Why workers forfeit direct reproduction is a crucial question in eusocial evolution. Worker reproduction provides an excellent opportunity to understand the mechanism of kin conflict resolution between queen and workers. We evaluated behavioral and physiological differences among females in the paper wasp Polistes chinensis antennalis to examine why some workers reproduce under queenright conditions. Reproductive workers were old and foraged less early in the season; their cuticular hydrocarbon (CHC) profiles overlapped with those of queens but were significantly different. The distinct CHC profile of the eggs of the queen likely represented a cue for policing against those by workers. Juvenile hormone (JH) and dopamine seemed to be associated with gonadotropic function, and the JH level of reproductive workers was similar to that of the queen. The high JH level of reproductive workers likely facilitated their reproduction even under queenright conditions. Gene expression levels of the queen and reproductive workers differed only in vitellogenin. These results suggest that worker reproduction is facilitated by an increase in JH level; however, CHC is not a fertility-linked signal, but a queen-linked signal; consequently, reproductive workers without a queen-linked signal might be allowed to stay within the colony.

Colony census in the field

Queens and nests attached under the eaves of buildings were collected in the field just before emergence of the first worker from the end of April to early June. The pedicels of each nest were vertically attached to the base of lidded transparent plastic cases (19 × 13 × 6 cm) with glue (Konishi G17). The individually marked queen (all colonies were single-foundress nests) was released on the original nest with weak immobilization on ice. After release, the case was fixed with metal wire within 15 cm of the original nest sites at midnight. No apparent queen loss was observed due to this manipulation. The colonies were collected after closing the lid at the midnight for individual marking and colony census. Adults emerging from the colonies were individually marked at 3-day intervals for behavioral observation, and every day for CHC analyses. There were few missing individuals that were foraging or off the nests because we marked every newly emerged individual at midnight.

Behavioral observations in the field

The behaviors of females (foraging, oviposition, dominance, and lateral vibration) in 12 queenright colonies were observed from late June to late July for 49 hours in 1993. Each of our observations was conducted for at least 55 minutes per day from 9 am to 3 pm by one person (KT) on sunny days. We counted the frequency of lateral vibration (wagging) separately from dominance acts, which are included in the category of dominance behavior (Gamboa and Dew 1981). Darting and biting opponents were classified as dominance behavior in this study. We counted occurrences of dominance behavior per individual because these behaviors are associated with reproduction in many paper wasps (e.g., West-Eberhard 1969; Jeanne 1972; De Souza and Prezoto 2012). We did not categorize or weight these dominance behaviors according to the level of escalation. We counted foraging trips that resulted in workers bringing water, honey, pulp, or flesh back to the nest. The data were analyzed using a log-linked Poisson general linear mixed model (GLMM), with the frequency of each variable of interest (lateral wagging, dominance act, foraging trip) and workers’ age (in days) included as an explanatory variable and the number of ovipositions as the response variable. Colonies and workers were treated as random effects, and observation time was treated as an offset variable. We used the lme4 package in R software (R Development Core Team) and selected the best models according to the Akaike information criterion (AIC). For GLMM analyses, the observation data were divided into four sessions with almost-identical day-period intervals according to colony development. The first session was before 7 July, the second was from 8 to 14 July, the third was from 15 to 21 July, and the fourth was from 22 July. The third and fourth sessions corresponded to the beginning of egg production by reproductive wasps of the next generation. A queen and her daughter workers were present in each colony in each session.

Colony CHC analyses

Each egg was collected using our handmade cell cups, which were previously inserted into empty cells to collect eggs without destruction or chemical contamination. The empty cells were produced by removal of the original eggs with fine forceps. The cell cup was made of filter paper, arranged using glue into the shape of a bucket without a bottom, by cutting off the tip of a triangular pyramid; using the inner wall of a 200-μL PCR tube, the filter paper was arranged in a pyramid, and the tip was cut off using scissors. The cups were immersed in anhydrous hexane (Sigma-Aldrich) for more than 24 hours. The cups were subsequently dried at room temperature and inserted into the empty cells. After oviposition on the cell cup by a wasp, the cup was gently withdrawn, and each egg was immersed in 50 μL of hexane for 24 hours at –20°C, after removing as much of the filter paper from around the egg as possible. The eggs were then removed from the hexane, and the remaining solution was kept at –20°C. Before hexane immersion, some of the collected eggs were gently placed within a controlled-temperature cabinet, maintained at 26°C ± 1°C with a 14L:10D photoperiod and 95% relative humidity, to determine the hatching rates. The differences in hatchability between the eggs produced by the queen and reproductive workers were analyzed using a log-linked binomial GLM. Colony was treated as a random effect. We collected and analyzed 220 eggs from 15 colonies from 2008 and 2010: we collected the eggs in both queenright and queenless conditions from eight of the colonies; we collected from five of the colonies under field conditions; and we collected eggs under laboratory conditions from ten of the colonies.

At the end of our experiment (end of July), each adult was collected in an individual clean glass vial and stored at –80°C. CHCs were extracted by dripping on the adult body (abdomen) with 100 μL of hexane, and the resulting solution was stored at –20°C. The abdomens were dissected and the ovarian status was categorized into three groups: well-developed (more than two matured eggs), developed (one or two matured eggs), or undeveloped (zero matured eggs). The former two classes corresponded to reproductive workers, while the latter corresponded to non-reproductive workers. We collected eggs produced by both the queen and by reproductive workers, and under both queenright and queenless conditions. We collected body CHCs (n = 184) from the queen, reproductive workers, and non-reproductive workers from 15 queenright colonies.

One-twelfth of the solution extracted from an egg, and 1/200 of that extracted from the body, were analyzed by gas chromatography-mass spectrometry (GC–MS) (HP6890; Hewlett-Packard) with an HP5973 mass spectrometer system (Hewlett-Packard) equipped with an HP-5MS column (0.25 mm ID × 30 m, 0.25 μm film thickness; Hewlett-Packard). The column temperature program for the former column was 150°C for 2 min, increasing by 7°C/min up to 325°C, followed by maintenance at this temperature for 6 min. The carrier gas was He. Identification of cuticular compounds was performed based on their mass spectra, as produced by electron impact ionization (70 eV).

The solutions extracted from the eggs and abdomens were analyzed by a gas chromatography-flame ionization detector (GC–FID) (GC-17A ver. 3.0; Shimadzu) under the same conditions as described for GC-MS, and the temperature of the FID was 325°C. The area of the peak on each chromatograph was estimated using CLASS-GC10 software (Shimadzu). We calculated the proportional area of each peak to determine the total area of all peaks for each individual and egg, and analyzed CHCs only when the mean proportion exceeded 0.5%.

We performed non-metric multidimensional scaling (NMDS) to visualize and analyze profiles of CHC differences between each egg and adult class. Permutational multivariate analysis of variance (PERMANOVA) was conducted to compare CHC profiles among classes using R software. We also analyzed CHC profiles using the random forest ensemble learning method, which applies decision trees to improve predictive performance (Breiman 2001), using R software. In these analyses, the relative importance of each CHC component among all CHCs was evaluated using the mean decrease in the Gini index. From these results, we extracted the five most important CHC components.

Analysis of colonies under laboratory conditions

Before the first worker emerged, 15 and 33 queenright nests were collected in the field around Gifu city in 2012 and 2013, respectively, and transported to the laboratory. The nests were attached to the wall of a cardboard box (30 × 30 × 60 cm). We cut lines 5 cm from the edge of each wall, except the wall to which the nest was attached (30 × 30 cm), and covered the remaining spaces with plastic film (Saran Wrap). The original foundress was released on the nest after weak immobilization on ice and fed with chopped mealworm (Tribolium castaneum) and silkworm (Bombyx mori) ad libitum. Tap water and honey were also supplied. The boxes were placed so that the top wall was 20 cm below a 20 W fluorescent light and maintained at 26°C ± 1°C with a 14L:10D photoperiod. Newly emerged adults discovered on daily inspection were collected, individually marked, and gently released back onto the nests.

JH extraction and liquid chromatography–mass spectrometry (LC–MS)

Each wasp in each class of workers, foundresses, and queens as described below was collected from the queenright colonies under room conditions and cooled on ice in 2013 and 2014. Hemolymph was subsequently collected by gently inserting a microcapillary tube between abdominal segments. After collection, the mouth of the remaining body was sealed with glue, centrifuged at 5,000 rpm for 5 min, and the remaining hemolymph was collected. The bodies were then dissected and the ovarian condition was checked. Only hemolymph that was clear yellow in color was used in further analysis. The amount of hemolymph collected from each wasp was measured, and JH was extracted from the samples following the methods of previous studies (Westerlund and Hoffman 2004; Cornette et al. 2008). Each hemolymph sample from 1 to 8 individuals was pooled prior to analysis. Totally, we used 172 individuals from 108 colonies for the extraction. Aliquots of 17 μL of hemolymph were homogenized in 400 mL methanol and allowed to stand at room temperature for 5 min. After centrifugation at 10,000 rpm for 1 min, the supernatant was collected and 30 ng fenoxycarb (Wako) was added as an internal standard. Then, 100 μL 2% NaCl solution and 300 μL hexane were added to the mixture, which was vortexed and allowed to stand at room temperature for 5 min. The mixture was centrifuged at 3,200 rpm for 5 min and the hexane phase was then transferred into a new glass vial. This process was repeated three times; finally, 900 μL hexane was obtained. The resulting mixture was stored at –80°C before vacuum drying in a CC-181 centrifugal concentrator (Tomy). Dried pellets were dissolved in 20 μL acetonitrile.

LC–MS was performed using 20 μL each of concentrated sample, following the procedure described by Cornette et al. (2008). Aliquots of 5 μL were separated on a 150 × 2-mm² inner diameter C₁₈ reverse-phased column (YMC-Pack Pro C-18.5 μm; YMC Co., Ltd.) protected by a guard column (YMC-Pack Pro, sphere ODS; YMC Co., Ltd.) eluted with a gradient of water/methanol (80–100% methanol over 0–15 min, 100% methanol for 5 min) at a flow rate of 0.2 mL/min using an Agilent 1100 high-performance LC (HPLC) system with an autosampler (Agilent Technologies). Mass spectral analysis was performed by electrospray ionization in positive ion mode on a micro-high-resolution time of flight (TOF-HS) spectrometer (Bruker Daltonik) with the electrospray capillary set at 4.5 kV and under a drying temperature of 200°C. The nitrogen pressure of the nebulizer was 1.6 bar and the drying gas nitrogen flow rate was 9 L/min. Quantification of JH III and fenoxycarb was performed by monitoring the [M+H]⁺ and [M+Na] ⁺ ions. For each sample, a calibration curve for JH III (Sigma-Aldrich) was plotted using the same internal standard concentration as for fenoxycarb. The JH III titer from each sample was then calculated after analysis of the chromatogram data using QuantAnalysis software (Bruker Daltonik). Data are expressed as ng/wasp. JH titers were measured for the foundress, queen, and non-reproductive workers at 2 days of age, and again for non-reproductive and reproductive workers at 30 and 32 days of age, respectively.

Topical application of JH and measurement of biogenic amine levels

Two 0-day-old wasps were collected from each of 5 of 48 colonies in 2013 and 2014 under room conditions with forceps and immobilized on ice. One of the wasps was topically treated with 2 μL acetone as a control, and the other was treated with 2 μL JH III (Sigma-Aldrich) solution (50 μg/μL in acetone) using a micro-syringe. After the treatments, the wasps were gently released into their original colonies. The two wasps were collected again at 4- or 8-day, and immediately frozen with liquid nitrogen. The brains of the wasps were dissected out from the heads on a Peltier cooling unit under a dissecting microscope. The dissected brains were homogenized in a micro-glass homogenizer in 50 mL ice-cold 0.1 M perchloric acid containing 12.5 ng/mL 3,4-dihydroxybenzylamine (DHBA) as an internal standard. Each sample was then transferred into a 1.5-mL Eppendorf tube, and centrifuged at 15,000 rpm for 30 min at 0°C. The supernatants were transferred to micro-vials for analysis by HPLC-electrochemical detection (ECD).

The HPLC system consisted of a solvent delivery pump (EP- 300; Eicom), a refrigerated automatic injector (231–401; Gilson), and a C₁₈ reversed-phase column (UG 120; Shiseido) maintained at 35°C in a column oven. An electrochemical detector (ECD-300; Eicom) with a glassy carbon electrode (WE-GC; Eicom) was used. The detector potential was usually set at 0.87 V against an Ag/AgCl reference electrode. The detector cell was held at a constant temperature of 35°C by placing it in the column oven. Signals from the ECD were recorded and integrated using data analysis software (PowerChrom; ADInstruments).

The mobile phase contained 0.18 M monochloroacetic acid and 40 mM 2Na-EDTA, which was adjusted to pH 3.6 with NaOH. Then, 1.62 mM of sodium-1-octanesulfonate was added into this solution as an ion-pair reagent and 7.4% CH₃CN as an organic modifier. The mobile phase buffer was filtered through a 0.22-μm filter (Millipore) and degassed. The flow rate was kept constant at 0.7 mL/min.

External standards were run before and after the sample runs. External standards [octopamine (OA), N-acetyldopamine (NADA), dopamine (DA), N-acetyltyramine (NATA), tyramine (TA), N-acetyl-5-hydroxytryptamine (NA5HT), tryptophan (TRP)], and serotonin [5-hydroxytryptamine (5HT)] were used for identification and quantification of biogenic amines. The peaks of each biogenic amine were identified by comparison of both the retention time and hydrodynamic voltammograms with those of the standards. Measurements based on the peak area of the chromatograms were obtained by calculating the ratio of the peak area of a substance to the peak area of the external standard. Amounts of biogenic amines were calculated in units of pmol/brain.

We fitted GLMMs to clarify the effects of JH treatment and workers’ age (4 or 8 days old) on each monoamine. We presumed the objective variables, i.e., each monoamine, followed a normal (Gaussian) distribution with an identity link function. We set each colony and sample as the random intercept of all models. The analyses were performed using R software and the MCMCglmm library (ver. 2.24; Hadfield 2010), which is a package for fitting GLMMs using Markov Chain Monte Carlo (MCMC) techniques, and the estimate of each explanatory variable is given by its posterior distribution. We used 50,000 iterations after 5,000 burn-in iterations and set 2% as the thinning rate (1,000 samples were picked for the posterior distributions) to run the calculation. As MCMCglmm has no facility for setting multiple chains, we ran three MCMC calculations independently and the results were merged for calculation of R-hat. An R-hat value < 1.1 confirms the convergence of MCMC sampling for each explanatory variable (Gelman et al. 2004). We accepted an explanatory variable as being significant if the 95% confidence interval of its posterior distribution did not contain 0.

RNA extraction and quantitative reverse-transcription polymerase chain reaction (qRT-PCR)

The heads of fresh queens (n = 6), reproductive workers (n = 10), and non-reproductive workers (n = 10) taken from the colonies under room conditions in 2013 and 2014 were immersed in liquid nitrogen and stored at –80°C. Total RNA was extracted from the heads with TRIzol reagent (Life Technologies) according to the manufacturer’s instructions. Briefly, each frozen head (kept at –80°C) of three adults’ castes was put into clean microtubes, and each head sample was roughly cut for filling with TRIzol reagent. Immediately, 300 μL of TRIzol reagent was added to each tube, and the samples were homogenized. The quality of total RNA was checked by spectrophotometry. For cDNA synthesis, 1 mg aliquots of total RNA were used in conjunction with the PrimeScript RT Reagent Kit with gDNA Eraser (TaKaRa). Each qRT-PCR mixture (12.5 μL) contained 0.5 μL of first-strand cDNA. Real-time detection and analysis of DNA were performed based on SYBR green dye chemistry using a SYBR Premix Ex Taq Perfect Real Time Kit (TaKaRa) and a Thermal Cycler Dice Real Time System (model TP700; TaKaRa). PCR was performed using primers for eight genes—insulin like-receptor 1 (ILR1), target of rapamycin (TOR), dopamine receptor D1 (DOP1), dopamine receptor D2 (DOP2), ultraspiracle (USP), vitellogenin (Vg), malvolio (malV), foraging (For)—designed based on the RNA-Seq data of Ferreira et al. (2013) or Toth et al. (2007). We selected these target genes because they are associated with social behavior (e.g., Ben-Shahar et al. 2003; Guidugli et al. 2005; Vergoz et al. 2007; Wheeler et al. 2014). The primer sequences are listed in Table S1. The mRNA values were normalized relative to ribosomal protein RPL37.

The wasp bodies were dissected and for evaluation of ovarian development; here, the ovarian development index was calculated as the sum of the first- and second-ranking scores of maximum length × width among basal oocytes in the ovarioles in each pair of ovaries according to Sasaki et al. (2009). The index was used to distinguish non-reproductive workers (index <0.4) from reproductive workers (≥0.4) in our experiments on gene expression and JH. This index-based bifurcation was consistent with the ovarian status used in our behavioral observation, as workers with undeveloped eggs, those with developed eggs, and those with well-developed eggs were estimated to have average ovarian index values of 0.1, 0.6, and 1.7, respectively (data not shown).