We developed lung-specific SARS-CoV-2 infectious mouse models that expressed the human ACE2 gene using SFTPB and Scgb1a1 promoters ( Supplementary Figures 1A, B ). In situ hybridisation confirmed hACE2 expression in the three transgenic mouse models (SFTPB-hACE2, SCGB1A1-hACE2, and K18-hACE2 mice). hACE2 expression was detected in the type I alveolar cells and bronchos region of K18-hACE2 mice, in contrast to the lack of expression in wild-type C57BL/6 mice, whereas hACE2 was detected in the alveolar region of SFTPB-hACE2 mice and in the bronchos region of SCGB1A1-hACE2 mice, in contrast to the wild-type FVB/NJ mice ( Figure 1A , Supplementary Figures 1C, D ).

After infection of K18-hACE2 mice with SARS-CoV-2, the body temperature gradually decreased by approximately 10°C and over 20% body weight loss was observed compared with those of mock-infected K18-hACE2 mice until 7 dpi. However, no such changes were observed in SARS-CoV-2-infected SFTPB-hACE2 and SCGB1A1-hACE2 mice ( Figures 1B, C ). Survival rate was correlated with body temperature and weight loss; SARS-CoV-2 infection was lethal to K18-hACE2 mice but not to SFTPB-hACE2 and SCGB1A1-hACE2 mice ( Figure 1D ).

The viral titre was measured as of 2 dpi, with the highest PFU detected at 2 dpi, which then gradually significantly decreased over time in all hACE2 transgenic mice ( Figure 1E ). Compared with those of the control mice, the liver weight decreased (P < 0.005) and the brain weight increased (P < 0.05) in SARS-CoV-2-infected K18-hACE2 mice at 7 dpi ( Supplementary Figure 2 ). The liver weight also increased at 1 and 7 dpi in SCGB1A1-hACE2 mice but decreased at 5 dpi. However, there were no changes in the organ weights of SARS-CoV-2-infected SFTPB-hACE2 mice.

Collectively, these data demonstrated that SARS-CoV-2 infection causes high mortality in K18-hACE2 mice, accompanied by body temperature and weight loss, but shows only rare and minor clinical signs in SFTPB-hACE2 and SCGB1A1-hACE2 mice.

Given the different expression patterns of hACE2 in the three mouse models ( Figure 1A ), we hypothesised that each model has distinct pathogenic features and different SARS-CoV-2 target cells. Thus, in-depth pathological analysis of the SARS-CoV-2-infected mice was performed in samples obtained at 1, 2, 5, and 7 dpi. H&E staining showed progression of lung disease following SARS-CoV-2 infection in K18-hACE2 mice over time ( Figure 2 ; Supplementary Figure 3 ). Immune cell infiltration from the blood vessels was detected in the lungs at 1 dpi, followed by gradual development of oedema around the blood vessels ( Figure 2A , asterisk; Supplementary Figure 3 ). Subsequently, capillary dilatation was detected at the inflammation lesion ( Figure 2A , arrowhead). At 7 dpi, there was an extensively damaged area with greater thickness of the alveolar septa due to infiltration of immune cells between alveolar cells. The pathological diagnosis was confirmed in each lobe, and in the upper and lower regions of the lung in all K18-hACE2 mice.

By contrast, in SFTPB-hACE2 mice, severe SARS-CoV-2-related lesions were observed at the early stage of infection, at 1–2 dpi ( Figure 2A ; Supplementary Figure 4 ). At 1 dpi, there was severe inflammation derived from immune cell infiltration near the blood vessels and bronchioles. Inflammation close to the blood vessels along with capillary dilatation in the alveolar region were prevalent at 2 dpi. However, there was a progressive tendency for the lesions to be alleviated at 5 and 7 dpi ( Figures 2B, C ). Neither oedema nor fibrosis was observed at any period of SARS-CoV-2 infection in SFTPB-hACE2 mice.

SCGB1A1-hACE2 mice showed a similar progressive pattern of lung lesion development to that of K18-hACE2 mice, with a time-dependent severity increase following SARS-CoV-2 infection ( Figure 2A , arrow; Supplementary Figure 5 ). Immune cell infiltration was detected in the peripheral bronchiole and blood vessels, and oedema developed at 1 dpi in SCGB1A1-ACE2 mice. Subsequently, inflammation and capillary dilatation steadily increased until 7 dpi. Although the SARS-CoV-2-related lesions of SCGB1A1-hACE2 mice progressed over time, the severity was relatively moderate compared with that of the SARS-CoV-2-infected K18-hACE2 mice ( Figures 2B, C ).

Since immune cell dynamics, including neutrophil accumulation, is a characteristic feature in patients with COVID-19 pneumonia (26–29), we performed IHC for macrophage and neutrophil markers as well as two inflammatory lymphoid cell markers, PTPRC and CD3, to confirm the lung-infiltrated immune cells in each mouse model. There was accumulation of innate and adoptive immune response cells, especially F4/80⁺ macrophages and CD3⁺ T cells, over time in SARS-CoV-2-infected K18-hACE2 mice ( Figure 3A ; Supplementary Figure 4 ). SARS-CoV-2-infected SFTPB-hACE2 mice showed drastic infiltration of F4/80⁺ macrophages and a mild increase of neutrophils at 2 dpi, which subsequently decreased to 38.3% and 26.9% neutrophils and 36.9% and 62.2% F4/80⁺ macrophages at 5 and 7 dpi, respectively ( Figure 3B ; Supplementary Figure 4 ). In addition, CD3⁺ T cells temporally increased at 5 dpi and there was no change in PTPRC⁺ B cells throughout the infection period. In SCGB1A1-hACE2 mice, SARS-CoV-2 infection increased the infiltration of macrophages and CD3⁺ T cells from 2 to 7 dpi. Neutrophils transiently increased at 2 dpi, and PTPRC⁺ B cells increased at 2 and 5 dpi and then decreased at 7 dpi ( Figure 3C ; Supplementary Figure 4 ).

The neutrophil-to-lymphocyte (NTL) ratio is crucial for prognosis and mortality in patients with COVID-19 (29–35); thus, we hypothesised that the composition of peripheral blood leucocytes would also vary in the SARS-CoV-2 infection mouse models. CBC analysis showed that the NTL ratio significantly increased to 7.34%, 20.64%, 11.94%, and 25.01% at 1, 2, 5, and 7 dpi, respectively, in K18-hACE2 mice ( Figure 3D ). However, the NTL ratio did not substantially change in SARS-CoV-2-infected SFTPB-hACE2 and SCGB1A1-hACE2 mice ( Figures 3E, F ).

In situ hybridisation of the S gene was performed to verify the primary infection site and viral distribution in the three hACE2 transgenic mouse models following SARS-CoV-2 infection ( Figure 4A ). The S gene was diffused in the alveolar region, and not in the bronchiole, in K18-hACE2 and SFTPB-hACE2 mice, but exhibited a local distribution in SFTPB-hACE2 mice ( Figure 4A , 2 dpi). In SCGB1A1-hACE2 mice, the S gene was detected in both the alveolar and bronchiole regions, demonstrating different regions of primary SARS-CoV-2 infection in the three models. However, the S gene was detected at the highest level at 2 dpi and then gradually decreased until 7 dpi in all models.

We further investigated the specific cell types that SARS-CoV-2 targets using various pulmonary epithelial cell lineage markers ( Figures 4B, C ). In K18-hACE2 mice, 93.71% of AGER-expressing type I pneumocytes co-stained with SARS-CoV-2 N protein at 2 dpi, whereas 95.21% of LAMP3-expressing type II pneumocytes co-stained with N protein at 2 dpi in STTPB-hACE2 mice ( Supplementary Figure 5 ). In SCGB1A1-hACE2 mice, N protein co-stained with SCGB1A1 (77.6%), MUC5AC (47.98%), and acetylated tubulin (47.23%), representing a bronchiole club cell marker, goblet cell marker, and ciliated cell marker, respectively ( Supplementary Figure 5 ). However, chromogranin A-positive pulmonary neuroendocrine cells did not express N protein in any mouse model ( Supplementary Figure 5 ). In addition, TEM showed that SARS-CoV-2 targeted type I pneumocytes in K18-hACE2 mice and ciliated cells in SCGB1A1-hACE2 mice ( Supplementary Figure 6 ).

Given that SARS-CoV-2 induced changes to the immune cell profiles of the mouse models with different degrees of pathological severity in the lung and a varied immune cell composition in the peripheral blood, we expected that SARS-CoV-2 infection could influence organs other than the lungs, as observed in human cases and mouse models (12, 36–39).

Since the spleen is the largest secondary lymphoid organ, we performed equivalent pathological analysis and immune cell profiling of the spleen as in the lungs. In K18-hACE2 mice, the lesion caused by SARS-CoV-2 developed white pulp necrosis, with the pulp size decreasing over time ( Figures 5A and B , Supplementary Figure 7 ). We also observed dramatic structural changes, such as increased T cell- and macrophage-positive areas, and a relative decreased B cell-positive area at 1 dpi ( Supplementary Figure 8 ). After 5 dpi, the T cell-positive area recovered to the control level. By contrast, white pulp necrosis and structural changes of the spleen were rarely detected in SARS-CoV-2-infected SFTPB-hACE2 and SCGB1A1-hACE2 mice ( Figure 5 ).

The SARS-CoV-2-related lesions of the small intestine, goblet cell hyperplasia, villi atrophy, and villi necrosis were detected only in K18-hACE2 mice, and the pathological score increased over time ( Figure 6 ). Additionally multifocal perivascular cuffing, brain lesions, and SARS-CoV-2 S gene distribution were observed in K18-hACE2 mice after 5 dpi ( Figure 7 ). In the brains of K18-hACE2 mice, 31.2% of NeuN⁺ cells, 79% of serotonin⁺ cells, 59.3% of nNOS⁺ cells, and 67.4% of DCX⁺ cells co-stained with SARS-CoV-2 N protein ( Figure 7C ), whereas CD68⁺ microglial cells and GFAP⁺ astrocytes did not express N protein ( Supplementary Figure 9 ).

K18-hACE2 mice [Tg(K18-ACE2)2 Prlmn/J], established in the C57BL/6 background, were purchased from Jackson Laboratory (Bar Harbor, ME, USA). SCGB1A1-hACE2 and SFTPB-hACE2 transgenic mice were newly established in this study from FVB/NJ mice. Human ACE2 cDNA (HG10108-UT; Sino Biological, Inc., Chesterbrook, PA, USA) was cloned behind the murine Scgb1a1 (MSCV Puro-CCSP : GFP, #67487, Addgene, Watertown, MA, USA) or human SFTPB promoter (#88868, Addgene), generating lung-specific hACE2 expression vectors. Linearized and purified DNA fragments were injected into pronuclei of mouse embryos at 1 ng/μL in 10 mM Tris-HCl and 0.1 mM EDTA (pH 7.5) or 40 ng/μL in 10 mM Tris-HCl and 0.25 mM EDTA (pH 7.4).

Founder mice (pups) were identified by PCR with the primers Scgb1a1 forward 5′- GAGAATGTCCAAAACATGAA-3′, Scgb1a1 reverse 5′-AGACCCATTTTGCTGAAGAG-3′; SFTPB forward 5′-CTTGTCTCTGACTCAGGGTATTT-3′ and SFTPB reverse 5′- CAACCGTTTGCTCTTGTCTTC-3′.

Animal experimental protocols were approved by the Institutional Animal Care and Use Committee (IACUC: 2020-0216, BA-2008-301-071-03) at Yonsei University College of Medicine and were executed in an Association for Assessment and Accreditation of Laboratory Animal Care (AALAC)-accredited unit (#001071).

For virus production, we used the Vero African green monkey kidney cell line (Korean Cell Line Bank #10081). Vero cells were maintained at 37 °C in a 5% CO₂ humidified incubator within DMEM supplemented with 2 mM l-glutamine, 100 units/mL penicillin, 100 μg/mL streptomycin, and 5% FBS. SARS-CoV-2 was obtained from the National Culture Collection for pathogens of Osong, Korea (NCCP 43326, S type), which was isolated from a Korean COVID-19 patient.

Viral culture and titre determination were performed by the plaque assay as described previously (46). In brief, serially diluted supernatants of lung homogenates were added to Vero cells in 6-well plates and incubated at 37 °C for 1 h with gentle agitation every 15 min. The cells were overlaid with DMEM and 1% SeaPlaque agarose (Lonza), 2% FBS, 100 units/mL penicillin, and 100 μg/mL streptomycin. After incubation for 72 h, the cells were fixed with 4% paraformaldehyde and stained with 0.5% crystal violet/20% methanol (v/v). The virus titre was quantified as PFU/g of tissue.

Nine-week-old mice were anaesthetised with a 30 mg/kg of Zoletil and 10 mg/mL of Rompun mixture and 10⁵ PFU SARS-CoV-2 was infected intranasally to K18-hACE2, SFTPB-hACE2, and SCGB1A1-hACE2 transgenic mice; for mock infection, we used an equal volume of PBS. The health, body weight, and body temperature of the mice were monitored daily. Body temperature was measured using an implantable programmable temperature transponder. All animal experiments with SARS-CoV-2 were conducted in a biosafety level-3 facility in accordance with safety guidelines.

The mice were euthanised by CO₂ and fixed with 10% neutral buffered formalin (Sigma, St. Louis, MO, USA) for 24 h. All tissues (lungs, spleen, brain, small intestine) were embedded in paraffin wax, routinely processed, and sectioned at 4 μm thickness for H&E staining and IHC. For H&E staining, de-paraffinized slides were dipped into 0.1% Mayer’s haematoxylin for 10 min and then into 0.5% eosin. The stained slides were dehydrated in an ascending serial grade of 50%, 70%, 95%, and 100% ethanol, and mounted with mount solution (Thermo Fisher Scientific Inc., Waltham, MA, USA). Stained slides were analysed by animal pathologists.

For IHC, the slides were passed through xylene; 100%, 95%, and 70% ethanol; and distilled water for de-paraffinization and rehydration. To perform the antigen retrieval step, pH 6.0 citrate buffer (Dako S1699; Agilent Technologies, Santa Clara, CA, USA) was used under high temperature in a high-pressure cooker and then cooled on ice for 1 h. The slides were treated with 3% H₂O₂ in PBS for 30 min to block endogenous peroxidase activity, followed by treatment with M.O.M reagent (Vector Laboratories, Burlingame, CA, USA) for 1 h, and incubated with the following mouse primary antibodies at 4 °C overnight: SARS-CoV-2 N protein (NB100-56576, Novus and 40143-MM08, Sino Biological), SCGB1A1 (ABS1673, Sigma), LAMP3 (DDX0191P, Novus), AGER (MAB1179, R&D Systems), CD3b (ab5690, Abcam), PTPRC (ab64100, Abcam), F4/80 (ab6640, Abcam), Ly-6G/Ly-6C (ab2557, Abcam), doublecortin (SC-271390, Santa Cruz), CD68 (ab31630), nNOS (ab1376), Neu N (ab104224), GFAP (SC-33673).

The slides were treated with a protein blocking solution (DAKO) for 1 h and then incubated with HRP-conjugated secondary antibody (DAKO) for 15 min. DAB substrate (DAKO) was used for development of the IHC signal, and slides were dipped into Mayer’s haematoxylin for nuclear staining. Alexa 488-conjugated anti-mouse, rat, and goat IgG, and Cy3-conjugated anti-rabbit IgG were used for immunofluorescence staining, and images were obtained on a Zeiss LSM980 confocal microscope.

We performed hACE2 and SARS-CoV-2 in situ hybridisation using the RNAscope kit (ACD) according to the manufacturer’s protocol. In brief, for deparaffinization and dehydration, the paraffin slides were passed through xylene and a 100% and 95% ethanol gradient twice. After air drying, the slides were treated with hydrogen peroxide, target retrieval solution, and protease K to enclose RNA. For amplification of the RNA signal, the slides were treated with amplifying reagent (ACD) for 2 h and the signal was detected with Fast Red reagent (ACD).

The lung, spleen, small intestine, and brain specimens were pre-fixed in 2% glutaraldehyde/paraformaldehyde in 0.1 M PBS and washed in 0.1 M phosphate buffer for 24 h at 4 °C for TEM. Post-fixation was performed with 1% OsO₄ in 0.1 M phosphate buffer for 2 h, followed by dehydration in an ascending serial grade of ethanol (50%, 60%, 70%, 80%, 90%, 95%, 100%) for 10 min each, and incubated with propylene oxide for 10 min. The specimens were embedded in Poly/Bed 812 kit (Polyscience) and the blocks were cut into 200-nm sections with an ultramicrotome. The sections were placed on a copper grid and imaged with TEM (JEM-1011, JEOL, Tokyo, Japan) at an acceleration voltage of 80 kV equipped with a Mega view III CCD camera (Soft Imaging System, Germany).

Peripheral blood samples were collected from the heart immediately following mouse euthanasia and complete blood count (CBC) analysis was performed using the BC-5000Vet system (Shenzhen Mindray Bio-Medical Electronics Co., China).

Statistical significance was calculated using PRISM v9.0 software (GraphPad Software, San Diego, CA).