Flower-derived eDNA holds great promise as a rapid and non-invasive tool for monitoring pollinators and their plant-associations. However, pollinators often only briefly interact with a plant and leave little eDNA, making them particularly challenging to detect. In addition, taxonomic biases in eDNA deposition and PCR amplification prevent quantitative analysis of pollinator diversity. These limitations have so far precluded the widespread use of eDNA in pollinator monitoring. Comparing flower-derived eDNA with conventional monitoring in flower strips, we here explore the utility of eDNA to detect community diversity, species abundances, and ecological specificity of plant-associated arthropods. We show that read abundances are a bad predictor of true abundances at the community level. Instead, the occupancy of individual species in replicated flower eDNA samples provides reliable quantitative estimates of pollinator biodiversity and detects their ecological specificity very well. Also, we find that pollinator eDNA can be collected non-invasively, by washing off from flowers in the field. Our work highlights eDNA analysis as a powerful tool for the rapid future monitoring of plant-arthropod interactions and plant-pollinator networks.

Comparing conventional monitoring and eDNA sampling in flower strips

The study was carried out on three flower strips in Rheinmünster (Baden-Württemberg, Germany, N 48° 44' 56.9076 E 8° 1' 38.7912, 125 m above sea level). They are all located relatively close together in an intensively farmed landscape where maize is the dominant crop. A small lake in the north-east, surrounded by small trees, and embankments along the roads are the only structural elements in the area. The three flower strips are 0.44 ha (7), 0.31 ha (15) and 0.27 ha (19) in size. The area was chosen because there is an ongoing conventional pollinator monitoring since 2010 by Bayer in cooperation with the Institute of Agroecology and Biodiversity (ifab) and the NABU forest institute (NABU-Waldinstitut - formerly Institut für Landschaftsökologie und Naturschutz (ILN) Bühl) (Buhk et al. 2018). It is important to note that the bee expert was only looking for wild bees and not for the Western Honeybee (Apis mellifera). Honeybees, however, were also not very prevalent on the studied sites.

To obtain comparable results, sampling dates were co-timed with the visual bee monitoring in the flower strips. Conventional sampling took place four times during our sampling period (06/02, 06/20, 07/06, 07/20/2023). We collected eDNA samples on four sampling dates (06/02, 06/27, 07/10, 07/21/2023). We sampled only after at least 24 hours of dry weather to assure eDNA was not washed away by rain. Hence, the sampling dates between eDNA sampling and morphological identification were not always identical. On all sampling dates two different methods for obtaining arthropod eDNA from flowers were tested:

1) The first method was to pick flower heads. Here, samples of the 12 most abundant plant species were collected per flowering strip on every date. If there were not enough abundant species in a flowering strip, replicates of very abundant species were made (for the species list see suppl. Table 1). Flowers were collected using single-use gloves, which were changed for each plant sample to prevent cross-contamination. Care was taken to collect only the flower head and not the stem. Furthermore, the flowers were gently shaken to dislodge any small arthropods present. Flowers of the different species were collected in roughly equal quantities in zip-lock bags.

2) The second method is a non-invasive approach. The flower heads were sprayed with deionized water using a pressure sprayer (Prima 5, Witten, Germany) and the water was collected in zip-lock bags, held under the flower while spraying. The same 12 plant species were used as for the collection of flowers, to make the results comparable. The same number of flowers were rinsed as were collected. Again, single-use gloves were worn, and the flowers were gently shaken to remove arthropods before rinsing.

All eDNA samples were stored in the field in an insulated box filled with ice and immediately frozen at -20°C on arrival at the laboratory until further processing.

For the conventional method, direct observation was chosen to monitor the bee species (see Buhk et al. 2018). The areas were walked two times on every visit for 30 min each and all bees identified on sight if possible. If this was not possible, they were caught and identified afterwards morphologically. 

DNA isolation, PCR amplification and sequencing

Collected flower heads were weighed. 10 times the amount of deionized water (in ml) was then added per plant weight (in gram) to the zip lock bags. The samples were then shaken for one minute and afterwards filtered through a 0.45 µm cellulose nitrate filter (Thermo Fisher Scientific Inc., Waltham, USA) using a vacuum pump. The same was done with the previously thawed water samples from the rinsing of the plants. Also, blank controls for the deionized water used in the field and laboratory were filtered. Filters were then stored in 1.5 ml tubes in a freezer at -20°C. DNA isolation was done using the Qiagen Puregene Blood & Tissue Kit (Qiagen, Hilden, Germany). In the first step 450 µl ATL buffer was added, and the samples were then bead-beaten at 1000 rm for 45s. This bead-beating was repeated one time. The rest of the DNA isolation followed the manufacturer's protocol.

We used the primer combination fNoPlantF_270 (forward primer, RGCHTTYCCHCGWATAAAYAAYATAAG) and mlCOIintR_W (reverse primer, GRGGRTAWACWGTTCAWCCWGTNCC) (Krehenwinkel et al. 2022) to amplify our samples. This combination targets a 116 bp fragment of the mitochondrial CO1 gene for a broad range of arthropod taxa. Samples were amplified using the Qiagen Multiplex PCR kit (Qiagen, Hilden, Germany). 1 µl each of the 5 µM primers, 5 µl multiplex master mix, 2 µl water and 1 µl DNA were used for the PCR. PCRs were run with 35 cycles and an annealing temperature of 46 °C. Three PCR replicates for each sample were run and treated as different samples afterwards. After checking the success of the PCR using a gel image, five cycles and an annealing temperature of 55 °C were used for the following dual index PCR (see Lange et al. 2014). During this, Illumina TruSeq Adapters (Illumina, California, USA) were attached to the 5' end. The indexed products were again checked on an agarose gel. The intensity of the band on the gel image was used to pool the samples in their approximate proportions. Afterwards magnetic beads (1:1 sample to beads ratio, AMPure XP, Beckman Coulter, California, USA) were used to clean the pools from leftover primers. For each isolation, PCR and index PCR, negative controls were run along with the samples and sequenced to check for contamination. Using an Illumina Miseq the samples were then sequenced with a Miseq Reagent Kit v3 (Illumina Inc, San Diego, California, USA).

Data processing and Statistical analysis

Samples were demultiplexes using CASAVA (Illumina Inc, San Diego, California, USA), with no mismatches allowed. PEAR (Zhang et al. 2014) was used to merge the resulting fastq files with a minimum quality score of 20% and a minimum overlap of 50 bp. Fasta files were created using the fastq filter (minimum 90% of bases >Q30) command of USEARCH (Edgar 2010). Primers were then trimmed of with the sed command in UNIX. Using USEARCH reads were dereplicated and then clustered into zero-radius operational taxonomic units (zOTUs). These were then compared against the full NCBI database with BLASTn (Altschul et al. 1990) and a minimum of 10 target sequences. Respective taxonomy was added using blast2taxonomy (Schöneberg 2024). Only zOTUs where all 10 hits were arthropods were retained. An OTU table was constructed using USEARCH. We annotated all sequences with a minimum of 90 % similarity to order and 100 % as species. For the following analysis, only zOTUs were kept, which occurred in all the three PCR replicates. Also, only zOTUs with a minimum of 11 reads in a sample were retained to remove index carry over between samples. In addition, one species of bumblebee (Bombus mucidus) was removed from the dataset as it is unlikely to occur in the sampling area (Westrich 2019).

References

Altschul, S. F.; Gish, W.; Miller, W.; Myers, E. W.; Lipman, D. J. (1990): Basic local alignment search tool. In: Journal of Molecular Biology 215 (3), S. 403–410. DOI: 10.1016/S0022-2836(05)80360-2.

Buhk, Constanze; Oppermann, Rainer; Schanowski, Arno; Bleil, Richard; Lüdemann, Julian; Maus, Christian (2018): Flower strip networks offer promising long term effects on pollinator species richness in intensively cultivated agricultural areas. In: BMC Ecol 18 (1), S. 55. DOI: 10.1186/s12898-018-0210-z.

Edgar, R. (2010): Usearch. Lawrence Berkeley National Lab. (LBNL), Berkeley, CA (United States).

Krehenwinkel, Henrik; Weber, Sven; Künzel, Sven; Kennedy, Susan R. (2022): The bug in a teacup-monitoring arthropod-plant associations with environmental DNA from dried plant material. In: Biology letters 18 (6), S. 20220091. DOI: 10.1098/rsbl.2022.0091.

Lange, Vinzenz; Böhme, Irina; Hofmann, Jan; Lang, Kathrin; Sauter, Jürgen; Schöne, Bianca et al. (2014): Cost-efficient high-throughput HLA typing by MiSeq amplicon sequencing. In: BMC Genomics 15 (1), S. 63. DOI: 10.1186/1471-2164-15-63.

Schöneberg, Yannis (2024): yschoeneberg/blast2taxonomy: Version 1.4.2: Zenodo. Available online at https://zenodo.org/records/10605493.

Westrich, Paul (2019): Die Wildbienen Deutschlands. 2nd ed. Stuttgart: Verlag Eugen Ulmer. Available online at https://livivo.idm.oclc.org/login?url=https://ebookcentral.proquest.com/lib/zbmed-ebooks/detail.action?docID=6969869.

Zhang, Jiajie; Kobert, Kassian; Flouri, Tomáš; Stamatakis, Alexandros (2014): PEAR: a fast and accurate Illumina Paired-End reAd mergeR. In: Bioinformatics 30 (5), S. 614–620. DOI: 10.1093/bioinformatics/btt593.