Impacts of proactive health management on cattle and horse diets and dung biodiversity in Danish rewilding areas

Reintroducing megafauna to reinstate missing top-down trophic interactions (trophic rewilding) is increasingly being applied as a tool to promote self-regulating, biodiverse ecosystems. Even though the theoretical background is clear, and megafauna effects are documented from prehistoric ecosystems, the effects of reintroduced herbivores in contemporary ecosystems remain understudied. This includes how reintroduced megafauna interact with each other and the ecosystem, but also how current management practices affect the processes they provide. In this study, we investigated the effects of proactive health management, i.e., winter feeding and anti-parasitic treatments, on the ecosystem by examining diets of large herbivores and dung-associated invertebrate communities. We used environmental DNA metabarcoding to yield community compositions of plants and invertebrates in dung from cattle and horses from five comparable nature sites in Denmark, which differed in management, and site/population-specific properties such as availability of woody plant species, herbivore densities, and provision of winter feeding and anti-parasitic treatments. We found different diet compositions between cattle and horses, highlighting their functional differences. For example, horse samples had higher relative read abundances of graminoid and tree DNA. Supplementary feeding affected diets, by decreasing consumption of graminoids and tree species relative to forbs and legumes, probably originating from fodder, and intense feeding seemed to almost eliminate consumption of local vegetation. However, more studies are needed to generalize these findings. Several invertebrate families were associated with either cattle or horse dung, suggesting complementary effects on dung-associated invertebrate biodiversity by these large grazers. The taxa that responded negatively to anti-parasitic treatments were mainly parasitic nematodes (e.g., the families Ancylostomatidae, Cooperidae, and Strongylidae), suggesting that the applied treatments work as intended, but these results should be interpreted with caution due to methodological limitations.

Synthesis and application. Our findings demonstrate functional differences between cattle and horses, which suggest complementary effects on vegetation development and consequently biodiversity. Also, our results indicate that this functionality is impacted by proactive health management actions. We suggest that potential effects on herbivory and biodiversity are carefully considered before supplementary feeding or anti-veterinary treatments are provided in year-round grazing systems and avoided if possible.

Methods
The dataset consists of DNA reads from high-throughput sequencing of 295 dung samples from cattle and horses collected at five sites in Denmark in 2022. From each of the five sites, 7 samples from cattle and 7 samples from horses (except the NM site where no horses were present) were collected in February, March, April, Jun,e and August, respectively. At each sampling event, a field blank was collected and sequenced alongside the other samples (see associated manuscript for details). Twenty-two of these samples (10 from the ML and SL sites respectively, and two field blanks collected in August) were sequenced alongside (and will also be used in) another project, and was thus not part of this raw data, but the filtered data from these samples are added as separate files (see description further below) and should be appended to the main dataset for the subsequent analysis.   

DNA was extracted from the samples with the Fast DNA Stool Mini Kit from Qiagen, and amplified by PCR reactions with two primer sets (see PCR reagents and thermal settings, etc. in the associated manuscript). One with the BF-1/BR-1 primers (Elbrecht & Leese, 2017, https://doi.org/10.3389/fenvs.2017.00011), targeting a 217 bp fragment of COI optimized for invertebrates, and one with the ITS2-S2F/ITS4 primers (Fahner et al., 2016, https://doi.org/10.1371/journal.pone.0157505), targeting the nuclear ITS region, and optimized for plants. 

During the laboratory pipeline, 32 extraction blanks (sample names including CNE) and four PCR blanks for each amplicon pool (sample names including NTC, 28 in total) were included, which were sequenced alongside the rest of the samples. Hence, in total, the uploaded raw sequencing data includes DNA reads obtained from 373 samples, which were separated into 7 batches. Each batch was used as a template for both primer sets and run through 4 replicate PCR reactions: L11-L74 for COI, and L081-L144 for ITS. For 1 library (L092), additional sequencing was performed, and thus, two separate raw data files exist from this library. See README.md for a description of how to treat these in the bioinformatic pipeline.

The raw sequencing data were run through the MetaBarFlow pipeline (https://github.com/evaegelyng/MetaBarFlow), with parameters following Thomassen et al. (2024) (https://doi.org/10.1111/mec.16847), and the exact scripts are located here: (https://doi:10.5281/zenodo.15296452). The pipeline produces an ASV list  (*DADA2_nochim.otus), and a matrix with read counts of each ASV in each sample (*DADA2_nochim.table), as well as a list with taxonomic assignment of all ASVs (*classified.txt) for each data set. The taxonomic identification of DNA sequencing reads for the ITS dataset were made by blasting (blastn) against the complete NCBI genbank nt database (https://www.ncbi.nlm.nih.gov/) downloaded locally, and for the COI dataset, blasting was performed against a custom build COI database containing all COI sequences from BOLD (www.boldsystems.org) and NCBI Genbank (https://www.ncbi.nlm.nih.gov/) See Klepke et al. (2022) (https://doi.org/10.1002/edn3.340) for further description of how the database was built, and the associated publication for details of Blast parameters.

For final taxonomic assignment (score_ID column in "*classified.txt") was defined as the last common ancestor of all blast hits within the range of sequence similarity of hits to the best match, including hits within a 2% margin of the best ID, and species ID was only assigned if the best match was >98% similar.

The list of taxonomic assignments was manually checked for errors resulting from spurious reference database sequences or similar, and when such errors were spotted, the taxonomic assignment of the given ASV was corrected manually. Also, for COI, ASVs identified at higher levels than species were assigned to "putative species", which were units including the same possible IDs. For ITS, aggregations were made at the genus level. These final manually edited IDs are found in the "final_ID" columns in the files "*classified_corrected.txt".

See the manuscript and associated GitHub (https://github.com/emilthomassen/RWDK_public) for further details about subsequent analysis.