DNA barcoding to identify leaf preference of leafcutting bees

Leafcutting bees (Megachile: Megachilidae) cut leaves from various trees, shrubs, wildflowers and grasses to partition and encase brood cells in hollow plant stems, decaying logs or in the ground. The identification of preferred plant species via morphological characters of the leaf fragments is challenging and direct observation of bees cutting leaves from certain plant species are difficult. As such, data are poor on leaf preference of leafcutting bees. In this study, I use DNA barcoding of the rcbL and ITS2 regions to identify and compare leaf preference of three Megachile bee species widespread in Toronto, Canada. Nests were opened and one leaf piece from one cell per nest of the native M. pugnata Say (N=45 leaf pieces), and the introduced M. rotundata Fabricius (N=64) and M. centuncularis (L.) (N=65) were analysed. From 174 individual DNA sequences, 54 plant species were identified. Preference by M. rotundata was most diverse (36 leaf species, H′=3.08, phylogenetic diversity (pd)=2.97), followed by M. centuncularis (23 species, H′=2.38, pd=1.51) then M. pugnata (18 species, H′=1.87, pd=1.22). Cluster analysis revealed significant overlap in leaf choice of M. rotundata and M. centuncularis. There was no significant preference for native leaves, and only M. centuncularis showed preference for leaves of woody plants over perennials. Interestingly, antimicrobial properties were present in all but six plants collected; all these were exotic plants and none were collected by the native bee, M. pugnata. These missing details in interpreting what bees need offers valuable information for conservation by accounting for necessary (and potentially limiting) nesting materials.

JSM, 0000-0002-2443-8192 Leafcutting bees (Megachile: Megachilidae) cut leaves from various trees, shrubs, wildflowers and grasses to partition and encase brood cells in hollow plant stems, decaying logs or in the ground. The identification of preferred plant species via morphological characters of the leaf fragments is challenging and direct observation of bees cutting leaves from certain plant species are difficult. As such, data are poor on leaf preference of leafcutting bees. In this study, I use DNA barcoding of the rcbL and ITS2 regions to identify and compare leaf preference of three Megachile bee species widespread in Toronto, Canada. Nests were opened and one leaf piece from one cell per nest of the native M. pugnata Say There was no significant preference for native leaves, and only M. centuncularis showed preference for leaves of woody plants over perennials. Interestingly, antimicrobial properties were present in all but six plants collected; all these were exotic plants and none were collected by the native bee, M. pugnata. These missing details in interpreting what bees need offers valuable information for conservation by accounting for necessary (and potentially limiting) nesting materials.

Background
The Megachilidae is the second largest bee family with over 3900 species and a worldwide distribution [1,2]. These bees

Sample collection
To sample leaves from nests I set up artificial nests (nest-boxes), which are a preferred habitat for many cavity-nesting bee species [8,17,44,45]. One nest-box was set up at each of 200 locations throughout the city of Toronto in residential gardens, community gardens, public parks and on green roofs, each year from 2011-2013, approximately 250 m apart in distance [46]. Each nest-box was made of a 30 cm piece of PVC pipe that was 10 cm in diameter into which 30 cardboard nesting tubes (10 of each of three widths: 7.6 mm, 5.5 mm, 3.4 mm) were inserted [47]. These were attached to fixed structures at each site (e.g. wooden stake, fence post, exposed tree limb). Nest-boxes were set up in April and retrieved in October. The nesting tubes were opened and all brood were stored in a walk-in fridge at 4 • C for the winter. In spring, all individual brood cells were placed in a growth chamber at 26 • C and 65% humidity so that they could be identified to species level after emergence. Thirty-six bee species (including five cleptoparasites) used the nest-boxes over the three seasons [46]. From this group, three common leafcutting species were selected for study, the native M. pugnata, and the introduced M. centuncularis and M. rotundata. M. pugnata uses mud and chewed leaves to line its brood cells, and makes partitions between adjacent cells using circular pieces of leaves laid one over the other (   To determine the identity of each leaf cut sample, using forceps I removed one leaf piece per brood cell, then each was cut into a 0.5 cm 2 square, washed in ethanol, air dried, and placed into an individual sampling vial in a 96-vial DNA extraction plate (AcroPrep) provided by the Canadian Centre for DNA Barcoding (CCDB). Two filled plates of leaf fragments were then sent to the CCDB for DNA extraction [48] and barcoding [37,49,50]. The primers rcbL and ITS2 were used to obtain nucleotide sequences for each sample. Identities of the leaf samples were determined by downloading sequence data (greater than 50 bp) from the Barcode of Life Data (BOLD) systems. A total of 47.4% of samples yielded incomplete sequence data (rcbL = 11.9% of sample, ITS2 = 35.6%). However, sequence quality was high (rbcL = 96.7% (<1% Ns); ITS2 = 93.75% (<1% Ns)) and sequence data were cross-referenced in the gene sequence database GenBank [51] for 100% matches using the Basic Local Alignment Search Tool (BLAST) [52]. Lastly, once the identities of the leaves were determined, the antimicrobial properties of each of the plant species were determined by surveying the literature that measures these values (table 1) [53-60].

Analysis
From the type and number of leaf species cut as determined by DNA barcoding, I determined the richness and Shannon diversity index (H ) of leaf species preference for each of the three bee species. Rarefaction curves for leaf preference of all bee species were interpreted using iNEXT software [61,62]. Using the package 'ecodist' [63] in the R statistical program [64], I calculated the Bray-Curtis dissimilarity index to compare how leaf preference differed between the leafcutting bee species. I then used a Spearman's rank correlation to determine similarity in leaf preference between each bee species pair. Bee species pairs that were positively correlated meant they overlapped in leaf preference. The leaf preferences of the three bee species were also compared by plant species grouped by status ('exotic' or 'native') and type ('tree', 'shrub', 'perennial', 'annual') using a set of Pearson's χ 2 -tests in the R package 'MASS'. Some plant species used by the bees were biennials (N = 3, see table 1) and these were included as perennials in the analysis. Another χ 2 -test was used to compare preference with plant species grouped as 'woody' ('tree' + 'shrub') or 'non-woody' ('perennial' + 'biennial' + 'annual').     To compare pd in leaf preference by each of the three bee species, I constructed a phylogeny of all leaf species identified from the nests of each species by using the prune function in the R package 'phytools' [65] from an existing rooted phylogeny of 912 plant species identified from the study region [66]. Faith's pd was calculated from the phylogeny of leaf preference for each of the three bee species using the R package 'picante' [67].     (table 1). Interestingly, all native plants collected by the bees are known to have antimicrobial properties (table 1). Only six plant species collected were known to contain no antimicrobial properties: Silene vulgaris (Moench) Garcke, L. corniculatus L., Stachys palustris L., Syringa vulgaris L., Rhamnus cathartica L. [54] and Ampelopsis japonica (Thunb.) Makino [59]. All six of these plants were exotic to the study region, and none were used by the native bee, M. pugnata (table 1)

Discussion
Leafcutting bees are reported to be selective in leaf preference, and to forage on only a few plant species [19,20]. Past studies reporting leaf preference of leafcutting bees did so using observational data [15,32,34]. In this study, I use DNA barcoding to identify leaf pieces from nests of three leafcutting bees and demonstrate that diversity in the leaves selected is far greater than previously reported. The leaf preferences of the two exotic bees overlap significantly and so I reject my first hypothesis that leaf preferences would be different among bee species. The pd and the species spectrum of selected leaves were higher in the two exotic bee species than in the native one. This is in accordance with the second hypothesis that leaf preference of the native bee is more restricted to phylogenetic groups. Finally, according to the χ 2 -tests none of the bee species exhibited a significant preference for native or exotic leaf species, indicating that the third hypothesis, e.g. that the native bee, M. pugnata, would prefer native plant species to exotic species, is not valid. Dependence of the native M. pugnata on nesting materials that are more phylogenetically related could make this species and other native bees more susceptible to environmental change [68]. Ecologists increasingly use phylogenetic measures to inform our understanding of community assembly as well as management practices to conserve biodiversity and ecological functioning [69,70]. Some studies have used phylogenetic relatedness to describe bees and their foraging preferences [71,72] and their nesting behaviour [73]. Future work could evaluate matching (or mis-matching) in phylogenetic relationships (e.g. [74]) among bees and nesting resources. These relationships examined along gradients of environmental change could aid in determining the extent to which these eco-evolutionary relationships depend on environmental filters, such as urbanization [75].
The leaves collected belonged to a variety of plant types: trees, shrubs and flowering plants, as well as one grass species. Only M. centuncularis exhibited a significant preference for woody plants (e.g. trees and shrubs) (table 2). None of the bee species exhibited significant preference for native or exotic leaf types. M. pugnata visited many exotic species including spontaneous urban plants (e.g. lamb's quarter (Chenopodium album)) and invasive species (e.g. purple loosestrife (L. salicaria)) (table 1). The exotic M. rotundata collected more exotic leaf types than the other two species (table 2), and among the 36 species recorded was the invasive dog-strangling vine (Cyananchum rossicum) [76] (table 1). This vine has invaded Southern Ontario, where it 'strangles' and suppresses the growth of native vegetation [77,78]. A number of other exotic plants were visited for nesting material by the three bee species ((e.g. bird's foot trefoil (L. corniculatus L.) and sweet clover (Melilotus albus Medik.)) (table 1). As exotic flora is abundant in urban areas [79,80], urban landscapes could provide a wider range of nest material for leafcutting bees, including native species [81,82].
Bee diversity and abundance are strongly linked to characteristics of the local environment, including the presence and quantity of foraging and nesting materials, as well as the amount of, and distance between, habitat containing these resources [83][84][85][86][87]. In urban landscapes containing many thousands of individually managed gardens that together form a rich diversity of flowering plants [88,89], leafcutting bees are not limited in these areas by nesting materials. Given how diverse leaf selection was for each of the leafcutting bees in this study, it is possible that further DNA barcoding would identify even greater numbers of plant species used. Citizen science to help record leaf cuts on plants (see figure 1) with DNA barcoding could be useful for examining leaf preference and diversity along urban-rural gradients where dominant factors affecting plant species assembly change from natural to anthropogenic [75,90].
Almost all the leaves identified in the nests are known to contain antimicrobial properties (table 1). Many ground-nesting bees waterproof and sterilize their brood cells using glandular secretions from the Dufour's glands [91,92]. These glands are significantly reduced in above-ground-nesting bees including megachilids [2], and so antimicrobial properties of leaves might inform choice [24][25][26]. One aboveground-nesting megachilid, Anthidium manicatum, collects leaf trichomes to make their nests which have antimicrobial properties [25] and their physical properties actively prevent attacking parasites [93]. One study showed leaf type and physical properties impact choice in M. rotundata, which preferred buckwheat unanimously over alfalfa and especially leaves greater than 1 cm 2 [19]. A combination of the antimicrobial and mechanical properties of leaves may inform choice among leaf types and why certain leaves are preferred over others by megachilid bees.
Missing from this study is an examination of the continuity and variability in leaf choice within single nests, which limits interpretation of the diversity of leaf types used by individual females. Identifying all leaves comprising a complete nest of individual bees (some complete nests contain greater than 100 individual leaf pieces) would contribute to knowledge on how diverse leaf preference is among individual bees, or even how it changes over a season. Knowing the identities of plants used by leafcutting bees can inform 'complete' pollinator gardening and broader actions supporting wild pollinators that include both foraging and nesting requirements [42,[94][95][96][97].
Data accessibility. DNA sequences: http://dx.doi.org/10.5883/DS-8669. Competing interests. I declare I have no competing interests. Funding. The work was funded by an NSERC scholarship (CGS D 408565) to the author and a discovery grant awarded to his supervisor, Dr Laurence Packer.