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

not peculiar. Amaranthaceae is one of the top ten families in the Tankwa National Park [42] and in the top 20 families within the Extra Cape Subregion, covering 92% of the Succulent Karoo [43]. Amaranthaceae is also common in arid areas in Australasia and Eurasia. It was also found to be one of the dominant families in the Gannaveld, near Oudtshoorn [44]. To date, intensive studies of plant-bee interactions in South Africa have focused on the western part of the country, and no studies of similar magnitude have been performed in the Savanna region. As Asteraceae is the largest plant family in South Africa and the Succulent Karoo [45], it was expected that it would be one of the most highly represented plant families identified in pollen from all biome groups, particularly in pollen from the Succulent Karoo specimens. Interestingly, in the bee subfamily Megachilinae varying rates of flower visitation has been observed between its three tribes. The Megachilini tribe visits Fabaceae flowers markedly more than the other two tribes, and the reverse is true for their visits to Asteraceae. A low flower visi- tation rate of only 20% was observed in the Megachilini for Asteraceae [39]. Our results showed that Fabaceae was present in the top five most visited families for all three groups but Asteraceae was only a highly represented family in pollen from widespread bees. Grasses (Poaceae) were another highly represented family in all three studied groups. There have been reports that bees also pollinate grasses [46–50]. The Poaceae family is very

highly represented in the Savanna and is one of the 15 largest families in the Succulent Karoo [39] and it is thus plausible that these Megachile species could collect pollen from this family. Pteridaceae was identified in almost all pollen samples and in high proportions as well. The legitimacy of this identification is under contention as it is possible that the entries in the underlying reference sequence database originating from NCBI were problematic [26]. As Pteridaceae is a family that has 7,643 species across South Africa and 331 species occur in the Succulent Karoo and 2,177 in the Savanna [35], it is quite plausible that airborne fern spores are included in pollen loads. Also, the presence of fern spores in pollen metabarcoding studies of honey [51, 52] confirm the possibility of bees foraging on spores, even though its presence was detected in less samples and in lower frequencies. This study illustrated once again the utility of historic insect specimens for pollen sampling in order to answer questions about floral host selection. Importantly, time since bee collection did not have a significant effect on the mean ranges of plant taxa observed, which indicates that pollen DNA was sufficiently preserved within the insect collection for analysis. Pollen loads from similarly stored museum specimens were analysed to investigate the effects of dif- ferent factors on bee species decline in the Netherlands [53]. They found a significant effect for time, where the size of floral ranges observed in pollen loads from bees prior to 1950 played a

key role in bee population trends. There are limitations in the present study that need to be taken into account when consider- ing the results presented. First, the results obtained were contingent upon the bee species chosen for study. All six of the species chosen were within the same family (Megachilidae), sub- family (Megachilinae) and tribe (Megachilini, [37, 38]). Choosing bee specimens from differ- ent subfamilies or tribes within the same family might produce different results. Second, not all specimens for the species of interest in the collection necessarily contained pollen for sam- pling, which means all specimens with pollen loads were selected for the study. In this study, it meant that some original sampling dates and localities were overrepresented in some species, in particular in M. karooensis and M. murina. In the widespread bee M. niveofasciata, speci- mens with pollen loads were coincidentally all caught in the western region of South Africa, whereas M. venusta had a broader sampling spread across the country. A limited spread of data hampers the ability to infer on floral choice differences on a species-wide scale. A uniform PLOS ONE Floral hosts of Megachilidae revealed by metabarcoding PLOS ONE | https://doi.org/10.1371/journal.pone.0244973 January 21, 2021 12 / 16