Influenza hemagglutinin subtypes have different sequence constraints despite sharing extremely similar structures

Despite sharing highly conserved structures and functions, influenza hemagglutinin subtypes exhibit significantly divergent sequence constraints, with approximately half of their sites showing distinct amino acid preferences driven by differences in buried residues and rewired local interactions.

The Gist

Imagine the Influenza virus as a master thief trying to break into a house (your cells). The tool it uses to pick the lock is a protein called Hemagglutinin (HA). Think of HA as the thief's specialized key.

There are many different versions of this key, called "subtypes" (H1, H2, H3, H5, H7, etc.). Even though these keys look almost identical when you hold them in your hand—they have the same shape, size, and mechanism to open the door—they are made from completely different materials. In fact, if you compared the "recipe" (the genetic code) for an H3 key and an H7 key, they would only share about 40% of their ingredients.

The Big Question:
Scientists have always wondered: If these keys look the same and do the same job, does it matter if we change a tiny part of the recipe? If we swap one ingredient in the H3 key, does it break? What if we swap that same ingredient in the H7 key? Does it break there too?

The Experiment:
The researchers in this paper decided to play a massive game of "What If." They took the H7 key and systematically swapped out every single ingredient (amino acid) at every single spot, one by one. They then tested if the modified key could still open the door (enter the cell). They did this for H7, and then compared their new results to previous tests they had done on H3 and H5 keys.

The Surprising Discovery:
They expected that since the keys look the same, the rules for changing them would be similar. They thought, "If you break a hinge on a door, it doesn't matter what color the door is painted; it still won't work."

But they were wrong.

They found that for about half of the spots on the key, the rules were completely different.

• The "Safe" Swap: In the H3 version, you could swap a specific ingredient for almost anything, and the key would still work.
• The "Fatal" Swap: In the H7 version, swapping that exact same ingredient for the same thing would break the key completely.

Why does this happen? The "Interior Design" Analogy
The paper explains this using a great metaphor about interior design.

Imagine two houses that look identical from the outside.

• House A (H3) has a heavy, wooden support beam in the living room. If you try to replace that wood with glass, the house collapses. But if you replace a brick on the outside wall, the house is fine.
• House B (H7) looks exactly the same from the outside, but the architect decided to use a steel beam instead of wood in that same spot. If you try to replace the steel with glass, it also collapses. But here's the twist: because the beam is steel, you could replace the outside wall with glass without it falling down.

Even though the houses look the same, the internal connections are different.

• In the H3 virus, certain parts of the protein hold each other together using "magnetic" forces (hydrogen bonds). If you change the magnet, the whole thing falls apart.
• In the H7 virus, those same spots are held together by "glue" (hydrophobic interactions). If you change the glue, it breaks. But if you change the magnet in the H3 house to glue, it might actually work fine!

The "Buried" Secret
The scientists found that the most dramatic differences happened in the middle of the protein (the "buried" parts), not on the surface.

• Think of the protein as a Swiss Roll cake. The frosting on the outside (the surface) is very flexible; you can change the flavor of the frosting, and the cake still tastes fine.
• But the sponge inside (the buried core) is delicate. If you change the sponge in one version of the cake, it might need to be chocolate. If you change it in another version, it needs to be vanilla. If you try to force chocolate into a vanilla recipe, the cake crumbles.

Why Should You Care?
This is a big deal for public health and vaccines.

1. Vaccines: We often try to predict how a virus will mutate to escape our vaccines. This paper shows that we can't just look at one virus (like H3) and guess how another (like H7) will behave. They have different "personalities" and different rules for survival.
2. Future Pandemics: If a bird virus (like H5 or H7) jumps to humans, we can't assume it will mutate the same way human viruses do. It might take a completely different path to become dangerous.
3. Evolution: It teaches us that nature is clever. Over thousands of years, viruses can completely rewrite their internal "instruction manuals" while keeping the same "cover page" (structure) and the same "job description" (function).

In a Nutshell:
Two viruses can look and act exactly the same, but their internal "wiring" is so different that a change that helps one might kill the other. It's like two cars that look identical and drive at the same speed, but one runs on diesel and the other on electricity. If you try to put diesel in the electric car, it won't just run poorly; it will explode. Understanding these hidden differences helps us stay one step ahead of the flu.

Technical Summary

1. Problem Statement

Influenza A virus hemagglutinin (HA) proteins from different subtypes (e.g., H1–H19) exhibit a paradoxical evolutionary dynamic: they share highly conserved three-dimensional structures and biological functions (receptor binding and pH-dependent membrane fusion) yet possess low amino acid sequence identity (as low as ~40%).

• The Core Question: How does such extensive sequence divergence on a nearly fixed structural backbone affect the tolerance to further mutations?
• The Gap: While it is known that closely related homologs often share mutational tolerance, it is unclear how "rewiring" of residue interactions over long evolutionary timescales alters site-specific evolutionary constraints. Previous deep mutational scanning (DMS) studies focused on single subtypes (H3, H5), but a comparative analysis across divergent subtypes was lacking.

2. Methodology

The authors employed Pseudovirus Deep Mutational Scanning (DMS) to systematically measure the functional effects of mutations across three distinct HA subtypes: H3, H5, and H7.

• Experimental System:
  • Target: The study focused on the H7 HA from the A/Anhui/1/2013 strain (a candidate vaccine virus with the Q226L mutation, allowing binding to both $\alpha2$-3 and $\alpha2$-6 sialic acids).
  • Library Construction: Two independent mutant libraries were created covering ~95–96% of all possible single amino acid mutations in the H7 HA ectodomain.
  • Pseudovirus Platform: Lentiviral particles were engineered to express only the HA protein (and a matching Neuraminidase, N9, to prevent self-aggregation). These particles are replication-incompetent (single-round entry), ensuring biosafety while measuring cell entry efficiency.
  • Cell Lines: Experiments were conducted on 293 cells engineered to express exclusively $\alpha2$-3 linked sialic acids, exclusively $\alpha2$-6 linked sialic acids, or a mix of both.
• Data Generation:
  • Mutant libraries were transduced into cells. The relative abundance of each barcode (representing a specific HA variant) before and after selection (cell entry) was quantified via high-throughput sequencing.
  • Metric: The effect of each mutation was calculated as the log2 enrichment ratio relative to the wildtype.
  • Comparison: These new H7 data were compared against previously published DMS datasets for H3 and H5 HAs.
• Analytical Framework:
  • Amino Acid Preferences ($\pi_{r,a}$): Raw mutation effects were converted into normalized preferences for each amino acid at every site using a Boltzmann-like distribution: $\pi_{r,a} = \exp(x_{r,a}) / \sum \exp(x_{r,a'})$. This allows direct comparison across subtypes with different wildtype residues.
  • Divergence Metric: The Jensen-Shannon Divergence (JSD) was used to quantify the difference in amino acid preferences between subtypes at each site.
  • Structural Analysis: Protein structures (PDB: 4O5N for H3, 4KWM for H5, 6II9 for H7) were aligned to correlate sequence constraints with structural features (solvent accessibility, backbone RMSD, and interaction networks).

3. Key Contributions

• First Comparative DMS across Divergent Subtypes: This study provides the first comprehensive, side-by-side comparison of mutational tolerance across three distinct influenza HA subtypes (H3, H5, H7) with ~40–47% sequence identity.
• Quantification of Epistatic Shifts: It demonstrates that evolutionary divergence leads to massive shifts in site-specific constraints, not just in antigenic regions but throughout the protein structure.
• Mechanistic Insight into "Rewiring": The paper identifies specific structural mechanisms (e.g., hydrogen bond networks vs. hydrophobic packing) that explain why the same structural position tolerates different amino acids in different subtypes.
• Public Health Data: It generates a high-resolution functional map for the H7 HA, a subtype of significant pandemic concern, aiding in vaccine immunogen design and viral surveillance.

4. Key Results

A. Extensive Divergence in Sequence Constraints

• Approximately 50% of HA sites display significantly diverged amino acid preferences across the subtypes (H3-H5, H5-H7, H3-H7).
• This divergence is not limited to surface-exposed antigenic sites; it is pervasive across the protein.
• Cell Type Independence: The divergence in preferences is largely independent of the target cell type ($\alpha2$-3 vs. $\alpha2$-6), indicating that the constraints are driven by intrinsic protein folding, stability, and fusion mechanics rather than receptor specificity alone.

B. Structural Determinants of Divergence

• Buried Sites are Key: The most divergent sites are predominantly buried (relative solvent accessibility $\le$ 0.2) and have different wildtype amino acid identities between subtypes.
• Biochemical Type Matters: Among buried sites with different wildtypes, those where the biochemical nature of the residue changes (e.g., hydrophilic in H3 vs. hydrophobic in H5/H7) show the highest divergence.
• Backbone Stability: Surprisingly, the divergence is not correlated with backbone structural deviations (RMSD). The protein folds remain nearly identical, yet the sequence constraints have shifted dramatically.

C. Case Study: Rewiring of Interactions

• The authors highlight a specific cluster of buried sites (123, 176, 178).
  • In H3: These residues form a specific hydrogen bond network (E/K/Y), creating a rigid constraint where only specific polar/charged residues are tolerated.
  • In H5 and H7: This network is absent; the environment is hydrophobic (I/L/I or M/A/I). Consequently, H5 and H7 tolerate hydrophobic residues at these sites but reject the polar residues required by H3.
• Conclusion: Evolution has "rewired" the local interaction network. The same structural position requires different chemical properties to maintain stability and function depending on the subtype's evolutionary history.

D. Comparison to SARS-CoV-2

• Unlike SARS-CoV-2 spike variants (which diverged over ~6 years and show mostly similar preferences with minor epistatic shifts in receptor binding), HA subtypes (diverged over millennia) show pervasive shifts in constraints across the entire protein, particularly in the core stability regions.

5. Significance and Implications

• Limits of Predictive Modeling: The findings suggest that experimental data from one genetic background (e.g., H3) cannot reliably predict the effects of mutations in a highly diverged background (e.g., H7), even if the structure is conserved. This challenges the assumption that structural conservation implies functional constraint conservation.
• Vaccine and Surveillance Strategy: For emerging threats like H7N9, relying on data from seasonal H3 strains may be insufficient. Surveillance and vaccine design must account for subtype-specific sequence constraints.
• Protein Evolution Theory: The study provides empirical evidence that over long evolutionary timescales, proteins can accumulate "cryptic" epistatic changes. Small, fixed mutations can rewire interaction networks, fundamentally altering the fitness landscape for future mutations without changing the overall protein fold.

In summary, the paper demonstrates that structure does not dictate sequence constraints in a static manner; rather, the evolutionary history of a protein determines a unique "fitness landscape" where the tolerance for mutations is heavily dependent on the specific residue interactions established in that lineage.