Structural principles within the human-virus protein-protein interaction network

Abstract

General properties of the antagonistic biomolecular interactions between viruses and their hosts (exogenous interactions) remain poorly understood, and may differ significantly from known principles governing the cooperative interactions within the host (endogenous interactions). Systems biology approaches have been applied to study the combined interaction networks of virus and human proteins, but such efforts have so far revealed only low-resolution patterns of host-virus interaction. Here, we layer curated and predicted 3D structural models of human-virus and human-human protein complexes on top of traditional interaction networks to reconstruct the human-virus structural interaction network. This approach reveals atomic resolution, mechanistic patterns of host-virus interaction, and facilitates systematic comparison with the host's endogenous interactions. We find that exogenous interfaces tend to overlap with and mimic endogenous interfaces, thereby competing with endogenous binding partners. The endogenous interfaces mimicked by viral proteins tend to participate in multiple endogenous interactions which are transient and regulatory in nature. While interface overlap in the endogenous network results largely from gene duplication followed by divergent evolution, viral proteins frequently achieve interface mimicry without any sequence or structural similarity to an endogenous binding partner. Finally, while endogenous interfaces tend to evolve more slowly than the rest of the protein surface, exogenous interfaces--including many sites of endogenous-exogenous overlap--tend to evolve faster, consistent with an evolutionary "arms race" between host and pathogen. These significant biophysical, functional, and evolutionary differences between host-pathogen and within-host protein-protein interactions highlight the distinct consequences of antagonism versus cooperation in biological networks.

Fig. 1.

The human-virus SIN. The network contains 3,039 endogenous (human-human) interactions among 2,435 human proteins alongside 53 exogenous (human-virus) interactions between 50 virus proteins from 36 viral species and their 50 human target proteins. See inset for symbol guide.

Fig. 2.

Endogenous-exogenous interface overlap in the human-virus SIN. (A) The most significant case of endogenous-exogenous interface overlap is plotted for each exogenous interaction. If an exogenous interface extensively overlaps with multiple endogenous interfaces, we plot the endogenous interface whose human binding partner shares the greatest structural similarity with the viral protein. If no such structural similarity exists, we plot the endogenous interface with the greatest Jaccard interface similarity to the exogenous interface. Filled points indicate confirmed common ancestry between viral protein and mimicked human binding partner. (B) An example of extensive interface overlap with significant sequence similarity. In the left rendering, based on PDB structures 2iw8 (41) and 1bi8 (42), human cell division protein kinase 6 (CDK6) is shown in complex with two of its endogenous binding partners, a D-type cyclin and a CDK inhibitor. In the right rendering, based on PDB structure 1g3n (43), the human binding partners have been removed, leaving their endogenous interfaces highlighted, and saimiriine herpesvirus 2 cyclin homolog is shown binding to the CDK. (C) An example of extensive interface overlap without significant sequence similarity. In the left rendering, based on PDB structure 3gxu (44), human ephrin-B3 is shown in complex with ephrin type-B receptor 3. In the right rendering, based on PDB structure 2vsk (45), the receptor has been removed, leaving its endogenous interface highlighted, and Nipah virus glycoprotein G is shown binding to ephrin-B3. (D) Interface overlap in the endogenous network is significantly more likely to be mediated by sequence homology than in the exogenous network (Fisher’s exact test, two-tailed P < 10^-7).

Fig. 3.

Functional properties of endogenous interfaces targeted by viral proteins. (A) Compared to all endogenous interactions in the SIN, those extensively overlapped by exogenous interactions are significantly less likely to be coexpressed (Fisher’s exact test, two-tailed P < 10^-6). (B) Distribution of interface residue occupancy (i.e., number of endogenous interactions in which the residue participates) for endogenous and mimicked interface residues. Compared to all endogenous interface residues in the SIN, residues mimicked by a viral protein participate in significantly more endogenous interactions (resampling-based one-tailed P < 0.001). (C) Human proteins binding to mimicked endogenous interfaces are significantly enriched for the GO slim term “Regulation of Biological Process” relative to generic human proteins with structural models (Fisher’s exact test, two-tailed P < 0.001).

Fig. 4.

Evolutionary properties of exogenous and endogenous interfaces. Levels of human-mouse amino acid sequence divergence are quantified for sets of exogenous and endogenous interface residues and compared with “surface residues” (i.e., residues which follow an interface residue-like unbound SASA distribution) from generic human proteins and human proteins targeted by viruses. Residues involved in exogenous interfaces consistently evolve faster than the protein surface. Statistical significance (***P < 0.001; ^nsP > 0.05) was determined from 1,000 rounds of rejection-resampling. Surface residues of both types are defined only during rejection-resampling, and so they have no absolute counts.