Contributions to Zoology, 86 (2) – 2017Gerrit Potkamp; Mark J.A. Vermeij; Bert W. Hoeksema: Genetic and morphological variation in corallivorous snails (Coralliophila spp.) living on different host corals at Curaçao, southern Caribbean
Materials and methods

To refer to this article use this url: http://contributionstozoology.nl/vol86/nr02/a02

Morphology

Shell length of the snails was measured with a digital calliper to the nearest 0.01 mm. Of 60 snails (7% of total), only the shell length was measured with a Vernier calliper to the nearest 0.05 mm. Shell length was defined as the length from the tip of the apex to the tip of the aperture (end of the anterior canal) (Fig. 2). To determine any measurement error that could arise from inconsistencies in the orientation of the shell between the calliper blades, the shell length of 65 snails was measured in triplicate. The measurement error was defined as the average distance to the mean of the three replicate measurements of each shell. The error for shells measured with the Vernier calliper was not calculated. Shell lengths and widths were log-transformed in all statistical analysis to achieve normal distributions and homogeneity of variance. Differences in shell length were tested using ANOVA models and in some cases using Kruskall-Wallis rank sum tests.

FIG2

Fig. 2. Shell length as used to measure all shells. Twelve black points represent the 12 landmarks used in the geometric morphometric analysis.

Landmark-based geometric morphometrics were used to assess the shape of each shell. After measuring shell length, the ventral side (aperture facing upwards) was photographed with a Nikon D7000 DSLR camera equipped with a Sigma 105 mm macro lens. The locations of 12 landmarks were recorded on each photo (Fig. 2) and chosen to capture the observed variation in shell shape during shell measurements. Most of these landmarks have been used before in the morphometrics of gastropods (Zelditch et al., 2004; Hollander et al., 2006; Mariani et al., 2012). Shells covered by encrusting algae were excluded from this analysis because their landmarks were hidden. To align landmarks and remove the effect of size, a generalized Procrustes superimposition was applied to the data (Gower, 1975; Rohlf and Slice, 1990). Replication errors landmark data were calculated as well. See Online Supplementary Material 1 for the methods followed.

To statistically test for differences in shell shape and potential relationships between shell shape and shell size, snail host species or depth (as well as the interactions between host species and both shell size and depth), distance-based Procrustes ANOVA models were used that are equivalent to other distance-based ANOVA methods, like PerMANOVA (Goodall 1991; Anderson 2001; Adams and Otárola-Castillo 2013). For all Procrustes ANOVA models, significance of the different factors was tested against 10,000 permutations. Host-associated differences in shell shape were tested through pairwise comparisons of the effect of host species in a full model (with all tested factors included) against a reduced model (with all factors except host species included). To account for multiple tests, p-values were corrected using a Bonferroni correction.

To define allometric patterns, a common allometric component (CAC) was calculated from the landmark data to express allometric patterns as one variable (Mitteroecker et al., 2004). Host-specific regressions between CAC and shell length were made (excluding hosts having less than five specimens with morphometric data). The vectors of shell length of snails associated with different host species were compared to reveal differences in allometric patterns in the amount of change in shell shape per unit of growth (corresponding to the distance among vectors of shell length) and the direction of shell shape change (corresponding to the correlation among vectors of shell length). Pairwise comparisons of both the distance and correlation among the vectors of shell length were made between a full model and a reduced model without the interaction between the factors host species and shell length (see Online Supplementary Material 2). As before, the p-values were corrected with a Bonferroni correction.

To visualize variation in shell shape, a principal component analysis (PCA) was performed using the landmark data. To separate real variation in shell shape from noise resulting from the error described above and calculate repeatability of axes, landmark data of the three replicated photos was included in the PCA. Intraclass correlation coefficients (ICC, model 2,1) were calculated between the PCA-scores of triplicates on all PCA-axes. An axis was considered repeatable when the ICC was higher than 0.80 (Burridge et al., 2015). To visualize differences in allometric patterns among snails associated with different host species, linear regressions between PCA scores and shell length were used to predict PCA scores (and therefore shell shape) of shells of specific lengths associated with specific host species

After the morphometric analysis, shells were crushed to remove the snail from its shell. Using a dissecting microscope, the sex of each snail was determined by presence or absence of a penis just above the left eyestalk. Differences in sex ratios among snails associated with different host species were assessed using Fisher’s exact tests. For tests on larger tables (to test for differences among host species), p-values were computed using Monte Carlo simulations, with 1 million replications. Pairwise differences among host species were assessed using pairwise Fisher’s exact tests; p-values were corrected using a Bonferroni correction. Linear regressions were made between sex ratio and mean shell length, and between mean male and female shell length. Individual points (corresponding to a single host species) were weighed according to the number of specimens. Only host species with more than five specimens were included in the analyses comparing snails among host species.

All statistics were done in R, using the package Geomorph 2.1.5 for all morphometric analyses (Adams and Otárola-Castillo, 2013; Adams et al., 2015; R Core Team, 2015).