Multicomplex Integrative Structural Modeling of a Human Histone Deacetylase Interactome

This study employs integrative structural modeling combining crosslinking mass spectrometry and AlphaFold to elucidate the assembly mechanisms of HDAC1/2 within NuRD, SIN3, and CoREST complexes, revealing that the intrinsically disordered C-terminal domain of HDAC1 folds into alpha helices upon complex formation.

The Gist

The Big Picture: Solving the "Lego" Puzzle of Life's Switches

Imagine your body is a massive, bustling city. Inside every cell, there are millions of tiny switches that turn genes on or off. These switches control everything from how you grow to how your cells repair themselves.

Two of the most important "switch operators" in this city are proteins called HDAC1 and HDAC2. Think of them as the master electricians. Their job is to tighten or loosen the wiring (DNA) to stop electricity (gene activity) from flowing.

But here's the problem: These electricians don't work alone. They are part of huge, complex construction crews (called NuRD, SIN3, and CoREST). For years, scientists knew what the electricians looked like in their "work boots" (the part that does the actual job), but the rest of their bodies were a mystery.

Specifically, the electricians have long, floppy tails (called Intrinsically Disordered Regions or IDRs). Imagine a construction worker wearing a long, frayed rope instead of a solid arm. In the past, scientists couldn't figure out how these floppy ropes behaved. Did they hang loose? Did they wrap around other workers? Did they turn into solid arms when they grabbed a tool?

This paper is the story of how the researchers finally figured out exactly how these "floppy ropes" behave and how the whole crew assembles.

---

The Detective Work: Using "Molecular Tape"

To solve this mystery, the researchers used a high-tech detective method called Crosslinking Mass Spectrometry (XL-MS).

The Analogy:
Imagine you have a giant, tangled ball of yarn with different colored strings (proteins) mixed together. You want to know which string is touching which.

1. The Glue: The researchers sprayed the proteins with a special "molecular glue" (a chemical crosslinker). This glue is sticky only for specific points. If two proteins are touching, the glue snaps them together.
2. The Snapshot: They then took the proteins apart and used a super-powerful microscope (Mass Spectrometry) to take a picture of the glued pieces.
3. The Map: By seeing which pieces were stuck together, they could draw a map of who was standing next to whom.

They did this for three different construction crews: NuRD, SIN3, and CoREST.

---

The Magic Trick: Turning "Floppy" into "Solid"

Once they had the map of who was touching whom, they needed to build a 3D model. This is where they used a new kind of "AI magic" combined with their glue-map.

The Analogy:
Think of the "floppy rope" (the C-terminal tail of HDAC1) as a piece of string.

• Old Way: If you just asked a computer to guess the shape, it would say, "It's just a string. It could be anywhere."
• New Way: The researchers told the computer, "Okay, we know this string is glued to this specific spot on the neighbor's shirt, and that spot on the other neighbor's hat."

By forcing the computer to respect these "glue spots," the floppy string suddenly snapped into a specific shape. It turned from a messy string into a solid, coiled spring (an alpha-helix).

The Big Discovery:
The researchers found that the "floppy rope" isn't just messy chaos. It's a chameleon.

• When HDAC1 joins the CoREST crew, the rope stays loose and wiggly (flexible). This helps the crew move around quickly to fix different problems.
• When HDAC1 joins the SIN3 crew, the rope snaps tight and becomes rigid (solid). This locks the crew into a specific, stable position.
• When HDAC1 joins the NuRD crew, the rope forms a solid coil that helps hold the whole team together.

The Lesson: The shape of the protein changes depending on who it is working with. It's like a person wearing a suit for a formal meeting but jeans for a hike; the person is the same, but their "outfit" (structure) changes based on the job.

---

Building the Master Model: The NuRD Subcomplex

The researchers didn't stop at just two people. They built a complete model of a major sub-team called NuRD.

The Analogy:
Imagine trying to build a model of a complex machine with 5 different parts, where 3 of those parts are made of jelly.

1. They took the "glue map" from their experiments.
2. They used a super-computer (Integrative Modeling Platform) to try billions of different ways to put the pieces together.
3. They filtered out the models that didn't match the "glue" evidence.

The result was a complete, 3D blueprint of the NuRD machine. They found that the "jelly" parts (the IDRs) actually folded into specific shapes to act as the glue holding the machine together. Without these floppy parts folding up, the machine would fall apart.

---

Why Does This Matter?

1. It solves a 20-year mystery: Scientists have known about these proteins for decades, but they couldn't see the "floppy tails." This paper finally shows us what they look like when they are doing their job.

2. It explains how diseases happen: Since these proteins control genes, if they get the wrong shape, they can cause cancer or other diseases. Now that we have the blueprint, we can design drugs to fix them.

3. It's a new way to do science: This paper proves that you can study "messy" proteins (which were previously impossible to study) by combining experimental glue-maps with AI predictions. It's like using a flashlight to see through the fog.

Summary in One Sentence

This paper used a special "molecular glue" and AI to take a picture of how the body's gene-switching machines assemble, revealing that their "floppy tails" actually fold into specific shapes to help them do their jobs, depending on which team they are working with.

Technical Summary

1. Problem Statement

Histone Deacetylases 1 and 2 (HDAC1/2) are critical enzymatic components of large chromatin remodeling complexes, including NuRD, SIN3, and CoREST. While the catalytic domains of HDAC1/2 have been well-characterized via X-ray crystallography, their C-terminal domains (CTDs) remain structurally uncharacterized. These CTDs are Intrinsically Disordered Regions (IDRs) comprising approximately 22–23% of the protein sequence. Furthermore, the precise architecture of how HDAC1/2 assemble with their various binding partners within these large, dynamic complexes is poorly understood. Traditional structural biology methods (like X-ray crystallography) struggle with IDRs and large, heterogeneous complexes, leaving a gap in understanding how these disordered regions fold and function within the cellular environment.

2. Methodology

The authors employed a multi-faceted Integrative Structural Modeling (ISM) approach combining experimental data with computational prediction:

• Sample Preparation: They generated stable HEK293T cell lines expressing C-terminally Halo-tagged HDAC1 and HDAC2. This tagging strategy was chosen to preserve native nuclear localization and interaction networks, avoiding artifacts seen with N-terminal tags.
• Crosslinking Mass Spectrometry (XL-MS):
  • Proteins were crosslinked using DSSO (a cleavable crosslinker) while immobilized on beads.
  • Samples were digested and analyzed via Orbitrap Fusion Lumos mass spectrometry.
  • Data was processed using Proteome Discoverer 2.4 (XlinkX node) to identify Crosslink Spectrum Matches (CSMs), yielding thousands of intra- and inter-molecular crosslinks.
• Integrative Modeling Platform (IMP):
  • The authors used the Integrative Modeling Platform (IMP) to build structural models of the NuRD, SIN3A, and CoREST complexes.
  • Data Integration: The models were constrained by:
    • XL-MS distance restraints (Bayesian crosslink restraints).
    • Cryo-EM density maps (for NuRD and CoREST).
    • Known atomic structures (X-ray/NMR) and homology models.
    • AlphaFold3 predictions for regions lacking experimental structures.
  • Sampling: They utilized Replica Exchange Monte Carlo (Gibbs sampling) to generate millions of models, clustering them to find the most populated, high-scoring ensembles.
• AlphaFold3 & HADDOCK:
  • To specifically investigate the folding of the HDAC1 CTD, they compared AlphaFold3 predictions (sequence-only) against ISM models constrained by XL-MS data.
  • They used HADDOCK for guided molecular docking to refine dimeric and trimeric assemblies.

3. Key Contributions

• Comprehensive Interactome Mapping: Generated the first extensive XL-MS interactome for endogenous HDAC1 and HDAC2, identifying direct vs. indirect interactors and mapping specific interaction hotspots (e.g., K50/51, K74/75, K89/90).
• Resolution of IDRs: Demonstrated that XL-MS coupled with ISM can successfully model the folding of intrinsically disordered regions (specifically the HDAC1 CTD) into ordered structures within the context of a complex.
• Complete Subcomplex Models: Built high-resolution integrative models for three major co-repressor complexes (NuRD, SIN3A, CoREST) and a specific NuRD subcomplex (HDAC1:MBD3:MTA1:GATAD2B:RBBP4) containing six modeled IDRs.
• Context-Dependent Folding: Revealed that the structural conformation of the HDAC1 CTD is not static but depends on its binding partner, folding into specific alpha-helical structures only when part of the larger complex.

4. Key Results

A. Protein Interaction Networks

• XL-MS identified 1,471 CSMs for HDAC1 and 1,187 for HDAC2.
• Analysis confirmed that HDAC1 and HDAC2 form distinct but overlapping networks, primarily interacting with members of the NuRD, SIN3, and CoREST complexes.
• Specific interaction hotspots were identified on HDAC1/2 (K50/51, K66/67, K74/75, K89/90) where they bind to scaffolding proteins like SIN3A, RCOR1, and MTA1.

B. Structural Models of Complexes

• NuRD Complex: The model (stoichiometry: 2 MTA1, 2 HDAC1, 4 RBBP4, 1 MBD3, 1 GATAD2B) achieved a precision of 55 Å. The MTA1-HDAC1 dimer forms the scaffold. The model successfully satisfied 99% of input crosslinks and correlated highly (0.86) with the EM map.
• SIN3A Complex: A 4-subunit model (SIN3A, HDAC1, SAP30, SUDS3) achieved a precision of 33 Å. SUDS3 spans the complex, while SIN3A, SAP30, and HDAC1 form a compact core hub.
• CoREST Complex: A 3-subunit model (RCOR1, HDAC1, KDM1A) achieved a high precision of 12 Å. It forms a bi-lobed structure with KDM1A in the upper lobe and HDAC1 in the lower lobe, connected by RCOR1.

C. Folding of the HDAC1 C-Terminal Domain (CTD)

• AlphaFold3 Limitations: When modeled in isolation or as simple dimers using only sequence data (AlphaFold3), the HDAC1 CTD remained largely disordered.
• ISM Success: When constrained by XL-MS data within the context of the full complexes (NuRD, SIN3A, CoREST), the HDAC1 CTD folded into compact alpha-helical structures.
• Contextual Plasticity:
  • In CoREST (with RCOR1), the CTD remained relatively flexible.
  • In SIN3A, the CTD adopted a more rigid, structured conformation.
  • In NuRD (with MBD3/MTA1), the CTD formed a stable alpha-helix.
• Trimeric Modeling: Modeling the HDAC1:MBD3:MTA1 trimer resulted in a more compact and ordered structure than the individual dimers, highlighting the necessity of modeling multi-protein assemblies to capture native conformations.

D. NuRD Subcomplex with Multiple IDRs

• The authors modeled a 5-subunit NuRD subcomplex containing six IDRs (from HDAC1, MBD3, MTA1, and GATAD2B).
• The model revealed that these IDRs form transient, locally ordered structural elements (PreSMos) dominated by alpha-helices, facilitating conformational selection and induced fit mechanisms essential for complex assembly and function.

5. Significance

• Methodological Advancement: This study validates a robust workflow for studying large, dynamic protein complexes containing IDRs. It demonstrates that combining XL-MS with ISM and AlphaFold3 can resolve structures that are inaccessible to crystallography or cryo-EM alone.
• Biological Insight: The findings challenge the view of IDRs as purely "floppy" tails. Instead, they show that the HDAC1 CTD undergoes context-dependent folding, adopting specific alpha-helical structures only when bound to specific partners. This suggests a regulatory mechanism where the cellular environment dictates the structural state and function of HDAC1.
• Therapeutic Implications: Understanding the precise architecture of HDAC-containing complexes and their disordered regions provides new targets for drug development, particularly for cancers where these complexes are dysregulated. The study highlights that the "disordered" nature of these regions is functionally critical for the adaptability of chromatin remodeling.