The operational sex ratio (OSR), i.e., the local ratio of fertilizable females to sexually active males at any given time, is of key importance for the strength of sexual selection and the reproduction of populations. We hypothesize that sex-specific cohort splitting, i.e., when one sex mostly metamorphoses while the other mostly enters diapause, may lead to OSR bias in nature. The OSR of an aquatic moth, Acentria ephemerella, has been shown to be strongly male-biased in situ. Here, we use a mesocosm experiment in which we determine the sexes of active, diapausing, and metamorphosing larvae to test whether the male bias in Acentria is due to sex-specific mortality or sex-specific cohort-splitting. Fish predation dd not result in a strong male bias of the whole population but increased male bias in pupae and female bias in diapausing larvae. The opposite effect of fish on pupal versus diapausing larval sex ratios suggests that fish-induced sex-specific cohort splitting, rather than sex-specific mortality, caused the OSR bias of Acentria observed in situ. Future research needs to study whether the OSR bias is an adaptive response to the presumably higher fish predation pressure on females or a maladaptive byproduct of sex-specific activity and growth responses to fish presence. Overall, shifts in OSR due to sex-specific cohort splitting could be a more common component of arthropod life-histories than previously thought.

The aquatic moth Acentria ephemerella (Crambidae) displays a pronounced sexual wing and size dimorphism (Berg, 1942). Female pupae and adult life stages are larger than male pupae and adult life stages (Berg, 1942; Miler et al., 2014). The adult females are predominantly rudimentarily winged (brachyptery > 99%), aquatic throughout their life and display morphological adaptations to the aquatic habitat such as swimming legs (Berg, 1942). Adult males in contrast are winged and emerge from the water column after metamorphosis (Berg, 1942). Acentria is a capital breeder with a short adult life span of 1–3 days and the number of eggs can be counted in advanced pupal stages (Berg, 1942; Miler et al., 2014). Mating occurs at the water surface, with the abdominal apex of the aquatic adult females reaching above the water surface (Berg, 1942). During mating, the terrestrial males transfer a spermatophore into the bursa copulatrix of the females which is located ventrally between the 7th and 8th abdominal segment (Berg, 1942). The larvae of Acentria feed from June to September on submerged aquatic macrophytes, mainly pondweed species Potamogeton spp. (Gross et al., 2002; Miler, 2009). A strong decrease in the biomass of its food plants occurs under high Acentria population densities during summer (Le Bagousse-Pinguet et al., 2012; Miler & Straile, 2010). The life-cycle of Acentria consists of one to three generations per year, of which one generation includes a diapause stage (Berg, 1942; Gross et al., 2002; Miler, 2009). Although information about the number of instars until metamorphosis of *Acentria *is scarce, previous research suggests five instars to be likely (Bänziger, 2000; Haenni, 1980), which fits to a typical Lepidoptera life cycle with five to seven instar stages (Williams & Feltmate, 1992). Second and third instar Acentria larvae of the last generation of the year enter a diapause stage in autumn, overwinter inside plant stems in a protective cocoon, the hibernaculum, and continue larval development until metamorphosis and emergence in the spring of the next year (Berg, 1942; Haenni, 1980; Gross et al., 2002; Miler et al., 2014).

The experiment was conducted in a large outdoor tank (area = 52.5 m², depth = 1.5 m) filled with fine sediment (~ 0.35 m) and water (~ 1 m) from Lake Constance (Fig. 1). 15 mesocosms (experimental units) were placed in this tank and consisted of a transparent plastic tube (volume ~ 0.48 m³, surface ~ 0.48 m², height ~ 1 m, length ~ 0.8 m, width ~ 0.6 m) attached to a metal frame in the sediment and to a floating frame used as a buoy to keep the plastic tube in a vertical position in the water column. Plastic tubes completely separated the experimental units from the surrounding water and there was no water exchange between the individual mesocosms and the tank. Frames were covered with metal gauze (mesh size ~ 1620 µm) approximately 0.1 m above the water surface, to allow for Acentria reproduction, but to prevent winged males from dispersing out of the replicates (for details see Miler et al., 2008). Both the plastic tubes and the gauze-covered frames (1) prevented any Acentria adults or larvae to move out of each of the mesocosms into the tank and vice versa and (2) prevented the exchange of any terrestrial organisms larger than 1620 µm in length or width and the exchange of any aquatic organisms between each of the mesocosms and the tank.

Five mesocosms were sampled at the time of fish introductions (10 August, i.e. start conditions), and ten experimental units were assigned to the fish and control (no fish) treatment, respectively and sampled on 20 August, i.e., altogether 15 mesocosms. The assignment of mesocosms to treatment and time of sampling was randomized using a random number generator.

Potamogeton perfoliatus shoots were sampled on 14 June 2006 in an Upper Lake Constance macrophyte patch. Macrophyte-associated* Acentria* larvae and pupae were removed and subsequently eight macrophyte shoots were planted in each mesocosm, i.e., a density of ~ 18 shoots m^-2^ typical for the early seasonal growing period (Wolfer & Straile, 2004). The shoots were allowed to root and establish new shoots for approximately five and a half weeks. From 20 to 22 July, a total of 79 Acentria pupae from field samples in Lower Lake Constance were introduced into each mesocosm (i.e., 1580* Acentria* pupae in total for all 20 mesocosms taken together). Since Acentria larvae breathe through a thin membrane and removing this membrane would be necessary to identify the sex of the larvae (Berg, 1942), we did not know the sex of the pupae introduced into the mesocosms. Based on a field study from 2005 in Lower Lake Constance, we can assume the sex of the pupae introduced into the mesocosms was slightly male-biased (~ 54% males, Miler et al., 2014), whereas Upper Lake Constance sites showed a considerably higher proportion of male pupae (Miler et al., 2014). Starting on 20 July, from these pupae adult individuals hatched, mated and laid egg clutches and on 31 July, feeding damage and small larvae could be observed in all mesocosms. Twelve sticklebacks (Gasterosteus aculeatus, total length range: 4.3–6.2 cm) were introduced into each mesocosm of the fish treatment on 10 August, i.e., after allowing Acentria to develop for ten days. In a previous mesocosm experiment at Lake Constance, a strong and significant predation of sticklebacks on Acentria was observed, with Acentria specimens found in the guts of sticklebacks (Miler et al., 2008). Sticklebacks were caught in a small pond near the University of Konstanz (geographic coordinates in decimal degrees: latitude 47.687, longitude 9.190), where they occurred in high abundances and were easy to capture with minnow traps. To our knowledge, the density of sticklebacks in Lake Constance has not been studied, but the fish density of 24 individuals m^-2^ used in our experiment was comparable to stickleback densities observed in nature (Thiel et al., 1995; Ward & FitzGerald, 1983).

Sampling occurred in 5 mesocoms on 10 August to assess* Acentria* densities prior to the onset of fish predation as well as Acentria densities, sizes and life stage distributions on 20 August, after ten days of fish predation and when the first adult individuals were observed. We did not observe any direct predation of Acentria by sticklebacks, but any such events would have been unlikely to be observed, since the mesocosms were only occasionally and briefly checked between sampling events. It is likely that active Acentria larvae are more vulnerable to fish predation than diapausing larvae, since the latter are better protected inside a hibernaculum and the plant stem (Berg, 1942; Gross et al., 2002; Haenni, 1980), although active Acentria often construct simple protective covers from pieces of leaves (Berg, 1942). We distinguished three developmental life stages of Acentria: active larvae (larvae outside macrophyte stems), diapausing larvae (larvae inside macrophyte stems) and pupae. Acentria larvae were washed through a sieve (mesh size 200 µm) and fixed in Carnoy’s solution (ethanol, chloroform, acetic acid; 6:3:1) for subsequent sex determination. The remaining material washed from macrophytes, including Acentria larvae overlooked during the fixation in Carnoy’s solution, was preserved in 70% ethanol in 100 ml plastic bottles. Since Acentria pupae are closely attached to the stems of P. perfoliatus shoots and larvae overwinter inside the stems, the remaining plant material was stored at 5°C in plastic bags and searched through for pupae and diapausing larvae within one week after sampling. Pupae were fixed in 70% ethanol and diapausing larvae in Carnoy’s solution. Macrophytes were dried at 90°C for three days and densities/span>of pupae and larvae were calculated as individuals per gram plant dry mass (Ind. g^-1^ dm).

Females of most lepidopteran species of the clades Ditrysia and Tischeriina (comprising together more than 98% of all species) contain a W sex chromosome forming a densely concentrated spot (the ‘W-body’) inside each somatic interphase nucleus, whereas males lack this feature (Fig. 2a, for reviews see Traut & Marec, 1996; Traut et al., 2007; Traut & Scholz, 1978). From each mesocosm up to 40 active and up to 40 diapausing larvae (Appendix Table 1) were dissected and their tissue, preferentially Malpighian tubules and silk glands, was stained with 1.5% lactic acetic orcein (Traut et al., 1986). After staining, larvae were inspected for the presence of sex chromatin (W chromatin) in highly polyploid nuclei (Fig. 2a) under a light microscope (Zeiss Axioskop 40) at 1000-fold magnification. In one of the five fish treatment mesocosms, the active larvae (n=8) could not be successfully sexed because these larvae were erroneously not preserved in Carnoy’s solution. In addition to active and diapausing larvae from the mesocosms, we analysed the primary sex ratio of 14 Acentria egg clutches that were sampled in Lake Constance and incubated at 20°C in the laboratory. From each clutch, the sex of 30 randomly selected freshly hatched larvae was determined within 24 h after hatching. Acentria pupae were sexed morphologically by measuring and comparing the length of wings and antennae as published in Berg (1942) (Fig. 2b). To convert length measurements into dry mass we established a head-capsule width – dry mass relationship for larvae (equation (1), n = 45, R² = 0.93, P < 0.0001) and a length – dry mass relationship for pupae (equation (2), n = 34, R2 = 0.93, P < 0.0001).

log₁₀ (weight [mg]) = -12.17 + (4.17 * log₁₀ (head capsule width [µm]))   (1)

log₁₀ (weight [mg]) = -1.96 + (2.93 * log₁₀ (body length [mm]))    (2)

In the following analyses, we will abbreviate the weight of larvae and pupae in log₁₀ (weight [mg]) as log dm. The developmental stages of pupae were assessed based on their eye development and eggs from specimens in an advanced developmental stage were counted after dissection.

All statistical analyses were performed with R 3.2.3 (R Development Core Team, 2015) and RStudio 0.99.489 (RStudio Inc., 2015). We used Mixed model ANOVA to study a) the effect of fish presence and developmental stage on Acentria densities, and b) the effect of fish presence and sex on sizes of active larvae, diapausing larvae and pupae. The analysis of pupal sizes revealed a singularity problem for estimating the random effects. In order to confirm the results for the fixed effects, we additionally ran a simple linear model (lm() function) using mean sizes for each mesocosm). This only marginally altered the F and p-values. Hence, we report in the case of pupal sizes also the statistical results from the mixed model ANOVA. The effects of fish presence on the % densities of active larvae, diapausing larvae, and pupae was analysed using a multinominal generalized mixed model using a Bayesian approach. Finally, the effect of fish presence and developmental stage on the sex ratio was analysed with mixed logistic regression. Type II ANOVA (F or Wald Chi-square) tests were used to test for significance of model coefficients (function Anova() in R). In all mixed models, the mesocosm ID was included as a random factor to avoid pseudoreplication (R packages lme4, functions: lmer(), glmer(), Bates et al., 2015, nlme, function: lme(), Pinheiro et al., 2024, brms, function: brm(), Bürkner, 2017 and stats (base R), function: glm(), R Development Core Team, 2015). Diagnostic plots using the R package DHARMa (Hartig, 2022) suggest no violation of the assumptions for the statistical models used in this study (dispersion, homogeneity of variances). The sex ratio bias of freshly hatched larvae was analysed with a Χ²² test (R package stats (base R), function: chisq.test(), R Development Core Team, 2015).