A novel immunity system for bacterial nucleic acid degrading toxins and its recruitment in various eukaryotic and DNA viral systems

Zhang, Dapeng; Iyer, Lakshminarayan M.; Aravind, L.

Contextual information gleaned from gene neighborhoods in prokaryotes and domain architectures of proteins, when combined with sequence analysis, can be a powerful means of discerning protein function (47) . Indeed, this method has proven particularly effective in both function prediction and identification of new analogous systems, using the organizational syntax of tightly linked genes, in case of toxin-antitoxin and restrictionmodification systems (9, 13, 14, 23, 48) . To better understand the role of the SUKH domain we performed a detailed analysis of the gene-neighborhoods of all bacterial genes encoding a protein with this domain ( Figure 2 ). Consequently, we were able to identify at least three striking themes among the gene-neighborhoods of this superfamily. Firstly, across the bacterial phylogenetic tree we found numerous genomic neighborhoods that linked two or more adjacent genes encoding SUKH domain proteins. In certain cases, e.g. B. grahamii (gi: 240850988), we found tandem arrays with up to six paralogous SUKH superfamily genes ( Figure 2 ). We found that in several instances these paralogous versions are not closely related and in certain cases adjacent paralogs might belong to completely different SUKH groups. For example, we found combinations of genes encoding proteins belonging to the Smi1-like (SUKH-1), Syd-like (SUKH-2), SUKH-3 and SUKH-4 groups in the same neighborhood in several bacteria such as B. cereus MM3 and various Streptomyces species (Figure 2 ). This observation suggested that there appears to be selective pressure for the diversification of the linked SUKH domain proteins encoded in a gene neighborhood either via sequence divergence, or independent assembly of neighborhoods from distantly related paralogs of different groups. This situation, wherein multiple paralogous genes are linked together as tandem arrays in a neighborhood, is relatively rare in bacteria (49) . Given that products of genes linked in conserved gene-neighborhoods physically interact, it is possible that these paralogs interact to form a single complex (47) . On the other hand, the multiple paralogs could also represent different alternative versions of the same component of a system which is under selection to display diversity. Given the great variability in the numbers and types of paralogous versions of the SUKH superfamily encoded by these neighborhoods, we favor the later explanation in this case (details see below). The second major feature that emerged from the analysis of gene neighborhoods was the linkage of genes encoding diverse SUKH superfamily members to genes encoding different types of nucleases ( Figure 2 ). Among these, we observed multiple linkages in distantly related bacteria, such as B. thuringiensis and M. marina and S. griseoflavus, to genes for nucleases of the metal-dependent NucA family, which includes the well-studied S. marcescens secreted endonuclease (50) and the Anabaena non-specific endonuclease NucA, which degrades both RNA and DNA (51) . Another prominent linkage observed in several bacteria, such as M.infernorum, various Bacillus species and N. mucosa, was to genes encoding proteins with a HNH superfamily nuclease domain ( Figure 3 ). Sequence analysis showed that several of the HNH domains were related to similar nuclease domains found in previously studied bacteriocins such as pyocin AP41 of P. aeruginosa, Klebsiella klebicin B and colicin E8 of E. coli (52) . These linkages involved members of