TL;DR
This paper introduces a novel pipeline that combines single-molecule force spectroscopy (SMFS) with AFM to unfold and identify membrane proteins directly within native cell membranes. By analyzing the force-distance curves and integrating this data with mass spectrometry and proteomic databases via a Bayesian framework, the researchers can characterize the mechanical properties and identify membrane proteins in situ, paving the way for single-cell membrane proteomics.
What Problem Does This Paper Address?
Understanding the structure and function of membrane proteins is crucial, but challenging due to their complex environment within cell membranes. While techniques like cryo-EM can determine static structures, they often require purified proteins and don't capture dynamic or heterogeneous states. Atomic Force Microscopy (AFM) coupled with Single-Molecule Force Spectroscopy (SMFS) can probe protein mechanics and unfolding pathways in near-native conditions, but has historically been limited to purified proteins. This paper addresses the significant challenge of applying SMFS to native membrane proteins within intact cells, enabling the study of their behavior in their natural environment without the need for laborious purification.
Methods
The core of this work is a multi-step pipeline:
- Membrane Isolation (Unroofing): A novel "unroofing" technique is employed to isolate the apical plasma membrane of single cells, preserving membrane proteins while removing cytoplasmic components. This method is adapted for both adherent and non-adherent cell types.
- Single-Molecule Force Spectroscopy (SMFS): Atomic Force Microscopy (AFM) is used to gently pull on individual membrane proteins embedded in the isolated membrane patches. This generates force-distance (F-D) curves, which act as "fingerprints" of the protein's unfolding process.
- Unsupervised Clustering: A density peak clustering algorithm is applied to the millions of collected F-D curves. This automatically groups similar unfolding patterns, effectively identifying distinct protein unfolding pathways without prior knowledge of the proteins involved.
- Bayesian Identification: A Bayesian framework is developed to probabilistically identify the membrane proteins corresponding to the identified clusters. This integrates information from:
- Mass Spectrometry Data: Provides protein abundance and length information for the cell type.
- Proteomic Databases (UniProt, PDB): Offers structural and topological information (e.g., secondary structure, domain lengths).
- SMFS Literature: Contains known unfolding characteristics of various membrane proteins.
- Validation: The pipeline's accuracy is validated by overexpressing specific membrane proteins with known signatures (like the N2B domain) and comparing their unfolding F-D curves to those identified in native membranes.
Datasets
- Native Cell Membranes: Data was collected from five cell types: hippocampal neurons, dorsal root ganglia (DRG) neurons, neuroblastoma (NG108-15 cells), rod outer segments, and rod discs.
- Purified/Reconstituted Proteins: Experiments were also performed on purified membrane proteins (Channelrhodopsin (ChR1), SthK, and Bacteriorhodopsin (bR)) reconstituted into lipid bilayers for initial benchmarking.
- Public Databases: UniProt and PDB were used for structural and topological information. Mass spectrometry data from existing studies was also integrated.
Key Insights and Ideas
- In Situ Protein Characterization: The ability to study membrane proteins directly in their native cellular environment, rather than relying on purified or reconstituted systems, is a major leap forward. This preserves crucial interactions and cellular context.
- Unfolding as a Fingerprint: The unique force-distance curves generated by SMFS serve as a "fingerprint" for individual membrane proteins, allowing for their identification even within complex mixtures.
- Bayesian Integration for Identification: Combining diverse data sources (SMFS curves, mass spec, structural databases) within a Bayesian framework provides a robust and probabilistic approach to identifying unknown proteins.
- Structural Heterogeneity Revealed: SMFS can capture variability in protein unfolding, particularly in loop regions, offering insights into structural heterogeneity that might be missed by static structural methods.
- Potential for Single-Cell Proteomics: The success in identifying proteins from just a few cells suggests the potential for developing single-cell membrane proteomics techniques.
Key Results
- The pipeline successfully isolated membrane patches and collected millions of F-D curves from native cell membranes.
- Clustering identified distinct unfolding patterns, revealing that even in native membranes, clean unfolding traces can be detected.
- The Bayesian identification method successfully narrowed down candidates for identified clusters to typically 1-4 proteins with high probability.
- Validation experiments using overexpressed constructs confirmed the identification of specific membrane proteins (TMEM16F, TRPC1, TRPC5, TRPC6).
- Structural insights were gained into the variable nature of loop regions in proteins like TRPC6, reconciling different cryo-EM structures.
- The protein profiles derived from SMFS showed a remarkable correlation with mass spectrometry data, suggesting a semi-quantitative proxy for protein length detection at the single-cell level.
Limitations and Future Work
- Cluster Merging: A limitation acknowledged is the possibility of merging F-D curves from proteins with different molecular identities but similar unfolding characteristics, especially for shorter proteins.
- Technical Improvements: The authors suggest future work could involve functionalizing AFM tips for targeted protein unfolding and developing methods to access the extracellular side of the membrane.
- Quantitative Reliability: While promising, the SMFS-based protein profiling is still considered semi-quantitative and requires further improvements for precise quantitative reliability compared to mass spectrometry.
- Extending the Approach: The authors envision extending this approach to characterize and distinguish membranes in healthy versus diseased cells, or in cells under different physiological conditions, particularly in scenarios where sample scarcity is an issue (e.g., patient samples).
Summary updated