COVID-19 and influenza infections mediate distinct pulmonary cellular and transcriptomic changes

K18 mice infected with a sub-lethal dose of CoV2 have persistent lung abnormalities

The K18 transgenic mice express human ACE2, a CoV2 receptor under the transcriptional control of the human cytokeratin 18 promoter in airway epithelial cells^{17,18,19,20,21}. We have shown that the K18 mice infected with a lethal dose (2 × 10⁵ TCID50) of CoV2 develop severe acute pneumonia, and that all infected mice meet euthanasia criteria within 7 days post-infection (DPI)¹⁵. To examine the chronic impact of CoV2 infection on the lung, we infected the mice with a twenty-fold lower (1 × 10⁴ TCID50) inoculum of CoV2 (Fig. 1a) (detailed mice information are shown in Supplementary Table 1). Both lethal and sub-lethal infections induced loss of body weight starting at 4 DPI (Fig. 1a); however, K18 mice infected with the sub-lethal dose of CoV2 failed to reach euthanasia criteria, re-gaining body weight starting at 8 DPI, and reaching the original body weight by 13 DPI (Fig.1a). We detected viral genomic RNA in the lungs of sub-lethal dose infected K18 mice at 21 DPI (Fig.1a). To further monitor the replication of CoV2, we assayed sub-genomic viral RNA load^15,22. There were no detectable CoV2 sub-genomic viral RNAs in the lungs or brains of the infected K18 mice at 21 DPI (Fig. 1c). Hematoxylin and eosin (H&E) staining showed patchy consolidation in all three K18 mice at 21 DPI (Fig.1d) and two K18 mice at 45 DPI infected with sub-lethal dose (Supplementary Fig. 1a–h). In the nasopharyngeal cavity of infected K18 mice at 3 DPI, we found that SARS-CoV-2 S protein co-localized with lung epithelial marker Pan cytokeratin (PanCK)²³ (Supplementary Fig. 2a). This result further demonstrates that SARS-CoV-2 initially infects the epithelial cells of the large airways in this infected K18 mice model. Immunofluorescent (IF) staining showed the co-localization of CoV2 spike protein with surface macrophages marker CD206 on lungs of sub-lethal CoV2 infected mice at 21 DPI (Fig. 1e). The co-localized cells take up about 80% among CD206 macrophages in these three sub-lethally infected mice (Fig. 1f).

a Male and female K18 mice were infected with 1 × 10⁴ TCID₅₀ (n = 3) or 2 × 10⁵ TCID₅₀ (n = 3) of SARS-CoV-2 intranasally and body weight changes were monitored. P < 0.01 vs 2 × 10⁵ TCID₅₀ by two-way ANOVA. Data are presented as mean and individual values. Total viral genomic N mRNA (b) and subgenomic N mRNA (sgmRNA) levels (c) in the lungs and brains of the K18 mice infected with a sub-lethal dose of SARS-CoV-2 (1 × 10⁴ TCID₅₀) and euthanized at 21 days post infection. The level of sgmRNA in the various organs were measured by qRT-PCR. d H&E images of lung sections show persistent, patchy pneumonia in the sub-lethal dose-infected K18 mice at 21 DPI. e, f Representative immunofluorescence staining of CD206 (green) and SARS-CoV-2 spike protein(red) in lungs from sub-lethal SARS-CoV-2 dose-infected K18 mice at 21 DPI. Nuclei are stained with DAPI (white). Staining demonstrates CD206 expression with (arrows) and without (arrowhead) SARS-CoV-2 spike protein. f Quantitative analysis of co-localization of CD206 macrophages and spike protein in the lungs shows the majority of SARS-CoV-2 positive cells in all three animals are CD206+ macrophages based on absolute number and percentage of total SARS-CoV-2 positive cells.

Lung transcriptomic analysis of acute and chronic CoV2 infection in K18 mice

To determine the transcriptional changes from acute, severe disease to chronic lung injury, we collected RNA from the lungs of CoV2-infected K18 mice at 21 DPI and performed bulk RNAseq analysis. K18 mice inoculated with phosphate-buffered saline were used as naïve control. Results were also compared with RNAseq analysis from K18 mice infected with a lethal CoV2 inoculum at 4 DPI from our previous study¹⁵. Compared to the naïve control, CoV2-infected K18 mice had increased expression of several collagen superfamily genes including Col1a1, Col28a1, Col5a2, Col3a1, Col15a1, Col1a2, Col5a1, Col6a3, Col4a5, Col27a1, Col8a1, Col7a1, Col16a1, Col4a4, and other extracellular matrix protein genes such as elastin gene Eln and clusterin gene Clu in lungs at 21 DPI (Fig. 2a, Supplementary Data 1, 2). Compared to infected K18 mice at 4 DPI, CoV2-infected K18 mice also had increased Col15a1 expression in the lung at 21 DPI (Fig. 2b, Supplementary Data 3). Moreover, we found evidence for higher activation of epithelial-to-mesenchymal transition pathways (Fig. 2c,d, and Supplemental Fig. 2b) at 21 DPI compared to lung tissue from naïve or infected K18 mice at 4 DPI. Furthermore, compared to lungs of naïve K18 mice, CoV2-infected K18 mouse lungs at 21 DPI also had increased activation of hallmark inflammatory response and interferon-gamma pathways in addition to increased complement pathways and Kras signaling (Fig. 2c). Since the production of collagens, elastin and clusterin play an essential role in fibrotic responses after tissue injury^24,25,26, these results indicate that sub-lethal infection with CoV2 induces a chronic inflammatory response and may mediate fibrosis in the lungs of K18 mice at 21 DPI.

Enhanced volcano plot of differentially expressed genes (DEGs) (a) and pathways (c) between sublethal SARS-CoV-2 dose-infected hACE2-K18 (1 × 10⁴ TCID₅₀) lungs (CoV K18 21 DPI) at 21DPI (n = 3) and K18 Naïve mice. Enhanced volcano plot of DEGs (b) and pathways (d) between sublethal dose SARS-CoV-2 infected K18 lungs at 21DPI (CoV K18 21 DPI) and lethal dose SARS-CoV-2-infected K18 lungs (2 × 10⁵ TCID₅₀) at 4DPI (CoV K18 4 DPI).

Lung transcriptomic analysis of CoV2 and A/PR/8/34 (H1N1)-infected K18 mice

Chronic lung injury from viral infection is not unique to CoV2. In a previous publication, pulmonary lesions were shown to persist for 60 days during experimental influenza A/PR/8/34 (H1N1) infection⁶. However, unlike the fibrotic state induced by COVID, flu-induced transcriptomic and epigenetic changes are more often associated with repair⁶. To identify unique gene signatures that may drive the lung from a reparative state to a fibrotic state, we conducted bulk RNA sequencing analysis from the lungs of K18 mice infected with either CoV2 (2 × 10⁵ TCID50) or H1N1 influenza (A/PR/8/34, 50 PFU)^6,15,27 at 4 or 6 DPI and compared them with these obtained from Naïve K18 mice. A heatmap of these findings shows that at 4 DPI and 6 DPI compared with Naïve K18 mice, both Flu and COV2 infections significantly induced a greater number of pro-inflammatory genes such as Saa3 and chemokines such as Cxcl11, Cxcl10, Ccl, Ccl2, Ccl7. We also observed that Flu infection mediated higher induction of those gene expressions than CoV2 infection (Fig. 3a,b, Supplementary Data 4, 5).

Pulmonary RNA of 2 × 10^5 TCID50 dose-SARS-CoV-2 infected K18-hACE2 mice and Influenza virus (50 PFU) infected B6 mice were collected and sequenced at 4 DPI (n = 3) and 6 DPI (n = 3). Heatmap shows the comparison of (a) Flu day 4 versus K18 naïve and SARS COV2 day 4 versus K18 naïve, gene sorted by Flu day 4 versus K18 naïve b Flu day 6 versus K18 naïve and SARS COV2 day 6 versus naïve K18, gene sorted by SARS COV2 day 6 versus naïve K18. The fold change threshold was 0–9. The top 20 genes were considered to generate all heatmaps. FC; fold change. c, e Enhanced volcano plot of DEGs (a) and pathways (c) between Flu (Flu K18 4 DPI) and SARS-CoV-2 (CoV2 K18 4 DPI) infected lungs at day 4 post infection. d, f Enhanced volcano plot of DEGs (b) and pathways (d) between Flu (Flu K18 6 DPI) and SARS-CoV-2 (CoV2 K18 6 DPI) infected lungs at day 6 post infection.

Differential gene expression was also analyzed from Flu and CoV2-infected K18 mice at 4 or 6 DPI (Fig. 3c,d, Supplementary Data 6, 7). At 4 DPI, both infections induced similar weight loss. However, compared to CoV2 infection, Flu infection significantly induced a greater number of pro-inflammatory genes (Irg1, Saa3, Il6) and chemokines (Cxcl11, Cxcl10, Ccl7, Ccl2, Cxcl2, Ccl1) (Fig. 3c, Supplementary Data 6). Pathway analysis revealed a greater induction of type I and type II interferon (IFN) pathways, as well as TNF alpha signaling via NFκB pathways in Flu infected mice compared to CoV2-infected mice at 4 DPI (Fig. 3e and Supplemental Table 2). TNF alpha signaling via the NFκB pathway and the inflammatory response pathways remained significantly elevated through day 6. In addition, Flu infection also significantly induced higher activation of P53, Kras signaling, epithelial-mesenchymal transition, and apical junction pathways compared to CoV2 infection at 6 DPI (Fig. 3f and Supplemental Table 3). Moreover, three keratin genes (Krt16, Krt15, Krt5) were also significantly induced in Flu-infected mice, suggesting epithelial changes (Fig. 3d). Of particular interest was Krt5, which is a marker for lung progenitor basal cell keratin 5, a protein that has been associated with Flu injury and repair^14,28.

Krt5+ cells are not found in pulmonary consolidation regions during CoV2 infection

The finding that CoV2 infection had substantially lower expression of Krt5 transcripts at 6 DPI prompted us to investigate whether this was due to a failure of Krt5+ cells to expand in CoV2 infection compared to influenza infection^12,13,14. We performed IF staining to compare the number of Krt5+ cells at 7, 14, and 21 DPI in CoV2 and Flu infected lungs (Fig. 4a and Supplementary Fig. 3). To monitor interstitial changes and fibrosis in consolidated lung regions, we co-stained for alpha smooth muscle actin (SMA) (Fig. 4a,b). Scattered Krt5 positive cells were found in Flu infected mice through 7–21 DPI, peaking at 14 DPI as defined by the highest percentage of proliferative Krt5+ cells and largest number of discrete Krt5+ pods (Fig. 4a–d and Supplementary Fig. 3). In striking contrast, we did not observe any Krt5+ “pods” or Krt5+ cells in consolidated areas in COVID lungs (Fig. 4a,c,d). Consolidated areas in Flu and COVID mouse lungs were shown to exhibit a mild increase in SMA expression by 14 DPI that resolves in Flu infected mice but persists at a low level by 21 DPI in CoV2-infected mice (Fig. 4a,b). We also found that Krt5+ cells express both transformation-related protein 63 (Trp63) and nerve growth factor receptor (NGFR), both pulmonary stem cell markers^12,14,29,30 (Supplementary Figs. 4a,b, 5), which further confirms the airway stem cells. Moreover, we documented that Krt5+ “pod” structures contain dense AT1 cells identified by E-Cad staining, and scattered AT2 cells identified by pro-surfactant protein C (pro-SPC) (Supplementary Fig. 6). Together, these data show that Krt5⁺ progenitor cells proliferate in Flu-infected lung but are absent in areas of injury in CoV2 infected lung.

a SMA-Krt5 double staining shows proliferation of Krt5+ cells in consolidated regions of Flu-infected lungs at 7 (n = 3), 14 (n = 4) and 21 DPI (n = 5). In SARS-CoV-2 infected lungs, consolidated pulmonary regions do not exhibit proliferation of Krt5⁺ cells (red) at 7 (n = 3), 14 (n = 3), or 21 DPI (n = 3). Comparison of the expression of SMA (b) and Krt5 (c) in regions of consolidated and normal lungs from the same animal as detailed in Supplementary Fig. 9. d Number of pulmonary regions with proliferation of Krt5+ cells (“pods”) per lung section. Data are shown as mean ± SEM. To compare values obtained from two groups, two-tailed unpaired Student’s t test was performed. ** indicates p < 0.01. * indicates p < 0.05.

We further used Picrosirius red (PSR) staining to investigate pulmonary collagen deposition between CoV2 and Flu infected lungs at different time points post infection (Fig. 5, Supplementary Fig. 7). We observed that CoV2-infected mice tended to increase in collagen deposition in the lungs as compared to Flu infected mice at 21 DPI only, but not other time intervals (Fig. 5a,b). In contrast, the flu-infected lungs had minor collagen deposition across the disease time course. The dynamics of influenza-induced collagen deposition correlate with the disease progression and recovery. Viral load peaks usually around day 5 which is followed by severe inflammation and disease (weight loss) from days 7–10. Injury peaks from 10–18 DPI and resolves 20–30 DPI (Fig. 5a,b)⁶. Collagen deposition in CoV2-infected lungs was sustained at 7, 14, 21, and 45 DPI (Fig. 5a) and 45 DPI. Similar findings were observed with trichrome staining which showed that CoV2-infected mice had significantly higher collagen deposition in the lungs than Flu infected mice (Supplementary Fig. 1e,h and Supplementary Figs. 8, 9). Together, these results demonstrate that there is a sustained collagen deposition in CoV2 but not Flu infected lungs and support the notion that sub-lethal infection with CoV2 induces persistent fibrosis in the lungs of K18 mice at 21 DPI.

a Representative image of PSR stainings showing collagen deposition in lungs of SARS-CoV-2-infected K18 mice (IN, 1 × 10⁴ TCID₅₀) at 7 (n = 3), 14 (n = 3), 21 (n = 3) and 45 DPI (n = 2) and Flu infected mice at 7 (n = 3), 14 (n = 4) and 21 DPI (n = 3) (50 PFU). b Quantification of collagen deposition in regions of consolidated and normal SARS-CoV-2-infected and Flu-infected lungs as detailed in Supplementary Fig. 7. Data are shown as mean ± SEM. Two-way analysis of variance (ANOVA) was used to compare collagen deposition level changes from 7 DPI to 21–45 DPI. One-tailed unpaired Student’s t test was performed to test the difference between two groups at one time point.

Krt5+ cell proliferation is absent in pulmonary consolidation area in CoV2 infected NHPs and human COVID-19 patients

We validated our finding with necropsied lung samples from CoV2-infected NHPs (Supplementary Table 4), and three fatal human COVID patients (case summary shown in Supplementary Table 5 and Supplementary Fig. 10). We compared the COVID NHP lungs acutely infected with a different respiratory virus, respiratory syncytial virus (RSV), to provide a comparator for Flu infected mouse model. Figure 6a shows that the acute RSV-infected NHPs at 7 DPI (left panel, n = 1) exhibited a strong Krt5 expression and cell proliferation in airways and in pulmonary parenchyma at site of consolidation with increased SMA expression. We previously reported that the CoV2-infected NHPs had typical COVID symptoms and pathological changes³¹. In acutely CoV2-infected NHPs between 14 and 30 DPI, Krt5 were only present in its normal niche within airways. We did not detect the proliferation of Krt5+ cells in regions of consolidation that displayed increased SMA expression within the pulmonary parenchyma (Fig. 6a, right panel, n = 4). Analysis of IF from three adult patients²³ showed a similar lack of Krt5+ cell proliferation and “pod” structure as well (patients 1 shown in Fig. 6b; and patients 2 and 3 shown in Supplementary Fig. 11). Using both NHPs and human samples, we demonstrated a lack of a Krt5+ progenitor cells response following acute CoV2 infection.

a Krt5⁺ progenitor cells (green) are normally identified in the basal layer of proximal conducting airways as seen in the SARS-CoV-2 infected NHP (top right). Following RSV infection in nonhuman primate (n = 1), there is proliferation of Krt5+ cells within airways and the adjacent pulmonary parenchyma which is not observed in SARS-CoV-2-infected nonhuman primates (n = 4). b Adult human COVID patient (patient 1) exhibits a similar reaction pattern to SARS-CoV-2 infected nonhuman primates with Krt5+ cells largely restricted to airways and rarely found within the pulmonary parenchyma.

scRNA-seq analysis of CoV2 and A/PR/8/34 (H1N1) infected K18 mice at 4 DPI

We further performed scRNA-seq analyses of flu-infected mice at an early phase of the infection (4 DPI). Pulmonary transcriptomic changes from Flu infection at the single-cell level were compared with those of CoV2 infection at 4 DPI, at the peak point of lung viral load when early pneumonia may appear¹⁵. The scRNA-seq analysis identified 10 distinct clusters of cells (Fig. 7a). We did not observe that Krt5 expression was increased in any of the identified 10 cell populations in CoV2 or Flu-infected mice in lungs at 4 DPI until 6 DPI (Supplementary Data 9). These results are consistent with the bulk RNA analysis of Flu and CoV2-infected mice at 4 and 6 DPI (Fig. 3a–d). The interferon-induced protein with tetratricopeptide repeats 1 (Ifit1), 2 (Ifit2), 3 (Ifit3), and the interferon-inducible chemokine Cxcl10 were significantly upregulated in all eight cell populations of Flu-infected lungs compared to the CoV2 infected lungs (Fig. 7b–e). These findings confirm the results from bulk RNA analyses and further demonstrate that CoV2 infection induced down-regulation of the interferon response pathway on all pulmonary cells but not specific cell populations. We further investigated the impact of CoV2 and Flu infection on pulmonary epithelial cells specifically in the scRNA-seq data. Transcriptomic differences and pathway activation within the pulmonary epithelial cell cluster from Flu- and CoV2-infected mice were compared at 4 DPI. There were 105 up- and 744 down-regulated genes in flu-infected pulmonary epithelial cells compared to in CoV2 infection. We observed significant up-regulation of several interferon-stimulated genes (ISGs), including Cxcl10, Ifit2, Gbp2, Rsad2, and Isg15 in Flu-infected lung epithelial cells compared to CoV2-infected cells (Fig. 8a and Supplementary Data 8). Both TNF receptor superfamily member 12 A (Tnfrsf12a), which may be involved in positive regulation of the extrinsic apoptotic signaling pathway and regulation of wound healing³², and Serpinb9, which may be related to oxidation-reduction reactions³³, were upregulated in influenza epithelial cells only and not observed in CoV2 infected mice (Fig. 8a and Supplementary Data 8). Among the down-regulated genes in Flu infected cells compared to CoV2, Sftpa1, Sftpb, Sftpc, and Sftpd, four surfactant proteins produced by type II pneumocytes³⁴, and Scgb1a1, Scgb3a1, and Scgb3a2, three cytokine-like secretory proteins of small molecular weight secreted by lung epithelial cells that serve as biomarkers for lung injury³⁵ were demonstrated in Flu-infected lungs. These down-regulated genes in Flu-infected lungs relative to CoV2 may reflect the extensive injury mediated by Flu infection. The interferon-gamma and alpha pathways, IL2-stat5 signaling pathway, apoptosis pathway, TNF α signaling via NFκB pathway, hypoxia pathway, UV response pathway, and P53 pathway were increased in the epithelial cells of Flu infected lungs compared to CoV2-infected lungs (Fig. 8b). These genes and pathways identified may contribute to the distinct lung injury and induced Krt5 progenitor cell proliferation between Flu and COVID infection-mediated lung.

a Major clusters and respective cell types for Flu and SARS-CoV-2 tissues at 4 DPI by scRNA-seq data. Uniform manifold approximation and projection (UMAP) for dimension reduction plot with major cell types of scRNA-seq. Single-cell suspensions from whole infected lungs at 4 DPI for both SARS-CoV-2 and Flu were processed and sequenced. The 8146 cells identified after being combined and processed, we identified ten major clusters including C1qa+ monocyte (n = 2,482), Endothelial (n = 1637), T cells (n = 892), Fibroblast (n = 804), S100a8+ monocyte (n = 633), Plasma (n = 280), NK cells (n = 258), Club (n = 220), and Epithelial (n = 207). b–e Expression of Ifit1, Ifit2, Ifit3, and Cxcl10 in the Fu-infected and SARS-CoV-2-infected lungs.

a Enhanced volcano plot for 4 DPI epithelial cells cluster DE genes. b Significant pathway for 4 DPI Flu-infected versus SARS-CoV-2-infected lungs (epithelial cells cluster).