Comparative analysis of fertility signals and sex-specific cuticular chemical profiles of Odontomachus trap-jaw ants

The cuticular chemical profiles of social insects encode information necessary for maintaining colony identity and social cohesion, providing the foundation through which eusociality is maintained (Blomquist and Bagnères, 2010). Signals of colony membership result from the blending of individual profiles through constant social interactions and assessments of nestmate profiles (van Zweden and d'Ettorre, 2010). Simultaneously, reproductive division of labor is maintained by a specific subset of cuticular chemicals expressed by queens and reproductive workers that signal their fertility (Liebig, 2010). Additionally, some species display sex-specific chemical profiles that distinguish males from female castes in the colony (Smith et al., 2014). Although these communicative chemical profiles have been generally described for a number of species (e.g. Van Oystaeyen et al., 2014), there have been few detailed comparisons among closely related species.

The cuticular chemical profiles of insects usually are composed primarily of straight-chain and branched-chain alkanes and alkenes of 19–35 carbons (Martin and Drijfhout, 2009; Blomquist and Bagnères, 2010). Lesser amounts of other classes of chemicals also have been documented on the cuticle including very long-chain hydrocarbons, triacylglycerides and oxygenated compounds (Cvacka et al., 2006; Buckner, 2010; Yew et al., 2011). Some of these non-hydrocarbon cuticular compounds (e.g. triacylglycerides) are important components of pheromone blends (Kuhbandner et al., 2012), interacting synergistically with cuticular hydrocarbons.

Cuticular chemical profiles of queens and workers have been compared for more than 60 species of social insects, and bioassays have resulted in the experimental identification of specific cuticular compounds responsible for fertility signaling in seven of these species (Van Oystaeyen et al., 2014). Phylogenetic comparisons indicate that particular cuticular compounds or classes of compounds (saturated alkanes) are highly conserved across genera as queen fertility signals (Van Oystaeyen et al., 2014; van Zweden et al., 2014). Several recent reviews highlight the theoretical implications that the observed high level of conservation of queen signals among species has for our understanding of signal evolution (Oi et al., 2015; Peso et al., 2015). Rapidly evolving queen signals are predicted to be a product of conflict between workers and queens when queen signals manipulate worker physiology against the interests of workers. In contrast, conserved queen signals are indicative of honest advertisement of queen fecundity rather than exploitative manipulation of worker reproductive physiology (Keller and Nonacs, 1993). However, comparative chemical data on more than one or two species within a genus are not available for the majority of social insect genera, and current phylogenetic comparisons and theories of fertility signal evolution rely heavily on equivocal comparisons across lineages rather than comparisons among closely related species (Van Oystaeyen et al., 2014). More finely focused, within-genera comparisons of queen and worker cuticular profiles are necessary to fully understand how these signals evolved.

Ants are the most well studied social insects in terms of the chemical profiles of queens (described for over 40 species), yet only two studies have directly compared the cuticular profiles of queens and workers of congeneric species. Brunner et al. (2011) found that the profiles of queen and worker ants of six Temnothorax species were dominated by linear and branched alkanes, but there was no clear trend in compounds that characterized queens across species. In contrast, Holman et al. (2013) found that queens of 11 Lasius species had greater relative amounts of 3-methylalkanes than did workers. For two of these species, L. niger and L. flavus, 3-methylhentriacontane was experimentally identified as the compound used for signaling queen fertility (Holman et al., 2010,, 2013). Beyond these directly comparative studies, queens of three species of Neoponera (formerly Pachycondyla) species were distinguished by relative amounts of particular alkenes, methylalkanes or dimethylalkanes (Heinze et al., 2002; Kellner, 2005; Evison et al., 2012). Species-specific blends of straight- and branched-chain alkanes also distinguished queens from workers in three Camponotus species (Bonavita-Cougourdan and Clement, 1994; Endler et al., 2006; Campos et al., 2012). These comparative studies of four genera from three ant subfamilies suggest that there is little evidence for conservation of ant fertility signaling compounds within genera, with the exception of the genus Lasius.

Cuticular chemical profiles of males have been reported for eight species of ants, only two of which, Diacamma ceylonense and Odontomachus brunneus (both in the subfamily Ponerinae), exhibit male-specific profiles (Cuvillier-Hot et al., 2001; Smith et al., 2014). Sexually dimorphic cuticular chemical profiles are thought to be the ancestral condition from which eusocial queen signals evolved, because of their widespread use as sex-specific mate recognition signals in solitary insects, including solitary and primitively eusocial Hymenoptera (Paulmier et al., 1999; Ayasse et al., 2001; Liebig et al., 2009; Kocher and Grozinger, 2011; Van Oystaeyen et al., 2014; Oi et al., 2015). Further investigation of the hydrocarbon patterns of males relative to those of female castes may prove essential in understanding the evolution of chemical signaling in social insects.

Here, we compared the cuticular chemical profiles of queen, worker and male ants of three species of Odontomachus ants that occur in the southeastern USA, and contrast these findings with earlier research on O. brunneus. The fertility signal of O. brunneus queens and workers was identified experimentally as (Z)-9-nonacosene (Smith et al., 2012, 2013, 2015). Cuticular profiles of male O. brunneus were dominated by pentacosenes and pentacosadienes, which were present in low amounts on workers (average combined relative abundance of 36% and 0.5%, respectively; Smith et al., 2014). Additionally, we report the identification of several novel 2,5-dialkyltetrahydrofurans that are major components of the cuticular lipids of O. ruginodis queens. We also present results of bioassays that tested the responses of workers to crude extracts of O. ruginodis queens, and to fractions of crude extracts that contained different classes of cuticular chemicals.

We compared the cuticular hydrocarbon profiles of four species of Odontomachus that co-occur in the state of Florida in the southeastern USA (see MacGown et al., 2014): (1) O. brunneus (Patton), a widespread native species, (2) O. relictus Deyrup and Cover, a native species restricted to two Florida populations, (3) O. ruginodis M. R. Smith, widespread throughout Florida and probably introduced, and (4) O. haematodus (Linnaeus), an introduced South American species that now occurs from Alabama to Florida. Phylogenetic research has revealed that O. relictus is closely related to O. brunneus, whereas O. ruginodis and O. haematodus are more distantly related but in the same New World clade (Spagna et al., 2008; Schmidt, 2009; Larabee, 2015). The cuticular chemical profiles of workers, queens and males of O. brunneus have been described previously (Smith et al., 2012, 2013, 2014, 2015).

Colonies and foundresses of O. ruginodis were collected in August 2012 from the MacArthur Agro-ecology Research Center in Lake Placid, Florida. Colonies of O. relictus were collected in August 2012 from the Archbold Biological Station in Venus, Florida, and colonies of O. haematodus were collected in March 2014 from the Weeks Bay Preserve in Fairhope, Alabama.

In the laboratory, multiple colonies of each species were separately housed in two to three interconnected 60×15 mm Petri dishes with plaster-lined bottoms that were moistened twice per week. Colonies received a constant supply of water and 20% sugar–water solution, and were fed twice a week on termites and freeze-killed crickets. All colonies were kept at an average temperature of 27°C. Colonies were used for chemical analysis after they had been kept under laboratory conditions for at least 4 weeks. Queens were sampled only if their colony had produced only female-destined brood and an egg pile was consistently present in the colony. For O. ruginodis foundresses, chemical samples were taken after the first clutch of 1–5 eggs was laid and before any of those eggs developed into larvae. Samples were taken from males present in the colonies when they were collected, and from males reared in the laboratory by queenless colonies containing reproductive workers.

The cuticular chemical profiles of queens, workers and foundresses were sampled from live ants by solid-phase microextraction (SPME). A SPME fiber (100 µm polydimethylsiloxane; Supelco Inc., Bellefonte, PA, USA) was lightly rubbed on the abdomen of a restrained ant (see bioassays section, below, for a description of the restraint) for 5 min, then thermally desorbed for 5 min in the injection port of a Hewlett-Packard 6890 gas chromatograph (Agilent Technologies, Santa Clara, CA, USA), equipped with a DB-5MS column (30 m×0.25 mm×0.25 µm film; J&W Scientific, Folsom, CA, USA), connected to a Hewlett-Packard 5973 series mass-selective detector. The GC injection port and transfer line were set to 280°C. The column temperature was held at 60°C for 2 min, increased to 220°C at 40°C min⁻¹, and then to 315°C at 4°C min⁻¹. Helium was the carrier gas at 1 ml min⁻¹, and samples were injected in splitless mode with a purge time of 2 min. Electron impact ionization mass spectra were obtained using 70 eV ionizing voltage, with a source temperature of 230°C.

The cuticular profiles of males and workers were compared by GC-MS analysis of solvent extracts of each sex. Ants were freeze-killed and submerged, individually, in 125 µl of hexane for 5 min. The extract was concentrated under a stream of nitrogen to a volume of ∼10 µl, and 1 µl was analyzed by GC-MS as described above, injecting in splitless mode.

Linear hydrocarbons were identified from their mass spectra, including the molecular ion if present, and by matching retention times with authentic standards. Branched chain compounds were identified from a combination of their mass spectra, looking specifically for enhanced ions due to fragmentation on either side of methyl branch points, and by comparison of their retention indices with those of straight-chain hydrocarbons (Carlson et al., 1998).

To identify the novel tetrahydrofuran compounds, a hexane extract of eight O. ruginodis queens was fractionated by liquid chromatography on a column of oven-dried silica gel (200 mg). The column was first wetted with hexane, then loaded with the extract, which had been concentrated to ∼100 µl under a stream of nitrogen, and the extract was flushed into the bed with a few drops of hexane. The column was then eluted sequentially with 1×1 ml hexane, 2×0.5 ml 10% ether in hexane, and 1×0.5 ml ether. The saturated and unsaturated tetrahydrofurans eluted cleanly in the second fraction. High resolution mass spectra of the tetrahydrofurans were obtained on a Waters GCT coupled gas chromatograph–high resolution mass spectrometer, run in EI mode (70 eV). The GC was equipped with a DB-5 column (30×0.25 mm i.d., J&W Scientific), programmed from 100 to 280°C at 10°C min⁻¹, with a final time of 20 min.

An aliquot of the fraction containing the tetrahydrofurans was diluted to ∼0.5 ml with hexane, and stirred for 2 h with a few milligrams of 5% palladium on carbon under a hydrogen atmosphere. The mixture was then filtered through a small pad of Celite to remove the catalyst, concentrated under a stream of nitrogen, and analyzed by GC-MS.

To determine the double bond positions in the unsaturated tetrahydrofuran compounds, a second aliquot of this fraction in a screw-cap vial was treated with 5 µl dimethyldisulfide and 10 µl of a 2% solution of iodine in tetrahydrofuran, warming at 50°C overnight with stirring. After cooling, the mixture was diluted with 0.5 ml each of hexane and 1 mol l⁻¹ aqueous sodium thiosulfate, and vortexed until the organic layer was colorless. The organic layer was pipetted off and passed through a plug of anhydrous Na₂SO₄ to remove traces of water, then concentrated under a stream of nitrogen, and analyzed by GC-MS using a medium polarity DB-17MS capillary column (25 m×0.20 mm, 0.33 µm film thickness; J&W Scientific, Inc.). Injections were made in splitless mode, and the oven was programmed from 100 to 280°C at 10° min⁻¹, with a final time of 20 min.

To identify the double bond positions in alkenes and dienes present in crude hexane extracts of queens, the extracts were first fractionated on AgNO₃-impregnated silica gel. Thus, crude extracts from 2–5 O. ruginodis, O. relictus and O. haematodus males were concentrated to ∼100 µl, and fractionated on a column of 10% AgNO₃ on silica gel (400 mg) contained in a Pasteur pipette with a glass wool plug. The column was dried immediately before use in an oven at 120°C for 2 h. After cooling to room temperature, the column was wetted with hexane, then loaded with the concentrated extract which was rinsed into the bed with a few drops of hexane. The column was then eluted sequentially with 2 ml hexane, 2 ml 10% cyclohexene in hexane, and ether to provide fractions containing saturated hydrocarbons, unsaturated hydrocarbons and more polar compounds, respectively. The unsaturated hydrocarbons fraction was then derivatized with dimethyl disulfide (DMDS) and the resulting adducts were analyzed by GC-MS using a medium polarity DB-17MS capillary column, as described above.

Compounds were included in data analyses if they appeared in ≥70% of one of the castes of individuals (workers, queens, foundresses or males) sampled. Statistical comparisons of the cumulative abundances of compound classes across caste and species were carried out with Mann–Whitney U-tests performed used JMP statistical software (SAS Institute Inc., Cary, NC, USA). For comparisons between workers and males, diagnostic power (DP) of individual compounds was calculated (van Zweden and d'Ettorre, 2010). DP is the standard deviation of the standardized peak area of all sampled individuals divided by the pooled standard deviation within the grouping of interest. Groupings of interest were population and sex and species; therefore, two measures of DP for each compound were calculated. Compounds with the highest DP are the most consistent within a group while being variable between groups; therefore, they are compounds most likely to be ‘diagnostic’ of the grouping (van Zweden and d'Ettorre, 2010).

Three workers from nine colonies (27 workers in total) were harnessed in a restraint fashioned from a 5×5 cm square of paper. A radial slit was made through the paper square, ending in a small hole at the center. The petiole (constriction between the thorax and the second abdominal segment) of the test ant was placed in the hole, so that the legs, thorax and head protruded from one side of the paper, and the abdomen from the other. Responses to stimuli were assessed by contacting the antennae with the stimulus and measuring antennal retraction, a stereotypical submissive posture displayed by workers of Odontomachus species when in proximity to a reproductive queen (Medeiros et al., 1992; Powell and Tschinkel, 1999). The worker pulls its antennae away from the queen such that the antennal scape is perpendicular to the long axis of the head. This assay has been used previously in studies of fertility signal perception in O. brunneus (Smith et al., 2015).

The first experiment presented restrained workers with cold-anesthetized non-nestmate queens or non-nestmate workers from the same colony. For the second experiment, three reproductive queens and three corresponding nest workers were freeze killed and extracted in 125 µl of hexane. The whole extract was then transferred in 10 µl increments to the tip of a glass rod (6 cm diameter), allowing the solvent to evaporate to coat the tip with the extract. The glass rod was then presented to non-nestmate ants as stimuli. For the third experiment, three additional queens and corresponding nest workers were freeze killed and extracted in 125 µl of hexane. Extracts of queens and workers were fractionated separately using a small column of 500 mg oven-dried 10% silver nitrate-impregnated silica gel in a Pasteur pipette with a glass wool plug. Columns were first wetted with 2×1 ml of hexane, then loaded with extract, and extract was eluted with 2.5 ml 20% cyclohexene in hexane to yield a complete hydrocarbon fraction followed by 2.5 ml of 20% ether in hexane to yield a fraction containing more polar constituents (including the 2,5-dialkyltetrahydrofurans). The worker extracts were run through an identical column, eluting with 20% cyclohexene in hexane. Fractions were concentrated under a stream of nitrogen, then resuspended in 125 µl of hexane, transferred to a glass rod, and presented to test ants as described above.

A different set of test workers was used for each of the three experiments. The stimulus queens, workers and extracts were presented in random order, and presented three times per trial. All trials were video recorded and given a coded title which ensured that the data transcriber was blind to the treatments and identity of test ants and stimuli. The consensus response (≥2 of 3 submissive or non-submissive responses) was recorded for each test ant. Data from experiments 1 and 2 were analyzed using McNemar's test, and data from experiment 3 were analyzed by Cochran's Q-test. All analyses were performed used JMP statistical software (SAS Institute Inc.).

GC-MS analysis on a medium polarity DB-17 column of the 10% ether in hexane fraction from silica gel chromatography of the crude cuticular extract of O. ruginodis queens showed two pairs of peaks (compounds 1–4, compound numbers in this section do not correspond to compound numbers in the figures or tables), each in approximately a 1:10 ratio. The EI mass spectrum of the first peak (compound 1) in the earlier eluting pair did not show a molecular ion, but had a base peak at m/z 183 and a large fragment at m/z 281. The larger, second peak in the pair (compound 2) had a base peak at m/z 181 and a large fragment at m/z 281, suggesting that 1 and 2 were analogs differing only by the presence or absence of a double bond. Compound 2 had a clearly visible molecular ion at m/z 392, with a small fragment at m/z 374 from loss of water, indicating that it contained at least one oxygen, giving a possible molecular formula for compound 2 of C₂₇H₅₂O, a formula requiring two rings or double bonds. This molecular formula was confirmed by high resolution mass spectrometry (calculated for C₂₇H₅₂O: 392.4018 amu; found 392.4027 amu). Given the evidence for a C=C double bond in compound 2 due to its differing in mass from compound 1 by 2 amu, and the fact that compounds 1 and 2 had the same ion at m/z 281, this suggested that the remaining site of unsaturation was a ring. Furthermore, the base peak of compound 2 at m/z 181 and the large fragment at m/z 281 suggested losses of C₁₅H₃₁ and C₈H₁₅, respectively, from the molecular ion at m/z 392, leaving C₄H₆O unaccounted for from the possible molecular formula, suggesting that the structure was a 2,5-disubstituted tetrahydrofuran ring, i.e. 2-octenyl-5-pentadecyltetrahydrofuran. This structure would be expected to provide the large fragments seen, by loss of the alkyl chains from the carbons attached to the oxygen, with concomitant stabilization of the resulting cations by the oxygen. The smaller, earlier eluting compound, lacking the double bond and with a base peak at m/z 183 and a prominent ion at m/z 281, would then correspond to the saturated analog, 2-octyl-5-pentadecyltetrahydrofuran.

The mass spectra of the later-eluting pair of peaks indicated that they were simply analogs of the first two compounds. Thus, the smaller, earlier eluting peak 3 had a base peak at m/z 183 and a large fragment at m/z 309, with no visible molecular ion at m/z 422, whereas peak 4 had a base peak at m/z 181, a large fragment at m/z 309, a clearly visible molecular ion at m/z 420, and a smaller peak at m/z 402 from loss of water from the molecular ion. The high resolution mass spectrum of compound 4 provided corroborating evidence for a molecular formula of C₂₉H₅₆O (calculated for C₂₉H₅₆O: 420.4312 amu; found 420.4326 amu). These data suggested that compounds 3 and 4 were 2-octyl-5-heptadecyltetrahydrofuran and 2-octenyl-5-heptadecyltetrahydrofuran, respectively.

Catalytic reduction of an aliquot of the fraction containing the tetrahydrofuran compounds with a catalyst of 5% palladium on carbon under a hydrogen atmosphere resulted in the disappearance of peaks 2 and 4, leaving only two peaks from compounds 1 and 3, confirming that 2 and 4 were analogs of the former compounds, with one double bond. To locate the positions of the double bonds in peaks 2 and 4, an aliquot of the fraction containing the tetrahydrofurans was derivatized with dimethyldisulfide. The resulting adduct from compound 2 gave a small molecular ion at m/z 486 (3% of base peak), a base peak at m/z 425 from loss of C₂H₅S, with a corresponding ion at m/z 61 for C₂H₅S⁺, indicating that the double bond was in the terminal position. Thus, compound 2 was conclusively identified as 2-(oct-7-enyl)-5-pentadecyltetrahydrofuran. There are four possible stereoisomers for this structure (two diastereomeric pairs of enantiomers), and it has not yet been possible to determine which isomer is produced by the ants.

Similarly, the DMDS adduct of compound 4 displayed a small molecular ion at m/z 514 (3% of base peak), and diagnostic ions at m/z 453 (base peak) from loss of C₂H₅S and m/z 61 for C₂H₅S⁺, again showing that the double bond was in the terminal position, proving the structure of compound 4 to be 2-(oct-7-enyl)-5-heptadecyltetrahydrofuran.

Profiles of O. ruginodis queens were distinguishable from those of workers and foundresses by increased relative abundances of several methylalkanes and dimethylalkanes as well as the heretofore undiscovered 2,5-dialkyltetrahydrofurans (Fig. 1, Table 1; see Fig. S1 for O. ruginodis foundress data, and Tables S1–3 for complete compound identification and abundance data for all species). The 2,5-dialkyltetrahydrofurans were only found in extracts from queens, and one of them [2-(oct-7-enyl)-heptadecyltetrahydrofuran, compound 53; Table 1] was the most abundant queen compound in this species.

Profiles of O. relictus queens were distinguishable from those of workers by increased relative abundances of several methylalkanes and dimethylalkanes, a small subset of linear alkanes, and some alkenes and dienes (Fig. 1, Table 1). The largest increase in the relative abundance of a single compound on queens was for (Z)-9-nonacosene. Queens of the closely related species O. brunneus were distinguished from workers by an increase in this same compound (Smith et al., 2012, 2013).

Profiles of O. haematodus queens were distinguishable from those of workers by increased relative abundances of several methylalkanes and dimethylalkanes, along with a subset of linear alkanes and a diene (Fig. 1, Table 1). The compounds with the largest proportional increases in queens were dimethylalkanes.

Comparisons of the collective total percentages of compound classes that were more abundant on the cuticles of queens of the four Odontomachus species revealed that queens have species-specific chemical profiles (Fig. 2). Worker profiles were also species specific, and whereas queens shared some compounds with their respective workers, they were differentiated from workers by diverse and species-specific sets of compounds.

Odontomachus ruginodis workers responded to anesthetized non-nestmate queens with antennal retraction significantly more frequently than they did when presented with non-nestmate workers (Fig. 3). Crude cuticular extracts coated onto the tip of a glass rod elicited analogous queen-specific antennal retractions (Fig. 3). However, when the crude extract of a queen was divided into hydrocarbon and dialkyltetrahydrofuran fractions, neither fraction alone elicited a queen-specific response when presented on a glass rod to conspecific workers (Fig. 3), indicating possible synergism between components in the two fractions.

Cuticular profiles of O. ruginodis, O. relictus and O. haematodus males were distinguished from those of all females by increases in relative abundances of several alkanes, alkenes and dienes (Fig. 4, Table 2). The most pronounced differences were in pentacosadienes, pentacosenes and pentacosane, which together constituted 39–58% of the profiles of males, across species. In comparison, that set of compounds constituted 1–3% of the profiles of workers of the same species. Compounds with the highest relative abundances on males (e.g. pentacosadienes, pentacosenes, tricosane) were the most diagnostic of sex, and among the compounds that were least diagnostic of species (Table 2), suggesting that the patterns of relative compound abundances that distinguish male profiles from females may be conserved across species. However, when examined more closely, species specificity was apparent in the patterns of alkene and diene isomer variation. For example, 5,9-alkadienes were present only in the cuticular extracts of O. haematodus, whereas 1,9-pentacosadiene was present only in extracts of O. ruginodis (Table 2). The remainder of the profiles of males consisted of relatively small amounts of the longer-chain hydrocarbons that make up the species-specific profiles of the respective workers (Fig. 4).

For social insects, the cuticular hydrocarbon profile encodes information permitting the identification of the sex, fertility status and colony identity of individuals (Blomquist and Bagnères, 2010). Previous work on O. brunneus found queens to be differentiated from workers primarily by a single fertility-signaling compound, (Z)-9-nonacosene (Smith et al., 2012, 2013). In contrast, queens of the three additional species surveyed here were distinguished from workers by a number of cuticular chemicals (Fig. 1, Table 1). Furthermore, profiles of queens of each of these species were characterized by an increased abundance of different classes of compounds (Fig. 2). The profile of queens of the species most closely related to O. brunneus, O. relictus, was somewhat similar in that the compound with the largest increase in relative abundance in comparison to workers was (Z)-9-nonacosene (relative abundance on workers: 1%, queens: 18%; Table 1). However, (Z)-9-nonacosene was not the only queen-distinguishing compound for O. relictus, nor did it differentiate queens from workers in O. ruginodis or O. haematodus.

Several dialkyltetrahydrofurans were found in cuticular extracts of O. ruginodis queens but not the corresponding extracts of workers or foundress queens (Table 1; see Table S1 and Fig. S1 for foundress data). To our knowledge, these compounds have not been found in nature before, and analogous 2,5-dialkyltetrahydrofurans have only been found in one other insect order, the Lepidoptera (Takabayashi and Takahashi, 1986a; Schulz et al., 1998). Braconid parasitoid wasps have been shown to exploit these compounds as kairomones for location and recognition of their lepidopteran hosts (Takabayashi and Takahashi, 1986b). 2-Octyl-5-pentadecyltetrahydrofuran was also found in low abundance co-eluting with 3-methylheptacosane in the profile of O. haematodus queens (see Fig. S3 and Table S3). Whereas all previously identified cuticular fertility signals in ants have been cuticular hydrocarbons (Van Oystaeyen et al., 2014), other social insects are known to signal their fertility with non-hydrocarbon compounds. For example, reproductive workers and breeding queens of the bumblebee Bombus terrestris produce long-chain esters and aldehydes (Sramkova et al., 2008). However, a recent bioassay demonstrated that queen-specific hydrocarbons were more effective than queen-specific esters at eliciting the typical physiological response of workers to a bumblebee queen (Van Oystaeyen et al., 2014). In a more concrete example, the queen pheromone of the honeybee, Apis mellifera, is a product of the mandibular gland and consists of (E)-9-oxodec-2-enoic acid and a blend of other non-hydrocarbon components (Slessor et al., 1988; Hoover et al., 2003). Furthermore, queens of the termite Reticulitermes speratus are reported to signal their fertility by emitting n-butyl n-butyrate and 2-methyl-1-butanol, and these compounds also have been found on their eggs (Matsuura et al., 2010; Yamamoto and Matsuura, 2011). However, neither the honeybee nor the termite pheromones are cuticular-based contact pheromones.

Our comparative data clearly show that the cuticular hydrocarbons that distinguish queens from workers are not conserved within the genus Odontomachus. Although recent phylogenetic comparisons suggested that queen fertility signals may be highly conserved throughout social hymenoptera (Van Oystaeyen et al., 2014), these comparisons may be misleading because they were usually based on data from single species from distantly related genera rather than from congeners or other closely related species. As an extreme example of a single chemical being used as a signal in distantly related insect groups, (Z)-9-nonacosene is a contact pheromone for females of the longhorned beetle Megacyllene caryae (Ginzel et al., 2006), but also the fertility signal of O. brunneus queens and workers (Smith et al., 2012). Nevertheless, this compound is not conserved as a fertility signal within the genus Odontomachus (Fig. 2, Table 1), or as a contact pheromone within the genus Megacyllene (Ginzel et al., 2003).

The apparent diversification of queen fertility signals in the genus Odontomachus suggests that these compounds manipulate worker reproductive physiology, and diversified as a result of conflicting interests of queens who chemically manipulate workers, and workers evolving resistance to those signals (Keller and Nonacs, 1993). However, our bioassays with O. ruginodis, and previously with O. brunneus, suggested that fertility signals are interpreted as such by workers only in the context of the other cuticular chemicals, rather than being chemicals that trigger responses in every worker that encounters the signal compound(s). Thus, cuticular extracts from O. ruginodis queens, when presented to workers on a neutral substrate, were as effective at eliciting a queen-specific releaser effect from workers as was a live queen (Fig. 3). However, neither the dialkyltetrahydrofuran nor the hydrocarbon fractions of extracts from queens elicited responses from workers (Fig. 3). (It should be noted that more polar compounds present in crude cuticular extracts may have been excluded from the fractions.) Furthermore, a recent series of experiments with O. brunneus revealed that its fertility signaling compound, (Z)-9-nonacosene, is only perceived by workers as a fertility signal when presented as a component in a nestmate or near-nestmate background chemical profile (Smith et al., 2015). Our current results suggest that, as with O. brunneus, O. ruginodis workers require a chemical context (some or all of the remainder of the cuticular chemical profile) within which to interpret queen signals, and do not respond to fertility signaling compounds that are presented out of chemical context. These results indicate the need for more systematic studies involving fractionating and recombining cuticular extracts to determine in bioassays how individual signal components are perceived, and how they might act synergistically with the remainder of the chemical profile.

To date, cuticular chemical fertility signals have been identified by manipulating concentrations of individual compounds, usually by applying them to the cuticle of workers (Smith et al., 2009) or to a neutral substrate (Holman et al., 2010). However, a single compound is unlikely to be solely responsible for the queen fertility signal of any of the three ant species studied here, as opposed to the single component fertility signal of O. brunneus. Queens of O. brunneus are distinguished from non-reproductive workers largely by a single compound (Smith et al., 2013), and applying the compound to the cuticle of a worker resulted in nestmates responding to the treated worker as though it were a queen (Smith et al., 2012, 2015). However, profiles of queens of the three species described herein were distinguished from those of workers by a number of compounds, suggesting that the queen signal in these species is likely to consist of multiple components. Work in other ant species, such as the carpenter ant Camponotus floridanus and the Argentine ant Linepithema humile, has shown that the hydrocarbon profiles of queens differ from those of workers in the relative abundances of the majority of their components (de Biseau et al., 2004; Endler et al., 2006). This suggests that the queen signals in these species might also consist of multiple components. Bioassays that start with tests of complete profiles followed by bioassays with fractions of the profiles may reveal that the context of a (more) complete chemical profile is necessary for perception of a fertility signal.

The hydrocarbon profiles of males of the three Odontomachus species studied here were broadly similar to those of male O. brunneus, with pentacosane, pentacosenes, pentacosadienes and tricosane representing more than half of the total hydrocarbons, whereas these compounds only constitute a small percentage of the hydrocarbons of females (Smith et al., 2014). These compounds were highly diagnostic of sex, rather than species (Table 2). Broadly, the classes of compounds that distinguished males from females in Odontomachus were conserved to a much greater degree than the classes of compounds that distinguished queens from workers. However, when examined in greater detail, the patterns of isomeric variation of these male-distinguishing compounds as well as the low-abundance portion of the profile that is shared with workers render the male profiles species specific (Fig. 4, Table 2; Smith et al., 2014).

Many studies of social insect fertility signals have noted potential homologies with the contact sex pheromones of females of solitary insect species (Liebig et al., 2009; Kocher and Grozinger, 2011; Van Oystaeyen et al., 2014; Oi et al., 2015). Indeed, sex-based differences in the cuticular chemical profiles of males and females are the basis for short-range mate recognition for many solitary insects (Tregenza and Wedell, 1997; Ferveur and Cobb, 2010; Ginzel, 2010). In contrast, sex-based differences in cuticular profiles have rarely been investigated in ants, and to date have only been reported for one Odontomachus and one Diacamma species, both of which belong to the basally derived ant subfamily Ponerinae (Cuvillier-Hot et al., 2001; Smith et al., 2014). Our results suggest that these sex-based differences are the most deeply conserved cuticular chemical patterns within Odontomachus. If these patterns are derived from signals from a solitary insect ancestral condition, in the evolutionary history of eusocial Hymenoptera, selection on females may have driven the divergence in hydrocarbon profiles among species that would be required for species-level mate recognition. Males might have been the first receivers of signals differentiating fertile from non-fertile females, which in turn may have driven the diversification of such signals. However, testing this hypothesis will require further investigation into male-specific signals in other social Hymenoptera, phylogenetic analysis of such signals, and a detailed knowledge of the reproductive biology of senders and receivers within each species.

In conclusion, our comparison of queen fertility signals among four Odontomachus species suggests that even within genera, these signals can be widely divergent. In contrast, male-specific hydrocarbon patterns appear to be more conserved within the genus, with species being differentiated by different patterns of isomers of the same chain lengths. Further comparative studies within genera will help to elucidate how these important eusocial signals arose and evolved, and investigations into sex-based cuticular chemical differences may provide insights into the evolutionary history of chemical signaling in social insects in general.

We thank the MacArthur Agro-ecology Research Center, Archbold Biological Station and Weeks Bay Preserve for access to field sites. We thank L. M. Hanks for access to equipment and comments on the manuscript. We are grateful to Rafael Achury, Fred Larabee and Bill Wills for assistance with collecting ants.