Age-dependent release of and response to alarm pheromone in a ponerine ant

Communication is essential to social insect societies, and chemical signalling is ubiquitous in governing individuals' behaviours and the overall organization within the colony (Leonhardt et al., 2016; Richard and Hunt, 2013). Such signals contribute to the maintenance of the social structure in many ways, including the suppression of worker reproduction by queen pheromones (Holman et al., 2010; Le Conte and Hefetz, 2008; Oi et al., 2016) and policing among workers based on recognition of worker-laid eggs (Endler et al., 2004; Oi et al., 2015; Smith et al., 2009). Chemical communication can also be vital for the survival and coherence of the colony. Trail pheromones enable recruitment to valuable resources (Hölldobler and Wilson, 1990; Jarau et al., 2010; Morgan, 2009), cuticular lipid profiles play a role in nestmate recognition (Pradella et al., 2015; Sturgis and Gordon, 2012; van Zweden and d'Ettorre, 2010), and alarm pheromones alert workers to danger (Hölldobler and Wilson, 1990). Alarm pheromones are common among social insects (Blum, 1969; Leonhardt et al., 2016) and workers' responses to such signals vary depending on species, caste and context (Hölldobler and Wilson, 2009; Schorkopf, 2016; Sasaki et al., 2014). In general, when alarmed individuals react by showing escape behaviours or relocating the brood, the reaction is termed panic alarm. If an alarm pheromone elicits aggressive alarm, individuals will move to confront the disturbance, prepared to attack (Wilson and Regnier, 1971).

Among the social Hymenoptera, ants are known to employ a remarkable diversity of chemical compounds for communication (Jackson and Morgan, 1993; Richard and Hunt, 2013). Ponerine ants, despite being thought to generally exhibit a more simple social structure (Peeters, 1997), are no exception and have been found to employ an array of chemical signals for essential social functions such as recruitment (Attygalle et al., 1991; Duncan and Crewe, 1994; Hölldobler et al., 1994; Leal and Oliveira, 1995; Witte and Maschwitz, 2000) and policing (Gobin et al., 1999; Kikuta and Tsuji, 1999), and as alarm pheromones (Longhurst et al., 1978, 1980; Wheeler and Blum, 1973; and see Schmidt and Shattuck, 2014).

The genus Platythyrea (Ponerinae, Platythyrini) appears to be exceptionally poor at producing and utilizing exocrine secretions. Of the 38 described species, there is only one report of characterized chemical compounds in an exocrine secretion, present in Dufour's glands of Platythyrea cribinodis, while no volatile compounds were detected in the venom gland of that species (Morgan et al., 2003). A related species, Platythyrea punctata (F. Smith 1858), possesses neither subepithelial glands (Gobin et al., 2003) nor a Dufour's gland, and samples taken from the venom gland and sternal glands were devoid of volatile material (Morgan et al., 2003). Nonetheless, we found this species to occasionally emit a lemon-like odour upon disturbance, indicating the presence of a volatile alarm pheromone. Apart from being a species for which no volatile signals have yet been reported, P. punctata is interesting to study because it is monomorphic and reproduces most often via thelytokous parthenogenesis (Hartmann et al., 2005; Heinze and Hölldobler, 1995). Thus, P. punctata is an ideal model organism for studies on age polyethism, as genetic effects within a colony can be expected to be minimal (Bernadou et al., 2015). The term age polyethism describes the phenomenon that age influences task allocation within the colony (Wilson, 1971; Robinson, 1992; Camargo et al., 2007). Younger individuals perform predominantly safe tasks such as nursing within the nest, and switch to more risky tasks outside of the nest, like foraging or guarding the nest, with increasing age (Wilson, 1971; Oster and Wilson, 1978; Seeley, 1982; Moroń et al., 2008; shown for P. punctata by Bernadou et al., 2015). Age polyethism, however, can not only govern the division of tasks but may also include differences in behavioural reactions to social signals (Cammaerts, 2014), as has been shown for morphological castes (Hölldobler and Wilson, 1990; Hughes and Goulson, 2001; Simola et al., 2016).

In this study, we aimed to comprehensively elucidate the putative alarm pheromone of P. punctata by characterizing the compound(s) underlying the lemon-like odour, identifying the tissue of origin and determining the quantity of the compound(s) therein. We then investigated the behavioural effect of the compound(s), and tested the hypotheses that there is an age effect on both the production and emission of the compound(s), and that age influences the behavioural reaction towards the putative pheromone.

All experiments were conducted with 10 colonies of P. punctata that were originally collected in Puerto Rico in 2012 (El Verde Field Station, El Yunque National Forest). They were kept in plastic boxes (20 cm×20 cm, 8.5 cm height, bottom covered with a layer of plaster of Paris) at 22–26°C and 75% humidity under a 12 h light:12 h dark cycle. The nest site in each box consisted of a chamber in the plaster floor that was covered with glass slides and kept dark by a sheet of black plastic on the glass. Water was provided ad libitum by means of a cotton wool-plugged plastic tube positioned in the box. Small cockroaches, fruit flies and diluted honey were provided as food 3 times a week after moistening the plaster floor to maintain high humidity in the boxes.

Colonies typically contained one, occasionally several, dominant reproductive individuals, with all other individuals acting as workers. Task allocation is influenced by an individual's age: young intranidals of P. punctata (YI, individuals with tan to brown coloration, aged <3 weeks) typically remain in the nest where they care for the brood, while foragers (FG, distinguishable from YI by the fully melanized black cuticula) are predominantly found outside the nest (Bernadou et al., 2015). All 10 colonies were tested in each of the experiments described below, providing n=10 replicates for every assay.

The results of headspace analyses (see ‘Pheromone emission’, below) supported the notion that the lemon-like odour is due to the putative alarm pheromone of P. punctata. In several other ponerine species, the mandibular glands are the source of alarm pheromone (Duffield and Blum, 1975; Longhurst et al., 1978), and we therefore considered the mandibular glands to be a likely source for the production and storage of this pheromone in P. punctata. In a preliminary analysis, we dissected a total of eight mandibular glands from FG that had been freeze-killed at −21°C. The glands were extracted singly in 20 µl of dichloromethane (DCM) containing 10 ng µl⁻¹ methyl undecanoate as an internal standard (IS) for a minimum of 1 h, and the extracts were then subjected to chemical analysis by gas chromatography and coupled mass spectrometry (GC/MS) as described below. The mandibular gland extracts were found to contain citronellal (a monoterpene aldehyde, the same compound that was detected in the headspace analyses) and, additionally, actinidine (a monoterpene alkaloid). For the quantification of mandibular gland contents, we used ants from each of the 10 colonies. Single ants cannot easily be retrieved from their colony without disturbance (potentially causing them to emit alarm pheromone). Therefore, a minimum of four P. punctata workers were carefully isolated from their colony, and kept together in a Petri dish for 17–24 h to allow for some replenishment of pheromone content. The dish contained a water source (moistened tissue paper) and a shelter (folded piece of red plastic). Ventilation was ensured by a gauze-covered opening in the lid of the dish. The Petri dishes were kept in the same climate chamber as the colony boxes. FG and YI from each colony were kept together in their respective Petri dish. After the isolation period, the ants were flash-frozen by immersion in liquid nitrogen and subsequently kept at −21°C until dissection of mandibular glands. For all 10 colonies, we analysed one mandibular gland each from one FG and one YI. A sample consisted of one intact mandibular gland still attached to the mandible (Fig. 1C) extracted in 20 µl of DCM containing 10 ng µl⁻¹ methyl undecanoate as IS for at least 3 h before chemical analysis (see below).

Extracts from the mandibular glands and the headspace analyses (see ‘Pheromone emission’, below) were analysed on a Shimadzu QP2010 SE GC/MS equipped with a polar BP20 column (30 m length, 0.25 mm inner diameter, 0.25 µm thickness, SGE Analytical Science/Trajan Scientific, Melbourne, VIC, Australia) with an injection volume of 1 µl in splitless mode using helium as carrier gas (linear velocity 40 cm s⁻¹). The injection temperature was 240°C, and the oven was programmed to increase in temperature from 60°C to 240°C at 5°C min⁻¹. The final temperature was held constant for 10 min. Chromatograms were analysed using the software package GCMS solution version 4.11 (LabSolutions, Shimadzu, Kyoto, Japan). Compounds were identified by comparison of mass spectra and retention indices with those of synthetic standards. Quantification was accomplished by relating the peak areas of pheromone components to the peak area of the IS (10 ng µl⁻¹).

For the determination of the absolute configuration of the pheromone compounds, the GC was equipped with a Beta-Dex 225 column (30 m length, 0.25 mm inner diameter, 0.25 µm thickness, Supelco/Sigma-Aldrich, St Louis, MO, USA). Samples were injected splitless at a temperature of 220°C. The temperature program started at 60°C (held for 3 min) and was raised at 1°C min⁻¹ to 100°C, which was held constant for 50 min. Subsequently, the temperature was raised to 220°C at 5°C min⁻¹ and held at the final temperature for 10 min. Injection volume, carrier gas and velocity were the same as described above. The retention times of the peaks detected in an extract of six pooled mandibular glands were compared with those of a racemic mixture of both citronellal and actinidine, as well as with those of a mixture of (S)-(−)-citronellal (SC) and (S)-(−)-actinidine (SA). The authentic standard of SC was purchased (Merck, Darmstadt, Germany) and that of SA was prepared as described previously (Beckett et al., 2010).

Seventeen to 24 h before being tested, P. punctata workers were carefully isolated from their colony to allow for potential replenishment of pheromone content. FG and YI from each colony were again kept together in their respective Petri dish, and tested on the same day.

As volatile emission had previously been observed upon disturbance, we tested the effect of physical distress on volatile release using dynamic headspace analyses (purge-and-adsorbent trapping; Asfaw et al., 2018). Experiments were conducted at room temperature (around 21°C) with a custom-made sampling apparatus consisting of a 50 ml Erlenmeyer flask that contained a clean magnetic stir bar (8 mm×40 mm, Carl Roth, Karlsruhe, Germany), fitted with a wash bottle insert. The apparatus consisted exclusively of glass, stainless steel and Teflon parts in order to minimize contamination. Prior to every headspace sampling, all parts were cleaned with ethanol and deionized water and then heated for a minimum of 60 min at 100°C in a convection oven (VWR VENTI-Line VL 115). Incoming air was filtered through a glass tube containing 150 mg of activated charcoal (Supelco Orbo 32, small, Sigma-Aldrich, Taufkirchen, Germany). Volatile compounds from the headspace were purged for 15 min with an airflow of 60±10 ml min⁻¹ and trapped on commercially available adsorption tubes (65 mm length×5 mm inner diameter, 5 mg activated charcoal, Gränicher and Quartero, Daumazan, France). Trapped volatiles were eluted with 40 µl DCM containing 10 ng µl⁻¹ methyl decanoate as IS and subjected to analysis by GC/MS as described above.

For each replicate, three FG or three YI were carefully transferred to the Erlenmeyer flask. To keep the ants in the flask, the glass tube reaching into the Erlenmeyer flask as well as the top of the flask were blocked off using cleaned stainless steel mesh and Teflon tape. As a control, headspace was sampled for the three ants with the enclosed stir bar remaining motionless. After 15 min, the adsorption tube was replaced by a new one and the same three ants were sampled again. Now, using a magnet on the outside of the flask to move the enclosed stir bar, physical disturbance was elicited by deliberately bumping the stir bar into the ants for the duration of volatile collection (15 min).

We investigated the detection efficacy of the headspace setup after finding that the headspace analyses yielded SC only, while two compounds, SC and SA, could be detected in the mandibular gland extracts. Thus, 750 ng synthetic SC and 385 ng synthetic SA (amounts based on the preliminary quantification of one gland equivalent), each dissolved in DCM, were applied onto the same piece of clean filter paper (disc of 6 mm diameter), which was then tested in the headspace sampling setup for 15 min using the same experimental parameters as described for the tests with ants. Replicates (n=10) were always conducted with freshly prepared chemical-treated filter paper discs.

We evaluated the reaction of P. punctata towards the pheromone, using five different treatments: (1) mandibular gland extract (amount equivalent to one mandibular gland), (2) a synthetic pheromone mixture consisting of a combination of 480 ng synthetic SC and 350 ng synthetic SA (amounts based on the equivalent of one mandibular gland as estimated from a pooled extract of five mandibular glands), (3) 480 ng synthetic SC alone, or (4) 350 ng synthetic SA alone; we additionally tested a control treatment (5), using pure solvent (DCM). Each assay was conducted with all 10 colonies before starting the next. Before testing, a freely accessible test arena was set up for each of the 10 colonies. Every arena was placed in the box of the respective colony for a minimum of 17 h prior to testing in order to allow the ants to familiarize themselves with the arena. The arena consisted of a glass plate (49 mm×49 mm) with a roughened surface, covering an equally sized piece of paper. The paper was printed with concentric rings (hereafter referred to as ‘lines’), dividing the area into different zones bordered by the respective lines. The lines and zones were labelled in the order: outer line (diameter d=48.5 mm), outer zone, middle line (d=34.5 mm), middle zone, centre line (d=20 mm) and centre zone. An additional ring (d=6 mm) printed around the centre point served to guide the placement of test compounds (see Fig. 3). Extracts and chemicals were applied to a clean filter paper disc (d=6 mm). After evaporation of the solvent (minimum of 5 s, until the paper appeared dry), the respective filter paper was mounted on the centre of the glass plate using a small piece of double-sided adhesive tape (about 3 mm×3 mm, Tesa SE, Hamburg, Germany) to prevent the ants from moving it from the original position.

For each replicate, the respective colony box was carefully placed under a video camera (Canon EOS M equipped with a Tamron 90 mm F/2.8 macro lens, recording from the top) in the climate chamber, and illuminated indirectly. Large pieces of white cardboard screen surrounded the setup to minimize distraction of the ants by external optical stimuli. Recording began 1 min after the box was set up. Introduction of the first filter paper (colony ‘baseline’ control, always the pure solvent: DCM) took place 3 min after the start of the recording, and the recording continued for another 3 min. After the recording was stopped, the first filter paper was removed, and the ants were given 3 min to settle. Subsequently, a filter paper impregnated with the respective test compound was introduced, and the second recording was started. The control treatment using pure DCM as test compound served to verify that the second placement of a filter paper per se would not influence the behaviour of the ants. Videos were analysed with The Observer XT 11 software (Noldus Information Technology, Wageningen, The Netherlands) in offline mode. Because of the volatility of the alarm pheromone, video analyses of colony assays covered the first minute after placement of a filter paper in the arena. The first 10 s were not included for analyses to avoid recording behaviour caused by the manipulation (introducing the filter paper), and to allow an odour gradient to develop. Two parameters were assessed: the number of times that ants crossed the printed lines (as a proxy for movement activity) and the number of ants present in each zone (a measure of ant density). An ant was defined to be present in a zone when more than half of its body was in the respective zone. Likewise, the crossing of a line was defined to have taken place when an ant moved from one zone to another in the designated time frame. The number of crossings over each of the three lines was documented for two consecutive seconds (to be able to assess the ants' movements) every 10 s, i.e. from 10th to 12th second, from 20th to 22nd second, from 30th to 32nd second, from 40th to 42nd second and lastly from 50th to 52nd second, resulting in a total observation time of 10 s. Density counts were taken by freezing the video every 5 s and counting the number of ants present in each zone, starting at the 10th second and ending with the last count at the 60th second, resulting in a total of 11 count events. Counts were taken without regard to ant identity, i.e. an ant present in the same zone during two consecutive density counts was counted once for each time point. The same principle was applied for the line crossings, with all crossings of every ant during the measuring time span being counted. The results were normalized by dividing the number of crossings over a line by its respective length, and by dividing the total number of ants for one zone by the respective zone's area.

YI are rarely seen outside the nest under normal circumstances, and never ventured out of the nest cavity during bioassays on the colony level. Thus, the setup had to be modified to be able to test whether P. punctata shows age polyethism in the reaction to the pheromone.

The same arena as described before was placed in the colonies for a minimum of 17 h prior to testing, though now it was additionally positioned on top of a larger glass plate (12 cm×12 cm). As for the colony assays, this was done to let ants familiarize themselves with the arena. As YI do not usually leave the nest and the arena would therefore be novel to them, this allowed the arena to acquire at least some of the colony-typical odour left by colony members. The whole construction was taken out of the colony box directly before testing, carefully removing ants currently present on it, and positioned under the video camera in the same fashion as for the colony assays. An outer barrier (PE, 4.1 cm high with rounded corners, 6.6 cm×6.6 cm) coated with Fluon (PTFE dispersion, Whitford GmbH, Diez, Germany) was positioned around the arena to prevent ants from escaping. The barrier was fixed in place from the outside using glue pads (patafix, UHU GmbH & Co KG, Bühl, Germany). Three ants of the same age class (either FG or YI) were carefully taken from the colony using flexible forceps and placed within the barrier directly prior to the experiment. Video recording was conducted as described for the colony-level assays, with the difference that the time between the end of the first video (baseline control) and the onset of the test video was reduced to 1 min, as the removal of the filter paper did not cause any disturbance. Only the synthetic pheromone (mixture of synthetic SC and synthetic SA) was tested in this setup.

Video analyses for age group bioassays were conducted in a similar fashion to that described for the colony-level assays, but were adapted for the lower number of ants. The observation period lasted 15 s after the introduction of the filter paper, omitting the first 1.5 s. Line crossings were documented continuously from 2 s onwards, resulting in a total of 13 s. Density counts were taken every 1.5 s with the final count at the 15th second, resulting in a total of 10 count events. The results were normalized as described for the colony assays.

Ant housing, treatment, and all experiments conformed to the relevant regulatory standards.

Statistical analyses were conducted with R version 3.6.1 (https://www.R-project.org). Wilcoxon paired tests were employed to compare the paired data of the headspace assays between the control situation and during physical distress for the peak areas of SC for FG and YI. Analyses of age polyethism were conducted with Wilcoxon paired tests for pheromone quantification (amounts of the single compounds as well as their ratio to each other), and using the values resulting from the test situation for FG and YI in the headspace analyses of pheromone emission under physical distress. Data for FG and YI were considered paired by colony origin.

The behavioural experiments testing the ants' reaction to the pheromone (for both colony and the age groups) were analysed using linear mixed effect models (LMM, R packages ‘lme4’: Bates et al., 2015; ‘car’: Fox and Weisberg, 2019). Movement activity and ant density were analysed separately, for both the colony assay and the age group assay. To test for the effect of treatment (colony assay, test compounds: mandibular gland extract, synthetic pheromone, SC, SA and DCM control) or age (age group assay: FG and YI), and the three lines/zones (outer, middle and centre) on the response of the ants, we first calculated the response variables by subtracting each baseline control value from the respective movement/density value. We always started with a full model, including lines/zones and treatment or age as fixed effect factors, as well as the interactions between these factors. Colony identity was included as random factor. We obtained minimal models by a stepwise backward elimination procedure (successively removing non-significant terms from the models), and comparing the nested models with a maximum likelihood method. The minimal models were then estimated using a restricted maximum likelihood method (Pinheiro and Bates, 2000). To comply with the assumption of normality and homoscedasticity as much as possible, data were transformed where necessary: first a fixed value was added to the response variables (raising the lowest value to 1) and the data were subsequently log₁₀ transformed. Post hoc multiple comparison procedures were conducted using Tukey all-pair comparisons with Bonferroni–Holm correction (R package ‘multcomp’: Hothorn et al., 2008).

The mandibular gland extracts contained two volatile compounds, SC and SA (Fig. 1A). The absolute configuration was determined by comparison with synthetic compounds (Fig. 1B). A single mandibular gland of FG contained on average 840±390 ng of SC and 510±220 ng of SA (Fig. 2A). Mandibular glands of YI contained significantly less of both: 340±120 ng SC and 140±70 ng SA per gland (Wilcoxon paired tests: v=51, P=0.014 for SC; v=0, P=0.006 for SA; Fig. 2A). Additionally, the ratio of the two compounds differed between the age classes (Wilcoxon paired test: v=0, P=0.006), with YI having relatively more SC (ratio of SC to SA: FG: 1.6±0.2, YI: 2.6±0.8; Fig. 2A).

When testing for pheromone emission with dynamic headspace analyses, SC could be detected in only one case (7 ng) in the control situation for FG. Upon physical distress, however, they released this compound in significantly larger amounts (170±300 ng in 7/10 replicates; Wilcoxon paired test: v=0, P=0.022; Fig. 2B). YI emitted SC in two out of 10 cases during the control setup (6 ng and 10 ng). In the physical distress situation, the pheromone was released in all 10 replicates, and in significantly higher amounts than during the control (540±370 ng, v=0, P=0.002; Fig. 2B). In comparison to the tested FG, YI tended to emit more SC (v=9, P=0.064; Fig. 2B). Headspace analyses testing the recovery rate for both compounds found in the mandibular glands could not recover any of the 385 ng of synthetic SA in any of the 10 replicates. Of the 750 ng of synthetic SC used in each replicate, 15 min of headspace sampling yielded on average 270±30 ng (recovery rate: 36%).

In the colony-level bioassays, the response values of the DCM control treatment did not differ from zero (for both movement activity and ant density), i.e. there was no change of behaviour in response to pure solvent (movement–0: z=0.367, P=0.999, density–0: z=0.355, P=0.999; Figs 3 and 4A, Table 1). All other treatments differed significantly from zero (Table 1), indicating that mandibular gland extract, as well as the synthetic pheromone mixture and synthetic SC or SA alone led to an increase of ant activity (Fig. 3). For further LMM analyses, the control treatment was removed from the dataset. The minimal model for movement activity included both fixed factors as significant (treatment: χ²=29.204, d.f.=3, P<0.001; line: χ²=9.328, d.f.=2, P=0.009). Pairwise comparisons revealed that the behavioural effect of SA alone was less pronounced than in the other treatments (mandibular gland extract–SA: z=4.885, P<0.001; synthetic pheromone–SA: z=3.640, P=0.001; SC–SA: z=3.861, P<0.001; Figs 3 and 4A, Table 1). Additionally, the reactions differed for lines between outer and centre (z=−2.716, P=0.019; Figs 3 and 4A, Table 1), suggesting a decreasing response with increasing distance from the odour source. For ant density in colony-level bioassays, only treatments had a significant effect, with reactions to the SA treatment being less distinct than those to mandibular gland extracts (z=−3.520, P=0.003) and synthetic SC (z=−2.809, P=0.025; Figs 3 and 4B, Table 1). In all but the control assay, the average behavioural change, as measured by the difference of movement activity and density between the baseline control and test situation was pronounced (Movie 1), and tended to be higher the closer the respective zone/line was to the source of the test stimulus (Figs 3 and 4).

In the age group bioassays confronting three ants with the mixture of synthetic SC and SA, behavioural responses did not differ from zero for ant density (FG–0: z=−1.303, P=0.592; YI–0: z=0.0, P=1). For movement activity, however, behavioural reactions of FG were significantly different from zero (z=3.765, P<0.001), while YI showed only a trend to move more over the lines (z=2.391, P=0.056; Table 1). The minimal model included both factors age class (χ²=3.844, d.f.=1, P=0.049; Fig. S1) and line (χ²=6.317, d.f.=2, P=0.042) as significant, with the reaction differing between outer and centre line (z=−2.716, P=0.019; Fig. S1; Table 1). An ant crossing the centre line in the test situation was observed in five of the replicates for YI while FG did so in all 10 cases. In general, YI were less active than FG and tended to remain huddled together beyond the outer line, showing hardly any movement. Nonetheless, a reaction (rapid antennal movements and sometimes single ants beginning to move around) could be observed in most cases upon exposure to the pheromone.

Platythyrea punctata workers released their volatile pheromone only upon infliction of physical distress, and the source of the pheromone was found to be the mandibular gland. Additional tests indicated that P. punctata does not emit the pheromone upon contact with prey items (not to familiar, unfamiliar or chemically defended prey, as might be expected if the pheromone had a general alerting purpose; Table S1). Our bioassays in ant colonies showed that SC and SA elicit an increase in movement activity and lead to a higher density of ants near the point of pheromone release. We thus conclude that SC and SA function as an alarm pheromone in P. punctata.

The mandibular gland is the source of alarm pheromones for several other ant species in different subfamilies (e.g. Ponerinae: Duffield and Blum, 1975; Longhurst et al., 1978; Formicinae: Bradshaw et al., 1979; Witte et al., 2007; Myrmicinae: Norman et al., 2017; Sasaki et al., 2014; Vander Meer et al., 2010; and see Hölldobler and Wilson, 1990). Both citronellal and actinidine have previously been reported as alarm/alerting pheromone components in ants. In Acanthomyops claviger, citronellal was found to be present in the mandibular glands and, in combination with the contents of Dufour's gland, to function as an alarm pheromone (Regnier and Wilson, 1968; Wilson and Regnier, 1971). Actinidine has been reported to be present in the pygidial gland contents of Megaponera foetens. It was suggested to have an alerting function, as more ants left a nest entrance treated with this compound than one treated with a solvent control (Janssen et al., 1995). In both cases, however, the absolute configuration of the compounds was not determined. The absolute configuration may be of importance as the behavioural reactions towards pheromone enantiomers and/or a racemic mixture can differ (see Riley et al., 1974; Wojtasek et al., 1998; Yu et al., 2014). In the present study, we demonstrated that P. punctata produces the (S)-enantiomers of both pheromone compounds. The dynamic headspace analyses yielded only SC, and the same was found in the recovery experiments with synthetic chemicals, clearly demonstrating that SA in natural doses was not detectable with this method. This was probably due to the compound's polar properties, which could lead to adsorption on the glass surfaces in the experimental setup.

The results of the colony-level bioassays showed that P. punctata reacts to both compounds of its mandibular pheromone (singly and in a mixture) in the same way. The ants showed increased movement and aggregated near the point of pheromone release. The reaction to synthetic SA alone, however, was less pronounced than the reaction to synthetic SC or a mixture of the two compounds. This could be due to the lower concentration of SA used in the assays (350 ng versus 480 ng of synthetic SC or the total of 830 ng in the pheromone mixture). Further studies would be needed to assess the potential effects of pheromone concentration on the behavioural reaction. The fact that the behavioural reaction of P. punctata to synthetic SC did not differ from that to the mixture of SC and SA indicates that there is no additive effect of the two single compounds. This is in contrast to reports on Formica rufa (Löfqvist, 1976), Atta colombica and two Acromyrmex species, where an extract containing all pheromone compounds elicited a stronger reaction than the most active single compound on its own (Norman et al., 2017). Multi-component alarm pheromones are common in ants (see Keeling et al., 2004) and other Hymenoptera (Brushini et al., 2010; Keeling et al., 2004; Slessor et al., 2005). In some cases, the various compounds have different functions that are related to their volatility, and thus come into play at different distances from the source (Bradshaw et al., 1975) or at different times after emission (Fujiwara-Tsujii et al., 2006). Although the vapour pressures of SC (0.2866 hPa at 25°C) and SA (0.1933 hPa at 25°C; both values from The Good Scents Company Information System: http://www.thegoodscentscompany.com/index.html) differ, the observed behavioural reactions were similar for the two compounds in the present study. Further studies could investigate whether the two alarm pheromone components differ in their effect over larger distances or with respect to time after emission. Furthermore, the potential of enantiospecificity in the alarm pheromone response of P. punctata deserves further investigation.

Species specificity is generally considered to be important for many types of pheromones, and can be achieved through unique compounds or specific mixtures of more common compounds (Wyatt, 2014). Alarm pheromones, however, can lack that characteristic. This is assumed to be caused by a comparatively low selection on specificity of alarm signals, as species can benefit from responding to alarm pheromones of sympatric species (Vander Meer and Alonso, 1998). It has also been found that congeneric species may produce the same main alarm compounds (e.g. eight species of Atta; see Norman et al., 2017). Although nothing is known about the alarm pheromones of other species of Platythyrea, two facts suggest that the alarm pheromone of P. punctata is not species specific. First, the two compounds have been reported for other ant species (although stereochemistry was not resolved: Regnier and Wilson, 1968; Wilson and Regnier, 1971; Janssen et al., 1995; see http://www.pherobase.com), and thus are probably not unique to P. punctata. Second, the reaction of the ants was comparable between SC alone and the mixture of SC and SA, i.e. the second compound does not seem to be required to elicit the full response. The effect of the alarm pheromone in bioassays with whole colonies was clearly visible (Movie 1). Ants were additionally observed to open their mandibles while approaching the pheromone-treated filter paper (but not the solvent control), making rapid sideways head movements with raised antennae as if searching for something to attack. In contrast, reactions of groups of only three FG were much less pronounced. This difference may not only be due to the higher number of ants in the colony, and thus a clearer increase in the reaction when exposed to the pheromone, but also be a consequence of context. Ants may react less strongly to alarm pheromones when fewer nestmates are around (Sakata and Katayama, 2001) or when they are not in the vicinity of the nest (Parmentier et al., 2015; Velásquez et al., 2006). It is also possible that within a colony the chemical alarm signal is reinforced by some of the newly approaching individuals (perhaps after contact with already alarmed individuals). The resulting increase in pheromone concentration could then lead to more pronounced or more prolonged reactions. Indeed, in the colony-level bioassays, the increased overall activity after pheromone exposure could be observed for a longer period of time than in assays with only three ants. A further explanation for a lower reaction in small groups compared with that at the colony level might be the subset of tested individuals itself. A study with Rhytidoponera confusa revealed that aggression was shown by only a small number of ants from a colony (Crosland, 1990). Similarly, a study on aggression in P. puncata found that individual ants varied in their level of aggression. While a minority of workers was involved in most aggressive interactions, several individuals avoided fighting altogether (Barth et al., 2010). The subsets of ants in the group bioassays may thus have contained relatively more individuals showing a less pronounced reaction/higher response threshold to the pheromone, while in the complete colony, many of the most responsive ants would have been attracted to the compounds presented in the arena and thus included in the counts.

We showed that P. punctata exhibits age polyethism not only in relation to performed tasks (Bernadou et al., 2015) but also in the production of and reaction to its alarm pheromone. YI had significantly less alarm pheromone in their mandibular glands than FG, and the ratio of the two compounds differed between the age classes. It seems highly unlikely that the compound ratios differ as a result of selective emission of one of the compounds. Rather, this effect may be due to a difference in onset times or rates of biosynthesis for the two compounds, as has been shown for Dufour's gland components in Formica sanguinea (Fadl Ali et al., 1988). Our results suggest that young workers begin their adult lives with little or no alarm pheromone, as has been found for the compounds of the poison gland in Harpegnathos saltator (Haight, 2012). Despite having less pheromone in their mandibular glands, YI tended to release both a higher total amount and a higher proportion of the available pheromone upon physical distress than FG. When taking into account the average amount of SC in one mandibular gland and the recovery rate of this compound with the dynamic headspace analysis, YI emitted on average about 74% of their total SC when under physical distress, while the tested FG emitted only 9% of their available pheromone. Such an age-related difference in the propensity for release of alarm pheromone could be adaptive, as YI, under natural circumstances, would only come into contact with an aggressive alien intruder if that ant had already reached the nest chamber, probably posing a severe threat to the colony. Whether a higher pheromone concentration actually elicits a stronger response, remains to be shown. Additionally, experience might influence how much alarm pheromone is emitted in a certain context and could also explain the lower amounts of pheromone released by FG.

Reaction to the alarm pheromone in colony-level bioassays was shown exclusively by FG. Nest defence is often performed by older ants, a typical consequence of age polyethism in which more risky tasks are taken up only later in life (Wilson, 1971; Oster and Wilson, 1978; Seeley, 1982; Moroń et al., 2008). Our results follow the prediction that it is the older workers that show a more pronounced reaction to the alarm pheromone (see Norman et al., 2017; Cammaerts, 2014; but see Parmentier et al., 2015). Nonetheless, there were some YI that reacted to the alarm pheromone with increased movement. It would be interesting to study whether such a higher propensity to react to alarm pheromone could be a permanent individual trait (e.g. a lower than average response threshold) that remains enhanced in relation to others from the same age cohort. However, the reaction of YI might also be more likely in ants from certain colonies (i.e. a genetic effect) or dependent on colony size.

To our knowledge, this is the first study to demonstrate the composition and effect of an alarm pheromone in the genus Platythyrea, and the first report of a volatile pheromone in P. punctata. The ants exhibited age polyethism not only in regard to pheromone emission but also in their behavioural reaction. A valuable extension of this work would be an investigation of the influence of pheromone concentration on behaviour, and determination of whether differing response thresholds or other physiological factors explain the observed behavioural age polyethism.

We thank Jürgen Heinze for providing the ant colonies and for comments on the study, and Bert Rivera Marchand and Bartosz Walter for their help with obtaining permits for collecting ants in Puerto Rico. We additionally thank two anonymous reviewers for their comments on a previous version of the manuscript.