Hornets (Vespa velutina) were collected on the campus of INRA-Bordeaux Aquitaine from July to November or were obtained at emergence from a comb artificially maintained in an incubator. They were obtained from an important natural population with high nest densities (Monceau and Thiéry, 2016). Males and females (workers) were sorted by observing the presence of an aedeagus or a sting, respectively, at the end of the last abdominal segment. For each experiment, hornets were cold anesthetized on ice for 10 min before further handling.

Antennae were obtained by cutting off the base of the scape. Samples were then fixed with 2.5% glutaraldehyde solution in 0.1 M phosphate buffered saline (PBS) at 4°C for 24 h. After three washes with PBS (10 min each), samples were dehydrated with increasing concentrations of ethanol (from 50% to 3 × 100%) at room temperature (10 min each). Samples were then dried at ambient air temperature under a hood and mounted on aluminum stubs with double-sided sticky tape. The antennae were sputter coated in argon plasma with platinum (~30 nm thickness) in a Polaron SC 7640 device (Elexience, Verrières-le-Buisson, France) at 10 mA and 0.8 kV for 200 s. Observations were performed in an FE-SEM Hitachi S4500 (Hitachi, Tokyo, Japan), with a low secondary electron detector, at 2 kV and 18 mm working distance, at the MIMA2 microscopy platform (http://www6.jouy.inra.fr/mima2).

Hornets were placed in Plexiglas holders and their antennae were fixated horizontally with low melting point wax (Deiberit 502, Schöps and Dr. Böhme, Goslar, Germany). Preparations were placed under a macroscope (Z16 APO A, Leica Microsystems, Wetzlar, Germany) to visualize and identify a basiconic sensillum from the fourth to the eighth antennal segment. Using a micromanipulator, sensilla were approached with a glass electrode filled with 2% micro-ruby (Dextran, Tetramethylrhodamine and biotin, 3,000 MW, D-7162; Invitrogen, Eugene, OR) in distilled water. A single basiconic sensillum on each antenna was perforated with the electrode's sharp tip, and then remained in contact to let the dye diffuse for 3 h. The electrode was then removed and the hornets were released in a breeding box with available food and water, for 48 h in the dark. Then, the brains were dissected out in 0.1 M PBS solution and plunged into fixative solution (4% paraformaldehyde in PBS) for 24 h at 4°C.

Hornet brains were dissected out in PBS and fixed for 24 h at 4°C in 4% paraformaldehyde. The brains were washed 3 times (10 min each) in PBS solution containing 0.2% of Triton X-100 (PBST) and preincubated for 3 h at room temperature in PBST with 10% normal goat serum (G9023, Sigma-Aldrich, Steinheim, Germany), henceforth NGS/PBST, to avoid unspecific staining. Tissues were probed with rabbit anti-serotonin primary antibody (S5545, Sigma-Aldrich, Steinheim, Germany) diluted (1:250) in NGS/PBST for 7 days at 4°C. Then, the brains were washed 3 times (10 min each) in PBST and incubated in Alexa-fluor 488-conjugated goat anti-rabbit secondary antibody (A-11008, life technologies; diluted 1:200 in NGS/PBST) for 7 days at 4°C. According to manufacturer data, pre-incubation of the primary antibody with 500 μM serotonin inhibits specific staining.

After selective staining of basiconic sensilla or after the immunostaining procedure, brains were washed in PBS (3 × 10 min), dehydrated in series of increasing ethanol concentrations (from 50% to 3 × 100% for 10 min each) and clarified in methylsalicylate (Sigma-Aldrich, Steinheim, Germany) for at least 3 days at 4°C. Brains were then mounted in the wells of aluminum slides filled with methylsalicylate and covered from both sides with cover slips. Antennal lobes were scanned with a laser-scanning confocal microscope (LSM-700; Carl Zeis, Jena, Germany) equipped with a water immersion objective (20 × plan-apochromat 1.0 NA). The brains were scanned at 1 μm intervals (z axis) creating confocal stacks of 1,024 × 1,024 (x,y) pixels, at a resolution of 0.45 μm/pixel. Micro-ruby was revealed using a 555 nm solid-state laser. Alexa fluor 488 or autofluorescence, depending on the experiment, were revealed using a 488 nm laser.

Serial optical sections were saved as LSM files and opened using ImageJ software with the Bio-Formats library plugin. Brightness and contrast of images were adjusted before being saved as TIFF files. Then, TIFF files were imported in three-dimensional analysis software (AMIRA 5.4.3, VSG, Berlin, Germany). Glomeruli were individually reconstructed by manual labeling in three planes (xy, xz, and yz) and using the Wrap function to obtain their 3D models. The number of stained glomeruli was visually assessed by overlapping the 555 nm wavelength image stacks with the background staining and corresponding 3D reconstruction.

Single sensillum extracellular electrophysiological recordings (SSR) were obtained from basiconic sensilla (type bs2) on the antenna flagellum. Hornets were held in Plexiglas holders and their antennae were fixed horizontally with low melting point wax. The body of the hornet was electrically connected to the ground by inserting a silver wire on or near the clypeus covered with a drop of electrocardiogram gel (Redux electrolyte Gel, Parker Laboratories, Fairfield, USA). One antenna was placed in a humidified constant air stream (15 mL/s). The end of an electrolytically sharpened tungsten wire was carefully inserted at the base of a sensillum on flagelomers 6–10 using a micromanipulator (Microstar, Scientifica UK). The tungsten electrode was connected to a custom-built preamplifier (×10), and further amplified (×100) and bandpass filtered at 10–2,800 Hz by a CyberAmp 320 amplifier (Axon Instruments, USA). The filtered signal was digitally sampled at 10 kHz (DT9816; Data Translation) and analyzed using a custom software to observe and detect spikes (dbWave, Marion-Poll, 1996).

Odorant stimulations were performed with a stimulus controller (CS05, Syntech, Germany) by blowing air during 5 s through a Pasteur pipette held ~2 cm from the tip of the sensillum being recorded. The Pasteur pipette contained a filter paper (1 cm²) loaded with 5 μL of odorant solution. We tested low-volatility long-chain hydrocarbons (alkanes) and more volatile aliphatic compounds (short chain alcohols, ketones and aldehydes). Alkanes (docosane, C22; pentacosane, C25 and heptacosane, C27) were dissolved in hexane (50 μg/μL) and the filter papers soaked with the solutions were maintained for 10 min under a constant air stream, to let the solvent evaporate. For stimulations, the pipettes were first heated at 60°C into an incubator, for a few minutes to volatilize the chemicals, and used immediately (see Carcaud et al., 2015). Control stimulations were performed with solvent alone (5 μL hexane) heated at 60°C, to rule out mechanosensory or thermosensory responses. Volatile aliphatic compounds (1-hexanol, C6ol; 2-nonanone, C9one and nonanal, C9al) were presented pure at ambient temperature.

In each sensillum we could record action potentials of different amplitudes in a continuous distribution, reflecting the large number of OSNs housed in this sensillar type (about 35 stained glomeruli). The spike detection threshold was adjusted by zooming on portions of the data, in order to discriminate spiking activity from random baseline fluctuations. The events detected were considered as an estimate of the total spiking activity and used as a measure of sensillum activity (Sharma et al., 2015). Spike frequency was measured in 100 ms bins for 15 s, 5 s before, 5 s during and 5 s after the stimulus. To decide whether a stimulus induced a significant activity, we compared the spike frequency during the stimulus to the activity recorded for 5 s before the stimulus. If activity during the stimulus presentation was above noise, defined as 2 standard deviations (SD) of spike frequency before the stimulus, it was considered as a response. For a general evaluation of response intensity, we also represented spike frequency as a number of SD above baseline at each 100 ms bin (Figures 6E,F). Responses (in SD units, defined as the average number of SD above baseline during stimulus presentation—gray area in Figures 6E,F) were compared statistically among stimulus types (long chain alkanes, volatiles, control) using a Kruskal-Wallis test followed by Dunn post hoc tests, which include a correction for multiple comparisons.

The antenna of V. velutina is classically composed of a scape, a short pedicel and a flagellum, which contains 10 flagellomeres in females and 11 in males (Figure 1A). The flagellomeres are profusely covered with cuticular sensory structures called sensilla (Figures 1B,C). A pair of bulging oval structures containing a few hair-like sensilla are nevertheless present on the ventral side of male flagellomeres (Figures 1C,D, 2B). These male-specific structures, the tyloids (Romani et al., 2005) contain large pores considered as excretory ducts involved in mating behavior. Sensilla can be classified in several different types based on their morphology (Zacharuk, 1980). We did not observe any marked sexual dimorphism with regards to the sensilla types and their distribution patterns over flagellar segments (Figure 2). Overall, 9 sensilla types were identified on the hornet antenna (detailed description in Supplementary Text 1). We found two types of trichoid sensilla (Figure 1E, olfactory and/or mechanosensory), two types of placode sensilla (Figure 1F, most probably olfactory), two types of chaetic sensilla (Figure 1G, possibly gustatory) and coeloconic sensilla (Figure 1G, possibly involved in hygro- and thermo-reception). Most importantly for this study, we observed two types of basiconic sensilla, which consist of a peg on a socket (Figures 1H,I). Basiconic sensilla 1 are long and slender with a smooth surface perforated by numerous minute pores especially in the tip region (Figure 1H). This sensillum type is found only on the dorsal antennal surface, mostly grouped in the proximal regions of the last flagellomeres (Figure 2). Basiconic sensilla 2 have a much larger base, exhibiting ridges (Figure 1I). The sensillum tip has a smooth, possibly porous surface, which was alternatively observed with a flat or inwardly bent shape, and often displayed a large terminal hole (Figure 1I). These sensilla are particularly densely represented on the dorsal surface of the last flagellomers and become progressively sparser toward proximal segments in both sexes (Figure 2). Given their dorso-apical distribution and their singular morphological features, we hypothesized that basiconic sensilla 2 may have a contact chemosensory function and be homologous to the CHC-sensitive basiconic sensilla of ants (Ozaki et al., 2005; Nakanishi et al., 2009; Sharma et al., 2015).

Mass staining of the antennal nerve revealed nine axon bundles which project to ~265 olfactory glomeruli in the V. velutina antennal lobe. These olfactory sensory tracts, termed T_A– T_I(Couto et al., 2016), innervate nine distinct clusters of glomeruli, with similar innervation pattern in males and females (Figure 3). We investigated the projection pattern of sensory neurons from basiconic sensilla 2, using specific single sensillum staining. Basiconic sensilla 2 were easily recognizable under the microscope, so that a glass electrode filled with fluorescent dye could be inserted into a single sensillum (inset in Figure 4A). When sensory neurons were particularly brightly labeled, OSNs could be traced from the antennal sensory tract to their glomerular termination (Figure 4A, Supplementary Movie 1). The stained axons run roughly in parallel within the antenna until they all suddenly intermingle and eventually split up forming an axon sorting-zone at AL entrance (SZ in Figure 4A). Within each glomerulus, each axon formed a claw-like innervation, penetrating the glomerulus' outer rim (cortex, Figures 4A,B). In all preparations (n = 5 males and 5 females), the stained OSNs projected to a restricted region on the dorso-caudal side of the AL, innervating many small glomeruli of the T_B cluster. We used the most brightly stained preparations to assess the number of glomeruli innervated by the OSNs contained in a single basiconic sensillum (n = 1 in each sex). At least 36 glomeruli were stained in the female, corresponding to ~37.5% of the 96 T_B glomeruli (Figures 4C,D). Similarly in the male, OSNs from a single basiconic sensillum projected to at least 29 glomeruli out of 80, corresponding to 36.2% of the T_B cluster (Figures 4E,F). These data suggest that basiconic sensilla 2 contain approximately 35 sensory neurons, which project exclusively into the T_B glomerular cluster.

The restricted innervation of a group of dorso-caudal glomeruli by sensory neurons from basiconic sensilla observed here in hornets is reminiscent of a similar structure in the antennal lobe of ant workers (Zube and Rössler, 2008; Mysore et al., 2009; Kelber et al., 2010; Nakanishi et al., 2010). In ants, the corresponding T6 cluster contrasts with other glomerular clusters by a lack of serotonin immunoreactive fibers (Zube and Rössler, 2008; Nakanishi et al., 2010). We thus checked whether this feature is also present in hornets. Using immunohistochemistry and an established antibody against serotonin, we observed widespread serotonin-like immunoreactivity throughout the hornet brain, and in the antennal lobe (Figure 5). All our stainings, both in females (n = 10, Figures 5A,C) and males (n = 5, Figures 5B,D), showed a clear dichotomy in the serotonin-like immunoreactivity of the V. velutina antennal lobe. Ventral and dorso-rostral glomeruli corresponding to T_A, and T_C to T_I exhibited a clear and homogenous serotonin-like innervation while no labeling was observed in the glomeruli situated in the dorso-caudal area corresponding to the T_B cluster (Figures 5C,D).

Using the single sensillum extracellular recording (SSR) technique (Sharma et al., 2015), a total of 45 basiconic sensilla 2 located between the 6th and the 10th flagellum segments were recorded in 15 workers. A panel of three cuticular hydrocarbons (docosane, pentacosane, and heptacosane), three volatile aliphatic odorants (1-hexanol, 2-nonanone, and nonanal) and a control were presented. In most cases (n = 37 sensilla), the sensillum did not respond to any of the tested stimuli. Because of the hard cuticle of the sensillum, electrode insertion was difficult and may sometimes have damaged the sensory neurons. Alternately, our odor panel may not contain odorants activating this sensillum. In the remaining cases (n = 8 sensilla), clear responses to one or a few stimuli were recorded (Supplementary Figure 1, Supplementary Table 1). These responses appeared mostly for the long-chain alkanes, with three sensilla responding to docosane, two sensilla responding to pentacosane and four to heptacosane (two sensilla responded to both C25 and C27). As observed in ants (Sharma et al., 2015), basiconic sensilla 2 sometimes also responded to volatile compounds, with two sensilla responding to 2-nonanone, one to 1-hexanol and two to nonanal (Supplementary Table 1). In the same conditions, no responses were observed to the controls, except for one sensillum that responded once to the stimulation with a heated pipette (Supplementary Table 1). Thus, the observed responses to alkanes were not mechano- or thermosensory responses and were not due to the hexane solvent. Two sensilla (#1 and #5), gave particularly robust recordings (Figures 6A–C). Sensillum #1, located on the 7th flagellomere responded to C25, C27 and 2-nonanone but not to the other odorants or to the control (Figure 6A). Sensillum #5, located on the 8th segment showed clear and reproducible responses to C27 but not to the other odorants or to the control (Figures 6B,D). We represented in Figure 6E the recorded responses to each stimulus and to the control relative to noise, in SD (standard deviation) units (number of SD of the signal relative to baseline, n = 8 sensilla). While the three tested long-chain alkanes induced peak responses above 3 SD, the three volatiles induced responses below 3 SD (Figure 6E). Consequently, the average response to alkanes during the stimulus was statistically higher than the responses to volatiles and to the control (Figure 6F; Kruskal-Wallis test, H = 13.55, p < 0.01; multiple comparisons (Dunn test): l.c. alkane vs. control, p < 0.01; l.c. alkane vs. volatile, p < 0.01; volatile vs. control, p = 1.0, NS). Thus, OSNs contained in basiconic sensilla 2 respond preferentially to long-chain alkanes that are typically found in the cuticular hydrocarbon profiles of hornets (Martin et al., 2009).

Neuroanatomical studies have pointed out striking similarities in AL organization across different Hymenoptera, especially concerning the division of the antennal nerve in different tracts projecting into distinct clusters of glomeruli (Zube et al., 2008; Nishino et al., 2009; Couto et al., 2016). A currently debated question is whether morphologically similar AL clusters across different species are indicative of conserved olfactory structures (homologous clusters), or if they are the result of convergent evolution. Our study unraveling compelling similarities between the hornet T_B cluster and the female-specific T6 cluster in the AL of ants, point to a homology. Indeed, in both species sensory neurons of basiconic sensilla exclusively project into a dorso-caudal cluster containing numerous small glomeruli (Nakanishi et al., 2009, 2010; Kelber et al., 2010). Olfactory information processed within this subsystem is further conveyed to higher-order centers by the median tract of uniglomerular projection neurons (m-ALT) in ants as well as in wasps (Zube et al., 2008; Couto et al., 2016). In addition to similar input-output connectivity, the rare immunoreactive profile of this glomerular cluster suggests that innervations of local AL neurons are alike. In ant workers, the T6 cluster is the only subdivision of the AL that lacks serotonin immunoreactive fibers (Zube and Rössler, 2008; Nakanishi et al., 2010). We found the same feature for the T_B cluster of both male and female hornets, as it was devoid of serotoninergic fibers whereas the rest of the AL was homogeneously stained. Similarly, a previous study showed differential reactivity to dehydrogenase and acetycholinesterase in two similar subregions of the antennal lobes of the ant, Camponotus vagus and the vespid wasp Polistes gallicus (Masson and Strambi, 1977). Although we cannot exclude that the resemblance between the ant T6 and the hornet T_B subsystems could be the result of convergent evolution, the striking similarities we observed, involving several different neuronal populations from the periphery to the central brain, make this hypothesis unlikely. Our data rather suggest that the vespid wasp and the ant ALs contain a homologous olfactory subsystem for CHC detection.

In ants, there is accumulating evidence that the basiconic sensilla subsystem is involved in CHC processing (Ozaki et al., 2005; Sharma et al., 2015). Using single sensillum electrophysiological recordings, we obtained odor-evoked responses in hornet basiconic sensilla 2, which respond more strongly to long-chain alkanes than to other tested stimuli. This suggests that this homologous subsystem in hornets and ants may have conserved a similar function, namely to detect and process CHC information. Interestingly, hornet basiconic sensilla also displayed responses to volatile aliphatic compounds (Figure 6, Supplementary Table 1). This pattern is similar to that observed in ants, although these insects' basiconic sensilla appear more broadly tuned than hornets' (Sharma et al., 2015). This difference may be explained by discrepancies between species in the number of sensory neurons housed in basiconic sensilla. We were not able to precisely assess the number of sensory neurons within a basiconic sensillum, but a range of 30–40 was reported in closely related Vespidae species (Lacher, 1964). This observation fits with our counts of ~35 labeled glomeruli after staining of a single sensillum (Figures 4D,F). Comparatively, carpenter ant basiconic sensilla contain at least 130 OSNs (Nakanishi et al., 2009). More sensory neurons participating in the recorded activity may broaden the apparent response profile of the sensillum.

What could be the function of a CHC-sensitive olfactory subsystem in the biology of hornets? Social wasps are known to use CHCs for the discrimination of nestmates from non-nestmates at the nest entrance (Gamboa et al., 1986; Ruther et al., 2002; van Zweden and d'Ettorre, 2010). In the common wasp, Vespula vulgaris, some CHCs have also been shown to act as queen pheromones, advertising the queen's fertility status (Van Oystaeyen et al., 2014; Oi et al., 2015). These signals also seem to be involved in worker policing, allowing the recognition by workers of eggs that were not laid by the queen (Foster et al., 2002; Oi et al., 2015). Lastly, these cues might also convey information about a workers' task and be involved in task allocation, as observed in ants (Greene and Gordon, 2003). Thus, an efficient CHC-processing system should be crucial for Vespidae social biology, as a basis for nestmate discrimination, caste differentiation, worker policing and task allocation.

One feature differed markedly between the CHC subsystems of hornets and ants: male hornets present numerous basiconic sensilla, whereas these are frequently absent in the males of numerous solitary and social Hymenoptera [Formicidae: carpenter ants (Nakanishi et al., 2009; Mysore et al., 2010); Apidae: honey bees (Esslen and Kaissling, 1976); bumble bees (Ågren and Hallberg, 1996); Eucera (Streinzer et al., 2013); Colletidae (Ågren, 1977); Andrenidae (Ågren, 1978); Trichogrammatidae (Amornsak et al., 1998)]. However, some spheciform wasps (Hymenoptera, Sphecidae, and Crabronidae) show the presence of basiconic sensilla in males (Walther, 1983; Herzner et al., 2003). Since the latest phylogenetic analyses indicate that vespoid wasps are basal to ants (Branstetter et al., 2017; Peters et al., 2017), our observations suggest that the presence of basiconic sensilla in males might be the ancestral trait. Accordingly, male hornets display a T_B cluster that was identical to that of females (Figures 3–5). In ants, the lack of a CHC-specific system in males was explained by their low involvement in social tasks, the males being utterly focused on mating (Nakanishi et al., 2009; Nishikawa et al., 2012). Mating in hornets is thought to take place on a hard substrate, possibly on the nest envelope (Batra, 1980). While they use volatile queen-emitted sex pheromones (Ono and Sasaki, 1987; Spiewok et al., 2006), copulation only occurs after the male has had the opportunity to touch the female with its antennae (Batra, 1980). This suggests that low volatile compounds, possibly CHCs, could be involved in the decision by the male to copulate with a given female, recognizing its fertility but also avoiding inbreeding. As different hornet species are thought to share the same sex pheromones (Ono and Sasaki, 1987) it may also participate in pre-mating reproductive isolation, avoiding mating across species.

A sensory system allowing the detection of kinship could have been a crucial preadaptation which facilitated the emergence of eusociality in Hymenoptera by preventing altruistic acts toward non-related individuals. Our study revealed that an olfactory subsystem that processes CHC information might be conserved in ants (Formicidae) and hornets (Vespidae), two families in which eusociality evolved separately. This olfactory subsystem may therefore have been present in their last common ancestor, a solitary predatory wasp (Johnson et al., 2013; Branstetter et al., 2017; Peters et al., 2017). An interesting possibility is that this ancestral CHC processing subsystem was initially involved in prey recognition and was later co-opted for kinship recognition. Indeed, some Crabronid wasps (Liris niger) recognize their prey by means of CHCs through basiconic sensilla (Anton and Gnatzy, 1998). Future work should now provide more examples of the existence of this olfactory subsystem in different hymenopteran families, aiming to understand its evolution. Similarly, it will be important to follow the evolution of the 9-exon olfactory receptor gene family, which is currently thought to be involved in the detection of CHCs, although this has not been demonstrated yet (Tsutsui, 2013; Engsontia et al., 2015; Zhou et al., 2015). The emerging model is that 9-exon ORs are expressed in the OSNs harbored by basiconic sensilla and that expansions of this OR class in some hymenopteran lineages went hand in hand with larger numbers of glomeruli within their CHC olfactory subsystem (McKenzie et al., 2016). More generally, our working hypothesis is that the CHC-processing subsystem is ancestral and more widespread among Hymenoptera than initially thought. We favor the idea that complex CHC recognition/discrimination abilities, possibly involving the basiconic sensilla/T_B subsystem, may have been a “spring-loaded preadaptation” playing a central role in the advent of eusociality (Kather and Martin, 2015).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.