Markers of limbic system damage following SARS-CoV-2 infection

Abstract

Alterations of the limbic system may be present in the chronic phase of SARS-CoV-2 infection. Our aim was to study the long-term impact of this disease on limbic system-related behaviour and its associated brain functional connectivity, according to the severity of respiratory symptoms in the acute phase. To this end, we investigated the multimodal emotion recognition abilities of 105 patients from the Geneva COVID-COG Cohort 223 days on average after SARS-CoV-2 infection (diagnosed between March 2020 and May 2021), dividing them into three groups (severe, moderate or mild) according to respiratory symptom severity in the acute phase. We used multiple regressions and partial least squares correlation analyses to investigate the relationships between emotion recognition, olfaction, cognition, neuropsychiatric symptoms and functional brain networks. Six to 9 months following SARS-CoV-2 infection, moderate patients exhibited poorer recognition abilities than mild patients for expressions of fear (P = 0.03 corrected), as did severe patients for disgust (P = 0.04 corrected) and irritation (P < 0.01 corrected). In the whole cohort, these performances were associated with decreased episodic memory and anosmia, but not with depressive symptoms, anxiety or post-traumatic stress disorder. Neuroimaging revealed a positive contribution of functional connectivity, notably between the cerebellum and the default mode, somatosensory motor and salience/ventral attention networks. These results highlight the long-term consequences of SARS-Cov-2 infection on the limbic system at both the behavioural and neuroimaging levels.

Keywords: MRI; emotion; functional connectivity; neuropsychological deficits; post-COVID syndrome.

Graphical abstract

Figure 1

Response format of the short form of the Geneva Emotion Recognition Test. After watching each video clip, the participant must choose the answer from the choices provided in this wheel.

Figure 2

Multimodal emotion recognition (as measured by the Geneva Emotion Recognition Test—short version) as a function of the severity of respiratory symptoms in the acute phase. Bars represent raw mean scores for each emotion, and whiskers represent standard deviations. The results showed patterns of poorer multimodal recognition abilities in patients with a moderate or severe form in the acute phase as compared with patients who had a mild form in the acute phase. *Significant Mann–Whitney U-test differences between groups corrected for multiple comparisons (Bonferroni methods).

Figure 3

Relationships between multimodal emotion recognition abilities (as measured by the total score of the Geneva Emotion Recognition Test—short version) and verbal episodic memory as measured by RL/RI 16-delayed free recall (top) and with olfaction as measured by the Sniffin’ Sticks test battery (bottom). Each dot represents a patient; lines represent the best least square linear fits. (A) The poorer the ability to recognize emotions, the poorer the performance on verbal episodic memory task. (B) The poorer the ability to recognize emotions, the poorer the performance on the olfactory recognition test.

Figure 4

Multivariate latent component from the group PLSC analysis. (A) Loadings of behavioural data. PLSC loadings were defined as correlation coefficients between a given feature and its weight in the latent component. Dots represent samples from the bootstrap procedure, and yellow highlights indicate the reliability of the scores’ contributions to the multivariate correlation components. (B) Bootstrap sampling ratios for functional connectivity (5% highest positive and negative values evaluated with bootstrap). This network representation illustrates the neuroimaging pattern, with red links indicating a positive influence of functional connectivity on the latent component and darker colours indicating a higher number of significant connections for each resting-state networks’ pair involved in the pattern. (C) Twenty brain networks and regions whose connections had the greatest impact [bootstrap sampling ratio (BSR)] on the latent component. Regions: Aud, auditory cortex; IPL, inferior parietal lobule; PCC, post cingulate cortex; pCun, precuneus; PFC, prefrontal cortex; RSP, retrosplenial cortex; TempOcc, temporal occipital area.