Fornix microstructural integrity and temporal cognition in multiple sclerosis: a DTI tractography study

Introduction Temporal cognition, including the ability to judge the order and distance in time of past events, is frequently impaired in multiple sclerosis (MS), yet its neural substrates remain unclear. The fornix, a major white matter (WM) tract connecting the hippocampus with diencephalic memory regions, is a plausible candidate due to its role in episodic and temporal relational processing [1]. It is therefore conceivable that associations between fornix microstructure and cognitive performance reflect the influence of neural integrity on temporal cognition. However, no study has directly examined this relationship in MS. Previous diffusion MRI (dMRI) studies have mainly investigated fornix lesion burden and global cognitive impairment [2,3], leaving its role in temporal order processing unexplored. This study tested whether fornix microstructure is associated with temporal cognition PwMS compared with HC. Methods Twenty three PwMS (age, years 38±9; 7M/16F) and twenty one HC (age, years 33±7; 10M/11F) underwent dMRI. Fibre orientation distributions were computed using constrained spherical deconvolution (MRtrix3), and TractSeg reconstructed the fornix (Fig.1) Indices of fractional anisotropy (FA), mean (MD), and radial (RD) diffusivities were obtained from tensor fitting and extracted along tractometry profiles using a 10% trimmed mean to minimise cerebrospinal fluid contamination. Temporal cognition was assessed with two temporal-order memory tasks: Time-of-Occurrence (error in dating scenes from a brief narrative) and Recency (accuracy in judging which scenes occurred first). Due to non-normal distributions, Mann–Whitney and Spearman tests were applied. In PwMS, associations between fornix metrics, EDSS, disease duration, and global T2 lesion load were examined. Partial correlations, controlling for lesion burden, assessed relationships between fornix microstructure and temporal cognition. Given the exploratory design and sample, p values were uncorrected. Results Whole-brain analysis showed that global T2 lesion load predicted widespread white-matter FA reduction (R²=.637, p<.001) but had no effect on grey-matter (GM) FA (R²=.098, p=.119); PwMS also showed lower whole-brain white-matter FA than HC (p=.048), with no GM differences (p=.233). At the fornix level, PwMS showed higher right-fornix MD (p=.013) and RD (p=.039) than HC, with no FA differences (Fig.2). Fornix diffusivity correlated with global T2 lesion load: higher lesion burden was associated with higher MD/RD and lower FA, but not with EDSS or disease duration, suggesting diffuse WM damage is independent of disability. In HC, higher fornix MD and RD tended to accompany poorer temporal cognition (greater Time of Occurrence error, lower Recency accuracy, though none reached significance. In PwMS, partial correlations controlling for lesion load revealed the opposite pattern: higher MD predicted better Recency accuracy bilaterally (right: ρ=.53, p=.009; left: ρ=.46, p=.028) and lower Time-of-Occurrence error on the left (ρ=-.43, p=.042; right: ns) (Fig.3). RD showed a similar, weaker pattern for Recency (right: ρ=.44, p=.035; left: ρ=.38, p=.074); FA showed no association with either task. Discussion The results show that the fornix is altered in PwMS and that microstructural changes relate to temporal cognition nonlinearly. Although PwMS showed higher fornix diffusivity than HC, structure–function associations diverged: in HC, higher diffusivity tended to indicate poorer temporal-order performance, whereas in PwMS, after controlling for global T2 lesion load, higher MD and similarly RD were associated with better Recency accuracy and lower Time-of-Occurrence error. This inversion was independent of lesion burden, and fornix diffusivity was unrelated to EDSS or disease duration, excluding clinical severity as a confound. Overall, the pattern suggests a reorganisation of temporal-judgement strategies, potentially reflecting compensatory neuroplasticity and cognitive reserve mechanisms in MS. These results remain preliminary and require replication in a larger cohort. Conclusion Fornix MD shows an inverted association with temporal cognition in MS compared with healthy controls, even after adjusting for total lesion load. This bilateral pattern positions the fornix as a potential marker of covert cognitive adaptation in MS, capturing compensatory processes not reflected in conventional disability measures.