Detection of Tarnished Plantbugs, Apple Maggots, and Codling Moths in Bats' Diet in Southern Michigan Apple Orchards

Field Collection                          

            I captured bats starting on 18 July 2019 and ended 12 September 2019. I used 4m, 6m, and 9m 38mm polyester mist nets (Avinet, Dryden, New York) on triple-high aluminum pole sets to capture bats in the orchards. Mist netting occurred 3-4 times per week, weather permitting, but only one orchard was sampled per night. Nets were placed in open corridors within or around the orchard and near advantageous spots such as water. Nets were active for 4 hours after sunset.  

            Nets were checked every 10 minutes for bat safety and to prevent escapes. After removing bats from the net, I placed each bat into its own brown paper bag for up to 1 hour to allow the bat time to defecate. Bats were then processed by identifying the species, measuring forearm length (mm), measuring weight (g), identifying its sex, estimating its age (juvenile or adult) using the metacarpal-phalangeal joint’s osteological development (Brunet-Rossinni and Wilkinson 2009), assessing its reproductive status (non-reproductive, lactating, post-lactating), and measuring wing score for white-nose syndrome damage (Reichard and Kunz 2009).

            I marked all captured bats with a non-toxic washable marker to identify recaptures throughout the sampling event. Any guano deposited in the bags were collected in a 2mL tube with desiccant beads and stored at -18°C for around a week and then stored at -20°C until processing (Corthals et al. 2015). I disinfected equipment using a 1:10 diluted bleach mixture after each use (Shelley et al. 2013).  

            In addition to capturing bats in mist nets, I collected additional guano samples underneath occupied roosts. EAFI and SPIR were the only sites that had accessible bat roosts on site. EAFI had a bat roost in the eaves of a house, and SPIR had an occupied bat house. I placed a tarp under each roost during the sampling event to collect additional guano samples. The tarps were placed at the start of netting and retrieved after 4 hours. Guano samples were collected using 2mL tubes with silica desiccant beads and stored at -18°C for around a week and then stored at -20°C until processing (Corthals et al. 2015). I disinfected tarps after each survey using a 1:10 diluted bleach mixture (Shelley et al. 2013).  

            In 2020, the US Fish and Wildlife Service did not permit bat handling because of the unknown risk of human-to-bat transfer of the SARS-CoV-19 virus during handling (AFWA 2020, Runge et al. 2020). Two new orchards were added to the study area that had accessible bat roosts under which I could collect freshly-deposited guano. Erie Orchard and Cider Mill (EROR; Erie, MI) and Crane Orchard U-Pick and Corn Maze (CROR; Fennville, MI) had E. fuscus roosting in a building and in barn rafters, respectively. EAFI and SPIR were the only original orchards retained from the 2019 field season since they had accessible bat roosts.

            Guano collection occurred from 19 May 2020 until 27 August 2020. Tarps were set under each roosting site to collect guano samples upon exiting and returning to the roost. Fecal samples that gathered on the tarp between sampling events were collected before the next sampling event began (Figure 2). Guano was collected again the day following the sampling event. Sampling events did not exceed 2 consecutive days. I collected the guano samples in 2mL or 15mL tubes with silica desiccant beads and stored at -18°C for around a week and then stored at -20°C until processing (Corthals et al. 2015). I disinfected tarps after each survey using a 1:10 diluted bleach mixture (Shelley et al. 2013).

Molecular Methods

           I extracted DNA from 130 guano samples using the QIAGEN PowerSoil Pro Kit (QIAGEN, Germantown, Maryland) following manufacturer instructions. I extracted DNA from all 2019 samples since they came from unique individuals and multiple species at different orchards. In 2020, only samples collected during a sampling event were extracted due to time and budget constraints. Two pellets from each tube (for both 2019 and 2020) were used for the isolation process to reduce the amount of PCR inhibitors in the final elution.

            I used the primer set LCOI-1490 (Folmer et al. 1994) and COI-CFMRa (hereafter ANML) as described in Jusino et al. (2019) to amplify 180bp of arthropod cytochrome oxidase I (COI) mitochondrial gene. Primers were modified to include the forward and reverse overhang designed for use with the Illumina MiSeq system for high throughput sequencing, reaching a total amplicon length of 239bp (LCO-1490) and 257bp (COI-CFMRa) (Table 6). Polymerase chain reaction (PCR) mix for DNA amplification used the following reagent volumes and concentrations per 30µL reaction: 15µL Q5® Hot Start High-Fidelity 2x Master Mix (New England BioLabs), 1.5µL each forward and reverse primers (10µM concentration), 2.4µL extracted template DNA, and 9.6µL deionized water. Thermocycler conditions followed those described by Jusino et al. (2019) for the ANML primer pair. Amplified DNA was visualized on a 2% agarose gel stained with GelRed® (Biotium, San Francisco, California), and visualized with a UV transilluminator. Samples positive for insect DNA were loaded into a 96-well plate. A minimum of 25µL amplified DNA was used in each well and sent to the core facility Annis Water Resources Institute.

            Library preparation, quality checking, and sequencing were performed by the core facility. These steps included the following. After the initial PCR amplification, Nextera XT indices (Illumina, San Diego, California) and adapters were added to each sample. The amplicon library concentration was checked using a Qubit fluorometer (ThermoFisher, Waltham, Washington), and the quality and average size of the amplicon library was determined using an Agilent Bioanalyzer (Agilent, Santa Clara, California). Individual libraries were normalized to 4nM and pooled together. The pooled library was denatured with 0.2N NaOH and diluted to 10pM. The amplicons were sequenced using a 2x300bp format, along with a 15% spike-in of Phi-X on the Illumina MiSeq system (Illumina, San Diego, California). The core facility removed Illumina MiSeq overhang sequences from the raw sequence reads prior to distributing raw data.

Bioinformatics Methods

I used QIIME 2 (Bolyen et al. 2019) to process my raw sequence data. My data were formatted as a Casava 1.8 paired-end demultiplexed FASTQ file with a forward and reverse read and the corresponding quality score for each base. I demultiplexed the reads which allowed the program to know which sample came from which read by using the unique barcode on the 5’ end of a sequence. 

            Because not all sequences were of equal quality and length, I used the cutadapt plug-in (cutadapt trim-paired; Martin 2011) to trim the ANML primer sequences off the 5’ end of each read. Cutadapt parameters were set with a minimum length of 180bp to filter only those reads matching the target amplicon length, an error rate of 0.1, and an overlap of 3. Reads that did not meet these criteria were discarded from the dataset.

            Next, I used the vsearch plug-in (vsearch join-pairs; Rognes et al. 2016) to join the trimmed demultiplexed forward and reverse paired reads together, allowing the program to identify staggered pairs. I produced an output file to visualize the best PHRED quality score for my data to further filter low-quality reads (quality-filter q-score; Bokulich et al. 2013). The standard minimum PHRED score for high-quality data is a minimum of 20, and the visualization indicated 20 was a good PHRED score cutoff for my data. I also chose to remove contigs with ambiguous base calls in the reads. All other reads with a PHRED score lower than 20 or with any ambiguous base calls were filtered out of the dataset.

            I then clustered similar reads together into operational taxonomic units (OTUs) using another vsearch plug-in (vsearch dereplicate-sequences; Rognes et al. 2016). Clustering the reads in vsearch summarizes the data into a feature table, which is a matrix of the number of times each OTU is observed in each sample. This table was then used to remove unfound chimeras de novo (vsearch uchime-denovo; Rognes et al. 2016). The remaining reads were clustered again, and the features were collapsed (vsearch cluster-features-de-novo; Rognes et al. 2016). I ran the to create OTUs where reads were 99% alike to each other. I further filtered my dataset to remove reads with less than 11 features, which maximized the data retained but removed as much of the low-quality reads as possible. The remaining reads were exported into a FASTA file.

            I used BOLDigger (Buchner and Leese 2020), an extension of BoLD (Barcode of Life Database; Ratnasingham and Hebert 2007) to identify each read. FASTA files were uploaded into BOLDigger. The program automatically searched for the matching sequences if there was a match available. It produced an output file of the top 20 matches per read.