Appendix 1—figure 3. Microbial groups are derived from taxonomic data and experimental data.

An example is given using a subset of the ASVs in the Burkholderiaceae family. (A) Phylogenetic clustering of ASVs based on 16S V4 alignment. ASVs are annotated at the genus and species level using the SILVA database. Note that for some ASVs, annotation at the species level is missing, although the phylogenetic tree suggests divergent groups at the species level. Overall abundance in the dataset (B) of each ASV and differential abundance in response to the N treatment (C) were used in tandem with sequence-based clustering to group ASVs with similar features into microbial groups at sub-genus resolution (labeled in green). In this example, the genus Ralstonia constitutes a monophyletic cluster of ASVs which were all successfully assigned to the species R. pickettii (A). This uniform group is also reflected in relatively uniform abundance (B) and positive response to N treatment (C). On the other hand, most ASVs in the Burkholderia genus could not be annotated at the species level, even though the phylogeny suggests at least 4 distinct groups below the genus level. The first group, Burkholderia insecticola was identified at the species level without fail and once again, this is reflected in uniform abundances as well as a consistently negative response to N treatment. Within the next cluster two ASVs are assigned to Paraburkholderia caffeinilytica, and we assigned all other ASVs in the same cluster to the same species because they showed consistent abundance and response to treatment. In the remaining two clusters, no ASVs could be annotated at the species level, hence we assigned a number to the unassigned species (Burkholderia sp 1 and sp 2). Experimental data confirms that the two clusters should be treated as separate microbial groups because Burkholderia sp 2 is roughly 10 times as abundant as Burkholderia sp 1 and we observe opposite responses to N treatment.