Proteome-wide comparison of tertiary protein structures reveal extensive molecular mimicry inPlasmodium-human interactions

Abstract

Molecular mimicry is a strategy used by parasites to escape the host immune system and successfully transmit to a new host. To date, high-throughput examples of molecular mimicry have been limited to comparing protein sequences. However, with advances in the prediction of tertiary structural models, led by Deepmind's AlphaFold, it is now possible to compare the tertiary structures of thousands of proteins from parasites and their hosts, to identify more subtle mimics. Here, we present the first proteome-level search for tertiary structure similarity between the proteins from Plasmodium falciparum and human. Of 206 P. falciparum proteins that have previously been proposed as mediators of Plasmodium-human interactions, we propose that seven evolved to molecularly mimic a human protein. By expanding the approach to all P. falciparum proteins, we identified a further 386 potential mimics, with 51 proteins corroborated by additional biological data. These findings demonstrate a valuable application of AlphaFold-derived tertiary structural models, and we discuss key considerations for its effective use in other host-parasite systems.

Diverse Genetic Conflicts Mediated by Molecular Mimicry and Computational Approaches to Detect Them

“…Examples of structural mimics have been known for decades, despite the difficulty in detecting them prior to accurate ab initio protein structure prediction tools, through protein crystal structures and binding data. Recent advances in protein structure and interaction predictions are enabling the discovery of many more structural mimics, both across genomes and taxa [17,58,59].…”

“…Protein structural predictions and structural alignment algorithms have been in development for decades, and have been instrumental in identifying similar protein structures between pathogens and hosts [15]. Structural mimic candidates are recognizable through 1) alignment between mimic and target protein structures [87], 2) interactions between the mimic and its interacting partners [16], and 3) competitive interactions between the mimic and targets for the interacting partners (e.g., [17,59]). Through combining these well-developed conceptual frameworks with recent innovations in structural prediction, more structural mimics can be identified.…”

“…By the 2010s, the power of whole genome sequencing and accumulating genomic data in public databases made in silico screens for molecular mimicry possible [12][13][14][15]. The power of these approaches strengthens every year, as more genomes across the tree of life are sequenced and as new tools, such as AlphaFold [16], are developed for computationally predicting and comparing molecular function [17]. Full knowledge about the distribution of molecular mimicry in nature awaits genome mining studies across the tree of life.…”