Inbreeding alters context-dependent reproductive effort and immunity in male crickets

2.1 Study animals

Inbred male G. sigillatus used in this study were selected from genetically distinct inbred lines that exhibit significant genetic variation in key life-history traits, including lifespan (Archer et al., 2012), body mass (Gershman et al., 2010b), male calling effort (Archer et al., 2012), female egg production (Archer et al., 2012), immune function (Gershman et al., 2010a), spermatophore mass (Gershman et al., 2010b), nuptial gift composition (Gershman et al., 2013) and cuticular hydrocarbon profiles (Weddle, Mitchell, Bay, Sakaluk, & Hunt, 2012). In this study, four lines were used from nine previously established inbred lines. These inbred lines were created by subjecting crickets, randomly selected from descendants of approximately 500 individuals collected at Las Cruces, New Mexico in 2001, to 23 generations of full-sib mating followed by multiple generations of panmixia within lines thereafter (Ivy, Weddle, & Sakaluk, 2005). Measurements made in 2007 revealed evidence of significant inbreeding depression in the inbred lines after 17 generations of inbreeding: inbred lines all showed lower hatching success, decreased offspring production and longer developmental times compared with the outbred population (Sakaluk, S. K., Oldzej, J., Hodges, C., Harper, J., Rines, I., Hampton, K. J., Duffield, K. R., Hunt, J., & Sadd, B. M., unpublished data). For this study, the four lines (A, F, G and I) selected were chosen because they were at the same stage of development at the time that the experiment was initiated. Outbred experimental males were selected from a large outbred colony, maintained at a population size of approximately 5,000 to ensure genetic heterogeneity and initiated with crickets collected from a wild population from Riverside, CA in 2014.

Newly hatched, experimental crickets were reared in 55-L plastic storage bins filled with egg carton to increase rearing surface area and provisioned with food (Envigo© 2018CM Teklad Certified Global 18% protein rodent diet meal) and water (glass vials plugged with cotton) ad libitum. When sex differences became apparent (4th or 5th instar), juvenile males were removed from stock colonies and housed individually in small (450 ml) plastic containers and provisioned with dry food pellets of the same Envigo© formulation and water (glass vials plugged with cotton) ad libitum. A small section of egg carton was also provided as a refuge. All individuals were housed in an environmental chamber at 32°C on a 16 hr:8 hr light:dark cycle. Males were checked twice weekly to determine the date of final moult to adulthood. Calling effort of males was used as a proxy for reproductive effort. Adult male G. sigillatus begin mating 4.5 days after eclosion on average (Burpee & Sakaluk, 1993) and, because calling is essential for mating, it follows that males are also calling from this time. Males were given an infection cue treatment 3 weeks (±2 days) following their adult eclosion, based on a previous study demonstrating that males at this age are more likely to show terminal investment in calling (Duffield et al., 2018). Body mass for each male was measured immediately prior to infection cue administration using an analytical balance (Mettler Toledo AG245).

2.2 Infection cue

A pathogenic infection may signal reduced residual reproductive value to the host, and this may be mediated through the infection-associated immune challenge. The preferred approach to investigate shifts in host life-history strategies as a consequence of infection is to simulate an infection through the use of nonpathogenic immune elicitors, as this technique eliminates the confounding effects of parasite proliferation, pathogenicity or parasite manipulation (Duffield et al., 2017). Earlier, we demonstrated that injection with heat-killed Escherichia coli stimulates an immune response and invokes a terminal investment response in male G. sigillatus (Duffield et al., 2015, 2018). This approach ensures the response is related to the host's strategy and not the result of alternative causes related to a live infection; thus, a similar protocol was implemented here. Males were randomly assigned to one of five treatments on an increasing infection cue spectrum: (a) naïve (unmanipulated control), (b) sham control (injection of 2 μl ringer saline), (c) low-dose infection cue (injection of 2 μl ringer saline with 5 × 10⁵/ml heat-killed E. coli), (d) moderate-dose infection cue (injection of 2 μl ringer saline with 5 × 10⁷/ml heat-killed E. coli) or (e) high-dose infection cue (injection of 2 μl ringer saline with 5 × 10⁸/ml heat-killed E. coli). We consider the sham control treatment not only a control for the effect of injection, but also as a low-level mortality threat because injection causes cuticle damage and induces an immune response that potentially signals a mortality threat to the individual (Ardia, Gantz, Schneider, & Strebel, 2012; Gillespie & Khachatourians, 1992; Wigby, Domanitskaya, Choffat, Kubli, & Chapman, 2008). Injections were performed using a 5-μl syringe with a 1 mm compression fitting (Hamilton® brand) within which a needle formed from a heat-pulled glass capillary tube was inserted. Crickets were injected between the 6th and 7th pleurite of the thorax. Pulled capillaries were cleaned in 70% ethanol, rinsed with ultrapure water and dried between each injection. Capillaries were not reused across treatments nor days. Treatments were always applied at the same time (09:00 hr ± 1 hr) throughout the experiment.

Escherichia coli (ATCC strain 23716) used to create our low, moderate and high infection cues were cultured at 30°C in 7 ml of liquid medium (10 g bacto-tryptone, 5 g yeast extract, 10 g NaCl in 1,000 ml of distilled water, pH 7). To prepare bacterial suspensions for challenge injections, 1 ml of an overnight culture was centrifuged (850 g, 4°C, 10 min), and the supernatant was discarded and replaced with sterile ringer saline. This procedure was repeated three times. The bacteria were then heat-killed (90°C, 5 min), and the concentration of bacterial cells was adjusted to the concentrations described earlier for each infection cue dose. Efficiency of the heat killing was confirmed by plating out samples of the suspension on media agar.

2.3 Assessing reproductive effort

We quantified calling effort (i.e., the amount of time males spent calling) over two consecutive nights following infection cue treatment. Calling effort was measured using a custom-built high-throughput sound monitoring array (Bertram & Johnson, 1998; Duffield et al., 2018) in which each male-containing individual container (250 ml) was fitted with a lid-mounted microphone (C1163, Dick Smith Electronics) and placed within a small Styrofoam box to prevent crosstalk between containers. Following administration of infection cues, males were isolated and given 7 hr (±1 hr) to acclimate prior to the start of recording trials and provisioned with water and a small piece of egg carton for refuge. Recording periods started at 17:00 hr and ended at 09:00 hr and ran again for 24 hr starting at 10:30 hr the day following treatment; these periods capture calling effort at a time most relevant to female mate attraction (Burpee & Sakaluk, 1993; Sakaluk, 1987). The sound monitor sampled each microphone throughout the recording periods every 2 s, and based on the binary output resulting from this protocol, total calling time was calculated for each male across recording periods (Duffield et al., 2018; Hunt et al., 2004).

2.4 Survival and post-mortem measurements

Following the end of the two-night calling trials, males were returned to their individual rearing boxes where they were provided water ad libitum. To make monitoring survival tractable, individuals were food-deprived during this time, as laboratory-reared G. sigillatus are known to live for 2–3 months in the laboratory when fed ad libitum (Burpee & Sakaluk, 1993; Ivy & Sakaluk, 2005; Sakaluk, 1987), far longer than their longevity under more natural field conditions (Sakaluk, Schaus, Eggert, Snedden, & Brady, 2002). Mortality was monitored and recorded daily. Upon their death, the pronotum width of experimental subjects was measured as a proxy for structural body size, using a stereomicroscope (Nikon SMZ800) equipped with a digital camera and imaging software (Nikon NIS-Elements Documentation v. 4.20).

2.5 Statistical analyses for reproductive effort and survival

The distribution of male calls indicated that calling effort measurements had an excess of zeroes and were over-dispersed. Therefore, the calling effort data were analysed in the same way as previously done in this system (Duffield et al., 2018). We analysed calling effort data with Bayesian methods using the R package MCMCglmm (Hadfield, 2010), fitting a zero-altered Poisson (ZAP) model in R (version 3.4.2, R Core Team, 2016). The ZAP model includes a logistic regression for the zeroes in the data and an over-dispersed Poisson regression for the zero-truncated counts. This type of model enabled us to answer two distinct questions within a single statistical structure: (a) what factors affect whether a male chooses to call and (b) if a male calls, what factors affect the amount of calling (Duffield et al., 2018; Houslay, Houslay, Rapkin, Hunt, & Bussière, 2017; Houslay, Hunt, Tinsley, & Bussière, 2015)?

We first wanted to know whether inbreeding influences an individual's propensity to terminally invest. For this “Inbreeding model,” we included the full data set and our predictors were treatment (naïve, sham, low, moderate and high infection cue doses), inbreeding status (inbred versus outbred) and the interaction between infection cue and inbreeding status. We then wanted to know whether genotype itself influenced an individual's terminal investment threshold. For this “Genotype model,” we included only inbred individuals and our predictors were treatment (naive, sham, low, moderate and high infection cue doses), inbred line (lines A, F, G and I) and the interaction between infection cue and inbred line. For both models, we used binary indicator variables (0/1) to specify whether or not an individual belonged to each treatment group (Gelman & Hill, 2007), using “naïve/inbred” and “naïve/line A” as our reference levels for the Inbred model and Genotype model, respectively. We ran each model for a total of 850,000 iterations, with a “burn-in” of 50,000 and retaining every 400th iteration thereafter, yielding an effective sample size of around 2,000 for each coefficient in the model. Fixed effects are considered statistically significant if the 95% credible intervals do not include 0. We checked for a lack of autocorrelation through plots of the thinned chains and ran each model three times to ensure convergence to a similar posterior distribution (tested using the Gelman-Rubin diagnostic; Gelman, Rubin, Gelman, & Rubin, 1992). Our main model used an uninformative prior, and we checked that results were robust to different prior specifications.

For both Inbreeding and Genotype models, male survival was analysed using a Cox proportional hazards model (SAS v. 9.4), with infection cue and inbreeding status (for Inbreeding model) or inbred line (for Genotype model), and the interaction between the two included as main effects and body condition (residuals derived from a regression of body mass on pronotum width) included as a covariate. Reported results derive from the best models as determined by corrected Akaike's information criterion (AICc; Hurvich & Tsai, 1989; Sugiura, 1978).

2.6 Quantifying immune function

We quantified immune function in a separate group of male crickets originating from the same inbred and outbred lines and subject to the same treatments described in the calling effort portion of the study; this was done so that we could quantify both immunity and calling effort at similar time points post-infection. We measured multiple immune parameters in an attempt to capture the suite of immune pathways that may differ between lines or be altered following infection cue treatment. Here, we analysed antibacterial activity, total circulating haemocytes and haemocyte microaggregations from haemolymph that had been collected 4 hr after treatment administration. General cell-free antimicrobial activity of haemolymph is a component of the humoral immune response of insects and includes the action of both lysozyme-like enzymes and antimicrobial peptides (Lemaitre, Reichhart, & Hoffmann, 1997). Additionally, haemocytes are a component of the cellular response of insect immunity and are involved in core processes that include coagulation, nodulation, phagocytosis and encapsulation (Lavine & Strand, 2002). Finally, the microaggregation of haemocytes is an early step in the process of nodule formation (Miller, Howard, Rana, Tunaz, & Stanley, 1999; Miller & Stanley, 2004), the predominant insect cellular defence reaction to bacterial challenges and responsible for clearing a large proportion of bacteria from circulation (Howard, Miller, & Stanley, 1998).

To collect haemolymph, males were cold-anesthetized, the membrane was pierced under the dorsal pronotum plate with a sterile 25-G needle, and 5 μl of outflowing haemolymph was taken with a prechilled glass microcapillary tube positioned at the puncture site. Collected haemolymph was then expelled into 11 μl of Grace's insect medium (MilliporeSigma) to be used in antibacterial assays; 5 μl of this mixture was then added to 15 μl of Grace's insect medium which was immediately used for circulating haemocyte and microaggregation counts. The samples for antibacterial assays were snap-frozen in liquid nitrogen and stored at −80°C for later analysis.

2.7 Zone of inhibition assay

Although immune-challenged individuals in this study were injected with E. coli, previous assays resulted in no measurable antibacterial activity on plates seeded with E. coli. Therefore, antimicrobial activity was assayed from zones of inhibition induced by samples on petri dishes containing agar seeded with Micrococcus luteus (ATCC 4698; see Sadd & Schmid-Hempel, 2007 for methodological details). Briefly, M. luteus from a single colony on a streak plate were incubated overnight at 30°C in 7 ml of media (2.5 g peptone and 1.5 g meat extract in 500 ml of nanopure water, pH 7.0). From this culture, bacteria were added to liquid media containing 1% agar held at 40°C to achieve a final density of 1.5 × 10⁵ cells/ml. Six millilitres of seeded medium was poured into a 100-mm diameter petri dish to solidify. Sample wells were made using a Pasteur pipette (Volac D810) fitted with a ball pump, 2 μl of sample solution thawed on ice was added to each well and negative (PBS) control wells were included on each plate. Plates were inverted, incubated for 48 hr at 30°C and then the diameter of inhibition zones was measured for each sample. Two diameter measurements of zones, perpendicular to one another, were measured for each sample using ImageJ (Schneider, Rasband, & Eliceiri, 2012) and averaged (measurements were performed blind to treatment). Because zone of inhibition diameter does not increase linearly with antibacterial activity, measured zone diameters were converted using a standard curve, to units (mg/ml) of lysozyme (from hen egg white, MilliporeSigma, CAS: 12650-88-3). Each haemolymph sample was tested in duplicate, with the mean of the duplicates being used in subsequent analyses.

2.8 Circulating haemocyte and microaggregation counts

Haemocytes and microaggregations were counted at 400× magnification under a phase-contrast microscope with a haemocytometer (Fast-Read 102® plastic counting chamber) to assess their numbers per individual as a proxy for cellular immunity (Duffield et al., 2018; King & Hillyer, 2013; Stoepler, Castillo, Lill, & Eleftherianos, 2013). Counting was performed blind to treatment.

2.9 Statistical analysis of immune measures

As with our analysis of reproductive effort, we compared inbreeding status (for Inbreeding model) and inbred lines (for Genotype model) for all immune measures. Infection cue and the interaction with inbreeding status or inbred line were also included as main effects, and body condition (residuals derived from a regression of body mass on pronotum width) included as a covariate for all immune measures. Total circulating haemocyte and zone of inhibition values were log-transformed to meet the assumptions of normality (reported means and confidence intervals were back-transformed) and analysed with general linear models. For microaggregation counts, we implemented a generalized linear model with a negative binomial error distribution. Reported results derive from the best models as determined by corrected Akaike's information criterion (AICc).

3.1 Reproductive effort

The calling effort of 609 males was measured over two consecutive nights following infection cue treatment (sample sizes included in Table S1). We found a significant interaction between the effects of inbreeding and infection cue on the time spent calling of males (Figure 1a). Inbred and outbred males diverged in how they responded to a low dose of heat-killed E. coli: outbred males decreased calling effort and inbred males increased calling effort (although nonsignificantly) relative to the naïve reference group (Figure 2), and there was a significant increase in the likelihood that a male would call at the moderate-dose treatment level (Figure 1a). Within inbred lines, we found no effect of line identity or any of the other predictors on either the time spent calling by males or the likelihood of calling (Figure 1b). Parameter estimate, confidence interval and p values are reported in Table S2.

3.2 Survival

The nonsignificant interaction term of inbreeding status by infection cue (Wald  = 1.90, p = 0.7544) was removed from the final Inbreeding model for survival, and there was also no significant effect of inbreeding status (Wald  = 0.48, p = 0.4903) or infection cue (Wald  = 0.88, p = 0.9271). There was, however, a significant effect of body condition (Wald  = 85.09, p < 0.0001), with males in better condition living longer than those in poorer condition (hazard ratio [lower, upper bounds of 95% confidence intervals] = 0.983 [0.980, 0.987]).

For the final Genotype model, we removed the nonsignificant interaction between inbred line and infection cue (Wald  = 13.85, p = 0.3106). Infection cue did not influence survival (Wald  = 0.83, p = 0.9348), but survival varied significantly among inbred lines (Wald  = 44.34, p < 0.0001), with lines F (hazard ratio [lower, upper bounds of 95% confidence intervals] = 0.539 [0.414, 0.703]) and G (hazard ratio [lower, upper bounds of 95% confidence intervals] = 0.438 [0.331, 0.579]) having increased survival relative to the reference line I and line A (hazard ratio [lower, upper bounds of 95% confidence intervals] = 0.904 [0.658, 1.234]). Body condition (Wald  = 73.17, p < 0.0001) significantly affected survival, with males in better condition living longer than males in poorer condition (hazard ratio [lower, upper bounds of 95% confidence intervals] = 0.984 [0.980, 0.988]).

3.3 Immune function

Immune function, based on antibacterial activity, circulating haemocytes and haemocyte microaggregations, was measured in 265 males 4 hr post-infection cue treatment (sample sizes included in Table S1).

3.3.1 Antibacterial activity of haemolymph

The interaction between inbreeding status and infection cue was nonsignificant (F_4,179 = 1.02, p = 0.3984) and was not included in the final Inbreeding model. We found a significant effect of inbreeding status on haemolymph antibacterial activity (F_1,183 = 4.77, p = 0.0303), with inbred males having greater antibacterial activity compared with outbred males (Figure 3a). There was also a significant effect of body condition (F_1,183 = 5.63, p = 0.0186), with males in better condition having increased activity. Infection cue did not influence antibacterial activity (F_4,183 = 0.19, p = 0.9457).

In the Genotype model, a nonsignificant line by infection cue interaction term (F_12,140 = 0.61, p = 0.8309) was removed from the final model. Within inbred lines, neither inbred line (F_3,152 = 0.78, p = 0.5096) nor infection cue treatment (F_4,152 = 0.37, p = 0.8499) significantly affected antibacterial activity, but we again found a significant effect of body condition (F_1,152 = 4.36, p = 0.0384), with males in better condition having greater activity.