In vivo and in silico alpha-synuclein propagation dynamics: The role of genotype, epicentre, and connectivity

This study demonstrates that while alpha-synuclein-induced atrophy in synucleinopathies is generalizable across different genotypes and preformed fibril species, the specific pattern of disease spread is critically dependent on the initial inoculation site (epicentre) and regional connectivity, as evidenced by the failure of computational models to predict atrophy following hippocampal versus striatal inoculation.

The Gist

The Big Picture: A "Bad Seed" Spreading Through the Brain

Imagine your brain is a massive, bustling city with millions of roads (neurons) connecting different neighborhoods. In Parkinson's disease and related conditions, a specific protein called alpha-synuclein gets corrupted. Think of this protein as a "bad seed" or a "zombie virus."

Normally, these seeds are harmless. But when they get misfolded (twisted into the wrong shape), they become infectious. They jump from one neuron to another, forcing healthy neighbors to turn into "zombies" too. This causes the neighborhoods to crumble (atrophy), leading to the symptoms of the disease.

This study asked three big questions:

1. Does it matter what kind of bad seed we use? (Human vs. Mouse)
2. Does it matter where we plant the seed? (The Striatum vs. The Hippocampus)
3. Can we predict exactly how the city will crumble using a computer simulation?

---

Experiment 1: Planting the Seed in the "Motor Control Center" (The Striatum)

The researchers took mice and injected these "bad seeds" (called Preformed Fibrils or PFFs) into the striatum, a part of the brain that controls movement. They used two types of mice:

• Wild-Type (WT): Normal mice with no genetic risk.
• M83: Mice genetically engineered to be very sensitive to this disease (like a city with very weak buildings).

They also used two types of seeds:

• Mouse Seeds: Made from mouse protein.
• Human Seeds: Made from human protein.

What Happened?

• The "Weak City" Collapsed: The M83 mice got sick much faster than the normal mice. It's like dropping a heavy rock on a house of cards; the cards (M83 mice) fell apart immediately, while the brick house (WT mice) barely shook.
• The "Mouse Seed" Was Deadlier: Surprisingly, the M83 mice injected with mouse seeds got sicker and died faster than those injected with human seeds.
  • Analogy: Imagine a city that is already stressed. If you introduce a local rumor (mouse seed) that everyone recognizes and panics over, it spreads faster than a foreign rumor (human seed) that people are slightly more skeptical of. The "local" seed triggered a stronger, more toxic reaction in the "weak city."
• The Damage Spread: Over time, the damage didn't just stay where the seed was planted. It spread to connected neighborhoods, causing the brain to shrink (atrophy) and the mice to lose their ability to move, balance, and grip.

Experiment 2: Planting the Seed in the "Memory Hub" (The Hippocampus)

Next, the researchers asked: "What if we plant the seed in a different place?" They injected the seeds into the hippocampus, a hub for memory and learning, which is highly connected but not usually the first place Parkinson's attacks.

What Happened?

• Local Damage Only: The damage stayed mostly stuck in the hippocampus. The mice lost some volume in that specific area, but they did not develop the widespread motor problems (like shaking or falling) seen in the first experiment.
• The "Wrong Neighborhood" Theory: This suggests that just because a seed spreads doesn't mean the whole city collapses. If you plant a virus in a library (hippocampus), the library might burn down, but the power plant (striatum) keeps running, and the city keeps moving. The location of the starting point determines the type of disaster.

The Computer Simulation: The "Weather Forecast" for the Brain

The team built a computer model (an "in silico" model) to see if they could predict this damage before it happened. They fed the computer two pieces of information:

1. The Road Map: How connected different brain areas are (structural connectivity).
2. The Fuel: How much "bad seed" protein each area naturally produces (gene expression).

The Results:

• Striatum Prediction: The computer was a genius at predicting the damage when the seed was in the striatum. It correctly guessed which neighborhoods would crumble based on the road map and the fuel levels.
• Hippocampus Prediction: The computer failed when the seed was in the hippocampus. It couldn't predict the damage pattern.
  • Analogy: The computer model is like a weather forecast that works perfectly for rain in the valley but fails completely when predicting a flood in the mountains. It suggests that the "rules" of how the disease spreads change depending on where you start.

The Takeaway: It's Not Just About the Virus, It's About the City

This study teaches us three main lessons:

1. Genetics Matter: If you have a genetic vulnerability (like the M83 mice), the disease hits you much harder and faster.
2. Location is Key: Where the disease starts determines what symptoms you get. A seed in the motor center causes movement issues; a seed in the memory center causes memory issues. The "epicenter" dictates the outcome.
3. One Size Doesn't Fit All: We can use computer models to predict Parkinson's progression if we start in the "classic" spots (like the striatum), but these models might fail if the disease starts elsewhere. We need to understand that different parts of the brain have different "vulnerabilities."

In short: Parkinson's isn't just a simple virus spreading everywhere. It's a complex interaction between your genetic "building materials," the specific "seed" that starts the trouble, and the "neighborhood" where it begins. Understanding this helps scientists design better treatments that target the specific weak points of a patient's brain.

Technical Summary

1. Problem Statement

Synucleinopathies, such as Parkinson's Disease (PD), are hypothesized to progress via the prion-like spread of misfolded alpha-synuclein (aSyn) protein. While this "spreading hypothesis" is widely accepted, critical gaps remain regarding:

• Generalizability: Does aSyn spreading and subsequent neurodegeneration behave consistently across different genetic backgrounds (genotype) and species-specific aSyn strains (human vs. mouse preformed fibrils)?
• Regional Vulnerability: Is the pattern of neurodegeneration solely determined by structural brain connectivity (the "connectome"), or does regional gene expression (specifically SNCA) and the specific disease epicenter (injection site) dictate vulnerability?
• Predictive Modeling: Can computational models accurately predict atrophy patterns resulting from different disease epicenters (e.g., striatum vs. hippocampus)?

2. Methodology

The study employed a multi-modal approach combining in vivo longitudinal animal experiments with in silico computational modeling.

A. In Vivo Experimental Design

• Subjects: Wild-type (WT) C57BL/6 x C3H mice and hemizygous M83 transgenic mice (expressing human A53T mutant aSyn).
• Inoculation: Mice received stereotaxic injections into two distinct brain regions:
  1. Striatal Inoculation (Experiment 1): Right dorsal striatum (caudoputamen). Groups received Phosphate Buffered Saline (PBS), Human aSyn Preformed Fibrils (Hu-PFF), or Mouse aSyn PFF (Ms-PFF).
  2. Hippocampal Inoculation (Experiment 2): Right dentate gyrus (hippocampus). Groups received PBS or Hu-PFF.
• Longitudinal Assessment:
  • MRI: T1-weighted imaging at -7, 30, 90, and 120 days post-injection (dpi) using a 7T Bruker scanner (100 μm³ isotropic voxels).
  • Behavioral Testing: Pole test, Rotarod, and Wire hang test to assess motor deficits.
  • Survival: Monitoring for humane endpoints.
• Image Analysis: Deformation-Based Morphometry (DBM) was used to calculate voxel-wise Jacobian determinants to quantify local volumetric changes (atrophy) over time.
• Statistical Analysis: Linear mixed-effects models for longitudinal data, Cox proportional hazards for survival, and Partial Least Squares (PLS) analysis to correlate brain atrophy with behavioral/demographic variables.

B. In Silico Computational Modeling

• Model Type: A Susceptible-Infected-Removed (S-I-R) agent-based model adapted from previous work (Zheng et al., 2019; Rahayel et al., 2022).
• Mechanisms:
  • Agents: aSyn proteins exist in three states: Susceptible (normal), Infected (misfolded), or Removed (metabolized).
  • Propagation: Driven by structural connectivity data from the Allen Mouse Brain Connectivity Atlas (AMBCA).
  • Synthesis: Proportional to regional SNCA gene expression levels (Allen Mouse Brain Atlas).
  • Atrophy: Simulated as a function of "Infected" agent accumulation and deafferentation from connected atrophied regions.
• Validation: Simulated atrophy maps were compared against empirical MRI-derived atrophy maps using Spearman correlations.

3. Key Contributions

1. Systematic Comparison of Strains and Genotypes: The study provides a comprehensive comparison of disease progression across WT vs. M83 genotypes and Human vs. Mouse PFF species, revealing nuanced interactions between host genetics and inoculum origin.
2. Decoupling Connectivity from Vulnerability: It challenges the notion that structural connectivity alone drives atrophy, demonstrating that regional SNCA expression is a stronger predictor of vulnerability.
3. Epicenter-Specific Dynamics: The study reveals that the "disease epicenter" (striatum vs. hippocampus) fundamentally alters the propagation dynamics and the ability of computational models to predict pathology.
4. Validation of Computational Models: It validates S-I-R models for striatal seeding but highlights their limitations when applied to non-dopaminergic, non-classical PD epicenters like the hippocampus.

4. Key Results

A. Disease Progression and Survival

• Genotype Effect: M83 mice exhibited accelerated disease progression and significantly reduced survival compared to WT mice.
• PFF Species Effect: Within M83 mice, those injected with Mouse PFF (Ms-PFF) showed faster weight loss, more severe motor deficits, and significantly lower survival rates compared to those injected with Human PFF (Hu-PFF).
  • Hazard Ratio: M83 Ms-PFF had a ~70% higher hazard of reaching a humane endpoint compared to M83 Hu-PFF.
• Sex Differences: In M83 Hu-PFF mice, males reached humane endpoints significantly faster than females; no such difference was observed in Ms-PFF mice.

B. Neuroanatomical Changes (MRI/DBM)

• Striatal Inoculation: Induced widespread, time-dependent atrophy.
  • Pattern: Atrophy was most severe in M83 Ms-PFF mice, affecting regions connected to the striatum (e.g., substantia nigra, motor cortex, periaqueductal gray).
  • Latent Dimensions: PLS analysis revealed a latent variable linking widespread atrophy, motor decline, and Ms-PFF injection in M83 mice.
• Hippocampal Inoculation: Resulted in focal atrophy restricted to the bilateral hippocampus (dentate gyrus, CA1, CA3) with no widespread network atrophy and no overt motor deficits.

C. Drivers of Vulnerability (Connectivity vs. Gene Expression)

• Connectivity Alone: Correlation between empirical atrophy and structural connectivity was negligible ($r = 0.017$).
• Gene Expression: Correlation between atrophy and SNCA expression was moderate ($r = 0.482$).
• Combined Model: The S-I-R model combining connectivity and SNCA expression successfully predicted striatal atrophy patterns (Peak correlation $r = 0.618$ at 90 dpi).

D. Model Failure in Hippocampal Seeding

• The computational model failed to accurately predict atrophy patterns following hippocampal inoculation.
  • Peak correlation dropped to $r = 0.33$.
  • Neither connectivity nor SNCA expression alone could explain the empirical data ($r < 0.17$).
• This suggests that while the striatum follows a predictable propagation rule based on connectivity and gene expression, the hippocampus (a highly connected hub) exhibits differential regional vulnerability not captured by current models.

5. Significance and Implications

• Refining the Spreading Hypothesis: The findings suggest that aSyn propagation is not a uniform process. It is heavily constrained by regional vulnerability (defined by SNCA expression and local cellular environment) rather than just axonal connectivity.
• Therapeutic Implications: The differential response to Ms-PFF vs. Hu-PFF in M83 mice highlights the importance of species-specific factors (e.g., immune response, fibril conformation/strain) in preclinical modeling. Therapies targeting aSyn aggregation may need to account for these strain-specific interactions.
• Limitations of Current Models: The failure of the S-I-R model to predict hippocampal atrophy indicates that current computational frameworks may be insufficient for modeling synucleinopathies originating outside the nigro-striatal pathway. Future models must incorporate additional factors such as local inflammatory responses, protein clearance mechanisms, or functional connectivity.
• Translational Value: By validating models for striatal seeding, the study supports the use of these in silico tools for drug discovery in PD. However, the hippocampal results serve as a critical warning against over-generalizing these models to all synucleinopathy presentations (e.g., dementia with Lewy bodies).

In conclusion, the study demonstrates that while aSyn spreading is a robust phenomenon, the resulting neurodegeneration is highly dependent on the genotype, the species of the inoculum, and critically, the disease epicenter, necessitating a more nuanced understanding of regional vulnerability in synucleinopathies.