Experimental design, wet- and dry-lab procedures
In August 2019 infested, wild ivy gourd fruits (Coccinia grandis(L.) Voigt) were collected in La Réunion at two sites: Bassin Plat (-21.321457, 55.485044, “BP”) and Manapany (-21.374674, 55.598343, “M”). They were brought to the laboratories at CIRAD (Centre de coopération internationale en recherche agronomique pour le développement) in Saint-Pierre, where collected fruits were placed in plastic boxes covered with fine-mesh clothes and containing sand as a pupation substrate. The fruits were kept in a climatic chamber (25 ± 1°C; 80 ± 10% HR; 12:12 light:dark photoperiod; with artificial light) until pupation. Boxes were regularly inspected for the presence of pupae. Pupae were transferred to 30 x 30 x 30 cm cages in the same climatic chamber. A few days later, adults emerged from the pupae. Adults were fed ad libitum with a diet of sugar and hydrolyzed yeast and had access to a wet sponge as a water source. We waited 3-4 weeks until females were mature and mated before the experiment was started. Adult gravid F0 females were randomly assigned to three experimental cages (four adult females per cage), each of them containing fruits of a different host plant: Coccinia grandis L. (Cucurbitaceae, “Co”), Cucumis sativus L. (Cucurbitaceae host, “Cu”), or Solanum melongena L. (Solanaceae host, “So”). In these cages, the females were let to oviposit. This resulted in six groups, one for each combination of site and host. After emergence, 25 third instar F1 larvae were taken from each host plant fruit and subjected to gut microbiome profiling. To evaluate whether microbiome differences between larvae reflect differences between adults, we also profiled the microbiomes of four F0adult females from each group. All specimens were preserved in 100% EtOH at -80 °C prior to microbiome profiling. As did not observe significant differences between whole body and gut microbiome profiles, we profiled whole body DNA extracts from larvae and adults. Laboratory procedures followed unless indicated otherwise. Larval microbiome profiles were obtained by pooling DNA extracts from five individual larvae, while adult F0 females were individually profiled. Metagenomic library preparation and sequencing were outsourced to Macrogen (https://www.macrogen.com). For library preparation, the Nextera XT kit was used (target: V3-V4 regions of 16S rRNA, insert size: 465 bp, primers: 341F and 805R . Libraries were sequenced on an Illumina MiSeq 150 PE platform (300 bp paired-end sequencing). A negative and a positive control (ZymoBIOMICS Microbial Community Standard D6300) were included to check for artifacts in library preparation and sequencing.
After verifying the quality of reads with FastQC , we used the DADA2 pipeline to remove primers and truncate reads (to a final length of 260 bp and 240 bp) and to fit the parametric error model for the identification and filtering of sequencing errors (2x106 reads used for model fitting). Filtering was based on maxEE = 1 in DADA2 (the complete DADA2 pipeline is available at https://github.com/wouterhendrycks/tephritid_microbiome_host_switch_project). For error estimation, the first two million sequences were used in the error model construction. Before pairing forward and reverse reads and filtering out chimeras, unique amplicon sequence variants (ASVs, 100% unique sequence identity) were extracted using the Bayesian classifier method of DADA2. The Silva v.132 database was used for the taxonomic classification of ASVs (percentage of identity = 97% similarity, p-min-consensus = 0.51). Chloroplast and mitochondrial sequences were removed using the R package decontam .
Core (stable associates) bacterial genera and ASVs were identified using the Abundance-Ubiquity method with a 50% minimal ubiquity threshold. This statistic evaluates whether a bacterial taxon is not more abundant than expected for its ubiquity. Significant deviations from expectation indicate that a bacterial taxon is not a stable core member . Statistical analyses were performed in R unless stated otherwise. Permutational Analysis of Variance (PERMANOVA) was conducted with PRIMER v7 using 9,999 unrestricted permutations of raw data).