Breaking Barriers: Transitioning from X-ray Crystallography to Cryo-EM for Structural Studies

This paper details a laboratory's successful transition from X-ray crystallography to cryo-electron microscopy for studying the ATAD2B protein, outlining practical strategies to overcome biochemical and technical challenges while providing a comprehensive workflow guide for structural biologists adopting this transformative technology.

The Gist

Imagine you are trying to take a perfect, crystal-clear photo of a tiny, complex machine inside a cell. This machine, called ATAD2B, is like a sophisticated molecular robot that helps organize the DNA in our bodies. For years, scientists tried to photograph this machine using a method called X-ray crystallography.

Think of X-ray crystallography like trying to take a photo of a spinning fan by first freezing it in a block of perfectly clear ice. If the ice is clear and the fan is frozen in the exact right position, you get a great picture. But if the fan is too big, too floppy, or just won't freeze into a neat block, you can't get the photo. That's what happened to the scientists in this paper: their "molecular robot" was too big and messy to freeze into a crystal.

So, they decided to switch to a new, high-tech camera called Cryo-EM (Cryo-Electron Microscopy).

The New Camera: Cryo-EM

Instead of freezing the machine in a block of ice, Cryo-EM is like taking thousands of quick, blurry snapshots of the machine as it floats in a thin layer of "flash-frozen water" (vitreous ice). Then, a super-computer acts like a photo editor, stacking all those blurry snapshots together to build a sharp, 3D movie of the machine.

The team in this paper was excited because Cryo-EM didn't require the machine to be frozen in a perfect crystal. It could handle the "messy" nature of their protein.

The Big Problem: The Unwanted Guest

However, there was a catch. When they prepared their protein sample, they accidentally invited a very large, very common "unwanted guest" into the mix.

Imagine you are trying to photograph a specific, rare bird in a forest. But, because you set up your camera in a busy park, every single photo you take is filled with pigeons. The pigeons are everywhere, and they look somewhat similar to your bird from a distance.

In this scientific story:

• The Rare Bird: The ATAD2B protein (the one they wanted to study).
• The Pigeons: GroEL, a helper protein that naturally lives inside the bacteria (E. coli) they used to make the protein.

Because they made the protein in bacteria, the bacteria's own "nanny" protein (GroEL) got stuck to their target. When they looked at their data, they saw thousands of images of these "pigeons" (GroEL) and very few of their "rare bird" (ATAD2B).

The Detective Work

At first, the scientists were confused. They built a 3D model, but it didn't look like their robot; it looked like a donut-shaped ring. A friendly expert looked at their data and said, "Hey, that's not your robot; that's the pigeon (GroEL)!"

They realized their sample was contaminated. The "pigeons" were so numerous that they were drowning out the "rare bird."

The Solution: Changing the Neighborhood

They tried to clean the sample, but the pigeons were too sticky. They realized they had two choices:

1. Keep trying: Take millions more photos hoping to find enough "rare birds" to ignore the "pigeons." This would take forever and cost a fortune in computer time.
2. Move the factory: Stop making the protein in bacteria (where the pigeons live) and start making it in insect cells (Sf9 cells).

They chose option 2. They switched their production line to insect cells. In this new environment, the "pigeons" (GroEL) didn't exist. When they made the protein again, it was pure. No pigeons, just the rare bird.

The Result

With the clean sample, they finally got the high-resolution photos they needed. They successfully built a detailed 3D model of the ATAD2B machine, showing exactly how it works.

What This Means for Everyone

This paper is a guide for other scientists who want to switch from the old "crystal" method to the new "Cryo-EM" method. It teaches us three main lessons:

1. Sample Purity is King: Even with the best camera, if your sample is dirty (full of "pigeons"), you won't get a good picture. Sometimes you have to change how you make your protein to get it clean.
2. Technology is a Team Sport: You need the right camera, the right computer, and the right software (like AI tools that can help pick out the "bird" from the "pigeons") to succeed.
3. Don't Give Up: Science is full of dead ends. The team hit a wall with their bacteria, but by learning a new skill (Cryo-EM) and changing their strategy (switching to insect cells), they broke through the barrier.

In short, this paper is a story about how a team of scientists learned a new way to take pictures of tiny machines, got distracted by a common contaminant, figured out a clever fix, and ultimately succeeded in revealing the secrets of a protein that helps control our genes.

Technical Summary

1. Problem Statement

The Glass laboratory aimed to determine the high-resolution structure of ATAD2B, a large (1,458 amino acids), multi-domain chromatin regulator containing AAA+ ATPase domains and a bromodomain.

• Initial Challenge: The team initially attempted to solve the structure using X-ray crystallography, a technique they were experienced with. However, they failed to obtain well-diffracting crystals.
• Root Cause: The target protein (a ~150 kDa monomer expected to form a ~900 kDa hexamer) proved difficult to purify in the high concentrations and homogeneity required for crystallization.
• Alternative Approach: The team transitioned to single-particle cryo-electron microscopy (cryo-EM), which is less sensitive to sample heterogeneity and does not require crystallization.
• New Obstacle: Upon initial cryo-EM screening, the team discovered their samples were heavily contaminated with GroEL, an abundant E. coli chaperonin. The contaminant co-purified with ATAD2B, and because both form large oligomeric complexes (GroEL is a 14-subunit heptamer; ATAD2B is a hexamer), they were difficult to separate via standard size-exclusion chromatography (SEC). The contaminant particles outnumbered the target particles by nearly 10:1, threatening the feasibility of the project.

2. Methodology

The paper details a comprehensive workflow spanning biochemical optimization, sample preparation, data collection, and computational processing.

• Sample Preparation & Expression:

  • Initial System: ATAD2B (residues 380–1458) was expressed in E. coli with an N-terminal GST tag. Despite optimization, the sample contained significant GroEL contamination.
  • Optimization: The team utilized negative staining to assess sample quality before plunge-freezing. They optimized buffer conditions (Tris pH 7.5, 150 mM NaCl, 5% glycerol) and grid types (UltrAuFoil) using a Vitrobot Mark IV.
  • System Switch: Recognizing that biochemical purification could not remove GroEL, the team switched expression systems to Sf9 insect cells (collaborating with the Trybus Lab). This successfully eliminated the bacterial chaperonin contamination, yielding >90% pure ATAD2B.

• Data Collection:

  • Data was collected at the National Center for Cryo-EM Access and Training (NCCAT) using a 300 kV Titan Krios microscope.
  • Three datasets were collected for the E. coli-expressed sample (to study the contaminant GroEL in different states): Apo, ADP-bound, and $\gamma$ATP-bound (using a slowly hydrolyzable analog).

• Computational Processing:

  • Software: The team utilized CryoSPARC, cisTEM, Topaz, and Phenix.
  • Particle Picking Strategy: Standard "blob" picking failed to isolate sufficient ATAD2B particles due to the overwhelming presence of GroEL. The team employed Topaz (a neural network-based picker). Crucially, they trained the Topaz model on the abundant GroEL particles. Surprisingly, this model also successfully picked the minority ATAD2B particles, recovering ~127,000 ATAD2B particles from the raw data.
  • Classification: Heterogeneous refinement and 3D classification were used to separate GroEL (D7 symmetry) from ATAD2B (hexameric).
  • Model Building: Atomic models were built using ChimeraX for rigid body fitting, Coot for manual adjustment, and Phenix for real-space refinement. AlphaFold predictions were used as initial guides.

3. Key Results

• The "GroEL" Detour: While the ATAD2B structure was not resolved to high resolution from the E. coli sample due to particle scarcity, the team successfully solved the structures of the contaminant GroEL in three states:
  • GroEL-$\gamma$ATP: 3.3 Å resolution (261,106 particles).
  • GroEL-Apo: 3.7 Å resolution (42,776 particles).
  • GroEL-ADP: 4.2 Å resolution (158,542 particles).
  • These structures were validated using MolProbity, showing excellent stereochemistry (e.g., 97.4% Ramachandran favored residues for the ATP state).
• ATAD2B Resolution Limit: Even with advanced computational sorting (Topaz picking + 3D classification), the E. coli dataset yielded only 45,000 ATAD2B particles, resulting in a map of only **5.1 Å resolution**. This was insufficient to resolve nucleotide coordination details in the ATPase pocket.
• Biochemical Solution: The switch to Sf9 insect cell expression successfully produced homogeneous ATAD2B free of GroEL. This confirmed that for large, complex targets, the choice of expression system is often the critical bottleneck, not just the imaging technique.

4. Key Contributions

• Practical Workflow for Contaminant Removal: The paper provides a case study on how to handle severe sample contamination. It demonstrates that while computational tools (Topaz) can recover minority species, there is a "contamination threshold" (approx. 70–80% contaminant) beyond which data collection becomes inefficient.
• Expression System Comparison: It highlights the superiority of insect cell (Sf9) systems over E. coli for expressing large, multi-domain eukaryotic proteins that are prone to chaperone co-purification.
• Educational Resource: The paper serves as a guide for labs transitioning from crystallography to cryo-EM, detailing:
  • Computational hardware requirements (GPU clusters, RAM).
  • Software pipelines (CryoSPARC, cisTEM, Topaz).
  • Common pitfalls (air-water interface damage, preferred orientation, chaperone contamination).
  • A curated list of learning resources (Table 1) and software tools (Table 5).
• Validation Standards: It outlines rigorous validation protocols for cryo-EM structures, including FSC curves, Q-scores, and MolProbity statistics, setting a standard for new practitioners.

5. Significance

• For Structural Biology: The study underscores that while cryo-EM is powerful for large complexes, it is not a "magic bullet" that bypasses the need for high-quality biochemistry. Sample purity remains the foundation of high-resolution structure determination.
• For the Community: By openly sharing the failure of the E. coli system and the success of the Sf9 system, the authors save other researchers time and resources. They provide a realistic assessment of the "learning curve" and computational costs involved in cryo-EM.
• Future Directions: The paper advocates for an integrated approach combining biochemical optimization, advanced vitrification, and machine learning-driven data processing. It suggests that as computational tools improve, the ability to handle heterogeneity will increase, but obtaining a homogeneous sample remains the most efficient path to near-atomic resolution.

In summary, this paper is a "lessons learned" narrative that documents the transition from crystallography to cryo-EM, emphasizing that overcoming biochemical hurdles (specifically chaperone contamination) is often more critical than the imaging technology itself.