Predicting Binding Affinities for the Binding Domain of Hyperpolarization-Activated Cyclic Nucleotide-Gated Channel Isoforms Using Free-Energy Perturbation

This study utilizes all-atom molecular dynamics simulations and free-energy perturbation calculations to determine the absolute binding free energy of cAMP to the cyclic-nucleotide-binding domains of HCN channel isoforms 1–4, thereby elucidating the molecular mechanisms underlying their differential sensitivities to cAMP modulation.

The Gist

The Big Picture: The Heart's "Pacemaker" and Its Key

Imagine your heart and brain have a built-in rhythm section, like a drummer keeping time. These drummers are tiny proteins called HCN channels. They act as the gatekeepers that decide when your heart beats or when your brain fires a signal.

Usually, these gates are locked tight. But there is a special key called cAMP (a molecule your body makes) that can unlock them. When cAMP fits into the lock (a part of the channel called the CNBD), it helps the gate open, allowing electricity to flow and keeping your rhythm steady.

The Problem:
Your body has four slightly different versions of these "drummers" (called isoforms 1, 2, 3, and 4). They are like four different models of the same car. Even though they look similar, they react differently to the key.

• Some models (like HCN1 and HCN3) are very sensitive; a tiny bit of cAMP opens them wide.
• Others (like HCN2) are stubborn; they need a lot of cAMP to budge.

Scientists want to design drugs that can target only the specific models causing trouble (like in heart failure or epilepsy) without messing up the others. But to do that, they need to understand exactly why the locks are different.

The Experiment: A Digital "What-If" Machine

Since these channels are too small to see with a microscope while they are working, the researchers used a super-powerful computer simulation. Think of this as a digital wind tunnel for molecules.

1. Building the Models: They built four digital versions of the channel's lock (CNBD) based on real data and computer predictions.
2. The Test Drive: They dropped the "key" (cAMP) into each lock and watched what happened for a long time (simulating microseconds of real-time movement).
3. The "Free-Energy Perturbation" (FEP): This is the fancy science part. Imagine you have a lock and a key. To measure exactly how hard it is to pull the key out, you don't just yank it. Instead, you slowly turn the key into "ghost dust" and then back into a solid key, step-by-step. By measuring the energy it takes to do this "ghosting" process, the computer can calculate the exact binding strength (affinity) between the key and the lock.

What They Found

After running thousands of simulations, the team discovered some surprising things:

1. The Ranking of the Locks
They found that the "locks" on HCN1 and HCN3 are the strongest. The key sticks to them the best. HCN4 is in the middle, and HCN2 is the weakest (the key slips out the easiest). This explains why different parts of the body react differently to the same chemical signal.

2. The Secret Handshakes (The Residues)
Inside the lock, there are tiny molecular "hands" (amino acid residues) that grab the key. The researchers looked at which hands were doing the grabbing.

• The "Strong Grip": In the sensitive models (HCN1 and HCN3), a specific Arginine hand (a positively charged amino acid) holds the key very tightly. It's like a strong magnet.
• The "Plan B": In the stubborn model (HCN2), that Arginine hand is missing or doesn't work well. So, the lock compensates by using a different hand, a Glutamate, to hold on. It's like a person who lost their left hand learning to write perfectly with their right.

Why This Matters

Think of drug design like trying to pick a specific lock in a house full of identical-looking doors.

• Before this study: Scientists knew the doors were different, but they didn't know exactly which tumblers were inside.
• After this study: They now have a blueprint. They know that to make a drug that targets the "stubborn" HCN2 channel without touching the "sensitive" HCN1, they need to design a key that interacts with the Glutamate hand but ignores the Arginine hand.

The Takeaway

This paper is a deep dive into the molecular mechanics of how our heart and brain keep time. By using advanced computer simulations, the researchers figured out exactly how the "keys" fit into the "locks" of different channel types.

In simple terms: They mapped out the molecular fingerprints of four different heart/brain switches. This map is a crucial step toward inventing better medicines that can fix a broken heart rhythm or stop a seizure without accidentally turning off the lights in the rest of the house.

Technical Summary

1. Problem Statement

Hyperpolarization-activated cyclic nucleotide-gated (HCN) channels are critical "pacemaker" channels regulating rhythmic electrical activity in the heart and brain. Their dysfunction is linked to arrhythmias, epilepsy, and pain perception. While the drug Ivabradine targets HCN channels, therapeutic progress is hindered by a lack of understanding regarding:

• The precise mechanistic details of dual activation by voltage and the modulatory ligand, cyclic adenosine monophosphate (cAMP).
• The structural and dynamic reasons behind the varying sensitivities and binding affinities of cAMP across the four HCN isoforms (HCN1–4).
• Existing structural "snapshots" (static crystal structures) fail to capture the dynamic conformational landscape required to explain isoform-specific differences in channel gating and modulation.

2. Methodology

The authors employed a rigorous computational approach combining all-atom Molecular Dynamics (MD) simulations and Free-Energy Perturbation (FEP) to calculate absolute binding free energies.

A. System Preparation:

• Models: Four simulation systems were constructed representing the monomeric Cyclic Nucleotide-Binding Domain (CNBD) of HCN1, HCN2, HCN3, and HCN4 bound to cAMP.
  • HCN1: Derived from cryo-EM structure (PDB: 5U6P).
  • HCN2 & HCN4: Derived from X-ray crystal structures (PDB: 3U10, 3U11).
  • HCN3: Derived from an AlphaFold model (AF-Q9P1Z3-F1-v4) due to lack of experimental data; cAMP was inserted by aligning with the HCN1 template.
• Simulation Setup: Systems were solvated in TIP3P water with 0.15 M NaCl (physiological conditions) using CHARMM-GUI.
• Force Fields: CHARMM36m for proteins/ions and CGenFF for the cAMP ligand.
• Protocol:
  • Equilibration in NVT, followed by production in NPT (310 K, 1 bar).
  • Replicates: 10 replicates per isoform (1 µs aggregate simulation time per isoform).
  • Restraints: Distance and RMSD restraints were applied initially to maintain ligand position relative to the $\beta$-jelly roll, followed by linear lifting.

B. Trajectory Analysis & Replicate Selection:
Before FEP, trajectories were screened to select stable replicates based on:

1. Ligand Center of Mass (COM) Displacement: Must be < 2 Å.
2. Ligand RMSD: Must be stable and < 3 Å.
3. Lid Distance: Distance between specific residues in the $\beta$-jelly roll loop and the C-D helix turn (indicating pocket closure).
4. Helix B-C Angle: To assess domain stability.
5. Hydrogen Bonding: Stability of H-bonds with deep-pocket residues (Gly, Glu, Arg).

C. Free-Energy Perturbation (FEP):

• Goal: Calculate the absolute binding free energy ($\Delta G_{bind}$) of cAMP to the CNBD.
• Process: Alchemical transformation (annihilation) of the cAMP ligand in both the protein-bound state and the solvated state.
• Parameters: 20 windows ($\Delta\lambda=0.05$), soft-core potentials to handle van der Waals singularities, and Bennett Acceptance Ratio (BAR) for free energy calculation.
• Per-Residue Analysis: Used the pairinteraction function in NAMD to decompose binding energy contributions by specific residues, focusing on conserved Arginine (offset +47) and Glutamate (offset +38) residues.

3. Key Results

A. Binding Affinity Ranking:
The study calculated the absolute binding free energies, revealing a distinct hierarchy in cAMP affinity among the isoforms:

1. HCN1 & HCN3: Highest binding affinity (most negative $\Delta G$).
   • HCN1: $\approx -11.1$ kcal/mol
   • HCN3: $\approx -11.7$ kcal/mol
2. HCN4: Intermediate affinity.
   • HCN4: $\approx -8.9$ kcal/mol
3. HCN2: Lowest binding affinity.
   • HCN2: $\approx -6.9$ kcal/mol

B. Structural Dynamics:

• Stability: HCN2 and HCN4 exhibited more stable "lid" distances and helix angles compared to HCN1 and HCN3, which showed higher variability.
• HCN3 Specifics: Due to the reliance on an AlphaFold model and manual ligand insertion, HCN3 required more replicates to ensure accurate representation, showing slightly higher RMSD fluctuations.

C. Per-Residue Interaction Insights:
The decomposition of interaction energies highlighted a compensatory mechanism between isoforms:

• HCN1 & HCN3: Rely heavily on a strong interaction with a conserved Arginine residue (offset +47), which contributes significantly to binding energy ($\approx -20$ kcal/mol).
• HCN2 & HCN4: Show a loss of the strong Arginine interaction. Instead, they compensate with a stronger interaction with a conserved Glutamate residue (offset +38) ($\approx -23$ kcal/mol for HCN2).
• This suggests that while the overall binding mode is similar, the specific residue interactions driving affinity differ, explaining the functional divergence.

4. Significance and Contributions

• Mechanistic Insight: The study moves beyond static structural comparisons to provide a dynamic, thermodynamic explanation for why HCN isoforms have different sensitivities to cAMP.
• Validation of Computational Methods: Demonstrates that combining long-timescale MD with FEP can accurately predict relative binding affinities for membrane protein domains, even for isoforms lacking experimental structures (HCN3).
• Drug Design Implications: By identifying the specific residue interactions (Arg vs. Glu) that dictate affinity, the study provides a blueprint for designing isoform-selective drugs. This is crucial for developing therapeutics that target specific HCN isoforms (e.g., for cardiac vs. neurological applications) to minimize side effects.
• Resolution of the "Sequence vs. Conformation" Debate: The results suggest that differences in channel activation are not merely a result of amino acid sequence differences but are driven by specific, dynamic interaction networks within the CNBD.

In summary, this work establishes a molecular basis for the differential modulation of HCN channels by cAMP, offering a computational framework to guide the development of next-generation, isoform-specific therapeutics.