*****
Phenotypic divergence in an island bee population: Applying geometric morphometrics to discriminate population-level variation in wing venation
Madeleine M. Ostwald*, Charles N. Thrift*, Katja C. Seltmann
Cheadle Center for Biodiversity and Ecological Restoration, University of California Santa Barbara, 93106-9615
*Equal contribution
Corresponding author: Ostwald.madeleine@gmail.com
 

 

Abstract

 
Phenotypic divergence is an important consequence of restricted gene flow in insular populations. This divergence can be challenging to detect when it occurs through subtle shifts in morphological traits, particularly in traits with complex geometries, like insect wing venation. Here, we employed geometric morphometrics to assess the extent of variation in wing venation patterns across reproductively isolated populations of the social sweat bee, Halictus tripartitus. We examined wing morphology of specimens sampled from a reproductively isolated population of H. tripartitus on Santa Cruz Island (Channel Islands, Southern California). Our analysis revealed significant differentiation in wing venation in this island population relative to conspecific mainland populations. We additionally found that this population-level variation was less pronounced than the species-level variation in wing venation among three sympatric congeners native to the region, Halictus tripartitus, Halictus ligatus, and Halictus farinosus. Together, these results provide evidence for subtle phenotypic divergence in an island bee population. More broadly, these results emphasize the utility and potential of wing morphometrics for large-scale assessment of insect population structure.
 
Keywords: wing morphology; island-mainland variation; Halictus tripartitus; California Channel Islands
 

Introduction

 
Insular conditions are major drivers of population-level phenotypic differentiation (Meröndun et al., 2019; Phillimore et al., 2008; Runemark et al., 2014; Velo-Antón & Cordero-Rivera, 2017). In particular, island populations can experience rapid evolutionary changes in morphological traits due to founder effect and subsequent genetic drift (Alsos et al., 2015; Barton, 1996; Hedrick et al., 2001; Jordan & Snell, 2008; de Souza et al., 2019; Sylvester et al., 1998; Velo-Antón et al., 2012). As such, islands have long been considered natural testbeds for evolutionary questions (MacArthur & Wilson, 1963; Warren et al., 2015), and have illuminated patterns of morphological variation, especially among birds and mammals (Cooper & Purvis, 2010; Grant, 1965; Millien, 2006). In contrast, trait variation in island populations of insects remains relatively unexplored, despite the prominence of these systems as ecologically important pollinators, pests, and invasive species (Fortuna et al., 2022; Hölldobler & Wilson, 1990; Klein et al., 2007; Traveset et al., 2013).
            Beyond these taxonomic biases, our understanding of phenotypic variation across populations is biased toward traits that are easily distinguished or quantified by human observers, such as body size and coloration (Doucet et al., 2004; Kraemer et al., 2019; Lomolino, 1985; Palkovacs, 2003). In contrast, variation in traits that present measurement challenges, such as morphological traits with complex geometries, tends to be underexplored. One such trait is the pattern of venation in insect wings. Veins provide the primary structural support for wings, and while the functional significance of variation in venation patterns remains largely unclear (Combes & Daniel, 2003), they are highly conserved in insect lineages and thus are useful in phylogenetic reconstructions and taxonomic determinations (Comstock & Needham, 1898; Sharkey & Roy, 2002). Indeed, many identifying characteristics in bee taxonomy are found in wing venation patterns, with characteristic variation distinguishing genera and species (Michener, 1994).
Within a species, however, wing venation may present subtler patterns of variation that are undetectable via traditional observation methods. Geometric morphometrics, a set of methods that allows for spatial analysis of biological forms, has emerged as a promising approach to quantifying variation in complex morphological traits (Mitteroecker & Gunz, 2009; Rohlf & Marcus, 1993). This approach has been successfully implemented to discriminate patterns of insect wing venation among (Baylac et al., 2003; Deregnaucourt et al., 2021; Francoy et al., 2009; Francoy et al., 2012; Kaba et al., 2017; Perrard et al., 2014; Rattanawannee et al., 2010, 2015; Santoso et al., 2018; Villemant et al., 2007) and even within species (Francisco et al., 2008; Francoy et al., 2011, 2016). Geometric morphometrics therefore has potential to assess the extent of phenotypic divergence among discrete insect populations by quantifying variation in this highly conserved trait.
We examined trait variation among island and mainland native bee populations in a Southern California coastal ecoregion. Santa Cruz Island is a 249 km2 Pacific island located 32 km due south of mainland Santa Barbara, California. It is the largest of the California Channel Islands, an eight-island archipelago notable for its biodiversity and endemic species and which has served as an ideal site for many microevolutionary studies of island-mainland variation (O’Reilly & Horn, 2004). Santa Cruz Island shares many of its bee fauna with mainland Santa Barbara (Seltmann, 2019), but the distance separating these locations generally precludes gene flow between populations. Bees typically forage within a few kilometers of their nesting sites, and dispersal distances are generally well under the 30 km water barrier separating Santa Cruz Island from the mainland (O’Reilly & Horn, 2004). Further, while stem- and wood-nesting bees have heightened island dispersal capabilities due to human transport of wood materials (Poulsen & Rasmussen, 2020), ground-nesting bees have limited opportunities for human-mediated island dispersal. Honey bees (Apis mellifera Linnaeus, 1758) were eradicated from the island by 2004 and have not been observed there since (Naughton et al., 2014; Seltmann, 2019; Wenner et al., 2009), suggesting that the channel is not easily crossed even by medium-sized bees. As such, we are confident that gene flow between island and mainland bee populations in this context is minimal to nonexistent, increasing the likelihood of phenotypic divergence between populations.
In this study, we investigate variation in wing venation in island and mainland populations of the sweat bee, Halictus tripartitus Cockerell, 1985. H. tripartitus is a widespread, ground-nesting social bee native to western North America and locally abundant both in mainland Santa Barbara and on Santa Cruz Island. We analyze museum specimens using a geometric morphometrics framework to assess the extent of variation in wing venation patterns between these two reproductively isolated populations. To contextualize the degree of variation, we additionally characterize variation in wing venation between H. tripartitus and two sympatric congeners, H. ligatus Say, 1837 and H. farinosus Smith, 1853. In doing so, we assess the role of reproductive isolation on population differentiation of morphological traits.

 

Methods

 

Specimens and wing imaging

 
To assess population-level variation in wing venation patterns, we imaged wings from three Halictus species: H. tripartitus (nisland = 149; nmainland = 149), H. ligatus (nmainland = 43), and H. farinosus (nisland = 3; nmainland = 40); (Figure 1). To achieve even sampling across species, we randomly selected 43 specimens of each species to analyze in our species-level comparison. We obtained bee specimens from natural history collections housed by the Cheadle Center for Biodiversity and Ecological Restoration in the University of California, Santa Barbara Invertebrate Zoology Collection. All specimens were female and were collected between 1956 and 2020, with the majority of specimens collected recently (mean: 2018, median: 2019; Figure 2); (specimen catalog numbers available in Supplementary Materials). Species-level identifications were confirmed by California native bee taxonomist Jaime Pawelek.
We removed left forewings from all specimens and imaged them with a stereo microscope digital camera along with a 1 mm scale slide (Dino-Lite AM3111T, Torrance, CA, USA; DinoXcope software 2.0.1). The basal tip of some forewings were removed if they were heavily sclerotized and prevented the wings from laying flat. We plotted 9 homologous wing venation landmarks (following Rattanawannee et al., 2015) onto each wing image using tpsDig software version 2.31 (Rohlf, 2015); (Figure 3). All analysis was conducted in R version 4.2.2.
 

Data analysis

 
We Procrustes-aligned landmark coordinates using R package ‘geomorph’ version 4.0.0 (Adams et al., 2022; Baken et al., 2021). To test for statistical differences between the two H. tripartitus populations and among the three species, we ran one-way multivariate analysis of variance (MANOVA) tests using R package ‘RRPP’ version 1.3.1 (Collyer & Adams, 2018, 2019). To visualize separation among groups, we generated density plots with discriminant analysis of principal components (DAPC) using R package ‘adegenet’ version 2.1.10 (Jombart, 2008; Jombart & Ahmed, 2011). To test the accuracy of using wing landmarks to predict an unknown bee’s species or population, we utilized DAPC cross-validation. Cross-validation also informed the number of principal components (PCs) retained in each analysis, which is non-trivial (Jombart & Collins, 2015[CT1] )
 
 

Results

 
Our analysis of wing landmark coordinates successfully discriminated between wings of H. tripartitus, H. ligatus, and H. farinosus (MANOVA: Pillai = 1.817, P < 0.001); (full MANOVA tables in Table 1; landmark coordinates in Supplementary Table 1). Based on cross-validation, 6 PCs were retained, and the density plot shows separation between species (Figure [CT2] 4a). The cross-validation test assigned 100% of Halictus specimens to their correct species (Supplementary Figure 1a).
Population-level discrimination was also successful. The two populations of H. tripartitus differed significantly in wing landmark coordinates (MANOVA: Pillai = 0.425, P < 0.001); (Table 1). Based on cross-validation, 13 PCs were retained. The density plot shows some separation between populations, with overlap (Figure 4b). The cross-validation test assigned 80.7% of H. tripartitus specimens to their correct population (Supplementary Figure 1b). This analysis accurately identified the two subgroups (populations) of H. tripartitus, although with a lower degree of separation (MANOVA, DAPC) and accuracy (cross-validation) than for the three congeneric Halictus species.
Level of Comparison df Residuals Pillai Z Pr(>Pillai)
Between Species (H. tripartitus, H. ligatus, H. farinosus) 2 126 1.817 12.925 <0.001
Within Species (Island vs. Mainland H. tripartitus) 1 296 0.425 7.715 <0.001
 
Table 1. MANOVA tables showing results of the comparison between three congeneric species of Halictus and two populations (island vs. mainland) of H. tripartitus.