SARS-CoV-2 omicron BA.5 and XBB variants have increased neurotropic potential over BA.1 in K18-hACE2 mice and human brain organoids

The reduced pathogenicity of the omicron BA.1 sub-lineage compared to earlier variants is well described, although whether such attenuation is retained for later variants like BA.5 and XBB remains controversial. We show that BA.5 and XBB isolates were significantly more pathogenic in K18-hACE2 mice than a BA.1 isolate, showing increased neurotropic potential, resulting in fulminant brain infection and mortality, similar to that seen for original ancestral isolates. BA.5 also infected human cortical brain organoids to a greater extent than the BA.1 and original ancestral isolates. In the brains of mice, neurons were the main target of infection, and in human organoids neuronal progenitor cells and immature neurons were infected. The results herein suggest that evolving omicron variants may have increasing neurotropic potential.

Herein we characterize BA.5 and XBB infection in model systems that allow robust investigation of neurotropic potential.We illustrate and characterize the increased levels of brain infection, pathology and mortality for BA.5 and XBB versus BA.1 infected K18-hACE2 mice.Infection of brain organoid systems with SARS-CoV-2 is well described (Song et al., 2021;Hou et al., 2022;Mesci et al., 2022), and we show herein that BA.5 productively infected human cortical brain organoids significantly better than BA.1.These results indicate that BA.5 and XBB may have increased neurotropic potential over BA.1.

Ethics statements and regulatory compliance
All mouse work was conducted in accordance with the Australian code for the care and use of animals for scientific purposes (National Health and Medical Research Council, Australia).Mouse work was approved by the QIMR Berghofer MRI Animal Ethics Committee (P3600).All infectious SARS-CoV-2 work was conducted in the BioSafety Level 3 (PC3) facility at the QIMR Berghofer MRI (Department of Agriculture, Fisheries and Forestry, certification Q2326 and Office of the Gene Technology Regulator certification 3,445).Breeding and use of GM mice was approved under a Notifiable Low Risk Dealing (NLRD) Identifier: NLRD_Suhrbier_Oct2020: NLRD 1.1(a).Mice were euthanized using carbon dioxide.

Mouse infection and monitoring
K18-hACE2 mice (strain B6.Cg-Tg(K18-ACE2)2Prlmn/J, JAX Stock No: 034860) were purchased from The Jackson Laboratory (2023), United States, and were maintained in-house as heterozygotes by backcrossing to C57BL/6 J mice (Animal Resources Center, Canning Vale WA, Australia) as described (Bishop et al., 2022).Heterozygotes were inter-crossed to generate a homozygous K18-hACE2 transgenic mouse line.Genotyping was undertaken by PCR and sequencing across a SNP that associates with the hACE2 transgene to distinguish heterozygotes [TTTG(A/C)AAAC] from homozygotes (TTTGCAAAC).The mice were held under standard animal house conditions [for details see (Yan et al., 2022)] and female mice (≈10-20 weeks of age) received intrapulmonary infections delivered via the intranasal route with 5 × 10 4 CCID 50 of virus in 50 μL RPMI 1640 while under light anesthesia as described (Dumenil et al., 2022).Each group of mice within an experiment had a similar age range and distribution, with the mean age for each group not differing by more than 1 week.Mice were weighed and overt disease symptoms scored as described (Dumenil et al., 2022).Mice were euthanized using CO 2 , and tissue titers determined using CCID 50 assays and Vero E6 cells as described (Rawle et al., 2021;Dumenil et al., 2022).

Maintenance and expansion of human induced pluripotent stem cells
The human-induced pluripotent cells (hiPSCs) used in this study were reprogrammed from adult dermal fibroblasts (HDFa, Gibco, C0135C) using the CytoTune-iPS 2.0 Sendai Reprogramming Kit (Invitrogen, A16518; Oikari et al., 2020).They were cultured on human recombinant vitronectin (Thermo Fisher Scientific) coated plates in StemFlex medium (Thermo Fisher Scientific) according to the manufacturer's guidelines.

Infection of cortical brain organoids
Organoids (30 days old) were transferred from each Clinoreactor into an ultra-low-binding 24-well plate (one organoid per well), infected with various SARS-CoV-2 strains at a multiplicity of infection (MOI) of 1 and placed within a humidified tissue culture incubator at 37C, 5% CO 2 for 2 h.Organoids were then washed twice with media, transferred into 50 mm LUMOX gas exchange dishes (SARSTEDT) (4 organoids per dish) containing 7 mL of differentiation media, and placed within a humidified tissue culture incubator at 37°C, 5% CO 2 for 4 days.Supernatants were collected at the indicated days and titers measured by CCID 50 assays as described above.At 4 dpi organoids were harvested and formalin fixed for histology and immunohistochemistry, or processed for RNA-Seq.

Histology and immunohistochemistry
H&E staining of formalin fixed paraffin wax embedded tissue sections was undertaken as described (Amarilla et al., 2021;Rawle et al., 2021).Immunohistochemistry was undertaken as described using the in-house developed anti-spike monoclonal antibody, SCV-1E8 (Morgan et al., 2023).
Sections were imaged using an Olympus UPLXAPO 10x/0.4NA air objective, 20x/0.8NA air objective and an UPLXAPO 60x/1.42NA oil-immersion objective mounted on a spinning disk confocal microscope (SpinSR10; Olympus, Japan) built on an Olympus IX3 body equipped with two ORCA-Fusion BT sCMOS cameras (Hamamatsu Photonics K.K., Japan) and controlled by Olympus cellSens software.Images were acquired as 3D Z-stack tile images and were deconvolved using Huygens Professional Deconvolution Software (Scientific Volume Imaging, Netherlands).

RNA-Seq and bioinformatics
RNA-Seq was undertaken as described using Illumina Nextseq 550 platform generating 75 bp paired end reads (Rawle et al., 2021;Bishop et al., 2022).The per base sequence quality for >90% bases was above Q30 for all samples.Mouse RNA-Seq reads were aligned to a combined mouse (GRCm39, version M27) and SARS-CoV-2 (Wuhan, NC_045512.2) reference genome using STAR aligner.Organoid RNA-Seq reads were aligned in the same manner except that the human (GRCh38, version 38) reference genome was used.Read counts for host genes and SARS-CoV-2 genomes were generated using RSEM v1.3.1, and differentially expressed genes were determined using EdgeR v3.36.0.To avoid missing type I IFN genes, which have low read counts (Wilson et al., 2017), a low filter of row sum normalized read count >1 was used.
IPA USR cytokine signatures obtained from BA.5-infected K18-hACE2 mouse brains were compared to gene expression data from two studies on COVID-19 patient brains (Fullard et al., 2021;Yang et al., 2021).For the Yang and Fullard studies there were 20 and 45 gene expression data sets, respectively, for the different tissues and cell-types.DEGs sets (n = 20 and 45) were derived by applying a q < 0.05 filter.These DEG sets were then concatenated to generate a single DEG list for each of the two studies.When a gene was present in more than one DEG set, the mean of the fold changes was used for the concatenated DEG list.IPA USR analysis was then performed as described above.

Statistics
Statistical analyzes of experimental data were performed using IBM SPSS Statistics for Windows, Version 19.0 (IBM Corp., Armonk, NY, United States).The t-test was used when the difference in variances was <4, skewness was > − 2 and kurtosis was <2.For non-parametric data the Kolmogorov-Smirnov test was used.
Kaplan Meier plots illustrate a highly significant difference between BA.1 vs. BA.5 and BA.1 vs. XBB, with no significant difference between BA.5 and XBB (Figure 1C).Mortality from BA.5 and XBB was delayed when compared with the original ancestral isolate, although the mean delay was ≤2 days (Figure 1C).The results for BA.5 contrast with a recent publication reporting that the reduced pathogenicity of early omicron sub lineages was retained for BA.5 (Uraki et al., 2022).
The brain tissue titers were determined for all mice that reach ethically defined disease endpoints for euthanasia (Figure 1D); there are only 3 data points for BA.1 as 80% of these mice did not reach ethically defined endpoints for euthanasia (Figure 1C), with surviving mice showing no detectable brain infections (Supplementary Figure S1C).Virus was detected in all brain samples, with two exceptions (Figure 1D, indicated by *).Of these, the BA.5infected mouse brain was subsequently found to be positive by IHC (the XBB-infected mouse was not tested).The brain titers in mice † mice reached ethically defined end points for euthanasia.(B) Disease scores for the indicated overt disease symptoms for animals described in "A".For BA.1 no overt symptoms were seen except in a single animal on 9 dpi, with this animal one of the 3 that required euthanasia due to weight loss.For BA.5 the remaining mouse (n = 1) on 6 dpi showed a Posture score of 2. For XBB the one surviving mouse recovered after 7 dpi.(C) Kaplan Meyer plot showing percent survival; n = 12 for original, n = 12 for XBB, n = 18 for BA.5 and n = 24 for BA.1 (data from 2 to 3 independent experiments).Statistics by log-rank tests comparing survival rates.(D) Viral tissue titers in brains of mice that had reached ethically defined end points for euthanasia (note that this was at a range of different dpi, see Supplementary Figure S1C).All mice with weight loss requiring euthanasia had detectable viral titers in brain (determined by CCID 50 assays), with two exceptions (*); although the BA.5 infected mouse emerged to be IHC positive (the XBB infected mice was not tested).Statistics by Kolmogorov-Smirnov tests; titers for the 3 omicron variants were not significantly different from each other.ND -not detected (limit of detection ≈ 2 log 10 CCID 50 /g).(Data from 5 independent experiments).
Frontiers in Microbiology 06 frontiersin.orginfected with the original strain isolate were significantly higher by ≈ 2 logs than the brain titers in mice infected with the omicron isolates (Figure 1D).Lung titers are shown in Supplementary Figure S1B and are lower in omicron infected mice compared to the original strain isolate, consistent with previous reports (Armando et al., 2022;Natekar et al., 2022;Lee et al., 2023;Ying et al., 2023).Lethality (generally associated with weight loss requiring euthanasia) in the K18-hACE2 model was previously associated with brain infections (Carossino et al., 2022;Fumagalli et al., 2022), an observation that would thus appear to remain largely true for omicron isolates (Figure 1D).Mice with no detectable brain titers did not reach ethically defined end points for euthanasia (Supplementary Figure S1C).However, our observations are in direct contrast to a previous report showing lethal BA.5 infections in K18-hACE2 mice in the absence of brain infection (Imbiakha et al., 2023).It is perhaps worth noting that in our hands C57BL/6 J mice infected with BA.1, BA.5 or XBB (Shuai et al., 2023) show 100% survival (Supplementary Figure S1D), despite robust lung infections (Shuai et al., 2023).We have not observed overt brain infection in these mice, arguing that C57BL/6 J do not provide a robust neuroinvasion model.Although age has been associated with lethal BA.5 infection in K18-hACE, we saw no significant correlation between age of mouse and lethality (Supplementary Figure S1E).

Immunohistochemistry of BA.5 and XBB brain infection in K18-hACE2 mice
Given the robust neuroinvasion seen for BA.5 and XBB that was not evident for BA.1 in K18-hACE2 mice, we sought comprehensively to characterize the brain infection and pathology for these emerging variants in this mouse model.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 (Carossino et al., 2022;Rothan et al., 2022;Seehusen et al., 2022;Vidal et al., 2022;Morgan et al., 2023).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) (Morgan et al., 2023;Supplementary Figure S2A).
IHC staining of brains of K18-hACE2 mice infected with BA.5 or XBB also showed widespread infection of cells in the cortex, as well as the hippocampus and the hypothalamus (Figures 2A-C).Viral RNA and protein have been detected in the cortex (Song et al., 2021;Shen et al., 2022) and hypothalamus (Stein et al., 2022) of post-mortem COVIDpatients.Viral protein has also been identified in the hippocampus of such patients (Emmi et al., 2023), with disruption of the hippocampus also reported (Douaud et al., 2022;Radhakrishnan and Kandasamy, 2022).In the hippocampus of K18-hACE2 infected mice, viral antigen could also be clearly seen in dendrites and axons (likely neural) (Figures 2B,C, right hand panels), with viral antigen staining in neurites previously shown in human brain organoids (Bullen et al., 2020).
As described above, brain infection was generally associated with weight loss and mortality in the K18-hACE2 model.Perhaps of note, even the low level of IHC-detectable SARS-CoV-2 infection (Supplementary Figure S2B) seen in one BA.5-infectedmouse (Figure 1D, *BA.5) was associated with weight loss that required euthanasia, suggesting that fatal outcomes may not always 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 (Figure 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 (Figure 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 (Martínez-Mármol et al., 2023).Despite being surrounded by infected neurons, most microglia retained their ramified morphology, although some cells with bushy and amoeboid morphology were present (Figure 3, Microglia) indicating activation-associated retraction of processes (Pinto and Fernandes, 2020).Microglia activation was also indicated by histology and RNA-Seq (see below).Although occasionally seen (Supplementary Figure S3A), anti-GFAP staining was minimal around infected neurons (Figure 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 S2).The apparent lack of astrocyte activation despite a fulminant SARS-CoV-2 brain infection (at least in mice), with their antiviral and neuroprotective activities thus presumably largely absent, is perhaps perplexing, but is consistent with SARS-CoV-1 studies (Netland et al., 2008).
A large series of annotations were associated with leukocyte migration and activation (Supplementary Table S2), with the top two overarching annotations shown (Figure 5A, Leukocyte migration, Activation of leukocytes).These annotations are consistent with the perivascular cuffing seen by H&E (Figure 4B).An additional series of neuropathology-associated annotations were also identified with high z-scores and significance (Figure 5A, Neuropathology).Activation of microglia and vascular lesions (Figure 5A) were consistent with the histological findings (Figures 4B-D).Apoptosis of neurons was reported in the NHP model (Rutkai et al., 2022), with pyroptosis in the CNS of COVID-19 patients also proposed (Sepehrinezhad et al., 2021).Gene Set Enrichment Analyzes (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 S2).In addition, a significant negative enrichment (negative NES) for olfactory neuroepithelium genes (MSigDB) (Durante et al., 2020) was also identified (Figure 5A), suggesting loss of cells in this tissue in BA.5-infected mouse brains.Infection of the olfactory epithelium likely provides the entry route into the brain in this model (Dumenil et al., 2022;Fumagalli et al., 2022).
To provide insights into the nature of the leukocyte infiltrates, cell type abundance estimates were obtained from the RNA-Seq expression data using SpatialDecon (Danaher et al., 2022;Figure 5B).The inflammatory infiltrate appeared primarily to comprise immature CD4 T cells (Hosseinzadeh and Goldschneider, 1993), macrophages, neutrophils, dendritic cells, CD8 T cells and NKT cells, with increased cell abundance scores seen with increasing viral RNA levels (Figure 5B,   TPM -transcripts per million; Supplementary Figure S5E).Although not substantial, increased cell abundance scores also increased with viral RNA levels for microglia (Figure 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.

Some concordance in cytokine gene expression patterns between brains of severe COVID-19 patients and brains from SARS-CoV-2 infected K18-hACE2 mice
We previously illustrated that inflammatory pathways identified by RNA-Seq of lungs from COVID-19 patients showed highly significant concordances with SARS-CoV-2 infected lungs from K18-hACE2 mice (Bishop et al., 2022).Two single-cell RNA-Seq data sets are publically available for selected human brain tissues (choroid plexus, medulla oblongata, and pre-frontal cortex) from deceased COVID-19 patients (Fullard et al., 2021;Yang et al., 2021).DEG sets from each tissue and cell-type were concatenated to create one overall DEG list for each of the two human studies.These DEG lists were analyzed by IPA as above, and the cytokine USR z-scores compared with those obtained from brains of BA.5-infected K18-hACE2 mice.Significant correlations emerged for both human studies (Figure 5D), with many of the prominent cytokines associated with SARS-CoV-2 infections (Bishop et al., 2022) identified in both species (Figure 5D, cytokines shown in red text).Overall, 80% of the cytokine USRs identified in humans also identified in K18-hACE2 mice (Figure 5E).However, the fulminant lethal brain infection likely explains the higher number of USRs identified for brains of K18-hACE2 mice (65 out of 97) that are not seen in COVID-19 patient brains (Figure 5E).
Thus both at the pathway level and the gene expression level, a level of concordance was apparent between cytokine mRNA expression data from (i) brain tissues of severe COVID-19 patients and (ii) brains of BA.5-infected K18-hACE2 mice.

Infection of 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 Figure S7A).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 Figures S7B,C).Such organoids were infected with the XBB, BA.5, BA.1 and the original ancestral isolate (MOI ≈ 1) and were cultured for 4 days.Dual labeling fluorescent IHC illustrated that BA.5 infected MAP2-negative cells, and some MAP2-positive cells (Supplementary Figure S8).The BA.5 virus infected substantially more cells in the organoids than the original ancestral (Figure 6A) or the BA.1 viruses (Supplementary Figure S9A).XBB also infected slightly more cells than BA.1 (Supplementary Figure S9B).The small area infected with the original ancestral isolate (Figure 6A, Original, insert) corresponded to an area of the organoid with IHC-detectable anti-hACE2 staining (Supplementary Figure S9C).The overall expression of hACE2 mRNA was low, with transmembrane protease serine 2 (TMPRSS2) mRNA often undetectable (Supplementary Figure S9D).Viral titers in the supernatants of the organoid cultures increased over the 4 day period, with BA.5 titers significantly higher than BA.1 titers by 1-2 logs (Figure 6B, p = 0.007, 2, 3 and 4 dpi).XBB titers were also up to ≈1 log higher, which reached significance if data from 4 and 5 dpi were combined (Figure 6B, p = 0.009, 4 & 5 dpi).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 (Figure 6C).
RNA-Seq of BA.5 infected human cortical brain organoids (4 dpi) compared with uninfected organoids provided 2,390 DEGs (q < 0.001), of which 575 were up-regulated genes (Supplementary Table S3).RNA-Seq of original ancestral isolate-infected organoids provided 252 DEGs (q < 0.001), of which 132 were up-regulated (Supplementary Table S3).Of the 132 up-regulated DEGs, 118 were also identified in the BA.5 infected organoids (Figure 6D), arguing that the original ancestral isolate is not inducing fundamentally different response in these organoids, and that the increased number of DEGs for BA.5 is likely due to more robust infection.
The 2,390 DEGs for BA.5 were analyzed by IPA (Supplementary Table S3), with "Coronavirus Pathogenesis Pathway" identified as a top canonical pathway (Figure 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 (Serafini et al., 2022), (ii) COPS5 (COP9 signalosome subunit 5), whose mRNA is bound by SARS-CoV-2 NSP9, perhaps resulting in suppression of host responses (Banerjee et al., 2020), (iii) LARP1, a translational repressor that binds SARS-CoV-2 RNA (Schmidt et al., 2021), (iv) ESR1 (nuclear estrogen receptor), which is important for ACE2 expression (Oner et al., 2022), (v) EGLN1, oxygen sensors that target HIF α subunits for degradation, with HIF-1α promoting SARS-CoV-2 infection and inflammation (Tian et al., 2021).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 (Graham et al., 2022).Consistent with the IHC data, a series of signatures describe disruption and death of neurons (Figure 6E).No significant up-regulation of classical inflammation or IFN signatures were identified, with the possible exception of oncostatin M (OSM) (Figure 6E).Serum concentrations of this IL-6 family pleiotropic cytokine show a strong positive correlation with COVID-19 severity (Arunachalam et al., 2020).However, OSM can also be secreted by neural cells, but in the brain it is thought often to play a neuroprotective role (Houben et al., 2019).

Discussion
We illustrate herein that the BA.5 and XBB variants show greater propensities to enter the brain and infect neurons in K18-hACE2 mice when compared to BA.1.In addition, BA.5 showed an increased capacity to infect human brain organoids.Taken together these results argue that these two more recent omicron variants of concern may have enhanced neurotropic properties when compared to an earlier omicron variant in these models.
The increased infection of brains by BA.5 and XBB over BA.1 in K18-hACE2 mice may be associated with the enhanced fusion activity of the later omicron variants (Tamura et al., 2023;Tang et al., 2023), which is usually associated with an enhanced ability to utilize TMPRSS2 and/or increased binding affinity for ACE2 (Aggarwal  S3)., 2022;Tamura et al., 2023).TMPRSS2 utilization is associated with virulence (Abbasi et al., 2021), even for omicron variants (Iwata-Yoshikawa et al., 2022), and is involved in neurotropism in K18-hACE2 mice (Li et al., 2021).An increased ability to infect, not just the TMPRSS2-positive sustentacular cells, but also TMPRSS2-low cells in the murine olfactory epithelium (Fodoulian et al., 2020), may thus promote entry of BA.5 and XBB into the brains of K18-hACE2 mice when compared with BA.1.In contrast to early omicron variants (Meng et al., 2022;Zhao et al., 2022;Qu et al., 2023), original ancestral isolates show a preference for TMPRSS2 utilization, and rapid fulminant infection of K18-hACE2 brains by such viruses is well described (Rothan et al., 2022;Seehusen et al., 2022).Interestingly, viral sequences from the brains of K18-hACE2 mice and hamsters infected with original ancestral isolates show loss of functional furin cleavage sites (Supplementary Figure S10).TMPRSS2 mRNA expression levels in K18-hACE2 mouse brains are very low (Supplementary Table S2), so TMPRSS2-independent infection (Qu et al., 2023) would likely be selected as the virus spreads within the brains.Such furin cleavage site deletions were not seen in BA.1 or BA.5 sequences from brains of K18-hACE2 mice, likely because omicron viruses can already effectively use the endosomal pathway (Meng et al., 2022;Zhao et al., 2022;Qu et al., 2023).In summary, more efficient use of TMPRSS2-dependent infection by BA.5 and XBB (and original ancestral isolates) compared to BA.1, may promote entry into the brain of K18-hACE2 mice via the olfactory epithelium.Once in the brain, utilization of the endosomal pathway by omicron viruses (and original ancestral isolates with non-functional furin cleavage sites) allows a spreading infection in TMPRSS2-low brain cells (Figure 1D, Brains).Infection of brain organoids represents a measure of neurovirulence, rather than neuroinvasiveness, as access to cells in this in vitro system clearly does not require transit across the cribriform plate (Jiao et al., 2021;Dumenil et al., 2022;De Melo et al., 2023).TMPRSS2 mRNA expression was even lower in the organoids (Supplementary Figure S9D) than in K18-hACE2 brains, with infection of human neurons shown to be TMPRSS2-independent (Kettunen et al., 2023).This is consistent with the poor infection of organoids by the original ancestral isolate (Figures 6A-C).BA.5 would thus appear to have an increased capacity for infection of brain organoids via a TMPRSS2-independent mechanism.This is not due to acquisition of hACE2-independent infection capabilities (Yan et al., 2022;Supplementary Figure S11).Nor is this likely due to an increased ability of BA.5 to counter type I IFN activities (Guo et al., 2022), as such responses were not detected (Figure 6E).hACE2 expression is low in the organoids, suggesting an increased affinity for hACE2 might promote BA.5 infection, with barely detectable levels of hACE2 able to support infection (Rawle et al., 2021).There are a number of differences in the spike protein between the BA.1 and BA.5 isolates including 9 amino acid changes in the receptor binding domain (Supplementary Figure S12), with BA.5 affinity for hACE2 reported as slightly higher in two studies (Tuekprakhon et al., 2022;Wang et al., 2022), but unchanged in a third (Cao et al., 2022).More efficient use by BA.5/XBB of co-receptors such as neuropilin (Cantuti-Castelvetri et al., 2020;Kong et al., 2022) or heparin sulphate proteoglycans (Guimond et al., 2022) might also be involved.Non-spike changes may also play a role (Chen et al., 2023).
To what extent are the observations herein relevant to human disease?Controversy remains regarding how useful the K18-hACE2 mouse models is for understanding human disease, although a considerable body of literature argues that many aspects of respiratory COVID-19 are recapitulated in this model (Yinda et al., 2021;Zheng et al., 2021;Bishop et al., 2022;Park et al., 2022;Ye et al., 2023).The fulminant lethal brain infection in K18-hACE2 mice is clearly not a feature of COVID-19.However, low level brain infections have been reported in hamster (Zhang et al., 2021;De Melo et al., 2023) and primate models (Beckman et al., 2022;Rutkai et al., 2022;Hailong et al., 2023), as well as in a range of human studies, including studies on long-COVID patients (Proal et al., 2023;Supplementary Table S1).The data herein also illustrates that certain features of the infected K18-hACE2 brains are also observed in some patients with severe COVID-19.As confirmed herein for BA.5 and XBB isolates, human neurons can be readily infected in vitro (Bullen et al., 2020;Ramani et al., 2020;Song et al., 2021;Hou et al., 2022;Mesci et al., 2022;Shen et al., 2022), with infection of neurons also seen in some COVID-19 patients (Song et al., 2021;Shen et al., 2022;Emmi et al., 2023), in a hamster model (De Melo et al., 2023) and in K18-hACE2 mice (Rothan et al., 2022;Seehusen et al., 2022), with BA.5 infection of neurons shown in Figure 3.A well described neurological manifestation of COVID-19 is anosmia, with infection of the olfactory epithelium implicated in humans (Khurana and Singh, 2022;Ziuzia-Januszewska and Januszewski, 2022), hamsters (De Melo et al., 2023) and in K18-hACE2 mice (Zheng et al., 2021).The olfactory epithelium may also provide access of virus to the brain (Fodoulian et al., 2020;Jiao et al., 2021;Dumenil et al., 2022), with a recent hamster study suggesting SARS-CoV-2 can travel along axons into the olfactory bulb (De Melo et al., 2023).SARS-CoV-2 may also access the brain via a breach in the blood brain barrier (Zhang et al., 2021); however, the lack of viremia (Yinda et al., 2021), and the ability to avoid brain infection by aerosolized delivery of virus into lungs, argues against this route of entry in the K18-hACE2 model (Dumenil et al., 2022;Fumagalli et al., 2022).
Brain infection is not a common feature of COVID-19, but low level infection does appear to manifest in a small group of COVID-19 and long-COVID patients (Supplementary Table S1; Proal et al., 2023).What comorbidities, injuries and/or other factors predispose to brain infection remain unclear.Perhaps of note, anosmia was recently closely linked to long-lasting cognitive problems in COVID-19 patients (RADC, 2022), with anosmia also predicting memory impairment in post-COVID-19 syndrome patients in a separate study (Ruggeri et al., 2023).A higher proportion of patients infected with BA.5 develop anosmia, when compared with BA.1 (Hansen et al., 2023) perhaps due to more rapid and fulminant infection of the upper respiratory tract (Carabelli et al., 2023).Taken together with the data presented herein, BA.5 and XBB may thus show increased risk of acute and long-term neurological complications over earlier omicron variants (Okrzeja et al., 2023).
FIGURE 1 K18-hACE2 mice infected with an original ancestral isolate, omicron BA.1, BA.5 and XBB isolates.(A) Percent weight change after intranasal infection with four SARS-CoV-2 isolates dose (5×10 4 CCID 50 ).Original ancestral isolate (n = 5-12 mice measured per time point), means and SEs are plotted for all mice (data from 3 independent experiments).BA.1, three mice showed weight loss >20% (requiring euthanasia, †) and are graphed individually; means and SE are plotted for the surviving mice (n = 4-15 mice measured per time point).BA.5 (n = 6) graphed individually.XBB (n = 6) graphed individually.†mice reached ethically defined end points for euthanasia.(B) Disease scores for the indicated overt disease symptoms for animals described in "A".For BA.1 no overt symptoms were seen except in a single animal on 9 dpi, with this animal one of the 3 that required euthanasia due to weight loss.For BA.5 the remaining mouse (n = 1) on 6 dpi showed a Posture score of 2. For XBB the one surviving mouse recovered after 7 dpi.(C) Kaplan Meyer plot showing percent survival; n = 12 for original, n = 12 for XBB, n = 18 for BA.5 and n = 24 for BA.1 (data from 2 to 3 independent experiments).Statistics by log-rank tests comparing survival rates.(D) Viral tissue titers in brains of mice that had reached ethically defined end points for euthanasia (note that this was at a range of different dpi, see Supplementary FigureS1C).All mice with weight loss requiring euthanasia had detectable viral titers in brain (determined by CCID 50 assays), with two exceptions (*); although the BA.5 infected mouse emerged to be IHC positive (the XBB infected mice was not tested).Statistics by Kolmogorov-Smirnov tests; titers for the 3 omicron variants were not significantly different from each other.ND -not detected (limit of detection ≈ 2 log 10 CCID 50 /g).(Data from 5 independent experiments).
hippocampal neurons (Figure 4G, dashed box) in the IHC positive region identified in Figure 2C.Despite fulminant brain infection not being a feature of human COVID-19, some histological lesions seen in COVID-19 patient brains are shared with K18-hACE2 mice, although it remains unclear which lesions require direct brain infection.

FIGURE 2
FIGURE 2 Immunohistochemistry of brains of BA.5 and XBB infected K18-hACE2 mice using an anti-spike monoclonal antibody.(A) Brain of BA.5-infected K18-hACE2 mouse 6 dpi showing IHC staining in the cortex and hypothalamus.Insert enlargements on the right.(B) As for "a" showing staining of the hippocampus 7 dpi.Far right shows staining of the axons (arrowheads).(C) XBB-infected K18-hACE2 mouse showing IHC staining in the cortex and hypothalamus, 6 dpi.Staining of the hippocampus for a mouse euthanized 7 dpi.

FIGURE 3
FIGURE 3 Dual labeling fluorescence immunohistochemistry of BA.5 infected brains from K18-hACE2 mice.(A) Sections from formalin fixed and paraffin embedded brains from BA.5 infected K18-hACE2 mouse brains (5 dpi) were analyzed by immunofluorescence.Sections were stained by indirect immunofluorescence with the anti-spike monoclonal antibody (red), and antibodies specific for a neuronal nuclear marker -NeuN (green).Dashed white box (left) indicate cells enlarged in the three inserts (top right).(B) As for "a" but costaining with the pan-microglial marker-Iba1 (green).Anti-Iba1 (middle panel); an enlargement of the dashed box (middle panel) is shown in the insert (bottom right), and indicates a microglial cell with amoeboid morphology.Another microglial cell with amoeboid morphology is indicated with a white unfilled arrowhead.White filled arrowhead indicates a microglial cell with bushy morphology.Merged plus DAPI (right panel); an enlargement of the dashed box is shown top right, with arrow indicating yellow (red green overlap), possibly associated with phagocytosis of infected cell debris.(C) As for "a" but costaining with a reactive astrocyte maker-GFAP (green).

FIGURE 4
FIGURE 4 Histological lesions in brains of BA.5 and XBB infected K18-hACE2 mice.H&E staining of brains from BA.5 (A-E) and XBB (F,G) infected K18-hACE2 mice.(A) Neuron vacuolation (hydropic degeneration) of neurons (arrowheads) in the cortex 4 dpi.A control brain from mice inoculated with UVinactivated BA.5 is shown on the right.(B) Perivascular cuffing.A venule (red blood cells in the center) is surrounded by leukocytes (two leukocytes are indicated by arrowheads), 7 dpi.A control venule is shown on the right.(C) Focal vasogenic edema; fluid filled perivascular space (arrowheads), 6 dpi.A control is shown on the right, with arrowheads showing normal perivascular spaces.(D) Microglial nodule; accumulation of microgliocytes (some typical microgliocytes indicated with arrowheads), 6 dpi.(E) Small hemorrhagic lesion, 7 dpi (arrowheads indicate some extravascular red blood cells).(F) Lesion in the cortex (from the IHC positive region in Figure 2C, left panel) showing perivascular cuffing (arrow), microgliosis (arrowheads) and vasogenic edema (as in "c"), 6 dpi.(G) Loss of hematoxylin staining of neurons in the hippocampus (chromatolysis) in the anti-spike positive region shown by IHC in Figure 2C (right panel).An IHC spike negative region from the hippocampus of the same mouse is shown as a control (bottom image).

FIGURE 5
FIGURE 5Transcriptome signatures in brains of K18-hACE2 mice infected with BA.5.(A) RNA-Seq of BA.5 infected brains (n = 6) compared with brains of mice inoculated with UV-inactivated BA.5 (n = 5) identified 437 DEGs.The DEGs were analyzed by Ingenuity Pathway Analysis (IPA) and GSEAs using the Molecular Signatures Data Base (MSigDB), with a representative sample of annotations shown and grouped by theme (a full list is provided in Supplementary TableS1).(B) The RNA-Seq expression data from brains of BA.5 infected K18-hACE2 mice were analyzed by SpatialDecon to provide estimates of cell type abundances.The BA.5 infected samples were ranked by viral RNA levels (highest to lowest).Cell types were clustered using the complete-linkage of Euclidean distance.(C) Relative cell type abundances for Microglia correlates with viral RNA levels.Statistics by Pearson correlation.(D) IPA cytokine USRs obtained from brains of BA.5 infected K18-hACE2 mice were compared with IPA cytokine USRs obtained using two DEG lists generated from publically available single-cell RNA-Seq data of selected brain tissues from deceased COVID-19 patients.Where a cytokine USR is identified in human but not mouse (or vice-versa), a z-score of zero is given to the latter.(E) Venn diagram showing overlaps between the cytokine USRs from the two human (Fullard and Yang) and the BA.5 infected K18-hACE2 mouse study (RNA-Seq data of brain tissues).
Collection of nasal swabs from consented COVID-19 patients was approved by the QIMR Berghofer Medical Research Institute Human Research Ethics Committee (P3600) and by the University of Queensland HREC (2022/HE001492).