Global analysis of thermal and chemical denaturation using CheMelt: Thermodynamic dissection of highly thermostable de novo designed proteins

This paper introduces CheMelt, an online tool for the global analysis of combined thermal and chemical denaturation data, which was used to demonstrate that highly thermostable de novo designed proteins exhibit distinct thermodynamic properties and that high thermal stability does not necessarily equate to high equilibrium stability.

The Gist

Imagine you are a chef trying to bake the perfect cake. You want a cake that doesn't crumble when you take it out of the oven, even if the oven is set to a scorching 200°C. In the world of science, researchers are doing something similar, but instead of cakes, they are baking proteins (the tiny building blocks of life) from scratch. These are called "de novo" proteins.

The problem? These new, super-strong proteins are so strong that they don't melt even when you boil them in water. This makes it very hard for scientists to measure how stable they really are. It's like trying to test the durability of a steel beam by throwing it in a pot of boiling water; the water just isn't hot enough to break it, so you learn nothing about its limits.

Here is a simple breakdown of what this paper is about, using some everyday analogies:

1. The Problem: The "Unbreakable" Proteins

Scientists have designed 35 new proteins that are incredibly heat-resistant. They are so tough that they stay folded (like a neat origami crane) even at temperatures above the boiling point of water.

• The Challenge: To understand how stable a protein is, scientists usually heat it up until it falls apart (unfolds) and watch what happens. But if the protein doesn't fall apart even at 100°C, the scientists are stuck. They can't see the "breaking point."

2. The Solution: The "Chemical Hammer" (CheMelt)

To break these tough proteins, the scientists used a two-pronged attack:

1. Heat: They turned up the temperature.
2. Chemical Solvent: They added a chemical called "Guanidine" (think of it as a super-strong soap or solvent) that makes the protein's structure slippery and weak.

By combining heat and this chemical "hammer," they could finally force the proteins to unfold.

To make sense of all this messy data, the team built a new digital tool called CheMelt.

• The Analogy: Imagine you have a pile of puzzle pieces from 15 different puzzles, all mixed together. You need to figure out the shape of the original picture for each one. CheMelt is like a super-smart, automated puzzle solver that takes all the scattered pieces (the data points from different temperatures and chemical levels) and snaps them together to reveal the full picture of the protein's stability. It's an online app that anyone can use, making this complex math easy to do.

3. The Discovery: The "Flat Tire" vs. The "Rock"

When they used CheMelt to analyze 15 of these proteins, they found something surprising.

Usually, when a natural protein (like one found in your body) unfolds, it exposes a lot of greasy, oily parts (hydrophobic residues) to the water. This causes a big "shock" to the system, measured as a high $\Delta C_p$ (a fancy way of saying "change in heat capacity").

• The Metaphor: Think of a natural protein like a rock. When it breaks, it shatters into many sharp, jagged pieces that splash water everywhere. Big splash = high energy change.

However, these new designed proteins were different. They had a very low $\Delta C_p$.

• The Metaphor: These proteins were more like a flat tire. When they deflated, they didn't explode or splash; they just slowly collapsed into a flat, quiet heap. They didn't expose many greasy parts to the water.

Why does this matter?
The scientists realized these proteins are so heat-stable not because they are "stronger" in a traditional sense, but because they are less sensitive to temperature changes.

• The Analogy: Imagine two cars.
  • Car A (Natural Protein): Has a very powerful engine but is sensitive to heat. If the temperature goes up a little, the engine runs wild.
  • Car B (Designed Protein): Has a weak engine, but it's built with materials that don't react to heat at all. It doesn't matter if it's 20°C or 100°C; the engine just hums along quietly.

The designed proteins are like Car B. They are stable at high temperatures because they are "boring" and unreactive, not because they are super-strong.

4. The Big Takeaway

The paper teaches us two main lessons:

1. Don't just look at the melting point: Just because a protein doesn't melt at 100°C doesn't mean it's "stable" in a useful way. It might just be insensitive to temperature. We need to look deeper (using tools like CheMelt) to see if it's actually stable or just "numb."
2. New Tools for New Problems: As we design more amazing, artificial proteins, we need better tools to measure them. CheMelt is that tool, acting as a bridge between complex data and clear answers.

In a nutshell: The scientists built a new digital calculator (CheMelt) to figure out why their super-tough, man-made proteins are so heat-resistant. They discovered these proteins are stable not because they are tough, but because they are "boring" and don't react much to heat, unlike our natural proteins which are more dramatic and sensitive.

Technical Summary

1. Problem Statement

• The Challenge of Thermostability: De novo designed proteins often exhibit extreme thermostability, with melting temperatures ($T_m$) frequently exceeding 100°C. This makes standard thermal denaturation analysis impossible because the proteins do not unfold within the accessible temperature range of water (up to ~95–100°C).
• Limitations of Current Methods: While chemical denaturation (using chaotropes like GdmCl) can lower the free energy of unfolding ($\Delta G_u$) to induce unfolding at lower temperatures, analyzing the combined effects of temperature and denaturant concentration is mathematically complex.
• Data Analysis Gap: Existing tools often require individual curve fitting, which is underdetermined for highly stable proteins. Global analysis (fitting multiple curves simultaneously) is more robust but lacks user-friendly, standardized software, particularly for the specific challenges posed by de novo designs (e.g., small fluorescence changes, non-linear baselines).

2. Methodology

The authors developed a comprehensive workflow combining experimental techniques with a new software tool:

• Experimental Approach (nanoDSF):

  • Screening: 35 de novo designed protein binders (mostly $\alpha$-helical bundles) were screened using Nanoscale Differential Scanning Fluorimetry (nanoDSF) at 0, 2, 4, and 6 M Guanidine Hydrochloride (GdmCl).
  • Titration: Proteins showing observable transitions were subjected to a full titration series (11 concentrations of GdmCl) to generate 2D unfolding data (Temperature vs. Denaturant Concentration).
  • Validation: Circular Dichroism (CD) was used on a subset of proteins that showed no fluorescence change to confirm they were indeed unfolding, distinguishing between "no unfolding" and "no signal change."

• Software Development (CheMelt):

  • Tool: A new web-based graphical user interface (GUI) named CheMelt (hosted at eSPC) was developed, based on the Python package pychemelt.
  • Algorithm: It performs a global fit of multiple melting curves simultaneously.
  • Model: It utilizes a two-state unfolding model where the free energy of unfolding ($\Delta G_u$) depends on both temperature ($T$) and denaturant concentration ($D$):
    $$\Delta G_u(T, D) = \Delta H_m \left(1 - \frac{T}{T_m}\right) - \Delta C_p \left[ (T_m - T) + T \ln\left(\frac{T}{T_m}\right) \right] + m \cdot D$$
  • Baseline Handling: The tool allows flexible modeling of baselines (native and unfolded states) using linear, quadratic, or exponential equations to account for non-linear signal drifts common in nanoDSF.

3. Key Contributions

1. CheMelt Platform: The release of an accessible, open-source tool for the global analysis of combined thermal and chemical denaturation data, lowering the barrier for high-throughput thermodynamic characterization.
2. Thermodynamic Dissection of De Novo Proteins: The first systematic thermodynamic analysis of a large set of highly thermostable de novo designed proteins, revealing their unique stability mechanisms.
3. Validation of Methodology: Demonstrated that CheMelt can accurately reproduce known thermodynamic parameters (using ACBP as a benchmark) and extract parameters for proteins that do not unfold in the absence of denaturant (simulated and experimental data).

4. Key Results

• Tool Validation:

  • Case I (ACBP): CheMelt successfully reproduced thermodynamic parameters ($T_m$, $\Delta C_p$, $m$-value) for Bovine Acetyl-CoA Binding Protein, matching previously published literature.
  • Case II (Simulation): The tool accurately extracted parameters for simulated hyper-stable proteins ($T_m$ 100–140°C) where no unfolding occurred without denaturant, proving the efficacy of extrapolation.
  • Case III (Experimental): Applied to 35 de novo binders. 20 showed no fluorescence transition (likely due to lack of buried aromatic residues or minimal environmental change upon unfolding). 15 showed clear transitions and were successfully fitted.

• Thermodynamic Findings:

  • Extreme Thermostability: The 15 fitted proteins had $T_m$ values ranging from 96°C to 152°C (average 122°C).
  • Low $\Delta C_p$ and $m$-values: The designed proteins exhibited significantly lower heat capacity changes ($\Delta C_p$) and $m$-values compared to natural proteins of similar size.
  • Correlation with Residue Count: While $\Delta C_p$ and $m$-values correlated with protein length (as expected), the slope was much shallower than in natural proteins (Myers et al., 1995 data).
  • Mechanism of Stability: The high thermostability is primarily driven by low temperature sensitivity (small $\Delta C_p$) rather than high equilibrium stability ($\Delta G_u$) at room temperature. This results in "flattened" stability curves where the protein remains folded at high temperatures but does not necessarily have a massive stability margin at 25°C.
  • Fluorescence Limitations: 57% of the designed proteins (20/35) showed no fluorescence change during unfolding, despite CD confirming they unfolded. This suggests the designed hydrophobic cores are less dehydrated or lack the specific aromatic packing required for large fluorescence shifts.

5. Significance and Implications

• Redefining Stability Metrics: The study highlights that a high $T_m$ does not automatically equate to high equilibrium stability ($\Delta G_u$) or robustness against aggregation/proteolysis at physiological temperatures. Stability achieved via low $\Delta C_p$ is distinct from stability achieved via high $\Delta G_u$.
• Design Insights: The low $\Delta C_p$ suggests that de novo designs may have hydrophobic cores that are not fully dehydrated or retain residual structure in the denatured state. This provides a feedback loop for improving protein design algorithms (e.g., RFdiffusion, ProteinMPNN) to better mimic natural hydrophobic burial.
• Methodological Advancement: CheMelt enables researchers to rigorously characterize the thermodynamics of hyper-stable proteins, which was previously difficult. This is crucial for industrial applications where thermal stability is required, ensuring that stability is not just a thermal artifact but a robust thermodynamic property.
• Future Directions: The authors emphasize the need to distinguish between thermal stability and general stability when engineering proteins, advocating for full thermodynamic characterization rather than relying solely on melting temperature.