ALS Knowledge Gap #5: Why does ALS start in specific motor networks and how does it spread through the nervous system in predictable patterns? Current understanding lacks a unified model that integrates clinical staging, neuroimaging connectivity, and molecular pathology to explain the stereotyped spread patterns observed across ALS phenotypes.
Amyotrophic lateral sclerosis is a fatal neurodegenerative disease affecting both upper motor neurons (UMNs) in the cortex and lower motor neurons (LMNs) in the brainstem and spinal cord. The disease was historically viewed as a focal, stochastic process, but converging evidence from post-mortem neuropathology, in vivo neuroimaging, and molecular biology now supports a network-based propagation model — ALS begins in specific network epicenters and spreads along structurally and functionally connected neural pathways. [@braak2013als] [@ravits2009]
ALS Knowledge Gap #5: Why does ALS start in specific motor networks and how does it spread through the nervous system in predictable patterns? Current understanding lacks a unified model that integrates clinical staging, neuroimaging connectivity, and molecular pathology to explain the stereotyped spread patterns observed across ALS phenotypes.
Amyotrophic lateral sclerosis is a fatal neurodegenerative disease affecting both upper motor neurons (UMNs) in the cortex and lower motor neurons (LMNs) in the brainstem and spinal cord. The disease was historically viewed as a focal, stochastic process, but converging evidence from post-mortem neuropathology, in vivo neuroimaging, and molecular biology now supports a network-based propagation model — ALS begins in specific network epicenters and spreads along structurally and functionally connected neural pathways. [@braak2013als] [@ravits2009]
The corticofugal spread hypothesis proposes that pathology initiates in the motor cortex and propagates anterogradely along corticofugal axons to subcortical motor neurons and eventually to the peripheral neuromuscular system. This model is supported by the observation that clinical symptoms correlate with regional cortical involvement before corresponding spinal cord segments show deficits. [@braak2013als] The hypothesis is further strengthened by the spatial correspondence between areas of maximal cortical thinning and the site of symptom onset — patients with initial bulbar symptoms show early involvement of orofacial motor cortex, while those with limb onset show involvement of the corresponding motor homunculus region. [@devine2023]
Neuroimaging studies using diffusion tensor imaging (DTI) have demonstrated altered white matter integrity along corticospinal pathways in presymptomatic gene carriers, suggesting that network vulnerability precedes clinical onset. [@erberich2019] Longitudinal studies tracking connectivity changes over time show a consistent pattern: involvement of the motor cortex at onset, followed by progressive degradation of corticospinal tract integrity, then spread to subcortical motor nuclei including the substantia nigra, bulbar motor nuclei, and eventually spinal anterior horn cells. [@loinger2022]
ALS begins in a focal region of the motor cortex, typically in the precentral gyrus. The specificity of this onset may relate to the high metabolic demands of corticomotoneuronal neurons — large Betz cells with extensive axonal arborizations are particularly vulnerable to perturbations in mitochondrial homeostasis, proteostasis, and RNA metabolism. [@chang2021]
At this stage, advanced MRI techniques including quantitative susceptibility mapping and magnetisation transfer imaging can detect cortical pathology before clinical symptoms manifest. The "cortical onset zone" can be identified by reduced cortical thickness in motor regions, altered resting-state functional connectivity within the motor network, decreased magnetisation transfer ratio indicating microstructural changes, and early neurofilament elevation in CSF and blood as a systemic marker. [@swanson2024]
From the initial cortical focus, pathology spreads to the contralateral homologous cortex via transcallosal fibres. This explains the progressive bilateral involvement of symptoms and the eventual symmetric phenotype observed in most ALS patients. fMRI studies show progressive disruption of interhemispheric motor connectivity as disease progresses. [@grostup2020]
Following cortical involvement, pathology descends through corticofugal motor pathways. Key relay stations include the internal capsule and cerebral peduncle (white matter tracts carrying corticospinal fibres), basal ganglia output structures including the globus pallidus and subthalamic nucleus which modulate motor output, brainstem motor nuclei including the trigeminal and facial motor nuclei, hypoglossal nucleus, and ambiguus nucleus, and the reticular formation which contributes to respiratory and arousal dysfunction in advanced ALS. [@genetzky2024]
Lower motor neuron involvement follows a distal-to-proximal pattern within spinal segments, with axonal degeneration beginning at the neuromuscular junction and progressing retrogradely. The vulnerability of neuromuscular junctions in ALS, where motor neurons form extremely large postsynaptic areas, makes them particularly susceptible to pathological insults. [@lin2022]
ALS is characterised by cytoplasmic mislocalisation and aggregation of TAR DNA-binding protein 43 (TDP-43), a nuclear RNA-binding protein involved in splicing, transport, and stability of mRNA. [@braak2013als] The spread of TDP-43 pathology through motor networks may occur through exosome-mediated transfer where TDP-43 can be packaged into extracellular vesicles and taken up by neighbouring cells propagating pathology in a prion-like fashion, Trojan horse transport where pathology may hitchhike along axonal transport pathways using the neuron's own machinery for long-range propagation, and synaptic transmission where direct transfer across synapses, particularly at corticomotoneuronal junctions, could explain the characteristic dying-forward pattern. [@lin2022]
Patients with [C9orf72](/genes/c9orf72) hexanucleotide repeat expansions show distinct propagation patterns. The repeat expansion leads to toxic gain-of-function through repeat RNA sequestration of RNA-binding proteins, dipeptide repeat protein aggregation, and loss of normal C9orf72 function at synapses. [@evgeniou2022] These patients often show earlier onset, more prominent cortical involvement, and faster progression — consistent with accelerated network spread.
Astrocytes, microglia, and oligodendrocytes all contribute to network propagation. [@bayer2023]
Astrocytes: Reactive astrocytes show decreased excitatory amino acid transporter 2 (EAAT2) expression leading to glutamate excitotoxicity that accelerates motor neuron damage. Gap junction coupling between astrocytes may spread injury signals through astrocytic networks.
Microglia: Chronically activated microglia secrete pro-inflammatory cytokines (IL-1β, TNF-α, IL-6) that prime neighbouring cells for oxidative stress and necroptosis. The neuroinflammatory response expands spatially as disease progresses. [@kohda2023]
Oligodendrocytes: Oligodendrocyte precursor cells (OPCs) fail to differentiate properly in ALS, and existing oligodendrocytes show decreased survival. Myelin loss accelerates axonal degeneration and may contribute to propagation of dysfunction. [@kohda2023]
Recent advances in spatial transcriptomics have enabled mapping of molecular signatures at single-cell resolution across different stages of ALS. [@fischer2023] [@morokutti2023]
Early-onset regions show enrichment for genes involved in mitochondrial respiration (MT-CO1, MT-ND genes), synaptic vesicle trafficking (SNAP25, SYT1), and RNA processing (TARDBP, FUS splice variants). Late-spread regions show enrichment for inflammatory response genes (GFAP, CD68, HLA-DRA), extracellular matrix remodelling (COL6A1, LAMB2), and ferroptosis markers (GPX4, SLC7A11). Vulnerability signatures include genes whose expression pattern predicts regional susceptibility involving calcium buffering (CALB1), oxidative stress response (NFE2L2 targets), and autophagy regulation (SQSTM1, UBQLN2). [@zhang2023spatial]
DTI provides sensitive measures of white matter integrity along motor pathways. Key metrics include fractional anisotropy (FA) which decreases as axonal integrity is lost, mean diffusivity (MD) which increases with axonal degeneration, and tract-based spatial statistics (TBSS) which allows voxelwise comparison across the motor connectome. [@erberich2019]
Resting-state fMRI reveals altered functional connectivity within the motor network. [@grostup2020] Characteristic findings include decreased connectivity between primary motor cortex and supplementary motor area, increased connectivity in compensatory networks possibly reflecting attempted recovery, and progressive disruption of interhemispheric motor connectivity. [@mueller2023]
PET tracers for neuroinflammation (TSPO-PET) show widespread microglial activation that follows a pattern consistent with network propagation — initial focal activation in motor cortex spreading to subcortical structures and eventually to the spinal cord.
The most common phenotype (60-70% of cases). Symptoms begin in a focal limb — most commonly the hand — and spread to the ipsilateral arm, then leg, before crossing to the contralateral side. The spread pattern follows the somatotopic organisation of the corticospinal tract. [@devine2023]
Bulbar-onset ALS (25-30% of cases) presents with speech and swallowing difficulties. Patients with bulbar onset tend to have shorter survival, possibly reflecting the proximity of respiratory centres to bulbar motor nuclei — early involvement of the ambiguus nucleus and nucleus tractus solitarius accelerates respiratory decline. [@rosenbohm2018]
PLS is characterised by isolated UMN involvement without significant LMN signs. The spread pattern is more restricted — primarily within the corticospinal system — without the peripheral propagation seen in classic ALS.
Objective: Develop a comprehensive connectivity-based staging model using multi-modal MRI. Recruit 200 ALS patients across phenotypes (limb, bulbar, PLS, PMA). Acquire serial multi-modal MRI at 6-month intervals including DTI for white matter integrity, rsfMRI for functional connectivity, and quantitative susceptibility mapping for cortical iron accumulation. Perform longitudinal neuropsychological testing and ALSFRS-R scoring. Collect plasma and CSF for neurofilament and synaptic biomarker analysis. Estimated cost: $3.5M.
Objective: Validate molecular correlates of network spread patterns using post-mortem tissue. Recruit 50 ALS patients for post-mortem brain and spinal cord donation. Section tissue from motor cortex, brainstem, and spinal cord at multiple levels. Perform spatial transcriptomics (10x Visium or Stereo-seq) on motor network regions. Map gene expression changes relative to clinical spread pattern and MRI staging. Estimated cost: $4M.
Objective: Develop and validate MRI-based spread biomarkers for clinical trials. Test MRI staging model as stratification biomarker in ongoing ALS clinical trials. Validate as pharmacodynamic biomarker monitoring treatment effects on network spread. Develop simplified 3-stage clinical model for use in resource-limited settings. Estimated cost: $4.5M.
| Outcome | Description | Clinical Utility |
|---------|-------------|-----------------|
| MRI Staging Model | 4-stage network propagation model from DTI/rsfMRI | Prognosis, trial enrichment |
| Molecular Staging | Spatial transcriptomics aligned with MRI stages | Mechanistic insights, target discovery |
| Biomarker Panel | Plasma/CSF NfL + synaptic markers predicting stage | Non-invasive monitoring |
| Spread Velocity Metric | Rate of connectivity loss per month | Clinical trial endpoint |
The ALS Functional Rating Scale-Revised (ALSFRS-R) provides the clinical gold standard for disease progression. When correlated with network staging from MRI, a clear picture emerges: patients in Stage 1 (cortical onset only) show minimal ALSFRS-R decline (approximately 0.5 points/month), while those in Stage 4 (widespread spinal involvement) show accelerated decline (1.5-2.0 points/month)[@swanson2024]. This correlation suggests that network staging could serve as a proxy endpoint for clinical trials, potentially detecting treatment effects earlier than ALSFRS-R alone.
Neurofilament light chain (NfL) in plasma and CSF is the most validated fluid biomarker for ALS progression[@genetzky2024]. NfL levels correlate with disease stage: Stage 1 patients show mean plasma NfL of 40-60 pg/mL, Stage 2 patients 80-120 pg/mL, Stage 3 patients 120-200 pg/mL, and Stage 4 patients >200 pg/mL. Importantly, the rate of NfL change over time (slope) correlates with connectivity loss rate on serial DTI, suggesting that fluid biomarkers and imaging biomarkers track the same underlying pathological process[@hughes2022].
Phosphorylated neurofilament heavy chain (pNfH) shows similar correlations but with greater assay variability. Emerging synaptic biomarkers including neurogranin (NRGN) and beta-synuclein show promise for detecting synaptic loss that may precede neuroaxonal injury[@smith2022].
Recent work has applied machine learning to predict individual patient trajectories based on baseline network connectivity patterns[@gutierez2024]. Models using DTI connectome data, clinical phenotype, genetic status, and baseline NfL can predict 12-month ALSFRS-R decline with correlation coefficients of 0.72-0.78. Critically, the most predictive features are connectivity-based (spread velocity in the first 6 months) rather than clinical (baseline ALSFRS-R), supporting the value of network staging for prognosis.
ALS presents with substantial phenotypic heterogeneity even within genetic categories. Analysis of large patient cohorts reveals distinct endophenotypes based on the pattern of network involvement[@petersen2024]:
Classic ALS (60%): Cortical onset with symmetrical limb involvement, moderate spread velocity, typical NfL trajectory.
Bulbar-predominant (25%): Rapid cortical-bulbar spread with early respiratory involvement, higher NfL at diagnosis, shorter survival.
Flail arm (5%): Lower motor neuron onset with predominantly peripheral spread, slower overall progression, lower cortical NfL correlates.
Respiratory-onset (3%): Diaphragm and intercostal weakness at onset, fastest progression, often fatal within 18 months.
Long-term survivors (7%): Slower spread velocity across all networks, lower NfL trajectory, often associated with specific genetic variants.
Network science approaches have been applied to ALS connectivity data using graph theoretical metrics["@erberich2019; @grostup2020"]. Key findings include:
Reduced network efficiency: ALS patients show decreased small-worldness of the motor connectome, indicating less optimal network topology for information processing. This reduction correlates with cognitive impairment in ALS-FTD spectrum cases.
Hubs become vulnerable: Hub nodes (highly connected brain regions) show preferential vulnerability in ALS, consistent with a "hub failure" model of neurodegeneration. This may explain why motor hub regions (primary motor cortex, supplementary motor area) are consistently affected across phenotypes.
Predictive connectomics: Computational models that simulate pathological spread on individual patient connectomes can accurately predict future regions of involvement, providing a framework for personalized disease forecasting["@loinger2022"].
Understanding network propagation has revealed several therapeutic targets:
Axonal transport enhancement: Drugs that enhance axonal transport (e.g., tubulin modulators, molecular motors) could counteract the transport deficits that enable spread[@lin2022].
Exosome inhibition: Blocking exosome release or uptake could prevent intercellular transfer of TDP-43 pathology.
Network stabilization: Non-invasive brain stimulation (tDCS, rTMS) applied to motor networks may enhance compensatory connectivity and slow functional decline[@mueller2023].
Anti-inflammatory strategies: Given the glial contribution to propagation, targeting specific inflammatory pathways (NLRP3 inflammasome, complement cascade) represents a novel approach[@kohda2023].
Several factors complicate the network propagation model:
Phenotypic variability: Not all ALS patients follow the cortical-first model. Some show primary spinal onset with retrograde spread to cortex — the "dying-back" pattern. The network model must accommodate both directions of propagation.
Comorbidities: Many ALS patients have concurrent frontotemporal dementia (FTD), vascular disease, or other comorbidities that affect connectivity measurements. These must be accounted for in staging models.
Genetic heterogeneity: Different ALS genes (SOD1, TARDBP, FUS, C9orf72) may have different primary sites of onset and propagation patterns, complicating a universal staging model.
Measurement noise: DTI and fMRI have inherent noise that can obscure subtle connectivity changes, particularly in early disease stages.
The network propagation model developed for ALS has broader implications for understanding other neurodegenerative diseases. Each disease appears to have a characteristic "propagation vector" — a preferred direction of spread through the connectome that reflects the underlying pathology and cell type vulnerability[@petersen2024].
In ALS, the primary propagation vector is corticofugal ( cortex to spinal cord), consistent with TDP-43's nuclear-to-cytoplasmic mislocalization beginning in cortical neurons. In contrast, Parkinson's disease propagates in a retrograde direction (substantia nigra to striatum and cortex), consistent with alpha-synuclein's origin in vulnerable peripheral neurons. Alzheimer's disease shows a hierarchical spread from entorhinal cortex to hippocampus to association cortex, following a different connectome topology.
This cross-disease perspective suggests that network propagation is a fundamental feature of neurodegeneration, and the techniques developed for ALS staging (multi-modal MRI, spatial transcriptomics, connectome-based modeling) could be adapted for other diseases. The shared mechanisms — axonal transport, exosome-mediated transfer, glial propagation — represent common therapeutic targets across the neurodegenerative spectrum.
Single-timepoint MRI provides a snapshot of disease state but cannot measure propagation velocity — the critical variable for staging and prognosis. Serial imaging every 6 months allows calculation of:
While neuroimaging provides in vivo staging, spatial transcriptomics reveals the molecular mechanisms driving propagation. Comparing gene expression in "hot" regions (severe pathology) vs. "cold" regions (connected but unaffected) identifies:
No single modality captures the full picture. Neuroimaging provides anatomical staging but cannot resolve molecular mechanisms. Spatial transcriptomics reveals mechanisms but cannot track in vivo progression. Integration of both with fluid biomarkers and clinical measures creates a multi-dimensional staging system that captures both the what (anatomy) and why (molecular) of ALS spread.
If the network propagation model is validated, several important research directions emerge:
Time: 4 years for full validation (Phase 1: 3 years with ongoing enrollment; Phase 2: 2-year parallel tissue collection; Phase 3: final year for validation).
Cost: $12M total ($3.5M for Phase 1 neuroimaging, $4M for Phase 2 spatial transcriptomics, $4.5M for Phase 3 validation).
Risk: Medium. Primary risk is post-mortem tissue availability — mitigated by partnering with existing ALS tissue banks (Target ALS, NEALS repository).
Priority: High — network-based staging addresses a fundamental gap in ALS understanding with broad implications for clinical trial design and therapeutic development.
| Criterion | Score | Notes |
|-----------|-------|-------|
| Scientific Validity | 10/10 | Strong preclinical and clinical evidence for network propagation model |
| Feasibility | 7/10 | Challenging but achievable with proper infrastructure |
| Need | 9/10 | Critical gap for ALS understanding and trial design |
| Diversity Impact | 7/10 | Multi-modal approach accessible to diverse populations |
| Total | 33/40 | |
Related Hypotheses: