Does ecology shape geographical parthenogenesis? Evidence from the facultatively parthenogenetic stick insect Megacrania batesii

Abstract Closely related sexual and parthenogenetic species often show distinct distribution patterns, known as geographical parthenogenesis. Similar patterns, characterized by the existence of separate sexual and parthenogenetic populations across their natural range, can also be found in facultative parthenogens – species in which every female is capable of both sexual and parthenogenetic reproduction. The underlying mechanisms driving this phenomenon in nature remain unclear. Features of the habitat, such as differences in host‐plant phenotypes or niche breadth, could favour sexual or asexual reproductive modes and thus help to explain geographical parthenogenesis in natural insect populations. Megacrania batesii is a facultatively parthenogenetic stick insect that displays geographical parthenogenesis in the wild. We aimed to explore whether sexual and parthenogenetic populations of M. batesii displayed niche differentiation or variations in niche breadth that could explain the separation of the two population types. To do this, we sampled host plants from across the range of M. batesii and quantified phenotypic traits that might affect palatability or accessibility for M. batesii, including leaf thickness, toughness, spike size and density, plant height, and chemical composition. We also quantified host‐plant density, which could affect M. batesii dispersal. We found little evidence of phenotypic differences between host plants supporting sexual versus asexual M. batesii populations, and no difference in host‐plant density or niche breadth between the two population types. Our results suggest that habitat parameters do not play a substantial role in shaping patterns of geographical parthenogenesis in wild populations of M. batesii. Instead, population sex ratio variation could result from interactions between the sexes or dispersal dynamics.


| INTRODUC TI ON
In the wild, parthenogenetic and closely related sexual species often show a pattern of differing reproductive modes over a geographic landscape, a phenomenon known as geographic parthenogenesis (Vandel, 1928;Lynch, 1984).The pattern of geographical parthenogenesis is highly variable.In some taxa, sexuals and parthenogens are clearly separated along climatic or ecological gradients.For example, parthenogenetic lineages are sometimes found in habitats at the edge of the range (Bell, 1982) or at higher latitudes or altitudes (Verduijn et al., 2004), and have been associated with habitats affected by Pleistocene glacial cycles (Glesener & Tilman, 1978;Hörandl, 2009;Suomalainen et al., 1987).In other taxa, sexuals and parthenogens are interspersed in spatial mosaics with no clear environmental gradient (reviewed in Tilquin & Kokko, 2016).While the existence of geographical parthenogenesis is well established empirically (Bierzychudek, 1989;Hörandl, 2006;Hörandl et al., 2008;Kearney, 2005;Verduijn et al., 2004), there is no accepted general explanation for how these patterns might arise and persist in natural populations.It has been suggested that differences in ecological niches available to sexual and parthenogenetic animals could explain the geographic distribution patterns of reproductive modes seen in the wild (Case & Taper, 1986;Halkett et al., 2006;Lehto & Haag, 2010;Meirmans et al., 2012), but the role of ecological niche in promoting sexual vs. parthenogenetic reproduction remains poorly understood.
Facultative parthenogens -organisms that can reproduce both sexually and parthenogenetically -offer a convenient system to better understand geographical parthenogenesis by investigating ecological niches available to sexually versus parthenogenetically produced conspecifics.While the literature on geographical parthenogenesis has mostly focused on separate obligate sexual and parthenogenetic species (reviewed in Kearney, 2005;Hörandl, 2009), some facultative parthenogens exhibit distinct sexual and parthenogenetic populations across their ranges (Burns et al., 2018;Miller et al., 2024;Morgan-Richards et al., 2010).Such variation in sex ratio and reproductive mode between populations in facultatively parthenogenetic species can be stable over many generations and can be regarded as an example of geographical parthenogenesis (Burns et al., 2018;Cermak & Hasenpusch, 2000;Morgan-Richards, 2023;Tsurusaki, 1986).However, the factors shaping geographic patterns of parthenogenesis in facultative parthenogenetic systems may be more nuanced than those involved in separate obligately sexual/ parthenogenetic species.Exploring the environmental factors shaping geographical parthenogenesis in facultatively parthenogenetic species could enhance our understanding of how ecological factors might influence the evolution of reproductive modes in nature while removing factors that may be due to speciation.Indeed, geographical parthenogenesis in facultative species could be especially informative of the early stages of transitions between reproductive modes (Miller et al., 2024).
Several theories have been developed to explain the reproductive mode variation found in natural populations of parthenogenic and related sexual animals (reviewed in Vrijenhoek & Parker, 2009;Tilquin & Kokko, 2016).In particular, it has been hypothesized that, in comparison to sexual lineages, parthenogens could evolve generalist strategies that could enable them to colonize marginal, extreme, or changing habitats (i.e. a 'general purpose genotype'; Lynch, 1984).
Alternatively, it has been hypothesized that parthenogens may experience selective sweeps that create populations of clones that are highly specialized for specific environments (i.e. a 'frozen-niche' mechanism; Vrijenhoek, 1979).Attempts to test these alternative predictions have mainly focused on study systems that are apomictic, hybrids (which are often also polyploid), or automictic parthenogens that reproduce via central fusion (Bierzychudek, 1989;Hersh, 2020;Oplaat & Verhoeven, 2015;Van Der Kooi et al., 2019).These reproductive pathways often lead to stable or even increased heterozygosity when compared to sexual relatives (Jaron et al., 2021;Marescalchi & Scali, 2003;Suomalainen et al., 1987).By contrast, few studies have investigated automictic parthenogens with diploidy restoration mechanisms that result in very low heterozygosity and allelic diversity (e.g.parthenogenesis via terminal fusion or gamete duplication).It remains unclear how automictic parthenogens compare ecologically to their sexual counterparts that have higher genetic diversity (see Larose et al., 2018).
The Peppermint Stick Insect, Megacrania batesii, is a facultative parthenogen with a restricted Australian range in tropical north Queensland rainforest (Cermak & Hasenpusch, 2000).Every female is capable of both sexual and parthenogenetic reproduction, with unfertilized eggs hatching exclusively into daughters and fertilized eggs hatching into approximately equal ratios of daughters and sons (Miller et al., 2024;Wilner et al., in prep.).Parthenogenetic reproduction in this species occurs via automixis, and results in very low or zero heterozygosity in offspring (Miller et al., 2024;Miller et al., in prep.).In the wild, despite the capacity of each female to switch between reproductive modes, distinct mixed-sex and all-female populations of M. batesii are found across the range (Miller et al., 2024).
Mixed-sex populations typically exhibit an equal sex ratio, and the individuals there mainly undergo sexual reproduction; all-female populations consist exclusively of females, reproducing parthenogenetically.All-female M. batesii populations predominate at the northern and southern edges of the known species range, while mixed-sex populations predominate in the central part of the range (Miller et al., 2024).Some mixed-sex and all-female populations exist in close proximity, and it is unclear what factors are preventing males from mixed-sex populations from invading nearby all-female In their natural range in Australia, M. batesii can be found on several species of host plant belonging to two genera, Pandanus and Benstonea (Miller et al., 2024).In addition to food, the host plants also offer protection to M. batesii, which rest within the median grooves of the leaves during the day (see Figure 1e However, given high phenotypic variation within each host-plant species and the lack of a full molecular phylogeny for this group, as well as a history of hybridization, it is difficult to identify these plants to the species level (see Buerki et al., 2012) therefore favour the establishment of all-female populations because only one female would need to reach the new host plant to colonize that area; whereas both a male and female would need to reach a habitat patch to form a new mixed-sex population.Thus, small-scale environmental heterogeneity in the form of host-plant phenotypic and distributional variation could contribute to the pattern of geographic parthenogenesis observed in natural populations of M. batesii.
We also observed that certain host plants appear to remain completely untouched by M. batesii, despite being located adjacent to, and sometimes even in contact with, similar-looking plants that have been heavily consumed by M. batesii.This suggests that there could be morphological or chemical differences between eaten and uneaten host plants.For example, plants may remain uneaten if they have thicker leaves or larger spikes that curb herbivory.Leaf thickness and toughness are common defences against herbivores and are often negatively correlated with herbivory (Caldwell et al., 2015;Clissold et al., 2009;Fine et al., 2004;Kursar & Coley, 2003;Peeters, 2002;Pringle et al., 2011).Pandanus and Benstonea species used as host plants by M. batesii possess spikes that line the edge of the leaves, and these spikes could function as a deterrent against herbivory by M. batesii because these insects initiate feeding at the edge of the leaf (Cermak & Hasenpusch, 2000).In particular, the density, length and angle of the spikes could affect how easily M. batesii can initiate feeding.Some of the host plants we observed in the field were mature trees, with edible leaves only occurring at the top of the tree (see Figure 1a), whereas other host plants were immature or growing as shrubs lacking a woody trunk, providing more easily accessible leaves (see Figure 1c,d).Thus, host-plant height could also affect M. batesii herbivory as taller trees may impede M. batesii dispersal to the section of the tree with edible leaves, requiring the insects to be exposed to predation for longer to get to their food source.Additionally, many plants produce chemical defences against herbivorous insects, with high carbon to nitrogen ratios (C/N) being associated with accumulation of carbon-based phenolics which can reduce herbivory (Bryant et al., 1983;Fürstenberg-Hägg et al., 2013;Lindroth et al., 2001;Wittstock & Gershenzon, 2002).Nitrogen content is especially important in plant-insect interactions as protein and nitrogen concentrations are positively correlated in leaves (Throop & Lerdau, 2004), and studies have shown that nitrogen concentrations directly affect growth and diet choice in many herbivorous insects, leading to insect preferences for higher nitrogen in leaves (Mattson, 1980;Osier & Lindroth, 2001;Rausher, 1981).
Thus, host plants may vary in morphological or chemical traits that affect palatability for M. batesii.
In this study, we aimed to quantify and compare host-plant phenotypes from habitats hosting sexual versus parthenogenetic stick insects in their natural range.Specifically, we aimed to understand whether parthenogenetic M. batesii show patterns consistent with the general purpose genotype theory or the frozen niche variation theory to explain the geographic parthenogenesis observed in the wild.Under the general purpose genotype theory, we would expect populations of parthenogenetic stick insects to be found in habitats with host plants that are more morphologically or chemically diverse compared to the host plants in habitats hosting sexual stick insect populations.This would suggest that parthenogenetic M. batesii lineages have a wider niche breadth than their sexual counterparts, reflecting broader tolerance for morphological and chemical variation in their host plants.In contrast, the frozen niche theory predicts that parthenogens are highly specialized with a narrow niche breadth.
Under this theory, we would expect that all-female populations of M. batesii will be found in habitats with little diversity of host-plant phenotypic and chemical traits, possibly with different but similarly narrow niches for each all-female population.Additionally, we aimed to test the prediction that parthenogenetic populations are more likely to establish and persist in low-density habitats.Lastly, we asked whether there were chemical or morphological differences between host plants that were eaten versus host plants that were avoided entirely by M. batesii, and whether parthenogenetic M. batesii have greater or lesser tolerances for host-plant morphological and chemical defences.To address these questions, we quantified variation in host-plant leaf morphology (leaf thickness, toughness, and slenderness, spine traits) and chemistry (carbon, hydrogen and nitrogen content), and host-plant height and density (in 10 × 10 m quadrats) in several mixed-sex and all-female populations of M. batesii within their natural range in far-north Queensland.

| Sampling locations
We studied the habitats of mixed-sex (sexual) and all-female (parthenogenetic) populations of M. batesii across a small latitudinal range (0.24°) along ~35 km of coastline between Cape Kimberley and Cape Tribulation, Queensland (see Figure 2).We also investigated two isolated all-female populations located ~160-180 km (1.56°) south of the Daintree River at Etty Bay and Bingil Bay.Mean daily maximum temperature varies from 27.9°C in the south to 29.4°C in the north (Bureau of Meteorology; bom.gov.au).The entire species range experiences very high rainfall, with mean annual rainfall ranging from 3363 mm in the south (BOM) to 4331 mm in the north (Daintree Rainforest Observatory https:// www.jcu.edu.au/ daint ree/ resou rces/ dro-rainf all-data), both well above the threshold of ~2000 mm per year for tropical rainforest (e.g.see Ciemer et al., 2019).The M. batesii populations included in this study are described in Miller et al. (2024), except for one population (MAD).In Miller et al. (2024) and in this study, populations were separated based on geographic (unpublished data).For all the populations included in the present study, a sufficient number of individuals was observed in the field to assign population type with high confidence (see Miller et al., 2024).
To confirm that the observed variation in sex ratio corresponded to variation in reproductive mode, eggs were collected from a subset of the populations and the hatchlings were sexed.As expected, hatchling sex ratios from mixed-sex populations were approximately even, confirming that reproduction in such populations is predominantly sexual, whereas hatchlings from all-female populations were all female, confirming that reproduction is exclusively parthenogenetic (Miller et al., 2024).One 'transitional' population appears to be in the process of transitioning from an all-female population to a mixed-sex population (see Miller et al., 2024).Host plants from the transitional population were excluded from analyses comparing all-female and mixed-sex populations as this population cannot be included in either of those two groups.However, this population was included in the analyses comparing eaten and uneaten host plants because M. batesii population type was not relevant to that question.We collected phenotypic data on a total of 89 Pandanus sp. or Benstonea sp.We also collected density data on a total of 28 quadrats, 14 from sites where mixed-sex populations occur, and 14 from sites where all-female populations occur (see Figure 2 and Table 1).

| Host-plant morphological traits
To quantify plant morphological traits, we measured plant height using a measuring tape for plants <2 m in height, and estimated height in m for taller plants.We measured the thickness of three leaves on each sampled plant with a digital micrometer and averaged these three measurements to obtain a leaf thickness value for the plant.We estimated leaf toughness by quantifying the pressure (in kg/cm 2 ) required to cut from the edge to the midline of each leaf using scissors and a penetrometer.To do this, a small divot was drilled into the handle of the scissors.Then, the base of the penetrometer was pressed into the divot until the scissors cut to the midline of the leaf.Three measurements of leaf toughness from the same leaves were obtained in this way for each plant, and the average computed.Thickness and toughness measurements, as well as photos of leaf spikes were taken at ~30 cm from the base of the leaf when possible and the oldest (closest to the ground, often tough, dry and brittle) and newest (furthest from the ground, often thin and soft) leaves on the plant were avoided.We took photos of the leaves of each sampled plant with a scale, and we measured the average length of 5 spikes along the leaf margin from base to tip (mm), as well as the average perpendicular distance between the tip of the spike and the edge of the leaf as a proxy for the angle of the spike (mm), and average distance between spikes (mm) using ImageJ (Schneider et al., 2012).Since M. batesii consumes their host plants by cutting in from the leaf edge, eating the spikes, and continuing inward towards the midline of the leaf, these spike characteristics may serve as a physical defence mechanism for the host plants.The shape of the leaves, specifically their slenderness, may also have functional significance as this could influence the ability of M. batesii to cling to or navigate around the leaves.Additionally, this species of stick insect is relatively large and thus slender leaves may not be able to support the weight of adult M. batesii.To calculate the average leaf slenderness ratio (leaf length/leaf width) of the host plants, we used ImageJ on photos of the entirety of each plant to measure three representative leaves per plant.

| Host-plant chemical traits
To quantify plant chemical traits, we obtained leaf clippings (~3 × 2 cm) from 39 of the sampled plants using scissors.The blades of the scissors were sterilized with ethanol prior to taking each sample.19 clippings were obtained from habitats hosting mixed-sex populations of M. batesii and 20 clippings were obtained from habitats hosting all-female populations; no clippings were taken from the transitional M. batesii population.The clippings were dried at 55°C for 48 h using a Thermoline Scientific Laboratory Oven (model: TEO-66F; https:// www.therm oline.com.au/ ) at the JCU Daintree Rainforest Observatory (https:// www.jcu.edu.au/ daintree).Dried leaf samples were then sent to the Mark Wainwright Analytical TA B L E 1 Sites and sample sizes for plant and habitat data.Site codes correspond to codes in Miller et al. (2024)

| Host-plant density
To quantify density of host plants in sample sites across the range, we sampled 10 × 10 m quadrats from 7 sites hosting mixed-sex populations of M. batesii and 9 sites hosting all-female populations of M. batesii, with some sites having multiple quadrats (see Table 1).Quadrats were sampled only in habitat patches with con-

| Statistical analyses
All analyses and data plots were done using R Statistical Software (v4.2.2; R Core Team 2022).Morphological analyses were based on measurements of plant height, average leaf thickness, average leaf toughness, spike angle (log-transformed to meet the assumption of normality), average distance between spikes, average spike length, and leaf slenderness ratio.Chemical analyses were based on nitrogen, hydrogen, and carbon content of leaves, expressed as percentages of sample mass.For both chemical and morphological analyses, we ran principal components analyses (PCA) using data matrices (either of all the plant functional traits or the plant chemical traits) that were centred, scaled, and then analysed using the function prcomp() from the package 'vegan' (Dixon, 2003).This function computes the principal components using a method based on the singular value decomposition (SVD) of the data matrix.Trait loadings and individual scores on the first two principal components were then visualized using the function ggbiplot() from the package 'ggbiplot' (https:// github.com/ vqv/ ggbiplot).
Both morphological and chemical data were first standardized to account for variation in scale between the different variables.This was done using the vegan function decostand() with the argument method='standardize'.This created a new multivariate object with a mean of zero and variance of one for each variable.Then, a Euclidean distance matrix was created using the function vegdist().We ran a permutation regression model (999 permutations) using this matrix as the response variable and the presence/absence of M. batesii chew marks as well as host-plant genus (to account for genera differences between host plants) and their interaction as fixed predictor variables with the function adonis2().This permutation model partitions sums of squares of a multivariate data set and is analogous to multivariate analysis of variance (permutational MANOVA, formerly known as nonparametric MANOVA; Anderson, 2001).This method can be used to test whether there are differences in the composition or structure of multivariate data (e.g.morphological traits or chemical traits) between groups (e.g.plants with chew marks present vs. absence, or Pandanus sp. vs. Benstonea sp.plants) by partitioning the total variation and comparing the observed differences to a distribution obtained through permutations of randomized group labels.To investigate the interaction between presence/absence of M. batesii chew marks and host-plant genus, we first separated genera and ran the PCA visualizations again for each genus.We paired this with two more permutational models (one for each genus) in adonis() to extract the coefficients and understand how each plant trait was affecting palatability in each genus.To understand how C/N ratios affected herbivory by M. batesii, we used a linear model with C/N ratio as the response variable and presence/absence of M. batesii chew marks as well as host-plant genus (and their interaction) as fixed predictor variables.
Density of host plants in each quadrat (plants/m 2 ) was logtransformed to meet the assumption of normality.A linear mixedmodel regression was performed on the transformed data using the function lmer() from the lme4 package (Bates et al., 2015), with density as the response variable, M. batesii population type (mixed-sex or all-female) as the fixed effect, and site as a random effect.
To understand whether M. batesii from all-female populations (parthenogenetic) occupy different niches in terms of morphological and chemical host-plant traits than their sexual counterparts, we conducted a PCA on the same set of morphological and chemical traits as our analysis on M. batesii preferences.Data collected from the habitat of the transitional M. batesii population, as well as plants that were not found to have M. batesii chew marks, were excluded from all analysis of niche differences, leaving 62 samples for this analysis.These data were centred, scaled, and then analysed with a PCA, using the function prcomp().Trait loadings and individual scores on the first two PCs of both PCAs were then visualized using the function ggbiplot().To test whether there were differences in host-plant chemical and morphological traits between host plants in mixed-sex versus all-female M. batesii populations, we ran the same permutation regression model as described above for M. batesii preferences.However, we adjusted the predictor variable from presence/absence of M. batesii chewmarks to M. batesii population type (mixed-sex or all-female).
We then tested for niche breadth differences, specifically whether a particular reproductive mode was found at sites where host plants exhibited a wider range of morphological or chemical traits.We first wanted to see if the host plants harbouring all-female M. batesii populations exhibit more consistent phenotypes across their range compared to host plants harbouring mixed-sex M. batesii populations.This would show us whether parthenogenetic M. batesii are more likely to persist in habitats with a specific host-plant phenotype.To do this, we first used the betadisper() function, which calculates distances between group members (host-plant phenotypes) and centroid of the group (whether that plant was hosting a mixed-sex or all-female population of M. batesii) by reducing the original Euclidean distances to principal coordinates, an analogue of Levine's test for homogeny of variants, but in a multivariate space (Anderson, 2001;O'Neill & Mathews, 2000).We did this for morphological and chemical traits separately.Then, to see if all-female populations of M.
batesii had a wider or narrower niche overall (i.e.across sites), we ran a permutational analysis using the function permutest() with default settings.This function compared the average group dispersions between phenotypes of host plants found in all-female versus mixed-sex M. batesii populations using a t-test, and then performed a permutation test based on the t statistic derived from that pairwise comparison.We also did this for the morphological and chemical traits separately.
Next, we investigated if habitats hosting all-female M. batesii populations had lower within-site host-plant phenotypic diversity (i.e.possibly different, but similarly narrow niches) than habitats hosting mixed-sex M. batesii populations.Similarly to the last analysis, we ran betadisper() again, this time with the group being site (e.g.CB, BK, TB, etc.) for both the chemical and morphological data sets.We modelled the distances to the within-site centroid using a linear mixed-effect model with M. batesii population type as a fixed effect, and site as a random effect.The distance to the group centroid (i.e.habitat site) was log-transformed to meet the assumption of normality.We did this for the morphological and chemical traits separately using the lmer() function.All visualizations were made using ggplot2.

| RE SULTS
In Plants of the family Pandanaceae are notoriously difficult to classify to species level, due to several key morphological traits having evolved independently and a rich history of hybridization (Buerki et al., 2012).Consequently, we were unable to confidently determine species identity of each plant based on morphology and did not include a species-level phylogeny in our analysis.batesii despite being located adjacent to, and sometimes even in contact with, morphologically similar and presumably conspecific plants that had been heavily consumed by M. batesii.Based on the set of morphological traits measured (plant height, leaf thickness, leaf toughness, spike length, spike angle, distance between spikes, slenderness ratio), there were no clear differences in morphology between eaten and uneaten plants (See Figure A1), with PCA showing nearly complete overlap between the two groups (See Figure 3a).
The first two principal component axes could explain a combined 68% of the total variance in the model (51% and 17% for PC axes 1 and 2, respectively; see Table A1a).Statistical analyses showed a significant interaction between genus and palatability (i.e.whether the plant showed evidence of feeding by M. batesii; PERMANOVA; p = .034;Table A2a).Further investigation into this interaction showed five of the seven morphological traits that we quantified had effects of opposite sign on palatability of Pandanus versus Benstonea plants (See Figure A2; Table A3).This suggests that the plant traits that affect palatability for M. batesii differ between Pandanus versus Benstonea spp., or they exhibit varying degrees of importance depending on genus.By contrast, we found clear differences in chemical composition between host plants that showed evidence of M.
batesii feeding versus host plants that were untouched by M. batesii (PERMANOVA; p < .001;See Figure 3b and Table A2b; for loadings see Table A1b).Eaten host plants had higher levels of carbon and hydrogen and lower levels of nitrogen than uneaten plants (see Figure 4).Likewise, eaten plants had higher C/N ratios when compared to uneaten plants (Table A2c and Figure 4d).
M. batesii sites varied greatly in density and size of host plants.~50 cm apart, with heavily overlapping leaves.We found no evidence for a difference in host-plant density between sites occupied by all-female versus mixed-sex M. batesii populations (see Figure 5; linear mixed model; p = .661;Table A4).
We found some evidence that all-female populations of M. batesii used host plants that were morphologically different from the host plants used by mixed-sex populations.PCA showed a pattern of highly overlapping data points on the first two principal components (See Figure 6a; for loadings see Table A1c) but permutational modelling revealed a difference between mixed-sex and all-female populations (Table A2d; p = .015).Relative to mixed-sex populations, all-female populations were more likely to be found on host plants with larger spikes and spikes angled further away from the leaf, as well as thicker and tougher leaves (Figure A3).Nonetheless, population type explained little variation in plant morphology (R 2 = 0.04).Comparing chemical traits of plants, we found no differences between plants in habitats occupied by all-female versus mixed-sex populations (see Figure 6b and Table A2e; for loadings see Table A1d).
PCA suggested a wider range of morphological trait values of plants hosting all-female populations of M. batesii, suggesting that parthenogenetic M. batesii occupy a wider niche in terms of hostplant morphology across the range (Figure 6a).Permutational statistical testing, however, did not support this result (Table A5a).The opposite trend was suggested by PCA on chemical traits, where plants hosting mixed-sex populations of M. batesii appeared to have a wider range of trait values (see Figure 6b), but this result was also not supported statistically (Table A5b).Some sites had much withinsite variability regarding morphological and chemical traits (see Figure 7).However, linear mixed-effects modelling did not support a difference in within-site host-plant variability in either morphological or chemical traits between sites hosting all-female vs. mixed-sex M. batesii populations (Table A6, Figure 7).Thus, we found no evidence of an overall difference across all sites in niche breadth between sexual and parthenogenetic M. batesii populations.

| DISCUSS ION
Environmental factors are believed to play a large role in the emergence and maintenance of geographical parthenogenesis in natural populations (Glesener & Tilman, 1978;Suomalainen et al., 1987;Tilquin & Kokko, 2016), but the factors driving differentiation in reproductive mode among populations of facultatively parthenogenetic animals remain poorly understood.Previous studies on facultative parthenogens in the wild have explored the role of historical range expansions of parthenogenetic lineages (see Law & Crespi, 2002;Morgan-Richards et al., 2010).However, the potential role of more subtle environmental factors in generating such patterns has seldom been explored.Here, we found some evidence of morphological differences between host plants in habitats inhabited by mixed-sex versus all-female M. batesii populations.Specifically, all-female populations utilized host plants with thicker, tougher leaves and larger spikes (see Figure A3).This could be due to the sexual dimorphism of M. batesii: females are substantially larger than males and might therefore possess a greater capacity to consume host plants with long spikes and thick leaves.However, while this difference was supported statistically, M. batesii population type explained little of the variance in host-plant morphology or chemistry.
Thus, phenotypic variation among host plants does not appear to exert a substantial influence on the geographical distribution of allfemale versus mixed-sex M. batesii populations.
Host-plant density is likely to affect the geographic distribution of M. batesii as they rely on their host plants for shelter, safety, and food (Cermak & Hasenpusch, 2000).M. batesii adults of both sexes  (Boldbaatar, 2022), rafting on vegetation during a flood event, or dispersal of eggs by ocean currents/flooding (Kobayashi et al., 2014).In contrast, the chances of both a male and a female successfully arriving at a new host plant in a low-density habitat patch would be notably lower.Moreover, even if a mixed-sex population arose in a small, isolated habitat patch, stochastic male extinction would be likely to occur eventually, leaving an all-female population (Miller et al., 2024).Our data suggest a weak trend consistent with that prediction (see Figure 5), but the difference in hostplant density between all-female and mixed-sex populations was not supported statistically (Table A4).The quadrat sizes in this study may have been too small to capture the variation between these two groups.It is also possible that the distance between habitat patches is more relevant to this question than distances between host plants within patches.Given the limited mobility of M. batesii nymphs and adults (Boldbaatar, 2022), the distance between clusters of host plants, which can be tens or hundreds of meters apart, could be a significant barrier to dispersal.
We also investigated whether predictions stemming from the general-purpose genotype (GPG) or frozen-niche variation (FNV) theories could account for the geographic distribution of all-female and mixed-sex M. batesii populations.The GPG theory predicts that parthenogenetic animals will have a wider niche breadth whereas the FNV theory predicts a narrower niche breadth in parthenogenetic animals compared to their sexual counterparts.We tested niche breadth generally in plants that supported parthenogenetic versus sexual M. batesii populations (overall host-plant variability).
We also tested niche breadth within each M. batesii population site The chemical composition of host plants could also be playing a pivotal role in interactions between M. batesii and their host plants as chemistry of host plants frequently exerts a significant influence on herbivory (Fritz & Simms, 1992;Howe & Westley, 1990;Nishida, 2014;Sánchez-Sánchez & Morquecho-Contreras, 2017;Wittstock & Gershenzon, 2002).Insect herbivores typically display a preference for host plants with low C/N ratios, meaning they contain relatively high nitrogen levels (Strong et al., 1984) including in old field plant communities, arable weeds, and alpine plant communities (Karley et al., 2008;Leingärtner et al., 2014;Schädler et al., 2003).Nitrogen has been shown to be important for growth and development of insect herbivores and therefore has often been used as a proxy for nutritional value of plants (Mattson, 1980;Wang et al., 2020;White, 1984).Our data suggest a different pattern in M. batesii: we found that M. batesii in their natural habitats preferred host plants with higher C/N ratios and lower nitrogen levels (Figure 4).This divergence from the general pattern seen in other species of herbivorous insects suggests that the traits governing herbivore preferences are likely to vary across distinct  plant-herbivore interactions (see also Tanentzap et al., 2011).In M. batesii, perhaps the preference for higher C/N ratios is due to the synthesis of the alkaloid actinidine.Actinidine is an important component of M. batesii's own chemical defences; which they synthesize from chemical precursors obtained from eating Pandanus and Benstonea host-plants (Chow & Lin, 1986;Prescott et al., 2009;Vasconcelos et al., in prep.).The chemical formula for actinidine is C 10 H 13 N, suggesting that chemical synthesis of actinidine might impose greater requirements for dietary carbon and hydrogen than for nitrogen.Nitrogen is also a known component of plant defensive compounds, including a defensive compound that serves as the chemical precursor to actinidine (Fürstenberg-Hägg et al., 2013;Prescott et al., 2009).Although M. batesii may therefore exhibit a preference for plants containing this defensive chemical, an excess of such compounds in plants could potentially render them toxic to M. batesii.Another possibility is that differences in chemical composition observed between chewed and unchewed plants are not a cause but a consequence of chewing by M. batesii.Plants often increase production of defensive compounds in response to herbivory (Chen, 2008;Chen & Markham, 2021;Wu & Baldwin, 2009).
Ruling out this possibility would require experimental tests.It is important to note that 6 of the 8 uneaten plant samples in the chemical analysis are morphologically very similar, strongly indicating they belong to the same species (B.lauterbachii).This suggests that the chemical composition of this species might be the reason it is avoided in the wild, although we cannot be certain without more samples from different uneaten species.
Ultimately, although we found strong evidence that peppermint stick insects exhibit discernible preferences for a specific host-plant chemical composition, and evidence that all-female populations can exploit host plants with tougher leaves, our findings suggest that the host-plant phenotypes investigated in this study do not provide a comprehensive explanation for geographical parthenogenesis in M. batesii.These results suggest that habitat-related factors, such as host-plant morphology and chemistry, may not be the key determinants of sexual or parthenogenetic reproduction in these populations.Instead, this geographical pattern may have resulted primarily from sexual conflict (Burke & Bonduriansky, 2018, 2019, 2022;Wilner et al in prep), from stochastic processes such as random dispersal events or male extinction in some mixed-sex populations (see Miller et al., 2024), or from some combination of these factors.
Megacrania batesii provides an informative study system for exploring questions regarding geographic parthenogenesis in nature.
While certain environmental factors such as host-plant phenotype differentiation cannot explain the patterns observed in the field, there are additional factors to consider in future studies.The presence of parasites, such as bacteria, fungi, biting midges, or parasitic wasps, may offer some explanation for these patterns.Parthenogenetic M. batesii populations located in swamp habitats were found to have more ectoparasites on them compared to sexual populations in the same type of environment (Miller et al., 2024).It is possible that certain sites may host a greater number of parasites than others.In environments where parasitic pressure is high, parthenogenetic lineages with limited genetic diversity may struggle to persist over time (i.e. the Red Queen Hypothesis;Bell, 1982;Hamilton et al., 1990).
Consequently, this could impact the geographic distribution of parthenogenetic lineages across the range.Competition with other insects and predation may also contribute to this.During field observations, we noted orthopterans feeding on the same species of host plants as M. batesii, but we did not identify any predators targeting this species.A deeper understanding of the biology and ecology of this system could provide additional clues to the factors that shape sex ratio variation in natural populations.Specifically, further investigation in the differences between sexual and parthenogenetic animals could provide valuable insights into the broader population dynamics of these two coexisting reproductive modes in nature.
ecology, Entomology, Evolutionary ecology, Landscape ecology, Population ecology populations.Thus, M. batesii offers an opportunity to test predictions about the role of ecological factors in the geographic separation of sexual and parthenogenetic lineages in natural populations.
), where they are both camouflaged and shielded from predators by spikes on the edge and bottom midline of each leaf.Both Pandanus and Benstonea are monocots within the Pandanacae family, and there are at least 5 species of Pandanus and at least 2 species of Benstonea within M. batesii's range (Atlas of Living Australia, 2024).
. The known range of M. batesii in Australia spans ~250 km of Queensland coastline, with most known populations clustered within an area spanning ~35 km of contiguous coastal rainforest.Variation in climate across the range of M. batesii is modest (see Methods), and unlikely to explain variation in reproductive mode given that allfemale populations predominate at both the southern and northern edges of the species range.However, we noticed that the host plants show considerable variation in morphology and density between habitats occupied by different M. batesii populations.Some M. batesii populations are found in high-density habitats where host plants are smaller (immature) individuals that occur in contiguous patches, often with overlapping leaves.Other populations are found in low-density habitats where many of the host plants are trees located several meters apart.Denser habitats may facilitate dispersal in this species as M. batesii are flightless and nymphs and adults can only disperse by walking short distances between host plants (Boldbaatar, 2022).Low-density habitats may F I G U R E 1 Host plants used by Megacrania batesii: (a) a mature Pandanus tectorius plant with a woody trunk, (b) a Benstonea lauterbachii plant growing in a swamp (c) an immature Pandanus sp.plant, (d) Benstonea monticola plants, and (e) M. batesii female resting in the groove of a Benstonea monticola leaf.
location, with clear genetic differentiation (based on whole-genome reduced representation sequencing) separating insects deriving from differing localities.Habitat for M. batesii was identified as aggregations of host plants of the genus Pandanus or Benstonea (or both).Miller et al. (2024) reported a strongly bimodal distribution of population sex-ratios based on observed counts of females and males in the field over 4 years (2019-2022): some populations consisted of approximately equal numbers of males and females ('mixed-sex populations'), whereas other populations consisted of females only ('all-female populations').At all but one location (see below), population type remained unchanged over the 4 years of observations included in Miller et al. (2024), as well as in 2023 and 2024 plants, including 27 plants from 6 mixed-sex populations, 49 plants from 8 all-female populations, and 13 plants from the transitional population.Chemical data were collected from 43 of these plants.

F
I G U R E 2 Habitat sampling sites sampled in this study.The sampled sites are a subset of the sites containing Megacrania batesii populations studied by Miller et al. (2024) with the addition of one new population (MAD).Colour indicates whether the M. batesii population at the site is mixed-sex or all-female.The patterned colour indicates the transitional population of M. batesii.
spicuous herbivory by M. batesii.The chew-marks of M. batesii on Pandanus and Benstonea host plants are readily distinguishable from leaf damage caused by other herbivores: M. batesii feeds by making distinctive, elongated cuts along the leaf edge (see Cermak & Hasenpusch, 2000).No quadrats were taken from habitat hosting the transitional M. batesii population.Within each quadrat, we counted the number of host plants (Pandanus and Benstonea) and divided by 100 to estimate the number of plants per square meter.Most of the quadrats sampled consisted largely of either Pandanus sp. or Benstonea sp.plants.Other vegetation within the quadrats typically consisted of grasses or shrubs that are not used by M. batesii.
However, based on records in the Atlas of Living Australia, there are at least 5 species of Pandanus in the sampled area (Pandanus tectorius; Pandanus spiralis; Pandanus solms-laubachii; Pandanus gemmifer; Pandanus cookii), the most common being Pandanus tectorius.There are also at least 2 species of Benstonea (Benstonea monticola; Benstonea lauterbachii), the most common being Benstonea monticola.These two Benstonea species have clearly distinct morphologies when mature (B.lauterbachii having much longer leaves and tending to grow directly in water) but are difficult to distinguish when immature.The genera Pandanus and Benstonea are clearly distinct morphologically, and we include genus identity for all sampled plants in the analyses reported below.Pandanus plants have relatively thicker and shorter leaves, with larger spikes, and grow as trees >3 m high when mature.Benstonea plants have relatively thinner, longer leaves with smaller spikes, and can develop thin woody trunks ~1-3 m high (see Figure 1; Callmander et al., 2012).Benstonea lauterbachii (Figure 1b) was found in multiple sites where M. batesii occurred, but mature B. lauterbachii plants never showed signs of being consumed by M. batesii.Other species, especially Benstonea monticola and Pandanus tectorius, were frequently used by M. batesii as host plants.However, in these and other host species, certain individual plants were completely untouched by M.
Some sites were characterized by numerous small-and mediumsized Pandanus or Benstonea plants densely packed with overlapping leaves, while other sites had large plants (usually, mature Pandanus sp.trees) spaced further apart.Density tended to be higher in Benstonea habitats (0.53 plants/m 2 ) than in Pandanus habitats (0.26 plants/m 2 ).At one site, many Benstonea monticola plants grew

F I G U R E 3
Herbivory on Pandanus and Benstonea spp.by Megacrania batesii.PCA of host plant (a) morphological and (b) chemical traits.Each point represents an individual host-plant sample, the shape indicates whether that sample was a Pandanus sp.plant or a Benstonea sp.plant, and the colour indicates whether M. batesii chew-marks were present or absent.
are flightless and have been shown to have a very limited capacity for dispersal, so distance between host plants may act as a barrier and a factor influencing the distribution of M. batesii populations (Boldbataar, 2022).We expected all-female populations to thrive F I G U R E 4 Chemical composition of (a) carbon, (b) hydrogen, (c) nitrogen (%), and (d) carbon/nitrogen ratio in individual plants with presence/absence of Megacrania batesii herbivory.The lower and upper hinges of the box correspond to the first and third quartiles (25th and 75th percentiles).The median line is shown.The whiskers extend from the hinges to the largest and smallest value within 1.5 times the inter-quartile range.Each point represents a host-plant sample, the shape indicates whether that sample was a Pandanus sp. or a Benstonea sp.plant, and the colour indicates whether M. batesii chew-marks were present or absent.patches of host plants are more distant from each other (i.e.low-density habitats) because, in such circumstances, the successful arrival of a single female to a distant host plant could initiate the establishment of a new all-female population.Dispersal could occur via walking

(
within-site host-plant variability) and compared the all-female to the mixed-sex population sites.If each all-female site contains a different specialist genotype, overall host-plant variability could be high for parthenogenetic M. batesii, but within-site variability would be low by comparison with sexually reproducing M. batesii.We found little evidence of differentiation in either overall or within-site hostplant variability in morphological or chemical traits (see TablesA5 and A6), suggesting that niche breadth does not differ substantially between parthenogenetic and sexual M. batesii populations.It is possible that the distributions of AF versus MS populations could be explained by other environmental factors that form part of the niche, such as temperature, precipitation, or humidity.However, M. batesii inhabits a tropical environment characterized by relatively uniform climactic conditions across its limited central range (latitude difference = 0.23°), and both population types occur throughout the central part of the range of M. batesii in Australia while all-female populations occur at both the southern and northern edges of the known species range.Large-scale climate parameters are therefore unlikely to play a major role in the distribution of mixed-sex versus all-female populations (although there could be unmeasured microclimate differences between sites that might contribute to this distribution).Overall, our study does not clearly support either the general-purpose genotype theory or the frozen-niche variation theory for M. batesii.

F I G U R E 5
Host-plant density in 10 × 10 m quadrats sampled in habitats occupied by all-female or mixed-sex Megacrania batesii populations.Each point is one quadrat sampled and the colour indicates whether that quadrat hosted an all-female population or a mixed-sex population of M. batesii.The lower and upper hinges of the box correspond to the first and third quartiles (25th and 75th percentiles).The median line is shown.The whiskers extend from the hinges to the largest and smallest value within 1.5 times the inter-quartile range.PCA results for host plant (a) morphological and (b) chemical traits.Each point represents an individual hostplant sample, the shape indicates whether the sample is a Pandanus or Benstonea, and the colour indicates whether that plant was within an all-female or mixedsex Megacrania batesii population.Distance to the group (site) centroid for each sampled plant from each site for (a) morphological traits and (b) chemical traits.Each point is one host plant and the colour indicates whether that site hosts an all-female population or a mixed-sex population of Megacrania batesii.Sample sizes for each site are in grey under the site label.

batesii site M. batesii population type No. plants in morphology analyses No. plants in chemical analyses No. quadrats for density analyses
, apart from MAD.
M.Note: Sample sizes reported here include samples taken from plants with no Megacrania batesii chew marks, thus they may differ from sample sizes in analyses where we excluded uneaten plants.Center at UNSW (https:// www.analy tical.unsw.edu.au) for analyses of carbon, hydrogen, and nitrogen concentrations (i.e. percentage composition) in each sample.