Morphological diversity in the sensory system of phyllostomid bats: implications for acoustic and dietary ecology
Data files
Mar 25, 2020 version files 640.55 KB
-
Acoustic-2017_raw.csv
16.24 KB
-
proc_alignear_coords.csv
103.65 KB
-
proc_alignhor_coords.csv
148.31 KB
-
proc_alignnl_coords.csv
247.02 KB
-
proc_alignsp_coords.csv
125.34 KB
Abstract
3D Imaging and shape analyses
We quantified the three-dimensional morphology of nose leaves and pinnae for 46 adult bats from 33 phyllostomids species that span the diversity in body size, nose leaf and pinnae morphologies, and dietary ecology within the family. The majority of specimens used (28 species) were collected by us in the field following approved methods (University of Washington IACUC protocol 4307-01), and the remainder (5 species) were fluid-preserved museum specimens in which the nose leaf and pinnae were preserved in their natural position (supplemental Table 1). Based on information and classifications from the literature, we grouped species into six taxonomic dietary categories (animalivores, insectivores, nectarivores, frugivores, omnivores, and sanguinivores; Giannini & Kalko, 2004), and two functional dietary categories: predators of non-mobile/non-evasive prey (nectarivores, frugivores, omnivores and sanguinivores), and predators of mobile, evasive prey (insectivores, animalivores). While assignment of species to these broadly defined dietary categories may be an over simplification of the breadth of their ecological roles (e.g., Glossophaga soricina; Clare et al., 2014), these classifications were necessary to overcome limitations due to sample sizes and the lack of quantitative dietary data that could inform more detailed analyses.
Unless the nose leaf and pinnae are adequately fixed during specimen preservation, this process can alter their shape (e.g., resulting in bent nose leaves). Furthermore, high-resolution imaging (such as µCT scanning, below) of these structures yields better results if they are scanned in isolation from denser structures like the skull. Thus, we captured pinnae and nose leaf morphology by taking casts from freshly euthanized animals. To do so, we used a President Jet dispenser gun to apply President dental molding epoxy (Epo-tek 301) to the pinnae and nose leaf (Fig. 2A). We allowed casts to dry on the specimen for a minimum of five minutes before carefully removing them. Due to limitations of field conditions and primarily using freshly collected specimens, we were not able to assess the repeatability of this technique. However, individuals of a species cluster closely together in morphospace, which indicates that this casting method is adequate for capturing interspecific variation.
To increase the size and taxonomic scope of our dataset, we were also able to use several fluid-preserved specimens that were specifically preserved to avoid deformation of soft tissues and could be destructively sampled (i.e., nose leaf could be dissected out for µCT scanning). This additional source of specimens did not seem to introduce errors in our quantification of morphology. We created 3D digital models of the nose leaf and pinnae by scanning either specimens or epoxy casts on a Skyscan 1174 µCT scanner (Bruker MicroCT, Kontich, Belgium) at a 17- 30.1 μm resolution, depending on the size of the cast or specimen. We used NRecon (Microphotonics, Allentown, PA) to convert CT shadow images into image stacks (“slices”), and imported these into Mimics 17.0 (Materialise NV, Leuven, Belgium, 2014) to segment nose leaf and pinnae and produce 3D surface (*.stl) files (Fig. 2B). We imported raw stl files into Geomagic Studio 2014.1.0 (3D Systems, SC, USA, 2014) to remove scanning artifacts (e.g., debris in molds) from the models.
To quantify nose leaf and pinnae shape, we used 3D geometric morphometric analyses (Bookstein, 1997; Zelditch, Swiderski, Sheets, & Fink, 2004). These were based on single point landmarks and surface patches, all placed on 3D models using Stratovan Checkpoint© (Stratovan Corporation, Davis, CA). For the nose leaf, we placed: (1) single-point landmarks at the base of each nostril and the apex of the spear, (2) evenly-spaced semi-landmarks around the nose leaf perimeter, and (3) two “patches” of semi-landmarks in a grid across the surface of the spear and the surface of the horseshoe, respectively (Fig. 2D). To analyze shape changes of subcomponents of the nose leaf separately (i.e., spear and horseshoe), we added landmarks to ensure each subcomponent had a sufficient number of true landmarks. For the spear, we placed a single-point landmark at the apex of the spear, two landmarks at the point where the spear meets the horseshoe, and a patch of semi-landmarks over the anterior surface of the spear (Fig. 2D). Some species lack a spear, and therefore were not included in analyses of that structure. For the horseshoe, we placed a single-point landmark on each nostril and one patch over the surface of horseshoe.
For pinnae, we placed two landmarks at the points where the pinna attaches to the head, and a patch of semi-landmarks across its surface (Fig. 2C). We exported landmark coordinates for each specimen as .csv files and computed species means for landmark coordinates in Excel. We then performed Procrustes superimposition analyses to scale, align and rotate landmark configurations (Rohlf, 1990), and obtain a set of variables describing the shape of the entire nose leaf, spear, horseshoe and pinnae across species. We used the package “geomorph” (Adams & Otárola-Castillo, 2013) in R v 99.903 (R Core Team, 2017) for geometric morphometric analyses.
Acoustics
Phyllostomid bats produce low-intensity calls (Brinkløv, Kalko, & Surlykke, 2009; Griffin, 1958) that are difficult to capture on passive recording devices. Consequently, call parameter data are sparse for most phyllostomid species. For this study, we collected 16-bit recordings of release calls using a microphone condenser (UltraSoundGate 116). Our sample included 101 individuals spanning 33 species. We held each bat in hand, placed a microphone approximately six inches from its face, and then released the bat away from environmental clutter while recording the calls it emitted as it flew away. Since bats had to be released to document their natural calls, we did not use these same individuals in morphological analyses. We measured call parameters for 3–7 individuals per species, with the exception of species that were rare or difficult to capture at our study localities (Chrotopterus auritus, Glyphonycteris sylvestris, Phyllostomus hastatus, Sturnira lilium), for which we were able to record 1 individual per species. We analyzed release calls using Avisoft SASLabPro v. 5.2.12 (Avisoft Bioacoustics, Berlin, Germany) to extract the following echolocation call parameters: minimum frequency (kHz), maximum frequency (kHz), peak frequency (kHz) (i.e., frequency with the highest amplitude), and total bandwidth (kHz) across the call. We averaged call sequences per individual (a minimum of 5) and calculated means and standard deviation of each parameter (supplemental Table 2). While release calls may not fully reflect the echolocation capabilities of the species, our own comparisons of release calls with foraging calls for one species (Carollia castanea) indicate that foraging call parameters fall well within the range of values recorded for release calls (Leiser-Miller, in press).
Statistical analyses
To test whether the nose leaf consists of two modules (spear and horseshoe; Fig. 2D), we used the function phylo.modularity (Geomorph package, Adams & Otárola-Castillo, 2013) to compute Covariance Ratio (CR) values for a two-module hypothesis based on the nose leaf landmark data, and estimate the p-value for this relationship over 1,000 random permutations. The CR ratio indicates the degree of covariation among landmarks within possible modules; values from 0 to 1 indicate less covariation between modules than within each module, supporting the modularity hypothesis, CR values greater than 1 describe greater covariation between modules than within modules, supporting the null hypothesis of no modules (Adams, 2016).
To identify major axes of shape variation across sensory structures, we conducted phylogenetic Principal Component Analyses (pPCA), using the Rojas, Warsi, & Dávalos (2016) phylogeny, on the Procrustes (shape) coordinates for each structure/substructure using the R package “phytools” (Revell, 2012). We assessed the significance of pPCA axes via Horn’s parallel analysis from the ‘paran’ function in R (Dinno, 2015). Nose leaf and pinna shape axes were not correlated with size (forearm length; Supplementary Table 5), and therefore size was not considered in subsequent analyses. To identify if shapes of external sensory structures are correlated with call parameters, we ran separate phylogenetic generalized least squares (PGLS) regressions under Brownian motion of acoustic parameters across the call (minimum frequency, maximum frequency, peak frequency, and total bandwidth) against significant pPCs shape scores (see Results; nose leaf pPC 1-5; pinnae pPC 1-4). Finally, we ran phylogenetic ANOVAs and post-hoc analyses to test for an association between diet category and nose leaf and pinnae shape, respectively. We used significant pPCs axes as response variables, and dietary category as the predicting factor.