Integration of metabolome and microbiota datasets
The rhizosphere microbiota dataset was taken from our previous
experiment (Vescio et al., 2021), where additional maize plants from the
present experiment (and the same pot) were used. In the previous work,
libraries targeting the V4 region of the bacterial 16S rRNA were
obtained from DNA extracted from rhizosphere soil (McPherson et al.,
2018), and sequenced using an Illumina MiSeq platform (Illumina, San
Diego, CA, USA) using the 300PE chemistry. De-multiplexed forward and
reverse reads were merged using the PEAR 0.9.1 algorithm using default
parameters (Zhang et al. 2014). Data handling was carried out using
VSEARCH 2.14.2 (Rognes et al. 2016) to quality-filtered reads, discard
chimeric sequences, bin OTUs and assign taxonomy by querying the SILVA
database (v. 132) (Quast et al. 2012).
Data analysis was performed using R statistical software 3.5 (R Core
Team 2013) (Supplementary material 1). The microbiota dataset was
processed to remove singletons and OTUs generated from the amplification
of plastidial rRNA. This dataset was then normalized using DESeq2 (Love,
Huber & Anders 2014). The metabolome dataset was also processed to
remove those metabolites that were not found in at least 25% of the
samples. Then, a framework that considers both metabolomic and
metabarcoding datasets was built to finely investigate the interaction
between root exudates and rhizosphere microbial community. This
framework has been built using MOFA+ (Argelaguet et al. 2018).
MOFA+ captured three factors that explain a high proportion of variation
in both molecules and taxa in the three different treatments
(Supplementary material 1). This information allowed identifying, for
each treatment, the top 10% molecules and 0.1% microbial taxa that
mostly contribute to explain the variation due to the treatments. Then,
their relative abundance was correlated (Two tails, Pearson’s
correlation) and significant correlations (p<0.05) were
recorded.