All mouse work was conducted in accordance with the “Australian code for the care and use of animals for scientific purposes” as defined by the National Health and Medical Research Council of Australia. Mouse work was approved by the QIMR Berghofer Medical Research Institute animal ethics committee (P3600), with infectious SARS-CoV-2 work conducted in a PC3 (Bio-Safety Level 3) facility at the QIMR Berghofer MRI (Australian Department of Agriculture, Water and the Environment certification Q2326 and Office of the Gene Technology Regulator certification 3445). 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.

Collection of nasal swabs from consented COVID-19 patients (to isolate circulating SARS-CoV-2 variants) was approved by the QIMR Berghofer Medical Research Institute Human Research Ethics Committee (P3600).

The omicron BA.5 isolate, SARS-CoV-2_QIMR03 (SARS-CoV-2/human/AUS/QIMR03/2022) belongs to the BE.1 sublineage (GenBank: OP604184.1) and was isolated from a nasal swab, voluntarily collected and donated by a de-identified, consented, adult COVID-19 patient with degree level education (53). Virus stocks were propagated in Vero E6 cells, stocks and tissue culture supernatants were checked for endotoxin (54, 55) and mycoplasma (MycoAlert, Lonza) (56). Virus titres were determined by CCID₅₀ assays (57).

mACE2-hACE2 mice (58) were generated as described (46) by Monash Genome Modification Platform (MGMP), Monash University and are freely available as heterozygotes through Phenomics Australia (MGMP code ET26). The strain was initially maintained in-house as heterozygotes by backcrossing to C57BL/6J mice. Heterozygotes were then inter-crossed to generate a homozygous mACE2-hACE2 transgenic mouse line. Genotyping was undertaken by digital droplet PCR (MGMP) to distinguish homozygotes from heterozygotes; hACE2 primers 5′-CCAGATGTACCCTCTGCAAG-3′/5′-TCGTGTTCAGGATGGTGTTC-3′, probe 6-carboxyfluorescein-5′-GCTCCAGCTGCAGGCTCTCCAGCA-3′-ZEN/IowaBlack; RPP30 reference primers CTTTGAACTTGTCTATGGTCCT/GCATCAAATTGAGGGCATTG, probe hexachlorofluorescein-TGTGTACCTTCTCATCGTTGCATC-ZEN/IowaBlack. A Bio-Rad QX200 ddPCR droplet generator was used to generate droplets, amplified products were analysed by QX200 droplet reader, and copy numbers were determined using QuantaSoft Analysis Pro Version 1.0 (Bio-Rad, USA). After 2 inter-crossing of homozygotes, all offspring were homozygotes, and a homozygous line was established.

Mice were infected as described (59), briefly, female mice (≈ 10-20 weeks of age) received intrapulmonary infections delivered via the intranasal route with 5×10⁴ CCID₅₀ of virus in 50 μl RPMI 1640, while under light anaesthesia. 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 (53). Mice were euthanized using CO₂, and tissue titres determined using CCID₅₀ assays and Vero E6 cells (57).

The MPs comprised internally dye loaded, Fluoresbrite^® yellow-green polystyrene-based microspheres with a diameter of 1 µm (Cat# 17154-10) purchased from Polysciences as 2.5% w/v in a sterile aqueous suspension without sodium azide. The zeta potential of ≈ 1 µm polystyrene beads has been estimated to be ≈ -20 mV (60). The rationale for choosing these MPs has been described previously (45); briefly, polystyrene MPs are frequently found in the environment, surface labelled microspheres have altered surface characteristics not recapitulate by MPs in the environment (and were thus not used), and the 1 μm size is approximately the size of a bacteria, with bacteria routinely phagocytosed by macrophages. MPs were diluted in PBS and administered alone or together with the viral inoculum in a single dose of 1 µg of MPs per mouse (≈ 40 µg/kg); the total inoculated volume was always 50 µl per mouse.

Lungs were fixed in 10% formalin for 2-3 days, tapped dry and embedded in O.C.T. (Tissue-Tek, Qiagen), with ≈ 7 µm cryosections, under a glass coverslip, viewed by fluorescent microscopy. MPs were counted by eye.

Lungs were weighed, manually chopped using scissors and digested in ammonium sulphate (50 mM), SDS (5 mg/ml) and proteinase K (1 mg/ml) overnight at 37°C as described (61). Digested suspensions were viewed by fluorescent microscopy and a haemocytometer, with fluorescent MPs counted by eye.

Histology was undertaken as described (59, 62), with lungs fixed in formalin, embedded in paraffin, sections stained by H&E and slides scanned by Aperio AT Turbo (Aperio, Vista, CA, USA). Image analysis (nuclear/cytoplasmic staining ratios) was undertaken using Positive Pixel Count v9 algorithm. White space analysis was undertaken using QuPath v0.2.3.

Immunohistochemistry (IHC) was undertaken as described (63) using the macrophage/monocyte monoclonal antibody F4/80 (Abcam, Cambridge, MA) and color developed using NovaRed (Vector Laboratories, Newark, CA, USA).

RNA-Seq and bioinformatic analyses were undertaken as described (46, 59). Briefly, mouse lung tissues were harvested into RNAlater, RNA was extracted using TRIzol (Life Technologies), and RNA concentration and quality measured using TapeStation D1kTapeScreen assay (Agilent). cDNA libraries were generated using Illumina TruSeq Stranded mRNA library prep kit and sequencing performed in-house using Illumina Nextseq 2000 platform (75-base paired end reads). Processed reads were aligned to GRCm39 vM31 (mouse genome) and the BA.5 genome using STAR aligner. Gene expression was calculated using RSEM and EdgeR. For the BA.5+MP vs. BA.5 data sets a term for viral reads was introduced into EdgeR to minimize the effects of viral loads on significance and fold change in the mRNA expression data. A filter was applied to the count matrix of counts per million (cpm) >1 for any given gene in at least 5 samples.

Differentially expressed genes (DEGs) were analysed using Ingenuity Pathway Analysis (IPA) (QIAGEN). Whole gene lists ranked by fold change were interrogated using Gene Set Enrichment Analyses (GSEA v4.0.3) (Broad Institute, UCSanDiego) using the “GSEAPreranked” module. Gene sets were obtained from the complete Molecular Signatures Database (MSigDB) v7.2 (31,120 gene sets) (msigdb.v7.2.symbols.gmt), Blood Transcription modules (BTMs) (64), and Xue at al., 2014 (65). Relative abundance of specific cell types was estimated via cellular deconvolution using the SpatialDecon package in R (66), with cell-type expression matrices obtained from Yoshida et al., 2019 (67) or the NanoString Cell Profile Library, either Mouse/Adult/Lung_MCA (Mouse cell atlas) or Mouse/Adult/ImmuneAtlas_ImmGen_cellfamily (Immune cell family) (available at https://github.com/Nanostring-Biostats/CellProfileLibrary/tree/master/Mouse/Adult).

Kraken metagenomic sequence classification was undertaken as described (68). The Interferome database (69) was used to identify interferon regulated genes among DEG sets.

Total RNA was used to synthesize cDNA using ProtoScript II First Strand cDNA Synthesis Kit (New England Biolabs) and qPCR performed using iTaq Universal SYBR Green Supermix (Bio-Rad) as per manufacturer’s instructions with primers Nr4a1 (Forward 5’-GTTGGGGGAGTGTGCTAGA-3’ and Reverse 5’-AATACAGGGCATCTCCAGCC-3’), Ccn1 (Forward 5’-AAGAGGCTTCCTGTCTTTGGC-3’ and Reverse 5’- AACTCGTGTGGAGATGCCAG-3’), Hspa1a (Forward 5’-TTTGTGTATTGCACGTGGGC-3’ and Reverse 5’-CCAGGGGAGAGTCCAAACAC-3’), Hspa1b (Forward 5’- AATGTTGGGAGCAGCACTGT-3’ and Reverse 5’-TGTCTTCCCAGGCTACTGGA-3’), Oas3 (Forward 5’-TGGCAATCCCATCAAGCCAT-3’ and Reverse 5’- CTGAGGGCTGGTGTCACTTT-3’), Irf7 (Forward 5’-ACCGTGTTTACGAGGAACCC-3’ and Reverse 5’-GTTCTTACTGCTGGGGCCAT-3’), Ccl4 (Forward 5’- GCCAGCTGTGGTATTCCTGA-3’ and Reverse 5’-TGAACGTGAGGAGCAAGGAC-3’). qPCR reactions were performed in duplicate and averaged. Gene expression was normalised using mRPL13a as the house-keeping gene (70). Log₂ fold-change was calculated using the 2-ΔΔCt method (71).

Statistical analyses of experimental data were performed using IBM SPSS Statistics for Windows, Version 19.0 (IBM Corp., Armonk, NY, USA). The t-test was used when the difference in variances were <4 fold (determined using Data Analysis ToolPak in Excel), skewness was >-2 and kurtosis was <2 (determined using SPSS). Otherwise, the non-parametric Kolmogorov-Smirnov asymptotic test was used (SPSS).

Experimental delivery of MPs into mouse lungs generally involves delivery of MPs in suspension into the lungs via the intranasal (i.n.) route (33, 35–38), as mimicking MP dust inhalation is currently technically, logistically and ethical difficult in a laboratory setting. Some of the challenges include consistent airborne MP dosing, staff safety considerations, and mouse eye irritation, respectively.

To assess the effects of MPs on uninfected lungs, a single dose of 1 µg of MPs per mouse (≈ 40 µg/kg) of azide-free, 1 µm diameter, fluorescent dye loaded, polystyrene beads (density 0.26 g/ml) was delivered in 50 µl of PBS via the i.n. route into the lungs of lightly anesthetized female C57BL/6J mice. Control mice received 50 µl of PBS. The anaesthesia prevents the mouse sneezing or coughing out the introduced material, but is light enough for the mouse to retain largely normal breathing, promoting deep lung delivery of the material; this is the same method used to infect mice with SARS-CoV-2 (46, 51, 59, 72, 73). The mice were observed daily and no overt clinical signs were observed. Mice were euthanized on day 2 and 6 post MP inoculation, and lungs analysed by RNA-Seq, with bioinformatic treatments comparing +MP day 2 vs. PBS day 2 and +MP day 6 vs. PBS day 6 ( Figure 1 ; Supplementary Tables 1 , 2 ).

RNA-Seq for +MP vs. PBS day 2 identified 341 differentially expressed genes (DEGs) at q (FDR) <0.05 ( Supplementary Table 1 ). Fold change (FC) was low overall, with only 9 genes showing FC>2 (log2 FC>1) and only 4 of these had a mean normalized counts per million (cpm) >10 across all samples ( Supplementary Table 1 ). Ingenuity Pathway Analysis (IPA) indicated a mild pro-inflammatory response, with IL-1β the highest Cytokine UpStream Regulator (USR) by z score ( Figure 1 , Inflammation; Supplementary Table 1 ). Some signatures that can be associated with stress responses were also evident, principally ERK5 (MAPK7) (74, 75) and NFkB (76, 77) ( Figure 1 , Stress). A series of annotations associated with cellular protrusions, proliferation, differentiation and survival were identified as the top annotations (by z score) by IPA Diseases and Functions ( Figure 1 ), with the largest negative z scores associated with growth failure or death ( Supplementary Table 1 ).

Relative abundance of specific cell types was analysed via cellular deconvolution, SpatialDecon (66), with cell-type expression matrices obtained from Yoshida et al., 2019 (67) or the NanoString Cell Profile Library, either Mouse/Adult/Lung_MCA (Mouse cell atlas) or Mouse/Adult/ImmuneAtlas_ImmGen_cellfamily (Immune cell family) ( Supplementary Table 1 ). These analyses suggested that the cells that were more transcriptionally active (expanding/repairing) in the +MP group were AT1 and AT2 type I and II alveolar epithelial cells ( Figure 1 , AT1, AT2). The positive z score for “Differentiation of stem cells” ( Figure 1 , Cellular protrusions, growth, differentiation, survival), may be associated with reduced abundance of Club cells (formally known as Clara cells), which are regional progenitor cells that repair bronchiolar epithelium in response to lung damage. Also identified was a reduced abundance of interstitial macrophages, cells that are associated with nerves and airways and appear to have important immunoregulatory roles (78, 79). A reduction in interstitial macrophages may also be consistent with the reductions in anti-inflammatory cytokines ( Figure 1 , IL-10 and IL-24) (79). Dendritic cell abundance was also down in the +MP group, perhaps consistent with mobilization to lymph/lymph nodes in response to inflammation (80).

In summary, MP inoculation into lungs generated a transcriptional signature indicating a mild inflammatory response, and mild increases in abundance scores (likely proliferation) for lung epithelial cells, and reductions in abundance scores for interstitial macrophages and dendritic cells.

Harvesting of lungs on day 6 after MP or PBS inoculation and comparing the transcriptome by RNA-Seq (+MP day 6 vs. PBS day 6) identified only 7 DEGs, all with low fold change ( Supplementary Table 2 ). This argues that the lung response to a single exposure to MPs is largely resolved within 6 days. GSEAs using MSigDB gene sets provided a series of significant annotations with high positive Normalised Enrichment Score (NES) associated with epithelial cells (e.g. Epithelial differentiation), suggesting some lung repair activities were still underway on day 6 ( Supplementary Table 2 ).

The best described mouse model for SARS-CoV-2 infection is the K18-hACE2 model which expresses the SARS-CoV-2 receptor, human ACE2 (hACE2), from the keratin 18 promoter (K18). This model is generally lethal within several days, as it usually leads to brain infection and ensuing weight loss that reaches ethically defined criteria for euthanasia (53, 59, 73). A non-lethal, less severe model is the mACE2-hACE2 model, wherein hACE2 is expressed from the mouse ACE2 promoter (mACE2) (58). We generated such a transgenic mouse on a pure C57BL/6J background by microinjection of the mACE2-hACE2 transgene into the pronucleus of C57BL/6J zygotes at the pronuclei stage (46) and generated a homozygous mACE2-hACE2 transgenic mouse line (see Materials and Methods). Given MP-mediated effects were generally mild ( Figure 1 ; Supplementary Table 1 ), we chose this mouse model so that any MP-mediated perturbations might be more readily detected.

Although infection of mACE2-hACE2 mice with an original strain isolate has been described (46, 58), infection of homozygous mACE2-hACE2 mice with an omicron BA.5 isolate has not. The latter did not result in significant weight loss (data not shown), with lung histology at 6 days post infection (dpi) showing a series of histopathological features that have been described previously in COVID-19 mouse models (49, 58, 72) ( Supplementary Figure 1 ). Lesions were less severe when compared with those seen after infection of K18-hACE2 mice with an original strain isolate (59, 81, 82), although significant loss of white space (unstained air-spaces) in H&E stained lung sections (indicating lung consolidation) (51, 72), was also seen in this model ( Supplementary Figure 2 ).

Infection of mACE2-hACE2 mice with BA.5 (BA.5 vs. PBS) was analysed by RNA-Seq, with lungs harvested on 2 dpi (peak viral load) and 6 dpi (peak lung pathology) and compared with mock infected lungs (PBS) harvested on days 2 and 6, respectively ( Supplementary Tables 3 , 4 ). When the lung cytokine response signatures (IPA Cytokine USR z scores) from BA.5-infected mACE2-hACE2 mice, were compared with those from K18-hACE2 mice infected with an original strain isolate (46), a highly significant correlation emerged for 2 dpi. However, at 5/6 dpi a less significant correlation was observed ( Supplementary Figure 3 ), reflecting the lower severity of lung disease in the BA.5 mACE2-hACE2 model.

IPA Diseases and Function annotations for BA.5 infected lungs for mACE2-hACE2 mice (BA.5 vs. PBS) pertinent to the analyses below include, top annotations for leukopoiesis/haematopoiesis, phagocytosis (engulfment of cells) and apoptosis & necroptosis ( Supplementary Tables 3 , 4 ).

To evaluate the effects of SARS-CoV-2 infection on MP clearance from lungs, mACE2-hACE2 mice were given a single inoculum of 50 µl containing both BA.5 (5x10⁴ CCID₅₀) and 1 µg (≈40 µg/kg) of azide free MP beads into the lungs via the intranasal route using the same procedure used herein and generally to infect mice with SARS-CoV-2 (59, 72, 83) (BA.5+MP). Control mice were not infected, but received the same 50 µl inoculum containing MP (+MP). Mixing the beads with the viral inoculum prior to infection prevented any complications that might arise from simple liquid occlusion of airways, if for, instance, the SARS-CoV-2 inoculum was given first and MPs were introduced later. Lungs were harvested at ≈ 4 hrs, 24 hrs, 48 hrs (day 2) and 144 hrs (day 6), were fixed in formalin, and the beads (MPs) observed by fluorescent microscopy of cryosections ( Figure 2A ). Quantitation suggested slower clearance of the MPs in BA.5-infected lungs ( Figure 2B ), consistent with SARS-CoV-2-mediated disruption of the ciliary layer, which is responsible for mucociliary clearance (59, 84). To provide better quantitation, lungs were digested in proteinase K and dissolved in SDS, and beads counted using a hemocytometer under a fluorescent microscope. The same trend was observed, with significantly more beads present at 24 and 48 hrs in infected mice ( Figure 2C ). Results from both assays indicated that MPs were largely cleared by day 6 ( Figures 2B, C ).

To evaluate the effects of MPs on SARS-COV-2 infection, mACE2-hACE2 mice were given a single inoculum of 50 µl containing both BA.5 (5x10⁴ CCID₅₀) and 1 µg (≈40 µg/kg) of beads, delivered into the lungs via the intranasal route (BA.5+MP). Control mice received the same 50 µl inoculum containing BA.5, but no MPs (BA.5). Lungs and nasal turbinates were harvested on 2 and 6 days post infection (dpi) and tissue titres determined by CCID₅₀ assays, with no significant differences in viral titres evident on 2 dpi, and below the level of detection by 6 dpi ( Figure 2D ).

Lungs were also analysed by RNA-Seq and reads aligned to the viral genome, with the number of viral reads not significantly different for groups with or without MPs at 2 or 6 dpi ( Figure 2E ; Supplementary Tables 5 , 6 ). Thus overall MP inoculation had no significant effects on viral loads in the respiratory track.

To assess the effects of MPs on the innate immune responses induced by BA.5 infection, the same groups described above (BA.5+MP vs. BA.5, 2 dpi) were compared by RNA-Seq, with reads aligned to the mouse genome. To minimize within-treatment variation in gene expression due to variability in viral load ( Figure 2E ; Supplementary Table 5 ), viral loads expressed in counts per-million were included as a term in the linear model used for estimating gene expression in EdgeR. Applying a q<0.05 filter, 596 DEGs were thereby identified ( Figure 3 ; Supplementary Table 5 ).

The DEGs were analysed by IPA as above, and the ‘All gene’ list, ranked by fold change, was used in a series of Gene Set Enrichment Analyses (GSEAs) using gene sets from the Molecular Signatures Data Base (MSigDB), Blood Transcription modules (BTM) and Xue et al., 2014 (65) ( Figure 3 ; Supplementary Table 5 ). Relative abundance of specific cell types was analysed as above using SpatialDecon (66), using gene expression matrices provided by NanoString Cell Profile Library, and Yoshida et al., 2019 (67) ( Figure 3 ; Supplementary Table 5 ).

The highest ranked IPA Canonical pathway was the ‘Coronavirus pathogenesis pathway’ ( Figure 3A , Pathology; Supplementary Table 5 ). MPs were recently reported to inhibit phagocytosis of apoptotic cells (known as efferocytosis) (42), and a top IPA Diseases and Functions annotation was ‘Phagocytosis’, with a high negative z score ( Figure 3A , Phagocytosis; Supplementary Table 5 ). During infection, macrophages phagocytose SARS-CoV-2-infected cells and are thereby activated, and in turn mediate strong activation of plasmacytoid dendritic cells (pDC) (85). The reduced phagocytosis is thus consistent with a reduction in activated M1 macrophages identified via a series of annotations ( Figure 3 , M1 macrophages); specifically (i) reduced abundance of polyinosinic:polycytidylic acid (pIC) stimulated macrophages (a stimulus that mimics viral double stranded RNA), (ii) a series of IPA Canonical pathways and Diseases and Functions macrophage annotations with negative z scores, and (iii) GSEAs with negative NES and q<0.05 using M1 macrophage gene sets from Xue at al., 2014 (65) ( Supplementary Table 5 ).

Significantly reduced numbers of pDC were identified in the BA.5+MP group by cellular deconvolution using NanoString expression matrices ( Figure 3 , pDC, Supplementary Table 5 ), consistent with the reduction in M1 macrophages (85). As might be expected, viral load and pDC abundance showed positive correlations, with pDC abundance lower for the BA.5+MP group across all viral loads ( Figure 3B ). pDC are dominant producers of type I interferons (IFNs) during SARS-CoV-2 infection, with low numbers of pDCs and low type I IFN levels generally associated with increased COVID-19 severity (86, 87). Consistent with reduced abundance of pDC in the BA.5+MP group, was a series of type I IFN IPA USR annotations with highly negative z scores ( Figure 3A , Type I IFNs; Supplementary Table 5 ). Reduction in type I IFN signatures was also evident from GSEAs using MSigDB gene sets ( Supplementary Table 5 ). The general overall reduction in innate sensing signatures ( Figure 3A , Innate sensing) was consistent with reduced type I IFN signatures, as well as the signatures associated with reduced phagocytosis, M1 macrophages and pDCs.

The lower type I and type III IFN signatures in the BA.5+MP group were not associated with significantly higher viral titres or higher levels of viral RNA ( Figures 2D, E ). This might be expected as even elimination of the type I or type III IFN receptor (in IFNAR^-/- and IL-28RA^-/- mice, respectively) had no significant impact on viral replication in the lungs (49). The ability of multiple SARS-CoV-2 proteins to mediate evasion of innate IFN responses is well described (88, 89), and explains the insensitivity of the virus to these responses. SpatialDecon, and GSEAs using MSigBD and Blood Transcription modules (BTMs), identified significantly reduced Natural Killer (NK) cell signatures in the BA.5+MP group ( Figure 3A , NK cells; Supplementary Table 5 ), with low early type I IFN levels linked to reduced NK activity during viral infections generally (90) and likely also SARS-CoV-2 infections (91). NK cells have the capacity to exert important early innate antiviral activities; however, SARS-CoV-2 shows a remarkable ability to evade this arm of the immune system (92). NK and NKT cells are important sources of early IFNγ (90), thus the high negative z score for the IPA USR annotation for IFNγ ( Figure 3A , Type I IFN; Supplementary Table 5 ) is consistent with reduced abundance of NK and NKT cells.

The general reductions in innate immune signatures for BA.5+MP vs. BA.5 ( Figure 3 , Innate sensing, Type I IFNs, NK cells) is likely to be responsible for the negative z scores for a series of IPA USR pro-inflammatory cytokine annotations ( Figure 3A , Proinflammatory cytokines; Supplementary Table 5 ). Overall, these data illustrate that MPs can significantly ameliorate SARS-CoV-2-mediated activation of innate immune responses in the lung at 2 dpi.

The Interferome database provides an open-access bioinformatic resource that allows identification of interferon regulated genes (IRGs). The 100 most down-regulated DEGs at 2 dpi (according to EdgeR) were interrogated using the Interferome database, with a total of 45 of these identified as type I IRGs ( Supplementary Table 5 ). Three of these (Ccl4, Irf7, and Oas3) were among the most down-regulated DEGs by RNA-Seq. Their significant differential expression was validated by RT-qPCR ( Supplementary Figure 4A ).

Three cytokine signatures associated with haematopoiesis were identified as down-regulated in the IPA USR analysis; CSF1/M-CSF, CSF2/GM-CSF and IL3 ( Figure 3 , Haematopoiesis; Supplementary Table 5 ). Cellular deconvolution using SpatialDecon and the Nanostring expression matrices also identified a higher abundance of haematopoietic stem and progenitor cells (HSPC) in the lungs of the BA.5+MP group at 2 dpi ( Figure 3A , Stem.Prog; Supplementary Table 5 ). M-CSF, GM-CSF and IL3 (as well as IFNα, IFNγ and IL1, and TLR) are some of the key cytokine signaling pathways that promote HSPC differentiation (93). These analyses thus suggest an accumulation of undifferentiated HSPC due to reduced differentiation of HSPC into myeloid and/or lymphoid precursors.

Dysregulated emergency myelopoiesis and immature myeloid cells are associated with poor outcomes in COVID-19 patients (94–96), although these are observations from peripheral blood during COVID-19 disease, rather than from lungs early post infection. Nevertheless, these data ( Figure 3A , Haematopoiesis & Haematopoietic stem and progenitor cells) suggest MPs in the lung suppress haematopoiesis, likely as a result of the overall reduction in the proinflammatory milieu for BA.5+MP. To clarify the terminology used in these annotations; haematopoiesis includes myelopoiesis, lymphopoiesis and erythropoiesis, whereas leukopoiesis encompasses myelopoiesis and lymphopoiesis.

IPA analysis of BA.5+MP vs. BA.5 indicated some signatures that are usually associated with anti-inflammatory activity ( Figure 3A , Anti-inflammatory). The high USR z score for TRIM24 ( Supplementary Table 5 ) argues that, although M1 macrophages are down, M2 macrophages are not increased, given that TRIM24 expression is suppressed in M2 macrophages (97).

Interstitial macrophages can have immunosuppressive properties (79, 98, 99); however, although identified as down in Figure 1 they were not identified as increased for BA.5+MP vs. BA.5. Conceivably, these anti-inflammatory signatures arose from the higher abundance of alveolar macrophages that are not M1 biased.

RNA-Seq was also undertaken for BA.5+MP vs. BA.5 at 6 dpi (the day of peak pathology in this model) with reads again aligned to the mouse genome. Using the same filter (q<0.05) 528 DEGs were identified and were analysed as above ( Supplementary Table 6 ). A number of proinflammatory cytokine IPA USR signatures were up-regulated ( Figure 4 , Proinflammatory cytokines), all are associated with increased COVID-19 severity; TNF (100, 101), IL1B (102), IL1A (103), OSM (104), IL17A (105), and IL6 (106). Upregulated activity of the transcription factor EPHAS is also associated with pulmonary inflammatory responses in lethal COVID-19 (107). The anti-inflammatory interleukin 1 receptor antagonist (IL1RN) signature was down-regulated, with reduced IL1RN recently associated with increased COVID-19 severity (108).

The increased proinflammatory cytokine signatures ( Figure 4 ) were not associated with reduced viral loads ( Figure 2E ). Of these responses, TNF is described as having anti-viral activity in some settings (109–111). However, anti-TNF therapy has not been associated with increases in SARS-CoV-2 replication or COVID-19 severity (100, 112), suggesting that TNF has minimal anti-viral activity against SARS-CoV-2 during COVID-19.

The mechanisms responsible for the increase in proinflammatory signatures ( Figure 4 ) remain unclear, with, for instance, reduced type I responses ( Figure 3 ) previously associated with reduced (not increased) lung inflammation (49). However, GSEAs using MSigDB gene sets provided a significant macrophage annotation with high NES for TLR3 stimulation, with an IPA USR annotation for TLR3 signaling also identified, albeit with a relatively low z score ( Figure 4 , Macrophage stimulation). Viral double-stranded RNA is a probable ligand for TLR3 (113). The reduced phagocytosis ( Figure 4 , Phagocytosis), and the increased proinflammatory responses ( Figure 4 , Proinflammatory cytokines), might suggest increased secondary necrosis due to a reduction in the phagocytosis of apoptotic cells (efferocytosis) (114–116). Multiple IPA Disease & Functions annotations suggest an increase in apoptosis signatures ( Supplementary Table 6 ), with secondary necrosis known to be proinflammatory via secretion of a number of mediators such as HMGB1 and ATP (114). These latter mediators were identified as IPA USRs, although again with relatively low z scores ( Figure 4 , Cell death and proinflammatory cell death signaling).

NK cell and haematopoiesis signatures remained, as at 2 dpi, down-regulated by MPs ( Figure 4 ). Tissue repair signatures were also identified ( Figure 4 , Tissue repair) as might be expected by 6 dpi, when virus has been largely cleared ( Figures 2D, E ).

Human lung RNA-Seq data for severe lethal COVID-19 infections was recently provided, with two signatures described, a ‘Classical signature’ and a ‘cytokine release syndrome’ (CRS) signature (117). We re-derived two DEG lists from the fastq files deposited for this study ( Supplementary Table 7 ; PRJNA1036279), with the lists then analysed by IPA. Significant cytokine USR z scores provided by this analysis were compared with the z scores of significant cytokine USRs identified for BA.5+MP vs. BA.5 ( Figures 5A, B ; Supplementary Table 7 ). A significant correlation emerged for CRS ( Figure 5A ). In both human (infected vs. uninfected) and mouse (BA.5+MP vs. BA.5) data sets, TNF, IL1A, IL1B, OSM, IL6 and IL17 signatures were prominently up-regulated, and CSF1, IL3 and EPO were prominently down-regulated ( Figure 5A , pink shading). These cytokines and their association with severe COVID-19 are described above for Figure 4 . Treatment with EPO (118) and CSF1 (119, 120) have been considered for COVID-19, with low IL-3 levels associated with increased severity (121).

Budhraja et al., 2022 also described a “Classical signature” as the more common pattern for lethal COVID-19 (117); a similar IPA analysis indicated no correlation when cytokine USR z scores for the ‘Classical signature’ were compared with cytokine USRs z scores identified for BA.5+MP vs. BA.5 ( Figure 5B ).

Another way of representing such data is by heat maps (46). When the top 30 cytokine USRs (by absolute z score) for BA.5 vs. PBS were ranked, and shown next to the aforementioned z scores for CRS and the Classical signature, there were no significant positive correlations ( Figure 5C , Pearson correlations). The correlation was also poor for comparisons with the mean cytokine USR z scores previously generated from 4 human studies of SARS-CoV-2 infected lung tissues (46) ( Figure 5C , Bishop, mean human, p=0.51). These poor correlations ( Figure 5C ) likely again reflect the relatively mild disease seen in the BA.5 mACE2-hACE2 mouse model.

When the top 30 cytokine USRs for BA.5+MP vs. BA.5 were ranked and compared with z scores from CRS in COVID-19 infected patients and the Bishop et al., 2022 study, significant positive correlations again emerged ( Figure 5D ). As in Figure 5B , there was no significant correlation with the classical signature ( Figure 5D ).

Taken together these analyses suggest that MPs in the SARS-CoV-2-infected lung pushed the cytokine signatures towards a CRS profile. However, MPs did not induce overt clinical disease in the BA.5-infected mice, nor were we able to detect significant histological changes in lung sections ( Supplementary Figures 2B , 5 ).

At the top of the DEG lists for BA.5+MP vs. BA.5 are a number of genes that are associated with stress, and/or genes that have previously been identified as being induced after smoke or diesel particle inhalation, or in other lung diseases/disease models ( Table 1 ). The top DEGs for BA.5+MP vs. BA.5 at 6 dpi where heat shock protein 70 (Hsp70) family members Hspa1a and Hspa1b, with Hsp70 induction well described in the MP literature (see Discussion).

We also calculated which genes were synergistically induced (i.e. genes for which the cpm for BA.5+MP > +MP alone plus BA.5 alone). Hspa1a and Hspa1b emerged as the most synergistically induced genes, with Cxcl5, Nr4a1 and Ccn1 also identified ( Table 1 ; Supplementary Table 8 ). RT-qPCR was used to validate differential expression of Nr4a1, Ccn1, Hspa1a and Hspa1b, with log₂ fold-changes and statistical significance consistent with the RNA-Seq data ( Supplementary Figure 4B ).

When the upregulated DEGs for BA.5+MP vs. BA.5 were compared with upregulated DEGs for +MP vs. PBS, only a small number of genes were found to be common to both DEG lists at both time points (n=12 for 2 dpi and n=2 for 6 dpi) ( Supplementary Figure 6 ). Thus MPs on their own induced a largely different set of genes, when compared with MPs in a SARS-CoV-2 infection setting. This observation supports the contention that the detrimental activity of MPs might best be observed in the dysregulation of the inflammatory processes during the course of a disease (45), rather than as an imposition of fixed MP-specific responses.

Despite the overall low level of overlaps, treatment with MPs was consistently associated with upregulated of one DEG, kruppel-like factor 2 (Klf2) ( Supplementary Figure 6 ).