Floral hosts of leaf-cutter bees (Megachilidae) in a biodiversity hotspot revealed by pollen DNA metabarcoding of historic specimens

ease of the process that does not require a trained palynologist [7]. Genetic analyses of pollen also allows for more accuracy, since pollen morphology is not diagnostic in all species [8]. Pol- len identification through metabarcoding has been shown to be consistent in comparison to microscopy-based identifications [2, 9, 10], but there is no consensus yet regarding the consis- tency of pollen quantification between the two methods [2, 11]. The applications of pollen metabarcoding are extensive, and include monitoring food- and air-quality, forensic studies, and reconstructing ancient plant communities among others [7]. Metabarcoding pollen loads sampled from pollinators, as opposed to conducting lengthy field experiments, can provide insight into plant-pollinator interactions. There is still a great deal to learn about interactions between most pollinators and the plants they visit and metabarcoding provides a molecular tool for the rapid identification of pollen with mixed origin. In particular, very little is known about the floral choices of many endemic bee species in South Africa, and their interactions with plants are currently only studied through direct observations (eg. [12]) or observations combined with micoscopic examination of pollen loads (eg. [13–16]). South Africa is exceptionally rich in plant and animal diversity, with insects accounting for the majority of animal species present within the country [17–19]. Plant and bee diversity and endemism is particularly high in the western part of the country [17, 20–22] with high levels of

specialisation in mutualisms reported for this biologically important area [12]. Specialist bee and plant species are more sensitive to changes in the environment and consequently more vulnerable to declines and extinction [23, 24]. This is particularly troubling in the face of global climate change. Given the susceptibility to changes in the environment and the high levels of endemism in South Africa, there is a drive to study and better understand the interactions between bees and plants of important biodiversity hotspots. Taxonomists, collecting bees for identification purposes, have filled insect collections with specimens from a wide range of taxa. These collections tend to be well maintained with infor- mation included with each specimen on when, where and by whom the specimen was col- lected [25]. Specimen labels can also include information about plant associations, but this is often lacking. Bees in these collections are not stripped of their pollen loads during their taxo- nomic identification and thus contain a wealth of information about the plants they visited before capture. This untapped resource can be taken advantage of by sampling pollen from insect collection specimens for metabarcoding. Recently, research showed that plant origin identifications could be made from pollen sampled from historic insect collection specimens dating back to 1914 using the internal transcribed spacer regions (ITS1 and ITS2; [26]). In this latter study, using ITS2 as a barcode, provided much better resolution for plant classification

than ITS1. Unfortunately, the DNA barcode reference library for South African plants is not complete and species- and genus-level plant classification were problematic, although family- level classification was possible [26]. Nonetheless, cataloguing the number of plant OTUs vis- ited by bees with different habitat preferences will provide valuable information. The use of historical specimens can greatly increase the scope of questions that can be asked, as well as increase sample sizes (both temporal and spatial) available for use in studies by supplementing data obtained from field observations. In this study, pollen loads from bee specimens housed in the national insect collection in South Africa was used to determine whether the floral choice of bees in the family Megachilidae differ between the Succulent Karoo biome in the west and the Savanna region in the east of South Africa that has fewer endemics. We expected higher levels of plant and pollinator endemism to be closely linked to increased pollinator specificity and so expected megachilid species collected from the Succu- lent Karoo biome to visit fewer plants during foraging in comparison to species collected in the less diverse Savanna biome. In addition, we expected megachilid species that are wide- spread and occur in multiple biomes across South Africa to display more generalist foraging PLOS ONE Floral hosts of Megachilidae revealed by metabarcoding PLOS ONE | https://doi.org/10.1371/journal.pone.0244973 January 21, 2021 2 / 16 12994067); M. karooensis (https://doi.org/10.

6084/m9.figshare.12994115); M. maxillosa (https://doi.org/10.6084/m9.figshare.12994148); M. murina (https://doi.org/10.6084/m9.figshare. 12994223); M. niveofasciata (https://doi.org/10. 6084/m9.figshare.12994238). Data for M. venusta are available in the European Nucleotide Archive (ENA) under project accession PRJEB14178 (http://www.ebi.ac.uk/ena/data/view/PRJEB14178). Funding: The authors received no specific funding for this work. Competing interests: The authors have declared that no competing interests exist.