Fixing Cryo-EM Sample Problems: Optimization Strategies & Workflows

We've diagnosed the 5 cryo-EM sample killers: preferred orientation, aggregation, denaturation, heterogeneity, and ice quality issues. Now comes the critical question: How do you fix it?

Nw, we'll cover systematic solutions, optimization workflows, modern AI-driven prediction, and a real-world case study of rescuing a failed ion channel project.

Solution 1: Fixing Preferred Orientation

The problem: Particles adopt the same orientation on the grid (only one view available).

The fix: Modify the grid surface or sample conditions to promote random orientations.

Strategy A: Graphene Oxide (GO) Supports

How it works:

• Replace holey carbon with graphene oxide

• GO is hydrophilic (reduces surface adsorption)

• Particles adopt random orientations

Success rate: 50-70% of preferred orientation problems solved

Pros:

• Most effective single solution

• Works for many protein types

Cons:

• GO introduces background noise (requires subtraction)

• Need to optimize GO coating thickness

Strategy B: Detergent Screening

How it works:

• Add mild detergent to reduce surface tension

• Prevents adsorption to air-water interface

• Particles stay in bulk solution (random orientations)

Detergents to try:

• CHAPS (0.01%)

• β-DDM (0.005%)

• Digitonin (0.02%)

• Amphipol A8-35 (0.01%)

Success rate: 30-50% improvement

Caution: Too much detergent creates micelles (background noise)

Strategy C: Nanodiscs (for Membrane Proteins)

How it works:

• Embed protein in lipid nanodisc (MSP1D1, MSP2N2)

• Nanodisc provides isotropic shape

• Reduces orientation bias

Success rate: 60-80% improvement for membrane proteins

Pros:

• Native-like lipid environment

• Reduces preferred orientation

• Removes detergent

Cons:

• Adds 20-30 kDa mass (visible in EM)

• Requires optimization (lipid type, ratio)

Strategy D: Affinity Grids

How it works:

• Functionalize grid with His-tag binding resin or antibody

• Protein tethered at specific site

• Reduces orientation bias

Success rate: 40-60% improvement

Cons:

• Requires protein tagging

• May introduce new bias (depends on tag location)

Solution 2: Preventing Aggregation

The problem: Particles clump together on the grid.

The fix: Optimize concentration and buffer conditions.

Strategy A: Optimize Protein Concentration

Test range: 0.5, 1, 2, 3 mg/mL

Sweet spot: High enough for particle density, low enough to avoid aggregation Typical optimal: 1-3 mg/mL for most proteins

How to find it:

• Make grids at each concentration

• Screen quickly

• Check particle distribution

Strategy B: Buffer Optimization

Increase salt:

• 100-300 mM NaCl screens charge interactions

• Reduces aggregation

Add stabilizers:

• Glycerol (5-10%) or trehalose (5%)

• Arginine (50-100 mM) suppresses protein-protein interactions

• BSA (0.01-0.1 mg/mL) as "sacrificial protein"

Adjust pH:

• Move away from pI (±1.5 pH units)

• Increased charge repulsion prevents aggregation

Strategy C: Predict and Eliminate Aggregation Hotspots

Use computational tools:

• Orbion, AGGRESCAN3D, CamSol

• Identify surface hydrophobic patches

• Design mutations (Leu→Ser, Ile→Thr)

Example:

• Hotspot: Leu47 on surface

• Mutation: L47S

• Result: Soluble at 5 mg/mL (vs 1 mg/mL wild-type)

Solution 3: Preventing Denaturation

The problem: Protein unfolds at air-water interface.

The fix: Stabilize protein or reduce interface exposure.

Strategy A: Stabilize with Ligands/Cofactors

Add saturating ligand:

• Substrate analog, inhibitor, or cofactor

• Stabilizes conformation

• Increases Tm

Example: Kinase

• Without ATP: Tm = 48°C

• With Mg²⁺ + AMP-PNP: Tm = 58°C

• Result: Protein stable on grid

Strategy B: Thermostabilizing Mutations

Design mutations to increase Tm:

• Target: +5-10°C increase

• Use Orbion or Rosetta

• Stable proteins resist denaturation

Example:

• Wild-type: Tm = 45°C, denatures on grid

• 3 mutations (predicted by Orbion): Tm = 55°C

• Result: Stable particles, 3.5 Å structure

Strategy C: Reduce Air-Water Interface Exposure

Use continuous carbon:

• Reduces interface area

• Or graphene/graphene oxide

Faster freezing:

• Reduce blot time (less time at interface)

• Self-blotting grids (Spotiton)

Detergent cushion:

• Add mild detergent (0.005-0.02%)

• Reduces surface tension

Solution 4: Eliminating Heterogeneity

The problem: Mixture of conformations, oligomers, or PTMs.

The fix: Homogenize sample or computationally separate.

Fix A: Remove PTM Heterogeneity

For glycosylation:

Step 1: Predict sites

• Use Orbion (predicts all glycosylation sites)

• Or NetNGlyc (free, consensus motifs)

Step 2: Mutate sites

• Asn → Gln (blocks N-glycosylation)

• Example: GPCR with 3 sites (N6Q, N15Q, N194Q)

Step 3: Test stability

• Measure Tm: If ΔTm > -3°C, acceptable

Alternative: Enzymatic deglycosylation

• PNGase F removes N-glycans

• Treat protein, then freeze for EM

Impact: Removing glycosylation enables uniform high-resolution reconstruction

Fix B: Stabilize Single Conformation

Add saturating ligand:

• Locks protein in one state

• Example: Kinase + ATP analog → DFG-in state only

Use nanobodies/Fabs:

• Binds epitope, stabilizes conformation

• Also increases particle size (easier alignment)

Fix C: Purify Single Oligomeric State

SEC separation:

• Collect only monomer peak

• Freeze immediately

Or crosslink:

• Mild glutaraldehyde or BS3

• Locks oligomeric state

Solution 5: Optimizing Ice Quality

The problem: Ice too thick, thin, or crystalline.

The fix: Optimize blotting parameters.

Blotting Parameter Optimization

Blot time:

• Test: 2, 3, 4, 5, 6 seconds

• Too short → thick ice

• Too long → thin ice, denaturation

Blot force:

• Vitrobot settings: -5 to +5

• Higher force → thinner ice

Humidity:

• 95-100% prevents evaporation before plunge

Temperature:

• 4°C for membrane proteins (prevents aggregation)

• Room temperature for soluble proteins

Glow Discharge Optimization

Timing:

• 15-30 seconds at 15 mA (easiGlow, PELCO)

• Freshly glow-discharged (within 1 hour)

Too much:

• Charged surface (particles stick, denature)

Too little:

• Hydrophobic surface (poor wetting)

Grid Type Screening

Test multiple types:

• Quantifoil R1.2/1.3 (standard)

• Quantifoil R2/2 (larger holes)

• UltrAuFoil R1.2/1.3 (gold, more stable)

• C-flat (different geometry)

Systematic Troubleshooting Workflow

Step 1: Check Ice Quality (5 min/grid)

Screen at low magnification:

• Holes filled with even, gray ice? → Good

• Very dark? → Too thick (increase blot time)

• Almost white? → Too thin (decrease blot time)

• Bright spots in FFT? → Crystalline (fresh ethane, clean sample)

Step 2: Check Particle Distribution (10 min)

Good: Particles evenly spread, 100-500 per hole Aggregation: Clustered → Lower concentration, optimize buffer Empty holes: No particles → Increase concentration, glow discharge All on carbon: → Increase concentration, add detergent

Step 3: Collect Test Dataset (2-4 hours)

Collect 200-500 micrographs:

• Check drift, CTF fit, particle count

• If poor → Optimize grid prep

Step 4: 2D Classification (1 day)

Good classes: Sharp features, multiple orientations, >80% good Preferred orientation: All same view → GO supports, detergent, nanodiscs Heterogeneity: Very different classes → Stabilize with ligands, remove PTMs Aggregation: Overlapping particles → Lower concentration, optimize buffer

The Modern Workflow: Predict Problems Before Making Grids

The old way (2015-2020):

1. Purify protein

2. Make 50-100 grids (blind testing)

3. Screen → Find most unusable

4. Collect data → Discover heterogeneity

5. Redesign construct

6. Repeat for 12-18 months

Cost: $200-400K, 12-24 months

The new way (2024+):

Step 1: Computational Prediction (1 day)

Orbion analysis:

• PTM prediction: Identifies glycosylation sites

• Aggregation analysis: Surface hydrophobic patches

• Stability analysis: Predicts Tm, suggests stabilizing mutations

• Cofactor prediction: What's required for stability

Output:

• "Remove N-glycosylation: N6Q, N15Q, N194Q"

• "Add thermostabilizing mutations: L124W, V168A"

• "Include antagonist during purification"

Step 2: Design Optimized Sample (1 week)

• Clone optimized construct

• Express with modifications

Step 3: Grid Optimization (3 weeks)

• Make 10-20 grids (systematic testing)

• High probability of success

Cost: $80-150K, 4-6 months

Success rate: 70-85% (vs 30-40% traditional)

Case Study: Rescuing a Failed Ion Channel

The Challenge

Target: Pentameric ligand-gated ion channel

Initial attempt (2018-2019):

• Construct: Full-length

• Expression: Sf9 cells, detergent-solubilized

• Grids: 80 made

• Data: 3 sessions ($150K)

• Result: Cannot refine beyond 6 Å

• Diagnosis: Severe preferred orientation (>95% top-down), glycosylation heterogeneity

The Rescue (with Orbion)

Step 1: PTM analysis

• Orbion predicts 4 N-glycosylation sites per subunit (20 total)

• Fix: Mutate all (N45Q, N89Q, N138Q, N276Q)

Step 2: Construct optimization

• AlphaFold: C-terminus (450-478) disordered

• Intracellular loop (340-385) flexible

• Fix: Truncate C-terminus, stabilize loop (V351A, L367S, F378A)

Step 3: Preferred orientation strategy

• Fix: Use nanodiscs (MSP1D1, POPC:POPG 3:1)

• Isotropic shape reduces orientation bias

Step 4: Stability optimization

• Orbion predicts agonist stabilizes open state

• Fix: Add saturating agonist (100 μM)

Final Construct & Strategy

• Residues: 1-449 (truncated)

• Mutations: N45Q, N89Q, N138Q, N276Q, V351A, L367S, F378A

• Nanodisc reconstitution

• Agonist-bound

Results

• Expression: 4 mg/L (2× improvement)

• Tm: 62°C (+8°C)

• Grids: 15 made, first grid showed good particles

• Preferred orientation: Solved (angular distribution uniform)

• Data: Single session (3 days)

• Processing: 800,000 particles

• Final: 2.9 Å structure

Cost: $60K, 5 months Outcome: Structure published

Lesson: Computational prediction identified all problems before grid optimization.

Practical Checklist: Before You Make Grids

☐ Sample Quality

• [ ] SEC-MALS monodisperse (>95%)

• [ ] DLS shows PDI < 15%

• [ ] Concentration optimized (tested 0.5-3 mg/mL on negative stain)

• [ ] Tm > 55°C

• [ ] Stable 24h at 4°C

☐ PTM Control

• [ ] PTM prediction complete

• [ ] Glycosylation addressed (mutated or deglycosylated)

• [ ] Mass spec confirms expected mass

☐ Construct Optimization

• [ ] AlphaFold pLDDT >70 for >90% residues

• [ ] Disorder prediction complete

• [ ] If membrane protein: Truncated loops optimized

☐ Conformational Homogeneity

• [ ] Ligand/cofactor added (if required)

• [ ] Single state confirmed

• [ ] Oligomeric state defined

☐ Grid Preparation Strategy

• [ ] Grid type selected

• [ ] Glow discharge protocol planned

• [ ] Blotting parameters planned

• [ ] Alternative supports available (GO, nanodiscs)

If you can check all boxes: Success rate 70-85%

If you skip: Expect 6-18 months with 30-40% success

The Economics of Prevention

Traditional Approach

Timeline: 18-30 months Cost: $370K (scientist, grids, microscope time) Success rate: 30-40% Cost per structure: $925K-1.2M (with failures)

Modern Approach

Timeline: 4-5 months Cost: $107.5K (prediction, optimization, single session) Success rate: 70-85% Cost per structure: $125-155K

ROI: 6-8× cost reduction, 4-6× faster

The Bottom Line

Cryo-EM sample preparation is predictable and fixable.

The 5 sample killers:

1. Preferred orientation → GO supports, detergent, nanodiscs

2. Aggregation → Optimize concentration, buffer, mutations

3. Denaturation → Stabilize with ligands, mutations

4. Heterogeneity → Remove PTMs, stabilize conformation

5. Ice quality → Optimize blotting parameters

The paradigm shift:

• Old: Make 100 grids → Hope → Fail → Repeat

• New: Predict problems → Fix sample → Optimize → Succeed

With modern computational tools, identify and fix problems before wasting grids.

Ready to Optimize Your Cryo-EM Sample?

If you're preparing for cryo-EM, Orbion can identify potential problems before you make your first grid.

Orbion provides:

• Complete PTM landscape (glycosylation causing heterogeneity)

• Construct optimization (disorder, flexible loops)

• Aggregation hotspot prediction

• Stability optimization (thermostabilizing mutations)

• Membrane topology analysis (for nanodiscs)

From sequence to cryo-EM-ready sample in <1 day.