Structure and Topology Prediction of Phage Adhesion Devices Using AlphaFold2: The Case of Two Oenococcus oeni Phages

Abstract

Lactic acid bacteria (LAB) are important microorganisms in food fermentation. In the food industry, bacteriophages (phages or bacterial viruses) may cause the disruption of LAB-dependent processes with product inconsistencies and economic losses. LAB phages use diverse adhesion devices to infect their host, yet the overall picture of host-binding mechanisms remains incomplete. Here, we aimed to determine the structure and topology of the adhesion devices of two lytic siphophages, OE33PA and Vinitor162, infecting the wine bacteria Oenococcus oeni. These phages possess adhesion devices with a distinct composition and morphology and likely use different infection mechanisms. We primarily used AlphaFold2, an algorithm that can predict protein structure with unprecedented accuracy, to obtain a 3D model of the adhesion devices’ components. Using our prior knowledge of the architecture of the LAB phage host-binding machineries, we also reconstituted the topology of OE33PA and Vinitor162 adhesion devices. While OE33PA exhibits original structures in the assembly of its bulky adhesion device, Vinitor162 harbors several carbohydrate-binding modules throughout its long and extended adhesion device. Overall, these results highlight the ability of AlphaFold2 to predict protein structures and illustrate its great potential in the study of phage structures and host-binding mechanisms.

Present Impact of AlphaFold2 Revolution on Structural Biology, and an Illustration With the Structure Prediction of the Bacteriophage J-1 Host Adhesion Device

“…From a structural biology view point, due to their flexibility, these machineries can rarely be studied as a whole (for exceptions see ( Sciara et al, 2010 ; Veesler et al, 2012 )) but mostly as isolated modules. However, we recently showed that the structure of a long and flexible Tal from the Oenococcus oeni phage Vinitor 162 could be predicted with AlphaFold2 ( Goulet and Cambillau, 2021 ).…”

“…Stunning results produced by AlphaFold2 at the CASP14 (Critical Assessment of Techniques for Protein Structure Prediction) contest indicated that deep learning-based methods are now able to predict protein structures with an accuracy comparable, in most cases, to that of experimental structures ( Pereira et al, 2021 ). This important achievement offers great perspectives to the life science community, including experimented structural biologists as well as untrained biologists, to address challenging biological questions ( Goulet and Cambillau, 2021 ; Tunyasuvunakool et al, 2021 ). In particular, the AlphaFold Protein Structure Database ( https://www.alphafold.ebi.ac.uk ), developed by DeepMind and EMBL-EBI, offers open access to structure predictions for the human proteome, for proteins from key model organisms (e.g., Mus musculus , Escherichia coli , Arabidopsis thaliana ), and for proteins from organisms related to global health ( Mycobacterium tuberculosis , Staphylococcus aureus , Trypanosoma cruzi ) ( Tunyasuvunakool et al, 2021 ; Varadi et al, 2022 ).…”

“…The reliability of predicted protein structures is such that they can be used as starting models for molecular replacement in X-ray crystallography, for the interpretation of cryo-electron microscopy (cryoEM) 3D reconstructions, and for the rational design of mutations and functional assays ( Kryshtafovych et al, 2021 ; Slavin et al, 2021 ; Alerasool et al, 2022 ; Ivanov et al, 2022 ; McCoy et al, 2022 ; Paul et al, 2022 ). Moreover, AlphaFold2 can go a step further and predict the structure of large and flexible machineries, thereby circumventing intrinsic limitations of experimental methods ( Goulet and Cambillau, 2021 ).…”

Structure Prediction and Analysis of Hepatitis E Virus Non-Structural Proteins from the Replication and Transcription Machinery by AlphaFold2

“…In particular, it is well adapted to predict structures of long, flexible and multi-domain proteins that cannot be analyzed as a whole using experimental approaches such as X-ray crystallography and electron microscopy. Recently, we applied this method to bacterial viruses (bacteriophages) to determine the structures of their whole host-binding machineries with several multi-domain proteins [ 31 , 32 ]. Moreover, as compared with homology modeling, AF2 produces structures that do not suffer from sequence bias, with an estimate of the prediction reliability for each protein residue given by a confidence score, the predicted local distance difference test (pLDDT, from 0 to 100).…”

Needle in a Whey-Stack: PhRACS as a Discovery Tool for Unknown Phage-Host Combinations

“…This may be an issue in cases where phages do not encode an easily recognizable RBP-encoding gene. To overcome this (at least for phages binding to carbohydrate receptors), identified carbohydrate binding domains (CBMs), responsible for phage host adhesion ( 49 ) and which have been shown to be host specific ( 50 , 51 ), could be employed.…”