Volume 92, Issue 5 p. 1055-1064
RESEARCH ARTICLE
Open Access

On the use of antibiotics in plasticity research: Gastropod shells unveil a tale of caution

Denis Meuthen

Corresponding Author

Denis Meuthen

Evolutionary Biology, Bielefeld University, Bielefeld, Germany

Correspondence

Denis Meuthen

Email: [email protected]

Search for more papers by this author
Klaus Reinhold

Klaus Reinhold

Evolutionary Biology, Bielefeld University, Bielefeld, Germany

Search for more papers by this author
First published: 03 March 2023
Handling Editor: Eric Vander Wal

Abstract

  1. Through phenotypic plasticity, individual genotypes can produce multiple phenotypes dependent on the environment. In the modern world, anthropogenic influences such as man-made pharmaceuticals are increasingly prevalent. They might alter observable patterns of plasticity and distort our conclusions regarding the adaptive potential of natural populations.
  2. Antibiotics are nowadays nearly ubiquitous in aquatic environments and prophylactic antibiotic use is also becoming more common to optimize animal survival and reproductive output in artificial settings. In the well-studied plasticity model system Physella acuta, prophylactic erythromycin treatment acts against gram-positive bacteria and thereby reduces mortality.
  3. Here, we study its consequences for inducible defence formation in the same species. In a 2 × 2 split-clutch design, we reared 635 P. acuta in either the presence or absence of this antibiotic, followed by 28-day exposure to either high or low predation risk as perceived through conspecific alarm cues.
  4. Under antibiotic treatment, risk-induced increases in shell thickness, a well-known plastic response in this model system, were larger and consistently detectable. Antibiotic treatment reduced shell thickness in low-risk individuals, suggesting that in controls, undiscovered pathogen infection increased shell thickness under low risk. Family variation in risk-induced plasticity was low, but the large variation in responses to antibiotics among families suggests different pathogen susceptibility between genotypes. Lastly, individuals that developed thicker shells had reduced total mass, which highlights resource trade-offs.
  5. Antibiotics thus have the potential to uncover a larger extent of plasticity, but might counterintuitively distort plasticity estimates for natural populations where pathogens are a part of natural ecology.

1 INTRODUCTION

The adaptive potential of natural populations in a changing world relies to a large extent on phenotypic plasticity, that is the ability of individual genotypes to express multiple phenotypes dependent on the environment (Diamond & Martin, 2016; Donelan et al., 2020). However, there is huge variation in the degree of plasticity within and across individuals. Neglecting possible sources for this variation contributes to incorrect inferences regarding the extent of plasticity in nature and its evolutionary significance (Meuthen et al., 2018).

Currently, the number of anthropogenic pollutants in natural ecosystems is steadily increasing (Rhind, 2009), and this is an increasing challenge for aquatic ecosystems (Häder et al., 2020; Küster & Adler, 2014). One of the major pollutants in natural water bodies are man-made pharmaceuticals that enter through sewage, which is often untreated (Wear & Thurber, 2015) and even when it is treated, wastewater treatment plants are often inefficient in removing pharmaceuticals (Corcoran et al., 2010). As previous research consistently suggests that exposure to pharmaceuticals can interfere with the ability of animals to display adaptive behaviours across contexts (Arnold et al., 2014; Brodin et al., 2014; Martin et al., 2021; Salahinejad et al., 2022), it appears likely that they might also impact their ability to cope with environmental change through phenotypic plasticity.

A major pharmaceutical class of concern are antibiotics, whose prescription rates for human use and the husbandry of livestock have increased over the last century (Ventola, 2015). Excretion and dumping of antibiotics and the subsequent incomplete elimination in most wastewater treatment plants (Corcoran et al., 2010) have led to antibiotics being nearly ubiquitous in natural environments (Martínez, 2008). At the same time, researchers increasingly use antibiotics in the laboratory as prophylactic medication to minimize subject losses in wild-caught animals (McEwen & Fedorka-Cray, 2002).

On one hand, while little is known about these mechanisms in invertebrates, at least from vertebrate research it is known that antibiotic treatment perturbs gut microbiota (Dudek-Wicher et al., 2018; Ramirez et al., 2020), which reduces hippocampal plasticity (Tang et al., 2021) and can compromise survival in novel environments (Qin et al., 2020). On the other hand, in both vertebrates and invertebrates, pathogen infection can damage and weaken hosts as it can reduce their foraging efficiency (Makin et al., 2020; Morton, 2018) and can increase their energy expenditure for immunological responses (Ganeshan & Chawla, 2014; Sheldon & Verhulst, 1996). This may constrain plastic responses, which usually carry an energetic cost as evidenced by reduced somatic growth of plastically responding individuals (Reimer & Tedengren, 1996; Storfer & White, 2004; but see DeWitt et al., 1998). Furthermore, some pathogens can directly manipulate host phenotypes to their own benefit (Heil, 2016; Poulin & Maure, 2015), which may suppress plastic responses or induce atypical phenotypes (Baldauf et al., 2007). Alternatively, these responses may take the same shape as adaptive plasticity to a different environmental factor (Cheng, 1986; DeWitt, 1998; Tariel et al., 2020b).

Predation risk often varies across time and space. As adapting to the presence of predators via phenotypic plasticity often generates visually distinct phenotypes that greatly benefit survival, antipredator plasticity is one of the first known cases of plasticity and remains to be one of the best-studied instances of phenotypic plasticity even today (Kishida et al., 2010). In particular, predator-induced morphological defences, also known as ‘inducible defences’, are a well-known instance of antipredator plasticity (Bourdeau & Johansson, 2012). A classic inducible defence is the risk-induced plastic increase in shell thickness that many gastropod species are capable of (Hoverman, 2010).

Here, we study how antibiotic exposure impacts the expression of inducible morphological defences in the invasive (Vinarski, 2017) cosmopolitan gastropod Physella acuta (Draparnaud, 1805). This small freshwater species feeds on detritus, bacteria and algae and is a well-established model system for behavioural (DeWitt et al., 1999; Kain & McCoy, 2016; Turner, 1996), morphological (Beaty et al., 2016; DeWitt, 1998, 2016; Gustafson et al., 2014) and life-history plasticity (Auld & Relyea, 2010b; Crowl & Covich, 1990). Predator-exposed P. acuta develop thicker shells (Auld & Relyea, 2010a; Gordon, 2019; Luquet & Tariel, 2016; Tariel et al., 2020b) and thick-shelled individuals survive better in the presence of predators (Auld & Relyea, 2011). As shell production is energetically costly in gastropods (Brusca & Brusca, 2003), the production of thicker shells as an inducible defence is resource-intensive as showcased by a trade-off with allocation into somatic growth (DeWitt, 1998). At the same time, P. acuta raised under high predation risk are younger and smaller at first reproduction (Auld & Relyea, 2008; Auld & Relyea, 2010b), which is a typical evolutionary outcome given high predation risk (Reznick & Endler, 1982; Roff, 1992). Recently, it has been proposed that prophylactic antibiotic treatment would benefit research on P. acuta, where subject losses have historically compromised plasticity experiments (DeWitt & Prestridge, 2022). In this comparative study, the broad-spectrum antibiotic erythromycin, which targets gram-positive bacteria, small gram-negative bacteria and protozoa (Derrick Jr. & Reilly, 1983; Williamson & Scott-Finnigan, 1978), has been found to be most effective in increasing survival rates of wild-caught P. acuta (DeWitt & Prestridge, 2022). Thus, here, we investigate the effect of a prophylactic erythromycin treatment on predator-induced shell thickness plasticity and growth of P. acuta to examine the consequences of antibiotic treatment in plasticity research.

2 MATERIALS AND METHODS

This study did not require ethical approval. Our experimental design has been preregistered (Meuthen & Reinhold, 2022). We had conducted an earlier pilot study with similar methodology (Meuthen & Reinhold, 2021; see Appendix S1 for results) but it had too many shortcomings to be able to draw reasonable conclusions. First, shell production requires a large source of calcium in the environment (Brusca & Brusca, 2003) but we had only used soft water (<1° dKH) without calcium supplementation, which caused individuals to have abnormally thin shells (mean ± SD; 0.028 ± 0.010 mm) in comparison to other studies such as Auld and Relyea (2010a) as well as in comparison to unpublished research from the same population that was raised in hard water (0.137 ± 0.050 mm). These thin shells often broke during handling and along with large unexpected mortality caused final sample sizes to be low (n = 36–43 per treatment). Second, in the pilot study, snails that were used as donors to simulate high perceived risk were not treated with antibiotics, and thus we could not rule out the possibility of re-infection in the high-risk treatments. Thus, here, we describe a repeat of the same study that better controls for these potentially confounding factors and has larger sample sizes.

Experiments were run from February to March 2022. To generate clutches for the present study, as parents we used 22 individuals from a laboratory P. acuta population whose ancestors were collected from a pond in Bonn, Germany (50°43′52.1″N 7°04′21.7″E) in 2019. Another species of the same genus, P. pomila, is able to store sperm for beyond 60 days (Wethington & Dillon Jr, 1991). At the same time, P. acuta is a preferential outcrosser (Henry et al., 2005; Jarne et al., 2000), which makes us unsure of paternity. Thus, we consider maternal lineage as family identity so as account for genetic effects. Immediately after hatching, we separated individuals into a fully factorial 2 × 2 split-clutch design, where we crossed a 5-day antibiotic treatment (25 mg L−1 erythromycin, MardFritz Aquaticsel Maracyn; Fritz Aquatics) as outlined by DeWitt and Prestridge (2022) or a sham water control treatment with a follow-up regular exposure (three times per week as described in Auld & Relyea, 2010b, 2011) to either perceived high-risk (conspecific alarm cues, one crushed snail per four litres of water as described in Tariel et al., 2020a, 2020b) or low-risk (clean water) for 28 days. Conspecific alarm cues are a reliable indicator of elevated predation risk across taxa (Chivers & Smith, 1998). Alarm cue donors had been antibiotic-treated with the same 5-day regimen prior to the study to avoid re-infection through exposure to snail homogenate.

During exposure of experimental subjects to treatments, P. acuta were held in groups of six within ø 5 cm, 10 cm high plastic tubes that were sealed with air-permeable foam stoppers and contained 64 mL of water. As aquatic medium, we collected water from artificial ponds, passed it through a 10 μm mesh to remove almost all microbes and then aged it at 25°C for at least 72 h before use to ensure that potentially present risk cues have degraded. We considered 72 h to be sufficient as P. acuta alarm cue has been estimated to decay after 41 h (Turner & Montgomery, 2003), and the half-life of other invertebrate alarm cues ranges between 0.71 and 14.4 h (Van Buskirk et al., 2014). Full water changes were performed three times a week. Crushed algae wafers (Spirulina Tabs Nature, sera) were provided ad libitum as food. As freshwater snails obtain calcium for shell construction from both the surrounding water and their diet (Van Der Borght & Van Puymbroeck, 1966; Young, 1975), we also supplied tanks with ad libitum crushed sepia shells. Light was provided in a 12:12 light:dark cycle (from 9:00 to 21:00 h) and temperature was kept at 24.69 ± 0.92 °C (mean ± SD). More details about the breeding and raising process are provided in Appendix S2. Over the experimental period, 58 snails died (dead snails per tank, median, interquartile range, range; antibiotic treated, high risk: 1, 0–1, 0–3; antibiotic treated, low risk: 0, 0–1, 0–3; control, high risk: 0, 0–1, 0–3; control, low risk: 0, 0–0, 0–2) but there was no statistically clear difference in mortality between treatments as outlined in Appendix S3. Furthermore, one antibiotic-treated low-risk tank containing three individuals was excluded as one individual was accidentally crushed during a water change, which released alarm cues.

At an age of 33 days, when sexual maturity is reached in our P. acuta as evidenced by observable spawning throughout tanks, we sampled the remaining 635 individuals with known parentage (157–166 individuals per treatment and 17–46 individuals per family). First, we dried them with dry wipes and weighed their total body mass (i.e. including the shell) to an accuracy of 0.1 mg (Quintix 124-1S; Sartorius). Then, we measured their shell thickness to the nearest 0.01 mm with a digital calliper (Model 108-4500; Imatec) three times: at the top, at the centre and at the bottom of the aperture. From these measurements, we then calculated average shell thickness. Preliminary data suggest that this method of measuring shell thickness offers high repeatability (R) as calculated according to Stoffel et al. (2017): n = 23, R ± SE: 0.924 ± 0.035, p < 0.001. In one instance, we found that we had accidentally raised seven snails instead of six in a single tank. Higher densities negatively impact individual mass in P. acuta (Beaty et al., 2016), but as we controlled for plasticity using body mass (see below), we decided to retain these individuals in our analysis. At the same time, we verified with preliminary analyses that our main results were not affected by the inclusion or exclusion of these seven individuals. We also assessed snail fecundity as the number of present clutches at the end of the experiment but did not find statistically clear differences between treatments as outlined in Appendix S4.

Using R 4.2.2, we applied linear mixed-effect models with maximum likelihood parameter estimation via the ‘lme4’ r package v.1.1-31 (Bates et al., 2015). First, we analysed main and interactive effects of the antibiotic and risk treatments on shell thickness and body mass. Because differences in holding densities are known to impact growth in P. acuta (Beaty et al., 2016), we entered body mass, which is correlated with density (Beaty et al., 2016), as a covariate during shell thickness analysis following previous research (Tariel et al., 2020b). As established during manuscript review, family identity nested in tank identity was entered as a random intercept throughout. To obtain familywise effect sizes, we also re-ran the same models without random effects for every family. Shell thickness had to be Box–Cox transformed (λ = 1.2745) to match the normality assumptions of model residuals, whereas body mass did not require transformation. With the ‘emmeansr package v.1.8.3 (Lenth et al., 2023), we extracted estimated marginal means for fixed factors from the full model. Means and standard deviations (SD) are provided as both raw values (yRaw), and additionally as transformed values (yTr) for shell thickness. Effect sizes are given along with 95% confidence intervals in brackets throughout. Power analyses are based on Monte Carlo simulations and were performed with the ‘simrr package v.1.0.6 (Green & MacLeod, 2016). For these power analyses, we did not enter tank identity as an additional random intercept as doing so generated singularity errors.

3 RESULTS

3.1 Shell thickness plasticity

Variation in shell thickness was explained by risk exposure (Figure 1). High-risk snails developed thicker shells (mean ± SD; yRaw: 0.136 ± 0.036 mm; yTr: 0.080 ± 0.026) than low-risk ones (yRaw = 0.130 ± 0.033 mm, yTr = 0.075 ± 0.023; estimate: 0.006 [0.001, 0.012], t99 = 2.247, p = 0.027). This effect was present throughout most families as only two out of 22 families decreased their shell thickness beyond error estimates in response to risk (Table 1).

Details are in the caption following the image
(a, b) Average shell thickness and (c, d) body mass (means ± SEs) of Physella acuta that were exposed from hatching onwards for 5 days to the antibiotic erythromycin at 25 mg L−1 or to a control treatment. Subsequent to this treatment, for 28 days, individuals were either exposed to conspecific alarm cues (high risk, black circles, nantibiotic-treated = 156, ncontrol = 157) or to a control treatment (low risk, white circles, nantibiotic-treated = 156, ncontrol = 166). (b) and (d) depict familywise reaction norms (means ± SEs), each of the 22 families is represented by a different colour, and colours consistently code for the same family across figures.
TABLE 1. Familywise effect size estimates ± SE for risk and antibiotic exposure on Physella acuta shell thickness as derived from linear models with individual total mass as covariate. Negative effect sizes (i.e. instances where treatments reduced shell thickness beyond error estimates on average) are highlighted in bold font, and neutral effect sizes (i.e. no changes in thickness beyond error estimates) are shown in italics.
Family ID Sample size Risk exposure Antibiotic exposure
A1 46 0.014 ± 0.008 −0.016 ± 0.007
A2 46 0.012 ± 0.006 0.001 ± 0.006
A3 44 −0.002 ± 0.006 −0.007 ± 0.006
A4 22 0.008 ± 0.006 0.011 ± 0.006
A5 23 0.010 ± 0.010 −0.020 ± 0.009
A7 22 −0.003 ± 0.008 −0.012 ± 0.008
A8 22 0.003 ± 0.007 0.022 ± 0.007
A9 21 0.012 ± 0.008 −0.014 ± 0.008
A10 17 0.010 ± 0.015 −0.014 ± 0.016
A11 46 0.022 ± 0.007 0.020 ± 0.007
A13 25 0.000 ± 0.007 −0.021 ± 0.007
A16 46 0.005 ± 0.005 −0.007 ± 0.005
A17 24 0.023 ± 0.007 −0.020 ± 0.007
A18 46 0.000 ± 0.005 −0.004 ± 0.006
A19 20 0.017 ± 0.013 0.004 ± 0.013
A21 35 0.013 ± 0.009 −0.008 ± 0.009
A22 24 0.015 ± 0.011 −0.030 ± 0.012
A30 23 0.001 ± 0.010 −0.013 ± 0.009
A31 21 0.022 ± 0.010 0.004 ± 0.011
A33 20 −0.018 ± 0.007 0.017 ± 0.006
A34 22 −0.007 ± 0.009 0.017 ± 0.010
A35 20 −0.021 ± 0.007 0.025 ± 0.007

While on average, we found no statistically clear difference in shell thickness between antibiotic-treated snails (yRaw: 0.131 ± 0.036 mm; yTr: 0.076 ± 0.026) and controls (yRaw = 0.135 ± 0.033 mm, yTr = 0.079 ± 0.024; estimate: −0.003 [−0.009, 0.002], t97.3 = −1.151, p = 0.253, Figure 1), a closer investigation of the data suggests that different families vary strongly in their response to antibiotics. Following antibiotic exposure, 6 families showed no plasticity beyond error estimates, 6 families increased shell thickness and 10 families decreased it (Table 1).

Furthermore, we could not find evidence for risk and antibiotic treatment to influence each other as there was no statistically clear interaction between antibiotic and risk treatment (estimate: 0.004, [−0.007, 0.015], χ21 = 0.493, p = 0.483).

Post-hoc investigation revealed that effect sizes of alarm cue-induced antipredator plasticity were more than twice as large in the antibiotic-exposed individuals (yRaw: 6.49% increase, yTr: 8.57% increase, estimate: 0.008 [0.000, 0.017], dCohen = 0.42 [0.02, 0.82], t98.4 = 2.070, p = 0.041) compared to the control individuals (yRaw: 2.82% increase, yTr: 3.44% increase, estimate: 0.004 [−0.004, 0.013], dCohen = 0.23 [−0.17, 0.62], t97.6 = 1.113, p = 0.269). The underlying cause for this effect was that while not being statistically different, in low-risk individuals, antibiotic exposure reduced shell thickness (yRaw: 5.22% decrease, yTr: 6.79% decrease, estimate: −0.005 [−0.013, 0.002], dCohen = 0.27 [−0.14, 0.67], t95.8 = −1.302, p = 0.196) to a greater extent than in high-risk individuals (yRaw: 1.24% decrease, yTr: 1.11% decrease, estimate: −0.001 [−0.009, 0.007], dCohen = 0.07 [−0.33, 0.46], t98.5 = −0.327, p = 0.745).

Lastly, heavier snails developed thinner shells, as suggested by a general negative relationship between weight and shell thickness (estimate: 0.457 [0.303, 0.0.611], t627 = 5.833, p < 0.001).

3.2 Power analyses on shell thickness plasticity

Power analysis revealed that for antibiotic-treated snails, statistical power of 80% for a plasticity-induced modification in shell thickness is achieved with a total sample size of 250 individuals, or 125 individuals per treatment (Figure 2a). In contrast, the detection of shell thickness plasticity in control snails with the same statistical power requires testing ca. 557 individuals, or 279 individuals per treatment (Figure 2b).

Details are in the caption following the image
Power analysis given α = 0.05 on detecting risk-induced shell thickness plasticity in the studied Physella acuta population (a) subsequent to antibiotic treatment and (b) in the absence of medical intervention. The y-axis represents statistical power, that is the probability that a statistical test correctly rejects the null hypothesis of no plasticity when the alternative hypothesis of a risk-induced change in shell thickness is true. Means and 95% confidence intervals are based on 1000 Monte Carlo simulations of linear-mixed-effect models with risk treatment and body mass as fixed effects, and family identity as random intercept. The dashed line indicates the standard adequate statistical power of 80%.

3.3 Body mass

High-risk snails (yRaw = 0.036 ± 0.012 g) weighed less than low-risk individuals (yRaw = 0.040 ± 0.013 g, estimate: −0.004 [−0.007, −0.002] g, t96.2 = −3.243, p = 0.002). Additionally, antibiotic-treated snails (yRaw = 0.039 ± 0.013 g) were slightly heavier than controls (yRaw = 0.037 ± 0.013 g, estimate: 0.002 [0.000, 0.005] g, t96.2 = 1.720, p = 0.089). These effects were consistent as we could not observe a statistically clear interaction between antibiotic and risk treatment (estimate: 0.000 [−0.004, 0.005] g, χ21 = 0.041, p = 0.840).

4 DISCUSSION

On average, high predation risk induced the commonly observed plastic increase in shell thickness that matches previous research (Auld & Relyea, 2011; Luquet & Tariel, 2016; Tariel et al., 2020b) and which is beneficial for survival in a predatory habitat (Auld & Relyea, 2011). Variation in this response across families was small, in accordance with other inducible defence studies across taxa (Meuthen et al., 2018; Ogran et al., 2020; Relyea, 2005). This confirms that predation risk is generally a strong selective force (Lima, 1998; Lima & Dill, 1990; Sih et al., 2000) which effectively eliminates genotypes incapable of mounting appropriate inducible defences in both gastropods (Auld & Relyea, 2011) and other taxa (Hoverman, 2010; Nilsson et al., 1995; Van Buskirk & Relyea, 1998).

At the same time, on average, antibiotic treatment induced a small decrease in shell thickness, but families strongly varied in their response. Follow-up analysis revealed that antibiotic-treated individuals had greater shell thickness plasticity than controls. Underlying this effect is that antibiotic treatment caused low-risk individuals to develop thinner shells, whereas this treatment did not affect the ceiling of the response under high risk. These results suggest that the investigated P. acuta population is infected by pathogens that in the absence of antibiotics increase shell thickness even under low-risk conditions and thereby make risk-induced shell thickness plasticity more difficult to detect. This follows previous research suggesting that pathogens alter host phenotypes across diverse taxa (Heil, 2016; Poulin & Maure, 2015), including gastropods. For example, in the snail Potamopyrgus antipodarum, infection by the trematode Microphallus sp. is correlated with a lack of defensive spines and wider shells (Levri et al., 2005) as well as with earlier reproduction (Jokela & Lively, 1995). In P. acuta, previous research suggests that phenotypic responses are parasite specific. Infection by the trematode Halipegus eccentricus does not affect P. acuta shell morphology and only modestly reduces crush resistance (Gustafson & Bolek, 2016). In contrast, infection with a different trematode, Echinostoma revolutum, was shown to release calcium from storage cells within the digestive gland, which leads to increased calcium deposition in the shell (Cheng, 1986). As exposure to predation risk likewise increases calcium deposition in shells so as to increase thickness as an adaptive defence (Auld & Relyea, 2011; Tariel et al., 2020b), this mechanism would explain the observed increase in shell thickness under low risk. However, as trematodes use snails only as intermediate hosts, and we transferred clutches into clean water for 13 generations, we believe it to be unlikely that trematodes or other macroparasites are involved in the observed responses here. Instead, we consider it to be more likely that the observed responses are caused by microparasites, mostly likely by gram-positive bacteria, small gram-negative bacteria or protozoa as these pathogens are sensitive to erythromycin (Derrick Jr. & Reilly, 1983; Williamson & Scott-Finnigan, 1978). Unfortunately, to our knowledge, there is no published research available on which specific bacterium or protozoan may induce greater shell thickness in untreated P. acuta similar to E. revolutum infection, and this constitutes an avenue for exciting follow-up research.

At the same time, there was large variation between families in their response to antibiotics (Table 1, Figure 1). First, in the absence of antibiotics, many families had an elevated shell thickness in the low-risk environment, leading to flat or non-adaptive reaction norms (i.e. unaltered or reduced shell thickness in response to high risk). For most families, antibiotic treatment then lowered the floor of the response in the low-risk environment and thereby reversed their reaction norm in an adaptive direction. From these observations, we consider it to be likely that some genotypes are more susceptible to infection, thus had the highest pathogen load, and therefore were the ones most responsive to antibiotic treatment. In contrast, other families likely had a low susceptibility to pathogens, were not infected or had a low pathogen load, and those were the ones largely unaffected by antibiotic treatment. Indeed, previous research highlighted genetic differences in susceptibility of the freshwater snail P. antipodarum to infection by the trematode Microphallus sp. (Gibson et al., 2016; Krist et al., 2004), and similar genetic differences may also cause variable levels of susceptibility against bacteria or protozoans that are sensitive to erythromycin. Taken together, when sampling a population, the proportion of genotypes with high susceptibility to pathogens (i.e. the proportion of genotypes infected with high pathogen loads) may explain variation between studies on inducible defences and exposure to antibiotics can remediate this issue as it makes responses more consistent across genotypes. This hypothesis is supported by the observation that results from our pilot study (Appendix S1) and the current ones were only consistent within antibiotic-treated individuals while control snails varied strongly in their response to risk between studies. This is presumably because the proportion of susceptible genotypes within sampled snails differed between studies. Additionally, as the absence of antibiotic treatment in alarm cue donors of the pilot study did not alter the observable response of antibiotic-treated individuals compared to the present one, we can exclude the possibility of re-infection of individuals from the cue-generating snail homogenate.

Moreover, increased shell thickness was associated with lower body mass, in accordance with suggested costs of trait production. As gastropod shell production is energetically expensive (Brusca & Brusca, 2003), there is a trade-off between allocation in inducible defences and allocation into somatic growth in P. acuta (DeWitt, 1998). Similar trade-offs have been observed in other taxa that have evolved morphological antipredator plasticity (Reimer & Tedengren, 1996; Storfer & White, 2004; but see Bourdeau & Johansson, 2012; Meuthen et al., 2019). This trade-off also explains why high-risk individuals (which developed thicker shells) had lower body mass than low-risk controls overall, and why antibiotic-treated snails (which had an overall lower shell thickness) were heavier than controls. At the same time, the observation that antibiotic-treated snails were overall heavier than controls suggests that antibiotic exposure did not negatively affect growth through, for example, a negative impact on the external food microbiome or the internal gut microbiome. Rather, the removal of pathogens through antibiotic treatment appears to be beneficial for overall somatic growth in P. acuta.

Taken together, our results suggest that antibiotic treatment is not only useful in reducing mortality when wild-caught gastropods are used (DeWitt & Prestridge, 2022), but can also remove undetected pathogens that can cause us to incorrectly estimate the full spectrum of plasticity that an animal is capable of. Power analyses on the present data suggest that in the presence of pathogens in sampled snail populations, reasonable statistical power for the detection of shell thickness plasticity is reached only with sample sizes of 279 individuals per treatment. This number is far beyond what has been reported in previous P. acuta shell thickness plasticity research: 20 individuals per treatment (Auld & Relyea, 2010a), 9–86 individuals per treatment (Tariel et al., 2020b) or 87–103 individuals per treatment (Luquet & Tariel, 2016). This suggests that the populations examined in these studies likely had a low pathogen load and/or genotypes with a low susceptibility to pathogens, and also that in the absence of antibiotic treatment, shell thickness plasticity may be undetectable using commonly applied sample sizes in some P. acuta populations. Given the general lack of researchers considering potential confounding effects of parasite and pathogen infection during animal studies (Chrétien et al., 2022), we do not know how many previous P. acuta studies remained unpublished because there were more genotypes with high susceptibility to pathogens among sampled snails. Even if some pathogens do not directly affect snail mortality, they may have prevented researchers from being able to detect plastic responses. Moving individuals into a laboratory environment at large holding densities along with the absence of natural selection may facilitate the accumulation of pathogen-susceptible genotypes as well as the prevalence of pathogens and thereby exacerbate this issue beyond what would be observable under natural conditions. In such instances, only antibiotic treatment may enable researchers to detect and study the full scale of evolved plasticity, as is useful when studying research questions related to whether plasticity in a trait has evolved, the costs of plasticity, or when correlations between plasticity and other traits such as personality traits are investigated (Betini & Norris, 2012). However, while antibiotic treatments may be useful in revealing the full extent of plasticity, counterintuitively, their application may also cause researchers to overestimate the amount of plasticity in natural environments where as part of natural ecology, pathogens, parasites and other interacting environmental factors are regularly present. This is especially risky during conservation-related research which aims to assess the adaptive potential of populations to persist in the face of stress from environmental change (Diamond & Martin, 2016; Donelan et al., 2020).

AUTHOR CONTRIBUTIONS

Denis Meuthen—Conceptualization, Funding acquisition, Data curation, Formal Analysis, Investigation, Methodology, Project administration, Resources, Software, Validation, Visualization, Writing—original draft; Klaus Reinhold—Conceptualization, Funding acquisition, Resources, Supervision, Writing—review & editing.

ACKNOWLEDGEMENTS

This research was financially supported by the Evolutionary Biology Department of Bielefeld University (awarded to Klaus Reinhold) and by a Freigeist Fellowship of the VolkswagenStiftung (awarded to Denis Meuthen). Open Access funding enabled and organized by Projekt DEAL.

    CONFLICT OF INTEREST STATEMENT

    The authors declare no competing interests.

    DATA AVAILABILITY STATEMENT

    Data supporting this article are available from the Dryad Digital Repository https://doi.org/10.5061/dryad.rn8pk0pgp (Meuthen & Reinhold, 2023a). The R code supporting this article is available from Zenodo https://doi.org/10.5281/zenodo.7687077 (Meuthen & Reinhold, 2023b).

      Journal list menu