Extended Data Fig. 1: Co-occurrence of guanidine-I riboswitch, APC superfamily permease, and guanidinase genes on genome scaffolds from comammox Nitrospira group A (a) and B (b).

We identified a guanidine-I riboswitch in 67% (56/83) of the screened comammox genomes and found that the presence of the riboswitch was more widespread in comammox group A genomes (52/63; including the two metagenome-assembled comammox genomes from the wastewater treatment plants Ribe and Haderslev) than comammox group B genomes (4/20). The majority (75/83) of screened comammox genomes encoded urease. All comammox genomes with a guanidine-I riboswitch (56/56) encode an APC superfamily permease immediately downstream from the guanidine-I riboswitch, and the majority (46/56) encode guanidinase immediately downstream from the permease. It should be noted that the majority of comammox genomes are MAGs that are incomplete, preventing the triad of a guanidine-I riboswitch, an APC transporter, and a guanidinase from being located together in every genome. More specifically, for nine cases in which guanidinase is not found immediately downstream from the permease, the scaffold edge can be found within the permease or immediately downstream from the permease. In the final case, the genes immediately downstream from the riboswitch and permease are genes involved in the guanidine carboxylase pathway. Labels indicate the genome and scaffold accession IDs as well as genomic coordinates. For scaffold neighbourhoods with a riboswitch, the genomic coordinates indicate the coordinates of the riboswitch. For entries without a riboswitch, the accession IDs and end-point coordinates of the displayed genomic segment region are indicated. All genome regions visualized have been oriented so that the riboswitch and permease would be transcribed from left to right. Grey (permease) and light red (guanidinase) indicate genes that contain extensive frameshifts and therefore could be pseudogenes. We also screened all comammox genomes for the presence of the other three known guanidine riboswitches (mini-ykkC, ykkc-III and IV⁶³ and examined whether genes involved in guanidine degradation were located nearby (Supplementary Table 3). Two additional guanidine-responsive riboswitches, the mini-ykkC and ykkC-III riboswitches, were identified in five and two comammox genomes, respectively, and no genes known to be involved in guanidine degradation, transportation nor nickel incorporation were found in the vicinity of these. More generally, no obvious, dedicated nickel loading protein could be detected in any of the comammox genomes.