Introduction

Colony-level behavioural phenotypes are an important feature of eusocial insects (Jandt et al. Reference Jandt, Detoni, Loope and Santoro2020). Traits such as defensiveness and foraging intensity may vary with context, such as brood investment or weather conditions (Pinter-Wollman et al. Reference Pinter-Wollman, Gordon and Holmes2012; Brito et al. Reference Brito, Aragao and Santos2018). However, consistent differences in these traits can persist between colonies, and colony behavioural phenotype may be linked to fitness advantages in some eusocial insects (Wray et al. Reference Wray, Mattila and Seeley2011; Jandt et al. Reference Jandt, Detoni, Loope and Santoro2020).

Within eusocial wasps, two types of social parasitism occur: temporary and inquiline (Sledge et al. Reference Sledge, Dani, Dapporto and Turilazzi2001). Temporary social parasites kill the resident queen of a host colony and use the remaining workers to raise the first generation of brood (Borowski et al. Reference Borowski, Daglio, Jacobson, Loope and Walker2021). Inquiline social parasitism involves invasion of the nest without killing the resident queen and using chemical or behavioural mimicry to remain in the host colony (Casacci et al. Reference Casacci, Barbero, Ślipiński and Witek2021). Some wasp species exhibit both temporary and inquiline social parasitism depending on circumstances (Sledge et al. Reference Sledge, Dani, Dapporto and Turilazzi2001). Although social parasitism has been studied extensively in ants (Hymenoptera: Formicidae) (Borowiec et al. Reference Borowiec, Cover and Rabeling2021) and paper wasps (Hymenoptera: Vespidae: Polistes Latreille) (Cervo Reference Cervo2006), social parasitism in yellowjackets (Hymenoptera: Vespidae: Vespula Thomson) has received comparatively less attention outside phylogenetic and taxonomic contexts (Carpenter and Perera Reference Carpenter and Perera2006).

Yellowjackets are common across North America and Europe as both native and invasive species (Lester et al. Reference Lester, Haywood, Archer, Shortall and Sanders2017). As generalist foragers with behavioural plasticity, they provide ecosystem services as predators and decomposers in their native range (Wilson-Rankin Reference Wilson-Rankin2014; MacLean et al. Reference MacLean, Van Akker and Van Hezewijk2021). Despite their ecological importance, public opinion of yellowjackets is generally negative, and they are typically only studied in invasive-species or pest-control contexts (Lester et al. Reference Lester, Haywood, Archer, Shortall and Sanders2017).

Here, we report field experiments and immunohistochemical analyses of defensive behaviour in two yellowjacket host species associated with the facultative temporary social parasite Vespula flavopilosa Jacobson. Our study included six host colonies in urban greenspaces in Winnipeg, Manitoba, Canada: four V. pensylvanica Saussure and two V. maculifrons du Buysson colonies. Colonies were located by placing sliced bologna as protein bait in urban greenspaces and following foragers carrying bait back to nest entrances. Individuals of V. flavopilosa were detected in all six colonies.

We measured colony-level defensive behaviour using a disturbance assay with a strike-counting device modified from Jandt et al. (Reference Jandt, Detoni, Loope and Santoro2020). The decoy was a flat, circular, black plastic target, 30 cm in diameter and 3 cm thick, constructed from two disposable plastic plates clipped together. The decoy contained an electret microphone sound-detection module (TS-US-115_5; DAOKI, Shanghai, China) and was mounted on a tripod as close to the nest entrance as terrain permitted, with a straight-line distance of approximately 40–50 cm; exact distances were not measured. We disturbed the nest with a scratch stimulus using a metal fork. Wasp strikes to the decoy were detected by the module, whose digital output was read by a generic ATmega328P-based Arduino Nano-compatible microcontroller; timestamps of threshold-crossing events were written to a microSD card. The threshold was adjusted so that strikes to the decoy, but not ambient noise, triggered event recording.

We measured three variables for colony-level defensive behaviour (Jandt et al. Reference Jandt, Detoni, Loope and Santoro2020): latency from disturbance to first strike, response duration from first to last strike, and total strikes to the decoy. To reduce the likelihood of habituation, only one disturbance trial was performed per colony per week, from colony discovery to colony death (late July through early October 2024), with timing kept as consistent as possible for each colony between 11:00 and 16:00, local time. Temperature and humidity were recorded for each trial.

Before each disturbance trial, five wasps were sampled at the nest entrance, anaesthetised on ice, and identified via dichotomous key (Kimsey and Carpenter Reference Kimsey and Carpenter2012). We calculated “current” parasite proportion as the proportion of V. flavopilosa in that weekly five-wasp sample, and we calculated “cumulative” parasite proportion as the proportion of V. flavopilosa among all wasps sampled from that colony up to and including that date. Because the current proportion was based on a weekly entrance sample, it could vary among weeks with colony demography, entrance activity, or sampling variation.

We analysed behavioural data using linear mixed-effects models in R, version 4.5.2, with the packages car, version 3.1.3, lme4, version 1.1.38, and lmerTest, version 3.1.3 (Bates et al. Reference Bates, Maechler, Bolker and Walker2015; Kuznetsova et al. Reference Kuznetsova, Brockhoff and Christensen2017; Fox and Weisberg Reference Fox and Weisberg2019; R Core Team 2025). For each behavioural response variable, we first compared candidate models containing date, temperature, humidity, date + temperature, temperature + humidity, or temperature × humidity using Akaike’s information criterion (AICcmodavg package, version 2.3.4; Mazerolle Reference Mazerolle2023). The selected covariate was then included in the final model for that behavioural response: temperature for latency and date for both response duration total strikes. Current and cumulative V. flavopilosa proportions were tested separately, and the colony of origin was included as a random effect to account for repeated measures. The retained models tested latency with host species, cumulative V. flavopilosa proportion, and temperature; response duration with host species, current V. flavopilosa proportion, their interaction, and date; and total strikes with host species, current or cumulative V. flavopilosa proportion, and date.

Vespula flavopilosa specimens were found in each tested nest, and the proportion of V. flavopilosa in the nests tended to increase through the season (Fig. 1A). Response duration decreased as the season progressed (P = 0.032; Fig. 1B) and appeared to vary with current V. flavopilosa proportion differently in the two host species, increasing in V. maculifrons but decreasing slightly in V. pensylvanica (interaction: P = 0.050; Fig. 1C). Vespula maculifrons colonies also responded more quickly than V. pensylvanica colonies in the latency model that included cumulative V. flavopilosa proportion and temperature (P = 0.030), and total strikes decreased over the season in both species (P = 0.006). Humidity was evaluated in candidate models but was not retained in the final behavioural models.

Figure 1.

Temporal changes in Vespula flavopilosa proportion and defensive-response duration in host yellowjacket colonies: A, cumulative V. flavopilosa proportion through time in identified nests; B, defensive-response duration over the colony cycle; C, defensive-response duration in relation to current V. flavopilosa proportion. Defensive-response duration is natural log-transformed from seconds in panels B and C. Orange and blue points/lines denote V. maculifrons and V. pensylvanica, respectively

Because serotonin has been linked to aggressive and defensive behaviour in insects, including social hymenopterans, we examined serotonin immunoreactivity in the optic lobe, mushroom body calyces, and lateral protocerebrum, regions with serotonergic circuitry implicated in sensory and behavioural regulation, including aggression (Bubak et al. Reference Bubak, Yeager, Renner, Swallow and Greene2016; Kamhi et al. Reference Kamhi, Arganda, Moreau and Trainello2017; Nouvian et al. Reference Nouvian, Mandal, Jamme, Claudianos, d’Ettore and Reinhard2018; Alekseyenko et al. Reference Alekseyenko, Chan, Okaty, Chang, Dymecki and Kravitz2019; Aonuma and Benelli Reference Aonuma and Benelli2023).

Each specimen collected for species identification was decapitated, and its brain was removed, fixed in 4% formaldehyde, and stored in phosphate-buffered saline (PBS). Final usable sample sizes for microscopy varied by colony and brain region because not all regions were usable in every specimen (Supplementary material, Table S1). At the end of the season, stored brains were washed three times in PBS, embedded in 7% agarose, and sectioned coronally at 100 µm using a Compresstome vibrating microtome (Precisionary Instruments VF-210-0Z; Precisionary Instruments, Ashland, Massachusetts, United States of America). We selected the middle two sections of each brain for mounting and imaging. The two sections were washed three times in PBS with 0.2% Triton X (PBST), blocked with 5% bovine serum albumin for 1 hour at room temperature, and incubated at 4 °C for 48 hours on a platform rocker with rabbit anti-serotonin antibody (Sigma-Aldrich S5545; 1:267; Sigma-Aldrich, St. Louis, Missouri, United States of America).

Next, slices were washed six times in PBST and incubated at 4 °C for 48 hours on a platform rocker, protected from light, with goat anti-rabbit Alexa Fluor 488 secondary antibody (Invitrogen A11034; 1:667; Thermo Fisher Scientific, Waltham, Massachusetts) and Alexa Fluor 555 phalloidin (Invitrogen A34055; 1:667; Thermo Fisher Scientific). After washing six times in PBS, slices were mounted in Vectashield mounting medium with DAPI (Vector Laboratories VECTH1500; Vector Laboratories, Newark, California, United States of America) under #1.5 coverslips and sealed with clear nail polish.

Specimens were coded by a third person, and microscopy and image analysis were performed blind to specimen identity, colony, species, and behavioural data. Mounted brains were imaged on a Zeiss LSM 700 laser confocal scanning microscope (Zeiss, Oberkochen, Germany), with 30-slice z-stacks with 0.4-µm z-steps, collected at 40× magnification distance from the optic lobe medulla, mushroom body calycal lip, and lateral protocerebrum. For each brain region, a region of interest was deconvolved using DeconvolutionLab2 in ImageJ (https://imagej.net/ij/), background-subtracted using a Rényi entropy-based threshold, and analysed with ImageJ particle analysis to measure immunolabelled particle count, particle size, and percent area covered by immunoreactivity (Supplementary material, Figure S1; Schindelin et al. Reference Schindelin, Arganda-Carreras, Frise, Kayning, Longair and Pietzsch2012; Sage et al. Reference Sage, Donati, Soulez, Fortun, Schmit and Seitz2017; Jirakitpuwapat et al. Reference Jirakitpuwapat, Kumam, Deesuwan and Sanguansuttigul2020).

Microscopy data were analysed with exploratory linear mixed-effects models fitted separately for each brain region, immunoreactivity measure, and predictor, with colony of origin included as a random effect. Duplicate measurements from the same brain were identified after unblinding and removed to prevent pseudoreplication. The only species difference observed in microscopy was in serotonin particle size in the optic lobe medulla, with smaller particles found in V. pensylvanica than in the other two species (P = 0.031). In the medulla, greater serotonin particle number and percent area coverage were associated with longer response latency (P = 0.010 and P  = 0.045, respectively; Fig. 2A, B) and fewer total strikes (P = 0.021 and P = 0.006, respectively; Fig. 2C, D). Other microscopy results were less consistent: response duration was negatively associated with region of interest percent area coverage in the lateral protocerebrum (P = 0.025), whereas in the mushroom body calycal lips, V. flavopilosa proportion was positively associated with region of interest percent area coverage (P = 0.029), and latency was negatively associated with particle size (P = 0.004).

Figure 2.

Defensive behaviour in relation to serotonin immunoreactivity in the optic lobe medulla: A, latency versus average serotonin-immunoreactive particle count; B, latency versus percent region of interest area coverage; C, total strikes versus average serotonin-immunoreactive particle count; D, total strikes versus percent region of interest area coverage. Particle count and percent area coverage are natural log-transformed. Orange, blue, and purple points denote Vespula maculifrons, V. pensylvanica, and V. flavopilosa, respectively. Lines and shaded areas show pooled simple linear fits to the plotted data with 95% confidence intervals for visualisation

These results should be interpreted cautiously because of the small number of colonies, absence of unparasitised control colonies, and exploratory nature of the neurobiological comparisons. Because V. flavopilosa was present in every examined colony, we cannot determine whether defensive patterns reflect intrinsic host-species differences, effects associated with V. flavopilosa presence or proportion, or both. Cumulative V. flavopilosa proportions varied widely among colonies, but because our sampling was limited to one year and six colonies, we cannot determine whether prevalence was unusually high or whether annual conditions contributed to the pattern. Nevertheless, these observations identify host-specific defensive responses and serotonergic correlates as targets for future work.

The observation that V. flavopilosa proportion in some colonies plateaued at low levels rather than continuing to rise may indicate the continued production of host workers by a living host queen (Fig. 1A). Under temporary social parasitism, the host queen is typically killed, host workers are no longer replaced, and the relative proportion of parasite-produced individuals would be expected to increase as the season progresses. By contrast, if the host queen remains alive, continued emergence of host workers could maintain a lower, more stable V. flavopilosa proportion, a pattern consistent with inquiline social parasitism. This interpretation remains speculative, however, because queen presence, brood composition, and colony demography were not directly measured, and inquilinism has not previously been observed in this species (Borowski et al. Reference Borowski, Daglio, Jacobson, Loope and Walker2021).

Differences between host species were observed for both latency and duration, with V. maculifrons responding more quickly to disturbance. Defensive behaviour was also related to current V. flavopilosa proportion, host species, and date, with a weaker host × V. flavopilosa proportion (parasite proportion: t ₃₀ = 2.75, P = 0.010; host species: t ₃₀ = 2.47, P = 0.019; date: t ₃₀ = −2.15, P = 0.040 interaction: t ₃₀ = −2.04, P = 0.050; Fig. 1B, C). Specifically, response duration tended to increase with current V. flavopilosa proportion in V. maculifrons colonies but decrease slightly in V. pensylvanica colonies (Fig. 1C). Vespula flavopilosa proportion was substantial in both V. maculifrons colonies (0.52 and 0.91) and ranged from 0.23 to 0.60 across V. pensylvanica colonies, providing some ambiguity over whether V. maculifrons or V. flavopilosa is the more defensive species (Fig. 1A). Colony senescence likely played a role in defensive behaviour, with both total strikes and defence duration decreasing as the season progressed. Because colony size and brood area were not measured, this seasonal decline could reflect reduced brood investment, worker decline, or other aspects of the yellowjacket annual cycle (Vetter and Visscher Reference Vetter and Visscher1997).

These results also represent, to our knowledge, the first confocal microscopy analysis of yellowjacket brains. In the optic lobe medulla, several measures of defensive behaviour appeared to decrease as serotonin immunoreactivity increased (Fig. 2). However, this relationship was correlational, serotonin immunoreactivity was measured in individuals collected before the behavioural assay, and other uncontrolled factors may have contributed to this pattern (e.g., colony identity, individual age, colony age, colony size, or time of year). Therefore, these results should be interpreted cautiously but are consistent with the broader view that serotonergic modulation of insect aggression can be context-dependent (Bubak et al. Reference Bubak, Watt, Yeager, Renner and Swallow2020). Results with the mushroom body calycal lips, particularly with respect to particle size, were more consistent with previous work that suggests a role for mushroom body serotonergic regulation in aggression (Alekseyenko et al. Reference Alekseyenko, Lee and Kravitz2010). Further studies should standardise and record decoy distance, measure colony size and brood investment directly, and compare parasitised and unparasitised colonies. Such comparisons would help test whether these behavioural and serotonergic patterns are consistently associated with social parasitism.