The Mammillary Body-Fornix Gate in Long COVID An exploratory structural and diffusion MRI study of tremor-like symptoms, internal vibrations, and neuromuscular fatigue

Technical Summary: The Mammillary Body-Fornix Gate in Long COVID

Problem Statement
Long COVID (post-COVID condition) presents with a heterogeneous clinical phenotype including post-exertional malaise (PEM), neuromuscular fatigue, internal vibrations, tremor-like symptoms, autonomic instability, and cognitive dysfunction. While neuroimaging studies have identified abnormalities in limbic, thalamic, brainstem, and white-matter systems, no single established circuit explains this specific constellation of symptoms. Existing literature suggests involvement of the hippocampus, thalamus, and brainstem but lacks a unified anatomical model linking these regions to the specific motor and autonomic symptoms observed in severe cases. This study addresses the gap by investigating whether abnormalities cluster at a specific anatomical convergence zone: the mammillary body-fornix-superior tuberal hypothalamic interface.

Methodology
This was an exploratory, cross-sectional structural and diffusion MRI study involving 122 participants: 88 with clinician-diagnosed Long COVID (33 of whom were bedridden) and 34 healthy controls.

• Imaging Acquisition: Data were acquired on a 3 Tesla Siemens Skyra scanner using high-resolution T1-weighted MPRAGE sequences for volumetric analysis and Diffusion Tensor Imaging (DTI) for microstructural assessment.
• Processing: Volumetric analysis utilized FreeSurfer 8.2 for segmentation of the mammillary bodies, superior tuberal/periventricular hypothalamic region, and fornix. DTI data were processed using FSL (v6.0.7.19) to derive Fractional Anisotropy (FA) and Axial Diffusivity (AD) maps. Tractography was performed using FreeSurfer Tracula.
• Operational Definitions: The study defined a "mammillary body-fornix gate" as the structural and functional interface where forniceal fibers approach the mammillary bodies within the periventricular hypothalamus. A key morphological observation was the "narrowing or loss of a visible internal passage" at the superior tuberal-mammillary interface.
• Clinical Measures: Participants underwent semi-structured clinical interviews based on the Canadian Consensus Criteria and physiological assessments including heart rate variability (RMSSD), surface electromyography, and classification of internal vibrations/tremor-like symptoms.
• Statistical Analysis: Group comparisons used one-way ANOVA with Welch corrections and Bonferroni-adjusted post-hoc tests. Effect sizes were calculated using partial eta-squared and Hedges' g. Correlations between imaging metrics and clinical variables were explored using Pearson correlations.

Key Contributions

1. Phenotypic Stratification: The study identified three distinct mammillary body volume phenotypes in the Long COVID cohort: reduced volume, enlarged volume, and control-range volume. This heterogeneity argues against a single uniform lesion model.
2. Anatomical Convergence: The paper proposes an operational "mammillary body-fornix gate" model. It highlights a specific morphological finding: segmentation-defined narrowing or loss of a visible internal passage at the superior tuberal-mammillary interface in a subset of patients.
3. Multi-System Clustering: The research links abnormalities in the hypothalamic-limbic interface with concurrent findings in brainstem-cerebellar pathways (specifically the dorsal raphe, midbrain reticular formation, and cerebellar peduncles), suggesting a network-level dysfunction rather than an isolated lesion.
4. Differentiation of Symptoms: The study explicitly treats "internal vibrations" and "tremor-like symptoms" as phenomenological descriptions requiring objective characterization, distinguishing them from established thermoregulatory shivering or specific movement disorder diagnoses.

Results

• Mammillary Body Volumes: Significant differences were found across the three phenotypic groups for both left and right mammillary body volumes (Left: $\eta_p^2 = .554$; Right: $\eta_p^2 = .559$). Segmentation revealed narrowing or loss of the internal passage at the superior tuberal-mammillary interface in 26 patients and 3 controls.
• Fornix Diffusion: The fornix showed altered diffusion measures and reduced tract coherence in the hypothesized gate region. Welch's test identified a significant group difference in right fornix FA ($p = .002$), with healthy controls showing higher FA than the Long COVID subgroup with enlarged mammillary body volume. Tractography suggested attenuated or discontinuous segments approaching the mammillary bodies.
• Brainstem and Cerebellar Findings: Long COVID participants exhibited significantly lower superior cerebellar peduncle volume ($p < .001$; Hedges' $g = 3.31$) and lower middle cerebellar peduncle FA ($p < .001$; Hedges' $g = 1.77$). Reduced volumes were also observed in the dorsal raphe and midbrain reticular formation.
• Clinical Associations: Exploratory associations were observed between the identified imaging abnormalities and motor deficits, proprioceptive dysfunction, autonomic dysregulation, fatigue, and internal vibrations.

Significance and Claims
The authors position this work as an exploratory study proposing a testable neuroanatomical model for a specific subgroup of severe Long COVID patients. The findings support a "mammillary body-fornix gate" hypothesis where a vulnerable periventricular hypothalamic-limbic interface may contribute to network dysfunction involving memory, homeostasis, arousal, and motor coordination.

The paper explicitly states what the data do not establish:

• Direct viral invasion or a specific coronavirus entry route (e.g., via the anterior commissure).
• Axonal destruction, mechanical compression, or tissue destruction.
• A thermoregulatory shivering mechanism for the reported symptoms.
• A single causal pathway or a diagnostic biomarker.

The authors conclude that the data suggest a convergence of biological processes (potentially inflammatory, vascular, metabolic, or neurodegenerative) affecting a specific anatomical cluster. They emphasize that prospective replication with standardized acquisition, preregistered regions of interest, correction for multiple comparisons, objective physiological measures, and longitudinal follow-up is required before the model can be validated as a biomarker or treatment target. The study advocates for mechanism-aligned phenotyping rather than attributing these symptoms to psychological causes.