Electronic Supplemental Material
Article title: Divergent responses of invaded and un-invaded populations of Pacific tree frog tadpoles to invasive predatory crayfish
Journal name: Oecologia
Author names: Katherine M. Pease and Robert K. Wayne
Corresponding author: Katherine M. Pease, Department of Ecology and Evolutionary Biology, University of California, Los Angeles, CA, 90095, USA;
Methods
Tadpole Morphology
Photographs
To better visualize the tail fin, we altered the lateral images by creating a “droplet” (automated action for batch processing) that set the image curves to specific levels, altering tones of the photograph. We created a tonal curve with four anchor points set at 1) 0, 0, 2) 78, 207, 3) 134, 74, 4) 241, 44. The tail fin below the tail muscle was still difficult to distinguish so was not used in morphological analyses.
Methods for semilandmark placement
We used MakeFan 6 (Sheets 2003) to place markers at the tip of the snout and tip of the tail and then made a comb between the markers. We used the alternative control screen (“AltComb3”) in MakeFan to make a comb with 40 lines evenly spaced between the two markers where the last 8 lines near the head were drawn as a fan with 20 radial lines. The comb and fan lines provided objective locations along the tadpole where semilandmarks could be placed.
Conventional Morphometrics
We measured a total of ten lateral traits – total length, body length, body depth, tail length, tail fin depth at three locations, and tail muscle depth at 3 locations (Figure 1). To adjust for differences in body size, we calculated a ratio of each measurement to total body length, resulting in nine traits for comparison between tadpoles groups. To examine differences between tadpole groups, we performed a MANCOVA on all ratio measurements followed by univariate tests (ANCOVA), correcting for multiple tests with a sequential Bonferroni (Rice 1989). Stream was included as a covariate in all tests. Because the use of proportions for size correction has been criticized (Arndt et al. 1991), we also adjusted for size differences by performing a PCA on all measurements to obtain the score from the first PC axis, which was used as a measure of overall size. We then regressed all measurements against the PC-1 scores and the residuals were used for further analysis as relative measurements. We used the residuals in a MANCOVA (stream as covariate), with all morphological measurements being compared between tadpoles from streams with and without crayfish. We then performed univariate analyses (ANCOVAs, stream covariate) on each morphological trait to determine significance, correcting for multiple tests. We tested for normality of all conventional traits and ratios and performed transformations if necessary.
Ventral Morphology
The total number of ventral photographs analyzed was 205 images (N=79 from streams without crayfish and N=126 from streams with crayfish). On the ventral image, landmarks 1-3 are traditional landmarks and 4-43 are semilandmarks (Figure S-2). Landmark 1 is the tip of the snout, landmark 2 is where the tail muscle meets the body, and landmark 3 is the tip of the tail. Landmarks were only placed on one half of the ventral side of the tadpoles because the shape is symmetric in this orientation. Placing landmarks on both sides would produce data points that are not independent from each other and would reduce the degrees of freedom (Zelditch et al. 2004).
We used TpsDig2 to measure the following conventional traits on the digital images (Figure S-2): ventral measurements –total length, body length, body width, tail length, tail width 1 (near the body), tail width 2 (mid-tail), tail width 3(end-tail).
Transformation of conventional traits
We transformed the following lateral conventional traits with a log transformation: body depth, tail muscle depth 1, tail muscle depth 2. We also transformed the following ratio (size-corrected) traits with a log transformation: body depth, tail muscle depth 1, tail muscle depth 3. We transformed tail muscle depth 3 with a square root transformation.
We transformed the following ventral conventional traits with a log transformation: tail width 1, tail width 2, tail width 3. We also transformed the following ratio (size-corrected) traits with a log transformation: tail width 2, tail width 3. We transformed tail width 1 with an inverse transformation.
Predation Methods
Morphology
We examined ventral morphology with geometric and conventional morphometrics as previously described. We transformed the following ventral variables with a log transformation: tail width 1, tail width 2, tail width 3, tail width 2 ratio, and tail width 3 ratio. For the size-corrected ratio calculation of tail width 1, we use the log-transformed values of tail width 1 divided by the untransformed values for total length. A total of 93 ventral images were analyzed, comprised of 33 tadpoles that survived, 18 that survived and were injured, and 42 that were killed and eaten.
We transformed the following lateral variables with a log transformation for normality: body depth, tail muscle depth 1, tail muscle depth 2, tail muscle depth 3, tail fin depth 3, tail muscle depth 3 ratio, and tail fin depth 3 ratio.
Selection
The best model for ventral traits included the traits of body width and tail width 2 (width near the middle of the tail). We analyzed lateral and ventral traits separately in order to increase the sample size in each analysis. Due to bad photographs, we had to discard certain samples from the analysis and thus the samples that were in the lateral analysis were not exactly the same samples that were in the ventral analysis. Therefore, we performed separate regressions on the traits to measure selection.
Results
Morphology
Conventional Morphometrics: Lateral
We found deeper tail fins and muscles in tadpoles from streams with no crayfish and deeper bodies in tadpoles from streams with crayfish (Table S-1; Figure S-3). We found a significant difference in all nine conventional morphometric traits between tadpoles from streams with and without crayfish, including stream as a covariate (MANCOVA: F9, 183 =7.05, p =<0.001). Examining morphometric ratios individually revealed that six traits were significantly different between tadpoles from streams with crayfish and tadpoles from streams without crayfish, including stream as a covariate (Table S-1); the six traits were body depth, tail fin depths 1, 2, 3, and tail muscle depths 2 and 3. After correcting for multiple tests, five traits (tail fin depth 1, 2 and 3; tail muscle depth 2 and 3) remained significantly different between the two groups (Table S-1; Figure S-3).
Adjusting for body size with a PCA and regression of PC1 scores against measurements showed similar results. The PCA was performed with ten lateral measurements, including total length, and the first PC explained 96% of the variation. When examined separately, five traits showed a significant difference between tadpoles from streams without crayfish and tadpoles from streams with crayfish after correcting for multiple tests. The five traits are the same as those in Table S-1 that were found to be significant after Bonferroni corrections (tail muscle depth 2; tail muscle depth 3; tail fin depth 1; tail fin depth 2; tail fin depth 3). Overall, the shape differences found with conventional morphometrics were similar to those found with geometric morphometric techniques. Both analyses showed that tadpoles from crayfish streams have shallower tail fins, shallower tail muscles, and deeper bodies compared to tadpoles from crayfish-free streams.
Geometric Morphometrics: Ventral
We found significant ventral shape differences between tadpoles from streams with and without crayfish in the canonical variates analysis. However, there was not a clear pattern of ventral shape change between the two groups and the shape changes were very minor even when the deformation was exaggerated by 100 fold in CVAGen6o (Figure S-4). The CVA revealed one distinct axis of differentiation between tadpoles from streams with and without crayfish (Wilk’s lambda = 0.52, d.f. =2, p= 0.025). Ventral shape differences were mainly in the uniform or affine direction. Shape deformation in this direction can also be described as shear in which parallel lines remain parallel with overall shape changes. This finding indicates that the shape differences are uniformly distributed throughout the tadpole body and there is little localized shape change within the tadpole.
We also found significant shape differences between the two groups when examining ventral partial warps between tadpoles from streams with and without crayfish, including stream as a covariate (F82, 118 =1.71, p=0.004. There was still a significant difference between the groups when centroid size was also included as a covariate (F82, 117= 1.73, p=0.003), indicating that size is not solely responsible for the shape differences between groups.
We detected ventral shape changes with centroid size in TpsRegr; the amount of shape change explained by centroid size was 28.88%. However, we did not find any difference in centroid size between the two groups of tadpoles, including stream as a covariate (ANCOVA: F1,199=0.030, p=0.467). This finding indicates that centroid size or allometry is not driving the shape differences that we found between the two groups of tadpoles.
Conventional Morphometrics: Ventral
We found wider tail muscles in tadpoles from streams without crayfish compared to tadpoles from streams with crayfish (Table S-1; Figure S-5). We found a significant difference in all six conventional morphometric traits between tadpoles from streams with and without crayfish, including stream as a covariate (MANCOVA: F6, 199=2.38, p= 0.030). Examining morphometric ratios individually with ANCOVAs, including stream as a covariate, revealed that three traits were significantly different between tadpoles from streams with crayfish and tadpoles from streams without crayfish (Table S-1); the three traits were tail widths 1, 2, and 3. After correcting for multiple tests, one trait (tail width 3) remained significantly different between the two groups (Table S-1; Figure S-5).
Adjusting for body size with a PCA and regression of PC1 scores against measurements showed similar results. The PCA was performed with seven measurements, including total length, and the first PC explained 93% of the variation. The MANCOVA (stream as covariate) with all six relative measurements between tadpole groups was significant (F6, 194=2.62, p=0.018). When examined separately as ANCOVAs (stream as covariate) and correcting for multiple tests, relative tail width 3 showed a significant difference between tadpoles from streams without crayfish and tadpoles from streams with crayfish (ANCOVA: F1, 199=8.57, p-=0.004). In summary, we found significant but not visually obvious ventral shape differences with geometric morphometrics between tadpoles from streams with and without crayfish. Using conventional morphometrics, we also found significant differences in the ventral shape of tadpoles from streams with and without crayfish. Tadpoles from crayfish streams have narrower tail muscles compared to tadpoles from crayfish-free streams.
Predation Experiments
Conventional Morphometrics: Lateral
Tadpoles that survived predation had relatively shallower tail fins near the body and relatively deeper tail muscles throughout the tail (Table S-2; Figure S-6). We found significant differences in all nine morphometric ratios between the two groups of tadpoles that survived predation and those that were killed or injured (MANCOVA: F9, 77=3.41, p=0.001). Examining ratios individually as ANCOVAs (stream as covariate), revealed three traits (tail fin depth 1, tail muscle depth 1, and tail muscle depth 2) that were significantly different between tadpoles that survived predation and those that were killed or injured, after correcting for multiple tests (Table S-2; Figure S-6).
Adjusting for body size by taking the residuals of measurements from a regression with the first axis of a PCA showed similar results. The PCA was performed with ten lateral measurements, including total length, and the first PC explained 93% of the variation. When examined separately as ANCOVAs (stream as covariate) and correcting for multiple testing, three traits were significantly different between tadpoles that survived predation and those that were killed or injured (relative tail muscle depth 1, relative tail fin depth 1, and relative tail muscle depth 2). In summary, both geometric and conventional morphometrics revealed significant lateral shape differences between tadpoles that survived predation and those that were injured or killed. Tadpoles that survived predation had shallower tail fins near the body and deeper tail muscles.
Geometric Morphometrics: Predation Experiments
Ventral
We did not find significant ventral shape differences between tadpoles that survived predation and those that were injured or killed. There was no significant axis of differentiation between tadpoles that survived and those that were injured or killed (CVA: p=0.060). We found no significant shape differences between tadpoles that survived predation and those that were injured or killed when we examined partial warps (MANCOVA (stream as covariate): F82, 5 =0.63, p= 0.823). When centroid size was also included as a covariate, there was still no significant difference (MANCOVA: F82, 4=1.02 p=0.575).
Conventional Morphometrics: Predation Experiments
Ventral
We did not find any significant differences in ventral conventional traits between tadpoles that survived and those that were injured or killed (MANCOVA (stream as covariate): F7, 80=0.94, p=0.484). Because no significant differences were found with the multivariate test, we did not perform univariate tests. Adjusting for body size with a PCA and regression of PC1 scored against measurements also showed no significant differences among the tadpoles from the predation experiments. In summary, we did not find any ventral shape differences using geometric and conventional morphometrics for tadpoles from the predation experiments, comparing tadpoles that survived to those that were injured or killed.
Selection: Ventral morphology
For ventral traits, we did not detect any significant selection (Table S-3). However, the cubic splines show a pattern of positive directional selection for tail width 2 (Figure S-7) and slight negative directional selection and stabilizing selection on body width (Figure S-7).
References
Arndt, S., G. Cohen, R.J. Alliger, V.W. Swayze, and N.C. Andreasen. 1991. Problems
with ratio and proportion measures of imaged cerebral structures. Psychiatry Research: Neuroimaging 40(1): 79-89.
Rice, W.R. 1989. Analyzing tables of statistical tests. Evolution 43: 223–225.
Sheets, H.D. 2003. IMP-Integrated Morphometrics Package. Department of Physics,
Canisius College, Buffalo, NY.
Zelditch, M.L., D.L. Swiderski, H.D. Sheets, and W.L. Fink. 2004. Geometric