5-6

[00:02.17] We then checked the association network.

[00:10.56] In the Klebsiella genus graph, the tailspike proteins are shown as white circles, and the serotypes are shown as colored circles.

[00:20.78] Generally, we found that each tailspike protein was linked with a single serotype, suggesting that these proteins tend to be specific for one type of glycan receptor.

[00:29.97] In situations where tailspike proteins linked with multiple serotypes, like in the cluster on the right, the associated glycans were usually similar.

[00:37.99] In this case, the K2 and K13 serotypes share an identical sugar backbone, differing only in one side chain.

[00:44.76] This suggests that the glycan backbone is likely the main determining factor for tailspike protein specificity.

[00:53.14] We observed similar trends across all five pathogens, implying that this specificity is a common trait of the tailspike proteins.

[01:00.17] Interestingly, this pattern persisted even when we examined associations across different pathogens.

[01:08.40] We found 26 cases where similar tailspike proteins appeared in both Salmonella and E. coli phages.

[01:16.99] In this example, the tailspike protein is a uniquely shared gene between the phages from two E. coli and two Salmonella.

[01:21.94] The glycans of these serotypes also share a similar backbone structure.

[01:31.13] The similarity of the glycan between Salmonella and E. coli in these cases might be due to their close evolutionary history or horizontal gene transfer.

[01:39.30] Similar patterns also emerged when we examined tailspike proteins in phages infecting E. coli and Klebsiella pneumoniae.

[01:45.50] Similar phages were found in E. coli and Klebsiella with the K63 serotype.

[01:54.97] It is also well-established that serotype-specific loci have been horizontally transferred from Klebsiella to E. coli multiple times during evolution.

[02:05.39] It is possible that bacteria can acquire new serotype loci through horizontal gene transfer, and as a result, they can gain or lose susceptibility to specific phages.

[02:11.08] This is a fascinating example of how phages and their bacterial hosts co-evolve.

[02:19.03] This dataset also presents an excellent opportunity to study the evolution and mechanisms of tailspike proteins.

[02:34.00] For example, the tailspike proteins that target E. coli O115 and O159 serotypes are clearly phylogenetically related, but they recognize glycans that differ by one monosaccharide in the backbone and one side chain.

[02:42.21] By comparing their sequences and structures, we can potentially uncover the differences that have led to shifts in their specificity.

[02:47.45] We are also interested in the application of this dataset in phage therapy.

[02:56.11] The associations we've identified can help predict the specificity of phages, which might be valuable for designing phage cocktails.

[03:05.74] In summary, we developed a deep-learning model called SpikeHunter, based on a large protein language model, that accurately classifies tailspike proteins.

[03:15.18] By analyzing prophages detected in five pathogen genomes, we built a comprehensive collection of tailspike protein sequences and predicted their substrates.

[03:22.62] We discovered the strong specificity of tailspike proteins for particular glycans, even across different species.

[03:33.45] Additionally, we found that the specificity of tailspike proteins is primarily determined by the glycan backbone motif, as they can tolerate some variations in the side chain.

[03:39.48] This work has significantly expanded our knowledge of the specificity of tailspike proteins.

[03:50.34] We hope that our resource and research will contribute to future studies in combating bacterial infections, understanding phage-host interactions, and advancing glycoscience.