A study led by researchers from the National Institute of Neurological Disorders and Stroke – U.S. National Institutes of Health and the U.S. Army’s Uniformed Services University of the Health Sciences and The Joint Pathology Center at the Defense Health Agency in Maryland have found that SARS-CoV-2 infections causes serious damage to the microvessels of the brain.
The study findings were published in the peer reviewed journal: Brain.
https://academic.oup.com/brain/article/145/7/2555/6621999
The cause of a hypercoagulable state is unclear but a generalized hyperinflammatory state11 and anti-phospholipid antibodies have been implicated.34 Rare cases of acute necrotizing haemorrhagic encephalopathy have also been described.5 Microcerebral haemorrhages may occur in critically ill patients.33
Using post-mortem high-resolution MRI, we previously found that most patients had widespread multifocal microvascular disease which corelated with vascular leakage and injury.16 In the present study, we characterized the pattern, mechanism and consequences of microvascular injury in patients with COVID-19. The loss of vascular integrity was evident by the presence of several large proteins in the perivascular regions that normally do not cross the blood–brain barrier.
These included fibrinogen, C1q, IgG and IgM. Fibrinogen was present in high concentrations around the blood vessels with a gradual decrease in concentrations at greater distances from the vasculature. This suggests a leaky blood–brain barrier. All markers of vascular injury were more common in the hindbrain. Similar deposition of fibrinogen in the lumen of the blood vessels has been described in active inflammatory lesions of multiple sclerosis. But even in these patients there was no perivascular leakage of fibrinogen as described in our patients.35
We found increased levels of PECAM-1 on endothelial cells. In contrast, an in vitro study showed a decrease in PECAM-1 on endothelial cells following treatment with recombinant spike protein of SARS-CoV-2.38 Thus, the mechanism of increased PECAM-1 levels in the endothelial cells remains unclear. However, this molecule can serve as an adhesion molecule for platelets and platelet aggregates that were adhered to the endothelial cells was a prominent observation in this study.
The platelets were activated and, in some instances, caused occlusion of the small blood vessels. These observations were supported by spatial transcriptomics data. The primary genes driving separation between CODIV-19 patients and controls, CD74 and TF, both contribute to thrombosis formation. In addition to its role in immune function, CD74 is known to be involved in thrombosis formation by contributing to the platelet cytoplasmic Ca2+ signalling pathway.39TF contributes to thrombosis formation by encoding coagulation factor III (TF, tissue factor) to initiate the coagulation cascade and has been reported to be increased in COVID-19.40
The semaphore signalling and the RhoGDI signalling pathways associated with vascular permeability were enriched in COVID-19 patients compared to controls. This may explain the microinfarcts seen in some COVID-19 patients on MRI.31 One study found that the lungs of patients with COVID-19 have unique vascular features consisting of endothelial damage, microthrombosis and intussusception angiogenesis.36
Several makers of endothelial cell function have been studied in COVID-19 patients. Meta-analysis showed that higher plasma levels of vWF antigen, tissue-type plasminogen activator, plasminogen activator inhibitor-1 antigen and soluble thrombomodulin were associated with poor outcome.41 In the liver, strong vWF staining in sinusoidal endothelial cells was associated with increased platelet adhesion.42 Consistent with these observations, we found increased immunostaining for vWF in the microthrombi. Two patients had a history of drug abuse; however, there was nothing unusual about their vascular pathology compared to the other patients.
To determine whether the compromise of the endothelial cells maybe an immune-mediated phenomenon, we looked for the deposition of immunoglobulins. Aggregates of IgG and IgM were found on endothelial cells and platelet aggregates that co-localized with several members of the complement cascade.
The presence of C1q, C4d and C5b-9 suggests activation of the classical complement pathway. We also found deposition of C1q and C3 in macrophages and endothelial cells, which has been shown to be induced by SARS-CoV-2 spike protein.43 Deposition of complement cascade and immunoglobulins suggests an immune-mediated injury to the endothelial cells. The antigen against which this immune response is targeted remains unknown.
Possibly, the antibodies are directed against an antigen on the endothelial cells, e.g. anti-idiotypic antibodies against the spike protein would bind to the ACE-2 receptor on endothelial cells.44 Alternatively, immune complexes formed by the antibodies and spike protein that may bind to the ACE-2 receptor on the endothelial cells. The spike protein has been shown to compromise the blood–brain barrier in vitro.45
Elevated levels of factors involved in the classical complement pathway have been found in the plasma and autoantibodies that cross-react with brain antigens and the spike protein have been described in the CSF of patients with neuro-COVID-19.46 Critically ill patients with neurological manifestations have also been found to have autoantibodies in plasma and CSF against a number of CNS antigens including endothelium of blood vessels.47 Thus, while the damage to endothelial cells may not be unique to the CNS, the consequences of the breakdown of the blood–brain barrier are unique to the CNS.
We found the cellular infiltrates of macrophages, CD4+ T cells and CD8+ T cells in COVID-19 patients using immunohistochemistry, which was consistent with other studies.11,16 Mirroring these observations, the results of our spatial transcriptomics data demonstrated that the signalling pathways involved in the migration or trafficking of these cells were enhanced in regions rich in PECAM-1+ cells or CD45+ cells. These included RhoGDI, PTEN and Gαi signalling pathways. The cellular infiltrates were predominantly in the perivascular region and largely composed of macrophages.
There were 10-fold more macrophages compared to T cells. Although there were few T cells, CD8+ cells out-numbered CD4+ cells and there were only rare B cells. This suggests that the inflammatory infiltrate was secondary to the leakage of serum proteins into the perivascular region as macrophages act as scavengers and help with the repair process. This is consistent with other studies that have found activated monocytes and macrophage markers in CSF48,49 and the brain.11
In one study, the CD8+ T cells were further characterized to show that they had both cytotoxic and exhaustion markers.50 We observed that the serum proteins such as fibrinogen and complement were taken up by glia and neurons. Similarly, astrocytosis was also most prominent in the perivascular regions suggesting that this was secondary to the vascular injury.
We also found multifocal loss of neurons in the hind brain including the cerebellum. This pattern of neuronal loss cannot be attributed to hypoxia where a more diffuse pattern of injury would be expected. The damaged neurons were often in close vicinity of the activated macrophages or microglia and there was evidence of neuronophagia suggesting that the neuronal injury was secondary to glial cell activation. Neuronophagia in the brainstem of patients with COVID-19 has been described previously.13
The underlying pathophysiological mechanisms of COVID-19-induced neuronal injury are unclear but may be explained by metabolic dysregulation and oxidative stress and DNA double-strand damage. Spatial transcriptomics of the brainstem, where neuronal injury was evident, indicated dysregulation of genes such as APOD, GSTT1, ATP5MC2 and MT1x and signalling pathways including PTEN signalling and PPAR signalling.
Taken together, these findings are consistent with metabolic dysregulation. For example, APOD has been shown to be upregulated in the brain following HCoV-OC43 infection and has a neuroprotective effect by controlling the levels of peroxidated lipids.51 We also found genes and signalling pathways associated with oxidative stress and DNA damage were upregulated in COVID-19. These included the genes ATP5MC2 and MT1x, Sirtuin, and the HIPPO and ATM signalling pathways. While the dysregulation of these genes and pathways may explain some of the pathological observations in this study, they may also represent potential therapeutic targets.
Interestingly, several of the pathological findings were more prominent in the hindbrain, which is consistent with other studies although the cause remains unclear.4 It has been hypothesized that the virus may reach the brainstem via the olfactory pathways or the vagus nerve that innervates the respiratory and gastrointestinal tracts.52 However, we were unable to confirm the infection in the brainstem. Involvement of the brainstem could have dire consequences since many vital functions are controlled by this region. It may also explain many of the acute and persistent manifestations seen in patients with COVID-19.53,54 Importantly, five patients in our study died suddenly, most while sleeping, hence the possibility of central apnoea needs to be considered although cardiac arrythmia or dysautonomia could be contributory.
We and others have failed to detect the SARS-CoV-2 virus using a variety of highly sensitive techniques.16,55 However, a previous study showed that human brain endothelial cells can be infected with the virus in vitro.37 Another study found small amounts of detectable SARS-CoV-2 RNA in a few individuals in the olfactory bulb, medulla oblongata and the cerebellum and viral protein was found in areas of acute infarcts.52 A subsequent study found double stranded RNA in pericytes of brain blood vessels of patients with COVID-19 and implicated it in the breakdown of the blood–brain barrier.56 Very low levels of SARS-CoV-2 were also detected in other studies by polymerase chain reaction only, and not by RNA in situ hybridization or by immunohistochemistry. They concluded that the neuropathological observations did not result from direct viral infection of brain parenchyma but were more probably due to systemic inflammation.13 Our spatial transcriptomics data supports this in that most of the pathways that we identified as differentially regulated are related to the immune response.
On the basis of these observations, we propose the following cascade of events. Infection with SARS-CoV-2 triggers the formation of immune complexes activating the classical complement pathway. The mechanism by which the viral infection causes the formation of immune complexes is not clear as we were unable to find the virus or viral proteins in the tissues. One possibility is that anti-idiotypic antibodies against the spike protein might bind to the ACE-2 receptor on endothelial cells triggering the cascade of events.44
These immune complexes may cause multifocal areas of endothelial cell activation resulting in platelet activation, aggregation and formation of thrombi. Injury to the endothelial cells may result in leakage of serum proteins into perivascular regions. This may set up several reparative cellular processes, which include infiltration of monocytes that differentiate to macrophages with phagocytic activity to clean up the proteins. The serum proteins are also taken up by glial cells and neurons.
The inflammatory process results in microglia activation causing neuronal injury and neuronophagia (Fig. 5B). Since several of the patients in our series died suddenly with very minor lung involvement, we believe that had these patients survived they would probably have progressed to develop long-COVID. Hence the pathological findings here are relevant to this population as well.
Our study has several limitations. Since several patients were found dead, medical histories and post-mortem intervals were not available. Even though we conducted extensive studies for detection of SARS-CoV-2, it is possible that our inability to detect the virus could be a technical or sampling artefact. Although the age groups of our patients were not perfectly matched, the controls were older and hence would have been expected to have more vascular pathology. Four of the controls had respiratory or systemic infections but did not demonstrate the pathological findings seen in the COVID-19 patients. We did not include any controls with microvascular disease hence we cannot be certain if the microvascular changes seen in the COVID-19 patients are specific for the infection.
Conclusions
Injury to the microvasculature by immune complexes with complement activation is the key central event that results in breakdown of the blood–brain barrier, microthromboses, perivascular inflammation and neuronal injury. We postulate that these events are central to the development of the neurological manifestations seen in acute COVID-19 and possibly in long-COVID. Importantly, these studies suggest that therapeutic approaches targeted against the development of immune complexes should be considered.
