SARS-CoV-2 infection is associated with neurological symptoms and complications that range from headache, anosmia and dysgeusia, to severe complications such as cerebrovascular events, encephalopathy, Guillain-Barré syndrome, and dementia-like syndrome (1). In addition, many COVID-19 patients develop a ‘post-COVID-19 syndrome’ defined as the persistence of a wide spectrum of symptoms beyond four weeks after infection (2). In symptomatic COVID-19 patients, a community-based study with over half a million people in the UK estimated that about one in three experienced at least one persistent symptom for 12 weeks or more (3). In a population-based study in Lombardy, the post-COVID-19 condition was associated with death, rehospitalization and use of health resources (4). Long-term neuropsychological impairments such as executive, attentional and memory deficits, are reported even after mild infection (5). While the exact causes of post-COVID-19 syndrome remain largely elusive, the prevalence of associated neurological symptoms with an increased risk of anxiety and depression at 16-month follow-up (6) suggests a brain origin (7, 8).

There is neurochemical evidence of neuronal injury in patients with COVID-19 (9, 10), with reports of a severity-dependent increase of neurofilament light chain (NFL) at 4-month follow-up, further supporting ongoing brain injury even weeks and months after acute infection (11). Not surprisingly, the neurological complications are associated with worse functional outcome, particularly in older subjects and those with comorbidities (12).

Hypotheses of pathogenic processes implicated in acute and delayed brain injury following a SARS-CoV-2 infection include: i) viral invasion, ii) bioenergy failure, iii) autoimmunity, and iv) innate neuroimmune responses (13). In all these processes the blood-brain barrier (BBB), which maintains the specialized microenvironment of the neural tissue by regulating the trafficking of substances between the blood and brain compartments, has a central role.

Brain endothelial cells are the primary unit in close association with pericytes and astrocytes (14). Pericytes, which are key cells in maintaining and supporting vascular homeostasis and barrier function (15), are also the main source of matrix metalloproteinase 9 (MMP-9) (16, 17). Inflammatory stimuli very rapidly activate MMP-9 at the pericyte somata, leading to degradation of the underlying tight junction complexes. Thus, MMP-9 can act as a toxic culprit of BBB disruption after acute (18, 19) and neurodegenerative diseases (20). Peptidyl prolyl cis-trans isomerase A (PPIA), also known as cyclophilin A, acts as an activator of MMP-9 (21, 22) through binding to its CD147 receptor, which in addition has been proposed as an alternative route for SARS-CoV-2 infection (23).

Mechanistically, it has been demonstrated that the severe COVID-19-related cytokine storm is induced by a “spike protein-CD147-PPIA signaling axis” (24). In vivo experiments using a preclinical mouse model indicated that an anti-CD147 antibody inhibited the cytokine storm of SARS-CoV-2 (24).

Astrocytic end-feet containing glial fibrillary acidic protein (GFAP) are an essential component of the BBB. High blood GFAP is a marker of structural damage in the acute phase of brain injury and a severity-dependent increase has been detected in COVID-19 patients (11, 25). These data highlight PPIA, MMP-9, and GFAP as key disease biomarkers, so their measurement in association with NFL, an established marker of brain injury, may help identify a blood signature for disease severity and neurological complications in COVID-19 patients (NeuroCovid).

We identified significant effects associated with SARS-CoV-2 infection in COVID-19 patients, with NeuroCovid subjects showing the highest levels of biomarkers associated with BBB disruption, while patients in the intensive care unit (ICUCovid) had higher levels of inflammatory response biomarkers.

This study examined the effects of SARS-CoV-2 infection on blood biomarkers of BBB disruption, neuronal damage and systemic inflammation by longitudinally monitoring two patient cohorts of COVID-19, with increasing disease severity and neurological complications. Blood biomarkers of BBB disruption were elevated in COVID-19 patients with levels comparable to or even higher than in ALS patients, pointing to neurological implications over a range of disease severities.

There was evidence of different temporal dynamics in ICUCovid compared to NeuroCovid patients with PPIA, the potent activator of the cytokine storm and MMP-9 inducer, and IL-10, the master regulator of immunity to infection, with the highest levels in ICUCovid patients in the acute phase (Figure 6). In contrast, MMP-9 was significantly higher in the acute phase in NeuroCovid patients, with severity dependency in the longer term. In line with previous findings, we found also clear severity dependency of NFL and GFAP levels with the highest levels in deceased patients, and severe NeuroCovid patients showing a tendency to maintain higher values than moderate patients in the longer term.

PPIA is a foldase and a molecular chaperone with multiple functions and substrates, including viral proteins essential for coronavirus replication (30). PPIA is a major target of redox regulation in activated lymphocytes (31, 32). Under stress conditions PPIA is secreted extracellularly by several types of cells, including pericytes, vascular smooth muscle cells and macrophages, behaving as a pro-inflammatory cytokine, with potent chemotactic activity toward leukocytes (21, 33, 34). Through the interaction with its CD147 receptor, in a NF-κB-dependent pathway, PPIA is an inducer of MMP-9 and of pro-inflammatory cytokines and chemokines (22, 35). High levels of PPIA have been seen in biofluids of several conditions associated with inflammation, including neurological and cardiovascular diseases (22, 36). Interestingly, high plasma concentrations of PPIA have also recently been reported in COVID-19 patients with mechanistic evidence for its involvement in the induction of the cytokine storm by activating CD147 (24).

A growing body of clinical data suggests that the cytokine storm is associated with COVID-19 severity, ICU admission, and is a crucial cause of death (37). In agreement with this, our ICUCovid patients had PPIA concentrations substantially higher than NeuroCovid patients. Also noteworthy is the extremely high PPIA plasma concentration in all COVID-19 patients. This may be linked to its up-regulation upon interaction of SARS-CoV-2 with CD147, as observed in animal models (24), and may favor viral replication (30, 38). Similarly, MMP-9 was very high in all COVID-19 patients. However, NeuroCovid patients had the highest levels of MMP-9 in the acute phase, with persistent high levels in most severe patients in the long-term. MMP-9 is a metalloproteinase with a wide substrate spectrum and is an important mechanism for fine-tuning cellular processes, but if aberrantly activated it is a key factor in BBB disruption and neuronal damage, by degrading tight junction proteins and laminin (18–20). MMP-9 can be induced by inflammatory signaling cascades with CD147 acting as the major upstream inducer in the CNS (39). CD147 is highly expressed in the brain capillary endothelium and various sub-regions of the brain (40). Brain pericytes are the main source of MMP-9 at the neurovascular unit and it is rapidly released in response to inflammatory stimuli (16, 17). It has also been demonstrated in vitro and in vivo that SARS-CoV-2 can infect the brain microvascular endothelial cells and cross the BBB by MMP-9-mediated disruption of basement membrane (41). Therefore, one can hypothesize that a local, early high MMP-9 concentration at the neurovascular unit in NeuroCovid patients, rather than extensive systemic inflammation as in ICUCovid patients, may be responsible for the BBB disruption that triggers neurological complications following SARS-CoV-2 infection.

Astrocytic end feet cover more than 99% of the neurovascular surface and directly affect BBB permeability (42). GFAP is a highly expressed protein of the CNS, almost exclusively in astrocytes. In neuropathological conditions, GFAP is released into the bloodstream either by direct venous drainage or through a compromised BBB (43). Blood GFAP can therefore serve as a useful biomarker and prognostic tool for numerous neurological conditions (25).

While classically considered a marker of astrogliosis, the presence of glial-derived proteins in peripheral body fluids has been suggested as indicating BBB disruption in acute CNS injury (44). In the case of traumatic brain injury, it has been recently suggested that high blood GFAP concentrations might reflect damage to astrocytic end feet enveloping the BBB, thus releasing GFAP directly into the blood when the BBB is injured (45). Elevated GFAP plasma levels have been reported in COVID-19 patients (10, 45) and were in line with neuropathological data indicating post-mortem evidence of BBB disruption and gliosis (46). In accordance with this, here we report high GFAP levels in a severity-dependent manner, with significantly higher levels at acute timepoints in deceased patients. GFAP only tended to be higher in NeuroCovid patients than in ICUCovid patients. However, the NeuroCovid cohort included several patients with Guillain-Barré syndrome in which blood-nerve-barrier (BNB) disruption is a key step (47). BNB lacks astrocytes and glia limitans, so the detection of barrier damage through GFAP in these cases is underestimated. Maladaptive microglia and monocyte activation may also exert a detrimental effect on BBB function and integrity in COVID-19 (14). Interestingly, a recent publication has shown that microglia−derived chemokine MCP-1 (also known as CCL2) seems to have a major role in neocortex neuroinflammation and BBB disruption in a mouse model of autoimmune encephalomyelitis (48). Moreover, high blood MCP-1 levels have been associated with disease severity and mortality in COVID-19 (49). Although not measured in our study, longitudinal analyses of MCP-1 in plasma and CSF of COVID-19 patients and correlation with neurological symptoms will shed light on this aspect in future studies.

NFL is an established marker of axonal injury (50). Although axonal degeneration is not a specific feature of ALS, NFL is considered its most characteristic biomarker since its concentration is higher than in any other neurological disease (51, 52). This may be because neurofilaments are abundantly expressed in the large myelinated axons involved in the degenerative process, which is particularly fast and severe in ALS compared to other diseases. The only other condition in which the NFL plasma concentration is as high as in ALS is HIV-associated dementia (HAD) (53). Interestingly, it seems that HIV-related CNS degeneration starts during primary infection and continues during subsequent stages of the disease. However, CSF NFL levels in primary infection are associated with CNS immune activation and BBB disruption but are not accompanied by high CSF total tau and low amyloid beta peptides, as in subjects with HAD (54). This indicates that this early neuronal injury is less severe and/or involves a different mechanism and can in fact be halted by antiretroviral therapy (55). Plasma NFL levels were high in all COVID-19 patients, with NeuroCovid patients reaching the same high levels as in HAD (53). Although the overall picture points to an increased risk for neurological dysfunctions in the long-term, the mechanism and extent to which acute axonal damage, in combination with systemic inflammation and BBB disruption, can predispose to neurodegeneration calls for further investigation.

Our observations may provide hints for a preventive approach. Should further evidence confirm that the neuronal damage found is secondary to, or exacerbated by, BBB disruption, therapies reducing BBB damage could serve as a valuable aid in attenuating the neurological damage in the acute phase and potential dysfunction in the longer term. Interestingly, MMP-9 stands as a druggable target since a set of potent MMPs inhibitors are already available for clinical use (56), furthermore drugs targeting the PPIA-CD147-MMP-9 signaling pathway are also under investigation. A PPIA inhibitor, cyclosporine A (CsA), a well-known immunosuppressive drug, and Meplazumab, an anti-CD147 monoclonal antibody, are being assessed in clinical trials up to phase 2/3 (NCT05113784) for the treatment of severe COVID-19 (57). There are some indications from observational studies of milder COVID-19 and lower mortality in solid organ transplant recipients and autoimmune disease patients under CsA treatment (58). Last, there is evidence that Annexin A1, an endogenous molecule endowed with resolving/protecting action on tight junctions, may have therapeutic potential in restoring cerebrovascular damage and BBB disruption in neurodegenerative diseases and metabolic disorders (59). Indeed, human recombinant annexin A1 has been recently shown to restore BBB integrity and reduce the expression and activity of MMP-9 in brain microvessels when administered in an experimental model of metabolic diabetic disorder (60). Thus, also Annexin A1 could be a therapeutic avenue for COVID-19 to explore in future studies.

There are limitations in this study that should be highlighted. First, neurocognitive assessment in these cohorts of patients was not performed, thus the question as to whether BBB biomarker changes may predict late cognitive dysfunction is still open and should be addressed in future studies. 13% NeuroCOVID patients had a known history of mild cognitive impairment possibly contributing to the altered neurologic state observed in the acute phase. However, in these patients acute NFL levels were comparable to or even lower than the other patients in the NeuroCovid group, making it unlikely that an already altered CNS homeostasis was the cause of the biomarker changes. Notably, in the ICU cohort there were no patients with pre-existing neurological conditions thus reinforcing the finding that COVID-19 per se may induce markers of BBB disruption and neurological damage. Last, patients in our study were recruited before the vaccination campaign. Although there is increasing evidence that COVID-19 vaccination may have a protective effect against the post-COVID-19 syndrome (61, 62), this aspect has not been fully explored, calling for follow-up studies to monitor distinct long-term consequences in vaccinated and non-vaccinated subjects.

Despite these caveats, our study may provide hints for upcoming therapeutic approaches for COVID-19 mitigating at the same time BBB disruption and neurodegeneration to reduce the neurological damage in the acute phase and potential dysfunction in the long-term.

The original contributions presented in the study are included in the article. Further inquiries can be directed to the corresponding authors.

The studies involving human participants were reviewed and approved by the ethics committees of the clinical centers involved: Fondazione IRCCS Ca’ Granda Ospedale Maggiore Policlinico, Milano (approval #868_2020, 28.10.2020), ASST Papa Giovanni XXIII, Bergamo (approval #123/20, 14.05.2020). The patients/participants provided their written informed consent to participate in this study.

VB, NS, MS, EZ: Drafting/revision of the manuscript, major role in the acquisition of data, study concept and design, analysis and interpretation of data. LP, IL: Drafting/revision of the manuscript, major role in the acquisition of data, analysis and interpretation of data. MC, NB: Drafting/revision of the manuscript, study concept and design. RZ, EF, VP, SL, GC, MG, NR, AC, CM: Drafting/revision of the manuscript, major role in the acquisition of data. BB, GF, RG, FBl, CC, NM, MT, PB, FBi, AB, AN: Drafting/revision of the manuscript. All authors contributed to the article and approved the submitted version.