Accurate interdomain contacts in mixed folded proteins from NMR-guided coarse-grained simulations

This study enhances the accuracy of coarse-grained molecular dynamics simulations for mixed folded-disordered proteins by integrating NMR-derived backbone dihedral terms, successfully generating conformational ensembles that correctly capture interdomain contacts and dynamic behaviors of the DNAJB6 chaperone.

The Gist

The Big Picture: The "Lumpy" Protein Problem

Imagine a protein as a machine. Some parts of the machine are rigid, solid gears (called folded domains) that do specific jobs. Other parts are like long, floppy strings of beads (called intrinsically disordered regions or IDRs) that wiggle around wildly.

In many important biological machines, these solid gears are connected by these floppy strings. The problem is that the strings don't just hang there; they often curl up and touch the gears in specific ways. This interaction is crucial for the machine to work (or stop working) correctly.

However, trying to see exactly how these floppy strings touch the gears is incredibly hard.

• The "Microscope" Problem: If you try to look at them with a standard microscope (like X-ray crystallography), the strings are too wiggly to get a clear picture.
• The "Computer" Problem: If you try to simulate them on a computer using every single atom (like a high-resolution video game), the computer gets overwhelmed. The strings move so fast and so far that it would take a supercomputer thousands of years to simulate just a split second of their movement.

The Solution: A "Coarse-Grained" Map

To solve the computer problem, scientists use Coarse-Grained (CG) simulations.

• The Analogy: Imagine you are trying to simulate a city. Instead of modeling every single brick, window, and person (which takes forever), you model every building as a single block and every person as a dot. This is a "Coarse-Grained" map. It's fast and shows you the big picture (where the traffic jams are), but it misses the tiny details (like which specific shop a person is walking into).

For a long time, these "block-and-dot" maps were great at showing the overall shape of the protein (how compact it is), but they were terrible at showing where the floppy string actually touches the gear. They often showed the string touching the wrong side of the gear, or touching itself instead.

The Breakthrough: Adding "Memory" to the Map

The authors of this paper (Billy Hobbs, Noor Limmer, and Theodoros Karamanos) wanted to fix this. They studied a specific protein called DNAJB6, which acts like a "chaperone" (a helper protein) that stops other proteins from clumping together.

This protein has a solid "J-domain" (the gear) and a floppy "GF-linker" (the string). They knew from experiments (NMR spectroscopy) that the string should be touching the front of the gear, but their standard computer maps kept showing it touching the back or curling up on itself.

Here is what they did:

1. The Clue: They looked at "chemical shift" data from NMR experiments. Think of this as a mood ring for the protein. The data told them exactly how "relaxed" or "tense" specific parts of the floppy string were. It revealed that certain parts of the string wanted to be straight (like a stretched-out arm) rather than curled up.
2. The Fix: They took their "Coarse-Grained" computer map and added a new rule. They told the computer: "Hey, based on the mood ring data, this specific part of the string needs to stay straight, and this other part needs to be a bit more rigid."
3. The Result: By forcing the string to follow these specific "mood" rules, the simulation suddenly changed. The floppy string stopped curling up on itself and reached out to touch the front of the gear, exactly where the experiments said it should.

The "Aha!" Moment: The Hidden Trap

When they looked at the new, improved simulation, they found something surprising.

Even though the protein was supposed to be in an "open" state (ready to do its job), the floppy string was actually curling around and blocking the gear's main interface.

• The Analogy: Imagine a door that is supposed to be open. But there is a long, sticky rope tied to the doorknob. Even though the door is technically "open," the rope is dragging across the floor and getting tangled in the door hinge, making it very hard to push the door all the way open.

The simulation showed that the "sticky" parts of the string (specifically some hydrophobic, oily amino acids) were grabbing onto the gear and creating a "partially closed" state. This explains why this protein is less efficient at its job than expected—it's constantly getting in its own way!

Why This Matters

This paper is a big deal for two reasons:

1. It's a New Tool: They showed that you don't need a super-expensive, high-resolution simulation to get the details right. You just need to feed the "low-resolution" map the right "local rules" derived from simple experiments. It's like taking a rough sketch of a city and adding a few traffic signs to make the navigation accurate.
2. It Explains Biology: They solved a mystery about how a specific protein works. They proved that the "floppy" part isn't just random noise; it has a specific, sticky personality that controls how the protein behaves.

Summary in One Sentence

The authors fixed a blurry, low-resolution computer map of a wiggly protein by adding specific "rules of behavior" derived from real-world experiments, revealing that the protein's floppy tail is actually a sticky trap that blocks its own function.

Technical Summary

1. Problem Statement

Proteins often contain a mix of folded domains and intrinsically disordered regions (IDRs), particularly low-complexity regions (LCRs). These IDRs are crucial for mediating protein-protein interactions but are notoriously difficult to characterize structurally due to their high dynamics and lack of a single stable conformation.

• The Challenge: While coarse-grained (CG) molecular dynamics (MD) simulations are used to extend timescales to capture IDR conformational ensembles, their minimalistic nature often fails to capture residue-specific secondary structure propensities.
• The Consequence: Standard CG models frequently produce inaccurate interdomain contact maps, leading to non-physical interactions between folded domains and disordered linkers. This limits the ability to understand how IDRs regulate function (e.g., autoinhibition).
• Specific Case: The study focuses on the DNAJB6 chaperone, specifically a construct containing a folded J-domain (JD) and a disordered glycine/phenylalanine-rich (GF) linker. Previous data suggested the "open" state (JD-GF) still exhibits reduced affinity for its partner Hsp70, implying hidden interdomain contacts, but the atomic-level nature of these contacts remained elusive.

2. Methodology

The authors employed a hybrid approach combining solution NMR spectroscopy with enhanced coarse-grained simulations.

A. Experimental Characterization (NMR)

• System: DNAJB6 JD-GF construct (open state) and JD-GF-α5 (closed/autoinhibited state).
• Relaxation Analysis: Measured $R_1$, $R_2$, and NOE at 600 MHz and 800 MHz.
  • Spectral Density Mapping: Extracted $J(0)$, $J(\omega_N)$, and $J(\omega_H)$ to characterize dynamics on the picosecond-to-nanosecond timescale.
  • Model-Free Analysis: Used extended model-free formalism to derive order parameters ($S^2$) and internal correlation times ($\tau_e$).
• Residual Dipolar Couplings (RDCs): Measured to confirm non-random coil behavior in the linker.
• Chemical Shifts: Used backbone chemical shifts (C$\alpha$, C$\beta$, CO, N, HN) to calculate secondary structure propensities via the TALOS-N algorithm.

B. Computational Modeling (CG Simulations)

• Framework: Used the CALVADOS force field (single-bead per residue) implemented in OpenMM.
• Baseline Simulation: Standard CALVADOS simulations reproduced global compaction (Radius of Gyration, $R_g$) but failed to capture specific interdomain contacts, predicting incorrect self-interactions within the GF linker.
• The Innovation (NMR-Guided Dihedral Terms):
  • The authors introduced a backbone dihedral potential into the CG model.
  • Parameterization: A residue-specific parameter ($\epsilon_d$) was tuned to match the NMR-derived secondary structure propensities of the GF linker.
  • Mechanism: This potential biases the single-bead representation to adopt specific dihedral angles (helical or extended) corresponding to the experimental chemical shifts, effectively encoding local structural preferences without requiring multi-bead representations.
  • Restraints: The folded J-domain was treated as a rigid body; dihedral terms were applied only to the GF linker (residues 75–98).

3. Key Results

A. NMR Dynamics Reveal Hidden Order

• Hydrophobic Rigidity: While the GF linker is largely disordered, specific hydrophobic residues (F87, E88, F89, F91, F93, R94) showed significantly reduced motions ($S^2 > 0.3$) on the nanosecond timescale.
• N-terminal Rigidity: Residues K70–N74 (N-terminus of GF) exhibited high rigidity ($S^2 > 0.85$) and helical propensity, behaving similarly to the folded J-domain, even in the "open" state.
• Implication: These residues are likely engaged in transient, long-range interactions with the J-domain.

B. Simulation Improvements

• Global Properties: The enhanced simulations maintained accurate global compaction ($R_g \approx 1.63$ nm), matching SAXS data ($1.73$ nm), confirming the dihedral term did not distort the overall ensemble size.
• Local Structure: The simulations successfully reproduced the NMR-derived secondary structure propensities (extended/$\beta$-strand-like conformations in the C-terminal GF), which standard CG models missed.
• Interdomain Contacts:
  • Standard CG: Predicted GF collapsing onto itself or the back of the J-domain.
  • Enhanced CG: Correctly predicted the GF linker interacting with the front face of the J-domain (specifically helices 2 and 3).
  • Mechanism: By enforcing an extended conformation via the dihedral potential, the hydrophobic GF residues were physically positioned to reach and contact the exposed hydrophobic patches on the J-domain.

C. Validation

• Contact Maps: The simulation-derived contact maps showed high correlation with experimental chemical shift perturbations (CSPs). Residues with high contact order in simulations matched those with reduced NMR dynamics.
• Controls:
  • JD-GSS: Replacing GF with a disordered GSS linker eliminated long-range contacts in simulations and CSPs, confirming the specificity of the GF interaction.
  • DNAJB1: Simulations of a related chaperone (DNAJB1) with a "less sticky" GF linker showed fewer interdomain contacts, consistent with weaker experimental CSPs.

4. Key Contributions

1. Methodological Advancement: Demonstrated that incorporating experimentally derived backbone dihedral potentials into single-bead CG models significantly improves the accuracy of interdomain contact predictions in mixed folded-disordered systems.
2. Structural Insight: Resolved the structural basis for the reduced Hsp70 affinity in the "open" JD-GF state. The study reveals that the linker samples partially closed states where hydrophobic GF residues interact with the J-domain, mimicking the autoinhibited state even without the C-terminal $\alpha$-helix.
3. Dynamic Interpretation: Linked NMR relaxation data (reduced $S^2$ values) directly to specific long-range interdomain contacts identified in the simulations, providing a unified view of local and global dynamics.

5. Significance

• Bridging Scales: This work bridges the gap between the computational efficiency of CG models and the atomic-level accuracy required to describe sequence-specific IDR behavior.
• General Applicability: The strategy is broadly applicable to any IDR with available NMR chemical shifts. Since chemical shifts are often the first NMR data obtained, this offers a low-barrier, high-impact route to refining CG simulations for signaling proteins, transcription factors, and phase-separating systems.
• Biological Implication: It challenges the binary view of "open" vs. "closed" states in chaperones, suggesting that "open" constructs may still sample autoinhibitory conformations via transient hydrophobic interactions, a mechanism critical for understanding chaperone regulation and aggregation prevention.