TL;DR
This paper investigates the subtle sequence and structural signatures that distinguish co-translational protein folding (folding as the protein emerges from the ribosome) from global folding (folding after the protein is fully synthesized). By analyzing millions of simulated protein sequences, they found that co-translational folding leads to a more compact N-terminus and a bias towards an N-terminal core, signatures also observed in real protein structures.
What Problem Does This Paper Address?
The fundamental question of how proteins fold into their complex three-dimensional structures is still being unraveled. Two primary mechanisms are proposed: global folding, where the entire polypeptide chain folds after synthesis, and co-translational folding (CT folding), where folding begins while the protein is still being synthesized by the ribosome. Understanding the interplay and distinct characteristics of these two mechanisms is crucial for a complete picture of protein structure formation. While experimental evidence supports both, identifying clear "signatures" of CT folding in sequences and structures has been challenging, especially in large-scale studies. This paper aims to bridge this gap by systematically classifying sequences based on their folding behavior and then analyzing the resulting properties to pinpoint these signatures.
Methods
The researchers employed a computational approach using the Hydrophobic-Polar (HP) model, a simplified representation of protein folding. They developed a classification algorithm (M1) to categorize HP sequences into three sets:
- Global-CT: Sequences that reach their unique global energy minimum (UGEM) conformation via co-translational folding.
- Kinetic-CT: Sequences that attain a unique final conformation co-translationally, but this is not necessarily the UGEM.
- Non-CT: Sequences that have a UGEM but cannot attain it or a unique final conformation co-translationally.
This classification was performed on over 1.5 million HP sequences. To further investigate folding behavior, they used Markov-chain simulations (M2 for global folding, M3 for CT folding) to assess folding rates and the propensity to reach UGEM or kinetic conformations. They also analyzed real protein structures from the SCOP database, applying similar sequence and structure measures to compare with their HP model findings.
Datasets
- HP Model Sequences: Over 1.5 million HP sequences were generated and classified. Specific subsets of length 25 (765,147 sequences) and longer sequences were analyzed.
- SCOP Database: A set of 10,311 protein domains from the SCOP database (release 1.75) was used, with a focus on specific classes (alpha, beta, alpha/beta, alpha+beta) for comparison.
Key Insights and Ideas
- Subtle Signatures: The paper challenges some previous assumptions, finding that CT folding signatures are more subtle than expected. CT conformations are not necessarily less compact or enriched in local contacts.
- N-terminal Bias: The strongest signature of CT folding is a restriction of conformational space in the N-terminal region, leading to a more compact N-terminus and an N-terminal bias in the core residues.
- Robustness to Mutation: Global-CT sequences, in particular, appear to be more robust to mutation, as evidenced by their enrichment in "hub nodes" within neutral net analyses.
- Directed Folding Pathways: The N-terminal restriction in CT folding likely guides the folding process, creating a directed pathway towards the UGEM structure, aligning with earlier theoretical suggestions.
- Underestimated Importance: The prevalence of CT folding signatures in real protein structures suggests that its importance in vivo is currently underestimated.
Key Results
- Classification: Over 1.5 million HP sequences were classified, with a significant portion (10-20% for longer sequences, 43% for length 13) showing co-translational folding properties.
- Folding Rate Improvement: Markov-chain simulations showed that starting from CT-folds consistently increased folding rates compared to random starting conformations, with Global-CT sequences benefiting the most.
- Real Protein Signatures: Analysis of SCOP domains revealed that a majority exhibit biases consistent with CT folding signatures, particularly in the alpha/beta class (71% with N-terminal core residues).
- PHI Score: The Possible Hydrophobic Interactions (PHI) score was found to be a strong indicator, with Global-CT sequences showing a low number of favorable interactions in N-terminal regions.
Limitations and Future Work
- HP Model Simplification: The HP model is a simplification. Transferring measures directly from HP models to real protein structures requires caution, especially regarding flexible termini.
- Kinetic Component: The primary classification method (M1) is deterministic and lacks a kinetic component, although kinetic simulations were used for validation.
- PHI Score Applicability: The PHI score, while effective in the HP model, is not directly applicable to real protein data in its current form.
- Future Work: The authors suggest further investigation into the applicability of their findings to real protein structures and the development of methods to directly assess CT folding potential in biological sequences. They also highlight the need to understand how sequence evolution might favor CT folding.
Summary updated