A Deep Quantitative Proteome Turnover Platform for Human iPSC-derived Neurons

This study establishes a comprehensive quantitative proteome turnover platform for human iPSC-derived neurons, enabling the deep measurement of over 10,000 protein half-lives and providing a publicly accessible resource (NeuronProfile) to advance neurological disease research and therapeutic development.

The Gist

Imagine your brain is a bustling, high-tech city. The neurons are the buildings, and the proteins inside them are the workers, machines, and maintenance crews keeping everything running. Just like any city, these buildings need constant repairs. Old, broken parts must be torn down (degraded), and new ones must be built (synthesized). This constant cycle of "tear down and rebuild" is called protein turnover.

For a long time, scientists could only watch this construction site in the "cities" of mice or rats. But human brains are different, and we needed to see the human construction site to understand diseases like Alzheimer's or ALS. The problem? Human neurons are stubborn. They don't divide like other cells, and they are incredibly hard to study in a lab. Plus, there are so many different types of proteins (some are super common, some are rare) that it's like trying to count every single brick in a skyscraper while it's being built.

Here is what this paper did, explained simply:

1. Building a Better Construction Site (The Method)

The researchers created a super-advanced way to track how fast these protein "workers" are replaced in human neurons grown from stem cells (iPSCs).

• The "Heavy" Switch: Imagine you have a construction crew wearing light-colored shirts. Suddenly, you switch them to heavy, dark-colored shirts. As the old light shirts wear out and get thrown away, new dark shirts appear. By counting the ratio of light to dark shirts over time, you can figure out exactly how fast the crew is being replaced. This is called dSILAC (a fancy way of saying "switching amino acid labels").
• The Deep Dive: Previous attempts were like looking at the city from a helicopter—you could see the big buildings but missed the small details. This team used a "deep dive" approach. They broke the proteins down into tiny pieces (peptides), sorted them into thousands of tiny bins (fractionation), and used a super-powerful microscope (Mass Spectrometry) to count them.
• The Result: They didn't just count a few hundred workers; they tracked 10,792 different proteins and over 160,000 unique parts. That is a massive upgrade from previous studies that could only see a few thousand.

2. Two Types of Neighborhoods: Cortical vs. Motor Neurons

The team studied two specific types of "buildings" in the brain city:

• Cortical Neurons: The "thinking" neurons in the outer layer (like the city's library and university).
• Motor Neurons: The "movement" neurons that send signals to your muscles (like the city's delivery trucks).

The Big Discovery:
Surprisingly, the overall speed of construction and demolition was almost the same for both types. Whether it was a thinking neuron or a movement neuron, the average protein lasted about 4 days before being replaced. The city runs on a similar schedule everywhere.

But, the specific jobs were different:

• The Delivery Trucks (Motor Neurons): Proteins needed for long-distance travel (like the long axons that reach your muscles) were replaced faster. It's like a delivery truck that drives 500 miles a day; its tires and engine parts wear out faster and need swapping more often.
• The Thinkers (Cortical Neurons): Proteins involved in complex chemical signaling and metabolism were replaced faster here. It's like a high-tech server room where the software updates and hardware patches happen constantly to keep the brain processing information.

3. The "City Map" (NeuronProfile)

The most exciting part isn't just the data; it's what they did with it. They built a free, interactive website called NeuronProfile.

Think of this as Google Maps for the brain's construction site.

• Before, if a drug developer wanted to know how long a specific protein lasts in a human neuron, they had to guess or use mouse data (which isn't always accurate).
• Now, anyone can go to the website, type in a protein name (like "Alpha-synuclein," which is linked to Parkinson's), and see exactly how fast it turns over, how much of it is there, and where it lives in the cell.

Why Does This Matter?

• Better Medicine: If you are designing a drug to fix a broken protein, you need to know how long that protein lasts. If it lasts 4 days, you might need to take the pill every day. If it lasts 4 weeks, you might only need it once a month. This map gives drug developers the exact schedule they need.
• Understanding Disease: Many brain diseases happen because the "demolition crew" stops working, and broken proteins pile up like trash. By knowing the normal speed of turnover, we can spot when the system is breaking down.
• Human Accuracy: Finally, we have a map of the human city, not the mouse city. This means the clues we find here are much more likely to lead to real cures for people.

In a nutshell: This paper built the first high-definition, real-time map of how human brain cells repair themselves. It shows us that while the general rhythm of the brain is consistent, the specific needs of different brain regions vary, and now, thanks to this new website, scientists everywhere can use this map to build better treatments for neurological diseases.

Technical Summary

1. Problem Statement

Understanding protein turnover (the balance between synthesis and degradation) is critical for elucidating neuronal homeostasis and developing therapies for neurological diseases. However, measuring protein half-lives in human postmitotic neurons presents significant challenges:

• Dynamic Range: Neuronal proteins exhibit a vast range of half-lives (from minutes to months), making simultaneous quantification difficult.
• Limited Coverage: Previous studies using Stable Isotope Labeling by Amino acids in Cell culture (SILAC) in human neurons were restricted to fewer than 4,000 proteins, missing low-abundance drug targets and signaling regulators.
• Species Differences: Most existing turnover data comes from rodent models, which may not accurately translate to human physiology.
• Technical Complexity: SILAC experiments in neurons suffer from missing values, complex data modeling, and high spectral complexity due to the presence of both light and heavy isotope peaks.

2. Methodology

The authors established a comprehensive, optimized workflow to measure global protein turnover in human induced pluripotent stem cell (iPSC)-derived neurons.

• Cell Models: Two distinct human neuronal subtypes were differentiated from wild-type iPSCs: glutamatergic cortical neurons and spinal motor neurons.
• Dynamic SILAC Labeling:
  • Cultures were switched from "light" media to "heavy" media containing $^{13}C_6^{15}N_2$-Lysine and/or $^{13}C_6^{15}N_4$-Arginine.
  • Samples were harvested at multiple time points (1, 2, 4, and 6 days) to capture degradation kinetics.
• Sample Preparation & Fractionation:
  • Proteins were digested using Trypsin/Lys-C.
  • To reduce spectral complexity and increase depth, the authors compared single-shot, HLB solid-phase extraction (SPE), and offline high-pH reverse-phase liquid chromatography (LC) fractionation. Offline LC fractionation was selected as the optimal method.
• Mass Spectrometry (LC-MS/MS):
  • Both Data-Dependent Acquisition (DDA) and Data-Independent Acquisition (DIA) were employed on a Thermo Fisher Q-Exactive HF-X.
  • DIA was found to provide superior quantification with fewer missing values and better reproducibility compared to DDA.
• Data Analysis Pipeline:
  • Identification: MaxQuant (DDA) and Spectronaut (DIA) were used against a human database and a custom neuron-specific contaminant library.
  • Quantification: Heavy-to-light ratios were calculated. Relative Isotope Abundance (RIA) was fitted to a first-order exponential decay model ($RIA = e^{-k_{loss}t}$) to determine decay rates ($k_{loss}$).
  • Half-life Calculation: Protein half-lives were derived as the harmonic mean of unique peptides belonging to each protein. Strict filtering (e.g., $R^2 > 0.8$, minimum 3 time points) ensured data accuracy.
• Bioinformatics: Statistical analysis (R), Gene Ontology (GO) enrichment, and subcellular localization mapping were performed.

3. Key Contributions

• Deep Proteome Coverage: The platform successfully quantified 10,792 protein half-lives and 162,854 unique peptides, representing a >2.5-fold increase over previous human neuron turnover studies.
• Optimized Workflow: The study validated a specific combination of offline LC fractionation and DIA-MS as the superior approach for deep, accurate SILAC quantification in complex neuronal samples.
• Comparative Analysis: It provided the first direct, deep comparison of protein turnover kinetics between two major human neuronal subtypes (cortical vs. motor).
• Public Resource (NeuronProfile): The authors launched NeuronProfile (www.neuronprofile.com), an interactive web platform allowing the community to query protein half-lives, abundances, and subcellular locations.

4. Key Results

• Global Conservation: Despite functional differences, global protein turnover is largely conserved between cortical and motor neurons, with a median half-life of approximately 4.0–4.2 days.
• Organelle-Specific Trends:
  • Fastest turnover (<1 day): E3 ubiquitin ligases, microtubule regulators, synaptic proteins, and SNARE complexes.
  • Slowest turnover (>10 days): Nucleosome (histones) and chromatin proteins.
  • Complex Consistency: Proteins within the same complex (e.g., v-ATPase, OXPHOS) exhibited tightly clustered half-lives.
• Subtype-Specific Differences:
  • Motor Neurons: Showed faster turnover of axonal fasciculation proteins (supporting long axon growth) and higher stability of potassium voltage-gated channels. They also exhibited higher abundance of metabolic proteins and cholinergic markers (e.g., CHAT).
  • Cortical Neurons: Showed faster turnover of glycosylation/metabolism proteins (consistent with high glutamate secretion energy costs) and higher stability of sodium channels. They were enriched in synaptic vesicle proteins and glutamate transporters.
• Turnover vs. Abundance: The study revealed complex relationships between protein abundance and half-life. For example, neurofilament proteins were more abundant but turned over faster in motor neurons compared to cortical neurons.
• Cross-System Comparison: When compared to rodent primary neurons and mouse brain tissues, human iPSC-derived neurons clustered closely with rodent primary neurons. However, synaptic proteins in in vivo tissue samples showed significantly longer half-lives than in culture models, likely due to developmental maturity.

5. Significance

• Therapeutic Development: The dataset provides critical parameters (half-lives, abundances) for rational drug design, including predicting target engagement duration, dosing frequency, and the efficacy of targeted protein degradation strategies for neurological disorders.
• Disease Modeling: By establishing a baseline for human neuronal proteostasis, this resource enables the identification of turnover defects in neurodegenerative diseases (e.g., ALS, Alzheimer's) where protein aggregation is a hallmark.
• Community Resource: The NeuronProfile platform democratizes access to high-quality human neuronal proteome data, facilitating hypothesis generation and cross-study comparisons without the need for specialized mass spectrometry infrastructure.
• Methodological Benchmark: The study sets a new standard for quantitative proteomics in postmitotic cells, demonstrating how to overcome the technical barriers of dynamic range and spectral complexity in SILAC experiments.