group and time-point).
Cuticular hydrocarbon analysis
A total of 70 workers were collected from one colony (N=70). Of these, 20 received a P. aeruginosa infected injury (OD 0.1), 20 received a sterile injury, and the remaining 30 did not receive any additional treatment as control (“healthy”). 10 individuals per group were frozen at -20°C, 2h, and 24h after manipulation, in addition to 10 healthy ants at 0h. Before freezing, the ants were kept alone in isolation in sterile petri dishes containing ad libitum water and honey water.