Facultative parthenogenesis is a form of reproduction in which females can either lay unfertilised eggs that typically develop into female offspring only, or mate and lay fertilised eggs that develop into male and female offspring. Facultative parthenogens often occur in mixed-sex populations where reproduction is mostly sexual, and all-female populations where reproduction is asexual. How all-female populations avoid invasion by males remains unknown. Here, we investigated the role of volatile and non-volatile (cuticular hydrocarbons, CHCs) pheromones in the persistence of all-female populations in the facultatively parthenogenetic stick insect, Megacrania batesii. We found that M. batesii exhibits slight sexual dimorphism in antenna morphology, and behavioural assays provided little evidence that males could locate females solely by volatile pheromones. However, CHC profiles differed substantially between different types of females. Analysis of CHC structure and abundance indicated a clear genetic difference between females from all-female versus mixed-sex populations, as well as a maternal effect of female parthenogenesis versus sexual development. Together, our results suggest that males might rely more on close-range chemical cues to differentiate females, and chemical communication could play a role in the persistence of all-female populations.

Animal maintenance

Individuals for all experiments were offspring derived from hatchlings or eggs collected from mixed-sex and all-female populations in northern Queensland between 2019 and 2022. All males and females from mixed-sex populations were collected from populations of the northern genotype, whilst females from all-female populations were collected from populations of the southern genotype. Hatchlings were first group-reared with up to four siblings on a live Pandanus tectorius host plant, then housed separately in a cylindrical plastic container (height = 39.5 cm, radius = 10 cm) covered with a mesh lid. The insects were all kept at approximately 27°C, with humidity of around 40%, and they were sprayed with reverse-osmosis water daily. Individuals housed without live plants were fed cut Pandanus leaves, which were replaced three times a week. All individuals used in experiments were reproductively mature adults. The paternity of individuals were manipulated by the control of mating status of their mothers.

Antenna morphology

To test for sexual dimorphism in antennal morphology, antenna static allometry and sensilla density were quantified in the two sexes. Antenna static allometry was quantified by measuring antenna and body size on digital photos of paternate northern males and females (40 individuals of each sex). Static allometry refers to the scaling relationship between a log-transformed trait (i.e., antenna) size and body size. Thorax length (measured from the anterior margin of the prothorax to the posterior margin of the mesonotum) was measured as a proxy for body size. Length of the entire antennae (from the first to the last antennomere) was measured in each individual. We included only one antenna per individual in the analyses: using only intact antennae, and selecting the longer of the two when both antennae were intact.

To compare sensilla density between paternate northern males and females, we removed one intact antenna from each individual (20 males and 20 females, which were all euthanised and kept in a -80°C freezer until the investigation). We used intact antennae only, selecting the longest of the two when both antennae of a specimen were intact. We took photos of the last antennomere using a camera (Leica, Flexacam C1) mounted on a stereo-microscope (Leica, MZ 16 A) at a magnification of 90x. Sensilla density was observed to increase from the base to the tip of the antenna, so we assumed that the last (i.e., most distal) antennomere had the highest density of sensilla. We used the densest part of the antenna because it effectively indicates olfactory sensitivity. We counted the number of sensilla (visible as bumps on the antenna surface) inside a randomly selected rectangular area (size approximately from 0.025 to 0.03 mm²⁾ on each antenna, and we divided the number of sensilla by the area of the rectangle to estimate sensilla density. We also measured the length of the last 20 antennomeres of these antennae to test for potential covariation of antenna length with sensilla density. All image measurements were performed in imageJ.

Volatile pheromone behavioural experiment

To test whether males respond to olfactory cues from females, we placed males in an open arena with hidden females and measured their activity levels and attraction to females. We ran two types of trials: North-South trials (impaternate northern females versus impaternate southern females) to test for genetic effects, and paternate-impaternate trials (impaternate northern females versus paternare northern females) to test for maternal effects. The equilateral triangular arena comprised four areas, including three smaller triangles at the corners (hereafter referred to as zones). Each zone contained a black mesh opaque hood (height = 49 cm, radius = 15.3 cm, mesh size = 0.15 mm²) with a smaller mesh container (height = 39.5 cm, radius = 10 cm, mesh size = 1.8 mm²) inside. In each trial, one of the three containers was empty (control), and the other two contained one female each. The black mesh hoods were used to prevent the use of visual cues by males but allow female volatiles to disseminate throughout the arena. The position of the containers was randomised for every trial to control for the effect of position. Each trial commenced when a focal male was placed on top of a plastic cylinder in the middle of the arena.

The arena was continuously video-recorded using tablets (Samsung T580) for 12 hours under red ceiling lights, during the night. We ran the trials at night because this is when these insects are most active. We put sackcloth on the arena floor to mimic rough topsoil and to make it easier for males to walk between zones, and we watered the arena, hoods and containers before every trial to increase humidity. Two trials (one of each type) were performed simultaneously in two identical controlled-temperature rooms (27°C) four nights per week, over seven weeks, alternating which trial type (North-South trials and paternate-impaternate trials) was in which room. The arenas had constant air-flow through vents on the ceiling (six air changes per hour) so that there was air movement for the focal males to follow pheromone plumes. The floor area of each room was 12 m². All individuals were well fed on the day before the trial. No Pandanus leaves were placed in the arenas to avoid males reacting to the smell of the leaves or the sound of feeding females, rather than to the smell of the females.

We measured the total time spent by the focal male in each zone, the time from the start of a trial until the male first reached each container (latency), and the total number of visits to all zones in one trial. We performed 30 trials for each trial type, using 16 paternate northern females, 17 impaternate northern females, 17 impaternate southern females, and 31 males in total. 29 of the males were used in both trial types, and each of the two remaining males was only used in one of the trial types. Individuals were reused in a subsequent trial after a minimum of three days from the previous trial. We randomised which trial type each male underwent first. Most paternate northern and impaternate southern females were used twice, whilst most impaternate northern females were used a total of four times (twice in each trial type). We were not able to standardise the mating status of all individuals used in this experiment, so we compared mated females to mated females, and unmated females to unmated females in each North-South trial to eliminate the effect of mating status. However, in paternate-impaternate trials, mated paternate northern females were compared with unmated impaternate southern females in 21 out of 30 trials. Our experiment might have therefore underestimated male preference for paternate females if males also preferred unmated females.

CHC extraction

After all behavioural assays had been completed, four females from each type were frozen at -80°C. Males used in the CHC assay were euthanised by freezing at the same time as females but were not involved in the behavioural assays described above. All individuals had mated at least once and were stored in a -80°C freezer until CHC extraction. Prior to immersion for CHC extraction, all individuals were thawed for 10 minutes at room temperature. Whole insect immersion in hexane is the most common way to extract pheromones from small insect species, but given the relatively large size of M. batesii, a custom procedure was used to extract their CHCs. Each individual was placed in a cylindrical flask, covered with aluminium foil and gently swirled for ten minutes in 5 mL of n-hexane (Sigma-Aldrich, Castle Hill, NSW, Australia, product no. 650552) to ensure the hexane washed over the whole insect. The hexane was spiked with 10 micrograms/mL of hexacosane (Sigma-Aldrich, product no.241687) as an internal standard. Water was removed from each extraction by filtering it through a glass Pasteur pipette packed with silane-treated glass wool (Alltech, Australia) and a small amount of anhydrous sodium sulphate (AjaxFinechem, Seven Hills, NSW, Australia, product no.503-500). Extracts were then transferred into 2 mL autosampler vials (Agilent Technologies, Mulgrave, VIC, Australia) and stored in a -20°C freezer until chemical analysis. 

Data analyses

All data analyses were carried out in R 4.1.3. To examine antenna length static allometry and relative antenna length, we built an ordinary least-squares (OLS) model using the lme4 package and, for comparison, a standardised major axis (SMA) linear model using the smatr package. SMA models assume equal error in the measurement of antenna length (y-variable) and thorax length (x-variable), so the true allometric slopes would not be underestimated, and there would not be bias that could produce misleading results when comparing male and female slopes. By contrast, an OLS model attributes all error to the y-variable, which could be more appropriate if the measurement error in curved antennae exceeds the measurement error in straight thoraxes. We conducted an analysis of covariance (ANCOVA) for OLS models and a likelihood ratio test for SMA models to compare the allometric slopes of the two sexes. The thorax length and antenna length were log-transformed, as the trait size-body size relationship is assumed to be exponential and log-transformation allows the exponent to be estimated as the slope from a linear model. A Welch’s t-test was performed to compare sensilla density between the sexes, and an ANCOVA was used to identify the relationship between sensilla density and antenna length of each sex.

Data from the pheromone behavioural experiment was analysed using generalised linear mixed models with a Gaussian distribution to test for the effect of female type on male responses (i.e., time spent in and latency to reach each zone) in each trial type separately, using the glmmTMB package. Zone type (i.e., the two female types and control) was modelled as the fixed effect, male identity as the random effect, and time spent in each zone and latency to each zone by focal males as response variables in separate models. The amount of time spent in each zone was log-transformed and latencies to each zone were square root-transformed to improve the distribution of residuals. We excluded 12 North-South trials and eight paterante-impaternate trials where the male did not visit any of the zones in these analyses. Likelihood ratio tests were performed by comparing each full model to the corresponding null model, where the fixed effect was removed. A near-significant effect was found for the paternate-impaternate trials. We therefore used a post-hoc (Tukey) test to compare the time spent in each zone using the emmeans package. Because focal females varied in age, which can affect female attractiveness, we also investigated the effect of female age (number of days since the adult moult) on male response in both trial types, by building generalised linear mixed models where female age was modelled as a fixed effect together with zone type. Female age and female type (i.e. zone type, excluding the control zone) were modelled as the fixed effects. Male identity was modelled as the random effect, and male response (i.e. either amount of time spent in zones or latency to zones) was the response variable. Amount of time spent in each zone and latencies were log-transformed, except for the latencies in North-South trials, to improve model fit (based on inspection of model residuals). Female age was standardised within female types to remove the correlation between female age and female type. All trials were included in analyses involving fixed effects of both zone type and female age.

In addition, we built another generalised linear mixed model as an exploratory analysis to investigate the effect of trial type on overall male activity (total number of visits to zones). This model was fitted with a Poisson distribution and a log link. Trial type was modelled as the fixed effect, male identity as the random effect, and the total number of visits as the response variable. We included trials where the male did not visit any of the zones in this analysis, but excluded two outliers, in which the two males made many more zone visits, to improve the distribution of residuals.

The effect of sex and female type on CHC profile was investigated by performing a principal component analysis (PCA) of centred log-ratio (CLR) coefficients of the proportional peak areas to identify covariation patterns of CHC profiles by insect type. CLR transformation was applied to make these compositional data suitable for analysis by standard multivariate techniques. CHC data is compositional, and different components can potentially act synergistically, making it more meaningful to analyse the entire profile instead of single components. The peak area of each component was first divided by the peak area of the internal standard of the corresponding sample to control for differences between sample runs on the GCMS. This relative peak area was further divided by the total relative peak area of the corresponding sample to avoid confounding body size with total CHC amount. CLRs were then calculated by dividing the proportional peak area by the geometric mean of the detected peaks in the corresponding sample. 10^-17 was added to all proportion peak areas before CLR transformation to avoid zero values because some peaks were present in only one type of insect.