Investigating anthropogenic and social influences on diet of semi-urban vervet monkeys using DNA metabarcoding

In human-dominated ecosystems, wildlife has been forced either to disappear or to adapt its behaviour in order to exploit the opportunities associated with anthropogenic activities. Vervet monkeys (Chlorocebus pygerythrus) are omnivorous primates whose natural habitats have been progressively encroached upon by expanding suburban development. Due to their generalist and opportunistic feeding behaviour, vervet monkeys have successfully adapted to semi-urban environments. Characterising the composition of their diet can therefore reveal how they exploit anthropogenic resources and uncover new foraging behaviours. However, accurately determining their diet through direct observation can be challenging, especially in semi-urban areas where numerous anthropogenic structures obstruct visibility. Environmental DNA (eDNA) has been proposed as a non-invasive complementary method to determine diet and foraging strategies by analysing the DNA mixtures present in faecal samples. In this study, we determined the dietary components of vervet monkeys using DNA metabarcoding of 447 faecal samples collected from two monkey groups over four months in a semi-urban neighbourhood in South Africa, and compared the results with observational foraging data to elucidate how vervet monkeys exploit anthropogenic resources. Subsequently, we evaluated whether dietary patterns can be distinguished between groups and within matrilineal levels. We found DNA metabarcoding data to be consistent with observational data, but the former revealed a broader diversity of consumed taxa. Additionally, we detected a difference in diet between the two groups, and a tendency for similar dietary patterns among matrilineal pairs compared to other group members. Our results support the use of the DNA metabarcoding methodology, both to determine the complex diet of omnivorous species in urbanised habitats and to address interindividual foraging behaviours.

Methods
DNA extraction: DNA was extracted from faecal samples using a phosphate buffer-based approach (Taberlet et al., 2018), based on the protocol from the NucleoSpin Soil Kit (Macherey-Nagel), with the following modifications. The faeces in the scintillation vials were directly transferred to 2 mL Eppendorf tubes with 1.3 mL of phosphate buffer. The samples were placed on a tube rotator for 15 minutes to facilitate DNA absorption and later homogenised by vortexing. The samples were then centrifuged for 5 minutes. The subsequent steps of the extraction required the use of QIAvac technology (Qiagen). DNA extractions were performed in a dedicated laboratory designed for handling low-template DNA samples (Laboratory for Conservation Biology, University of Lausanne). For subsequent analyses, all DNA extracts were dilutedfivefold.

DNA metabarcoding assay: DNA extracts were amplified in triplicate using two sets of primers, one for plants and one for vertebrates. The first primer pair (Sper01) targets plant components of the diet by amplifying the P6 loop of the trnL (UAA) intron of chloroplast DNA (10-220 bp; Taberlet et al., 2018). The second primer pair (Vert01) amplifies a fragment of the mitochondrial rDNA 16S, and is highly specific for vertebrates (56-132 bp; Taberlet et al., 2018). For the latter, a blocking oligonucleotide (5'-CTATGCTTAGCCCTAAACCTCAGTAGTTAAACCAACAAAACTACT-C3-3') was added to specifically inhibit the amplification of vervet monkey DNA. PCR primers included 5? tags, consisting of an 8-nucleotide sequence with at least 3 nucleotide differences between each tag, are used for assigning sequences to their respective sample. PCR reactions were performed in a final volume of 20 µL in 96-well plates. For the Sper01 primers, the mix contained: 1× AmpliTaq Gold 360 (Applied Biosystems), 0.5 ?M of forward and reverse primers, 0.16 mg/mL of Bovine Serum Albumin (BSA; Roche Diagnostics), and 2 µL of template DNA. For the Vert01 primers, the mix contained 1 × AmpliTaq Gold 360 (Applied Biosystems), 0.2 ?M of forward and reverse primers, 0.16 mg/mL of Bovine Serum Albumin (BSA; Roche Diagnostics), 2 ?M of blocking oligonucleotide, and 2 µL of template DNA. PCR cycling conditions were 10 minutes at 95°C, followed by 40 cycles of 30 seconds at 95°C, 30 seconds at 52°C or 49°C (for Sper01 and Vert01, respectively), and 60 seconds at 72°C, with a final extension of 7 minutes at 72°C and 10 minutes at 4°C. For each PCR plate, a negative extraction control, a negative PCR control (ultrapure water), positive controls, and blanks were included. Positive controls contained DNA of known concentration from plant or vertebrate species not expected at the study site, serving to validate the amplification success (Table S5). Amplification success was verified by 2% agarose gel electrophoresis for a subset of samples. Finally, PCR products for each primer pair were pooled for library preparation.

DNA sequencing: Amplicons were purified using the MinElute PCR Purification Kit (Qiagen) and quantified with a Qubit 4 Fluorometer (Invitrogen). Fragment sizes and relative abundance of the amplicons were quantified using a Fragment Analyzer (Agilent Technologies). Libraries were then prepared using a protocol based on the TagSteady approach (Carøe & Bohmann, 2020). Library quantification was performed using a qPCR Real-Time System (Bio-Rad) to ensure accurate quantification. Following quantification, the size of the DNA fragments in the libraries was measured again using a Fragment Analyzer. Sequencing was performed on an Illumina MiniSeq System, using the Mid Output Kit, generating 8 million 150 bp paired-end reads.

Local plant database: A list of plants present in the study area and potentially consumed by vervet monkeys was established based on input from local botanical experts. For each of these plants, we verified their presence in GenBank. Those that were missing or had low alignment scores to the targeted metabarcode (Identity <98%, Coverage <70%, E-value >5) were identified morphologically in the field and collected for later sequencing, to create a custom database for the missing elements. In total, 37 plant species were sampled (Table S4). For each species, two pieces of one cm² leaf fragments were dried with silica gel beads in 20 mL HDPE scintillation vials (Carl Roth GmbH) and stored until DNA extraction, except for nine succulent plant species, which were preserved in 95% ethanol. DNA extraction was performed using the DNeasy Plant Mini Kit (Qiagen). PCR reactions were then carried out on the entire chloroplast trnL (UAA) intron region using primers c and d (Taberlet et al., 2007). They were performed in a total volume of 25 µL, containing: 1 × PCR Gold Buffer (Thermo Fisher Scientific), 2 mM of MgCl2, 0.2 mM of dNTPs, 0.5 ?M of forward and reverse primers, 1 U of AmpliTaq Gold 360 (Applied Biosystems), and 2 µL of template DNA diluted 250-fold. PCR cycling conditions consisted of an initial denaturation of 5 minutes at 95°C following by 45 cycles of 30 seconds at 95°C, 30 seconds at 5°C, and 60 seconds at 72°C, with a final elongation of 5 minutes at 72°C. Subsequently, purification and Sanger sequencing were conducted at Microsynth AG (Balgach, Switzerland). The sequences were aligned using MEGA11 v. 11.0.13 (Tamura et al. 2021; Figure S1).

Bioinformatics processing: Sequencing results from each library were handled separately using the OBITools package (Boyer et al., 2016). First, forward and reverse reads were assembled with a minimum quality score of 40 and assigned to each corresponding sample based on unique tags and primers. Identical sequences were then clustered. Sequences that could not be aligned, as well as those with fewer than 10 reads per library or those not meeting the appropriate primer length,h were removed. After correcting for PCR/sequencing errors, remaining clusters were reduced based on a 97% similarity threshold using the sumaclust algorithm (Mercier et al., 2013). Taxonomic assignment of sequences was based on three sources: the local plant database we created (see above), and two databases generated from in silico PCR simulations using our two sets of primers with a 95% similarity threshold. These simulations were queried against the GenBank database maintained by the National Center for Biotechnology Information (NCBI) using the ecoPCR software (Ficetola et al., 2010). Additionally, each operational taxonomic unit (OTU) identified for Sper01 and Vert01 was manually verified using BLAST on GenBank to confirm the results. Further sequence cleaning and analyses were conducted using the metabaR package (Zinger et al., 2021). In addition, because experiments involving peanuts as food rewards were conducted during the study period, we removed all DNA reads assigned to the o genus Arachis in the faecal samples collected the day after these experiments took place.