Kynurenine Study in MS Found Lower Levels Across Both Branches

The standard account of the kynurenine pathway in multiple sclerosis (MS) goes like this: chronic inflammation stimulates the pathway, the kynurenine-to-tryptophan ratio rises, and activity shifts toward the neurotoxic branch dominated by quinolinic acid. Kupjetz and colleagues tested that account in the largest kynurenine study yet conducted in MS, pooling 353 patients from five cohorts across Denmark, Germany, and Switzerland against 111 healthy individuals, with 11 metabolites measured by targeted mass spectrometry in a single laboratory. Participants had mild to moderate disease, disability scores at or below 6.5 with a mean of 3.1, none were on corticosteroids at the time of blood sampling, and relapsing patients were in remission. The two groups matched closely on age, 46.1 years against 45.7. What came back does not fit the standard account cleanly, and the way the paper handles that is the reason it is worth reading.

The Individual Metabolites Run the Wrong Way
Patients had a lower kynurenine-to-tryptophan ratio, not a higher one, and lower concentrations of most kynurenines: tryptophan, kynurenine, kynurenic acid, anthranilic acid, xanthurenic acid, picolinic acid, and quinolinic acid. That list spans both branches, the neurotoxic and the neuroprotective. Only three measures were higher in patients: neopterin, the quinolinic-to-kynurenic acid ratio, and 3-hydroxyanthranilic acid. Quinaldic acid and 3-hydroxykynurenine did not differ at all. The authors do not smooth this over. They write that earlier studies produced discrepant results on systemic kynurenine concentrations in MS, and that what their data reproduce is the single most consistent finding from human MS trials, the raised quinolinic-to-kynurenic acid ratio. Their explanation for the lower absolute levels is bioenergetic rather than inflammatory: liver and skeletal muscle contribute substantially to circulating kynurenines, so patients with lower physical activity and reduced fitness may simply turn the pathway over less. Reduced IDO1 expression in patient immune cells is offered as a second route.

Two Patterns, and How Much of Each Is One Measurement
Factor analysis on the patient cohort returned two factors explaining 54.2% of the variance. The first, labelled NeuroTox, loaded neopterin, kynurenine, the kynurenine-to-tryptophan ratio, 3-hydroxykynurenine, anthranilic acid, quinolinic acid, and the quinolinic-to-kynurenic ratio. The second, NeuroPro, loaded kynurenic acid, quinaldic acid, xanthurenic acid, and picolinic acid, with the quinolinic-to-kynurenic ratio loading inversely. Reading the loadings closely changes how much weight the word pattern can carry. Kynurenine at 0.991 and the kynurenine-to-tryptophan ratio at 0.998 essentially define NeuroTox, and the quinolinic-to-kynurenic ratio at −1.012 with kynurenic acid at 0.920 dominate NeuroPro. Each factor is driven by two measurements with the rest contributing modestly. The authors report the sampling adequacy statistic as 0.715 and describe it themselves as middling, and they set aside tryptophan as a precursor and 3-hydroxyanthranilic acid because it loaded on both factors.

The Clinical Associations Are Small and Do Not Survive Adjustment
NeuroTox correlated with disability score at r = 0.13 (p = 0.032, 95% CI 0.01–0.24) and NeuroPro at r = −0.17 (p = 0.003, CI −0.28 to −0.06). Squared, those describe roughly 2% and 3% of the variance in disability, and both confidence intervals come within a few hundredths of zero. In proportional odds regression the association reached significance in only one of five subcohorts. The paper states that in adjusted analyses the associations became nonsignificant, possibly because of the relationship between age and disability, and that significance returned only when age and study cohort were dropped from the models. Neither pattern was associated with walking capacity, fatigue, cognitive performance on either test, MRI metrics, or serum neurofilament light chain. The authors attribute those nulls to a well-functioning cohort at a mean disability score of 3.1 without extensive structural brain damage, which is a fair reading and also means the study cannot speak to the disability range where these questions matter most.

The Strongest Signals Are About the Body, Not the Disease
The largest correlations in the paper are not with MS. NeuroTox rose with body mass index at r = 0.30 (p < 0.001), with body fat percentage at 0.27 by bioimpedance and 0.38 by dual-energy X-ray absorptiometry (p = 0.027), and with age at 0.23. NeuroPro fell with body fat percentage at −0.20 and rose with cardiorespiratory fitness at 0.23. Every one of those exceeds the correlation with disability. The authors read it as adipose tissue driving proinflammatory stimulation of the pathway while regular exercise counteracts it, and they connect it to the established finding that obesity in childhood and adolescence raises MS risk and that obese patients accumulate disability faster. The reading is coherent. It also means these patterns track modifiable physiology at least as closely as they track the disease, which is the more actionable of the two conclusions.

What Is Not in the Models
Of the patients, 73.7% were receiving a disease-modifying treatment, including 65 on teriflunomide and 39 on ocrelizumab, while none of the healthy individuals were on anything. The covariates used to test the between-group metabolite differences were sex, age, body mass index, and sample storage time. Treatment was not among them, although it does enter the later regression models as a yes-or-no term. Since these drugs act on the immune activation that is supposed to drive the pathway, treatment status sits directly on the mechanism being measured, for the study's central comparison. The authors name two further omissions themselves: diet and time since last relapse, both of which they say would have been optimal to include and might have raised the variance their factors explain. The cohorts were also not matched on sex, at 67.1% female among patients against 53.2% among controls, a choice made deliberately to keep the smaller control group intact.

What the Design Cannot Say, and What Would Settle It
The authors are direct about the limits. A cross-sectional design supports no statement about the cellular or tissue origin of these metabolites, about exchange across the blood-brain barrier, or about causality. They did not sample cerebrospinal fluid, which they call the preferred material for studying kynurenines in central nervous system function, though they cite work in which serum kynurenines outperformed spinal fluid ones for predicting Parkinson disease. They also concede the interpretive move at the centre of the paper: the neurotoxic and neuroprotective labels were assigned from previous literature rather than measured here, and need validating in follow-up work using molecular markers of excitotoxicity or oxidative stress. Their own proposal for what comes next is specific and correct, calling for participants spanning the full disability range rather than clustering at a mean of 3.1, and repeated blood sampling rather than a single draw. Adding a treatment-matched comparison and dietary records would close the two confounds this dataset leaves open.

Disclaimer: This blog post is based on the cited research article and is intended for informational purposes only. It is not intended to provide medical advice. Please consult with a healthcare professional for any health concerns.

Reference:
Kupjetz, M., Langeskov-Christensen, M., Riemenschneider, M., Inerle, S., Ligges, U., Gaemelke, T., Patt, N., Bansi, J., Gonzenbach, R. R., Reuter, M., Rosenberger, F., Meyer, T., McCann, A., Ueland, P. M., Eskildsen, S. F., Nygaard, M. K. E., Joisten, N., Hvid, L., Dalgas, U., & Zimmer, P. (2025). Persons with multiple sclerosis reveal distinct kynurenine pathway metabolite patterns: a multinational cross-sectional study. Neurology: Neuroimmunology & Neuroinflammation, 12(6), e200461. https://doi.org/10.1212/NXI.0000000000200461