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.

Sequence profile searches and structural comparisons reveal a vast superfamily of Smi1-related proteins As a first step to computationally characterize the Smi1/Knr4 protein, we analyzed it using the SEG program to identify potential globular regions in it (42) . This indicated the presence of a single globular domain that was then used as a seed in iterative sequence profile searches of the nr database with PSI-BLAST and JACKHMMER from the HMMER3 package. In addition to recovering other eukaryotic proteins with a homologous region, such as FBXO3 from animals, SKIP16 from plants and PGs2, a subunit of tubulin polyglutamylase complex, the search also recovered a large number of bacterial proteins such as the Bacillus subtilis YobK. Given the great diversity of sequences recovered prior to convergence from bacteria, we initiated transitive sequence profile searches with several distinct bacterial starting points to achieve maximal coverage in terms of detection. We also noted that a crystal structure for YobK has been solved by the joint structural genomics initiative (PDB: 2PRV). We used this structure as a query for structure similarity searches using the DALIlite program and recovered hits to four other homologous structures (3FFV, 2PAG, 2ICG, 3D5P; Z > 7.5). Of these, 3FFV was the structure of the earlier characterized protein Syd from E. coli which interacts with SecY, a key component of the Sec-dependent protein secretion system that traffics proteins across the bacterial inner membrane (43) (44) (45) . Consistent with this, we also found that Syd homologs were recovered with borderline e-values (e $ 0.03-0.05) in the above JACKHMMER and PSI-BLAST searches. Hence we included the Syd homologs in the profiles to further expand the relationships of the group of proteins homologous to Smi1/Knr4. At convergence, some of these searches also recovered with borderline e-values proteins (e $ 0.05) from certain DNA viruses such as FPV250 (gi: 9634920) from the fowl poxvirus, and the US22 family of proteins (e.g. US22, UL26, IRS1 and TRS1) from herpesviruses. To confirm the relationship of these proteins to Smi1 we used them in a profile-profile comparison search with the HHpred program against a library of HMMs created using the sequence of polypeptides in the PDB database as a query. These searches recovered the structures 2PRV, 3FFV and 2ICG as the best hits with significant P-values (P = 10 À4 to À8 ). Furthermore, examination of the hits produced by the viral proteins in profile-profile comparisons showed that most of the versions from herpesviruses possessed two tandem repeats of the domain homologous to Smi1. Additional transitive searches with these viral proteins revealed that homologous proteins are present in a number of distantly related or unrelated DNA viruses. Finally, the above searches also recovered hits to two distinct groups of proteins each with over 100 representatives in the nr database, predominantly from bacteria, typified respectively by CA_C3700 (gi: 15896931) from C. acetobutylicum and SGR_4389 (gi: 182438182) from S. griseus. Profileprofile comparisons with the HHpred program using alignments of each of these groups of proteins also confirmed their relationship to the Smi1-like proteins via recovery of significant hits (e = 10 À4 to À6 ) to HMMs generated using the sequences of 2PRV and 3FFV as best hits. Thus, it became clear that Smi1/Knr4 defines a large superfamily of conserved domains that is widespread in bac