COVID-19 Neuropathology: Evidence for SARS-CoV-2 invasion of Human Brainstem Nuclei

Aron Emmi University of Padova Stefania Rizzo University of Padova Luisa Barzon University of Padova Elisa Carturan University of Padova Alessandro Sinigaglia University of Padova Silvia Riccetti University of Padova Mila Della Barbera University of Padova Rafael Boscolo Berto University of Padova Patrizia Cocco University of Padova Veronica Macchi University of Padova Monica De Gasperi University of Padova Cristina Basso University of Padova Raffaele De Caro (  rdecaro@unipd.it ) University of Padova https://orcid.org/0000-0002-2307-0277 Andrea Porzionato University of Padova

Symptoms range from anosmia, ageusia, dizziness and headache, which are commonly reported by patients with mild disease, to altered mental status, neuropsychiatric disorders, stroke, and, rarely, meningitis, encephalitis, and polyneuritis, which occur in hospitalized patients with severe disease 1,5 . Between 10 to 30% of people with SARS-CoV-2 infection experience long-term sequelae, referred as "long COVID", including neurological manifestations such as hyposmia, hypogeusia, headaches, fatigue, sleep disorders, pain, and cognitive impairment 3 . Despite some reports of detection of SARS-CoV-2 in the brain and cerebrospinal uid of patients with COVID-19 [2][3]6 , it is still unclear whether the virus can infect the central nervous system (CNS). In particular, it still remains to be elucidated whether neurological manifestations and neural damage are a direct consequence of viral invasion of the CNS, are due to post-infectious immune-mediated disease, or are the result of systemic disease 1,6-11 . Studies on human neural cell cultures and brain organoids report con icting data on SARS-CoV-2 neurotropism 12 . Overall, they suggest that SARS-CoV-2 does not infect and replicate e ciently in human neural cells, while it can replicate at high rates in choroid plexus epithelial cells 7,[13][14] . At variance, intranasal inoculation of SARS-CoV-2 in transgenic mice overexpressing human ACE2 under the K18 promoter resulted in brain invasion and widespread infection of neurons, radial glia and neuronal progenitor cells [15][16] . Other coronaviruses, such as SARS-CoV and MERS-CoV, appear to be able to infect the CNS in both humans and animal models 17 .
Data deriving from large autopsy studies in patients who died from COVID-19 suggest for the neuroinvasive potential of SARS-CoV-2 in the CNS 8-9,17 , even though infection appears to be limited to sparse cells in the brainstem and not associated with encephalitis or other speci c changes referable to the virus 8 . However, other studies failed to detect SARS-CoV-2 antigens or genomic sequences in brain tissues of COVID-19 patients 11,[17][18] . Neuropathological changes in the brain of COVID-19 patients are mild and mainly represented by ischaemic lesions, astrogliosis, microglial nodules, and cytotoxic T lymphocyte in ltrates, most pronounced in the brainstem, cerebellum, and meninges [8][9]11,18,21 . Deep spatial pro ling of the local immune response in COVID-19 brains through imaging mass spectrometry revealed signi cant immune activation in the CNS with pronounced neuropathological changes (astrocytosis, axonal damage, and blood-brain-barrier leakage) and detected viral antigen in ACE2-positive cells enriched in the vascular compartment 18 . Microglial nodules and the perivascular compartment represented COVID-19-speci c, microanatomic-immune niches 18 . Single-nucleus gene-expression pro ling of frontal cortex and choroid plexus tissues from severe COVID-19 patients showed broad perturbations with upregulation of genes involved in innate antiviral response and in ammation, microglia activation and neurodegeneration 20 . As evidenced by the above case series, and considering the different case reports available [22][23][24] , SARS-CoV-2 infection of CNS seems to be limited to isolated cells within the brainstem and cranial nerve axons of the lower medulla and have been reported in few cases of the various autopsy series, while widespread neuropathological sequelae (such as astrogliosis, microgliosis, lymphocyte in ltration, microvascular injury, brinogen leakage) have been documented in most examined specimen.
In the preset study, we assess the neuropathological changes of 24 patients who died following a diagnosis of SARS-CoV-2 infection in Italy during the COVID-19 pandemic (from March 2020 to May 2021) and 10 age-matched controls with comparable medical conditions.

Study design and Materials
Hospitalized patients who died following a diagnosis of SARS-CoV-2 infection in the Veneto Region, Italy, during the peak incidence of COVID-19 (from March 2020 to May 2021) were autopsied according to established COVID-19 infection security protocols. Inclusion criteria for the study were: a) diagnosis of SARS-CoV-2 infection con rmed by molecular testing of naso-pharyngeal swabs and b) high-quality brain tissue samples available for histopathological and immunohistochemical analysis. A total of 24 patients were included in the study.
10 age-matched control cases with similar general medical conditions, predating the COVID-19 pandemic in Italy, were included to compare for activated microglial density in the brainstem while also serving as controls for viral protein immunohistochemical staining.

Methods
Sampled brains were xed in 4% phosphate-buffered formalin solution following autopsy and subsequently sectioned for histopathological and immunohistochemical analysis. Samples of the cerebral cortex, basal ganglia, hippocampus, cerebellar cortex, deep cerebellar nuclei, choroid plexuses and meninges were obtained, while the brainstem was isolated at the level of the rostral extremity of the mesencephalon and extensively sampled in its whole cranio-caudal extent. The 12 cranial nerves, where available, including the olfactory bulb, tract and bifurcation, were also sampled.
Haematoxylin and eosin stain was employed for histopathological evaluation. Immunoperoxidase staining was performed on a Dako EnVision Anti-nucleocapsid and anti-spike antibodies were validated through SARS-CoV-2 infected Vero E6 cells and autopsy-derived lung tissue from SARS-CoV-2 infected patients as positive controls; non-infected cells and lung sections deriving from autopsy cases predating COVID-19 pandemic (2017) were used as negative controls (Supplementary Figure 1). Peroxidase reactions were repeated at least three times to ensure reaction consistency.
Real-time RT-PCR analyses were performed to detect SARS-CoV-2 genome sequences. Brie y, total RNA was puri ed from this selected material using a RecoverAll™ Total Nucleic Acid Isolation kit (Thermo Fisher Scienti c) following the manufacturer's instructions. One-step real-time RT-PCR assays targeting SARS-CoV-2 nucleocapsid (N) coding region were run on ABI 7900HT Sequence Detection Systems (Thermo Fisher Scienti c), as previously reported 25 .
Slides were examined by experienced neuropathologists and morphologists blind to patient clinical ndings. Disagreements were resolved by consensus. The degree of astrogliosis and microgial proliferation was classi ed using a four-tiered semi-quantitative approach for each evaluated section, while microglial activation was quanti ed by the means of digitally-assisted immunoreactivity quanti cation by experienced morphometrists.

Quanti cation of Activated Microglia
The degree of microgliosis was assessed through a digitally-assisted quanti cation approach at the level of the medulla, pons and mesencephalon. For each subject, standard sections passing through the area postrema (medulla), facial colliculus (pons) and red nucleus (midbrain) were processed for HLA-DR immunoperoxidase staining. Photomicrographs were acquired under a Leica DM4500B microscope (Leica Microsystems) connected to a Leica DFC320 high-resolution digital camera (Leica Microsystems) and a computer equipped with softwares for image acquisition (QWin, Leica Microsystems) and analysis (ImageJ). Immunoreactivies were evaluated according to morphology and counted manually within six counting elds ( elds of view, FOV) spanning across the dorsal-to-ventral axis of the sections; FOV structures and boundaries are summarized in Supplementary Table 1 for each level of sectioning. The number of immunoreactivites per mm 2 was calculated for each counting eld and assigned to one anatomical compartment (i.e. tegmentum, tectum and basis), based on their topography according to Mai and Paxinos 26 . Comparisons and statistical evaluations were conducted per individual FOV, anatomical compartment and level of section (medulla, pons, mesencephalon).

Statistical Analyses
Statistical analyses and visualizations were performed using GraphPad Prism 9. Differences in activated microglia counts (microglia / mm 2 ) between COVID-19 and control patients in Figures 2B, 3B and 4B, were analyzed by t tests with Welch's correction. Microglial density in the different FOV in Figures 2C, 3C, and 4C were determined by ordinary one-way ANOVA tests for each level of sectioning. Tukey multiple comparisons test was performed. Further statistical details for each plot can be found in the corresponding gure legend. Throughout the manuscript * indicates p < 0.05, ** p < 0.01, *** p < 0.001 and **** p < 0.0001.

Clinical Data
Twenty-four COVID-19 patients were included in the study. In all patients, SARS-CoV-2 RNA was detected by molecular testing in nasopharyngeal swabs. Eleven were females, while 13 were males. The mean age of the included subjects was 73±13.7 years. Most included subjects were affected by preexisting chronic medical conditions, such as hypertension (N=13, 7 females, 6 males). Eleven patients (7 female, 4 male) were affected by neurological or neurodegenerative disease prior to SARS-CoV-2 infection. Twenty-three patients were hospitalized prior to death. Patients were hospitalized for 14.5±11.3 days and died 1 to 34 days following admission. The available clinical data for our cohort is reported in Table 1.
Ten age-matched control subjects were included in the study. All patients were negative for SARS-CoV-2 infection or died prior to the COVID-19 pandemic in Italy. Two were female, while 8 were male. The mean age of included controls was 74±14 years. All patients presented chronic medical conditions including hypertension, diabetes, and hypercolesterolemia. Three patients died due to pneumonia. Three patients had a clinical diagnosis of dementia. The available clinical data for the control group are reported in Table 2.
Histopathological evaluation revealed haemorragic injury in four patients and small vessel thromboses in nine patients; thromboses were identi ed mainly at the level of the pons, deep cerebellar nuclei and cerebral cortex, with one patient presenting thromboses in multiple sites.
Enlarged perivascular and perineuronal spaces, indicative of mild to moderate edema, were present in 22 subjects and were more pronounced at the level of the brainstem and basal ganglia, with particular regard to the medullary tegmentum. Small vessels were congested in most subjects with moderate perivascular extravasation at the level of the medulla, pons and deep cerebellar nuclei in six cases. Fresh territorial ischaemic injuries were evident in ve patients. Old territorial ischaemic lesions were found in six subjects.
Variable degrees of astrogliosis were evident in all subjects in all assessed regions, but were more pronounced at the level of the medullary tegmentum, pons and substantia nigra ( Figure 2). Reactive Bergmann Glia was found in the cerebellar cortex of 5 patients; for the detailed assessment of astrogliosis within sampled regions, refer to Table 3.
Parenchymal and perivascular microglia appeared activated and with increased phagocytic activity, as testi ed by HLA-DR / CD68+ immunoreactivity, in 23 assessed subjects, with particular involvement of the brainstem and basal ganglia. Moderate to severe perivascular CD68+ macrophage in ltration was found in 23 subjects, while parenchymal macrophages were particularly evident at the level of the substantia nigra of 12 subjects. Microglial stars associated with perineuronal CD68+ and HLA-DR+ cells were suggestive of neuronophagia in 18 subjects and were identi ed at the level of the substantia nigra (N=14), dorsal motor nucleus of the vagus (N=12), medullary reticular formation (N=9), area postrema (N=6) and basal ganglia (N=5).

Microgliosis: Quanti cation and Distribution of Activated Microglia
Microgliosis was more pronounced within the medulla, pons and brainstem in COVID-19 patients and statistically signi cant differences (p<0.001; p<0.001; and p<0.0001, respectively) were found when compared to age-matched controls, as seen in Welch's corrected T-test plots in Figure 3B, 4B, 5B. The topographical distribution of activated microglial cells within the anatomical boundaries of the brainstem and its nuclei can be appreciated in Figure 3A, 4A and 5A.
At the level of the medulla of COVID-19 patients, single-way ANOVA of individual FOVs ( Figure 2C) revealed statistically signi cant differences (p<0.001) between FOVs located within the boundaries of the tegmentum, when compared to FOVs of the Medullary Pes; no differences were found between FOVs located within the same anatomical compartment. At the level of the Pons, no signi cant differences in activated microglial density were found between FOVs or anatomical compartments ( Figure 3C).
In the Mesencephalon, statistically signi cant differences were found when comparing FOVs of the Tegmentum (FOV1, FOV2) to FOVs of the Tectum (FOV3,FOV4) and Pes (FOV5,FOV6) as seen in Figure 4C. Furthermore, statistically signi cant differences were found between FOV3 and FOV4, with the latter displaying higher activated microglial counts (p<0.001), suggesting an increasing dorsal-to-ventral gradient of microgliosis within the structure.

SARS-CoV-2 Tropism
Immunoperoxidase staining for SARS-CoV-2 spike protein (Spike Subunit 1) and nucleocapsid protein was performed on all samples of included subjects and controls and showed positive results in cases with SARS-CoV-2 infection, but not in controls. In particular, viral proteins were detected in four subjects (#1-4) at the level of the cerebellar meninges, in seven subjects (#3, #7, #9, #10, #11, #17, #18) within CNS parenchima, in ve subjects (#3, #7, #9, #10, #17) with immunoreactive neurons within the anatomically de ned boundaries of the solitary tract nucleus, nucleus ambiguus and substantia nigra. Some of these subjects (#7, #11, #17, #18) also displayed endothelial cell immunoreactivity in small vessels of the cerebral cortex (subject #11), deep cerebellar nuclei (#17-18) and hippocampus (#7) (Supplementary Figure 2); small vessel thromboses, perivascular extravasation and haemorragic injury were found within affected regions of these cases. In case #7, ischaemic injury of the right rostral hippocampal formation due PCA occlusion was associated to perivascular extravasation, oedema, brinogen leakage and viral protein immunoreactivity within small vessel endothelium, further con rmed by RT-PCR. Haemorragic injury in the territory of the right MCA in #11 was associated to marked endothelitis within perilesional tissue, presenting both viral protein immunoreactive endothelium and positive RT-PCR. Similarly, the deep cerebellar white matter and dentate nuclei in cases #17-18 presented small vessel thromboses and extensive haemorragic injury. Conversely, in some cases with small vessel thromboses within the pons and frontal cortex (e.g. #19-20), viral proteins and RNA was not detectable. The distribution and topography of SARS-CoV-2 protein immunoreactivities is summarized in Figure 6A-D. Histopathological evaluation for each subject and in each assessed region is reported in Table 3.
Molecular testing by real-time RT-PCR detected SARS-CoV-2 RNA in 10 out of 24 COVID-19 subjects, 9 of whom had also SARS-CoV-2 S and/or N protein-positive IHC (Table 3

Discussion
In the present study, the neuropathological ndings of 24 COVID-19 patients were examined and compared with age matched-controls with comparable medical conditions. Our ndings indicate, in line with some of the previous autopsy reports, speci c neuropathological alterations in the brains of COVID-19 patients, with particular regard to topographically-de ned microgliosis within anatomical compartments of the brainstem and viral immunoreactivity in speci c loci of the CNS, either within the boundaries of brainstem nuclei or in the context of ischaemic and haemorragic injuries. Platelet and brin microthrombi, in particular, were characteristic ndings of the COVID-19 cohort, and often affected multiple organs, such as the lungs, liver, intestine, and hypopharynx and even the carotid body 10,27−28 , as summarized in Table 1.
Microthromboses were more frequent within the pons, deep cerebellar nuclei and cerebral cortex. In some cases, haemorragic injury and microthromboses were found in regions with viral protein immunoreactivity in vascular endothelial cells.
SARS-CoV-2 neuronal tropism, on the other hand, was con ned to speci c loci of the CNS. As seen in Figure 6A While previous studies identi ed viral protein immunoreactivity in sparse cells throughout the brainstem 8-18 , our ndings appear to be in line with available animal studies on other coronaviruses, i.e. SARS-CoV and MERS-CoV, which are known to be able to infect the brainstem, and particularly the solitary tract nucleus and nucleus ambiguus, so that an analog pattern of neuroinvasion for SARS-CoV-2 has been suggested 17,29−31 . Internalization of SARS-CoV-2 is known to cause inhibition of ACE-2 activity and progressive depletion of membrane-bound ACE-2 30-31 , with subsequent ACE1/ACE2 imbalance and increase in Angiotensin II (AngII). Circulating AngII may in turn increase the sympathetic output both centrally, at the level of the circumventricular organs (area postrema and subfornical organ) 32 , and peripherally, by acting on the carotid body 10,30−32 . As ACE-2 is also expressed in the solitary tract nucleus, sympathetic activation may be furtherly increased by local ACE1/ACE2 imbalance and AngII stimulation. Thus, COVID-19-induced increase in AngII may represent an additional way to furtherly worsen sympathoactivation, which may exert signi cant detrimental effect through its actions on lungs, heart, vessels, kidney, metabolism, and/or immune system, representing a so-far undervalued mechanism at the basis of the vicious circle between COVID-19 and known comorbidities [30][31] .
However, while the hypothesized effects of SARS-CoV-2 invasion of the CNS remain to be investigated, post-mortem evidence of direct viral invasion in humans, with analog topographical distribution to animal models, represents a relevant step towards the elucidation of COVID-19 pathophysiology.
As for HLA-DR reactive activated microglia, we found signi cant differences between COVID-19 patients and controls in all assessed levels of the brainstem, however the COVID-19 group was also characterized by higher density of activated microglia within speci c anatomical loci, such as the medullary and mesencephalic tegmentum. Our ndings appear to be in line with Schwabenland et al. 18 , who identi ed microglial nodules and perivascular HLA-DR+ reactive microglia as hallmark for COVID-19, in contrast to both controls and ExtraCorporeal Membrane Oxygenation (ECMO) patients. Conversely, Deigendesh et al. 33 found signi cant differences in HLA-DR+ activated microglia when comparing COVID-19 subjects to non-septic controls, but no differences were found with patients who had died under septic conditions; according to Deigendesh et al. 33 this may represent a histopathological correlate of critical illness-related encephalopathy, rather than a COVID-19-speci c nding. In our study, the evidenced pattern of microgliosis appears to match with the distribution of SARS-CoV-2 immunoreactivities, localized mainly within the vagal complex and the substantia nigra. However, when comparing COVID-19 patients with and without viral immunoreactivity, no statistically signi cant differences in the overall degree of brainstem microgliosis were found, as seen in Figure 5. Hence, microglial activation does not appear to be directly related to neuronal invasion of SARS-CoV-2 within affected regions, but could represent the consequence of systemic infection / cytokine storm ongoing during COVID-19 affecting topographic compartments of the brainstem with an intrinsic anatomical vulnerability, such as the medullary tegmentum, and the substantia nigra [34][35] . Furthermore, COVID-19 is characterized by different evolutionary phases and heterogeneous individual responses, and the short interval between infection and death in our cohort (mean hospitalization time = 14 days) and the fact that included patients died during the acute phase of the disease, may not be su cient to determine detectable neuropathological alterations as a direct consequence of viral invasion, which may require more time to develop 3,31 . In this perspective, future studies on "long COVID" patients 3 may be able to shed a light on the long-term consequences of COVID-19, particularly concerning the detection of SARS-CoV-2 within the CNS after the acute phase of the disease, and whether or not this leads to speci c neuropathological alterations as a consequence of viral invasion.

Conclusions
The present study contributes to de ne the spectrum of neuropathological alterations in COVID-19, as well as the neuroinvasive potential of SARS-CoV-2 within the CNS. Unlike previous ndings, we have documented several cases in which viral proteins and RNA were clearly detectable within anatomically de ned regions of the CNS. Similarly, microglial activation in the brainstem seems to be signi cantly different between COVID-19 and controls, with the former also presenting a pattern of increased microglial density in speci c compartments of the medulla and midbrain.
In line with available literature, however, SARS-CoV-2 direct invasion does not appear to directly correlate with the severity of neuropathological changes, such as microglial activation. Hence, the spectrum of neuropathological alterations described could be ascribed to systemic infection, rather than direct viral invasion, and require further con rmation from other studies.

Declarations
Data Sharing: data is available from the corresponding author upon request.
Con icts of interest: the authors declare no con icts of interest.
Ethical approval: All procedures were carried out in accordance to the Declaration of Helsinki. Samples were anonymous to the investigators and used in accordance with the directives of the Committee of the Ministers of EU member states on the use of samples of human origin for research.       A) Anatomical heatmap of activated microglia within the mesencephalon. B) Welch's corrected T-Test plot of manually-counted activated microglia within the mesencephalon in COVID-19 subjects (red) vs controls (green), revealing statistically signi cant differences (p<0.0001) between groups. C) one-way ANOVA of activated microglial densities per Field of View (FOV) reveals statistically signi cant differences between FOVs of the Tegmentum, Tectum and Pes, with higher densities of activated microglia fund within the Tegmentum. D-E) Activated microglia at the level of the Substantia Nigra, CD68 immunohistochemistry. Scale Bars: 100µm.