Blood Proteins Predict ALS Years Before Symptoms Strike

A single blood draw taken years before any symptom appears may one day flag who is on the path to amyotrophic lateral sclerosis, according to a large prospective proteomic study drawing on the UK Biobank. Researchers tracked 52,584 middle-aged and older adults who had 2,920 plasma proteins measured at baseline, then waited to see who developed ALS. Over a mean follow-up of 11.8 years, 291 participants received an ALS diagnosis, 34 of whom carried a pathogenic hexanucleotide repeat expansion in C9orf72, the most frequent genetic cause of ALS in people of European ancestry. The prospective design matters enormously: unlike cross-sectional studies that compare symptomatic patients with healthy controls, this approach captured the biological state of the body during the silent, presymptomatic period when motor neurons are already degenerating but weakness, dysphagia and dysarthria have not yet emerged.

The headline result is that nine plasma proteins showed statistically significant associations with future ALS risk after false discovery rate correction. Neurofilament light polypeptide, or NEFL, a structural protein released when neurons are damaged, dominated the signal with a hazard ratio of 2.897 per standardized increase, meaning participants with higher baseline levels faced nearly a threefold elevated risk. Elevated CSRP3, MYBPC1, MEGF10, MYOM3 and THBS4 also predicted increased risk, while higher levels of EGFLAM, ITGAV and CD248 were associated with reduced risk. When the researchers combined these nine proteins with age and sex into a prediction model, it achieved an area under the curve of 0.867 for predicting ALS onset within three years, and exceeded 0.78 across all evaluated time horizons. In a randomly allocated internal validation cohort of roughly half the participants, NEFL, CD248, EGFLAM, ITGAV and MYBPC1 remained significantly associated with ALS risk.

Stratifying by C9orf72 expansion status revealed strikingly different proteomic landscapes between genetic and non-genetic forms of the disease. For the 257 incident cases without the C9orf72 expansion, four proteins were significant: NEFL again led with a hazard ratio of 2.742, followed by MYBPC1, MYOM3 and CSRP3, all muscle-related proteins. For the 34 C9orf72-associated cases, NEFL showed an even stronger association with a hazard ratio of 4.043, and COL5A1, a collagen protein supporting structural integrity in skin, joints and blood vessels, emerged as an additional significant predictor with a hazard ratio of 3.072. The integrated model for predicting C9orf72-associated ALS onset within three years reached an area under the curve of 0.955, an exceptionally high figure for a disease that is usually diagnosed only after irreversible damage has accumulated.

The C9orf72 analysis tackled one of the most puzzling features of this genetic variant: incomplete penetrance. Of the 128 participants carrying expansions exceeding 30 repeats, only some developed ALS during follow-up, and the study confirmed that disease risk is not strictly proportional to repeat length. Carriers with 30 to 100 repeats developed ALS, while some individuals with thousands of repeats remained disease-free, and no significant correlation existed between expansion size and age at onset. Within this carrier subcohort, only NEFL remained significantly associated with ALS risk after false discovery rate correction, with elevated baseline levels predicting subsequent onset at a hazard ratio of 3.378. Carriers who later developed ALS had significantly higher baseline NEFL levels than nonprogressors, though no linear relationship linked repeat size to plasma NEFL concentration, suggesting that NEFL reflects the biological transition from carrier status toward symptomatic disease rather than the underlying genetic burden itself.

Temporal modeling of plasma protein trajectories across the disease course produced perhaps the most vivid finding of the entire study. Plasma NEFL levels in future ALS patients begin a gradual increase approximately 9.27 years before diagnosis, offering a window of presymptomatic detection that far exceeds the 12-month predictive window reported in earlier cross-sectional work. In non-C9orf72 cases, NEFL peaked one to two years before diagnosis before stabilizing, whereas C9orf72-associated cases showed a steeper, more sustained rate of increase without a distinct prediagnostic peak, consistent with prior observations that this genetic subtype produces higher blood NEFL concentrations. The researchers also linked higher prediagnostic NEFL levels to a younger age at onset in all-type ALS and non-C9orf72 ALS, with regression coefficients of -1.501 and -1.440 years respectively, meaning each standardized NEFL increase predicted onset more than a year earlier. No protein showed a significant association with postdiagnostic survival after multiple testing correction in the all-type or non-C9orf72 cohorts, while in the small C9orf72-associated subgroup, 33 candidate proteins showed potentially significant associations with survival that the authors urge be interpreted with caution, particularly given possible violations of the proportional hazards assumption.

Functional enrichment analysis of the nominally associated proteins pointed toward muscle-related pathways, including actomyosin structure organization and muscle cell development, alongside cell adhesion molecules and PPAR signalling pathways shared across all ALS subtypes. The authors caution that circulating muscle-derived signatures do not conclusively demonstrate primary muscle pathology or cachexia, since these proteins may originate from neural, hepatic or inflammatory sources, but they cannot exclude the possibility that primary alterations in muscle physiology contribute to ALS pathogenesis. For the C9orf72 carrier subgroup, the most prominent enriched pathways involved immune receptor and cytokine receptor interactions, particularly IL-10-associated signalling, reinforcing emerging evidence that CD4-positive T cells releasing IL-5 and IL-10 contribute to C9orf72-associated pathology. This immunological signature had not previously been established with certainty in ALS proteomic studies and suggests that targeting cytokine or kinase-related pathways might help delay onset in expansion carriers.

Several of the identified proteins are entirely novel candidates in ALS research. EGFLAM, also known as Pikachurin, mediates synaptic signal transmission between photoreceptor neurons and bipolar cells in the retina, and its implication in ALS risk for the first time points to a potential role in neurodegeneration beyond the visual system. ITGAV, an integrin elevated in microglia and macrophages in the SOD1 G93A mouse model of ALS, showed therapeutic promise when neutralized with antibodies in preclinical settings, ameliorating motor deficits and prolonging survival in that animal model. THBS4, an adhesive glycoprotein mediating cell-matrix and cell-cell interactions, has been implicated in promoting pro-inflammatory differentiation and apoptosis. CD248, or endosialin, showed reduced levels associated with increased ALS risk in this study, and prior work has suggested elevated CD248 may correlate with more favourable prognosis, hinting at a possible protective role during the preclinical stage.

Mendelian randomization using protein quantitative trait loci as instrumental variables identified a nominal causal role for MYOM3 in ALS, with genetically predicted elevated levels associated with increased risk at an odds ratio of 1.101. MYOM3 is involved in myopathies and muscle adaptation to eccentric exercise, and given epidemiological evidence linking strenuous physical activity to elevated ALS risk, dysregulation of MYOM3 or its pathways may provide a molecular mechanism connecting mechanical stress to motor neuron susceptibility. However, colocalization analysis found no significant evidence for shared causal variants between the MYOM3 protein loci and ALS genetic signals, meaning horizontal pleiotropy cannot be excluded and causal claims remain limited. For the other proteins, the lack of significant genetic support suggests their circulating abnormalities are more likely downstream consequences of disease pathology than genetically driven causes.

The study carries substantial strengths, including its prospective design, large sample size, long follow-up and the use of the Olink proximity extension assay, which offers high sensitivity and specificity across a broad dynamic range. But the authors emphasize important limitations: proteomic measurements were available only at baseline, so longitudinal profiling is needed to validate the temporal protein trajectories; there was no independent external validation cohort, making the novel associations discovery-phase findings; the C9orf72 carrier subcohort was small, rendering subtype-specific conclusions preliminary; and the cohort was 93.3 percent White, potentially limiting generalizability. Even so, the work demonstrates that a markedly smaller protein panel than previously proposed can prospectively stratify ALS risk years before symptoms, providing a foundation for biomarker-based risk stratification, early intervention trials and personalized monitoring strategies in high-risk individuals, including C9orf72 expansion carriers who currently have no way of knowing whether their genetic burden will ever translate into disease.

Subject of Research: Prospective plasma proteomic biomarkers for predicting incident amyotrophic lateral sclerosis stratified by C9orf72 repeat expansion status

Article Title: Plasma Proteomic Signatures Predict Incident Amyotrophic Lateral Sclerosis: A Prospective Study Stratified by C9orf72 Expansion

Article References: Su, W.-M., Duan, Q.-Q., Zhang, N., He, S.-Y., Liu, R.-Y., Wen, X.-J., Gu, X.-J., Wang, Y., Cao, B., & Chen, Y.-P. (2026). Plasma Proteomic Signatures Predict Incident Amyotrophic Lateral Sclerosis: A Prospective Study Stratified by C9orf72 Expansion. Journal of Cachexia, Sarcopenia and Muscle, 17(5), Article e70405. https://doi.org/10.1002/jcsm.70405

Image Credits: AI Generated

DOI: 10.1002/jcsm.70405

Keywords: ALS, biomarkers, proteomics, NEFL, C9orf72, UK Biobank, neurodegeneration, prediction model, Mendelian randomization, plasma proteins, prospective cohort, Plasma