Beetle cultures

Tenebrio molitor used in this study originated from an outbred stock culture of more than 100,000 individuals, maintained under standard laboratory conditions (24 ± 2 °C, 70% RH; permanent darkness) and allowed to breed randomly. Ten inbred lines (designated A to J) were created by subjecting a random subset of beetles from this population to five generations of full-sib mating (only one inbred pair per generation), followed by free mating in panmixis among the offspring within each line. Each inbred line was then maintained in two 47 L containers, each hosting several hundred individuals. Individuals from both containers were mixed every generation to maintain genetic unity while introducing environmental variation. All the experimental beetles were reared and kept in an insectary at 24 ± 2 °C, 70% relative humidity, under constant darkness, and were supplied ad libitum with bran flour, water, and apple as a supplement. Beetles used in this study were virgin adults of controlled age (10 ± 2 days post-eclosion), randomly sampled as pupae from both containers from each inbred line or the outbred stock culture. Before being used in the study, all beetles were weighed to the nearest 1 mg using an OHAUS balance (Discovery Series, DU114C).

Experimental design

Groups of 17 females per inbred line were used to characterize maternal investment in TGIP following a standard maternal immune challenge. TGIP was analysed as a function of female body mass, fecundity, and post-reproductive survival. The data allowed us to quantify genetic variation among beetle lines for these maternal traits by calculating their respective heritabilities and testing for correlations among them. Since our focus was on genetic variation in maternal investment in TGIP across the beetle lines, all females were immune-challenged to induce TGIP, as it is a prerequisite for its expression. Indeed, there is no maternal transfer of immunity if females are not immune-challenged (Zanchi et al. 2012; Dubuffet et al. 2015; Dhinaut et al. 2018b). Consequently, for the purpose of our study, control females (either injected with saline solution or unmanipulated) were not necessary.

Using inbred organisms in evolutionary research can be problematic, particularly when inbreeding depression leads to a significant reduction in fitness or fitness-related traits, or when the inbred lines do not accurately represent a random sample of genotypes from the original outbred population (Archer et al. 2012). Conversely, purging of recessive deleterious alleles during inbreeding may further bias the surviving lines toward genotypes with higher fitness, potentially altering trait distributions and reducing genetic variance. Inbreeding can also distort quantitative genetic parameters, especially the direction and magnitude of genetic correlations between traits (Rose 1984). Therefore, to illustrate the potential impact of inbreeding on trait expression, we included 17 females from the outbred population, alongside the inbred lines, and measured their investment in TGIP following a standard maternal immune challenge, as well as body mass, fecundity, and post-reproductive survival, as outlined above.

While we did not specifically aim to test for inbreeding effects, we compared traits between outbred and inbred lines (see below) to establish a baseline for comparison. The females from the outbred population were not included in further statistical analyses but were incorporated into the figures for visual comparison, helping determine whether the trait means for the inbred lines fell within the natural range observed in the outbred population. However, there may still be effects of inbreeding on quantitative genetic estimates. Therefore, as with any study involving inbred lines, caution is needed when interpreting our quantitative genetic estimates.

Maternal investment in offspring immune protection, following a standard benign bacterial immune challenge, was estimated based on the proportion of eggs that show antibacterial activity and the level of antibacterial activity in protected eggs (Zanchi et al. 2012). To conduct the immune challenge, we used the gram-positive bacterium Bacillus thuringiensis, known to be a common bacterial pathogen of coleopteran insects (Jurat-Fuentes and Jackson 2012). For each inbred line, 17 virgin adult females were weighed to the nearest mg and underwent an immune challenge by being injected with a 5 µL suspension of inactivated B. thuringiensis in sterile phosphate-buffered saline (PBS, 10 mM, pH 7.4) after being chilled on ice for 10 minutes (Zanchi et al. 2012). Inactivation of the bacteria was done by fixation in formaldehyde solution (see below for the method). Following the immune challenge, the females were paired with a virgin and immunologically naive male from the outbred stock culture. They were then allowed to produce eggs for 8 days after the immune challenge in Petri dishes supplied with wheat flour, apple, and water under standard laboratory conditions. Similarly, 17 virgin adult females from the outbred stock culture were used as a control group to compare with the inbred lines. The total number of eggs produced by each female during the 8 days following their immune challenge was recorded. However, only the eggs produced between days 2 and 8 after the maternal immune challenge were used to test their antibacterial activity, as previous studies have shown that immune-challenged females of T. molitor mainly protect their eggs within this period (Zanchi et al. 2012). Furthermore, it was found that the eggs of females immune-challenged with B. thuringiensis require 3 days of development after being laid to exhibit antibacterial activity (Dhinaut et al. 2018b). As a result, the eggs were kept under standard laboratory conditions for 3 days post-oviposition before being frozen in liquid nitrogen and stored at –80 °C, pending measurement of their antibacterial activity. On the 8th day after their immune challenge, females were isolated in grid boxes (boxes with 10 compartments; each compartment: L x W x H, 4.8 ×3.2 ×2.2 cm) without food. The females were checked once a week for survival under starvation to assess their remaining resources for survival (Moret and Schmid-Hempel 2000). Therefore, this approach allowed us to estimate female body mass, fecundity within 8 days post-challenge, investment in egg protection, and survival under starvation after reproduction.

Bacterial cultures for immune challenges

Bacterial cultures and immune challenges were performed as described by Dhinaut et al. (2018b). The bacterium B. thuringiensis (CIP53.1) used for the immune challenges was obtained from the Pasteur Institute. The bacteria were grown overnight at 28 °C in liquid Broth medium (10 g bacto-tryptone, 5 g yeast extract, and 10 g NaCl in 1000 mL of distilled water, pH 7). Afterward, the bacteria were inactivated for 30 minutes in 0.5% formaldehyde prepared in PBS, rinsed three times in PBS, and their concentration was adjusted to 108 bacteria per mL using a Neubauer improved cell counting chamber under a phase-contrast microscope (magnification x 400). The success of the inactivation was verified by plating a sample of the bacterial solution on sterile Broth medium with 1% bacterial agar and incubating it at 28 °C for 24 hours. Aliquots of the bacterial suspension were kept at −20 °C until use. Immune challenges of the beetles were performed by injection through the pleural membrane between the second and third abdominal tergites using sterile glass capillaries that had been pulled out to a fine point with an electrode puller (Narashige PC-10).

Egg antibacterial activity

The antimicrobial activity of each collected egg was measured using a standard zone-of-inhibition assay (Dhinaut et al. 2018b). Individual egg samples were thawed on ice, and egg extracts were prepared by homogenizing each egg into an acetic acid solution (0.05%, 5 μL per egg). After centrifugation (3500 g, 2 min, 4 °C), 2 µL of the supernatant was applied to zone-of-inhibition plates seeded with Arthrobacter globiformis (CIP105365), obtained from the Pasteur Institute. A. globiformis is commonly used as the target bacterium in antibacterial assays due to its high sensitivity to insect-derived antibacterial compounds, which enhances assay resolution (Dubuffet et al. 2015). This sensitivity is crucial for detecting antibacterial activity at the level of individual eggs. Zone inhibition plates were prepared from an overnight culture of A. globiformis, which was added to Broth medium containing 1% agar to reach a final concentration of 105 cells per mL. Six millilitres of this seeded medium were poured into Petri dishes and allowed to solidify. Sample wells were made using a Pasteur pipette fitted with a ball pump. Two microliters of the sample solution were added to individual wells. A positive control (Tetracycline: Sigma-Aldrich, St. Louis, MO, USA, T3383; 2.5 mg/mL in absolute ethanol) was included on each plate (Dhinaut et al. 2018b). Plates were incubated overnight at 28 °C, after which the diameter of each inhibition zone was measured. Due to slight variations in inhibition zones in tetracycline controls, sample inhibition zones were normalized. This was achieved by adjusting each sample’s zone diameter based on the ratio between the tetracycline control on a reference plate (the plate showing the largest inhibition zone) and the tetracycline control on the plate where the sample was measured (Dhinaut et al. 2018a, 2018b).

From this assay, we estimated female investment in egg protection using two complementary measures. First, the proportion of protected eggs per female was calculated as the number of eggs exhibiting antibacterial activity (i.e., showing a detectable zone of inhibition) divided by the total number of eggs produced. Second, among the subset of eggs that showed antibacterial activity, the mean diameter of the inhibition zone was used to estimate the amount of antimicrobial compounds transferred to the eggs. These two measures respectively captured (1) the proportion of offspring receiving protection and (2) the level of protection provided to those offspring.

Statistics

We compared maternal trait measures between outbred and inbred females with linear mixed-effects models (LMMs) or generalized linear mixed-effects models (GLMMs) implemented in the lme4 package (Bates et al. 2015) in R (R Core Team 2015), including a binary fixed effect for inbreeding status (inbred vs. outbred) and controlling for body mass. All inbred lines were pooled into a single category (‘inbred’), while the outbred population was coded as ‘outbred’. Body mass was included as a covariate, as it is a proxy for female quality in T. molitor (Jehan et al. 2020, 2021, 2022; Crosland et al. 2022; Zanchi et al. 2022), and line identity was included as a random factor. The proportion of protected eggs laid by each female was analyzed with a binomial GLMM, taking into account the number of protected eggs and the number of eggs tested per female. Fecundity (total egg number) and survival time under starvation were analyzed using Poisson GLMMs, whereas levels of antibacterial activity in protected eggs were analyzed with an LMM. Finally, we tested the effect of inbreeding on female body mass itself with an LMM that included line identity as a random effect.

Phenotypic relationships between the two measures of maternal investment in egg protection (the proportion of protected eggs among those tested, and the average antibacterial activity of protected eggs from each female) and female body mass, fecundity, and starvation resistance were tested using LMMs or GLMMS as above. Female body mass, fecundity, and starvation resistance were included as covariates, with the inbred line treated as a random factor. The outbred population was not included in this analysis.

To assess the potential heritability of maternal investment in egg protection (measured both as the proportion of protected eggs among those tested and as the average antibacterial activity of protected eggs from each female), body mass, fecundity, and post-reproductive survival, we calculated the repeatability of each trait across lines using the rptR package in R (Nakagawa and Schielzeth 2010; Stoffel et al. 2017). Each of the 10 inbred lines (the original outbred population from which the lines were derived was not included in these analyses) consisted of 17 individuals, and all individuals were measured once for the traits listed above. In this context, repeatability reflects the proportion of phenotypic variance explained by differences among lines and sets an upper bound on broad-sense heritability (H²), which in turn represents an upper bound on narrow-sense heritability (h²), particularly if additive genetic effects predominate. This approach assumes that lines are genetically homogeneous and that environmental variation is randomly distributed across them. Repeatability was estimated using linear mixed-effects models with “Line” as a random effect and 10,000 bootstrap iterations to compute 95% confidence intervals (CIs). Additionally, we used a likelihood ratio test (LRT), as implemented in the rptR package (REML-based), to assess whether repeatability estimates for each trait were significantly greater than zero. P-values from the LRT were adjusted for multiple testing using the Benjamini-Hochberg false discovery rate (FDR) procedure (Benjamini and Hochberg 1995). However, as a conservative criterion, we considered repeatability to be significant only when the 95% CI did not include zero.

Genetic correlations between maternal traits were estimated by accounting for within-line variation for each pair of traits. To this end, we employed a repeated sampling procedure: one female was randomly sampled from each line, and a Pearson correlation coefficient (along with its standard error) was calculated for all pairs of traits. For each correlation estimate, the jackknife resampling method was applied following the approach of Roff and Preziosi (1994). Specifically, a series of N (in this case 10) pseudo values was computed by sequentially excluding each line one by one and re-estimating the correlation using the formula:

$${S}_{N,{i}}=\,{N}_{{rN}}-\,{\left(N-1\right)}_{{rN}-1,{i}}$$

where SN,i is the ith pseudo value, rN is the correlation coefficient estimated using one randomly sampled female per inbred line across all N lines, and rN−1,i is the correlation coefficient calculated after excluding the ith inbred line (Archer et al. 2012; Letendre et al. 2021). The jackknife estimate of the genetic correlation (rj) is the mean of the pseudo values, and an estimate of the standard error (SE) is given by:

$${SE}=\frac{{\sum }_{i=1}^{i=N}{\left({S}_{N,i}-{r}_{j}\right)}^{2}}{N\left(N-1\right)}$$

The entire procedure, comprising random sampling of one female per line, correlation estimation, jackknife resampling, and computation of rj and SE, was repeated 10,000 times. From these iterations, we calculated the mean genetic correlation coefficient, its average standard error, and the proportion of iterations in which the correlation was statistically significant.

According to simulation models, the jackknife method yields more precise genetic estimates than traditional inbred line means when fewer than 20 inbred lines are used (Roff and Preziosi 1994). These genetic (co)variance estimates from inbred lines include dominance or epistasis variance and are thus considered broad-sense estimates (Falconer and Mackay 1996). By rearing beetles individually, we reduced the variance between lines caused by shared environmental factors and interactions (Archer et al. 2012; Letendre et al. 2021). Average genetic correlations were considered statistically significant if their ratios to average standard errors exceeded 1.96, rejecting the null hypothesis of no correlation based on a two-tailed t-distribution with infinite degrees of freedom.