Single-cell proteomics reveals proteome remodeling and cellular heterogeneity during NGF-induced PC12 neuronal differentiation

This study establishes an optimized single-cell proteomics workflow to overcome PC12 cell aggregation challenges, revealing that NGF-induced neuronal differentiation involves global proteome remodeling and the emergence of distinct cellular subpopulations with unique protein expression patterns that are masked in bulk population analyses.

The Gist

The Big Picture: Listening to the Orchestra vs. The Soloist

Imagine you are trying to understand how a symphony orchestra learns a new, complex piece of music.

• The Old Way (Bulk Proteomics): Traditionally, scientists would record the entire orchestra playing together and analyze the single, loud sound wave. They could tell you the orchestra was playing a "sad song" or a "fast song," but they couldn't tell you that the violinist in the back row was playing a slightly different tune than the one in the front, or that the drummer was still figuring out the rhythm. The unique, individual voices get lost in the crowd noise.
• The New Way (Single-Cell Proteomics): This paper is about putting a tiny, high-tech microphone on every single musician in the orchestra. This allows scientists to hear exactly what each cell is doing, revealing that not everyone is playing the same note at the same time.

The Challenge: The Sticky, Shy Musicians

The scientists used a specific type of cell called PC12 cells. Think of these cells as "actor trainees." When you give them a special signal (a chemical called NGF), they stop acting like generic cells and start transforming into neurons (nerve cells), growing long arms called "neurites" to talk to other cells.

However, these "actors" are difficult to work with:

1. They are sticky: Once they start acting (differentiating), they cling to each other and the floor like super-glued gum.
2. They are fragile: If you try to pull them apart roughly, they break.
3. They are small: They are tiny, so getting a clear "recording" of their proteins is like trying to hear a whisper in a hurricane.

The Solution: A High-Tech "Pipette" and a Magic Detergent

To solve these problems, the team built a specialized workflow:

1. The Gentle Detacher: Instead of using harsh chemicals to rip the cells apart, they used a gentle solution (Versene) that acts like a "non-stick spray," allowing the cells to float apart without getting hurt.
2. The Thermal Inkjet Printer: They used a device similar to a high-precision printer (the HP D100). Instead of ink, it shoots tiny droplets of water containing exactly one cell into a tiny well. It's like a robot bartender that can pour exactly one drop of wine into a specific glass without spilling a drop.
3. The Magic Detergent (DDM): Some proteins are like oil; they don't mix well with water and hide inside the cell's membranes. The scientists added a mild soap called DDM. Think of this as a "grease-cutting agent" that helps wash out those hidden, oily proteins so the machine can see them.

The Discovery: Not Everyone Changes at the Same Speed

Once they started recording the "music" of individual cells over 6 days, they found something surprising that the old "orchestra recording" missed:

The "Two-Track" Reality
By Day 4 and Day 6, the scientists expected all the cells to look like mature neurons. Instead, they found the population had split into two distinct groups:

• Group A (The Stars): These cells were fully transformed. They had grown long arms, were making the right proteins for communication, and were ready to work as neurons.
• Group B (The Stragglers): These cells were still stuck in the middle. They hadn't grown their arms yet and were still holding onto their old "generic cell" proteins.

Why does this matter?
If you had only listened to the whole orchestra (bulk analysis), you would have heard a "medium" sound and assumed everyone was halfway there. But by listening to individuals, they realized the transformation is messy and uneven. Some cells are ready to graduate, while others are still in the classroom.

The "Hidden" Proteins

The study also found that some proteins are like "secret agents."

• In the bulk (average) data, these proteins looked very common because the "Star" cells were making so many of them.
• But in the single-cell data, they realized these proteins were actually only present in a few specific cells. The "average" was lying; the proteins weren't everywhere, they were just concentrated in the successful cells.

The Takeaway

This paper is a triumph of precision. By building a better way to handle fragile, sticky cells and listening to them one by one, the scientists showed us that biological change isn't a smooth, uniform wave. It's a chaotic, individual journey where every cell decides its own pace.

In short: They turned a blurry group photo into a high-definition video of every single person, revealing that while the group is moving forward, some are sprinting, some are walking, and some are still tying their shoes.

Technical Summary

1. Problem Statement

Biological Challenge: Neuronal differentiation is a dynamic, asynchronous process where individual cells within a population may progress at different rates. Conventional bulk proteomics averages protein signals across thousands of cells, obscuring biologically meaningful heterogeneity, transitional states, and rare subpopulations.
Technical Challenge: Applying single-cell proteomics (SCP) to neuronal cells (specifically PC12 cells) is technically difficult because:

• Differentiated PC12 cells are highly adherent, fragile, and prone to aggregation.
• Standard cell isolation methods (like FACS) can damage fragile neurites or cause cell loss.
• Low protein abundance in single cells requires highly sensitive workflows to detect thousands of proteins without significant sample loss or contamination.
• Existing SCP workflows often struggle with membrane-associated and low-solubility proteins, which are critical in neuronal biology.

2. Methodology

The authors developed and optimized a robust, label-free single-cell proteomics workflow tailored for adherent neuronal cells.

A. Sample Preparation & Cell Handling:

• Cell Model: PC12 cells treated with Nerve Growth Factor (NGF) to induce differentiation over 6 days (Day 0, 2, 4, 6).
• Dissociation: Used Versene (EDTA-based) instead of enzymatic trypsin to gently dissociate cells without damaging fragile neurites or inducing stress responses.
• Anti-Adhesion: Cells were washed with DPBS containing EDTA and filtered through a 35 µm strainer to remove aggregates.
• Dispensing: Utilized a Thermal Inkjet (TIJ) dispenser (HP D100/Tecan Uno) for non-contact, precise single-cell deposition into 384-well plates. This minimized mechanical stress and sample loss.
• Lysis & Digestion: A one-step digestion workflow was employed in 384-well plates using a mild nonionic surfactant, n-dodecyl-β-D-maltoside (DDM), to improve the recovery of membrane-associated and low-solubility proteins.
• Reagents: 1 ng of Trypsin/LysC mixture per cell.

B. Mass Spectrometry Acquisition:

• Instrument: Bruker timsTOF SCP coupled with a NanoElute 2 UHPLC system.
• Acquisition Mode: dia-PASEF (Data-Independent Acquisition with Parallel Accumulation-Serial Fragmentation). This combines the depth of DIA with the sensitivity and speed of ion mobility separation.
• Parameters: 15-minute gradient, 300 nL/min flow rate, and specific TIMS settings (1/K0 range 0.66–1.42 V·s/cm²) to maximize ion accumulation and resolution.

C. Data Analysis:

• Processing: DIA-NN (v1.8.1) against a Rattus norvegicus database.
• Normalization: Sum normalization per cell followed by log2 transformation and Z-score scaling.
• Statistical & Computational Analysis:
  • Dimensionality Reduction: PCA and UMAP for visualization.
  • Clustering: K-means and Non-Negative Matrix Factorization (NMF) to identify distinct proteomic states.
  • Enrichment: Metascape for pathway analysis.
  • Comparison: Bulk vs. Single-cell percentile ranking to identify proteins masked in population averages.

3. Key Contributions

1. Optimized Workflow for Adherent Neurons: Established a gentle, high-precision TIJ dispensing protocol specifically for fragile, adherent neuronal cells, achieving ~85% single-cell viability and accuracy while minimizing doublets.
2. Enhanced Membrane Protein Recovery: Demonstrated that the inclusion of the mild surfactant DDM significantly improves the detection of membrane-associated and low-solubility proteins (e.g., vesicular and cytoskeletal proteins) in single-cell lysates.
3. High-Depth Profiling: Achieved consistent quantification of 2,000–3,000 proteins per cell across all differentiation stages, providing sufficient depth to resolve complex biological states.
4. Unmasking Heterogeneity: Showed that bulk proteomics averages out critical subpopulations, whereas SCP reveals distinct cellular states (e.g., cells with neurite outgrowth vs. those without) within the same time point.

4. Key Results

A. Workflow Performance:

• The optimized workflow yielded high reproducibility with precursor-level coefficients of variation (CV) peaking at 30–40%.
• DDM-treated samples showed significantly higher protein counts (p=0.011) and better recovery of structural and membrane proteins compared to detergent-free samples.

B. Temporal Proteome Remodeling:

• Day 0 vs. Day 2: Early changes included downregulation of lysosomal proteins and lipid metabolism, and upregulation of adhesion-related proteins, signaling the exit from the proliferative state.
• Day 4 vs. Day 6: Significant downregulation of cell cycle markers (Mki67, Cdk1) and upregulation of neuronal cytoskeletal proteins (NEFM, NEFL, TUBB2A, Prph).
• Dynamic Range: The dataset covered a dynamic range of four orders of magnitude, detecting both high-abundance structural proteins and low-abundance signaling molecules.

C. Discovery of Cellular Heterogeneity:

• Subpopulations: UMAP and NMF analysis revealed that at Days 4 and 6, the cell population splits into distinct subgroups that are invisible in bulk data.
  • Group A (Differentiated): High abundance of endosomal trafficking proteins (SNX1, VPS26B), translational machinery, and neuronal maturation markers (NEFL, VGF). These cells exhibited extensive neurite outgrowth.
  • Group B (Less Differentiated): Lower abundance of trafficking and cytoskeletal proteins, lacking neurite extension.
• Bulk vs. Single-Cell Discrepancy: Many proteins associated with neuronal growth (e.g., Cend1, Baiap2, Pick1) appeared highly abundant in bulk data but were actually restricted to specific subpopulations in single-cell data. Bulk measurements masked this heterogeneity, creating a "false" average signal.
• Pathway Shifts: Early differentiation was characterized by a shift from ribosomal/translation machinery (Cluster 1) to RNA processing and nucleocytoplasmic transport (Cluster 6), reflecting the transition from proliferation to specialized neuronal function.

5. Significance

• Methodological Advancement: This study provides a validated, accessible pipeline for single-cell proteomics of difficult-to-handle neuronal cells, overcoming barriers of adhesion, fragility, and low protein yield.
• Biological Insight: It challenges the view of neuronal differentiation as a uniform process. Instead, it reveals a bimodal or multimodal progression where cells within the same culture adopt distinct proteomic trajectories.
• Clinical/Research Implication: The ability to detect rare, functionally distinct subpopulations (e.g., cells successfully committing to a neuronal fate vs. those failing to differentiate) is crucial for understanding neurodevelopmental disorders and optimizing regenerative medicine strategies.
• Proteome vs. Transcriptome: By focusing on the proteome, the study highlights post-transcriptional regulation and protein-level heterogeneity that transcriptomics alone might miss, offering a more accurate functional map of neuronal maturation.

In conclusion, this work demonstrates that single-cell proteomics is essential for resolving the complex, heterogeneous nature of neuronal differentiation, uncovering molecular states and regulatory pathways that are completely obscured by traditional population-averaged approaches.