Molecular diet analysis
Field and lab methods To examine bird diets, between April-June 2016 we caught 103 birds and successfully collected 41 faecal samples at 200m and 2000m elevations at the Bengal sites (Supplementary table S2). We caught an additional 107 birds and collected 82 faecal samples at 1200m, 2300m, 2700m and 3200m elevations in May 2017. All faecal samples were stored in 95% ethanol until DNA extraction. We used Qiagen QIAamp DNA Stool Mini kit to extract DNA with some minor modifications in the protocol, following Zeale et al. (2011).
To examine weaver ant diet, we collected items of food that the weaver ants were carrying to their nest between April-June 2016. We collected food items from 25 different colonies, either for an hour or until we had collected 10 items, whichever happened first. We extracted DNA from these samples by taking a small part of each of the prey items, crushing it and then using the Qiagen DNeasy blood and tissue kit and following the manufacturer’s protocol for DNA extraction. We carried out a 25μL PCR (polymerase chain reaction) to amplify a segment of 16S rDNA using the lns16S-short primer pair (5’-3’ TRRGACGAGAAGACCCTATA; ACGCTGTTATCCCTAAGGTA) described by Clarke et al. (2014). We used this primer set because it is invertebrate specific, so successful DNA amplification indicates presence of prey DNA in the bird faecal DNA extract. We sent successful PCR products to the Sequencing core at the University of Illinois at Chicago, Chicago, Illinois, for barcoding, pooling, size selection and sequencing on an Illumina Miniseq platform. All PCR products had strong primer dimer bands that were removed at the sequencing core before sequencing with a procedure that selected for fragments longer than 150 bp. (see Supplementary methods for further details)
Data analysis We followed the dada2 pipeline (Callahan et al. 2016) to construct an amplicon sequence variant (ASV) table and then created a FASTA file with all the unique sequences recovered from our samples. We used blastn in blastplus to BLAST this FASTA file against the NCBI nr database and used the NCBI-taxcollector script (Dias et al. 2014) to get detailed taxonomic information for the top hit of our BLAST results. After filtering out bacteria and other contaminating sequences, we estimated overlap between the diets of weaver ants and birds at low and mid-elevations at various taxonomic levels using EcoSimR (Gotelli et al. 2015) and visualized the overlap using venn diagrams and non-metric multidimensional scaling (NMDS) in R (R Core Team 2014). EcoSim R is used to estimate niche overlap between species and compare it to a null distribution of niche overlap given information on resource utilization (in columns) by each species (rows). The algorithm randomizes resource utilization for each species by reshuffling the row values and generates a null distribution of niche overlap. In our case, we considered birds at low elevations (N = 18), birds at middle elevations (N = 15) and weaver ants (N = 25) at low elevations as “species” and used the frequency of occurrence of different orders or families in diets as the “resources”. All the analyses were done in the R programming environment and in the shell using the R packages dada2 (Callahan et al. 2016), DECIPHER (Wright 2016), VennDiagram (Chen and Boutros 2011), ggplot2 (Wickham 2011), vegan (Oksanen et al. 2018) and EcoSimR (Gotelli et al. 2015). All scripts and details of the steps are available as supplements to this paper.