Long-term lung scarring after Omicron BA.5 COVID-19 infection in mice

Using reverse genetics, we generated a mouse-adapted SARS-CoV-2 MA10 recombinant bearing the BA.2 or BA.5 subvariant spike (BA.2 MA and BA.5 MA) and a related reporter virus where we replaced ORF7a with nanoluciferase (BA.2 nLuc and BA.5 nLuc) (15–17). Genome schematics of the infectious clone constructs are shown in Fig. S1. To assess replication kinetics, the growth of the two BA.5 recombinant viruses was compared to that of the ancestral strains SARS-CoV-2 MA10 and D614G nLuc, at an MOI of 0.001 on Vero E6 cells. The growth kinetics of BA.5 MA and BA.5 nLuc were the same, indicating that replication was not impacted by the inclusion of the reporter gene. The ancestral pandemic strain (SARS-CoV-2 D614G) had a growth advantage over both BA.5 viruses at most times assessed (Fig. 1). Importantly, parental SARS-CoV-2 MA10 virus grew similarly to both BA.5 viruses early in the kinetic, yet parental MA10 titers exceeded those of BA.5 viruses at later timepoints (Fig. 1). Altogether, these data demonstrate that our BA.5. spike recombinant viruses grew efficiently in Vero cells.

We have previously established a model of SARS-CoV-2 pathogenesis using mouse-adapted SARS-CoV-2 MA10 in female BALB/c mice (15). To evaluate BA.5 MA pathogenesis in a similar model, we infected 14- to 16-week-old adult female BALB/c mice intranasally with different doses, 10⁴ or 10⁵ plaque-forming units (PFUs), evaluating dose-dependent disease outcomes. While both cohorts experienced weight loss, the low-dose cohort only experienced a 10.7% group mean weight loss nadir at 4 days post-infection (dpi). In contrast, the high-dose cohort continued to lose weight throughout the study, eventually exceeding 20% of starting group mean weights (Fig. 2A). Gross lung discoloration (GLD) score is a semi-quantitative measure of acute lung damage associated with emerging CoV replication, indicative of edema and diffuse alveolar damage (15, 25). Like weight loss, GLD scores were significantly increased in the high-dose cohort as compared to the low-dose group (Fig. 2B). Similarly, mortality was only observed in the high-dose group (Fig. 2C). Regardless of inoculum dosage, the levels of viral replication within the lungs and nasal turbinates showed no significant difference between the two cohorts (Fig. 2D and E). Acute lung injury (Fig. 2F) and diffuse alveolar damage (Fig. 2G) were assessed using histological scoring schema that have been utilized for multiple emerging CoVs (15, 25). Both scoring metrics (Fig. 2F and G) revealed that acute lung injury was significantly elevated regardless of virus dose, which was evident in photomicrographs of representative mock-infected (phosphate buffered saline [PBS] inoculation) (Fig. 2H), 10⁴ PFU (Fig. 2I), and 10⁵ PFU (Fig. 2J) infected animals sacrificed at 4 dpi.

Viral pathogenesis was then evaluated in 10- to 12-month-old mice infected with 10⁴ or 10⁵ PFU of BA.5 MA. Both high- and low-dose cohorts rapidly lost weight approaching 80% of starting weight by 4 dpi (Fig. 3A). GLD scores were elevated in the high-dose cohort as compared to the low-dose group (Fig. 3B). Although trends in body weight loss were similar up to 5 dpi, high-dose infection was uniformly lethal (Fig. 3C), while the majority of low-dose infected animals survived out to 15 dpi. As observed in the younger mice above, lung and nasal titers between the infected groups did not differ significantly, regardless of dose (Fig. 3D and E). As expected, infection was associated with increases in the histopathological measures of acute lung injury (Fig. 3F) and diffuse alveolar damage (Fig. 3G). Lung tissue sections were evaluated for fibrotic lesions by Picrosirius red staining (26). Like acute lung injury scores, the frequency of profibrotic lesions increased with infection (Fig. 3H). At later times post-infection (7, 15 dpi), we evaluated tissue sections for the prevalence of pro-fibrotic regions (Fig. 3H). Like before, only the low-dose group had elevated fibrosis scores over mock-infected animals, and importantly, no high-dose animals survived to 15 dpi and thus could not be included in this analysis. Counterintuitively, the low-dose group had elevated histological scores (Fig. 3F through H) as compared to the high-dose group, but this was likely driven by survivor bias since the few high-dose infected animals that survived to later times post-infection likely had diminished disease severity compared to those that perished at earlier times. Histologic manifestations of acute lung injury, including accumulation of proteinaceous debris in the airways, inflammatory cell infiltration, and alveolar septal thickening are shown in Fig. 3I through K.

From a selected panel of representative SARS-CoV-2 reactive human monoclonal antibodies (mAbs), we identified two that showed robust neutralization titers against BA.5 nLuc (Fig. 4A). Two mAbs (COV2-3605 and COV2-3678) were compared against a recombinant version of a neutralizing SARS-CoV-2 monoclonal antibody (rLY-CoV1404) that served as a positive control, and an isotype-matched recombinant version of a human mAb antibody recognizing the unrelated dengue virus envelope protein (rDENV-2D22) that served as a negative control. In addition, a mock treatment group received an equivalent volume of PBS. In a neutralization assay against BA.5 nLuc, all three SARS-CoV-2 antibodies potently neutralized the virus, with IC₅₀ titers of 9.7 ng/mL for COV2-3605, 51.7 ng/mL for COV2-3678, and 3.2 ng/mL for rLY-CoV1404 (Fig. 4A). Both, COV2-3605 and COV2-3678 are encoded by the human antibody variable gene segment IGHV3-53. Many IGHV3-53- and IGHV3-66-encoded mAbs constitute a public clonotype, and this class of antibodies has been recurrently isolated from human subjects following SARS-CoV-2 infection or vaccination (27–30). Negative-stain electron microscopy (EM) revealed COV2-3605 Fab and COV2-3678 Fab bound to RBDs in the open conformation of the spike, consistent with the fact that members of this public clonotype target the semi-cryptic class I antigenic site (Fig. 4B) (31). Negative-stain EM data collection statistics are provided in Table S1.

To assess the prophylactic efficacy of mAbs, 10- to 12-month-old female BALB/c mice were treated with antibodies via intraperitoneal injection of 200 µg of mAb, and 12 hours later, mice were inoculated with a lethal dose of 10⁵ PFU of BA.5 MA. Mice treated with COV2-3605 or rLY-CoV1404 showed minimal or no weight loss. In contrast, ~10% transient weight loss was observed at 5 and 6 dpi for the COV2-3678-treated cohort (Fig. 4C). Importantly, negative control group mice (i.e., PBS or isotype-matched control mAb rDENV-2D22) showed rapid weight loss through 7 dpi, with all animals succumbing to infection or reaching humane endpoints for euthanasia by 7–8 dpi (Fig. 4D). Similarly, only negative control group animals had elevated GLD scores (Fig. 4E). Minimal breakthrough infection was noted in the COV2-3678-treated mice, with one of five mice exhibiting low replicating virus at 4 dpi (Fig. 4F). In contrast, no live virus was detected in the COV2-3605 and rLY-CoV1404 mAb-treated cohorts. Viral titers in mAb-treated cohorts were significantly reduced compared to PBS-treated cohorts, which had titers approaching 10⁵ PFU/lobe (P < 0.0001).

We have previously developed a model of PASC-like lung pathology using a mouse-adapted ancestral pandemic strain SARS-CoV-2 MA10 in aged mice (26). Omicron-related infections are reported to cause less frequent PASC disease in humans as compared to ancestral VOCs (24). To understand the potential role of spike protein variation in this disease process, we generated a mouse-adapted recombinant BA.2 Omicron spike virus (BA.2 MA) that is different in 4 amino acids in the spike gene from the BA.5 MA spike-containing virus. To understand BA.2 MA PASC potential, we infected cohorts of 10- to 12-month-old female BALB/c mice with PBS (mock) or 10⁵ PFU of BA.2 MA and followed animals through 120 dpi (BA.2 MA) to evaluate virologic and pathologic outcomes. Unlike with the BA.5 MA infection noted above, BA.2 MA-infected mice exhibited minimal weight loss (Fig. 5A), despite viral replication in both the lung and nasal turbinates achieving comparable levels to those seen in the BA.5 MA-infected mice described above, demonstrating the role of spike protein variation on disease progression and severity (Fig. 5B). Congruent with body weight loss, gross pathology (Fig. 5C), and histologic measures of lung fibrosis (Fig. 5D) were largely absent and non-remarkable post-BA.2 MA infections. We next measured the magnitude and durability of the neutralizing antibody response in BA.2 MA challenged mice using antigenically homologous (BA.2 nLuc) or heterologous (SARS-CoV-1 nLuc, SARS-CoV-2 D614G nLuc, SARS-CoV-2 BA.1 nLuc, SARS-CoV-2 BA.5 nLuc, and SARS-CoV-2 XBB.1.5 nLuc) Sarbecoviruses (20). As expected, the homotypic neutralization response against BA.2 was the most robust (~10,000 IC₅₀) (Fig. 5E). The magnitude of the heterotypic responses varied by genetic proximity to BA.2, where responses were 10-fold lower to ancestral pandemic strains (D614G) and were also reduced against more future emerging Omicron variants (e.g., BA.5, XBB1.5) (Fig. 5E). To understand the long-term consequences of BA.5 MA infection, we infected similarly aged mice noted above with PBS or 10-fold less virus (10⁴ PFU) to ensure disease with the majority of animals surviving. Congruent with Fig. 3, BA.5-infected animals lost an average of 17.4% ± 7.6% by 7 dpi but recovered by 30 dpi (Fig. 6A) with approximately 60% survival (Fig. 6B). Gross pathology GLD scores, which are most evident during the acute phase of infection (Fig. 3B), were low but measurable at 15 dpi and waned over time (Fig. 6C). Neither replication-competent virus via plaque assay nor viral RNA via quantitative reverse transcription-PCR was detected in lung tissue, and viral nucleocapsid antigen was not detected in lung, liver, kidney, or spleen tissue sections at any times assessed (data not shown). Lung fibrosis was evident by Picrosirius red staining with mean scores averaging between ~2 and 3 over time, indicating fibrotic lesions involving 6% to 50% of the lung parenchyma (Fig. 6D). By day 107, about 20% of the BA.5 MA virus-infected mice had resolved the most prominent long COVID lesions. Unlike mock-infected animals, BA.5 infection was associated with subpleural chronic alveolitis with scattered tertiary lymphoid structures and areas of PSR-stained interstitial fibrosis in the alveolar parenchyma at both 30 and 107 dpi (Fig. S2), thus confirming an association between chronic inflammation and fibrosis. Like BA.2 infection, homotypic neutralizing antibody responses were most robust at 30 dpi and waned through 107 dpi (Fig. 6E). As before, BA.5 serum poorly neutralized ancestral and more contemporary SARS-CoV-2 VOC and zoonotic strains. Overall, the magnitude of the heterotypic responses was driven by genetic proximity to the homotypic antigen like BA.2 above and demonstrated that Omicron spike variation can have a profound impact on acute and post-acute pathogenesis in our mouse model.

Using reverse genetics, we recovered the BA.5 and BA.2 wild-type spike gene (S) sequence in the background of previously described mouse-adapted mutations (BA.5 MA and BA.2 MA) (GenBank under accession numbers PV800150↗ and PV800152↗) (20, 21). A second set of viruses encoding the BA.5 or BA.2 S protein sequence that expressed nanoluciferase gene in place of ORF7a as previously described (BA.5 nLuc and BA.2 nLuc) (GenBank under accession numbers PV800151↗ and PV800153↗) (20, 21). Infectious virus recovery was performed as previously described (17, 20). Viral growth curve analysis was performed by infecting Vero E6 cells at an MOI of 0.001.

For recombinant protein expression, Expi293F cells (Thermo Fisher Scientific; cat # A1435101) were maintained at 37°C in 8% CO₂ in Expi293F Expression Medium (Thermo Fisher Scientific; catalog number A1435102), while ExpiCHO cells were maintained at 37°C in 8% CO₂ in ExpiCHO Expression Medium (Thermo Fisher Scientific, cat # A2910001).

Female 14- to 16-week-old or 10- to 12-month-old BALB/c mice were obtained from Envigo (Inotiv) (strain 047). Mice were inoculated intranasally under ketamine/xylazine anesthesia with either 1 × 10⁴ or 1 × 10⁵ PFU BA.5 MA, or 1 × 10⁵ PFU BA.2 MA in 50 µL PBS as indicated, as previously described (15, 17, 20).

cDNAs encoding mAbs of interest were synthesized (Twist Bioscience) and cloned into an IgG1 monocistronic expression vector (designated as pTwistmCis_G1) and used for production in mammalian cell culture. This vector contains an enhanced 2A sequence and GSG linker that allows for the simultaneous expression of mAb heavy and light chain genes from a single construct upon transfection. For antibody production, we performed transfection of ExpiCHO cell cultures using the Gibco ExpiCHO Expression System as described by the vendor (71). IgG molecules were purified from culture supernatants using HiTrap MabSelect SuRe (Cytiva) columns on a 24-column parallel protein chromatography system (Protein BioSolutions).

To express SARS-CoV-2 S proteins for ELISA binding and EM studies, we introduced the mutations of the BA.2 variant into the context of a previously described stabilized S protein construct (VFLIP) (72). In addition to a C-terminal T4 fibritin foldon domain, an 8 × His tag, and a TwinStrep tag, this construct contains an inter-protomer disulfide bond, a shorter glycine-serine-rich linker between the S1 and S2 domains, and five proline substitutions relative to the native SARS-CoV-2 S sequence. Plasmid encoding the BA.2_VFLIP antigen was transiently transfected into Expi293F cells, and culture supernatants were collected 4 to 5 days following transfection. After clarification by centrifugation and the addition of BioLock (IBA LifeSciences), antigen was purified using affinity chromatography with StrepTrap XT columns.

EM imaging was performed with COV2-BA2 spike protein in complex with either COV2-3605 or COV2-3678. A recombinant form of the COV2-BA.2 spike was expressed and purified by affinity. Fabs were generated from purified, recombinantly expressed mAbs via enzymatic digestion using a FabALACTICA kit (Genovis, cat # A2-AFK-005). Antigen-Fab complexes were generated by incubating BA.2 S_VFLIP antigen with COV2-3605 Fab or COV2-3678 Fab in a 1:4 (antigen:Fab) molar ratio for 2 hours at room temperature.

Three microliters of the complex sample at ~10 µg/mL was applied to a glow-discharged grid with continuous carbon film on 400 square mesh copper EM grids (Electron Microscopy Sciences). Grids were stained with 2% uranyl formate (73)⁠. Images were recorded on a Gatan US4000 4 k ´ 4 k CCD camera using an FEI TF20 (TFS) transmission electron microscope operated at 200 keV and controlled with SerialEM (74)⁠. All images were taken at 50,000 magnification with a pixel size of 2.18 Å/pixel in low-dose mode at a defocus of 1.5 to 1.8 µm. The total dose for the micrographs was ~33 e/Å2. Image processing was performed using the cryoSPARC software package (75)⁠. Images were imported, contrast transfer function estimated, and particles were picked automatically. The particles were extracted with a box size of 256 pix and binned to 128 pix (4.36 Å/pixel), and multiple rounds of 2D class averages were performed to achieve clean data sets. The final data set was used to generate an initial 3D volume, and the volume was refined for a final map at the resolution of ~18 Å. Fab Model docking to the EM map was done in Chimera. PDB: 12E8 was used for the Fab. ChimeraX⁠ software was used to make all the figures (76). Data collection statistics are provided in Table S1.

Antibody studies were conducted with mice treated prophylactically (12 hours prior to infection) with 200 µg of the indicated mAbs or isotype-matched IgG controls. Virus-challenged mice were inoculated with BA.5 MA at 1 × 10⁵ PFU intranasally. A mock-infected control cohort received an equivalent volume of phosphate-buffered saline intranasally.

Moderate-throughput nanoluciferase assays in a 96-well format were conducted as previously described (20).

Following harvest, mouse lungs were fixed for ≥7 days in 10% phosphate-buffered formalin at 4°C and prepared for histological examination and scoring as previously described (17, 20).

All statistical analyses were performed using GraphPad Prism 10. Statistical significance was determined by two-way analysis of variance (ANOVA) with Tukey's multiple comparison test for weight loss, Kruskal-Wallis non-parametric test with Dunn's correction for lung discoloration, and one-way ANOVA with Tukey's multiple comparison test for tissue titers. Symbols denote statistically significant relationships at the respective levels: *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001. All cohorts started with five mice per harvest timepoint at the time of infection. Error bars represent standard error of the mean.