Metabolic changes in the brain's communication pathway of people after COVID measured by MR spectroscopy

Coronavirus disease 2019 (COVID-19) is an infectious disease caused by the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) [1]. Patients with COVID-19 may present a variety of symptoms from asymptomatic infections, mild symptoms such as fever, cough, and shortness of breath, to severe life-threatening pneumonia and multiorgan involvement [2, 3]. SARS-CoV-2 also affects the central nervous system (CNS), resulting in neurological symptoms such as loss of smell and taste [4]. A substantial proportion of affected patients do not fully recover from COVID-19, going on to develop long COVID (a postviral syndrome) [5]. The patients experience neurological and psychological symptoms associated with the disease, including headaches, brain inflammation, fatigue, muscle pain, depression, and anxiety [6, 7].

Systematic reviews and MR brain studies report parenchymal brain abnormalities, subcortical micro- and macro-bleeds, cortico-subcortical swelling, and non-specific deep white matter changes visible on MR images [2, 8]. Signal and diffusion abnormalities are extended predominantly to the corpus callosum, cingulate cortex, and insula, jointly implicating the olfactory brain network [2, 9]. Sawlani et al. also reported white matter abnormalities especially involving the splenium of the corpus callosum and brainstem in COVID-19 patients with a range of medical and neurological comorbidities [10].

Corona viruses can invade the CNS by two mechanisms: viral replication into glial or neuronal cells of the brain or autoimmune reaction [11, 12]. Autopsy studies of patients with COVID-19 show prominent neuroinflammation and activated microglia [13, 14]. Glial and neuronal metabolic abnormalities in post-COVID-19 patients may be non-invasively assessed by proton magnetic resonance spectroscopy (1H MRS). N-acetylaspartate and N-acetylaspartylglutamate (NAA) are considered to be markers of neuronal viability and density [15, 16]. Neuroinflammation is typically represented by elevated glial marker myo-inositol (mI), concomitantly elevated choline-containing compounds (Cho), and total creatine [17], due to their higher concentrations in glia than in neurons [14]. Creatine together with phosphocreatine, referred to as total creatine (Cr), is considered as a marker of cerebral bioenergetics. Choline-containing compounds comprising mainly phosphocholine, glycerophosphocholine, and free choline reflect cell membrane constituents and membrane turnover [16]. Myo-inositol is considered as a glial cell marker, it is connected to the osmoregulation of astrocytes, and it is important for the integrity of the cells. Increased mI probably reflects a glial activation and proliferation; however, its exact function remains unclear [18].

Post-COVID-19 patients with persistent neuropsychiatric symptoms revealed in the anterior cingulate cortex lower NAA indicating neuronal injury and lower mI reflecting glial dysfunction, possibly related to mitochondrial dysfunction and oxidative stress [14]. Their subset of non-hospitalized group showed decreased mI compared to hospitalized post-COVID-19 patients. The non-hospitalized group further showed significantly decreased mI, Cho, Cr, and Glx in frontal white matter (WM) compared to healthy controls. Reda et al. [12] reported decreased NAA and NAA/Cr and elevated Cho, Glx, and lactate (Lac) in non-hemorrhagic lesions, increased Lac/Cr in necrotizing leukoencephalopathy, and no changes in apparently normal WM in patients with acute COVID-19 infection with neurological symptoms.

COVID-19-associated necrotizing leukoencephalopathy showed NAA/Cr reduction and elevation of Cho/Cr, Glx/Cr, Lac/Cr signal ratios, and Lac [12, 19]. Similar metabolic patterns, but with less pronounced changes and without an increased Lac/Cr ratio, were also seen in patients with COVID-19 postcardiac arrest and in non-COVID-19 patients with delayed posthypoxic leukoencephalopathy [19]. Sklinda et al. did not found any changes in values of NAA, Cho, and Cr in patients hospitalized due to symptoms of severe brain fog (i.e., insomnia, sudden impairment of cognitive function, headache, and depression) [20]. Lin et al. [21] reported increased Cho in the cingulate gyrus, which may be reflective of neuroinflammatory changes due to long COVID.

Although many MRI studies report signal abnormalities on MR images in the corpus callosum [2, 9, 10], to our knowledge, no MRS study evaluating metabolic changes in the splenium of the corpus callosum (CCS) has been performed yet. The corpus callosum forms a bridge between the cerebral hemispheres, containing crossing axonal fibers from both hemispheres [22]. Its primary role lies in facilitating interhemispheric collaboration [23]. The CCS is the most posterior part of the corpus callosum [22] and its lesions may result in the disconnection of the cerebral hemispheres, with disruption of higher cortical function, loss of conscious processes, and delirious behavior [24]. As many post-COVID-19 patients experience neuropsychological symptoms [5 –7, 14], knowledge about possible changes in the metabolic profile in the CCS may contribute to a better understanding of the pathophysiology of long COVID.

The majority of studies dealing with patients after suffering from COVID-19 with neuropsychiatric symptoms commonly report microstructural changes in the brain, predominantly in the corpus callosum (anterior part as well as splenium), manifesting as signal and diffusion abnormalities [2, 7, 9, 10]. Although the changes in the microstructure may also imply metabolic abnormalities, to our knowledge, to-date, metabolism using MRS has been studied only in the anterior cingulate cortex, frontal WM, and COVID-19-related lesions visible on MRI [12, 14, 19 –21, 30, 31]. For this reason, our study focused specifically on the metabolism in the splenium of the corpus callosum.

Our study revealed an age-dependent effect of SARS-CoV-2 on Cho concentrations in the splenium of the corpus callosum (see Fig. 4). While all metabolic concentrations showed significant linear dependence on age, as indicated by Pearson's r, only Cho in CCS exhibited a significantly higher slope in the post-COVID-19 patient group compared to the healthy control group. This suggests that SARS-CoV-2 disproportionately increases Cho concentration in the CCS among older post-COVID-19 patients compared to younger ones. Our analysis demonstrated that Cho concentrations, adjusted for age in the CCS, differed significantly between all PAT and HC groups. However, considering a subjective threshold of 40 years of age, Cho was significantly increased only in the older PAT group compared to the older HC group. Our results are consistent with the findings of other studies showing a higher vulnerability of older populations to COVID-19 than younger adults [32]. Due to the strong effect of SARS-CoV-2 on Cho-age dependence and no effect on NAA-age dependence, NAA/Cho was also significantly more decreased in older patients than in older healthy subjects. NAA-age dependence in the PAT and HC groups were similar (see Fig. 4). The concentrations of NAA, mI, and Cr were stable. Interestingly, the PAT group did not reveal lower NAA indicating neuronal injury nor lower mI reflecting glial dysfunction in the CCS as post-COVID-19 patients with persistent neuropsychiatric symptoms in the anterior cingulate cortex [14]. It is possible to speculate that increased choline-containing compounds may be attributed to the microstructural changes in the CCS. Unfortunately, MRS relaxometry was not included in the examination protocol. Several studies previously reported lesions in the CCS with restricted diffusion and low values of apparent diffusion coefficient in hospitalized patients [9, 10]. Ibrahim et al. found increased values of mean diffusivity (MD) and decreased fractional anisotropy in post-COVID-19 patients in the forceps major without apparent signal abnormalities on standard MR images [7]. Diffusion abnormalities may show membrane reorganization of the axonal fibers between both hemispheres. As the MR visible Cho signal only reflects water-soluble choline metabolites such as phosphocholine and glycerophosphocholine, relatively immobile membrane components as sphingomyelin and phosphatidylcholine do not contribute to the total Cho signal intensity in the healthy tissue [16]. However, in the case of microstructural changes in the tissue, it is possible to speculate that changes in the arrangement of cell membranes may occur and tightly bound molecules may be released, become visible using MR spectroscopy, and cause an increase in the choline signal. As other metabolites are stable in the CCS, this explanation is more probable than neuroinflammation, glial dysfunction, or increased membrane turnover reported in other brain regions [12, 14, 19, 21].

Interestingly, 22 subjects with positive IgG antibodies to the nucleocapsid of SARS-CoV-2 believed they had never contracted COVID-19 (according to an anamnestic questionnaire) indicating they had passed COVID-19 infection completely asymptomatically without any post-COVID-19 neuropsychological symptoms. Simultaneously, the low levels of IgM and IgA classes indicated a non-acute phase of infection. Although these patients were not examined by a neurologist, and information about their health status during the COVID-19 pandemic is only subjective, we divided the PAT group into symptomatic and asymptomatic. At the same time, we would like to add that we believe that dividing subjects into PAT and HC groups based on IgG antibody levels or a positive qPCR in the past is more objective than the subjective division into symptPAT, asymptPAT, and HC. AsymptPAT revealed similar metabolic patterns to symptPAT (Fig. 4, ). Surprisingly, a subgroup of asymptomatic patients also revealed an age-dependent increase of choline-containing compounds in the CCS. Increased Cho was found in both symptomatic and asymptomatic older patients compared to healthy older controls. These results imply that the increased Cho in the CCS should not be related to neuropsychological symptoms. Although lesions in the CCS may result in neuropsychiatric problems [24], none of the subjects in our study were found to have lesions in the CCS.

In this study, we also evaluated the NAA/Cho ratio. This ratio is often used in metabolic studies and routine clinical practice as a parameter indicating metabolic changes in damaged brain tissue. Its calculation does not require special software or corrections as in the case of the calculation of individual metabolite concentrations. Although this is not a specific marker or characteristic only for COVID-19, a significantly decreased NAA/Cho ratio in older patients (driven by increased Cho in older patients) shows that it is the appropriate parameter for evaluating metabolic changes in the CCS in patients after COVID-19 infection even using a routine clinical protocol containing 1H MRS.

The present study has several limitations. Mainly, low SNR related to small VOI (1.4 ml). It would be desirable to have a minimum SNR = 10 which is important for accurate and precise metabolite quantification. However, only slight variations in absolute metabolite concentrations with SNR in the case of singlet signals in clinical conditions have been shown [33 –35]. Increasing the number of acquisitions by a factor of 2 would increase the measurement time by a factor of 2 (more than 16 min only for one spectrum resulting in the risk of patient movement), but would result only in a 1.4-fold change in SNR (from mean SNR = 7 to 10). Narrow peaks (good B₀ shimming), the efficiency of water suppression, and the elimination of unwanted coherences using a smaller voxel volume are more important than the SNR [33, 34, 36]. As the CCS is a small structure, small VOI dimensions were used to prevent contamination of metabolic signals from adjacent brain structures and CSF and to prevent loss of specificity. Low SNR is problematic in the case of multiplets such as Glx with more sophisticated line shapes. Therefore, these metabolites were not evaluated, although some MRS studies reported changes in Glx values in the frontal gray and white matter in long COVID patients [15, 21, 22, 24]. This study focused on an accurate quantification for precise detection of even small changes in NAA, Cr, Cho, and mI. Secondly, the data were acquired only from the CCS using a 3 T MR tomograph. In future studies, it is advisable to perform the measurement at 7 T to improve the sensitivity of MRS in small regions of interest and to also study other brain regions such as frontal white and gray matter. Thirdly, a methodological weakness is a variable of the subject's symptoms, their severity, and duration, and therefore, many more subjects are needed. However, in our case, this seems to be unimportant as the changes in Cho concentrations and NAA/Cho were found even in asymptomatic post-COVID-19 subjects. Finally, it is also possible to see a limitation in missing information regarding relaxation times. In this study, it was not necessary to calculate T1 and T2 relaxation times due to the application of a short TE together with a long TR in the MRS sequence. However, relaxometry may help create an idea of the extracellular space geometry. Despite these limitations, MR spectroscopy remains the most suitable non-invasive method for evaluating in vivo metabolic changes in the brain tissue.

An age-dependent effect of SARS-CoV-2 on concentrations of choline-containing compounds in the splenium of the corpus callosum has been found. Significantly increased Cho concentrations and decreased NAA/Cho were found in the CCS in the older post-COVID-19 subjects. Metabolism in the CCS of younger post-COVID-19 patients seems to be unaffected by SARS-CoV-2. The patients revealed similar metabolic patterns regardless of the severity of the symptoms. These changes should therefore not be related to neuropsychological problems of the post-COVID-19 patients. The observed changes in choline-containing compounds may be related to the microstructural reorganization in the corpus callosum (also reported in diffusion measurements) rather than increased membrane turnover. The CCS revealed a normal neuronal function, normal bioenergetic and glial metabolism in post-COVID-19 patients, as concentrations of total N-acetylaspartate, total creatine, and myo-inositol remained stable. Nevertheless, further metabolic studies are recommended to confirm these observations.