The 36 autopsy cases of COVID-19 include adults from close to middle age to elderly, there is almost equal gender representation, and presentation is often with hypoxia and frequently with the diagnosis of 2019 novel coronavirus-infected pneumonia. Clinical course varies considerably in length and complexity. Most patients are hypertensive with adventitial sclerosis as evidence of microvascular wall injury, over half have diabetes mellitus, half are obese, many have chronic heart and/or pulmonary disease, and a few have a history of cancer or other comorbidities. These are the findings in most COVID-19 patients in this age range [1, 3, 18]. African Americans account for over half of the cohort cases, but given the small sample size no effect of race is inferred.
Brain microcirculatory system and microvasculopathy
Normal CNS blood vessels 40–400 µm in diameter generally are referred to as microvessels, and when including capillaries the term microcirculation is used [19]. In normal physiological and in pathological conditions, capillaries and microvessels readily dilate and the microcirculation may require further enhancement of its normal physiological remodelling. This can include segmental capillary atrophy (pruning of regressed capillaries) and the formation of IA [20, 21]. Regressed capillaries that appear during physiological remodelling have been described as ‘strings’ or ‘empty sleeves’ [22]. The CNS microcirculation is below direct detection by magnetic resonance imaging (MRI), although through specific imaging methods the occurrence of brain microvasculopathy in COVID-19 patients has been suggested [23].
Both reactive and acute inflammatory microcirculatory alterations are present in all 36 COVID-19 cohort cases. Principal conditions speculated to underlie microcirculatory injury from COVID-19 are hypoxia [1, 10, 13, 24], hypercytokinemia, and RAS dysfunction [1, 3, 21]. Autoimmunity associated with hypercytokinemia or expressed as type 3 hypersensitivity vasculitis in the CNS has been postulated as a cause of vascular damage in COVID-19 [1, 3, 16, 17]. Predisposing conditions causing chronic vascular-wall injury (e.g., hypertension, diabetes mellitus, chronic hypoxia), as found in this cohort, might leave microcirculatory channels prone to autoimmune damage [17].
Significant direct SARS-CoV-2 infection of the CNS remains an open question. In one immunostaining study, SARS-CoV-2 spike and membrane proteins co-localize in the brain with vascular endothelial staining and with caspase 3, suggesting microcirculatory endothelial-cell viral attachment and perhaps infection with apoptosis. Other organs in that study have similar microcirculatory findings, while the additional step of investigating co-localization of viral RNA in extracranial blood vessels is negative. The conclusion, which may apply to the CNS microcirculation, is that endocytosed pseudovirions formed only of SARS-CoV-2 protein may be associated with vascular endothelial cells [4].
The reactive microvasculopathy in the COVID-19 cohort is of note for its frequency and morphology. Microcirculatory channels with waists, those with starburst profiles, and others with a sinuous, frond-, or tuft-like appearance most likely have a similar origin following mural injury. The IA component is well known in hypoxic brain and in brain tumors [20, 25], and IA has been found in COVID-19 pulmonary tissue more often than in controls [26]. While not being disease-specific, the combination in most of our cohort cases of widespread dilated, distorted, thin-walled microcirculatory channels and of frequent IA in so many microscopic sections is unusual.
Ampoule-like waists similar to those in COVID-19 reactive microvasculopathy have been reported in ataxia-telangiectasia (A-T) in the brain [27, 28], in experimental episodic brain arteriolar network occlusion [29], and in an ischemia-perfusion model in which pericyte contraction causes persistent waists that last even into a return to normoxia [30]. More complex frond-like microvessels, somewhat similar to those in COVID-19, have been described in A-T [31]. The A-T mutation prevents proper microcirculatory remodelling after physiological stress because of the absence of anti-angiogenic mediation that would have been the required brake on the growth and permeability phase of normal microcirculatory healing [32, 33].
The morphology of microcirculatory mural irregularities in the brain in COVID-19, particularly sinuous, frond-, or tuft-like channels, suggests the possibility of stalled microcirculatory remodelling following mural injury when comparing these changes to findings in A-T [31] and to models of experimental hypoxia [20]. These animal models of microcirculatory stress directly address stalled healing.
In general, microcirculatory injury causes hypoxia. In the milieu promoted by microcirculatory injuries, endothelial cells are sensitive to pro-angiogenic stimuli that cause endothelial cells to lose contact with pericytes, which makes endothelial cells prone to apoptosis. This results in segmental regression (atrophy, ‘string’, ‘ghost’, or ‘empty sleeve’ capillaries). In the healing phase (remodelling, pruning), there is competition between pro-angiogenic stimuli and anti-angiogenic mediators such as cytokines. Anti-angiogenic mediation predominates in this phase, allowing vascular regression and healing even in the continued presence of growth stimulation. The inability of pro-angiogenic stimuli to overpower anti-angiogenic mediators avoids a constant proliferative and permeability phase during normal microcirculatory restitution. However, during severe hypoxic stress, including conditions in which IA might develop, the microcirculatory remodelling phase can stall because competing mediators are out of physiological balance [20].
Stalled microcirculatory healing in the brain during severe hypoxia might be exacerbated by hypercytokinemia in COVID-19. The result may be the development of sinuous and other distorted microvascular walls that are reminiscent of A-T microvascular fronds. Furthermore, hypercytokinemia might be one factor in COVID-19 that presumably could stall brain microcirculatory healing even when hypoxia is not severe by upsetting the physiological balance during remodelling mediation.
IA formation appears to be more straightforward as a response to chronic hypoxia. However, IA in the preponderance of sections within the cohort is very uncommon if not unique in a viral disease. ‘Mini-glomeruloid’ microvascular formations resulting from pronounced IA further suggest an effect induced by hypoxia [20, 25]. However, in COVID-19 it is possible that additional aberrant capillary growth mediation might be involved including perhaps influences of hypercytokinemia or perhaps RAS activity.
Acute neutrophilic endotheliitis involves various organs in COVID-19, but it has not been demonstrated in the CNS in this disease [11, 17]. In our cohort, acute endotheliitis is found in two of the three major brain regions in half of the cases, including in the brainstem in all but one case. Acute endotheliitis is an autoimmune vasculitis that is the early phase of type 3 hypersensitivity vasculitis. In acute neutrophilic endotheliitis, karyorrhexis is seen as a tuft-like, beaded, or ‘nuclear dust’ PMN nuclear fragmentation within microcirculatory channels, as found throughout our case cohort. PMN activation in this phase of autoimmune vasculitis is the main vascular damage effector along with hypercytokinemia [15, 34]. Many karyorrhectic intraluminal PMNs mark the microcirculation affected by acute endotheliitis in our cohort. This finding is beyond the occasional karyorrhectic PMNs found within the CNS microcirculation in some autopsy cases, particularly in patients with hypertension where the intermediate fibrinoid necrosis stage of type 3 hypersensitivity vasculitis may develop [34–36]. It is also of note that karyorrhectic PMNs in autoimmune vasculitis precede mononuclear cell recruitment in small-vessel vasculitis, and that these vasculitides can include eosinophils [34], as identified in our COVID-19 cohort.
Dehiscent microcirculatory channels might be consistent with type 3 hypersensitivity-induced small-vessel necrosis [34, 37]. However, dehiscent channels may also form in a different manner in this cohort such as following hypercytokinemia, RAS dysfunction, and microcirculatory remodelling, as well as following thromboembolism [38] and possibly the effects of viral capsid proteins or direct infection. Dehiscent microcirculatory channels might also develop after stalled healing of capillaries when pericytes are lost and endothelial cells undergo apoptosis.
Neutrophilic extracellular traps (NETs), although proposed as a possible mechanism for cerebrovascular injury [39], have yet to be demonstrated in the CNS in COVID-19. NET morphology involves extracellular chromatic material with a different morphology than apoptosis with pyknosis or nuclear fragmentation [40].
Findings in this cohort proximate to death consistent with the early phase of type 3 hypersensitivity vasculitis, in addition to acute endotheliitis, include PMN mural transmigration which leads to acute perivasculitis [34, 36, 41, 42]. Acute perivasculitis is present in 72.2% of our cohort cases. Fibrinoid necrosis and intimal fibrosis that occur in subsequent stages of type 3 hypersensitivity vasculitis are not identified. The early neutrophilic phase may occur days to weeks after circulating immune complexes form in response to a foreign antigen or through molecular mimicry (often to a viral infection). The immune complexes can deposit on a tissue site, which is typically a vascular wall in a viral infection. In some instances, the circulating antibody will complex with a ‘planted’ antigen [15, 17]. The possibility of circulating and ‘planted’ SARS-CoV-2 protein in brain microcirculatory channels requires further scrutiny.
An interesting ‘planted’ antigen scenario has been suggested for SARS-CoV-2 infections. Sequence analysis demonstrates that some human chaperone proteins might be able to participate in molecular mimicry with SARS-CoV-2 because of shared amino acid sequences. It has been postulated that chaperones might become localized in vascular endothelial-cell plasma membrane following signalling from shear and metabolic stress such as that related to risk factors for hypertension and diabetes mellitus [43].
Specific inhibitors of anaphylaxis might provide effective prophylactic treatment for terminal complement component generation in type 3 hypersensitivity vasculitis [17]. Treatment for acute endotheliitis is likely to include anti-inflammatory and immune-modulating drugs and also inhaled nitric oxide to induce vasodilation and for its anticoagulant and direct antiviral activity [44]. Dexamethasone treatment during at least part of the disease course had no appreciable effect in our cohort.
Other potential CNS effects due to microvasculopathy in COVID-19
Cardiopulmonary pacing regions in the brain include the major integrating nucleus of the tractus solitarius and its many connections, such as the hypothalamus, the sensorimotor cortex, and the insular cortex [45]. All of these brain regions exhibit microcirculatory injury in this COVID-19 cohort, including both reactive microvasculopathy and less frequently acute endotheliitis. From a viewpoint centered on COVID-19 microvasculopathies, a variety of functional microcirculatory problems might arise in any brain area from any of the aforementioned origins in COVID-19 to result in local neuronal dysfunction reflected as isolated, multifocal, fleeting, recurring, minor, or devastating CNS or peripheral nervous system change.
Vascular endothelial cells, microvascular-wall pericytes, perivascular astrocytes, resident CNS microglia, and neurons form neurogliovascular units (NVUs) in the CNS. Neuronal activity induces this cellular complex, which is coupled together in each NVU, to cooperate in the regulation of blood flow in support of uninterrupted neuronal transmission. Microcirculatory remodelling following infarcts, hemorrhage, and viral infections is accompanied by NVU uncoupling during which activation of pericytes, astrocytes, and resident microglia occurs, including the production of pro-inflammatory cytokines. NVU uncoupling decreases the client neuron’s energy availability, thereby decreasing neuronal activity. Reactive, thromboembolized, or acutely inflamed CNS microcirculatory channels thus provide a negative effect on neuronal activity by denying the metabolic support of episodically-required blood flow [30, 45]. Functional MRI performed on COVID-19 patients has shown early results regarding brain regions with neuronal activity/vascular flow mismatches that would be the expected finding in brain circulatory lesions [46, 47].
Hypoxia sufficient enough to induce a sustained contraction of pericytes results in ampoule-like waists in the microcirculation until after NVU recoupling begins [30, 48]. Therefore, the waists in dilated microvessels in the COVID-19 cohort might serve as signs of recent or continuing NVU uncoupling. This appears to apply to Case 1 with only minutes of survival upon hospital arrival as well as to the remainder of the cohort with survival up to 84 days.
In severe hypoxia, these pericyte-mediated microvascular contractions trap red cells [30, 48]. This microcirculatory finding is mimicked during temporary arteriolar network occlusion in animals wherein the microcirculation is distorted by ampoule-like waists and there is red cell and serum protein accumulation in non-flowing microvessels until functional shunting of retrograde flow begins [29]. The dilated microvessels filled by red cells and serum protein in our COVID-19 cohort might only be signs of a terminal event. However, it is possible that such functional shunting to provide oxygen and energy to deprived microvascular beds occurs in COVID-19 cases considering the extent of the reactive microvasculopathy. This set of microcirculatory alterations could be an additional burden on the maintenance of neuronal activity that would likely already be burdened by hypoxia, hypercytokinemia, microthrombi, thromboembolism, acute endotheliitis, and possibly pseudovirion endocytosis or a viral infection. A further effect postulated to “stall” capillary blood flow in COVID-19 patients is the presence of increased mononuclear cells that are part of the inflammatory response [24]. Together, these concomitant microcirculatory stress factors might underlie some of the reported neurological symptoms in COVID-19 [1, 3, 24].
A further factor that could induce microcirculatory stress has been the observation of possible megakaryocytes in cerebral capillaries in COVID-19 [49]. There is a similar finding in 18 of our cohort cases (50%). It should be noted, however, that large, hyperchromatic capillary mural nuclei, occasionally with a syncytial appearance, can be found in CNS and skeletal muscle endothelial cell infections caused by some bacteria and viruses [50, 51]. Enlarged but often clear capillary mural nuclei are reported in animal models of cerebral hypoxia [52]. Intussusceptive extension of endothelial cells which have large dark nuclei early in IA have a similar appearance [26], as seen in our cohort. It is likely that these scattered, single-capillary findings in tissue sections are not of a single cause, and further study is warranted.
Finally, note is taken of the possibility of a central hypoventilation syndrome which has been suggested to occur in COVID-19 involving the brainstem’s central cardiopulmonary pacing network [53]. In this syndrome, there is failure of the switch from automatic to voluntary breathing around daybreak, and thus respiratory effort ceases as normal pacing fails from a variety of causes [54]. Failure of the switch from automatic to voluntary breathing remains a question that is not reliably supported in our small case cohort.