Omicron BA.5 is lethal in K18-hACE2 mice
Infection of K18-hACE2 mice with original ancestral isolates of SARS-CoV-2 is well described and results in weight loss and mortality by ≈ 5 days post infection (dpi) 62,63,89−92. We re-illustrate this phenomena herein using an original ancestral isolate (SARS-CoV-2QLD02) and our K18-hACE2 mice, with the ethically defined end point of > 20% weight loss reached by 4–5 dpi (Fig. 1a, Original). An omicron BA.1 isolate (SARS-CoV-2QIMR01) was substantially less virulent, with only 20% of mice showing weight loss > 20% requiring euthanasia by 9/10 dpi (Fig. 1a, Omicron BA.1; Supplementary Fig. 1). The reduced pathogenicity of BA.1 isolates in K18-hACE2 mice is consistent with previous reports 70–72, 93.
Infection of K18-hACE2 mice with an omicron BA.5 virus (SARS-CoV-2QIMR03) resulted in severe weight loss requiring euthanasia in 100% of mice by 4–6 dpi (Fig. 1a, Omicron BA.5). BA.5 infected mice also showed more overt symptoms than BA.1, with original ancestral isolate showing slightly higher disease scores than BA.5 (Fig. 1b). Kaplan Meier plots illustrate a highly significant difference between BA.5 and BA.1 (Fig. 1b, p = 1.7x10− 9). Although mortality from BA.5 was significantly delayed when compared with the original ancestral isolate (p = 0.003), the mean delay was only 0.88 days (Fig. 1c). These results contrast with a recent publication reporting that the reduced pathogenicity of omicron sub lineages was retained for BA.5 78.
On 5 dpi there were no significant differences in viral titers in lungs and nasal turbinates between BA.1 and BA.5, whereas the original ancestral isolate had ≈ 2 logs higher lung titers and ≈ 4 logs higher nasal turbinate titers (Fig. 1d, p ≤ 0.034 Kruskal-Wallis tests). Brain titers were also ≈ 4 logs higher on 5 dpi for the original ancestral isolate when compared with BA.5, with BA.5 titers not significantly different from the BA.1 brain titers for the 3 mice that were euthanized due to weight loss 9/10 dpi (Fig. 1d, Brains). However, consistent with previous reports 68,94, the majority (80%) of BA.1 infected mice showed no symptoms, with no virus detected in the brains of symptom-free animals (Fig. 1d, Brains).
Immunohistochemistry Of Ba.5 Brain Infection In K18-hace2 Mice
The fulminant brain infection seen after infection of K18-hACE2 mice with original ancestral isolates is well described, with widespread infection of neurons in various brain regions, including the cortex 59,60,62,64,68. A similar pattern of brain infection was observed using our K18-hACE2 mice and an original ancestral isolate, with immunohistochemistry (IHC) undertaken using a recently developed anti-spike monoclonal antibody (SCV2-1E8) 64 (Supplementary Fig. 2a).
IHC staining of brains of K18-hACE2 mice infected with BA.5 also showed widespread infection of cells in the cortex, as well as the hippocampus and the hypothalamus (Fig. 2a,b). Viral RNA/protein has been detected in the cortex of post-mortem COVID-patients 40,41, with disruption of the hippocampus also reported 21,95. The hypothalamus does not appear to be the site of infection in COVID-19 patients 96, although the hypothalamic-pituitary axis does appear to be affected 97. Viral antigen was also clearly evident in dendrites and axons (likely neural) (Fig. 2b), with viral antigen staining in neurites previously shown in human brain organoids 98.
In the K18-hACE2 model, brain infection is generally associated with weight loss and mortality 99 (generally via euthanasia after reaching ethically defined end points). Perhaps of note, even low levels of IHC-detectable BA.5 infection was associated with weight loss that required euthanasia (Supplementary Fig. 2b), illustrating that such fatal outcomes do not require a fulminant brain infection.
Ba.5 Infects Neurons In K18-hace2 Mice
To confirm infection of neurons by BA.5, the cortex of infected K18-hACE2 mice were co-stained with the anti-spike monoclonal antibody and anti-NeuN, a neuronal nuclear antigen marker. Extensive co-localization within the same cells was observed (Fig. 3, Neurons) illustrating that neurons are a primary target of BA.5 infection in K18-hACE2 mouse brains. Co-staining with anti-spike and anti-Iba1, a pan-microglia marker, showed minimal overlap with anti-spike (Fig. 3, Microglia), arguing that microglia are not a major target of infection. Occasional overlap (yellow) may be due to phagocytosis of debris from virus infected cells. Despite being surrounded by infected neurons, most microglia retaining their ramified morphology, although some cells with bushy and amoeboid morphology were present (Fig. 3, Microglia) indicating activation-associated retraction of processes 100. Microglia activation was also indicated by histology and RNA-Seq (see below). Although occasionally seen (Supplementary Fig. 3a), anti-GFAP staining was minimal around infected neurons (Fig. 3, Reactive astrocytes), arguing that astrocytes are largely not being activated. RNA-Seq also did not identify GFAP as a significantly up-regulated gene, nor did bioinformatics identify an astrocyte activation signature (see below; Supplementary Table 1).
Lesions In The Brains Of Ba.5 Infected K18-hace2 Mice Identified By H&e Staining
The brains of BA.5 infected K18-hACE2 mice showed a number of lesions that have also been observed in COVID-19 patients and/or primate models. Neuron vacuolation (hydropic degeneration) was clearly evident (Fig. 4a), and has been reported previously for infection of K18-hACE2 mice with an original ancestral isolate 59, and was also observed in non-human primate (NHP) model of SARS-CoV-2 infection 35. The presence of viral antigen in the cortex was associated with apoptosis (Supplementary Fig. 3b) and a high intensity of H&E-detectable lesions (primarily vacuolation), but was not associated with local immune cell infiltrates (Supplementary Fig. 4). The lack of infiltrates around areas of infection has also been noted in COVID-19 patients 41. Perivascular cuffing (Fig. 4b) is well described in histological examinations of brains from deceased COVID-19 patients 42,43,66,101,102. Other lesions observed in the BA.5-infected K18-hACE2 mouse brains, that have also been described in post-mortem COVID-19 patients, include perivascular edema (Fig. 4c) 103–105, occasional microglial nodules (Fig. 4d) 42,66,102,106, and occasional small hemorrhagic lesions (Fig. 4e) 101,107.
Rna-seq Of Ba.5-infected K18-hace2 Mouse Brains
Mice were infected as in Fig. 1 (BA.5) and euthanized when weight loss reached the ethically defined endpoint of ≈ 20% (Supplementary Fig. 5a). Control mice received the same inoculation of UV-inactivated BA.5. Brains were examined by RNA-Seq (BioProject ID: PRJNA911424), with the PCA plot shown in Supplementary Fig. 5b and viral RNA levels in Supplementary Fig. 5c. Differentially expressed genes (DEGs) (q < 0.05, n = 437) were analyzed by Ingenuity Pathway Analysis (IPA) 56,63 (Supplementary Table 1). Selected representative IPA annotations, grouped by themes, are shown in Fig. 5a. The dominant annotations illustrate a cytokine storm, with the top cytokine Up Stream Regulators (USRs) including interferons both type II (IFNγ) and type I (IFNα2, IFNλ1 IFNβ1), as well as TNF, IL-1 and IL-6, all previously well described for SARS-CoV-2 infections 56. The concordance for cytokine USRs for brain and lung infection, and for the three virus isolates, was high (Supplementary Fig. 5d), arguing that inflammatory responses are generally very similar for brain and lungs and for the different SARS-CoV-2 variants.
A large series of annotations were associated with leukocyte migration and activation (Supplementary Table 1), with the top two overarching annotations shown (Fig. 5a, Leukocyte migration, Activation of leukocytes). These annotations are consistent with the perivascular cuffing seen by H&E (Fig. 4b). An additional series of neuropathology-associated annotations were also identified with high z-scores and significance (Fig. 5a, Neuropathology). Activation of microglia and vascular lesions (Fig. 5a) were consistent with the histological findings (Fig. 4b,c,d). Multiple sclerosis-like features, myelitis, demyelination 108,109 and encephalitis 9,26−28 are also features described for COVID-19 patients. Apoptosis of neurons was reported in the NHP model 35, with pyroptosis in the CNS of COVID-19 patients also proposed 110. Gene Set Enrichment Analyses (GSEAs) using gene sets provided in MSigDB (≈ 50,000 gene sets) and in Blood Transcription Modules, generated broadly comparable results to those obtained from IPA (Supplementary Table 1). In addition, a significant negative enrichment (negative NES) for olfactory neuroepithelium genes (MSigDB) 111 was also identified (Fig. 5a), suggesting loss of cells in this tissue in BA.5-infected mouse brains. COVID-19-associated anosmia (loss of smell) in humans is likely associated with infection of the olfactory epithelium 63.
To provide insights into the nature of the leukocyte infiltrates, cell type abundance estimates were obtained from the RNA-Seq expression data using SpatialDecon 112 (Fig. 5b). The inflammatory infiltrate appeared primarily to comprise immature CD4 T cells 113, macrophages, neutrophils, dendritic cells, CD8 T cells and NKT cells, with increased cell abundance scores seen with increasing viral RNA levels (Fig. 5b, TPM - transcripts per million; Supplementary Fig. 5e). Although not substantial, increased cell abundance scores also increased with viral RNA levels for microglia (Fig. 5c).
In summary, the bioinformatic analyses illustrate that the inflammatory responses in BA.5-infected K18-hACE2 mouse brains are largely innate (4–6 dpi) and typical of acute SARS-CoV-2 infections, with many annotations consistent with histological findings and a series of studies in COVID-19 patients.
Human Cortical Brain Organoids
Human, induced pluripotent cells (hiPSCs), derived from a primary dermal fibroblast line (HDFa) from a normal human adult, were used to generate approximately spherical, ≈ 2–3 mm diameter, “mini-brains” using a rotating incubator (Supplementary Fig. 6a). RNA-Seq and IHC illustrated that 30 day old organoids were comprised primarily of neural progenitor cells (expressing SOX2 and nestin) and immature neurons expressing MAP2 (Microtubule-Associated Protein 2) and TUBB3 (tubulin beta 3 ) (Supplementary Fig. 6b,c). Such organoids were infected with the BA.5, BA.1 and the original ancestral isolate (MOI ≈ 1) and were cultured for 4 days. Dual labelling fluorescent IHC illustrated that BA.5 infected MAP-2-negative cells, and some MAP2-positive cells (Supplementary Fig. 7). BA.5 infected substantially more cells in the organoids than the original ancestral (Fig. 6a) or the BA.1 viruses (Supplementary Fig. 8a). The small area infected with the original ancestral isolate (Fig. 6a, Original, insert) corresponded to an area of the organoid with IHC-detectable hACE2 staining (Supplementary Fig. 8b). Overall expression of hACE2 mRNA was low, with TMPRESS2 mRNA often undetectable (Supplementary Fig. 8c). Supernatants showed a significant increase in viral titers after BA.5 infection over the 4 day period, and virus titer in the supernatant was significantly higher for BA.5 compared to BA.1 at 4 dpi (Fig. 6b). RNA-Seq of organoids harvested 4 dpi also illustrated that viral RNA levels were ≈ 25 fold higher for organoids infected with BA.5 than those infected with an original ancestral isolate (Fig. 6c).
RNA-Seq of BA.5 infected human cortical brain organoids (4 dpi) compared with uninfected organoids provided 2390 DEGs (q < 0.001), of which 575 were up-regulated genes (Supplementary Table 2). RNA-Seq of original ancestral isolate-infected organoids provided 252 DEGs (q < 0.001), of which 132 were up-regulated (Supplementary Table 2). Given the higher level of infection, more DEGs might be expected for BA.5. Of the 132 up-regulated DEGs, 118 were also identified in the BA.5 infected organoids (Fig. 6d), arguing that the original ancestral isolate is not inducing fundamentally different response in these organoids.
The 2390 DEGs for BA.5 were analyzed by IPA (Supplementary Table 2), with “Coronavirus Pathogenesis Pathway” identified as a top canonical pathway (Fig. 6e). The top USRs were (i) PTPRR (a protein tyrosine phosphatase receptor), which was recently identified in a study of brains from SARS-CoV-2 infected hamsters and is associated with depression in humans 114, (ii) COPS5 (COP9 signalosome subunit 5), whose mRNA is bound by SARS-CoV-2 NSP9, perhaps resulting in suppression of host responses 115, (iii) LARP1, a translational repressor that binds SARS-CoV-2 RNA 116, (iv) ESR1 (nuclear estrogen receptor), which is important for ACE2 expression 117, (v) EGLIN, oxygen sensors that target HIF α subunits for degradation, with HIF-1α promoting SARS-CoV-2 infection and inflammation 118. IPA Diseases and Functions feature identified a series of neuropathology-associated annotations, including a motor dysfunction signature, with motor deficits documented for severe COVID-19 patients 20. Consistent with the IHC data, a series of signatures describe disruption and death of neurons (Fig. 6e). No significant up-regulation of classical inflammation or IFN signatures were identified, with the possible exception of oncostatin M (OSM) (Fig. 6e). Serum concentrations of this IL-6 family pleiotropic cytokine show a strong positive correlation with COVID-19 severity 119. However, OSM can also be secreted by neural cells, but in the brain it is thought often to play a neuroprotective role 120.