Microbiome diversity and predictive functional profiling
Three α diversity metrics were calculated: the Abundance coverage
estimator (ACE) to assess ASV richness, the Inverse Simpson index (ISI)
to assess ASV evenness and Faith’s phylogenetic diversity (FPD) to
investigate phylogenetic richness. In order to evaluate differences in
microbial α diversity between larvae raised on different host plants
and/or between sites, we used two-way ANOVAs which for
F1 larvae included Site (BP: Basin Plat, M: Manapany) as
a random factor and Host Plant (C. grandis, C. sativus, S.
melongena ) as a fixed orthogonal factor and for F0adults included Site (BP: Basin Plat, M: Manapany) as a random factor
and Parental Goup (group 1, 2, 3) as a random factor nested in Site.
ANOVA was implemented using the GAD package . Count data from which
diversity metrics were calculated were not normalised as all rarefaction
plots reached a plateau (S1). To ensure homoscedasticity, a log
transformation was applied to the ISI and a fourth root transformation
was applied to the ACE and FPD. Cochran’s C tests were used to test for
homogeneity of variances with the GAD package . Pairwise comparisons
were done by using a F test with Holm correction for multiple
comparisons using the phia package .
Before calculating β diversities, we first removed all ASV’s that
occurred in only one sample and we normalized counts by transforming
them into proportions to represent community structure .
Generalized Unifrac distances using the d5 matrix and Unweighted Unifrac
distances were calculated as a β diversity metric . As Unifrac distances
take into account the phylogenetic relationships, we constructed a
midpoint rooted maximum likelihood tree of the bacterial relationships
using a general time reversible substitution model in the program
Fasttree . Bacterial 16S sequences were aligned with the DECIPHER
algorithm .
Differences in microbiome β diversity between larvae raised on different
host plants and/or between sites were tested using a two-way
Permutational Analysis of Variance (PERMANOVA, ) with Site (BP: Basin
Plat, M: Manapany) as a random factor and Host Plant (C. grandis,
C. sativus, S. melongena ) as a fixed orthogonal factor.. The False
Discovery Rate (FDR) correction with experiment-wise p < 0.05
was used to correct for multiple testing. Differences between larvae
raised on different host plants were visualized with a Principal
Coordinate Analysis (PCoA) and 95% confidence ellipses were drawn using
the ggplot2 package .
To test for differential abundance of microbial genera (i.e. genera with
relatively more sequences assigned to them) among larvae raised on
different host plants and from different sites, we used ALDEx2 . ASVs
that could not be classified were assigned to distinct, unidentified
genera. Genera that showed differential abundance between two treatments
with an effect size difference between 1 and -1 were filtered out to
reduce the false positive rate . Significance was assessed by both the
Welch t test and the Wilcoxon rank sum test followed by
FDR correction
with experiment-wise p <
0.05 as FDR is better suited for exploratory analyses .
We predicted the possible metabolic functions of the microbiomes by
applying Tax4Fun2 to the 16S rRNA sequences
. We used the standard 16S rRNA
sequence reference dataset Ref100NR of Tax4Fun2 for the functional
prediction with a cut-off of 97% sequence similarity threshold to
reference genomes. The predicted KEGG (Kyoto Encyclopedia of Genes and
Genomes) orthology groups were further cleaned by removing all KEGG
groups that are unrelated to metabolic pathways involved in bacterial
metabolism/host-microbe symbiosis (e.g. removal of pathways related to
human diseases). Functional profiles were visualized using heatmaps with
the pheatmap package .