Bacteria use protein-protein interactions to infect their hosts and hijack fundamental pathways, which ensures their survival and proliferation. Hence, the infectious capacity of the pathogen is closely related to its ability to interact with host proteins. Here, we show that hubs in the host-pathogen interactome are isolated in the pathogen network by adapting the geometry of the interacting interfaces. An imperfect mimicry of the eukaryotic interfaces allows pathogen proteins to actively bind to the host’s target while preventing deleterious effects on the pathogen interactome. Understanding how bacteria recognize eukaryotic proteins may pave the way for the rational design of new antibiotic molecules.

To fulfil their function, proteins need to interact with each other forming complexes. Understanding how pathogen proteins bind their host counterparts is important to explain how bacteria can infect, survive and proliferate inside cells. To achieve that, pathogen proteins mimic eukaryote interfaces to interact with the host. However, we discovered that such mimicry is imperfect, and pathogen proteins display particular features that are not found in eukaryotic complexes. This imperfect mimicry would allow pathogen proteins to actively bind to the host targets while preventing deleterious effects on the pathogen network. Indeed, we show that highly connected proteins (hubs) in the host-pathogen networks are mostly isolated in the pathogen network. The existence of imperfect mimicry opens the door to the design of new molecules aimed to target host-pathogen complexes with reduced side effects. Hence, in the long term, our results may lay the foundation of a new class of antimicrobials.

Citation: de Groot NS, Torrent Burgas M (2020) Bacteria use structural imperfect mimicry to hijack the host interactome. PLoS Comput Biol 16(12): e1008395. https://doi.org/10.1371/journal.pcbi.1008395

Editor: Arne Elofsson, Stockholm University, SWEDEN

Received: July 7, 2020; Accepted: September 23, 2020; Published: December 4, 2020

Copyright: © 2020 de Groot, Burgas. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.

Data Availability: All relevant data are within the manuscript and its Supporting Information files.

Funding: Ministerio de Economía y Competitividad (MEC):Marc Torrent RYC-2012-09999; Ministerio de Economía y Competitividad (MEC):Marc Torrent SAF2015-72518-EXP; Ministerio de Economía y Competitividad (MEC):Marc Torrent SAF2017-82158-R; European Society of Clinical Microbiology and Infectious Diseases (ESCMID):Marc Torrent ESCMID-2016 The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Competing interests: The authors have declared that no competing interests exist.

In nature, bacteria do not exist in isolation but within communities of multiple species that require the bacteria to communicate and organize [1]. In particular, pathogenic bacteria need to interact with host cells to ensure their survival and proliferation [2]. Most of these interactions are mediated by protein effectors that are involved in microbial virulence [3, 4]. These effectors, delivered to the host by secretion systems [5], outer membrane vesicles [6] or other specific mechanisms [7], allow bacteria to bind to host cells, replicate and spread within the host and subvert its immune system.

The collection of all these interactions between the host and the pathogen are known as the host-pathogen interactome. In the interactome, proteins with high connectivity degree are known as interactome “hubs” and are associated with essential and conditional phenotypes [8, 9]. The correlation between node degree and gene essentiality is known as the centrality-lethality rule [10] and has been observed in many organisms and interspecies interactomes [11]. In infectious diseases, pathogens selectively target hubs in the host interactome to rewire specific pathways for their own benefit [8]. The elucidation of these interactions is of utmost importance to understand how pathogens hijack the host systems.

Protein interactions are mostly mediated by domain-domain pairs. However, the number of protein folds [12] and interface geometries [13] is limited and these structures are frequently ‘reused’, allowing a single protein to bind many partners. This recycling strategy is necessary because the interaction space is limited [14] and the distribution of domains in bacteria and eukaryotes has major differences. For example, the most common folds in prokaryotes are those involved in housekeeping functions, such as P-loop-containing NTPases and TIM barrels, whereas the distribution in eukaryotes is dominated by domains with regulatory functions, such as protein kinases and β-propellers [15]. Hence, these constraints set a diversity upper limit and impose restrictions on how bacteria can bind eukaryote proteins.

Despite recent advances in the characterization of host-pathogen interactions, the structural knowledge of these protein complexes is very limited and largely restricts our ability to understand pathological systems. Here, we analyzed the degree centrality of bacterial proteins in the pathogen and the host-pathogen interactomes and investigate the structural characteristics of the interactions involved. We observed that hubs in the host-pathogen interactome are largely isolated in the pathogen interactome. This behavior can be explained by an imperfect mimicry of host interfaces that allow bacteria to minimize the toxicity of these proteins by restricting the number of interactions while maintaining an affinity for host proteins.

Because effectors target specific host interactions, their interfaces need to be fine-tuned to interact with a precise set of targets [16]. If proteins are not controlled for promiscuous interactions, this can lead to toxic effects [17]. In the case of effectors, optimized interfaces must be controlled to avoid undesired pathogen self-interactions that may compromise cell performance. A safe ward strategy against unwanted interactions is the timely expression of proteins under certain circumstances [18]. If bacterial effectors were only produced when bacteria are in close contact with the host, their deleterious effects could be minimized. However, this does not seem to be the case. The expression of effectors delivered by secretion systems is not triggered upon infection in cases where data is available (S1A Fig). Also, other mechanisms of delivering virulence factors, such as extracellular vesicles, encapsulate large amounts of proteins from the bacteria cytoplasm and periplasm and do not specifically select their cargo [19]. This strategy would also be non-optimal as detection and killing of bacteria by the innate immune cells is very fast [20] while altering the expression of a protein may take longer [21].

In this context, either these proteins are physically isolated from the rest of the proteome or are integrated into the network in a controlled manner. In eukaryotes, spurious interactions are prevented by segregating proteins into different compartments. However, bacteria lack these compartments. A possible alternative would be the use of protein condensates [22], non-membrane bound structures formed by liquid-liquid phase separation [23]. These condensates would allow proteins to be relatively isolated from the cell milieu, but ready to be delivered when required. However, we did not find a clear difference in condensate propensity between secreted effectors and the rest of the proteome (S1B Fig). Besides, bacteria are more proficient in forming solid aggregates, that are less dynamic than liquid condensates [24, 25].

Based on these evidences, we hypothesized that effectors could be integrated into the bacteria network in such a way that they were not deleterious when pathogens replicate outside the host but can be effectively deployed upon infection. By isolating effectors from the pathogen network, they would have less control over the interactome, minimizing the side effects. In agreement with this reasoning, we observed that effector proteins are significantly depleted of pathogen hubs (Fig 1A). These results were also validated in an orthogonal database to control for potential methodological bias (Fig 1A, see Materials and Methods for a detailed description of the control database).

(A) We computed the degree of interaction for all proteins in Y. pestis, B. anthracis and F. tularensis and asked whether the number of hubs (here defined as the 5% of most connected proteins) differ between effectors and non-effectors. In all cases, the number of hubs observed is significantly lower than expected. Similar results were obtained in the control database. The p-values were computed using a χ2-square test of independence to assess the probability of observing such a large discrepancy (or larger) between observed and expected values. Effectors were defined using EffectiveDB with a stringency threshold of 0.95 [52]. All comparisons remain statistically significant in the threshold interval 0.90–0.99 (p