One Bat’s Waste is Another Man’s Treasure: A DNA Metabarcoding Approach for the Assessment of Biodiversity and Ecosystem Services Using Bat Faeces

Thomas Curran (  thomascurran303@gmail.com ) Waterford Institute of Technology https://orcid.org/0000-0001-5184-4498 Samuel Browett Waterford Institute of Technology David O'Neill Waterford Institute of Technology Aidan O'Hanlon National Museum of Ireland Catherine O'Reilly Waterford Institute of Technology Andrew Harrington Waterford Institute of Technology Allan McDevitt University of Salford Denise O'Meara Waterford Institute of Technology


Taxonomic Assignment 148
Taxonomic assignment was made by assigning MOTUs generated to species level with a minimum 149 identity of 98% requiring at least 90% coverage using the GenBank and BOLD databases, the latter of 150 which was used to confirm identification when MOTUs presented more than one possible species-151 level identification and were removed from the dataset when more than one species was assigned to 152 the same MOTU ( Supplementary Information 1,2,3). If multiple MOTUs were assigned to the same 153 species, they were agglomerated together using the sum of their sequence reads. 154

Dietary Diversity Measures 155
Using the R packages ggplot2, tidyverse, and knitR a "donut chart" was constructed to graphically 156 present taxonomic data for each MOTU detected within the R. hipposideros diet (donut chart script 157 source at https://github.com/ShrewlockHolmes/Taxa_Donut_Chart_Visual). The donut chart was 158 separated into three levels, each representing a different taxonomic rank i.e. order, family and genus. 159 The outermost level also contained a number providing an indication of the number of species within 160 that genus that was identified. 161 Associations between dietary composition at the levels of sex and location were assessed using 162 multiple statistical measures. The data were transformed into relative read abundance (RRA) using 163 the transform_sample_counts function within the R package phyloseq to provide an indication of how 164 common or rare certain taxa are in relation to other taxonomic groups. Stacked bar plots were 165 constructed in the R package ggplot2 using the RRA for each order. 166 Using RRA, a distance matrix was created using the Bray-Curtis dissimilarity method. Permutational 167 multivariate analysis of variance (PERMANOVA) was performed using the adonis2 function in the R 168 package vegan (Oksanen et al., 2019) with 10,000 permutations to determine compositional 169 difference in the prey taxa identified within the R. hipposideros diet by sex and location. To ensure 170 that the homogeneity of variance within the groups was not affecting the compositional differences, 171 the function betadisper() was used to measure the multivariate distance of samples to the group 172 centroid. All diversity measures described here were repeated with MOTUs agglomerated to order, 173 family, genus, and species taxonomic ranks. The data were then visualised using a non-metric 174  These 348 MOTUs represented ten arthropod orders (Araneae, Coleoptera, Diptera, Glomerida, 213 Hemiptera, Hymenoptera, Isopoda, Lepidoptera, Neuroptera, and Trichoptera), and one Annelida 214 order (Opisthopora: Crassiclitellata); consisting of 60 families, 120 genera, and 161 species (Fig. 2). 215 The most dominant order in the diet was Lepidoptera, followed by Diptera (Table 1), which accounted 216 for 55.23% and 18.01% of species in the diet, respectively. The orders Araneae, Hymeoptera, and 217 Trichoptera occurred less frequently in the diet and accounted for six, seven, and fifteen of the 218 identified species respectively (17.4% of the overall species level diet) ( Table 1). Species identified 219 within rarely occurring orders / suborders, such as Coleoptera (1.24%), Crassiclitellata (1.24%), and 220 Glomerida (0.62%) contributed marginally to the overall diet of R. hipposideros. Furthermore, several 221 species were recorded in this study that have not previously been documented in Ireland (see 222 discussion and Supplementary Information 4 for further details). 223 Barplots were constructed based on RRA to represent the variations of R. hipposideros diet according 224 to roost site location and sex (Fig. 3). At the roost level, Lepidoptera and Diptera were found to be the 225 most dominant orders overall with the exception of roost 3 (Co. Kerry), where the order Hymenoptera 226 was dominant. When diet was investigated by sex, Lepidoptera and Diptera were again the dominant 227 orders. Female R. hipposideros tended to consume more Lepidoptera than males. Less frequently 228 occurring orders including Neuroptera, Trichoptera and Hymenoptera were also more common in the 229 female diet, with Trichoptera only occurring in the female diet and Neuroptera and Hymenoptera 230 rarely occurring in males. 231 The PERMANOVA showed that sex did not have a statistically significant effect on the diet of R. 232 hipposideros (R 2 : 0.00273-0.0236, Pr(>F): >0.05). However, roost location was found to be a 233 statistically significant factor impacting the R. hipposideros diet (R 2 : 0.26115-0.3276, Pr(>F): <0.01) 234 ( Table 2). The R 2 values showed that between 26% and 32% of distance variation (depending on the 235 taxonomic rank assessed) was caused by the roost location. This data, at each taxonomic rank, was 236 also visualised using NMDS plots (Fig. 4). The NMDS plots showed that at order level there was an 237 overlap in most of the roost locations, with slight variation. However, roost 3 (Co. Kerry) formed its 238 own cluster outside of the other locations. This pattern can be seen at all taxonomic ranks, where 239 some overlap of each roost was observed, with slight variation, except for roost 3, showing that the 240 diet of R. hipposideros at this roost differed to the others. 241 The Permutest and Tukey analysis showed that sample homogeneity did not influence the 242 compositional difference detected via PERMANOVA as all p-values at both sex and roost for all 243 taxonomic ranks were >0.05. 244 The ANOSIM results also corroborated the trend observed via PERMONVA as sex differences were not 245 found to influence dietary composition. Statistic R values for sex ranged from -2.11 x 10 -2 to 1.61 x 10 -246 2 , and significance at all taxonomic ranks was >0.05 showing that sex did not significantly impact diet. 247 However, roost location was again found to have a statistically significant effect on the diet of R. 248 hipposideros, with statistic R ranging from 0.19 to 0.40, and significance values for all taxonomic ranks 249 <0.01. 250 251

Identification of pest species 252
A total of 38 potential pest species were identified, representing almost 24% of the overall species 253 identified in the diet (Table 3). Pest species were mostly Lepidopteran species, with 35 of the 38 254 (~92%) pest species identified as Lepidoptera. The rest of the potential pest species identified 255 consisted of two Diptera species (~5%) and one Hemiptera species (~2%) (Supplementary Information 256 5). 257 Of the 38 species listed in Table 3 something not normally achievable via hard-part analysis. This highlights the sensitivity of the DNA 282 metabarcoding approach over traditional hard-part methods and the resolution of the data 283 generated. 284

Location-and Sex-based Dietary Variation 285
Roost location was found to be the most informative variable to explain dietary differences across the 286 dataset, which was also found to be the case in R. ferrumequinum when studied in France (Tournayre 287 et al. 2021). Here, the diet of R. hipposideros was dominated by Diptera and Lepidoptera, but their 288 frequencies and composition varied according to location. The order Hymenoptera was relatively 289 abundant at roost 3 (Co. Kerry) and was also detected at roost 1 (Co. Mayo), but at a lower abundance. 290 Some less frequently occurring orders were also identified, including Araneae, Coleoptera, 291 Crassiclitellata, Neuroptera, and Trichoptera. Araneae, Coleoptera, and Trichoptera were all identified 292 in Co. Kerry. Dietary variation, particularly for the Kerry site, was evident in Fig. 4, where the points 293 around the group centroid for the Kerry samples clustered separately to the other five locations. Even 294 though the other roosts are located near woodland areas, most are in agriculture-dominated areas, 295 whereas the Kerry site is located in the centre of a heavily wooded area, considered as ideal habitat 296 for R. hipposideros in Ireland. The site in Co. Kerry is of international interest as it is a Special Area of 297 Conservation (SAC) for a range of priority habitats listed on Annex I and II of the European Habitats 298 Directive. This suggests that R. hipposideros diet is representative of what arthropods are present at 299 the time of sampling (i.e. opportunistic foraging) and that variable habitats play a role in influencing 300 bat diet. This is a factor which should be considered for future studies intending to use DNA 301 metabarcoding as a tool to investigate arthropod diversity and presence/absence of target 302 organisms/groups (Thomsen and Willerslev 2015). 303 Our analysis showed that the sex of the bat did not significantly impact their diet, with both male and 304 female R. hipposideros having a heavy Dipteran and Lepidopteran based diet, but again at varying 305 frequencies, but were not statistically significant. Females appeared to prefer Lepidoptera over 306 Diptera, while males predated more often on Diptera (Fig. 3). The female diet was also found to 307 include less frequently occurring orders (i.e. Hymenoptera, Neuroptera, and Trichoptera). Similar focus on larger prey items with a higher energy content to support their nutritional demands. These 315 subtle but important differences could be further investigated using the molecular approach outlined 316 in this study combined with an increased sample size to provide more statistically robust insights into 317 sex-biased dietary preferences. 318

Ecosystem Services 319
A total of 38 potential pest species were detected in this study, but the magnitude of the risk posed 320 by each of these species in Ireland is not well known, as the species were identified by comparing the 321 data generated from this study with studies from Spain and France The dataset generated here suggested the presence of 14 arthropod species not previously reported 354 in Ireland ( Figure S4). However, further investigation revealed uncertainties that these identifications 355 were truly new, and more likely caused by an inadequate reference database. A little over 10.5% of 356 the species level identifications generated from this study provided inconclusive results, despite using 357 internationally accepted thresholds for identification (   The application of DNA metabarcoding here has also allowed for the detection of potential vector 391 organisms that have been implicated in the spread of disease. In this study, several mosquito (Diptera: 392 Culicidae) and midge (Diptera: Ceratopogonidae) species were identified including Culex pipiens, Cx. care has to be taken that the originally deposited sequence was also accurately identified. However, 405 the approach of using a predator diet to indirectly survey potential airborne vectors has shown great 406 promise in this study and has the potential to be a powerful surveillance tool. have been in the east of Ireland. However, a closely related species P. faxinella is present in Ireland 418 and P. rucifeps was formerly considered to be a dark variant of this species, but DNA barcoding has 419 enabled the distinction between these two species. When the MOTU generated in this study was 420 compared to P. faxinella it was found to only be 97% similar, providing good confidence that both P. 421 rucifeps and P. faxinella are present in Ireland as has been recognised in Britain (Barnett 2017), and 422 that it is more common and widespread than previously thought. 423

Conclusion 424
In this study, DNA metabarcoding of relatively few bat faecal pellets provided a large arthropod 425 dataset. We found that the location of the bat roost was an important factor to explain dietary 426 variation in R. hipposideros, a finding which could be adapted in future studies aiming to investigate 427 the impact of land use on biodiversity. Our findings were not limited by the methodology we 428 employed, but by the lack of available DNA sequences present on reference databases to compare 429 Irish insect diversity. Our study was relatively small in scale but as a result, we were in a position to