Deep axonal proteomics of human iPSC-derived neurons by microfluidic separation and DIA-MS

Mapping the Long-Distance Wires of the Brain

Neurons are highly specialized cells that rely on extreme structural compartmentalization to function. They consist of a soma (the cell body), dendrites (receivers), and axons—long, slender "wires" designed to transmit electrical signals over vast distances. Because these axons can span more than a meter in humans, they require their own dedicated supply of proteins and organelles to maintain health far away from the central cell body.

For years, scientists have struggled to study the specific molecular makeup of these axons. Most proteomic studies—the large-scale analysis of all proteins in a sample—rely on "whole-cell" lysates. These mix the axon with the soma and dendrites. This is like trying to understand the specialized components of a city's power lines by analyzing a sample taken from the entire metropolitan area. The signal from the wires gets lost in the noise of the buildings. Consequently, the unique vulnerabilities of axons in neurodegenerative diseases like ALS and hereditary spastic paraplegia (HSP) have remained difficult to pin down.

A new study combines microfluidic separation with ultra-sensitive mass spectrometry to solve this isolation problem. By physically separating the axons from the cell bodies before analysis, the researchers have produced the deepest human axonal proteome reported to date.

The struggle to isolate the wire

The primary obstacle to studying axonal biology is the difficulty of harvesting enough pure axonal material. Axons and cell bodies are densely intermingled in standard cultures. This makes physical separation a massive technical hurdle. Historically, researchers have turned to proximity labeling—tagging proteins near the axon with a chemical marker—to identify axonal components. While useful, the authors note that this method often misses low-abundance proteins. It also risks introducing artifacts by overexpressing non-native tags.

Other methods attempt physical separation but often suffer from inconsistent fluidic isolation or contamination from dendrites (the branching receivers of the neuron). Even when successful, the amount of protein recovered from these isolated axons is incredibly small. Researchers estimate only about 50 nanograms are available. This tiny sample size is typically insufficient for conventional Data-Dependent Acquisition mass spectrometry (DDA-MS). DDA-MS is a method that selects only the most abundant ions for analysis. This effectively ignores the "quiet" parts of the proteome.

Separating compartments with microfluidics

To overcome these limitations, the researchers implemented a workflow that integrates microfluidic compartmentalization with high-sensitivity Data-Independent Acquisition mass spectrometry (DIA-MS). The process begins with the use of XonaChips. These are microfluidic devices featuring parallel chambers connected by long, narrow tunnels. These tunnels are roughly 900 $\mu$m long and only 7 $\mu$m wide. They act as a physical filter. They allow axons to grow through while preventing the larger somata and dendrites from passing .

Once the neurons have matured, the researchers perform a discrete harvest. They use a controlled lysis (the breaking down of cell membranes to release contents) process. They apply buffer to the axonal compartment while maintaining a larger volume on the somatodendritic side. This prevents backflow. This ensures that the resulting protein lysate is specifically enriched for axonal components.

The crucial leap in depth comes from the choice of mass spectrometry hardware. Instead of traditional DDA-MS, the authors utilized an Orbitrap Astral mass spectrometer operating in DIA-MS mode. In DIA-MS, the instrument does not pick and choose specific ions. Instead, it systematically fragments all ions within defined mass windows. This approach, combined with the high sensitivity of the Astral analyzer, allows the team to extract maximum information from the precious 50 ng of axonal material.

Achieving unprecedented proteomic depth

The results of this integrated workflow show a massive increase in the number of proteins identified. The authors report quantifying approximately 9,676 proteins in the somatodendritic compartment and 7,871 in the axonal compartment [Figure 2A]. This represents a significant improvement over previous human axonal studies. Those earlier studies identified only about 500 protein groups.

The study finds that the axonal and somatodendritic compartments are fundamentally distinct. Principal Component Analysis (PCA)—a statistical technique used to simplify complex datasets by identifying the main axes of variation—showed a clear separation between the two compartments [Figure 3A]. While many proteins are shared, the researchers identified 1,250 proteins that are significantly enriched in the axon. These proteins are heavily involved in vesicle-mediated transport and synaptic vesicle dynamics. These are the processes that allow neurons to communicate at their terminals [Figure 3G].

Beyond mere identification, the authors used the tool to compare different types of neurons. They compared cortical neurons to lower motor neurons. They identified a "core" axonal proteome of 417 proteins. These remain consistent across different cell types [Figure 4E]. However, they also found subtype-specific signatures. Lower motor neuron axons were enriched in mitochondrial components. Cortical neuron axons showed a surprising enrichment in lysosomal and autophagic machinery [Figure 5K].

Understanding the limits of the map

While this represents a significant technical milestone, the study has notable constraints. First, the researchers note that the somatodendritic compartment is not a "pure" soma. It contains various axonal elements. Therefore, it should be viewed as somatodendritic-enriched rather than strictly somatic. This nuance is vital for anyone using this data to build comparative models.

Second, the proteomic depth varies by cell line. In the CN2 cortical neuron line, the axonal samples showed higher variability and fewer protein identifications than the CN1 line. The authors attribute this to differences in axonal growth rates [Figure 2B]. For a practitioner, this means that the "depth" of your data may be highly dependent on the specific biological performance of your cell culture.

Finally, the subtype-specific differences cannot be definitively attributed to neuronal identity alone. The cortical and motor neuron lines used in the study differed in both genetic background and differentiation strategy. These variations might be influenced by the way the cells were grown rather than purely by their biological classification.

A new reference for neurodegeneration

Is this workflow ready for the lab? For researchers investigating neurodegeneration, the answer is a qualified yes. The authors have provided a quantitative reference map for 100 of the 112 known genes associated with ALS and HSP [Figure 4G]. This allows investigators to move beyond asking "is this protein mutated?" They can ask how a mutation changes the protein's distribution between the soma and the axon.

The paper establishes that many disease-linked proteins do not show preferential axonal enrichment in healthy cells. This suggests that the "selective vulnerability" of axons in diseases like ALS might not be caused by proteins simply being located there. Instead, it may arise from how those proteins behave once they arrive. This workflow provides the necessary resolution to test that hypothesis. It moves the field toward a more spatially aware understanding of how brain cells fail.

Figures from the paper