In heterogeneous environments, mobile species should occupy habitats in which their fitness is maximized. Mangrove rivulus fish inhabit mangrove ecosystems where salinities range from 0 to 65 ppt, but are most often collected from areas with salinities of ∼25 ppt. We examined the salinity preference of mangrove rivulus in a lateral salinity gradient, in the absence of predators and competitors. Fish could swim freely for 8 h throughout the gradient with chambers containing salinities ranging from 5 to 45 ppt (or 25 ppt throughout in the control). We defined preference as the salinity in which the fish spent most of their time, and also measured preference strength, latency to begin exploring the arena, and number of transitions between chambers. To determine whether these traits were repeatable, each fish experienced three trials. Mangrove rivulus spent a greater proportion of time in salinities lower (5–15 ppt) than they occupy in the wild. Significant among-individual variation in the (multivariate) behavioral phenotype emerged when animals experienced the gradient, indicating strong potential for selection to drive behavioral evolution in areas with diverse salinity microhabitats. We also showed that mangrove rivulus had a significantly greater probability of laying eggs in low salinities compared with control or high salinities. Eggs laid in lower salinities also had higher hatching success compared with those laid in higher salinities. Thus, although mangrove rivulus can tolerate a wide range of salinities, they prefer low salinities. These results raise questions about factors that prevent mangrove rivulus from occupying lower salinities in the wild, whether higher salinities impose energetic costs, and whether fitness changes as a function of salinity.

An animal's survival and reproductive success depend on its ability to either operate in variable environments or relocate when conditions become suboptimal. The decision to stay or leave is ultimately based on which option maximizes the animal's fitness (Nguyen et al., 2013; McManus et al., 2014). If the benefits of relocating outweigh the costs, then the animal should disperse (Caughley, 1994). Costs of dispersal include use of energy, risk of injury or death, and outbreeding depression (Bonte et al., 2012), while benefits include escaping unfavorable conditions, obtaining new resources, and decreased chances of inbreeding (Caughley, 1994). However, for an animal to gauge the magnitude of benefits that it might receive from moving, it must have information about alternative habitats, which can be obtained by exploring new areas, contingent upon the aforementioned costs (Nguyen et al., 2013).

Coastal ecosystems are characterized by diverse microhabitats that are relatively close in proximity, making it possible for mobile aquatic organisms to gather information about surrounding habitats. Aquatic species often have a particular range of salinities that they can tolerate, but also a salinity in which their fitness is highest (Boeuf and Payan, 2001). When the ability to disperse is limited, the habitats in which animals settle can significantly constrain fitness. For example, growth and survivorship of the barnacle Balanus amphitrite are negatively impacted when the animal occupies salinities greater than or equal to 10 ppt (Qiu and Qian, 1999). Because barnacles remain attached to a substrate during adulthood and cannot readily escape unfavorable environmental conditions, they must endure these consequences in the event of salinity fluctuations. Fishes, in contrast, have the ability to disperse throughout their lifetime, which allows them to move to areas with more favorable salinities, if available (Bonte et al., 2012). Salinity preference thus plays an important role in habitat selection for aquatic organisms living in brackish environments, which can vary in salinity both spatially and temporally, creating distinct microhabitats (Surge and Lohmann, 2002). Many organisms that inhabit variable environments tend to have wide tolerance ranges (Gabriel, 2005; Schultz, and McCormick, 2012). For example, the killifish Fundulus heteroclitus is able to tolerate shifts in temperature, pH, salinity and oxygenation; each of these factors varies significantly within their salt marsh habitat (Schulte, 2014). In any case of habitat selection, there are most likely a set of environmental conditions in which the animals experience highest fitness, and a preference for these habitats should be selected for (Kearney and Porter, 2004).

When organisms occupy a particular habitat, it could be because the conditions in that habitat confer the highest fitness, or it could reflect biotic and abiotic factors (e.g. competition, predation, salinity and temperature) that limit occupancy of optimal habitats (Svärdson, 1949; Kearney and Porter, 2004). To determine where an animal achieves the highest absolute fitness in a multidimensional environment, all factors other than the one of interest must be controlled for. In addition, animals should be exposed to the full, ecologically relevant range of the abiotic factors (i.e. its fundamental niche) (Pearman et al., 2008). When other things that limit dispersal are present (e.g. competition, predators), the fundamental niche narrows to the realized niche (Morse, 1974), which is typically where organisms are found in their natural environments. Although individuals of a given species are often found within a given niche, among-individual variation around that average niche space can exist, and this variation is then subject to selection. Preference studies can provide insights into whether habitat selection is constrained by other factors, and reveal whether, in the wild, the animals occupy their realized or fundamental niche. In the laboratory, extraneous variables can be controlled while examining preferences, which should provide information about the conditions under which the animal might experience highest absolute fitness. Furthermore, laboratory settings also allow for assessment of repeatability of that preference, which can be difficult in field-based studies on organisms that show low site fidelity. Quantifying repeatability, the proportion of total phenotypic variation that is due to among-individual differences (Falconer and Mackay, 1996; Lessells and Boag, 1987; Boake, 1989), is essential for understanding the potential for selection to drive the evolution of salinity preferences (Brodie and Russell, 1999; Boake, 1989; Arnold, 1994).

Mangrove rivulus fish, Kryptolebias marmoratus Poey 1880, are small, self-fertilizing hermaphroditic vertebrates found in a wide range of microhabitats within mangrove ecosystems of Florida, the Bahamas, parts of the Caribbean and Central America (Huber, 1992). Field data have shown that they exist in a broad range of salinities. Based on data from 274 different field sites, we have collected mangrove rivulus in salinities ranging from 0 to 65 ppt, with a mean±s.e.m. salinity of 26±0.44 ppt (see also Taylor, 2012). However, very few mangrove rivulus eggs have been collected in the field, so habitat preferences for egg-laying remain unknown (Taylor, 1990). Mangrove rivulus can exist in a wide range of salinities in the field, even over small spatial scales (Sutton et al., 2018), despite the apparent costs that the animals might accrue at both low and high salinities. For example, Lin and Dunson (1999) showed that exposure to different salinities early in life significantly affected adult mass; treatment animals raised at 12 and 40 ppt had significantly higher final masses than those raised at 1 ppt. Mortality rates of mangrove rivulus living at 12 and 40 ppt were also significantly greater than those living at the lower salinity (Lin and Dunson, 1999). Overall, mangrove rivulus raised in the lowest salinity (1 ppt) matured at a slower rate, grew to a smaller size and produced fewer eggs than those reared in higher salinities (Lin and Dunson, 1995). When salinities deviate from the isosmotic point, animals can incur significant costs; however, this is not always the case and it is not always straightforward to predict which salinities are associated with elevated physiological costs (e.g. Ern et al., 2014). While many freshwater fishes do best (e.g. have lower metabolic rates) in freshwater or have metabolic rates indistinguishable from those at an isotonic salinity, some saltwater fishes do best at the isosmotic point and others show no increase in physiological costs at higher salinities (Ern et al., 2014). It appears that the most pronounced costs are experienced when salinities differ from those the fish was reared in (Ern et al., 2014), but very low or very high salinities might require that the animal dedicate more energy towards osmoregulation, perhaps at the expense of growth, resulting in a smaller fish. Indeed, Sutton et al. (2018) showed, in mangrove rivulus, that acute exposure to increased salinity concentrations caused elevated metabolic rates and activity levels. Different salinities significantly affect growth rate in a variety of other fish species, both freshwater and marine, which may be due to the relative amounts of energy being devoted to osmoregulation versus somatic processes (Boeuf and Payan, 2001).

Because salinity seems to have such a large impact on the growth, mortality and reproduction in mangrove rivulus, it seems reasonable that selection might have acted to shape relatively narrow salinity tolerances and strong salinity preferences in this species. This motivated a controlled laboratory study to identify the preferred salinity, which is likely the salinity at which the lowest costs would be incurred. Although previous studies have demonstrated the ability for mangrove rivulus to tolerate various salinities, and the effects that those salinities can have on reproduction and survival (Lin and Dunson, 1995; Lin and Dunson, 1999; Frick and Wright, 2002; Taylor, 2012; Sutton et al., 2018), none have attempted to determine whether the species has a salinity preference and if there is among-individual variation around the species-level average preference.

Environmental conditions in the area where eggs are laid can have significant impacts on offspring survival and phenotype. Specifically, the salinities that aquatic species are exposed to during early life can have considerable effects on the phenotype, which are driven largely by the increased energy demands of osmoregulation (Urbina and Glover, 2015). Some salinities result in reduced hatching success and larval survival, as well as decreased size at hatching and growth rate (Berlinsky et al., 2004; Mihelakakis and Yoshimatsu, 1998; Zhang et al., 2010; Ramee and Allen, 2016). Because osmoregulation comes at a cost, brackish water species can conserve energy by inhabiting areas in which they are isotonic (Boeuf and Payan, 2001); it would then be expected that oviposition sites should also be selected for in this manner to reduce the potential energy cost sustained by offspring during development. During any given egg-laying bout, mangrove rivulus lay very few eggs (often just one) (Harrington, 1963; Lomax et al., 2017).

In this study, we controlled for extraneous environmental variables such as water and air temperature, light, food availability, competition and predation to determine how mangrove rivulus would distribute along a salinity gradient when salinity was the only difference among available microhabitats. The objectives of this study were to: (1) determine whether mangrove rivulus exhibit salinity preferences, including measurements of strength of preference, number of transitions between salinities, latency to begin exploring the salinity gradient, and the covariance among these traits; (2) establish whether these behavioral traits are repeatable; (3) determine whether the preference of mangrove rivulus in the laboratory corroborates field collection data; (4) examine whether mangrove rivulus have a salinity preference for oviposition sites; and (5) determine the effects of developmental salinity on hatching success. We hypothesized that mangrove rivulus would have a salinity preference and that their preference would be repeatable. A preference was inferred if the fish spent significantly more time in one salinity compared with others. It was predicted that the fish would show a preference for 25 ppt because this is the salinity in which mangrove rivulus are found most commonly in the wild and at which they are raised in our laboratory colony. Previous work suggests that the isotonic point for mangrove rivulus is nearer to 15 ppt (Frick and Wright, 2002; Bielmyer et al., 2012), leading to an alternative prediction that the fish would prefer salinities lower than 25 ppt. Additionally, we hypothesized that they would have a preferred salinity in which to lay eggs and that the salinity experienced during development would influence hatching success. We predicted that mangrove rivulus would prefer to lay eggs in 25 ppt and that eggs laid in 25 ppt would have the highest hatching success compared with eggs laid in any other salinity. Alternatively, mangrove rivulus might choose to lay their eggs, and the eggs might fare better, at salinities closer to the isosmotic point of 15 ppt.

Housing conditions

When not being tested, all fish were housed in ventilated, 1.2 liters Rubbermaid® containers filled with 25 ppt salt water (Instant Ocean® salt and aged tap water). All individuals were kept under a 12 h:12 h light:dark photoperiod at a temperature of 25.42±0.0043°C (mean±s.e.m.), and were fed 2 ml of live brine shrimp (Artemia) nauplii reconstituted in water 6 days per week. All fish were adult hermaphrodites, aged between 133 and 379 days old when entering their first treatment (mean±s.e.m. age=252±8.62 days). The University of Alabama Institutional Animal Care and Use Committee approved all procedures described herein (IACUC no. 15-10-0111).

Salinity preference in a lateral gradient

Genotype selection

Mangrove rivulus are self-fertilizing hermaphrodites and are able to produce isogenic lineages, wherein all individuals share the same genotype. The genotypes used in this study were derived from a broad geographical range, including Belize, the Bahamas, the Florida Keys, East Florida and West Florida. Sixty-three genotypes with a field-caught progenitor (F0 generation) that was homozygous at a minimum of 31 out of 32 microsatellite loci (Avise and Tatarenkov, 2015; Tatarenkov et al., 2012) were selected for salinity preference trials. All fish used in this experiment were one (F1) or two (F2) generations removed from wild-caught progenitors. A sample of 16 genotypes from across the geographical range had animals represented in the salinity preference trials and the control trials (see details below). Separate individuals were used in the control and experimental group. The total sample size for salinity preference trials was thus 79 individual fish from 63 genotypes. Animals were selected in this way so as to maximize genetic diversity in the study. We did not have replicates of the same genotype within a treatment, thus the study focused on among-individual variation (repeatability) rather than variation among specific genotypes (heritability).

Experimental design

Lateral salinity gradients were built according to Staaland (1969) with modifications as outlined by McManus et al. (2014) using 74 liter aquaria (Fig. 1). Each aquarium was divided length-wise using black corrugated plastic to create two gradients per tank, each measuring 74.6×14.6×29.8 cm, henceforth referred to as half-tank. All dividers were constructed from 6.35 mm black corrugated plastic and secured using marine aquarium silicone, such that each side of the tank was completely separated from the other. Each half-tank salinity gradient consisted of five U-shaped chambers (13.8×14.6×8.8 cm) containing the following experimental salinities: 5, 15, 25, 35 and 45 ppt. The salinity gradient remained stable for at least 7 days, as indicated by a low coefficient of variation for salinity measurements across an 8-day trial, even with a fish allowed to swim freely through the gradient (Table S1). The outside of each aquarium was covered with light green paper to minimize disturbance and to easily visualize fish on videos. Webcams (Logitech, Suzhou) were suspended above the aquaria to monitor the fish's location throughout the salinity preference trials. Two days before each trial, salinity concentrations of 5, 15, 25, 35 and 45 ppt were prepared using aged tap water and Instant Ocean® Aquarium Sea Salt (Spectrum Brands, Blacksburg, VA, USA). Salinity concentrations were then checked for accuracy using a handheld refractometer. An air stone was placed in each salinity reservoir to aid in the removal of chlorine from tap water. A pump was placed in the 45 ppt reservoir to prevent the salt from settling at the bottom; agitation from the air stones was sufficient to prevent settling in the other salinities. The temperature of the water was recorded before each trial and maintained an average of 26.8±0.35°C.

Fig. 1.

Experimental setup. Side view of salinity gradient tank with (A) and without (B) barriers (gray squares) in place. Rubber barriers were used when filling the tank to prevent mixing.

Fig. 1.

Experimental setup. Side view of salinity gradient tank with (A) and without (B) barriers (gray squares) in place. Rubber barriers were used when filling the tank to prevent mixing.

Before beginning a trial, all salinity mixtures were checked with a handheld refractometer again for accuracy. Rubber barriers were placed on top of the dividers that were connected to the bottom of the half-tank (Fig. 1A) (McManus et al., 2014). To randomize the direction of the gradient each time it was set up, a coin was flipped (heads right, tails left) to determine whether the left or right side of the half-tank would contain the lowest salinity concentration. For example, if the left side had the lowest salinity, then the chambers would increase left to right in the following fashion: 5, 15, 25, 35 and 45 ppt. Then, each chamber was filled with 1.8 liters of the premixed salinity. For the control group, the half-tank was filled with 25 ppt water in each chamber to evaluate chamber preference independent of salinity concentration. After all of the chambers were filled, the fish was gently placed in the center chamber (25 ppt) with a small fish net and allowed 30 min to acclimate to the half-tank. Following acclimation, video cameras were turned on and rubber barriers were removed. After 8 h, cameras were turned off, and the fish was removed from the half-tank with a small net and returned to its original housing. The half-tanks were then emptied using a siphon and rinsed with freshwater. Videos were then analyzed using JWatcher 1.0 (www.jwatcher.ucla.edu) to determine the amount of time spent per chamber and the number of transitions between chambers. All salinity preference fish were tested three times with 21 days between trials to minimize learning or habituation effects (the direction of the gradient was determined randomly for each trial).

Salinity preference trails were run between September and December 2015. Fish were then fed 4 ml of brine shrimp per day and monitored for egg laying. Once a fish had laid eggs, they were then used for egg laying preference trials from January through June 2016, as described below. Additional fish were added to the egg laying experiment to account for those from the salinity preference experiment that never laid eggs.

Statistical analysis

The time that it took the fish to first transition out of the 25 ppt central chamber (latency to emerge) was removed for each trial to avoid biasing data by an individual's motivation. We generated an average preference score by multiplying the number of seconds spent in a chamber by the assigned chamber number (centered at 25 ppt with 1 unit difference between chambers, i.e. 5 ppt=−2, 15 ppt=−1, 25 ppt=0, 35 ppt=1 and 45 ppt=2) and then dividing the result by the total number of seconds. A more strongly negative score thus indicates a preference for lower salinities, while a more strongly positive score indicates a preference for higher salinities. This score is then defined as an individual's preference within a given trial. Variance for an individual's preference scores was also calculated, which describes the strength of preference (where low variance indicates high strength of preference for a given salinity). To avoid confusion in the graphics, the negative of the variance was plotted such that higher values indicate stronger preference. The number of transitions between chambers was used to calculate the transition probabilities for each individual's trial using Python 2.7 (http://www.python.org) with code developed by one of the authors (H.L., Script 1). To determine the effect of treatment on latency to emerge, preference, strength of preference and number of transitions, we used lme4 package in R (Bates et al., 2015) to conduct general linear mixed models (LMM) for each variable. For each response variable in turn, we ran models with and without the fixed effect of treatment (note that a random effect of Fish ID is retained in all models; equation shown for preference only):
formula
(1)
formula
(2)
where β1 is the estimate associated with the treatment effect and e is the residual error. We examined the among-individual covariance structure between preference, strength of preference, number of transitions and emergence latency separately for control and experimental treatments using multivariate mixed models in the ASReml package in R (https://www.vsni.co.uk/software/asreml/). For each treatment, two multivariate models were compared that differed in the among-individual covariance structure. Both models fitted an intercept for each trait, and a trait-specific fixed effect of the round of trials. Each model also included an unstructured covariance matrix for the residual (co)variation between the four traits. The first multivariate model used an unstructured covariance matrix (indicated in Eqn 3 below as us:FishID), enabling the partitioning of all among-individual variances and covariances between the four response traits. The second multivariate model was constrained such that there was no among-individual covariance (indicated in Eqn 4 as idh:FishID) as follows:
formula
(3)
formula
(4)

The models were then compared using a log-likelihood test to determine whether there was any evidence for among-individual correlation structure. This was done following Houslay and Wilson (2017), where a chi-square value was calculated as −2×(log likelihood of model 1−log likelihood of model 2) to determine whether among-individual correlation structure existed. To determine the significance of pairwise correlations, we calculated a z-score (estimate/s.e.), where z-scores >|1.96| were considered significant. However, we were unable to estimate the among-individual correlations in the control treatment as only one trait (number of transitions) had any measurable among-individual variation.

To determine repeatability, we then used the ASReml package in R (https://www.vsni.co.uk/software/asreml/) to run general linear mixed models (GLMM) with treatment [gradient versus control (just 25 ppt)] and round (first, second or third trial) as fixed effects and Fish ID as a random effect. A separate model was run for each dependent variable – preference, strength of preference, latency to emerge from the acclimation chamber, and transitions between chambers. Both latency to emerge and transition variables were log transformed to achieve normality of model residuals. To determine whether any of the behavioral responses were repeatable, two models were run for each response in each treatment condition separately to parse the variance into total variance, among-individual variance and within-individual (residual) variance as follows (shown for preference only):
formula
(5)
formula
(6)

The variance components from each model were then compared to determine whether behavioral responses were repeatable. This was done following Houslay and Wilson (2017), where a chi-square value was calculated as −2×(log likelihood of model 1−log likelihood of model 2) to determine whether among-individual variance (random effect of Fish ID) was significant. The pin function in the nadiv package (Wolak et al., 2012) was then used to calculate the adjusted repeatability by dividing among-individual variance by the sum of among-individual and residual variance; this function also provided the standard error for the repeatability estimate.

Egg laying preference

Genotype selection

A group of 67 individuals were selected as experimental fish, and an additional 33 were selected as control fish. These animals were derived from 53 genotypes, all of which were also represented in the salinity preference experiment. Field-caught progenitors were homozygous at a minimum of 31 out of 32 microsatellite loci (Avise and Tatarenkov, 2015; Tatarenkov et al., 2012). All fish used in this experiment were F1 or F2 generation and laid viable, fertilized eggs (i.e. perivitelline space present) prior to the trial to ensure that they were reproductively active and capable of effective self-fertilization.

Experimental design

We modified the lateral salinity gradient for egg laying by adding to each chamber Poly-Fil fiber situated at the air–water interface as an egg laying substrate. Fish were placed in the gradient for 2 weeks to lay eggs. After 1 week, the fish was removed from the gradient and placed into a 1.2 liter Rubbermaid® container of the same salinity as the chamber they were located in at the time of capture; this allowed us to check the chambers and Poly-Fil for eggs. The gradient was then emptied and refilled. Once the gradient was re-established, the fish was returned to the chamber in which it was located prior to the egg check. The location and number of eggs per chamber were recorded. At the end of the second week, fish were returned to their original housing area and the gradient and Poly-Fil were checked again for eggs. While in the gradient, fish were fed by adding 2 ml of brine shrimp to each chamber daily (so as to avoid chamber preferences associated with food). All eggs were stored in containers with the same salinity as they were found in until hatching. To determine hatchling success, eggs were checked daily to record the date of hatching and received weekly water changes with the same salinity in which they were laid.

Statistical analysis

The presence or absence of an egg in each chamber was used for our analysis so as to not bias a particular salinity if a fish laid multiple eggs in a single chamber. This is important because the number of eggs laid by each individual was highly variable, ranging from 0 to 20 eggs across the 2 week period. Using the lme4 package in R (Bates et al., 2015), we ran a GLMM to test egg laying preference in a salinity gradient. In the following model, treatment refers to either the salinity gradient or the control where all chambers were filled with 25 ppt, while chamber refers to the location the egg was laid. To account for a possible edge effect (see Fig. 4B), we included whether a given chamber was an ‘edge’ in our model. Additionally, because the gradient remained stable for only 1 week, there were two egg checks for each fish during the 2-week period. Thus, we included ‘time’ as a fixed effect to account for any variance in egg laying between the two egg-checking periods but, because ‘time’ itself was not central to the hypotheses that we were testing, we did not include its interactions with other fixed effects:
formula
To determine the effects of salinity on hatching (yes or no), a general linear mixed model with a binomial distribution and logit link function was used as follows, with parent ID as a random effect:
formula
(8)

Comparisons between the salinities were then made by least squares means independent contrasts.

Salinity preference

When in a salinity gradient, mangrove rivulus showed a significant preference for lower salinities (Table 1, Fig. 2A,B). In the control group, where there was no salinity gradient, individuals spent more time in chambers at each edge. Strength of preference was significantly higher in the salinity gradient than in the control (Table 1, Fig. 2C). There was no difference between experimental and control groups in the total number of transitions between chambers (Table 1, Fig. 3) or in latency to emerge from the central chamber at the start of the trial (Table 1). For the experimental group, strength of preference, number of transitions and latency to emerge were significantly repeatable, with between 38 and 53% of the total behavioral variance being attributed to among-individual differences (Table 2). Preference was not repeatable in the experimental group (Table 2). In the control group, only the number of transitions between chambers was repeatable, with ∼40% of the behavioral variance being attributed to among-individual differences (Table 2). Multivariate model comparisons of among-individual correlation structures showed strong among-individual correlation structure in the experimental treatment (χ26=26.698, P=0.0001; Table 3) but not the control (χ26=0.021, P=0.999).

Table 1.

Summary of general linear mixed models for salinity preference, strength of preference (negative of the variance), transitions between chambers and latency to emerge from the central chamber

Summary of general linear mixed models for salinity preference, strength of preference (negative of the variance), transitions between chambers and latency to emerge from the central chamber
Summary of general linear mixed models for salinity preference, strength of preference (negative of the variance), transitions between chambers and latency to emerge from the central chamber
Fig. 2.

Salinity preference of mangrove rivulus, Kryptolebias marmoratus, in control and experimental treatments. (A) Average time spent in each salinity, (B) preference and (C) strength of preference. For preference, the scores have been converted to ppt for ease of interpretation, and strength of preference is graphed as the negative of the variance such that higher scores for strength (=1/lower variance) indicates a stronger preference.

Fig. 2.

Salinity preference of mangrove rivulus, Kryptolebias marmoratus, in control and experimental treatments. (A) Average time spent in each salinity, (B) preference and (C) strength of preference. For preference, the scores have been converted to ppt for ease of interpretation, and strength of preference is graphed as the negative of the variance such that higher scores for strength (=1/lower variance) indicates a stronger preference.

Fig. 3.

Probability of transitioning from one chamber to the next. Probabilities are derived from the number of transitions between each chamber by each fish. Wider arrows indicate greater likelihoods of transitioning between adjacent chambers, and the arrowhead indicates the direction of transition. Actual transition probabilities are associated with their respective arrows.

Fig. 3.

Probability of transitioning from one chamber to the next. Probabilities are derived from the number of transitions between each chamber by each fish. Wider arrows indicate greater likelihoods of transitioning between adjacent chambers, and the arrowhead indicates the direction of transition. Actual transition probabilities are associated with their respective arrows.

Table 2.

General linear mixed model by residual maximum likelihood for repeatability of preference, strength of preference, transitions between chambers and latency to emerge from the central chamber

General linear mixed model by residual maximum likelihood for repeatability of preference, strength of preference, transitions between chambers and latency to emerge from the central chamber
General linear mixed model by residual maximum likelihood for repeatability of preference, strength of preference, transitions between chambers and latency to emerge from the central chamber
Table 3.

Multivariate among individual correlation structure across all rounds in lateral salinity gradient for experimental treatment

Multivariate among individual correlation structure across all rounds in lateral salinity gradient for experimental treatment
Multivariate among individual correlation structure across all rounds in lateral salinity gradient for experimental treatment

Egg laying preference

When given the opportunity to lay eggs along a salinity gradient, individuals laid eggs with greater frequency in lower salinities (Fig. 4). There was no significant overall effect of treatment (χ2=0.07, P=0.79, d.f.=1) or time (χ2=0.20, P=0.66, d.f.=1), but there was a significant main effect of chamber (χ2=16.6, P=0.002, d.f.=4), which was treatment dependent (treatment×chamber: χ2=9.45, P=0.05, d.f.=4). In the control group, individuals were more likely to lay eggs in the edge chambers compared with the central chambers, as indicated by a significant chamber effect (χ2=12.82, P=0.012, d.f.=4) and a priori contrasts (Table 4); these results correspond to the edge effect that was observed in the salinity preference experiment, conducted on a separate set of individuals. Individuals in the experimental group were more likely to lay eggs in 5 ppt than in any other salinity, with a significant salinity effect (χ2=13.26, P=0.01, d.f.=4), and significant contrasts between 5 ppt and all other salinities (Table 4). Additionally, there was a significant effect of salinity on hatching (χ2=13.99, P=0.0013, d.f.=3, Fig. 5). Eggs laid in the lowest salinity had a significantly higher probability of hatching that those laid at higher salinities (5 ppt vs 15 ppt: χ2=3.59, P=0.058, d.f.=3; 5 ppt vs 25 ppt: χ2=4.14, P=0.04, d.f.=3; 5 ppt vs 35 ppt: χ2=7.60, P=0.005, d.f.=3, and 5 ppt vs all other salinities: χ2=10.40, P=0.001, d.f.=3).

Fig. 4.

The number of mangrove rivulus that laid eggs in each chamber for the control and experimental groups; some fish laid eggs in multiple chambers. (A) Control group; (B) experimental group. In the control group, each chamber contained 25 ppt. A total of 32 of 67 experimental fish laid eggs and 18 of 33 control fish laid eggs while in the gradient.

Fig. 4.

The number of mangrove rivulus that laid eggs in each chamber for the control and experimental groups; some fish laid eggs in multiple chambers. (A) Control group; (B) experimental group. In the control group, each chamber contained 25 ppt. A total of 32 of 67 experimental fish laid eggs and 18 of 33 control fish laid eggs while in the gradient.

Table 4.

Differences in the probability of laying eggs between chambers in control and salinities in experimental groups

Differences in the probability of laying eggs between chambers in control and salinities in experimental groups
Differences in the probability of laying eggs between chambers in control and salinities in experimental groups
Fig. 5.

Hatching success at each salinity in the experimental group, where mangrove rivulus had the option of laying eggs in any of the five salinities.

Fig. 5.

Hatching success at each salinity in the experimental group, where mangrove rivulus had the option of laying eggs in any of the five salinities.

The ability to exist and be phenotypically flexible in a variable environment can come with significant costs (Piersma and Drent, 2003). In aquatic habitats, that cost is often the energy devoted to osmoregulation (Boeuf and Payan, 2001). By investigating salinity preferences and repeatability of those preferences, we can gain insight into whether these traits might evolve in response to natural selection (Boake, 1989). In addition, salinity preferences provide clues into the habitats in which individuals' fitness might be highest. We initially hypothesized that mangrove rivulus would prefer to occupy salinities of 25 ppt and that their preference would be both repeatable and would align with field observations (as they are most often found at 25 ppt). Additionally, we hypothesized that mangrove rivulus would prefer to lay eggs in 25 ppt and that hatching success would be highest in 25 ppt. Our findings support the hypothesis that mangrove rivulus exhibit salinity preferences, both in terms of where they spend their time and where they lay their eggs. However, in a laboratory environment, free of anything that might constrain their movement (e.g. predators, competitors, physical factors such as temperature), mangrove rivulus preferred to occupy salinities below 25 ppt and laid eggs with greatest frequency in 5 ppt. Hatching success also was highest at 5 ppt and decreased precipitously as salinity increased. Moreover, salinity preference was not repeatable, but the strength of preference and latency to emerge were repeatable in the salinity gradient.

We investigated repeatability because the opportunity for selection to drive the evolution of behavior in environments with salinity microhabitats hinges on there being considerable variation among individuals (Boake, 1989; Wolak et al., 2012). Given that the repeatability of a trait is, arguably, the upper bound of its heritability (Boake, 1989; Falconer and Mackay, 1996; but see Dohm, 2002), it is likely that strength of preference and latency to emerge – both of which showed high repeatabilities in the experimental group – could evolve in response to selection. This might be especially true in highly heterogeneous habitats, which can reveal consistent among-individual differences in behavior. These findings are notable because environments with microhabitat options (experimental group) exposed behavioral variation among individuals that was not present in environments with only one option (control group). Given that mangrove environments are replete with microhabitat variation, we expect that such variation would be available in wild mangrove rivulus populations for natural selection to act upon. If among-individual differences are underlain by genetic variation, strong selection on exploratory behavior might drive phenotypic divergence between populations with different degrees of microhabitat variation. As a result of changes in the influx of both freshwater and saltwater to coastal systems owing to climate change, which is likely to alter microhabitat structure, mangrove forests might provide a unique opportunity to catalog the evolution of behavioral and physiological responses to changing salinity niches (Brennan et al., 2015). For example, when high spatiotemporal variation in salinity occurs, individuals that tend not to explore and have a strong preference for a specific salinity could have reduced fitness compared with those that quickly seek new microhabitats and show a relatively weak preference for a particular salinity.

Mangrove rivulus were equally active in control and experimental groups; in each treatment, individuals explored the full experimental apparatus and transitioned between chambers a similar number of times. The number of transitions was repeatable in each group, indicating consistent among-individual differences in activity levels and/or willingness to explore an unfamiliar area. These findings are consistent with Edenbrow and Croft (2012), who showed exploration within a maze to be repeatable in mangrove rivulus. However, the pattern of movement was different between control and experimental groups. While individuals in the experimental group varied in their activity levels, movement was concentrated in the lower salinities (25 ppt and below; Fig. 3). This significantly reduced among-individual variance in the number of chambers visited during a given trial and resulted in low repeatability for salinity preference in animals exposed to the salinity gradient. The control group also showed very low repeatability for chamber preference in the absence of salinity variation but likely for a different reason; in that group, the vast majority of the variance was within rather than among individuals (Table 2, Fig. 3), indicating that individuals are inconsistent in the chambers they visit most often from trial to trial.

Edenbrow and Croft (2011) also showed significant variation in the expression of behavior among ages and genotypes, indicating some context dependence. It has been previously documented that both abiotic and biotic factors (e.g. temperature, food availability, predation) can influence behavior and behavioral consistency (Nussey et al., 2007; Bell et al., 2009; Edenbrow and Croft, 2013), which was observed in our repeatability analysis for latency to emerge. The time it took mangrove rivulus to emerge from the central chamber was not repeatable in the control but was highly repeatable in the salinity gradient treatment (see also Kluen and Brommer, 2013). Strong repeatability for latency to emerge in the experimental group was due to the fact that, when faced with a salinity gradient, some fish sampled their environment quickly and others more slowly. Without the gradient (i.e. controls, all chambers 25 ppt), consistent among-individual differences in latency to emerge disappeared.

Within the experimental group, strength of preference differed consistently among individuals, which could reflect variation in the ability to flexibly adjust physiology along a salinity gradient. Some fish spent the majority of their time in one of the few low salinity chambers, while others transitioned between chambers with salinities ranging from 5 to 25 ppt. Such differences might depend on the individuals' physiology and capacity to respond to the challenges of osmoregulation in fluctuating salinity conditions. Fish rely on multiple structures (gills, gut, kidney; Edwards and Marshall, 2012) for ion and water exchange with their environment. Some individuals could be more efficient at regulating changes in ionocyte function within the gills, aquaporin expression in the intestine, or glomerular filtration rates (Edwards and Marshall, 2012; Cutler and Cramb, 2002). This opens the possibility to explore empirically how individuals with low versus high strength of preference cope with living in different salinities, and whether among-individual differences in physiological flexibility are underlain by genetic variation.

Adult mangrove rivulus are most often found at 25 ppt in the wild, but we found strong preferences for lower salinities under controlled laboratory conditions. It is also relatively rare to find eggs or hatchlings in the wild (Taylor, 1990; Taylor, 2012). While adult mangrove rivulus can clearly tolerate salinities ≥25 ppt, their preference for lower salinities indicates that they are found most often in salinities that are suboptimal, i.e. where they are likely to incur physiological costs of osmoregulation. Based on the findings of this study, our inability to find eggs in the wild is likely due to individuals selecting areas of lower salinity for egg laying and then returning to areas of higher salinity to possibly avoid predators and competitors. Mangrove rivulus will actively navigate their microhabitat options via swimming, but also have additional means of exploring their environment. Mangrove rivulus can traverse land by terrestrial tail-flip jumping and can survive out of water, as long as they are moist, for 66 days (Taylor, 1990; Pronko et al., 2013; Styga et al., 2017). With a greater ability to explore their environment via terrestrial tail-flip jumping, and having a broad tolerance to many factors that make the mangrove ecosystem a hostile environment for many fish, mangrove rivulus are able to take advantage of the many variable microhabitats available in this system (Taylor, 2012). We observed mangrove rivulus navigating to low salinities to lay eggs and these eggs had the highest hatching success at 5 ppt (Figs 4 and 5). The developmental environment can affect not only survival, but also the resulting phenotype owing to the increased energy demands of osmoregulation (Brown et al., 2012). Independent of energy requirements, developmental plasticity in response to salinity might also change the phenotype in adaptive or non-adaptive ways (West-Eberhard, 2003; Albecker and McCoy, 2019). An important area of future research might thus be to examine the extent to which exposure to different salinities early in life might drive physiological, behavioral and morphological variation.

Coordinated behavioral responses to environmental cues were evident in this study. Some among-individual correlations were expected. For example, if some individuals consistently took longer to emerge than others, they then had less time to explore the apparatus, leading to a negative among-individual correlation between latency to emerge and the number of transitions. There was also a negative among-individual correlation between number of transitions and strength of preference (negative of the variance, such that high values indicate higher strength of preference; Fig. 2); individuals that had strong preferences made fewer transitions. The positive, but not statistically significant, among-individual correlation between strength of preference and latency to emerge indicates that more ‘cautious’ individuals (those that take longer to emerge) also tend to find their preferred salinity and stay there. There was only among-individual correlation structure in the experimental treatment, and this structure remained consistent over time, perhaps representing a behavioral syndrome (Sih et al., 2004). In this context, the syndrome reflects a gradient of exploratory phenotypes. On one end are individuals that are quick to explore novel environments, actively move through the area, and exhibit weak preferences. On the other end are individuals that take a more restrained approach to novel environments, move around less and exhibit strong preferences. We used genotypes from across the mangrove rivulus' expansive geographic range, leaving two primary explanations for the behavioral variation that we observed: (i) individuals were derived from populations under divergent selection (e.g. those with and without significant microhabitat variation), and we used a representative sample of genotypes from these areas; and/or (ii) mangroves exhibit considerable spatiotemporal variation in microhabitat characteristics and/or stability such that fluctuating selection maintains behavioral variation within populations.

Some facets of our experimental treatment (whether the exposure to a salinity gradient or increased microhabitat heterogeneity) revealed consistent differences among individuals that were not present in a uniform environment (control). When phenotypic variation emerges as a result of microhabitat heterogeneity, it suggests variation among individuals in phenotypic flexibility across environments (i.e. slope of the reaction norm). If this emergent variation is heritable, there should be increased opportunity for selection to drive evolutionary change in heterogeneous environments.

In addition to present-day conditions, the evolutionary history of a species can impact how they will respond to future selection (Crowley et al., 2019). Species with a previous history of inhabiting variable environments should be able to respond to the changing environment appropriately given that habitat selection can have significant fitness consequences. When only considering osmoregulation demands, aquatic species that inhabit brackish environments should select microhabitats in which they are isotonic and where the metabolic cost of osmoregulation is minimal (Boeuf and Payan, 2001; Sutton et al., 2018). Based on our field observations indicating that mangrove rivulus are most frequently observed at salinities close to 25 ppt, and the fact that the isotonicity point of mangrove rivulus is nearer to 15 ppt (Frick and Wright, 2002; Bielmyer et al., 2012), it does not appear that mangrove rivulus are occupying ideal habitats in the wild. This could be because mangrove rivulus are excluded from lower salinity microhabitats by competitors or predators. Other factors also could impact their ability to osmoregulate efficiently (e.g. diet, temperature, dissolved oxygen) such that inhabiting an area of higher salinity may result in higher overall fitness (Hammerschlag, 2006). More work is needed to: (i) identify the possible abiotic and biotic factors that exclude mangrove rivulus from their preferred salinity in the wild; (ii) understand the physiological costs associated with occupying higher salinities; and (iii) understand whether physiological differences among genotypes might explain among-individual variation in strength of preference.

We would like to thank Sarah Blackwell, Brady Hudson, Marly Lowery, Mack Padgett, Judson Russell and Erin Yepsen for their assistance.

Author contributions

Conceptualization: S.C.M., S.K., R.L.E.; Methodology: S.C.M., S.K., R.L.E.; Validation: S.C.M., R.L.E.; Formal analysis: S.C.M., T.M.H., A.J.W., H.L.; Investigation: S.C.M.; Resources: R.L.E.; Data curation: S.C.M.; Writing - original draft: S.C.M., R.L.E.; Writing - review & editing: S.C.M., T.M.H., A.J.W., H.L., R.L.E.; Visualization: R.L.E.; Supervision: S.C.M., R.L.E.; Project administration: R.L.E.; Funding acquisition: S.C.M., R.L.E.

Funding

This study was supported by a Grants-in-Aid of Research from the Society of Integrative and Comparative Biology.

Data availability

Data are available from the Dryad digital respository (McCain et al., 2020): https://doi.org/10.5061/dryad.63xsj3v03

Albecker
,
M. A.
and
McCoy
,
M. W.
(
2019
).
Local adaptation for enhanced salt tolerance reduces non-adaptive plasticity caused by osmotic stress
.
Evolution
73
,
1941
-
1957
.
Arnold
,
S. J.
(
1994
).
Is there a unifying concept of sexual selection that applies to both plants and animals?
Am. Nat.
144
,
S1
-
S12
.
Avise
,
J. C.
and
Tatarenkov
,
A.
(
2015
).
Population genetics and evolution of the mangrove rivulus Kryptolebias marmoratus, the world's only self-fertilizing hermaphroditic vertebrate
.
J. Fish Biol.
87
,
519
-
538
.
Bates
,
D.
,
Mächler
,
M.
,
Bolker
,
B.
and
Walker
,
S.
(
2015
).
Fitting linear mixed-effects models using lme4
.
J. Stat. Softw.
67
,
1
-
48
.
Bell
,
A. M.
,
Hankison
,
S. J.
and
Laskowski
,
K. L.
(
2009
).
The repeatability of behaviour: a meta-analysis
.
Anim. Behav.
77
,
771
-
783
.
Berlinsky
,
D. L.
,
Taylor
,
J. C.
,
Howell
,
R. A.
,
Bradley
,
T. M.
and
Smith
,
T. I. J.
(
2004
).
The effects of temperature and salinity on early life stages of Black Sea bass Centropristis striata
.
J. World Aquacult. Soc.
35
,
335
-
344
.
Bielmyer
,
G. K.
,
Bullington
,
J. B.
,
DeCarlo
,
C. A.
,
Chalk
,
S. J.
and
Smith
,
K.
(
2012
).
The effects of salinity on acute toxicity of zinc to two euryhaline species of fish, Fundulus heteroclitus and Kryptolebias marmoratus
.
Integr. Comp. Biol.
52
,
753
-
760
.
Boake
,
C. R. B.
(
1989
).
Repeatability: its role in evolutionary studies of mating behavior
.
Evol. Ecol.
3
,
173
-
182
.
Boeuf
,
G.
and
Payan
,
P.
(
2001
).
How should salinity influence fish growth?
Comp. Biochem. Phys. C Toxicol. Pharmacol.
130
,
411
-
423
.
Bonte
,
D.
,
Van Dyck
,
H.
,
Bullock
,
J. M.
,
Coulon
,
A.
,
Delgado
,
M.
,
Gibbs
,
M.
,
Lehouck
,
V.
,
Matthysen
,
E.
,
Mustin
,
K.
,
Saastamoinen
,
M.
, et al. 
(
2012
).
Costs of dispersal
.
Biol. Rev.
87
,
290
-
312
.
Brennan
,
R. S.
,
Galvez
,
F.
and
Whitehead
,
A.
(
2015
).
Reciprocal osmotic challenges reveal mechanisms of divergence in phenotypic plasticity in the killifish Fundulus heteroclitus
.
J. Exp. Biol.
218
,
1212
-
1222
.
Brodie
,
E. D.
, III
and
Russel
,
N. H.
(
1999
).
The consistency of individual differences in behaviour: temperature effects on antipredator behaviour in garter snakes
.
Anim. Behav.
57
,
445
-
451
.
Brown
,
C. A.
,
Galvez
,
F.
and
Green
,
C. C.
(
2012
).
Embryonic development and metabolic costs in Gulf killifish Fundulus grandis exposed to varying environmental salinities
.
Fish Physiol. Biochem.
38
,
1071
-
1082
.
Caughley
,
G.
(
1994
).
Directions in conservation biology
.
J. Anim. Ecol.
63
,
215
-
244
.
Crowley
,
P. H.
,
Trimmer
,
P. C.
,
Spiegel
,
O.
,
Ehlman
,
S. M.
,
Cuello
,
W. S.
and
Sih
,
A.
(
2019
).
Predicting habitat choice after rapid environmental change
.
Am. Nat.
193
,
619
-
632
.
Cutler
,
C. P.
and
Cramb
,
G.
(
2002
).
Branchial expression of an aquaporin 3 (AQP-3) homologue is downregulated in the European eel Anguilla anguilla following seawater acclimation
.
J. Exp. Biol.
205
,
2643
-
2651
.
Dohm
,
M. R.
(
2002
).
Repeatability estimates do not always set an upper limit to heritability
.
Funct. Ecol.
16
,
273
-
280
.
Edenbrow
,
M.
and
Croft
,
D. P.
(
2011
).
Behavioural types and life history strategies during ontogeny in the mangrove killifish, Kryptolebias marmoratus
.
Anim. Behav.
82
,
731
-
741
.
Edenbrow
,
M.
and
Croft
,
D. P.
(
2012
).
Sequential hermaphroditism and personality in a clonal vertebrate: the mangrove killifish
.
Behav. Process.
90
,
229
-
237
.
Edenbrow
,
M.
and
Croft
,
D. P.
(
2013
).
Environmental and genetic effects shape the development of personality traits in the mangrove killifish Kryptolebias marmoratus
.
Oikos.
122
,
667
-
681
.
Edwards
,
S. L.
and
Marshall
,
W. S.
(
2012
).
Principles and patterns of osmoregulation and euryhalinity in fishes
.
Fish Physiol.
32
,
1
-
44
.
Ern
,
R.
,
Huong
,
D. T. T.
,
Cong
,
N. V.
,
Bayley
,
M.
and
Wang
,
T.
(
2014
).
Effect of salinity on oxygen consumption in fishes: a review
.
J. Fish Biol.
84
,
1210
-
1220
.
Falconer
,
D. S.
and
T. F. C.
and
Mackay
, . (
1996
).
Introduction to Quantitative Genetics
, 4th edn.
Essex
:
Longman
.
Frick
,
N. T.
and
Wright
,
P. A.
(
2002
).
Nitrogen metabolism and excretion in the mangrove killifish Rivulus marmoratus I. The influence of environmental salinity and external ammonia
.
J. Exp. Biol.
205
,
79
-
89
.
Gabriel
,
W.
(
2005
).
How stress selects for reversible phenotypic plasticity
.
J Evo. Biol.
18
,
873
-
883
.
Hammerschlag
,
N.
(
2006
).
Osmoregulation in elasmobranchs: a review for fish biologists, behaviourists and ecologists
.
Mar. Freshw. Behav. Phy.
39
,
209
-
228
.
Harrington
,
R. W.
Jr
. (
1963
).
Twenty-four-hour rhythms of internal self-fertilization and of oviposition by hermaphrodites of Rivulus marmoratus
.
Physiol. Zool.
36
,
325
-
341
.
Houslay
,
T. M.
and
Wilson
,
A. J.
(
2017
).
Avoiding the misuse of BLUP in behavioural ecology
.
Behav. Ecol.
28
,
948
-
952
.
Huber
,
J. H.
(
1992
).
Review of Rivulus Ecobiogeography Relationships: The Most Widespread Neotropical Cyprinodont Genus
.
Paris
:
Société Française d'ichtyologie
.
Kearney
,
M.
and
Porter
,
W. P.
(
2004
).
Mapping the fundamental niche: physiology, climate, and the distribution of a nocturnal lizard
.
Ecology
85
,
3119
-
3131
.
Kluen
,
E.
and
Brommer
,
J. E.
(
2013
).
Context-specific repeatability of personality traits in a wild bird: a reaction-norm perspective
.
Behav. Ecol.
24
,
650
-
658
.
Lessells
,
C. M.
and
Boag
,
P. T.
(
1987
).
Unrepeatable repeatabilities: a common mistake
.
Auk
104
,
116
-
121
.
Lin
,
H.-C.
and
Dunson
,
W. A.
(
1995
).
An explanation of the high strain diversity of a self-fertilizing hermaphroditic fish
.
Ecology
76
,
593
-
605
.
Lin
,
H.-C.
and
Dunson
,
W. A.
(
1999
).
Phenotypic plasticity in the growth of the self-fertilizing hermaphroditic fish Rivulus marmoratus
.
J. Fish Biol.
54
,
250
-
266
.
Lomax
,
J. L.
,
Carlson
,
R. E.
,
Wells
,
J. W.
,
Crawford
,
P. M.
and
Earley
,
R. L.
(
2017
).
Factors affecting egg production in the selfing mangrove rivulus (Kryptolebias marmoratus)
.
Zoology
122
,
38
-
45
.
McCain
,
S. C.
,
Kopelic
,
S.
, Houslay,
T. M.
Wilson
, A. J.,
Lu
,
H.
and
Earley
,
R. L.
(
2020
).
Choice consequences: salinity preferences and hatchling survival in the mangrove rivulus (Kryptolebias marmoratus)
.
Dryad Digital Repository
.
McManus
,
L. C.
,
Yurek
,
S.
,
Teare
,
P. B.
,
Dolan
,
T. E.
and
Serafy
,
J. E.
(
2014
).
Killifish habitat suitability as a measure of coastal restoration performance: integrating field data, behavioral trials and simulation
.
Ecol. Indic.
44
,
173
-
181
.
Mihelakakis
,
A.
and
Yoshimatsu
,
T.
(
1998
).
Effects of salinity and temperature on incubation period, hatching rate and morphogenesis of the red sea bream
.
Aquacult. Int.
6
,
171
-
177
.
Morse
,
D. H.
(
1974
).
Niche breadth as a function of social dominance
.
Am. Nat.
108
,
818
-
830
.
Nguyen
,
L. P.
,
Nol
,
E.
,
Abraham
,
K. F.
and
Lishman
,
C.
(
2013
).
Directional selection and repeatability in nest-site preferences of semipalmated plovers (Charadrius semipalmatus)
.
Can. J. Zool.
91
,
646
-
652
.
Nussey
,
D. H.
,
Wilson
,
A. J.
and
Brommer
,
J. E.
(
2007
).
The evolutionary ecology of individual phenotypic plasticity in wild populations
.
J. Evol. Biol.
20
,
831
-
844
.
Pearman
,
P. B.
,
Guisan
,
A.
,
Broennimann
,
O.
and
Randin
,
C. F.
(
2008
).
Niche dynamics in space and time
.
Trends Ecol. Evol.
23
,
149
-
158
.
Piersma
,
T.
and
Drent
,
J.
(
2003
).
Phenotypic flexibility and the evolution of organismal design
.
Trends Ecol. Evol.
18
,
228
-
233
.
Pronko
,
A. J.
,
Perlman
,
B. M.
and
Ashley-Ross
,
M. A.
(
2013
).
Launches, squiggles and pounces, oh my! The water–land transition in mangrove rivulus (Kryptolebias marmoratus)
.
J. Exp. Biol.
216
,
3988
-
3995
.
Qiu
,
J. W.
and
Qian
,
P. Y.
(
1999
).
Tolerance of the barnacle Balanus amphitrite amphitrite to salinity and temperature stress: effects of previous experience
.
Mar. Ecol. Prog. Ser.
188
,
123
-
132
.
Ramee
,
S. W.
and
Allen
,
P. J.
(
2016
).
Freshwater influences on embryos, hatching and larval survival of euryhaline Gulf killifish Fundulus grandis and potential constraints on habitat distribution: early survival of F. grandis in fresh water
.
J. Fish Biol.
89
,
1466
-
1472
.
Schulte
,
P. M.
(
2014
).
What is environmental stress? Insights from fish living in a variable environment
.
J. Exp. Biol.
217
,
23
-
34
.
Schultz
,
E. T.
and
McCormick
,
S. D.
(
2012
).
Euryhalinity in an evolutionary context
.
Fish Physiol.
32
,
477
-
533
.
Sih
,
A.
,
Bell
,
A.
and
Johnson
,
J. C.
(
2004
).
Behavioral syndromes: an ecological and evolutionary overview
.
Trends Ecol. Evol.
19
,
372
-
378
.
Staaland
,
H.
(
1969
).
A device for the study of salinity preference in mobile marine animals
.
Comp. Biochem. Phys.
29
,
853
-
857
.
Styga
,
J. M.
,
Houslay
,
T. M.
,
Wilson
,
A. J.
and
Earley
,
R. L.
(
2017
).
Ontogeny of the morphology-performance axis in an amphibious fish (Kryptolebias marmoratus)
.
J. Exp. Zool. A
327
,
620
-
634
.
Surge
,
D. M.
and
Lohmann
,
K. C.
(
2002
).
Temporal and spatial differences in salinity and water chemistry in SW Florida estuaries: effects of human-impacted watersheds
.
Estuaries
25
,
393
-
408
.
Sutton
,
A. O.
,
Turko
,
A. J.
,
McLaughlin
,
R. L.
and
Wright
,
P. A.
(
2018
).
Behavioral and physiological responses of an amphibious, euryhaline mangrove fish to acute salinity exposure
.
Copeia
106
,
305
-
311
.
Svärdson
,
G.
(
1949
).
Competition and habitat selection in birds
.
Oikos
1
,
157
-
174
.
Tatarenkov
,
A.
,
Earley
,
R. L.
,
Taylor
,
D. S.
and
Avise
,
J. C.
(
2012
).
Microevolutionary distribution of isogenicity in a self-fertilizing fish (Kryptolebias marmoratus) in the Florida Keys
.
Integr. Comp. Biol.
52
,
743
-
752
.
Taylor
,
D. S.
(
1990
).
Adaptive specializations of the cyprinodont fish Rivulus marmoratus
.
Fla. Sci.
53
,
239
-
248
.
Taylor
,
D. S.
(
2012
).
Twenty-four years in the mud: what have we learned about the natural history and ecology of the mangrove rivulus, Kryptolebias marmoratus?
Integr. Comp. Biol.
52
,
724
-
736
.
Urbina
,
M. A.
and
Glover
,
C. N.
(
2015
).
Effect of salinity on osmoregulation, metabolism and nitrogen excretion in the amphidromous fish, inanga (Galaxias maculatus)
.
J. Exp. Mar. Biol. Ecol.
473
,
7
-
15
.
West-Eberhard
,
M. J.
(
2003
).
Developmental Plasticity and Evolution
.
Oxford University Press
.
Wolak
,
M. E.
,
Fairbairn
,
D. J.
and
Paulsen
,
Y. R.
(
2012
).
Guidelines for estimating repeatability
.
Methods Ecol. Evol.
3
,
129
-
137
.
Zhang
,
G.
,
Shi
,
Y.
,
Zhu
,
Y.
,
Liu
,
J.
and
Zang
,
W.
(
2010
).
Effects of salinity on embryos and larvae of tawny puffer Takifugu flavidus
.
Aquaculture
302
,
71
-
75
.

Competing interests

The authors declare no competing or financial interests.

Supplementary information