Integrating TSPO-PET imaging with metabolomics for enhanced prognostic accuracy in multiple sclerosis

Abstract

Background: Predicting disease progression in multiple sclerosis (MS) remains challenging. PET imaging with 18 kDa translocator protein (TSPO) radioligands can detect microglial and astrocyte activation beyond MRI-visible lesions, which has been shown to be highly predictive of disease progression. We previously demonstrated that nuclear magnetic resonance (NMR)-based metabolomics could accurately distinguish between relapsing-remitting (RRMS) and secondary progressive MS (SPMS). This study investigates whether combining TSPO imaging with metabolomics enhances predictive accuracy in a similar setting.

Methods: Blood samples were collected from 87 MS patients undergoing PET imaging with the TSPO-binding radioligand ¹¹C-PK11195 in Finland. Patient disability was assessed using the expanded disability status scale (EDSS) at baseline and 1 year later. Serum metabolomics was performed to identify biomarkers associated with TSPO binding and disease progression.

Results: Greater TSPO availability in the normal-appearing white matter and perilesional regions correlated with higher EDSS. Serum metabolites glutamate (p=0.02), glutamine (p=0.006), and glucose (p=0.008), detected by NMR, effectively distinguished future progressors. These three metabolites alone predicted progression with the same accuracy as TSPO-PET imaging (AUC 0.78; p=0.0001), validated in an independent cohort. Combining serum metabolite data with PET imaging significantly improved predictive power, achieving an AUC of 0.98 (p<0.0001).

Conclusion: Measuring three specific serum metabolites is as effective as TSPO imaging in predicting MS progression. However, integrating TSPO imaging with serum metabolite analysis substantially enhances predictive accuracy. Given the simplicity and affordability of NMR analysis, this approach could lead to more personalised, accessible treatment strategies and serve as a valuable tool for clinical trial stratification.

Keywords: MULTIPLE SCLEROSIS; PET.

Figure 1. Extralesional ¹¹C-PK11195 DVR is higher in multiple sclerosis patients with increased EDSS scores. DVR values for (A) the NAWM, (B) perilesional tissue, (C) thalamus and (E) T2 lesions were greater in progressors (ΔEDSS within 1 year of sampling, n=23), compared with non-progressors (n=52). No differences were observed in (D) the T1 lesion or (F) whole brain. (G) ROC curve analysis showed that NAWM DVR could distinguish between progressors and non-progressors, similarly to (H) the multi-region model. *p<0.05, **p<0.01 ***p<0.001. DVR, distribution volume ratio; EDSS, expanded disability status scale; NAWM, normal-appearing white matter.

Figure 2. Selected serum metabolites vary with brain inflammation levels classified by ¹¹C-PK11195 DVR. (A) Patients with high (n=39) and low (n=33) brain inflammation were classified by plotting NAWM DVR against perilesional DVR. Serum samples were measured by NMR and analysed using ROC curves and Student’s t-test. (B and C) Acetate, glutamate, glucose and glutamine showed moderate discriminatory capacity between high and low brain inflammation. (D) Acetate levels were lower in high inflammation (p=0.0099), while (E) glutamine (p=0.0064), (F) glucose (p=0.0076), and (G) glutamate (p=0.024) levels were increased. *p<0.05, **p<0.01, ****p<0.0001. DVR, distribution volume ratio; NAWM, normal-appearing white matter; NMR, nuclear magnetic resonance.

Figure 3. Serum glutamate, glucose and glutamine, but not acetate, independently predict disease progression. (A) Cross-validated accuracy of glutamate, glucose, glutamine and acetate, to predict progression, (B) combined metabolite model ROC AUC, (C) combined metabolite model ROC AUC with ¹¹C-PK11195 DVR data. (D) Serum glutamate, (E) glucose, and (F) glutamine levels in progressors and non-progressors. Pearson correlations between baseline EDSS and serum (G) acetate, (H) glucose, (I) glutamine, and (J) glutamate. **p<0.01, ***p<0.001, ****p<0.0001. DVR, distribution volume ratio; EDSS, expanded disability status scale.

Figure 4. Serum metabolites predict disease progression in an independent cohort (SMSC). Serum (A) glutamine, (B) glutamate, and (C) glucose levels in progressors (n=12) and non-progressors (n=25). (D) ROC AUC of logistic regression for individual metabolites: glutamate, glutamine, and glucose. (E) Combined metabolite model ROC AUC, (F) combined metabolite model ROC AUC including MRI data - T2 lesion number. *p<0.05, **p<0.01.