Cognitive impairment (CI) is one of the common symptoms of post-acute SARS CoV-2 infection (PASC), otherwise known as long covid (1). The pathogenesis, however, remains unclear with possibilities ranging from immune dysregulation, autoimmunity, microthrombi, and blood–brain barrier (BBB) impairment (2, 3). To date, there is only one study which explored the role of BBB disruption using MR imaging, but the study did not have a longitudinal design (4).

Brain imaging has provided a limited contribution to understanding of CI development following COVID-19. Studies using standard MRI brain scan sequences have mostly focused on severely affected patients during acute infection showing ischemia, hemorrhage, venous sinus thromboses (2, 5, 6). There is, however, limited literature for brain imaging using more advanced techniques in PASC. Studies using structural and diffusion MRI in the post-acute phase found alterations in grey matter (GM) volume, including the hippocampus (7–10), and white matter (WM) changes (8, 10). One volumetric MRI study, which followed 401 COVID-19 patients at the onset of infection and 4 months from the acute phase, demonstrated reduced brain volume in the orbitofrontal and parahippocampal gyrus related to the primary olfactory and gustatory systems (7). A recent cross-sectional study explored BBB disruption as the cause of CI in PASC patients (4). This study used dynamic contrast enhanced (DCE) perfusion imaging and demonstrated that BBB impairment correlated with neuroinflammatory markers. Other studies have used diffusion imaging and resting state fMRI to assess brain changes in PASC with CI and showed widespread diffusivity changes along the main axis of WM tracts, and abnormal functional connectivity (FC) among resting state networks (11–16). Taken together, these findings suggest that chronic inflammation can potentially alter the BBB, brain microstructure, and neurochemicals, which may accelerate and perpetuate a chronic neuropathological change in PASC in a self-sustaining cycle, leading to cognitive impairment.

Herein we used a longitudinal design and investigated mechanistic pathways associated with the development of CI in PASC subjects. Firstly, we test using serial imaging a triple-hit hypothesis where the convergence of three deleterious factors: (1) changes in BBB, (2) altered brain neurochemicals associated with neurotoxicity, and (3) disturbed WM microstructure contribute to CI. Finally, we also explored whether these three metrices are associated with each other and the cognitive scores in the PASC participants.

Fourteen subjects with PASC who had persisting symptoms of anosmia, ageusia, fatigue, and CI who were clinically evaluated at the Neurology Department at SVH between July 2021-August-2022 were consented for the use of their clinical data for research. All 14 patients have been referred for CI post COVID and all underwent a full neurological assessment (BJB). MR imaging was conducted as part of the neurological assessment. All participants underwent routine and advanced MR imaging early in the disorder (3 months ± 2 weeks) referred as Time Point 1 (TP-1) and 10 repeated the MRI scan at 12 months (± 2 weeks) here after referred as Time Point 2 (TP-2). Seven of those participants had been enrolled into the ADAPT study, a prospective cohort of 128 SARS-CoV-2 positive patients and had received serial measurements of cognition with the Cogstate Brief Battery (1). In brief, the neurocognitive score (NCS) was derived from CogState: individual task z-scores were averaged into a mean z-score (higher equates to a better objective global performance) corrected for normative age, education and sex effects. Individuals with a prior history of drug use, significant head injury, psychiatric illness, and hepatitis C virus co-infection were excluded. Cognitively, between TP-1 and TP-2 most patients did not significantly improve suggesting that even if not strictly cognitively impaired at baseline, that most may have performed below their optimal level of cognitive functions. As such, two patients significantly improved, 5 remained stable and 3 declined TP-1 and TP-2. Out of 14 subjects only one is left handed and all the subjects including controls are right handed. Ten healthy age and sex matched controls were recruited and underwent the same neurological assessments at one time point (BJB) (Table 1). Local ethics approval was obtained (2022/ETH0022).

We have undertaken a comprehensive proof-of-concept prospective case–control study to investigate parallel changes in blood brain barrier permeability, neurochemicals, and white matter microstructure changes in PASC CI patients. We found significant differences in all three measures in PASC CI patients compared to controls: K-Trans was increased and Glx showed reduction in both the FWM and brainstem. Furthermore, we also found increased FA and decreased MD in FWM. However there is no statistically significant correlation was found between K trans and cognitive performance. Over 12-months, Glx values were normalized in the PASC CI patients, whereas the K-Trans did not significantly change. Interestingly, over 12-months, FA values further increased, and MD decreased in in the superior longitudinal fasciculus in the region of the FWM. Taken together, these findings support a pathogenic mechanism for PASC CI in which excitotoxicity is driving neuroinflammation.

Our central finding is BBB impairment as assessed by the DCE-derived metric K-trans. To the best of our knowledge, there is only one other study that has used K-trans to measure the BBB integrity in PASC (4). It also showed diffuse BBB impairment at four months post COVID-19 infection but change over the subsequent months was not examined (4). Subtle BBB disruption cannot be identified on routine MR studies with gadolinium administration. K-trans has been used in other neurological diseases characterized by BBB impairment (24) and is accepted as a marker of subtle BBB impairment at the capillary level before there is gadolinium enhancement on standard MRI sequences (25).

Both cortical grey and white matter injury have been a consistent feature of PASC (11, 26, 27), even in those with mild acute COVID-19. However, these studies have shown variable outcomes in DWI metrics with some showing increased FA and decreased MD (16) and others the opposite (7, 14). We used skeletonized FA and MD and found that FA increased, and MD decreased at baseline. There was further increase in FA at 12 months especially in the superior longitudinal fasciculus, a white matter bundle that has been shown to be closely linked to poorer working memory (28). Skeletonized MD can be used to detect small vessel disease that have been implicated in the pathogenesis of long COVID (3, 10). Interestingly, although several of these studies have shown increase in MD, we showed decreased MD and increased FA at four months. This paradoxical increase in the FA and MD has been hypothesized to be associated with axonal degeneration (29), microgliosis and inflammation (30), and was also observed in the Lu et al. study (16). This study also showed an increase in MD in a sub-group of patients who had COVID-19 encephalitis. Our study has shown similar observations, with increased FA and decreased MD across several white matter tracts in the PASC CI participants as compared to the controls at four months. The further increase in FA in the superior longitudinal fasciculus is an interesting finding, as we would have expected any white matter microstructural injury to have normalized by 12 months. Notably, this change was not in the white matter area showing significant increase compared to controls at baseline. Thus, whether this longitudinal change reflects further inflammatory changes or improved integrity (31) of the white matter over time remains unclear. This is an important consideration for future research, as the superior longitudinal fasciculus has long bidirectional projections among the prefrontal, temporal, occipital, and parietal cortices and strongly associates with perceptual organization and working memory (28).

Our results can point to pathogenesis. The spectral metabolite changes with reduced Glx and NAA point to excitotoxic neuronal injury which may be secondary to BBB impairment given the relationship to K-trans. Excitotoxicity is implicated in a number of neurocognitive disorders (25, 32) and reflected in MRS metabolites as a reduction in Glx: intracellular Glx levels are increased leading to cell injury and consequent Glx depletion within the cell. To our knowledge, there is only one other study that has examined MRS in PASC CI. This study reported increased Glx compounds. However, it did not specify the precise site of interrogation used to derive the metabolites, was cross-sectional and only expressed each individual metabolite level as a ratio to creatine. Nonetheless, it is known that excess intracellular Glx compounds lead to cell injury and consequent intracellular Glx depletion (33, 34).

There are several limitations in our study. One notable limitation of our study is the small sample size, which may reduce the generalizability of the findings to a broader population. The absence of control data at 12-months is another significant limitation, as it precludes us from making direct comparisons between patient changes and potential normative brain alterations. Additionally, the small number of patients with neuropsychological assessment and the narrow mild range of impairment likely precluded correlation of MR parameters and cognitive impairment. Despite these limitations, our study provides valuable insights into the intra-individual changes over time; however, future research with larger sample sizes and longitudinal control groups is warranted.

In conclusion, our findings have relevance to both pathogenesis and clinical evaluation. MRI techniques addressing BBB disruption, white matter integrity and inflammation may offer further insights into the steps in PASC CI pathogenesis. Further, these techniques may serve as a biomarker for the disorder, assist in the appropriate timing of a therapy, and in assessing therapeutic response.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

The studies involving humans were approved by the Ethics Committee of St Vincent’s Hospital (2022/ETH0022). The studies were conducted in accordance with the local legislation and institutional requirements. The participants provided their written informed consent to participate in this study.

JC: Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Project administration, Software, Writing – original draft, Writing – review & editing. GP: Formal analysis, Investigation, Writing – review & editing. LC: Formal analysis, Project administration, Writing – review & editing. GD: Supervision, Writing – original draft. AK: Investigation, Writing – review & editing. GM: Writing – review & editing, Investigation. DD: Writing – review & editing. AB: Investigation, Validation, Writing – review & editing. DJ: Visualization, Writing – review & editing. XZ: Writing – review & editing, Data curation, Methodology. ML: Data curation, Writing – review & editing. NJ: Formal analysis, Writing – review & editing. BB: Conceptualization, Investigation, Supervision, Writing – original draft, Writing – review & editing.