Cellular mechanisms underlying neurological/neuropsychiatric manifestations of COVID-19

Abstract

Patients with SARS-CoV-2 infection manifest mainly respiratory symptoms. However, clinical observations frequently identified neurological symptoms and neuropsychiatric disorders related to COVID-19 (Neuro-SARS2). Accumulated robust evidence indicates that Neuro-SARS2 may play an important role in aggravating the disease severity and mortality. Understanding the neuropathogenesis and cellular mechanisms underlying Neuro-SARS2 is crucial for both basic research and clinical practice to establish effective strategies for early detection/diagnosis, prevention, and treatment paradigms. In this review, we will comprehensively examine current evidence of SARS-CoV-2 infection in various neural cells including neurons, microglia/macrophages, astrocytes, pericytes/endothelial cells, ependymocytes/choroid epithelial cells, and neural stem/progenitor cells. Although significant progress has been made in studying Neuro-SARS2, much remains to be learned about the neuroinvasive routes (transneuronal and hematogenous) of the virus and the cellular/molecular mechanisms underlying the development/progression of this disease. Future and ongoing studies require the establishment of more clinically relevant and suitable neural cell models using human induced pluripotent stem cells, brain organoids and postmortem specimens.

Introduction

SARS-CoV-2 virus has caused a global pandemic affecting 218 countries and territories with 51,791,296 confirmed cases and 1,278,460 confirmed deaths, as of November 10, 2020 (www.worldometers.info/coronavirus), and large numbers of new cases still occur daily. The unprecedented scale of the pandemic continues to push the boundaries of scientific research and development as the community proceeds to further understand the virus and its complexity involving multiple organ systems with varying severities. The most common and predominant presentations of illness are respiratory symptoms. However, there are increasing reports of neurological and neuropsychiatric symptoms^1–8. Such symptoms may be nonspecific, such as somatoform disorders including muscle sore, dizziness, headache, anxiety, sleep disturbances, depression, fatigue, anger/stress, loneliness, stigmatization; or specific, such as central nervous system (CNS) disorders including stroke, seizures, encephalitis, ataxia, and myelitis; peripheral nervous system (PNS) disorders including loss of smell and taste (anosmia/hyposmia and ageusia/dysgeusia), Guillain-Barré syndrome and Miller Fisher syndrome; or psychiatric and neuropsychiatric syndromes including major depressive disorder, bipolar disorder, obsessive-compulsive disorder and post-traumatic stress disorder. The exact mechanisms of SARS-CoV-2 underlying neurological/neuropsychiatric dysfunctions are not fully understood but seem to be due to a combination of potentially direct viral infection, or indirectly systemic disease complication plus para/post-infectious inflammation in the nervous system. Notably, knowledge of the long-term effects of SARS-CoV-2 on the nervous system in surviving COVID-19 patients is virtually absent.

In this review, we will focus on the current understanding of cellular mechanisms underlying COVID-19 neurological/neuropsychiatric pathogenesis and symptoms. Accumulated evidence has shown that SARS-CoV-2 may directly infect various neural cells. Additionally, systemic immune activation and chronic neuroinflammation may also contribute to neurological dysfunctions. Although severe neurological complications seem to be rare, the global scale of the pandemic can amount to a large proportion of cases in the long-term as symptoms and latent effects of the virus continue to manifest, potentially leaving patients with prolonged existing of neurological and/or neuropsychiatric disturbances^{7, 8}. Presumably, neurological and neuropsychiatric manifestations in surviving COVID-19 patients may have equal, if not more significant, adverse consequences when compared to pulmonary manifestations. Understanding the cellular mechanism will pave the way for basic research directions and clinical therapeutic strategies. For better clarity and understanding, we tentatively assign SARS1 for the original 2003 SARS caused by SARS-CoV-1 and SARS2 for the 2019 COVID caused by SARS-CoV-2, while SARS defines both SARS1 and SARS2, and SARS-CoV represents SARS-CoV-1 and SARS-CoV-2.

1. Clinical evidence of SARS-CoV-2 infection in the nervous system

Neurologic and neuropsychiatric dysfunctions in SARS2 patients have recently been well discussed^{1–6, 9}. However, many questions remain to be clarified, particularly if these neurologic/psychiatric disturbances are caused by a direct infection of SARS-CoV-2 to the brain or the downstream consequences via indirect mechanisms caused by SARS-CoV-2 systematically. The previous clinical findings of the brains from patients infected with SARS-CoV-1 have suggested that these SARS-CoVs are able to invade the brain. Subsequent clinical reports have documented both CNS and PNS involvement including stroke, seizures, smell loss, Guillain-Barré syndrome, and myopathies (manifested as rhabdomyolysis)¹⁰. Autopsies from several SARS1 patients revealed the presence of SARS-CoV-1 in the brain, primarily in the cytoplasm of neurons in the cortex and hypothalamus¹¹. Additionally, SARS-CoV-1 could be detected in the cerebrospinal fluid (CSF) of SARS1 patients who were experiencing seizures¹⁰. Similarly, neurological complications were reported in some patients with MERS, including peripheral neuropathies, encephalopathies, intracerebral hemorrhage, stroke, seizures, and confusion¹². However, there is no confirmatory evidence from post-mortem or CSF viral detection studies to support MERS-CoV directly infecting the brain.

In SARS2 cases, SARS-CoV-2 has been detected by RT-PCR assay in the CSF of numerous patients with various neurological symptoms^13–17 although many anecdotal clinical reports showed a negative test for SARS-CoV-2 by RT-PCR assays of CSF samples¹⁸. The reason why no viral load was detectable in the CSF of the majority of SARS2 patients with an active neurological condition remains unclear. This discrepancy could be due to the differences in sensitivity of RT-PCR assays, the stages of infection and the timing when the CSF sample is manipulated and the assay is performed, or the clearance of SARS-CoV-2 from the CSF. Nonetheless, detection of SARS-CoV-2 by RT-PCR in CSF from patients with active neurological symptoms has provided undeniable evidence that SARS-CoV-2 could invade the nervous system. Undoubtedly, more studies are needed to secure convincible evidence.

A case series study of 18 post-mortem SARS2 patients with various neurological symptoms has shown evidence of acute hypoxic ischemic damages in the brains of all patients, including neuronal loss¹⁹. Of these patients, the virus was detected by RT-PCR in 5 brains. SARS-CoV-2 spike (S) protein was detected in cortical neurons and endothelial cells in SARS2 brain samples²⁰ but not the viral nucleocapsid (N) protein by immunohistochemical analyses¹⁹.

Other neuropathological case studies have found histological evidence of vascular etiologies, encephalitis (both global and brain stem-specific) and/or meningitis, demyelination, and other encephalopathies^{21, 22}. Postmortem analysis using transmission electron microscopy and RT-PCR assay in a SARS2 patient with Parkinson’s disease revealed the presence of SARS-CoV-2 in neurons and capillary endothelial cells in the frontal lobe²³. However, the question remains whether the outcome neuropathology resulted from direct infection or a downstream event following a systemic infection; or whether the lack of evidence of vital direct infection was an outcome of low sensitivity of detection threshold in methodologies. New technologies such as RNAscope have been utilized to detect SARS-CoV-2 RNA in cultured cells²⁴ and lung specimen of SARS2 patients, and are expected to validate the existence of SARS-CoV-2 in brain tissues. Digital droplet PCR, viral outgrowth assay, and single cell RNA sequencing of brain samples from SARS2 patients may also help to validate the neurotropism of SARS-CoV-2. Further multilabeled immunohistochemical analyses are needed to better understand the cellular distribution and mechanisms of SARS-CoV-2 in the nervous system.

In summary, accumulated evidence from clinical studies has shown a variety of neurological and neuropsychiatric manifestations in SARS2 patients (Table 1). However, more convincible evidence for direct viral infection to the nervous system remains to be found in addition to the indirect contribution of systemic immune/inflammatory responses after SARS-CoV-2 infection²⁵. The availability of the advanced new technologies for neuropathogenic studies on human brain specimen are expected to continuously provide in-depth findings of SARS-CoV-2 infection in the nervous system.

Table 1:

Clinical Features of Neurological and Neuropsychiatric Symptoms in SARS2 Patients (Percentages derived from the clinical studies with >100 patients; The others noted the case numbers)

Manifestation Type	Symptom/Feature	Percentage	Reference
General Neurological Symptoms	Any nervous symptoms	36.4%–82.3%	97–99
	Headache	6.5%–70.3%	79, 97–105
	Muscle pain (myalgia)	22%–62.5%	79, 97–105
	Dizziness	8%–29.7%	79, 98, 99, 101, 105
	Fatigue	50%–69.6%	79, 101, 103, 105
	Seizures	<1%	98, 99, 106–108
	Impaired/disordered consciousness	4.3%–9%	98, 99, 101, 109
	Rhabdomyolysis	3.5%	98
Sensory Impairments	Anosmia/hyposmia	5.1%–70.2%	76, 79, 97–99, 103, 104
	Ageusia/dysgeusia	5.6%–54.2%	76, 79, 97–99, 103, 104
	Vision impairment	1.4%–7.4%	97, 99, 103
Neuropathies	Peripheral neuropathy	0.4%–8.9%	98, 99, 107, 108
	Gullian-Barre Syndrome and variants	23 cases	107, 108, 110
	Nerve pain	2.3%	99
	Focal weakness and/or motor deficit	0.6%–2.5%	97, 98
Encephalopathies	Encephalopathy (undefined)	7.2%–31.8%	98, 107, 108
	Confusion/disorientation/altered mental state	2.0%–27.9	97, 100, 106, 108, 109
	Acute necrotizing encephalopathy (ANE)	3 cases	17, 111
	Leukoencephalopathy	2 cases	21, 107
	Hypoxic encephalopathy	9%	101
Inflammations	Acute demyelinating encephalomyelitis (ADEM)	10 cases	22, 107
	Myelitis (CNS)	3 cases	107
	Encephalitis	15 cases	98, 107, 108, 112
	Lymphocytic meningitis	6 cases	112
Acute cerebrovascular disease (cumulative cases and/or unspecified)	Ischemic stroke	1.4%–5.0%	21, 99, 106–108, 112, 113
	Intracerebral hemorrhage	25 cases	21, 98, 106, 108, 112, 113
	Depression	32.6%	109
	Anxiety	35.7%	109
Neuropsychiatric Presentations	Sleep disorders/insomnia	37.2%–41.9%	97, 109
	Anger/stress/irritability	3.9%	109
	Psychosis	0.8%–4.4%	108, 109
	Behavioral disorder (not specified)	2.0%	97
	Frequent recall of traumatic memories		109
	Cognitive disorder	0.5%	108
	Acute hypoxic ischemic damage		19, 101
	Astrogliosis		22, 114, 115
Neuropathologic Features	Microgliosis		115
	Demyelination		22
	Neuronal Cell Loss/Injury		22, 112
	Cytotoxic T-cell infiltration		115
	Macrophage infiltration		22

2. The neuropathogenic modeling for SARS2

Due to the limited sources of human brain specimen from SARS2 patients, various models have been developed to explore the neurovirulent infection of SARS-CoV-2 in various types of neural cells and recapitulate the neuropathogenesis and neurological manifestation as well as their correlations with viral neuroinvasion (Table 2).

Table 2:

Experimental Models for SARS2 Neuropathogenesis

	Model	Description	Model Characteristics	References
In vivo Mouse Models	Transgenic K18-hACE2 mice	Human ACE2 expression driven by the epithelial cell-specific K18 promoter	High susceptibility to SARS-CoV-2 infection (2×103~1×105 plague formation unit) and best recapitulation of several SARS2 clinical features; Several studies also report presence of viral RNA and/or histopathological findings in the brains of some mice20, 32, 35, 36, 80, 116	20, 32, 33, 35, 36, 80, 116, 117
	Transgenic HFH4-hACE2 mice	Human ACE2 expression driven by the lung ciliated epithelial cell-specific HFH4/FOXJ1 promoter	Susceptibility to SARS-CoV-2 infection and presentation of interstitial pneumonia; High viral RNA levels were detected in the brain of a portion of mice and was highly correlated with increased mortality	34
	Transgenic hACE2 mice	CRISPR/Cas9 edited knock-in mice that express human ACE2 under the mouse genomic Ace2 promoter	Human ACE2 replaces mouse Ace2 expression; High susceptibility to SARS-CoV-2 infection, with particularly high viral RNA levels found in the brain	31
	Adenovirus delivery of hACE2	Human ACE2 packaged in adenoviral particles under the CMV promoter and delivered to wildtype mice for exogenous expression of hACE2	Human ACE2 expressing cells show susceptibility to SARS-CoV-2 infection, with high viral titers in lung tissues; Mice develop some pathologies but do not display severe SARS2 features	32, 118, 119
In vitro Human iPSC-Derived Culture Models	Monolayer neural cell cultures	Human iPSC-derived neural cell cultures patterned into specific cell types.	Monolayer cultures of mature neurons41, 42, 47 of various subtypes (e.g.- dopaminergic, cortical, etc.), NSCs/NPCs20, 40, 73, and glial cell cultures41, 42, including microglia and astrocytes; Susceptibility to SARS-CoV-2 infection varies by cell type	20, 40–42, 47, 73
	Cerebral organoids	3D cerebral organoids generated from human iPSCs	Mature organoids present as a heterogenous 3D structure consisting of multiple neural cell types, including neurons, NSCs/NPCs, and glial cells; Susceptible to SARS-CoV-2 infection to various extent	20, 40, 44, 45, 47, 120
	Region-specific cerebral organoids	3D cerebral organoids generated from human iPSCs that are patterned to specific regional lineages	Organoids of hippocampal, midbrain, hypothalamic, and forebrain subtypes with heterogenous mixture of neural cells; Varied susceptibility to SARS-CoV-2 infection	41
	Choroid plexus organoids	3D cerebral organoids generated from human iPSCs primarily composed of choroid plexus epithelial cells	High susceptibility to SARS-CoV-2 infection; SARS-CoV-2 infection leads to transcriptional dysregulation and impaired barrier function	41, 46
	Blood vessel organoids	3D organoid structures derived from human iPSCs composed of vascular networks of endothelial cells	High susceptibility to SARS-CoV-2 infection that can be inhibited in an ACE2-dependent manner	65

a. Preclinical animal models

Angiotensin-converting enzyme-2 (ACE2) has been well recognized to serve as the major receptor for SARS-CoV-2 to gain entry into host cells²⁶. The homology conservation and the expression patterns of ACE2 in various species have been identified²⁷. Therefore, numerous non-rodent species have been tested as natural animal models for SARS2, including monkey, ferret, hamster, cat, bat, and pig²⁸. Cynomolgus macaques infected with SARS-CoV-2 exhibit virus shedding and replication in lung tissues and certain pathological changes, but no clinical signs²⁹, while another study showed moderate clinical symptoms in rhesus macaques infected with SARS-CoV-2³⁰. Thus, non-human primate models may be useful for further preclinical tests; however, monkey models in general fail to manifest clinical illness and are very expensive and inconvenient. Mouse models are more practical and feasible for elucidating virus pathogenicity and therapeutic targets. However, both SARS-CoVs exhibit a poor cellular tropism in wild type mice due to significant species difference in ACE2 sequence and function. Thus, various types of human ACE2 (hACE2)-expressing mouse models have been established for SARS-CoV-2 infection (Table 2). Interestingly, only hACE2 mouse models among the currently available SARS2 animal models fully recapitulated SARS2 clinical scenarios, particularly the severity and mortality of the illness. The successfully established hACE2-transgenic mouse model with high susceptibility to both SARS-CoV-1 and SARS-CoV-2 infection provided a useful tool in SARS-CoVs studies^31–34, particularly the K18-hACE2 transgenic mice, in which hACE2 expression is driven by the epithelial cell-specific promoter human cytokeratin 18 (K18) and is capable of recapitulating many features of SARS-CoV-2 infections^{32, 33, 35}, even better than the other hACE2-expressing mouse models^{32, 35}.

Although numerous reports on SARS2 animal models are available, the understanding of neuroinvasive pathogenesis and resultant neurological changes caused by the virus remain limited. In SARS2 animal studies, the virus is delivered intranasally, which may readily infect brain via olfactory nervous system. The brain cellular expression pattern of ACE2 in these animal models may address the susceptibility of brain tissues/cells to SARS-CoV-2. SARS-CoV-2 RNAs are detected in the brain of hACE2 mice after intranasal inoculation^{31, 34, 36}.

b. Human brain organoid models

To get the most clinically relevant data for whether and how SARS-CoV-2 may infect human brain, it is important to use materials from humans. The simplest and most direct approach to test human cellular susceptibility to SARS-CoV-2 is to utilize immortalized human neural cell lines or primary neural cells. SARS-CoV-2 can moderately infect and replicate in the human glioblastoma cell line U251³⁷. Primary olfactory sensory neurons of human⁴ and hamsters³⁸ are highly susceptible to SARS-CoV-2 infection. However, another study did not detect any presence of SARS-CoV-2 in primary olfactory neurons in hamsters, but did identify viral infection in a large proportion of the supporting sustentacular cells³⁹. In cultured human neural stem cells (NSC)/neural progenitor cells (NPC), ACE2, TMPRSS2, cathepsin L, and furin were readily detected and cells were highly susceptible to infection with SARS-CoV-2⁴⁰, but not SARS-CoV-1⁴⁰. Human primary astrocytes are rarely (0.18%) infected by SARS-CoV-2⁴¹. Other human primary neural cells have not yet been examined for their susceptibility to SARS-CoV-2 infection.

Fortunately, the rapidly evolving technology surrounding stem cell culture has allowed for the generation of human induced pluripotent stem cells (iPSCs) that can be differentiated into virtually any cell lineage. This has led to the generation of pure human neural cells such as NSCs/NPCs, neurons, astrocytes, microglia, oligodendrocytes, endothelial cells, etc. Human iPSC-derived neurons, particularly dopaminergic neurons, are highly susceptible to S-pseudotyped SARS-CoV-2 infection, but iPSC-derived microglia and macrophages are moderately infected by S-pseudovirions⁴². In contrast, another report showed that iPSC-derived neurons and astrocytes were sparsely infected, and microglia had no infection; however, interestingly, the choroid plexus epithelial cells underwent robust infection by live SARS-CoV-2 virus⁴¹. The inconsistent reports in neural cell viral susceptibility may result from variable differentiation protocols, the neural cell maturity, and virus infection dosage.

Three-dimensional tissue cultures known as organoids contain multiple cell types and recapitulate tissue structures and development. Cerebral organoids (COs) have been used to model several neurodevelopmental and neuropathological disorders as they accurately resemble the molecular, cellular, and structural features of the developing brain (Figure 1a). COs can be readily infected with neurotropic viruses and develop phenotypes similar to that seen clinically, as has been shown in several models studying Zika virus (ZIKV)⁴³.

Figure 1: Human brain organoids to model SARS-CoV-2 infection.

(a) Example timeline of an experiment to generate cerebral organoids (COs) with SARS-CoV-2 infection. Human fibroblasts or peripheral blood mononuclear cells (PBMC) are reprogrammed to generate induced pluripotent stem cells (iPSCs), which are then grown in 3D suspension culture as embryoid bodies (EBs). After neural induction, COs are embedded in Matrigel to promote formation of neuroepithelial lobes and then transferred to spinning culture, using an orbital shaker or similar method. At around one month, microglia begin to mature in microglia-containing COs. By around 50–60 days, astrocytes can be detected in COs. At this point, COs are in a state of growth and neural cell maturation; this is typically the timepoint where COs are infected with SARS-CoV-2 for experiments. (b) Examples of different types of COs. After pre-patterning to a neuroectodermal fate using dual SMAD inhibition, EBs can be induced into region-specific organoids by adding different patterning factors. Several studies have used this technique to examine the effect of SARS-CoV-2 infection on different brain regions, particularly the choroid plexus. MPC, microglia progenitor cells; NSC/NPC, neural stem/progenitor cells; OPC, oligodendrocyte progenitor cells.

As clinical samples from SAR2 patients are scarce, COs make an attractive experimental model for studying the neuropathogenic effects of SARS-CoV-2 infections in human live brain. To date, several labs have utilized COs to study the neurotropic potential of SARS-CoV-2; all have consistently found that the virus can readily infect and replicate within COs derived from human iPSCs^{20, 40–42, 44–46}. In most studies, SARS-CoV-2 preferentially infected neurons and viral particles were primarily localized to the soma, sometimes extending into neurites and axons. This infection of COs was further confirmed in two studies using electron microscopy, which detected viral particles in neurons and viral budding in the endoplasmic reticulum^{20, 40}. An additional study using a SARS-CoV-2 pseudovirus instead of live virus also found that neurons of COs readily uptake viral particles, with a similar expression pattern⁴⁷. While this evidence does not take into consideration how SARS-CoV-2 would cross the BBB to enter the brain, it does show that human brain organoids are susceptible to the SARS-CoV-2 infection.

3. Neural cell susceptibility to SARS-CoV-2

SARS-CoV-2 is primed by proteases such as transmembrane protease serine 2 (TMPRSS2) and requires expression of its receptor angiotensin-converting enzyme-2 (ACE2) to gain entry into host cells²⁶. However, several other receptors have been identified that serve as potential receptors or co-receptors for SARS-CoV-2 infection such as neuropilin-1⁴⁸, CD147^{49, 50}, CD26^51–53, CD133⁵³, CD209⁵⁴, furin^{40, 50}, and cadherin-17⁵³. Therefore, studying the expression of ACE2 and other potential receptors throughout the body helps bring understanding to cell susceptibility to SARS-CoV-2 infection. Earlier studies found ACE2 mRNA expression nearly ubiquitous but its protein expression more restricted, with high levels found in the heart, kidneys, lung epithelium, small intestine, and vascular endothelium⁵⁵. In the brain, neural cells include neurons, astrocytes, oligodendrocytes, microglia/macrophages, NSCs/NPCs, ependymal cells, pericyte, endothelial cells and various type of immune cells (Table 3). These neural cells also exhibit enormous diversity in their subtypes, sizes, shapes, regional locations, and functions (Figure 1b). The expression pattern of ACE2 and other potential receptors in various neural cells remains to be determined and their susceptibility to SARS-CoV-2 may vary with physiological and pathological conditions. Exploration of SARS-CoV-2 neurotropism and cellular localization is important in elucidating SARS2 neuropathogenesis and developing therapeutic and prophylactic countermeasures.

Table 3:

Neural Cell Susceptibility to SARS-CoV-2 Infection

Cell Type	Receptor Expression	Susceptibility to Infection	Consequences of Infection in vitro	Consequences of Infection in vivo
Neurons	Varied expression of ACE2 (subtype and region specific)	Varying reports of susceptibility	Increased cell death20, 41, 44, 45; impaired synaptogenesis20, 40, 44; Tau mislocalization45	Evidence of direct SARS-CoV2 infection in neurons of K18-hACE2 and hACE2 transgenic mice20, 31, 36, 80 and select clinical SARS2 cases20, 23; Neuronal injury17, 19, 21, 22, 59, 112, 114, 121, 122, neuronal necrosis and/or neuronophagia19, 22, 115, 121, 123 detected in clinical SARS2 cases
Microglia	Inconsistent reports	Inconsistent reports	Not available due to low susceptibility to SARS-CoV-2 infection	Reactive microglia/microgliosis detected in SARS-CoV-2 infected K18-hACE2 transgenic mice36 and clinical SARS2 cases19, 59, 115, 121–123, but no evidence of direct infection of microglia; Presence of microglial nodules in select SARS2 cases19, 115, 121, 123
Pericytes/ Endothelial Cells	High expression of ACE255, 62, 68 and TMPRSS2	High susceptibility	High levels of viral replication and shedding	Vasculitis and other cerebrovascular etiologies detected in K18-hACE2 transgenic mice20, 36, 80; Acute cerebrovascular injuries in numerous clinical SARS2 cases (see Table 1)
Choroid Plexus Cells	High expression of ACE2 and TMPRSS241, 46, 68, 69, 124	High susceptibility41, 46	Increased cell death, downregulation of genes related to BCSFB integrity and upregulation of inflammatory genes41; Impaired barrier function46	One clinical case report of direct SARS-CoV2 infection in endothelial cells of the choroid plexus and tanycytes of the median emminence124
NSCs/NPCs	Moderate expression of ACE2; expression of TMPRSS2, furin, and cathepsin L40	Conflicting reports of low to moderate susceptibility20, 40, 44, 46	Increased cell death20, 40, 44	No evidence available
Astrocytes	Weak to little ACE2 expression69	Low susceptibility36, 41, 46, 58	Not available due to low susceptibility to SARS-CoV-2 infection	Reactive astrocytes/astrocytosis detected in SARS-CoV-2 infected K18-hACE2 transgenic mice36, 80 and clinical SARS2 cases17, 22, 114, 115 but no evidence of direct infection of astrocytes

a. Neurons

Both in vitro and in vivo studies have demonstrated the neuronal susceptibility to SARS-CoV-2, although the infection efficiency varied with neuronal subtypes, regional difference, and entry receptor expression patterns. Database mining showed weak expression of hACE2 in human brain cortical neurons but strong expression of CD147 and neuropilin-1. Experiments using RT-PCR, immunostaining and immunoblotting validated hACE2 expression in human neurons to various extents. The expression of entry receptors destinates neuronal infectivity of SARS-CoV-2. Human neuronal cell line U251³⁷ and primary olfactory sensory neurons⁴ are susceptible to SARS-CoV-2 infection. In hamsters, one report identified high infection of primary olfactory neurons, but another study found viral infection only in sustentacular cells instead of olfactory neurons (Figure 2a)³⁹. In mice, pervasive expression of Ace2 protein in olfactory epithelial sustentacular cells and olfactory bulb pericytes but not olfactory sensory neurons was detected by immunohistochemistry, suggesting that olfactory bulb neurons are likely not a primary site of infection, but that vascular pericytes may be sensitive to SARS-CoV-2⁵⁶. Interestingly, SARS-CoV-2 infected COs at Day 60 showing a significantly higher rate than at Day 15, suggesting that the virus prefers infecting more mature neural cells⁴⁵ (Figure 1). Neurons in hypothalamus and with dopaminergic properties are highly susceptible to SARS-CoV-2, indicating its possibly potential affinity to some neuroantomic structures underlying pathogenesis for clinical Neuro-SARS symptoms⁵⁷. The S protein-immunoreactivity is detected in brain neurons of K18-hACE2 mice after SARS-CoV-2 infection³⁶ by intranasal delivery, suggesting the possibility that SARS-CoV-2 live virus can inffect brain neurons via the intranasal route.

Figure 2: Diagrams for some potential routes of SARS-CoV-2 neuroinvasion.

There are two primary routes by which SARS-CoV-2 may invade the brain: (1) transneuronally or (2) hematogenously. Potential transneuronal routes include nasally via the olfactory system, orally via the gustatory/trigeminal nerves, spinally via the dorsal root ganglia (either via sensory receptors in the skin or in the gut), or through the enteric nervous system via the vagus nerve. Potential hematogenous routes can occur with an intact blood-brain barrier (BBB) via the choroid plexus or circumventricular organs (CVOs), or with an impaired BBB via damaged endothelial cells. (a) SARS-CoV-2 may invade the brain via the olfactory system. Viral particles enter the olfactory epithelium after passage through the nasal cavity. Within the olfactory epithelium, SARS-CoV-2 can create a reservoir within support cells, such as sustentacular cells. The virus may enter the olfactory bulb transneuronally, via axons of olfactory receptor neurons, or by alternative routes, such as infection of vascular pericytes (not pictured) or entry into the cerebrospinal fluid (CSF) in the subarachnoid space. Basal cells of the olfactory epithelium and olfactory ensheathing cells within the lamina propria and subarachnoid space are also susceptible to SARS-CoV-2 infection. (b) SARS-CoV-2 may invade the brain by crossing the BBB or blood-CSF barrier. Viral particles travel through the blood and exit through fenestrated capillaries within the choroid plexus. SARS-CoV-2 may infect choroid plexus epithelial cells and be shed into the CSF, from which they could cross into brain parenchyma. Additionally, viral particles may pass through fenestrated capillaries of CVOs directly into brain parenchyma and/or CSF. (c) SARS-CoV-2 may invade the brain by crossing the BBB. Viral particles travel through the blood and exit into the brain parenchyma by infecting or passing through damaged endothelial cells and/or their surrounding pericytes or hijacking by infected immune cells (not pictured). Within the parenchyma, SARS-CoV-2 may directly infect neural cells or elicit an immune/inflammatory response. Schematic created using BioRender.com.

b. Microglia/macrophages

Microglia in the brain play a critical role in neurodevelopment, neural innate immunity, and neuroinflammation. Microglia, like macrophages, may predominantly contribute to the inflammatory response during Neuro-SARS2⁵⁸, as significant activation of microglia and neuronophagia exists in the brainstem (pons, medulla oblongata), and olfactory bulb of SARS2 patients⁵⁹. Whether SARS-CoV-2 can infect and replicate in microglia remain elusive. In iPSC-derived microglia, one study showed no infection of live SARS-CoV-2 virus, but another study demonstrated moderate susceptibility to S-pseudovirions⁴². SARS-CoV-2 has been shown to infect monocytes and macrophages, resulting in host immunoparalysis⁶⁰. In human macrophages from lymph nodes and lung, SARS-CoV-2 was detected by immunohistochemistry and electron microscopy⁶¹. Whether infected macrophages infiltrate the brain, serving as a hijacking mechanism for SARS-CoV-2, remain to be determined.

c. Pericytes and endothelium cells (EC)

The blood–brain barrier (BBB) functions to protect the brain against harmful invasion such as viruses. The BBB permeability is exquisitely regulated by the neurovascular unit formed by highly specialized endothelial cells, pericytes, astrocytes, and others (Figure 2c). Early studies with immunohistochemistry showed that ACE2 expression was limited to vascular endothelial cells in the human brain⁵⁵. Interestingly, a recent study demonstrated that ACE2 in the brain is not expressed in the vascular endothelial cells but instead in pericytes⁶². Several single cell RNA-seq studies have identified high level of ACE2 in pericytes⁶³. Pericytes wrap themselves around vascular endothelium and play a critical role in maintaining vascular integrity. However, it is unlikely that the virus would be able to cross the BBB by passing from the blood to pericytes unless the vascular endothelium itself was damaged in the first place (Figure 2c), whether it be from pre-existing conditions, direct viral infection, or post-infection immune response. There has been direct evidence for SARS-CoV-2 infection of vascular endothelial cells and endothelialitis reported in clinical cases⁶⁴. This evidence is further corroborated by a recent study in which SARS-CoV-2 was able to readily infect engineered human blood vessel organoids, which could be reduced using clinical-grade soluble ACE2⁶⁵. iPSC-derived endothelial cells are also susceptible to SARS-CoV-2 infection⁴². The severity of vasculopathy in the brain correlates with the severity of SARS2 neurological symptoms such as in stroke or mental alterations⁶⁶. A clinical case report identified profound pericyte apoptosis in the lungs of SARS2 patients, which might be the initial trigger of micro-vasculopathy⁶⁷. The direct evidence for SARS-CoV-2 infection in brain pericytes and the sequela is still lacking.

d. Choroid plexus epithelial cells and ependymal cells

The choroid plexus is a complex network of epithelial cells surrounding capillaries in the ventricles of the brain and is the primary region of CSF production. The epithelial cells of the choroid plexus also have a unique role as serving as the major barrier between the blood and CSF (termed the blood-CSF-barrier or BCSFB) (Figure 2b). The BCSFB acts similarly to the BBB as a selectively permeable barrier for transport to and from the brain, but likewise is also susceptible to pathogen invasion. One study looking at SARS-CoV-2 infectability in region-specific COs (including hippocampal, hypothalamic, midbrain, and cortical) (Figure 1b) found that choroid plexus epithelial cells expressed in hippocampal organoids were susceptible to SARS-CoV-2 viral infection at a much higher rate than that of neurons and other neural cell subtypes⁴¹. This led the authors to generate choroid plexus specific organoids, which displayed very high rates of infection even at low viral titer, strongly suggesting that SARS-CoV-2 may be able to invade the brain by infecting choroid plexus epithelial cells in humans. Another study validated the high susceptibility of the choroid plexus to SARS-CoV-2 S-pseudovirus and live virus in iPSC-derived brain organoids and demonstrated the leakage of BCSFB due to SARS-CoV-2-induced choroid plexus epithelial damages⁴⁶. This possibility is further supported by recent evidence finding high expression of ACE2 and TMPRSS2 in the choroid plexus of both humans and mice^{46, 68, 69}.

The ependymal cells are unique neuroepithelial cells lining the brain ventricles and are critical for CSF generation and regulation. Their basal membranes contact astrocyte endfeet via tentacle-like extensions. SARS-CoV-2 has been detected in the CSF of SARS2 patients^{13–17, 70, 71}. Whether SARS-CoV-2 in the CSF can infect ependymal cells and render neuroinvasion remains to be determined (Figure 2b). Given ependymal cells have weak or no expression of ACE2⁶⁹ and TMPRSS2⁶⁹, they may express other viral entry receptors for SARS-CoV-2, or may allow neuroinvasion under pathological conditions.

e. NSCs/NPCs

Neurogenesis occurs throughout life in both animals and humans. NSCs are self-renewing, multipotent, and long-lived cells that generate the main cell types of the nervous system. NSCs represent a small population of quiescent and slowly dividing cells, whereas their intermediate NPCs are a large population of amplifying, rapidly dividing cells. NSCs/NPCs are susceptible to various virus infections as such as ZIKV, Dengue virus, CMV, etc.^{43, 72}. ACE2 is highly expressed in human iPSC-derived NSCs/NPCs⁷³; thus, three studies validated infection of SARS-CoV-2 in these cells^{40, 41}. However, another study did not detect ACE2 expression or SARS-CoV-2 infection in NSCs/NPCs⁴⁶. NSCs share properties with pericytes, ependymal cells, and tanycytes. In circumventricular organs (CVO), NSCs may have direct contact with circulating pathogens. The functional states and cell fate decisions of NSCs/NPCs after SARS-CoV-2 infection warrant further investigation.

f. Astrocytes

Since astrocytes are the most abundant cell type in the CNS, their role in neural virulence has been well studied for different viruses. The endfeet of astrocytes play a key role in the formation and function of the BBB. However, SARS-CoV-2 infection efficiency is very low or absent in human primary astrocytes^{41, 46, 58}, which may be due to high heterogeneity of astrocyte subtypes as well as weak/rare expression of ACE2 and TMPRSS2⁶⁹. As is the case with HIV infection in astrocytes, a minority of GFAP-positive cells infected by SARS-CoV-2 may represent NSCs. A better understanding of the similarity and diversity of SARS-CoV-2 infection between NSCs and astrocytes may provide new therapeutic routes for Neuro-SARS2.

4. Potential neuroinvasive routes for SARS-CoV-2 infection

Neurological symptoms and neuropsychiatric disorders in SARS2 patients have been extensively reported and evaluated in numerous review papers^{1–6, 9}. The viral entry receptor expression and infection susceptibility of neural cells strongly support the preliminary conclusion that SARS-CoV-2 can effectively infect various types of neural cells in human brain and cause neurological/neuropsychiatric pathogenesis and symptoms. However, the routes for SARS-CoV-2 neuroinvasive infection remain to be determined. The possibility of the following potential routes is explored based on currently available literature (Figure 2).

a. Retrograde/anterograde axonal transport and transneuronal invasion:

Neurotropic viruses may invade neurons via virion trafficking in a retrograde or/and anterograde manner by interacting neuronal cytoskeletal proteins. After neuronal infection, the virions may be released and transneuronally spread to neighboring or presynaptic neurons. Many RNA viruses have been shown to invade the CNS via this transneuronal pathway including SARS-CoV-1, thus SARS-CoV-2 is extrapolated to follow similar transneuronal routes to infect the nervous system.

a-1. Intranasal epithelium and olfactory nervous system

One of the most plausible proposed routes of neuroinvasion of SARS-CoV-2 is via the olfactory system. Invasion by this route would involve the virus infecting cells of the olfactory epithelium, potentially establishing a viral reservoir (Figure 2a) ^{4, 74}. This is because (1) olfactory epithelial cells have high ACE2 and TMPRSS2 expression⁶⁸; (2) several clinical studies identified SARS-CoV-2 infection in the olfactory nerve system¹⁹, including MRI hyperintensity in the olfactory bulb of SARS2 patient⁷⁵; (3) a large majority of SARS2 patients present loss of smell/taste^76–79; and (4) intranasal delivery of SARS-CoV-2 effectively induces brain infection in animal models⁸⁰. Once cells in the olfactory epithelium are infected, the virus could either travel transneuronally via axons of olfactory receptor neurons or by non-neuronal cells up to the olfactory bulb⁷⁴. However, it is still less clear if SARS-CoV-2 can infect neuronal cells within the olfactory epithelium in order to invade the brain via the olfactory bulb, as olfactory receptor neurons express relatively low levels of ACE2⁸¹. If the olfactory bulb and/or olfactory nerves are infected by SARS-CoV-2, it may provide the mechanisms for the widely reported loss of smell and/or taste in SARS2 patients^{76–79, 82}. Alternatively, impaired sense of smell may be the result of an immune response in the olfactory epithelium, as these cells also co-express high levels of innate immune response genes and olfactory epithelial biopsies from SARS2 patients show significantly increased levels of proinflammatory cytokines⁸³.

a-2. Oral health and gustatory/trigeminal nervous system

Oral health plays an important role in the severity of SARS2 development. Oral bacterial infection may increase the risk of SARS-CoV-2 infection. The majority of SARS2 patients present taste disturbances such as ageusia, dysgeusia, xerostomia^{82, 84}. In some cases this may present as a prodromal symptom or as the sole manifestation of SARS2, particularly at the early stage⁸⁵. A large cohort study which enrolled > 2 million participants concluded that the loss of smell and taste is more predictive of SARS-CoV-2 infection than all other symptoms, including fatigue, fever, or cough. Similar to olfactory neuronal terminals, the gustatory and trigeminal nerve endings may be highly susceptible to SARS-CoV-2 infection, especially at the early stage of infection (Figure 2)⁸⁴. Understanding the mechanism of these neuroinvasions will provide a better strategy for the early diagnosis, prevention, and treatment of SARS2 patients, particularly in regard to neurological manifestations.

a-3. Skin and the peripheral nervous system

The human dorsal root ganglion (DRG) neurons express high levels of ACE2 at both mRNA and protein levels⁸⁶. Database mining showed broad expression of entry receptors in human DRG at the lumbar and thoracic level. In a subset of nociceptive neurons, ACE2 is coexpressed with MRGPRD (Mas-related G-protein coupled receptor member D), a polymodal nociceptive receptor on the nerve endings at the outermost layers of skin and luminal organs. SARS-CoV-2 can survive on the skin for 9 hours⁸⁷. Thus, SARS-CoV-2 may infect human nociceptors, causing not only peripheral pain but also neuroinvasion into the CNS via DRG sensory neurons (Figure 2). However, the direct evidence for the DRG infection is still lacking.

a-4. Gut infection and enteric nervous system

In addition to respiratory symptoms, gastrointestinal symptoms are also paramount in SARS2 patients due to abundant expression of ACE2 and other potential entry receptors in mature enterocytes⁸⁸. Presence of SARS-CoV-2 viral RNA can be identified in the stool of SARS2 patients⁸⁹. The intrinsic enteric nervous system is enriched with neurons and glial cells and exquisitely innervates mucosal epithelial cells. Its location makes the enteric nerve endings be readily exposed to the invading viruses. Notably, high level of ACE2 and TMPRSS2 expression have been identified in a large number of enteric neurons and enteric glia cells⁶⁹. Thus, the enteric nervous system is assumed to be susceptible to SARS-CoV-2 infection. The virus infects enteric neurons and travels along with the well-established gut-brain axis such as the vagus or splanchnic nerve routing into the CNS (Figure 2). Several viruses have been shown to take the gut-brain transneuronal pathway to infect the CNS, such as herpes, varicella zoster virus, and influenza⁹⁰. Additionally, infection in the enteric glial cells could induce various degree of inflammatory responses, which may aggravate systemic and neurological symptoms.

b. Hematogenous invasion to the nervous system

b-1. SARS-CoV-2 hematogenous dissemination to circumventricular organ (CVO) and choroid plexus under physiological condition

Another potential route of SARS-CoV-2 infection into the brain could be via circumventricular organs (CVOs) and/or choroid plexus, where the BBB is absent, particularly at the early stage of SARS2 wherein the BBB and BCSFB remains intact elsewhere (Figure 2b)⁵⁸. The BBB under physiological conditions functions to prevent harmful pathogens from entering the brain. However, CVOs have no BBB structure, allowing the circulating pathogens, along with nutrients and metabolites, cytokines and hormones freely travel through to and out from the brain⁹¹. Specific CVOs include the median eminence, subfornical organ, and organum vasculosum lamina terminalis (around the third ventricles at the mid-level hypothalamus), as well as the area postrema in the brain stem. Direct imaging to track SARS-CoV-2 reporter virus passing through the CVO into the brain in human brain organoids or animal models is warranted.

The theory of the choroid plexus serving as an invasion route for neurotropic viruses was recently highlighted in a study showing that ZIKV can infect pericytes and cross the BCSFB via the choroid plexus in mice⁹². Interestingly, ZIKV’s presence in the CSF and infection of the choroid plexus preceded infection in the rest of the brain, which could be attenuated by neutralizing the virus in the CSF. If SARS-CoV-2 utilized a similar neuroinvasion mechanism, it could explain the inconsistency in CSF detection of the virus. SARS-CoV-2 may enter the brain via the CSF prior to infection of CNS neurons and then the CSF viral levels may decrease over time as it is cleared by the immune system; however, this could be when neurological symptoms begin to manifest as the virus spreads in the brain. Both CVOs and the choroid plexus are characterized by fenestrated and highly permeable capillaries (Figure 2b)⁵⁸. SARS-CoV-2 in the CSF may infect ependymal cells and subsequently infect CVOs then spread to neural cells.

b-2. Transendothelial invasion via damaged BBB under pathological condition

The BBB is formed by the highly specialized endothelial cells, pericytes, and astrocytes, and is regulated by the neurovascular unit. BBB could be damaged by virus infection of any cell components of BBB or the post-infectious inflammatory responses. SARS-CoV-2 virus in the blood may directly infect brain endothelial cells and transendothelially disseminate virus into brain parenchyma via the damaged BBB, which could be aggravated by viral infection and inflammatory storm (Figure 2c)⁹³.

b-3. Hijacking (trajectory) route via infected immune cells

Several neurotropic viruses such as HIV and Western Nile virus have been shown to enter brain through a “Trojan horse” mechanism, in which infected immune cells traffic across the permeabilized BBB and spread the hidden viruses to neurons and glial cells⁹⁴. SARS-CoVs have been shown to infect monocytes/macrophages, lymphocytes, and leukocytes^{60, 95}. Thus, SARS-CoV-2 is likely able to traffic across the BBB and BCSFB to infect neural cells⁹³.

5. Neural cell consequence after SARS-CoV-2 infection

While elucidating the mechanisms of how SARS-CoV-2 may enter the brain is highly important to our understanding of neurological/neuropsychiatric manifestations of SARS2, it is equally critical to understand the cellular consequences that result from the virus once in the brain, such as neural cell survival/death, proliferation, migration, virus production, viral spread, etc. Several studies report increased neuronal cell death in COs after SARS-CoV-2 infection as determined by TUNNEL staining and Caspase-3 immunostaining^{20, 41, 44, 45}. Histopathological examination of human brain autopsy specimens also characterized significant neuronal damage¹⁹. NSCs/NPCs also experience significant cell death^{20, 40, 44}. While less data is available for glial cells, such as astrocytes and microglia, if infected, they may produce various extent of immune/inflammatory responses or neuroprotective responses. Intriguingly, cells infected with SARS-CoV-2 may promote the death of uninfected neighboring cells^{20, 41}. Robust syncytial formation of infected cells found in the lung of SARS2 patients⁹⁶ may also occur to neural cells, being a source neuropathogenesis in the brain.

Transcriptomics analysis identified large scale transcriptional dysregulation, particularly in metabolic genes and cytokine storm genes as well as genes related to hypoxia response in SARS-CoV-2-infected cells or choroid plexus organoids^{20, 41, 44, 45}, confirming the local hypoxic environment in neurons surrounding SARS-CoV-2 infected cells via HIF1α expression. The upregulation of vascular remodeling genes and the downregulation of genes related to CSF secretory function as well as cell junction genes suggest potentially impaired function of the BCSFB and BBB.

Significantly decreased expression of vGLUT1, an excitatory pre-synapse marker, has been identified in human brain organoids, suggesting impaired synaptogenesis in SARS-CoV-2 infected cells^{20, 40, 44}. Another study identified Tau mislocalization in the neuronal soma of SARS-CoV-2-infected mature neurons as opposed to the typical localization within axons of normal neurons⁴⁵. Further investigation showed that the phospho-Tau mislocalization is specific to pT231Tau, which correlated with a fraction of cells that also expressed caspase-3. Thus, SARS-CoV-2 infection results in aberrant pT231Tau phosphorylation, which may trigger programmed cell death pathways. Intriguingly, the neuronal apoptosis induced by SARS-CoV-2 infection appears dependent on Type II interferon response (IFNγ), but not Type I interferon response (IFNα/β)^{20, 41, 44, 45}.

6. Conclusions and future directions

Accumulated evidence from clinical observations and laboratory studies helps approach to understanding the neuropathogenesis underlying SARS-CoV-2 infection causing various neurological and neuropsychiatric conditions in SARS2 patients. It remains to be determined whether SARS-CoV-2 directly or indirectly induces neurological/psychiatric damages. Studies have shown that many of SARS2 patients with an active neurological symptom had no detectable virus in the CSF. In addition, no virus could be recovered from the brain tissues of patients who died of SARS2. These observations suggest the possibility that indirect effects such as inflammation induced by SARS-Cov-2 infection contribute to the neurological/neuropsychiatric manifestations. However, failure to detect the virus may be due, at least partly, to low sensitivity of the assays; stages of infection and timing when the samples were examined. Notably, studies from different groups reported the significant differences in SARS-CoV-2 infection of various neural cells, which may be responsible for the conflicting data reported. Hopefully, with the availability of more advanced technologies such as RNAscope^{55, 56}, digital droplet PCR^57–59, viral outgrowth assay^{60, 61}, single cell RNA sequencing^{62, 63}, and multilabeled immunohistochemical analysis as well as better in vivo and in vitro models such as preclinical animals, human induced pluripotent stem cells, and human brain organoids for various neural cells, we will be able to determine the mechanisms for SARS-Cov-2 infection-caused neurological injury, which is crucial in developing strategies for treatment of SARS2 patients with the neurological/neuropsychiatric manifestations. Although several new studies using brain organoids offer strong and straightforward evidence for the neuroinvasive effect on different neural cells with particularly high susceptibility of choroid plexus epithelial cells, there are still limitations of current brain organoid technology to recapitulate clinical patients, including immaturity of neural cells, unclear dose of neurovirulent infection in vivo, and lack of additional cell types such as microglia, pericytes, endothelial cells, oligodendrocytes, and immune cells⁴¹. Given that the SARS2 pandemic remains raging and seasonal resurgence is impending, more extended efforts and investigation into Neuro-SARS2 mechanisms remain urgently needed. Long-term follow-up for potential neurological/neuropsychiatric manifestations as well as development of novel prophylactic and therapeutic treatments for Neuro-SARS2 are equally critical.

Abbreviations:

ACE2

angiotensin-converting enzyme-2

BBB

blood-brain barrier

BCSFB

blood-CSF barrier

CoV

coronavirus

COVID-19

coronavirus disease 2019

CNS

central nervous system

CO

cerebral organoid

CSF

cerebrospinal fluid

CVO

circumventricular organ

DRG

dorsal root ganglion

iPSC

induced pluripotent stem cell

K18

human cytokeratin 18

MERS

Middle East respiratory syndrome

MERS-CoV

Middle East respiratory syndrome coronavirus

Neuro-SARS

neurological/neuropsychiatric manifestation of severe acute respiratory syndrome

NPC

neural progenitor cell

NSC

neural stem cell

PNS

peripheral nervous system

RT-PCR

reverse transcription polymerase chain reaction

SARS

severe acute respiratory syndrome

SARS-CoV

severe acute respiratory syndrome coronavirus

TMPRSS2

transmembrane protease serine 2

ZIKV

Zika virus

Data availability statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.