Divergent landscapes of positive and negative selection signatures across residue-resolved human-virus protein-protein interaction interfaces

By integrating human-virus protein-protein interaction maps with residue-resolved contact data, this study reveals that positive and negative selection signatures exhibit distinct spatial patterns across virus-targeted host proteins, with positively selected residues clustering more prominently on interfaces shared between viral and endogenous partners, thereby highlighting these "mimic-targeted" sites as focal points of adaptive evolution.

The Gist

Imagine your body's proteins as a bustling city. Some parts of this city are the core infrastructure (like power plants and water pipes) that must never change, or the whole city collapses. Other parts are the front doors and windows, which are constantly being tested by intruders.

In this scientific paper, the authors are studying how these "front doors" (protein interaction interfaces) of human cells evolve when they are under attack by viruses. They are looking at a high-stakes game of "Evolutionary Chess" between humans and viruses.

Here is the story of what they found, broken down into simple concepts:

1. The Two Rules of the Game

Every part of a protein follows one of two rules:

• The "Do Not Touch" Rule (Negative Selection): Some parts of the protein are so critical for keeping the cell alive that they cannot change. If they change, the cell dies. These are like the foundation of a building; you can't repaint them or move them without risking a collapse.
• The "Adapt or Die" Rule (Positive Selection): Other parts are under attack by viruses. To survive, these parts must change to trick the virus or block its entry. These are like the locks on your front door; if a burglar learns how to pick them, you have to install a new, better lock immediately.

2. The Big Discovery: Clumps vs. Sprinkles

The researchers mapped out exactly where these changes happen on the surface of the proteins. They found a fascinating pattern:

• The "Adaptation Clumps": The parts that are changing to fight viruses (Positive Selection) aren't scattered randomly. They are clumped together in specific neighborhoods. Imagine a neighborhood where everyone is painting their houses different bright colors to confuse the burglar. The changes are concentrated in one spot.
• The "Stability Sprinkles": The parts that must stay the same (Negative Selection) are spread out evenly across the protein, like sprinkles on a donut. They are everywhere, holding the structure together.
• The "No-Go Zone": The "Adaptation Clumps" and the "Stability Sprinkles" rarely touch. It's as if the city planners decided to keep the colorful, changing houses far away from the fragile foundation to avoid causing chaos.

3. The Two Types of Front Doors

The researchers realized that not all "front doors" are the same. They found two distinct types of interfaces on human proteins:

Type A: The "Exclusive Viral Door" (Exogenous-Specific)

This is a door that only viruses use to get in.

• What happens here: The human body builds a fortress wall in just one small corner of this door. It changes only that specific spot to block the virus, while the rest of the door stays exactly the same.
• The Metaphor: It's like putting a heavy, reinforced steel plate over just the keyhole of a door, leaving the rest of the wood untouched.

Type B: The "Shared Door" (Mimic-Targeted)

This is the most interesting type. It's a door that viruses use, but our own human proteins also use to talk to each other. The virus tries to mimic our own proteins to sneak in.

• What happens here: Because the virus is pretending to be us, we can't just change one small spot (like the keyhole). If we change that, we might break our own internal communication. Instead, the human body changes the entire door surface.
• The Metaphor: Imagine a security guard (the virus) wearing a fake uniform (mimicry) to walk through a checkpoint. You can't just change the lock on the gate; you have to change the entire uniform policy, the badge design, and the handshake routine for everyone. The changes are spread out evenly across the whole interface.

4. The "Super-Connector" Effect

Here is the most surprising finding: The "Shared Doors" (Type B) act like a hub or a magnet.

Even though the changes on the "Shared Door" are spread out, they seem to pull the "Exclusive Viral Doors" (Type A) into their orbit. The researchers found that the "Adaptation Clumps" on the Exclusive Doors are actually more connected to the Shared Doors than they are to other parts of the Exclusive Door itself.

The Metaphor: Think of the Shared Door as a busy town square. Even if you live in a quiet suburb (the Exclusive Door), the activity in the town square influences how you decorate your house. The "Shared Door" is the focal point of the evolutionary war, dragging the changes from the rest of the protein toward it.

Why Does This Matter?

This study changes how we understand the war between humans and viruses.

1. It's not random: Evolution isn't just throwing changes everywhere. It's a highly organized strategy.
2. Context is King: How a protein evolves depends entirely on who is trying to bind to it. If it's just a virus, we change a small spot. If it's a virus pretending to be us, we change the whole surface.
3. The "Shared" Danger: The most dangerous spots for viruses to attack are the ones where they mimic our own biology. These spots become the epicenters of evolutionary change, driving the entire protein to adapt in a coordinated way.

In a nutshell: The human body is a master architect. When a virus attacks a specific spot, we reinforce that spot. But when a virus tries to sneak in by pretending to be us, we redesign the whole neighborhood to keep the peace, all while making sure the foundation of the house never cracks.

Technical Summary

1. Problem Statement

Proteins targeted by viruses evolve under dual selective pressures: negative selection to preserve essential endogenous functions (e.g., binding to human partners) and positive selection to adapt and evade viral engagement. While it is established that virus-targeted host proteins contain both types of selected residues, the spatial organization of these pressures at the residue level within Protein-Protein Interaction (PPI) interfaces remains poorly understood.

Key gaps addressed by this study include:

• Whether positively selected residues are clustered within specific sub-regions of an interface or distributed uniformly.
• How positive and negative selection pressures are spatially arranged relative to one another on the same interface.
• How these patterns differ across distinct interaction contexts: exogenous-specific (virus-only), endogenous-specific (human-only), and mimic-targeted (shared by both virus and human partners).

2. Methodology

The authors developed a multiscale framework integrating evolutionary rates with structural contact maps to analyze virus-targeted human proteins.

• Data Source: Utilized a previously established human-virus structural interaction network containing experimentally determined 3D structures (from PDB) or homology-based templates.
• Residue-Level Evolutionary Rates: Calculated site-specific $dN/dS$ ratios using the HyPhy Fixed Effects Likelihood (FEL) method. Orthologs from 19 other mammals were used for comparative analysis.
• Residue Classification: Interfacial residues were categorized based on $dN/dS$ values:
  • Pos (Positively Selected): $dN/dS > 1$.
  • SNeg (Strongly Negatively Selected): $dN/dS < 0.1$ (a stricter threshold than the conventional $<1$ to ensure balanced comparison).
  • Other: $0.1 \le dN/dS \le 1$.
• Contact Graph Construction:
  • Constructed a residue-residue contact graph where nodes are interfacial residues and edges represent atomic contacts (distance $< 5$ Å).
  • Edges were classified by the types of residues they connect (e.g., Pos-Pos, SNeg-SNeg, Pos-SNeg).
• Statistical Analysis:
  • Observed-over-Expected (O/E) Ratios: Computed by comparing the actual count of specific contact types against a null distribution generated by 10,000 intra-protein node-label permutations (shuffling residue labels while keeping the graph structure intact).
  • Sub-interface Analysis: The contact graph was partitioned into sub-interfaces (exogenous-specific, endogenous-specific, mimic-targeted) to analyze context-specific patterns.
  • Cross-interface Analysis: Evaluated clustering of positive selection between different sub-interface types (e.g., between mimic-targeted and exogenous-specific regions).

3. Key Contributions

• Residue-Resolution Mapping: First study to map selective pressures onto PPI interfaces at single-residue resolution rather than averaging rates across entire interfaces.
• Spatial Segregation Discovery: Demonstrated that positive and strong negative selection are not only distinct but spatially segregated, with positive residues clustering and negative residues dispersing.
• Context-Dependent Evolutionary Strategies: Revealed that the spatial distribution of positive selection depends on the interaction context. Specifically, mimic-targeted interfaces exhibit a unique "distributed adaptation" pattern, unlike the "focused adaptation" seen in exogenous-specific interfaces.
• Mimic-Targeted Interfaces as Evolutionary Hubs: Identified mimic-targeted interfaces as focal points that drive adaptive clustering not just within themselves, but also across boundaries into neighboring exogenous- and endogenous-specific regions.

4. Key Results

A. Global Spatial Organization

• Clustering of Positive Selection: Pos-Pos contacts are significantly enriched (O/E ratio $\approx 1.40$), indicating that positively selected residues form spatial clusters.
• Dispersion of Negative Selection: SNeg-SNeg contacts show little to no enrichment (O/E $\approx 1.06$), indicating strongly negatively selected residues are broadly and uniformly distributed.
• Spatial Segregation: There is a significant depletion of Pos-SNeg contacts (O/E $\approx 0.74$), proving that adaptive sites and highly constrained sites occupy distinct structural niches within the same interface.

B. Interface Context Differences

• Exogenous-Specific Interfaces: Positively selected residues are clustered in specific sub-regions. This suggests a strategy of "focused adaptation" to evade specific viral binding sites while leaving the rest of the interface constrained.
• Mimic-Targeted Interfaces: Positively selected residues are uniformly distributed across the entire interface (no significant Pos-Pos enrichment). This suggests a "distributed adaptation" strategy where the host spreads adaptive changes across the shared interface to evade viral mimicry without compromising the essential endogenous binding function.
• Evolutionary Rates vs. Spatial Patterns: Despite having similar proportions of positively selected residues and similar overall evolutionary rates, exogenous-specific and mimic-targeted interfaces exhibit fundamentally different spatial organizations.

C. Cross-Interface Clustering

• Focal Points: The study found that Pos-Pos contacts are significantly enriched between mimic-targeted interfaces and their neighboring exogenous- or endogenous-specific interfaces.
• Gradient Effect: The clustering between mimic-targeted and other interfaces is stronger than the clustering within the exogenous/endogenous interfaces themselves. This implies that mimic-targeted interfaces act as anchors or focal points that coordinate adaptive changes across the broader interaction landscape.

D. Case Studies

• TRAIL-R2: Pos residues cluster at the viral binding tip (UL141), while SNeg residues are dispersed to maintain endogenous TRAIL binding.
• TAP2: Pos residues cluster in the inner cavity (ICP47 binding site), while SNeg residues are dispersed to maintain the TAP1-TAP2 complex stability.

5. Significance

This paper provides a high-resolution view of the evolutionary "arms race" between hosts and viruses. It challenges the view of PPI interfaces as monolithic units, revealing them as heterogeneous landscapes where adaptive flexibility and functional constraint are intricately balanced.

• Mechanistic Insight: It elucidates how hosts reconcile the need to maintain essential functions (negative selection) with the need to evade pathogens (positive selection) by spatially separating these pressures.
• General Strategy: The discovery of "distributed adaptation" in mimic-targeted interfaces suggests a general host strategy to counteract viral molecular mimicry, which may be applicable to understanding immune evasion and drug resistance in other contexts.
• Methodological Advance: The integration of residue-resolved evolutionary rates with structural contact graphs offers a powerful new framework for studying protein evolution, moving beyond protein-wide averages to uncover localized evolutionary dynamics.

In summary, the study demonstrates that the host's evolutionary response to viral pressure is not random but is structurally coordinated, with mimic-targeted interfaces serving as critical hubs that orchestrate adaptive changes across the protein interaction network.