Larval settlement and metamorphosis in a marine gastropod in response to multiple conspecific cues

Crepidula fornicata collection and husbandry

We collected adult Crepidula fornicata from Crab Meadow Beach (Northport, New York, USA: 40°55′46′′N, 73°19′8′′W) in July 2013 and returned them to the lab the same day. No permit is required for collecting this species in New York State, and animals were not returned to the field. Adult females were removed from their substrates to check for incubating egg capsules. Capsules with larvae that were ready to hatch (stage IV veligers sensu Leroy et al., 2013) were selected and hatched by physically agitating them in a bowl of filtered seawater at room temperature. Larvae from three females were combined for rearing in cultures of 800 ml of 1 µm-filtered seawater (FSW) at a concentration of one larva per four ml. We collected seawater from an underground well at Flax Pond Marine Laboratories, Old Field, New York, USA (40°57′49′′N, 73°08′26′′W). We fed larvae 40,000 cells/ml of the alga Isochrysis galbana (clone T-ISO) daily. We maintained larval cultures at 20 ° C and replaced FSW via reverse filtration every three to four days. We tested for larval competence (ability to metamorphose) every two days once the larvae developed shell brims (Pechenik, 1984) and were at least 750 µm long (Pechenik & Heyman, 1987). Competence was tested by placing 12–24 larvae (1–2 larvae from each culture) in 20 mM KCl solution for 8 h. It is not possible to work with a larval culture that has reached 100% competence, because at this point many larvae have spontaneously metamorphosed and are no longer available for experiments. We therefore started the experiment within 24 h of the larvae reaching 75% competence (75% of larvae metamorphosed in response to KCl; Pechenik & Heyman, 1987), which occurred 19 days post-hatch.

Preparation of test solutions

We tested three factors for induction of larval settlement: seawater conditioned with conspecific adults, conspecific pedal mucus, and 20 mM elevated KCl. Prior to the start of the experiment, we acid-cleaned all glassware for ten minutes in 10% concentrated HCl, rinsed them in deionized water, and then autoclaved them.

To create adult-conditioned seawater (ACW), we placed 100 g of adult C. fornicata (shell and wet tissue mass) and one liter of FSW into a beaker, then oxygenated the water for twelve hours. Large epibionts (e.g., barnacles, macroalgae) were removed from the shells, but shells were not otherwise treated. One liter of FSW without C. fornicata, to be used as a control, was also oxygenated. After twelve hours, adults were removed and ACW and FSW were filtered to 40 µm with a Nitex mesh filter. Temperature, dissolved oxygen, salinity, and pH were measured at all preparation steps. The biotic and abiotic parameters for the bioassay (salinity, trial length, mass of adults used for cue preparation, etc.) are listed in Table 1. These values were chosen based on optimization studies for each parameter where settlement was measured over a range of values. The optimal value was chosen based on settlement rates. When multiple values gave similar settlement rates, we made our choice based on logistical concerns (details in Fig. S1).

We prepared pedal mucus treatments (replicates with pedal mucus left by adult snails; PMG) at the same time that ACW was prepared. Individual replicates of the experiment were conducted in 60 ml drinking glasses (shot glasses or shooters; hereafter “glasses”). For each PMG replicate, a single small (∼15 mm) adult C. fornicata was added to all glasses for 12 h at the same time that the ACW was prepared. Glasses were filled with 35 ml FSW and covered to prevent evaporation and snail escape. All snails began the 12 h period at the bottom of the glass, and so even though some crawled above the water line during this period, they still left mucus footprints in the glasses. During the ACW preparation, glasses not receiving the PMG treatment had adults removed and were acid-cleaned and autoclaved to remove the mucus. To prevent desiccation of the mucus, we did not remove adults from glasses receiving the PMG treatment until immediately before the start of the experiment; these glasses were then drained.

For KCl treatments, we prepared a concentrated solution of 200 mM KCl in distilled water, which was stored until use in the experiments (following Pechenik & Heyman, 1987).

Bioassay

Because larvae are expected to become more likely to metamorphose as they develop, we used a randomized complete block design to account for larval age, using the start date of the experiment as a blocking factor (i.e., blocks were run through time). All blocks used the same batch of larvae and therefore larvae in the later blocks were older. All blocks were run within a ten-day period. All three factors (ACW, PMG, KCl) had two levels (present or absent) and were tested using a factorial design (eight possible treatments; Fig. 1). The treatment combination where all factors were absent corresponded to FSW and served as a negative control. Each treatment combination had three replicate glasses for 24 replicates per block (72 replicates total).

Each glass in the bioassay contained 20 ml of the test solution (Fig. 1). To make the KCl treatments, we added 2 ml of the concentrated KCl solution to the test solution of the replicate (either FSW or ACW; Fig. 1) for a final KCl concentration of 20 mM elevated above FSW. Ten larvae were then individually pipetted into each glass. Larval growth and development in many marine larvae, including C. fornicata, varies among cultures (rearing beakers). This variation was accounted for by placing one larva from each rearing beaker (ten total beakers) into each replicate glass. The same set of rearing beakers was used in all three blocks of the experiment.

Every 12 h, we counted the number of larvae metamorphosed in each replicate, and recorded any mortality. Metamorphosis was measured by the loss of the velar lobes, meaning that it was an irreversible step in development. We removed metamorphosed juveniles and dead larvae from the trial at each time point. To limit bacterial growth and the buildup of waste products, after 24 h we replaced the test solution in the glasses with clean glasses containing new ACW, KCl, and PMG prepared as described above, and individually pipetted larvae into the clean glasses. The total time of the experiment was 48 h, which included five time points and two different preparations of test solutions. Larvae were not fed during the experiment.

Data analysis

We conducted an analysis based on settlement rates rather than the proportions of larvae settled at a fixed time point, since proportions are sensitive to the underlying mathematical function of settlement rate and the timepoint selected for the analysis. We modeled larval settlement by predicting the proportion of larvae settled (y) at each timepoint (t, in hours) using the cumulative distribution function for the single-parameter exponential model: ${\displaystyle y={1-e}^{-\lambda t}.}$ Given a constant probability of settlement, the waiting times for a single individual to settle (t_i) in a given treatment are exponentially distributed. Note that this model assumes that all larvae in the experiment are developmentally capable of settling (competent), although trials began when only 75% of larvae were competent; it was not possible to wait for 100% competence due to high rates of spontaneous settlement under these conditions (see above). Incomplete competence will not affect the overall results if competence is equal in all treatments, a reasonable assumption given our random assignment of larvae to treatments. The exponential distribution is defined by the single parameter λ, which can be estimated as ${\displaystyle \hat{\lambda }=n/\sum _{i=1}^n{t}_i.}$ However, because not all larvae settled during the first 48 h, we calculated λ incorporating Type I censoring with the following equation: ${\displaystyle \hat{\lambda }=\frac{r}{\sum _{i=1}^r{t}_i+T\left(n-r\right)},}$ where n is the total number of larvae tested and r is the number of larvae that settle during the bioassay. Thus, (n − r) is the number of non-metamorphosed larvae at time T which represents the end of the bioassay (fixed at 48 h for all blocks). By modeling larval settlement in this way, the settlement rate for each replicate could be summarized with a single value ($\hat{\lambda }$) that used time-course data and also accounted for Type I censoring.

To analyze the experimental data, we corrected the number of larvae tested (n) for mortality (average mortality per block = 2%, or approximately five larvae; mortality was consistent among treatments) and then calculated $\hat{\lambda }$ for each replicate. We used data from the first block of the experiment to calculate the correlation between the predicted number of larvae settled for each replicate at each timepoint (based on λ) and the observed number of larvae settled.

The full experiment was then analyzed using $\hat{\lambda }$ as a response variable in an analysis of variance. Blocks were treated as random effects and experimental treatment factors (ACW, PMG, KCl) were treated as fixed effects. The treatment where all factors were absent was equivalent to FSW and served as a control (Fig. 1). Planned comparisons were conducted of each factor against this control (H₀: KCl = FSW, ACW = FSW, and PMG = FSW). The significance of planned comparisons was assessed using the 95% confidence intervals of the mean difference between the treatments. Confidence intervals that did not overlap with zero indicated a significant difference between the treatment and the control (Motulsky, 2010).

Finally, in order to compare our results using rates to results that would be obtained by using proportions, we calculated an ANOVA on the arcsine-squareroot transformed proportions of larvae settled at each timepoint in the analysis (four separate ANOVAs). Statistics were conducted using JMPIN (Version 4.0.4, 2001; SAS Institute, Cary, NC, USA) and R 3.0.1 (R Core Team, 2013).

Modeling settlement rates

The proportion of larvae expected to settle over successive twelve-hour intervals was predicted by modeling settlement with the rate parameter λ from the exponential distribution. The fit of the estimated values of λ ($\hat{\lambda }$) to the observed cumulative proportions of larvae settled is illustrated with data from the first block of the experiment (Fig. 2). The correlation of predicted and observed values was high (overall r = 0.930, p < 0.001) and consistent across treatments (Fig. 2A). The slope of the best-fit line of predicted and observed data was less than one (slope = 0.810 ± 0.030 SE), indicating that the model slightly underpredicted at most time points (Fig. 2B). Modeling settlement as the rate parameter λ was a more informative statistical analysis than using proportions of larvae settled at a given time. Performing the analysis on arcsine-transformed proportions using ANOVAs yielded inconsistent results, such that the significance of both main effects and interaction terms depended on the time point selected for the analysis (Table 2, Table S1).

Larval settlement rates

All factors showed an increased settlement rate (λ) relative to the filtered seawater (FSW) control (Figs. 3 and 4). The linear model contained two statistically significant treatment effects (KCl, F_1,62 = 19.43, p < 0.001; KCl*ACW, F_1,62 = 14.57, p < 0.001; Table 3), with block effects through time accounting for 24.6% of total variation. The strength of the artificial cue was likely responsible for significant interaction effects (KCl*ACW, Table 3), as complete induction by KCl allowed for no additional effect of the ACW treatment.

The mean settlement rates (λ, in units of ln(number larvae settled per hour)) for KCl, ACW, PMG, and FSW were 0.01796, 0.01218 0.007234, and 0.000715, respectively. The 95% confidence intervals of the difference in mean settlement rates between each treatment factor and the FSW control did not overlap with zero (difference in λ between KCl − FSW = 0.01075 − 0.02374; ACW − FSW = 0.005248 − 0.017572; PMG − FSW = 0.002583 − 0.010456). This indicates that all factors induced settlement in C. fornicata (Fig. 4).