Cellular Heterogeneity and Molecular Reprogramming of the Host Response during Influenza Acute Lung Injury

ABSTRACT

An exuberant host response contributes to influenza A virus (IAV) (or influenza)-mediated lung injury. However, despite significant information on the host response to IAV, the cellular framework and molecular interactions that dictate the development of acute injury in IAV-infected lungs remain incompletely understood. We performed an unbiased single-cell RNA sequencing (scRNAseq) analysis to examine the cellular heterogeneity and regulation of host responses in the IAV model of acute lung injury. At the cellular level, IAV infection promoted the overwhelming recruitment of monocytes that exhibited the cell differentiation trajectory to monocyte-derived macrophages. Together, monocytes and monocyte-derived myeloid cells constituted over 50% of the total immune cells in IAV-infected lungs. In contrast, IAV infection resulted in a significant loss of nonhematopoietic cells. Molecularly, our data show the multidimensional cell-cell communication dynamics of interferon and chemokine signaling between immune and nonimmune cells and the cell-specific molecular pathways regulating the host responses during IAV-induced lung injury. Our data provide a foundation for further exploring the mechanistic association of the IAV host response with acute lung injury.

IMPORTANCE A dysregulated host response develops acute lung injury during IAV infection. However, the pathological immune mechanism(s) associated with acute lung injury during IAV infection is yet to be elucidated. In this study, we performed scRNAseq to examine the dynamics of host responses during the peak of IAV-mediated lung injury. At the cellular level, our data reveal significant myelopoiesis predominated by monocytes and macrophages and the simultaneous disruption of the nonhematopoietic cell framework, crucial for regulating inflammation and barrier integrity in IAV-infected lungs. Molecularly, we observed a complex cellular network involving cell-cell communications and a number of unique regulons dictating the outcome of interferon and chemokine responses during peak lung injury. Our data present a unique atlas of cellular changes and the regulation of global and cell-specific host responses during IAV infection. We expect that this information will open new avenues to identify targets for therapeutic intervention against IAV lung injury.

INTRODUCTION

The resolution of influenza A virus (IAV) infection requires the coordinated function of immune and nonimmune cells that work in tandem to develop an antiviral host response in the lung. Resident and infiltrating innate immune cells mount early host defense to control the viral load and promote the development of antigen-specific CD8⁺ T cell responses that lead to the complete elimination of IAV in the lung (1). Thus, significant interdependence and cooperation among myeloid and T cells are required to resolve IAV infection in the lung. However, defects in immune regulation develop into an exuberant and dysregulated host response that promotes lung injury and compromises respiratory functions (2–6).

The resident leukocytes and nonleukocyte cells, i.e., epithelial cells, develop a chemokine gradient that recruits immune cells to IAV-infected lungs. While substantial data from mouse models show the significance of leukocytes in IAV control, how these cells develop a coordinated response that either leads to viral clearance or promotes host response-driven lung pathology remains to be determined. For instance, while CD8⁺ T cells are crucial for IAV control (7, 8), emerging evidence suggests their pathological role during severe IAV infection (9, 10). Furthermore, CD8⁺ T cells have the potential to cause bystander damage to noninfected epithelial cells (11). Similarly, the renewed emphasis on nonepithelial cells, i.e., endothelial and fibroblast cells, highlights the significant role of these cell types in regulating lung barrier integrity during IAV infection. Therefore, a more comprehensive understanding of cellular communications and the development of a pathological host response to IAV is warranted.

Recent advances in single-cell RNA sequencing (scRNAseq) technology have opened a new way to identify and characterize immune cell landscapes in health and disease (12, 13). Furthermore, the technology enables the identification of differentially expressed genes (DEGs) and molecular pathways at the single-cell level, which might drive complex pathophysiological responses to various diseases (14). To examine the cellular heterogeneity and regulation of host responses during the peak of IAV acute lung injury, we performed scRNAseq analysis on total lung cells from mock- and IAV-infected mice. We found that monocytes and macrophages represented the most predominantly recruited immune cells in IAV-infected lungs, which exhibited a lineage differentiation trajectory. Lung myelopoiesis coincided with a loss of nonhematopoietic cells, which correlated with acute lung injury in IAV-infected lungs. Finally, our data identify novel cellular communications between immune and nonimmune cells and cell-specific molecular pathways during IAV infection.

RESULTS

Macrophages and monocytes are predominant interdifferentiating immune cells in IAV-infected lungs.

To interrogate the cellular landscapes during IAV lung injury, we performed scRNAseq from the lungs of mock- and IAV-infected mice (Fig. 1A) on day 7 after IAV infection. The day 7 time point was chosen because of the peak body weight loss (Fig. 1B), lung inflammation/pathology (hematoxylin and eosin [H&E] staining of lung sections) (Fig. 1C and D), and IAV load (Fig. 1E) associated with 7 days postinfection (p.i.). A total of 7,672 cells (mock infected, 4,082 cells; IAV infected, 3,590 cells) with 27 distinct clusters were obtained and defined as seven major cell types, which were further characterized into 17 cell populations (Fig. 1F; see also Fig. S1A and Table S1 in the supplemental material), by using canonical lineage-defining markers, including endothelial cells and fibroblasts (Pecam1, Flt1, and Mfap4), B cells (Cd79a, Ms4a1, and Cd19), T cells (Cd3g and Cd3e), natural killer (NK) cells (Nkg7 and Gzma), macrophages and monocytes (Fcgr1, Ly6c2, and Mafb), neutrophils (S100a8 and S100a9), and epithelial cells (epithelial cellular adhesion molecule [EpCAM] and Tmem212) (Fig. 1G). We then compared the relative distributions of immune cell compartments between the mock and IAV infection groups.

FIG 1.

Influenza disease and scRNAseq landscapes between mock- and IAV-infected lungs. Mice were infected with 250 PFU of IAV, and bronchoalveolar lavage fluid (BALF) and lungs were aseptically isolated for further processing. (A) Schematic representation of the flow of the experiment. (Created by modifying illustrations provided by Vecteezy.com.) (B) Percent weight changes up to 14 days after IAV infection. dpi, days postinfection. (C) H&E images of lung sections (magnification, ×20). (D) Lung inflammatory burden by H&E analysis at the indicated time points after IAV infection. (E) Viral titers in lung homogenates by a TCID₅₀ assay. Data in panels B to E are shown as means ± standard errors of the means (SEM), representative of results from 2 different experiments (n = 5). Statistical analysis was performed using one-way analysis of variance (ANOVA). ****, P < 0.0001. (F) Uniform manifold approximation and projection (UMAP) embedding of single-cell transcriptomes from 7,672 cells from mock (n = 2) and IAV (n = 2) infection, annotated by cell type. (G) Dot plot of the mean expression levels of canonical marker genes for 7 major lineages from tissues of each origin, as indicated.

Among immune cells, we observed the largest proportional increase in cell clusters belonging to monocytes and macrophages (Fig. 2A, Fig. S1B, and Table S1). Further analysis revealed the heterogeneity in macrophage cell clusters, grouped as M1, M2, and M0 macrophages (Fig. 2A and B). Compared to mock infection, IAV infection resulted in a significant increase in the M1 macrophage cluster. In contrast, a contraction of the M2 cluster was observed during IAV infection. While the cell number in the M0 macrophage cluster increased during IAV infection, the changes in cell proportions did not differ between mock and IAV infection (Fig. 2A and Fig. S1B). Together, monocyte and macrophage cellular clusters represented over 50% of the total leukocytes in IAV-infected lungs. No statistical difference in the levels of NK cells, dendritic cells, and neutrophils was observed between mock and IAV infection (Fig. 2A). Flow cytometry validated the scRNAseq findings showing monocytes and macrophages as the most abundant myeloid cells in IAV-infected lungs (Fig. 2C). Consistent with the scRNAseq results, flow cytometry data revealed that IAV-infected mice had significantly increased proportions of M1 macrophages (CD11b⁺ CD38⁺ CD163⁻ CD86⁺ iNOS⁺) and monocytes (CD11b⁺ Ly6C⁺ CCR2⁺), with comparable levels of M2 macrophages (CD11b⁺ CD38⁻ CD163⁺ CD206⁺ Arg-1⁺) and neutrophils (CD11b⁺ Ly6G⁺) between mock- and IAV-infected mice at 7 days p.i. (Fig. 2C). Therefore, our subsequent analysis focused mainly on macrophages and monocytes, the most significant innate immune cell types in our model.

FIG 2.

Innate immune cell landscapes in mock- and IAV-infected lungs. Mice were infected with 250 PFU IAV, lungs were aseptically isolated, and single-cell suspensions were prepared for scRNAseq and flow cytometry. (A) UMAP plot of innate immune cells and relative cell proportions of seven subsets (M1/M2 macrophages, macrophages, monocytes, dendritic cells, neutrophils, and NK cells) from tissues of each origin, color-coded by clusters and cell subsets, as indicated. (B) Dot plot of the mean expression levels of canonical marker genes for 7 cell types, as indicated. (C) Flow cytometry of M1 and M2 macrophage, CCR2⁺ monocyte, and neutrophil cell counts in the lungs of mock- and IAV-infected mice at 7 days postinfection. Cell counts are represented as fold changes. Briefly, cells are defined as monocytes (CD11b⁺ Ly6C⁺ Ly6G⁻ CCR2⁺ cells), M1 macrophages (CD11b⁺ CD38⁺ CD86⁺ CD163⁻ iNOS⁺ cells), M2 macrophages (CD11b⁺ CD163⁺ CD38⁻ CD206⁺ Arg-1⁺ cells), and neutrophils (CD11b⁺ Ly6C⁺ Ly6G⁺ cells). Data are shown as means ± SEM, representative of results from 2 different experiments. Statistical significance was determined by two-tailed Student’s t test (n = 5). ****, P < 0.0001. (D) Pseudotime trajectories of macrophage and monocyte cells in mock and IAV infection along with Seurat clusters (top). Cells were sorted along pseudotime. Heat maps show comparisons of the percentages of cells from mock and IAV infection occupying individual states (top) and significantly enriched GO terms from differentially regulated genes of each state (bottom). Differentially expressed genes were identified as significant at an adjusted P value of <0.05 and an |avg_logFC| value of >0.25. Enriched terms were identified as significant at an adjusted P value of <0.05. NA, not applicable. (E) Heat map depicting the relative expression levels of shared differentially expressed genes (DEGs) in monocytes, M1 macrophages, and M2 macrophages between mock and IAV infection. (F) Gene set enrichment analysis (GSEA) of KEGG pathways for innate immune cells. Color stands for genes upregulated (red) or downregulated (blue) after IAV infection. Enriched terms were identified as significant at an adjusted P value of <0.05. NES, normalized enrichment score; PPAR, peroxisome proliferator-activated receptor; MAPK, mitogen-activated protein kinase; PKG, protein kinase G.

Because monocytes and macrophages exhibit overlapping functional phenotypes (15), we used the scRNAseq cell trajectory approach and reconstructed lineage differentiation between monocyte and macrophage subsets. The cells from the mock- and IAV-infected groups were ordered along pseudotime (Fig. S1D), and the vast majority of monocytes and M1 macrophages aligned in a continuum along inflammatory state 1 (Fig. 2D), characterized by the enrichment of cellular pathways such as the regulation of the apoptotic process, the response to interferon gamma (IFN-γ), and myeloid cell differentiation. Furthermore, a significant proportion of non-M1 macrophages (M0) were found in states 7, 8, and 9, associated with positive regulation (Gene Ontology [GO] annotations) of cell migration, negative regulation of cell death, and regulation of the cellular catabolic process. Monocyte and M1 macrophage populations also exhibited common differentially expressed genes (DEGs) belonging to the proinflammatory host response (Fig. 2E, Fig. S1C, and Table S2). Furthermore, the gene set enrichment analysis (GSEA) of the Kyoto Encyclopedia of Genes and Genomes (KEGG) revealed that monocytes and M1 macrophages showed significant functional overlap (Table S3) based on common cellular pathways in the two cell types associated with oxidative stress and cytokine and chemokine signaling (Fig. 2F). Overall, these data suggest that monocytes and M1 macrophages act as dominant innate immune cells during IAV infection and that these cells exhibit significant overlap in cellular differentiation and functional responses based on their quantitative transcriptomes.

Finally, we compared the T cell landscapes in mock- and IAV-infected mice. We used lineage-defining markers to annotate T cell clusters (Fig. S1E). Our data show that compared to mock infection, IAV infection resulted in only a marginal (nonsignificant) increase in the quantitative frequency of CD4⁺ T cells (Fig. S1E). In contrast, a significant increase in the CD8⁺ T cell frequency was observed during IAV infection (Fig. S1E). Furthermore, CD8⁺ T cells exhibited a potent cytotoxic response based on the expression of perforin-1 (Prf1) and granzyme b (Gzmb) (Fig. S1F). These data suggest that in a primary IAV infection model, CD8⁺ T cells develop into cytotoxic cells as early as day 7, which is crucial for controlling the viral load.

IAV causes a loss of nonhematopoietic cells in the lung.

Being that it is the target of the host response to resolve IAV infection, nonhematopoietic cells are crucial for viral clearance and the regulation of pulmonary inflammation during IAV infection. We used scRNAseq to investigate the differences in the cellular landscapes of lung nonhematopoietic cells between mock and IAV infection. Lineage-defining markers annotated six distinct cell clusters, i.e., endothelial cells, fibroblasts, myofibroblasts 1, myofibroblasts 2, pericytes, and type II pneumocytes (Fig. 3A and B). While all six cell clusters were present in mock- and IAV-infected lungs, endothelial cells represented the most predominant cell type, followed by fibroblasts and pericytes (Fig. 3A and Fig. S2C). IAV infection led to a significant contraction of all six cellular clusters. Compared to other nonhematopoietic cells, type II pneumocytes (epithelial cells) represented the smallest cell clusters with fewer cells. However, in previous scRNAseq studies, the lower cell numbers in clusters similar to ours were reliable for data analysis and interpretations (16–18).

FIG 3.

scRNAseq landscapes of nonhematopoietic cells in mock- and IAV-infected lungs. Mice were infected with 250 PFU IAV. On day 7, mice were euthanized, and lungs were aseptically isolated for immunofluorescence staining or single-cell suspension preparation for scRNAseq and flow cytometry. (A, left) UMAP plot of barrier cells color-coded by clusters and cell subsets, as indicated. (Right) Cell numbers of each cell type among mock- and IAV-infected lungs. (B) Dot plot of the mean expression levels of canonical marker genes for 6 cell types, as indicated. (C) Immunofluorescence staining of fibroblasts (top) and endothelial cells (bottom) in lung sections of mock- and IAV-infected mice at day 7 postinfection. Lung fibroblasts were stained with vimentin (red) and ColI (magenta), airway and vascular smooth muscle cells were highlighted by αSMA staining (green), and nuclei were counterstained with DAPI (blue). Shown are representative images from 2 different experiments (n = 4 per group). Images were taken at a ×40 magnification. IAV infection induces vascular inflammation with damage to the integrity of the endothelial cell layer. Endothelial cells were stained with CD31 (red), smooth muscle cells were stained with αSMA (green), and nuclei were counterstained with DAPI (blue). The loss of vascular endothelial integrity in IAV-infected lung sections, exposing the smooth muscle layer, is highlighted with white arrows. Representative images are from 2 different experiments (n = 4 per group). Images were taken at a ×63 magnification. (D) Flow cytometry of lung fibroblasts 7 days after IAV infection. Lung fibroblasts are gated as CD31⁻ EpCAM⁻ CD90.2⁺ CD140a⁺. (E) Heat map depicting the relative expression levels of the top 50 shared differentially expressed genes in endothelial and fibroblast cells between mock and IAV infection. Differentially expressed genes were identified as significant at an adjusted P value of <0.05 and an |avg_logFC| value of >0.25. (F) GSEA of KEGG pathways for barrier cells. Only the top 20 significantly changed canonical pathways in each cell type are shown. Color stands for genes upregulated (red) or downregulated (blue) after IAV infection. Data are shown as means ± SEM. Significance was determined by two-tailed Student’s t test. Data are representative of results from 2 different experiments (n = 5). ***, P < 0.005.

To validate the scRNAseq data, we used lung sections from mock- and IAV-infected mice and performed immunofluorescence staining for fibroblasts (collagen type I [ColI] and vimentin) (Fig. 3C, top), endothelial cells (CD31) (Fig. 3C, bottom), and epithelial cells (EpCAM) (Fig. S2E). The combination of ColI and vimentin has been described to be a marker for fibroblasts (19, 20). Consistent with the scRNAseq data, immunofluorescence staining demonstrated significant reductions in the protein expression of CD31 and EpCAM, indicating the reduced endothelial and epithelial cell presence in IAV-infected lungs (Fig. 3C and Fig. S2E). Furthermore, compared to mock infection, immunofluorescence staining showed damage and a loss of continuity of the bronchial airway epithelial layers in IAV-infected lungs (Fig. S2E). However, in contrast to the scRNAseq results, we observed an increased expression of ColI in IAV-infected lungs (compared to mock) by immunofluorescence staining (Fig. 3C). The flow cytometry results were in contrast to the immunofluorescence findings and supported the scRNAseq findings showing a significantly reduced number of fibroblasts in the lungs of IAV-infected mice (Fig. 3D).

Endothelial cells and fibroblasts exhibited shared expression of several key DEGs associated with interferon and chemokine signaling, cellular inflammation, and the regulation of barrier integrity (Fig. 3E; Fig. S2A, B, and D; and Table S4).

Next, we investigated cell-specific differentially regulated pathways between mock and IAV infection (Fig. 3F) by GSEA (Table S5). A dot plot was generated for the top 20 significantly changed pathways identified from the comparisons for each cell type between mock and IAV infection (Fig. 3F). Our data reveal significant functional overlap, as well as distinct cell-specific molecular responses, among nonhematopoietic cells. For instance, compared to mock infection, endothelial cells and type II pneumocytes downregulated axon guidance and fatty acid metabolism pathways during IAV infection. Interleukin-17 (IL-17) signaling was uniquely expressed in myofibroblasts following IAV infection. However, all cell types commonly upregulated the pathways associated with antiviral defense, i.e., influenza, HIV-1 infection, herpes simplex virus 1 (HSV-1) infection, Epstein-Barr virus infection, and NOD-like receptor signaling. These data suggest that IAV infection causes significant cell death of nonhematopoietic cells, serving as a potential causative link to exacerbated pulmonary inflammation and injury.

Cell-cell communications identify differences in the flow of biological information between mock and IAV infection.

The development of the host response during IAV infection manifests from complex cell-cell communications allowing the flow of biological information. Cell-cell communications could occur through either direct physical interactions or unique ligand-receptor interactions. The magnitude of these interactions determines the fates of infection resolution, lung inflammation, and injury. We detected 459 significant ligand-receptor pairs among the 17 cell groups, which were further categorized into 113 signaling pathways, including type I IFN (IFN-I), IFN-II, CD45, IL-1, and transforming growth factor β (TGF-β) (Fig. S3A and B). Compared to mock infection, monocytes and M1 macrophages showed the most significant changes in outgoing/source (ligand expression) and incoming/target (receptor expression) biological communications during IAV infection (Fig. 4A). We then compared the most dynamically changing pathways between mock and IAV infection. The pathways implicated in pattern recognition (galectin and C-type lectin-like receptor [CLEC]), inflammation (platelet endothelial cell adhesion molecule [PECAM], tumor necrosis factor [TNF], IL-1, CD45, IFN-I, SEMA7, and complement), and cellular repair and regeneration (TGF-β, collagen, and non-canonical WNT [ncWNT]) represented the top 30 most dynamically changing pathways between mock and IAV infection (Fig. 4B).

FIG 4.

Cell-cell communications between mock- and IAV-infected lungs. Mice were infected with 250 PFU IAV. At 7 days postinfection, lungs were aseptically isolated, and single-cell suspensions were prepared for scRNAseq and flow cytometry. (A) Scatterplot showing the major sources and targets of IAV infection compared to mock infection. Colors indicate the cell types. The dot size is proportional to the number of inferred links (both outgoing and incoming) associated with each cell group. (B) The overlapping signaling pathways between mock and IAV infection were ranked based on their pairwise Euclidean distances in the shared two-dimension manifold. A larger distance implies a larger difference in the communication networks between mock- and IAV-infected mice. (C) All of the significant signaling pathways were ranked based on their differences in overall information flow within the inferred networks between mock and IAV infection. The signaling pathways in red were more enriched in mock-infected mice, and the ones in green were more enriched in IAV-infected mice. (D) Heat map showing the relative importance of each cell group based on the computed network centrality measures of IFN-I and IFN-II signaling for mock and IAV infection. (E) Flow cytometry of IFN-γR expression on CCR2⁺ monocytes (top) and M1 macrophages (bottom) at 7 days postinfection. Data are shown as means ± SEM, representative of results from 2 different experiments. Statistical significance was determined by two-tailed Student’s t test (n = 5). ****, P < 0.0001.

To investigate the differences in cellular communications between mock and IAV infection, we compared the overall communication probabilities between mock and IAV infection (Fig. 4C). While most pathways were active in mock- and IAV-infected lungs, NPY, TRAIL, IL-10, PD-L2, LIGHT, CALCR, and PD-L1 represented seven core immunoregulatory pathways uniquely activated during IAV infection. A significant number of active pathways that exhibited a dynamic shift from the mock (>50) (Fig. 4C, dark orange) to the IAV (38) (Fig. 4C, green) infection state involved pathways regulating the antiviral response, cell adhesion, inflammation, apoptosis, and tissue repair (Fig. 4C). Next, we determined the cell-specific changes in the outgoing signals (ligand expression) across all significant pathways using pattern recognition analysis (Fig. S3B). Monocytes and M1 macrophages were the most significant sources of ligand expression between mock and IAV infection (Fig. S3B).

Next, we analyzed IFN-I and IFN-II cellular communications between mock and IAV infection based on ligand-receptor expression. Our data show that monocytes and M1 macrophages were most significantly associated with outgoing (ligand expression) and incoming (receptor expression) IFN-I signaling during IAV infection. In contrast, endothelial and CD8⁺ T cells were the primary sources of outgoing and incoming IFN-I signaling during mock infection (Fig. 4D). Notably, while most IFN-I interactions among mock-infected cells were paracrine, the IFN-I interactions during IAV infection were predominantly autocrine and, to a lesser extent, produced paracrine signaling to fibroblasts, monocytes, dendritic cells, and CD8⁺ T cells. In contrast to IFN-I, NK and CD8⁺ T cells were the most predominant cell types producing the IFN-II response during mock and IAV infection, respectively. M1 macrophages and fibroblasts were the recipient cells for the IFN-II response during IAV infection (Fig. 4D). Since M1 macrophages represented a dominant cell recipient of the IFN-II signal and our cell trajectory data demonstrated the monocyte-M1 macrophage differentiation fate, we analyzed IFN-II receptor (IFN-γR) expression on CCR2-expressing monocytes and M1 macrophages by flow cytometry. Compared to mock infection, the CCR2⁺ monocytes and M1 macrophages had significantly high expression levels of IFN-γR during IAV infection (Fig. 4E), supporting the scRNAseq findings. These findings suggest that monocytes and M1 macrophages are the most dynamic immune cells, which regulate cellular communications and host responses during IAV infection.

Cellular heterogeneity and regulation of chemokine signaling.

Chemokines play a crucial role in initiating the host response against IAV by regulating the recruitment of immune cells to the lung. Because monocytes, macrophages, and CD8⁺ T cells were the most abundant immune cells and correlated with lung injury, we investigated the cellular sources and regulation of chemokine signaling during IAV infection. We first determined the cellular sources of CCL and CXCL chemokine expression between mock and IAV infection. Figure 5A shows the expression levels of genes belonging to four gene families identified as DEGs (adjusted P value of <0.05 and absolute values of log2 fold-change of the average expression between the two groups |avg_log2FC| of >0.25 in at least one cell type). Although a number of cells expressed CCL chemokines, macrophages and fibroblasts exhibited a significant upregulation of CCL chemokines (Ccl2, Ccl4, Ccl5, and Ccl7) upon IAV infection (Fig. 5A). Compared to mock infection, several immune and nonimmune cells, i.e., macrophages, endothelial cells, myofibroblasts, fibroblasts, pericytes, and B cells, expressed significantly high levels of the CD8⁺ T cell recruitment factors Cxcl9 and Cxcl10 during IAV infection. Furthermore, the expression levels of the chemokine receptors Ccr1, Ccr5, and Cxcr4 were significantly higher during IAV infection (Fig. 5A).

FIG 5.

Cellular heterogeneity and host response regulation in mock- and IAV-infected lungs. (A) Dot plot showing the DEGs for chemokine and chemokine receptor signaling between mock and IAV infection. The squares indicate that the genes were identified as DEGs in a specific cell type. (B) Heat map showing the relative importance of each cell group based on the computed network centrality measures of CCL and CXCL signaling for mock and IAV infection. (C) Relative contribution of each ligand-receptor pair to the overall communication network of the CCL and CXCL signaling pathways between mock and IAV infection. (D) Heat map of regulon activity analyzed by SCENIC with default thresholds for binarization. “Regulon” refers to the regulatory network of TFs and their target genes. “On” indicates active regulons. “Off” indicates inactive regulons. The top rows represent the cell types and the group information. The numbers in parentheses next to the regulon names indicate the numbers of genes (g) enriched in regulons. (E) Gene expression levels of Stat1, Ifitm3, Irf7, and Irf8 in the lungs at day 7 postinfection, as indicated by qPCR (fold change).

Next, we investigated the dominant ligand-receptor interactions involved in the overall communication network of CCL and CXCL signaling pathways between mock and IAV infection (Fig. 5B). While neutrophils and NK cells were the primary sources (expressors) of CCL ligands during mock infection, monocytes and M1 macrophages were the most significant cells expressing CCL during IAV infection (Fig. 5B). Neutrophils also served as significant receivers of the CCL signaling pathway under the mock conditions (Fig. 5B). In contrast, monocytes and M1 macrophages served as the most significant receivers of the CCL response during IAV infection (Fig. 5B). The communication networks for CXCL signaling differed substantially from those for CCL signaling, with endothelial cells being the primary source of the CXCL response during mock infection. Neutrophils were the most prominent sources and receivers of CXCL signaling during IAV infection (Fig. 5B). We then focused on the relative contribution of each ligand-receptor pair to the overall communication network of CCL and CXCL signaling pathways (Fig. 5C). Notably, among all known ligand-receptor pairs, the CCL signaling pathway was dominated by the Ccr1 receptor interacting with Ccl6, Ccl3, and Ccl5 ligands during mock infection. However, Ccr5 was the primary receptor interacting with Ccl4, Ccl3, and Ccl5 driving CCL signaling during IAV infection. Ccl2 was the main ligand interacting with Ccr2 during IAV infection. In the CXCL signaling pathway, while the Cxcl12-Cxcr4 pair was the dominant contributor to CXCL signaling during mock infection, the Cxcl2-Cxcr2 ligand-receptor pair was the major contributor to the CXCL communication pathway during IAV infection (Fig. 5C). These findings highlight the cell-specific contributions of CCL and CXCL chemokine ligands and the unique profiles of ligand (chemokine)-receptor (chemokine receptor) interactions in regulating chemokine signaling during IAV infection.

Identification of unique transcription factors (regulons) during IAV infection.

The activation of transcriptional factors (TFs) as a result of ligand-receptor interactions determines the outcome of the cell-specific host response. We used a gene regulatory network (GRN) inference approach (21) to investigate the regulon activities of cell-specific transcriptional factors between mock and IAV infection. We identified 300 activated regulons by using single-cell regulatory network inference and clustering (SCENIC) with 10,392 filtered genes and default filter parameters among all of the cell types (Fig. S4A). The SCENIC analysis revealed that the genes regulated by the transcriptional factors Hif1a and Irf7 were the most active in all innate cells following IAV infection (Fig. 5D). However, the genes regulated by the transcriptional factors Brf2, Tgif1, and Taf7 were downregulated during IAV infection (Fig. 5D). Transcriptional factors such as Bach1, E2f4, and Hcfc1 were associated primarily with the macrophage response during IAV infection. Compared to mock infection, IAV infection led to a significant upregulation of endothelial cell-specific Stat2, while the transcriptional factor Tef was substantially reduced following IAV infection (Fig. S4A and B). Quantitative PCR (qPCR) validated the scRNAseq findings showing the significant upregulation of interferon-regulated genes, such as Stat1, Irf7, Irf8, and Ifitm3, during IAV infection (Fig. 5E). These data identify the crucial differences in regulon activities between mock and IAV infection regulating the development of the host response.

DISCUSSION

In this work, we utilized an unbiased scRNAseq analysis to investigate the cellular dynamics of host responses in an IAV infection model of significant weight loss and lung pathology. The findings reveal that monocytes and monocyte-derived macrophages are the most predominant immune cells, coinciding with the loss of nonhematopoietic cells and increased lung injury during IAV infection. Furthermore, our data show the dynamics of cellular communications between hematopoietic and nonhematopoietic cells and the unique regulons associated with cell-specific host responses during IAV infection. Our findings highlight the application of scRNAseq to establish an atlas of the cellular heterogeneity and dynamics of host responses during IAV lung injury.

Monocytes, macrophages, and CD8⁺ T cells constitute a crucial antiviral apparatus against IAV (22–24). Our data demonstrate that monocytes and M1 macrophages are the two most predominant innate cell types in IAV-infected lungs, which follow a common cell lineage differentiation trajectory with an overlapping host response phenotype. Molecularly, monocytes and M1 macrophages exhibited significant overlap in host response regulation, especially the activation of transcriptional factors associated with interferons (Irf7), HIF-1, and oxidative stress. The increase in monocyte and M1 macrophage cell clusters coincided with a contraction of the M2 macrophage cell cluster, suggesting that IAV suppresses anti-inflammatory responses to facilitate lung injury. Furthermore, the positive correlation between the quantitative levels of CD8⁺ T cells and myeloid cells (day 7 p.i.) suggests a relationship between CD8⁺ T cells and myeloid cells in the orchestration of IAV host response-driven lung injury, which warrants further exploration. The balance between M1 and M2 macrophages is crucial for IAV clearance and the regulation of lung inflammation (25). The reduced M2 macrophage response in our model suggests that IAV creates an inflammatory loop dominated by monocytes, M1 macrophages, and CD8⁺ T cells that promote exuberant inflammation by suppressing anti-inflammatory M2 macrophages at the peak of IAV lung inflammation (7 days p.i.). In our IAV model, we did not observe differences in the frequencies of NK cells or neutrophils (compared to mock infection). The neutrophil data in our model are supported by a recent study showing that IAV-mediated suppression of the neutrophil response is a likely causative link to enhanced susceptibility to secondary Aspergillus fumigatus infection (26). However, some reports have also shown the crucial role of NK cells and neutrophils during IAV infection (27). Because our data are representative of day 7 after IAV infection, we do not rule out the role of NK cells or neutrophils at earlier time points after IAV infection. We speculate that these differences are likely attributable to the differences in the IAV infection doses and time points of investigation.

Nonhematopoietic cells are crucial for regulating lung inflammation during infections and inflammatory diseases (28, 29). We compared the relative cellular frequencies and host responses in nonhematopoietic cells between mock and IAV infection. Of six nonhematopoietic cell clusters in mock-infected lungs, endothelial cells and fibroblasts were the most abundant cell types, and IAV infection caused a significant contraction of endothelial cells and fibroblast cell clusters. A protective role of endothelial cells and fibroblasts has been shown in the regulation of lung inflammation and the maintenance of barrier integrity during IAV infection. Because the lung is a heavily vascularized organ for gas exchange (30, 31), endothelial cells, in particular, are crucial for maintaining vascular integrity, which is vital for gas exchange and the maintenance of lung function. Our findings support previous work showing a significant loss of endothelial cells and fibroblasts during peak IAV lung inflammation (32). Furthermore, our data show dominant interferon signaling in endothelial cells and fibroblasts correlated with cellular injury and a disrupted lung barrier. The pathological effect of interferon signaling on lung epithelial cells has been reported in mouse models of influenza infection (33). Our data further suggest that interferon signaling likely modulates the function of nonepithelial (nonhematopoietic) cells, impacting lung inflammation and barrier integrity during IAV infection.

The development of the host response is mediated by complex cell-cell communications via ligand-receptor interactions that dictate the strength of the biological signal during infections or inflammatory diseases. A balanced inflammatory response that favors IAV clearance and tissue repair is central to restoring inflammation-induced lung injury and homeostasis (34). However, despite significant information on the influenza host response and inflammation, there remains a knowledge gap on the cellular pathways and molecular events that dictate the outcome of immune-mediated lung pathology during IAV infection. Our cell-cell communication data show that monocytes and M1 macrophages are major recipients of IFN-I and IFN-II signaling (based on the upregulation of IFN-I and IFN-II receptors) during IAV infection. These data importantly suggest that the inflammatory contributions of IFNs to IAV lung injury are mediated, at least partly, by their ability to regulate the inflammatory phenotype of myeloid cells. Our findings on cell-cell communication and information flow further show that inflammatory pathways such as IFN-I, IFN-II, major histocompatibility complex class I (MHC-I), Toll-like receptors (TLRs), and IL-1 represented the top 30 enriched pathways during IAV infection. In contrast, cellular pathways involved in tissue repair, such as TGF-β, WNT, and Notch, were significantly downregulated during IAV infection.

Our data have limitations in deciphering the relative contribution of cell-specific host responses driven by the viral load or by a bystander effect. Steuerman et al. demonstrated differences in cell-specific host responses between IAV-infected and uninfected (bystander) lung cells based on dual-transcriptome analysis of viral and host single-cell mRNA profiles of influenza-treated lungs (22). They chose an early infection time point (48 or 72 h) and showed that all major immune and nonimmune cell types manifested substantial fractions of infected cells albeit at low viral transcriptome loads relative to epithelial cells. While early infection time points are crucial for investigating the role of the viral load as an initial trigger of the host response, the focus of our study was on the host response during peak lung injury (day 7). In our model, the heightened inflammatory response and body weight loss at day 7 p.i. are likely manifestations of the combined effect of the viral load (infected cells) and a bystander response triggered by noninfected cells (26).

Collectively, our findings establish the differences in cellular landscapes and host response regulation between mock and IAV infection. We show that IAV infection promotes the recruitment and development of monocytes and monocyte-derived cells while simultaneously causing the loss of nonhematopoietic cells, which disrupts lung barrier integrity. Finally, our data identified several regulatory pathway networks uniquely regulated during IAV infection that can create a foundation for further exploring the mechanistic relationship between the host response to influenza and acute lung injury.

MATERIALS AND METHODS

Influenza infection model.

C57BL/6 mice were purchased from The Jackson Laboratory and bred in-house. Equal proportions of 6- to 8-week-old male and female mice were included in all experiments. To establish IAV infection, mice were lightly anesthetized with a 4% (vol/vol) isoflurane-oxygen mixture, intranasally inoculated with 250 PFU of influenza A virus (IAV) (A/PR/8/34 [PR8]) (Charles River) in a 50-μL volume, and monitored for body weight loss and signs of morbidity. At the indicated time points (days 1 to 14), mice were euthanized by CO₂ exposure followed by cervical dislocation and processed for downstream applications.

Histopathology.

Lung sections were prepared, processed, and stained by the Histology Core, Department of Biomedical Sciences, University of North Dakota. Whole lungs were perfused and fixed with 10% formalin for 24 h before being transferred to 70% ethanol prior to processing. Tissues were embedded in paraffin and sectioned into 5-μm sections. Each lung specimen was stained with hematoxylin and eosin, and inflammation was evaluated by three pathologists, in a blind fashion, on a scale of 0 to 4 with increments of 0.5. For scoring, mock-infected lungs were defined as having a score of 0, and lungs with severe cell infiltration and tissue pathology were defined as having a score of 4 (35).

Flow cytometry.

The lungs were aseptically collected from mock- and IAV-infected mice, and single cells were prepared as previously described (35). One million cells were stained with Ghost Dye-Brilliant Violet 510 (BV510) (Tonbo Biosciences, San Diego, CA) and anti-mouse CD16/CD32 (BD Biosciences, San Jose, CA) at 4°C for 30 min and then stained for surface markers with CD11b-allophycocyanin (APC)/Cy7, Ly6C-BV711, Ly6G-fluorescein isothiocyanate (FITC), CCR2-phycoerythrin (PE)/Cy7, F4/80-APC, CD3-APC/Cy7, IFN-γR–BV421 (BD Biosciences, San Jose, CA), CD11c-peridinin chlorophyll protein (PerCP)/Cy5.5, CD38-FITC, CD86-PE/Cy5, CD206-PE/Cy7, CD163-BV421, CD8-PE/Cy7, and CD4-BV711 for 30 min at 4°C. For M1 and M2 macrophages, cells were further fixed and permeabilized using a BD Cytofix/Cytoperm kit according to the manufacturer’s instructions (BD Biosciences) and stained with inducible nitric oxide synthase (iNOS)-PE and arginase 1-APC (R&D Systems, Minneapolis, MN). Unless otherwise specified, all fluorochrome-labeled antibodies were purchased from BioLegend (San Diego, CA). To quantify lung fibroblasts, lung single cells were stained with CD45-BV605, CD31-PE, EpCAM-APC, CD90.2-pacific blue (PB), and CD140a-PE/Cy7 antibodies. Lung fibroblasts were gated as CD45⁻ CD31⁻ EpCAM⁻ CD90.2⁺ CD140a⁺ cells as previously described (23, 24). The complete list of antibodies used for flow cytometry (Table S6) and the gating strategy for M1 and M2 macrophages and fibroblasts (Fig. S5 and S6) can be found in the supplemental material. A BD FACSymphony cytometer was used to acquire 100,000 events, and data were analyzed using FlowJo (TreeStar). For the generation of flow cytometry t-distributed stochastic neighbor embedding (t-SNE) using FlowJo, cells were gated for singlet and live cells before downsampling to 15,000 events per sample. t-SNE was generated using an iteration number of 1,000, a tradeoff of 0.5, and a perplexity of 30 (36).

Immunofluorescence staining.

Formalin-fixed and paraffin-embedded lung sections from mock- and IAV-infected mice were prepared as previously described (35) and probed for epithelial or endothelial cell detection using anti-mouse alpha-smooth muscle actin (αSMA) (1:10,000) (Sigma-Aldrich, Darmstadt, Germany), rabbit anti-mouse CD31 (1:50) (Abcam, Cambridge, UK), or rabbit anti-mouse EpCAM (1:50) (Abcam, Cambridge, UK) antibodies (Table S7). To detect lung fibroblasts, lung sections were stained with rabbit anti-mouse collagen type I (ColI) (1:100) (catalog number AB765P; Millipore-Sigma) and chicken anti-mouse vimentin (1:500) (catalog number AB5733; Millipore-Sigma). Tissues were incubated with the corresponding secondary antibodies goat anti-mouse IgG2a-Alexa Fluor 488 (AF488) (1:200) (catalog number A-21131; Thermo Fisher), anti-rabbit IgG-AF546 (1:200) (Invitrogen, Carlsbad, CA) or anti-IgG-AF633 (catalog number A-21070; Thermo Fisher), and anti-chicken IgG-AF594 (1:200) (catalog number A-11042; Thermo Fisher) in 5% goat serum–phosphate-buffered saline (PBS) for 1 h at room temperature (RT), and nuclei were counterstained with 4′,6-diamidino-2-phenylindole (DAPI). All of the images were acquired using an Olympus total internal reflection fluorescence microscope containing a Hamamatsu Orca-Flash4.0 camera.

Viral load.

A 50% tissue culture infective dose (TCID₅₀) assay was performed as previously described (37, 38). Briefly, Madin-Darby canine kidney (MDCK) cells were seeded into a 96-well plate at a density of 1.5 × 10⁵ cells/mL in Dulbecco’s modified Eagle’s medium (DMEM) and incubated overnight at 37°C in a 5% CO₂ incubator. Meanwhile, lung tissue (partial lobe of the right lung) homogenates were prepared and stored until cells reached 90% confluence. Once confluence was reached, 10-fold serial dilutions of the lung tissue supernatants were prepared in infection medium and added to empty wells containing confluent cells. The plates were incubated, as described above, for 3 to 6 days (or until the observation of cytopathic effects). Once observed, cells were fixed in 10% paraformaldehyde, washed with PBS, stained with crystal violet, and analyzed for the TCID₅₀ using the Spearman-Karber method, as described previously.

scRNAseq.

Seven days after IAV infection, mice were euthanized, lungs were perfused, and whole lungs were excised and immediately placed into 10% fetal bovine serum (FBS) in RPMI 1640. After collection, lungs were minced in medium and mixed with 5.0 mL of digestion buffer containing 5% FBS, 1.0 mg/mL collagenase, and 0.5 mg/mL DNase. After 30 min of incubation at 37°C, cells were passed through a 70-μm cell strainer and spun at 300 × g for 7 min, and the cell pellets were resuspended in 1 mL of ACK lysing buffer for 3 min. After lysis, 10 mL of 10% FBS–RPMI 1640 was added to the samples, and the samples were washed twice. Single-cell suspensions of mock (n = 2 [1 male and 1 female])- and IAV (n = 2 [1 male and 1 female])-infected lungs from two independent experiments were counted, and 2 × 10⁶ cells were resuspended in medium containing 10% dimethyl sulfoxide (DMSO) and 20% FBS and allowed to slow freeze in Mr. Frosty. 3′ single-cell gene expression libraries (v3.0) were constructed using the 10× Genomics Chromium system. Single-cell library preparation was done by Singulomics Corporation (https://singulomics.com/). Paired-end 150-bp sequencing was performed to produce high-quality data on an Illumina (San Diego, CA, USA) HiSeq platform. Therefore, two rounds of 3′ single-cell library preparation and paired-end sequencing were performed for the study.

scRNAseq data integration and cell trajectory construction.

Reads with sequence quality of <30 were filtered out and then mapped to the GRCm38 mouse reference genome using the CellRanger toolkit (version 3.0.2). Individual sample output files were read in Seurat v3 (39), and cells of low quality were further excluded with the following criteria: a gene number of between 200 and 7,000 and a unique molecular identifier (UMI) count of >1,000. Next, data from mock and IAV treatments were integrated, and the LogNormalize method was used for data normalization. Dimension reductions, including principal-component analysis (PCA), clustering, and uniform manifold approximation and projection (UMAP) (resolution = 0.8), were then performed on the integration data (39). FindAllMarkers was used to determine the cluster-specific expressed genes. Cell types were then identified by using canonical lineage-defining markers included in CellMarker and PanglaoDB, and the cell clusters annotated as the same cell type were merged. Differentially expressed genes between mock and IAV infection for each cell type were identified by MAST (30). Genes were considered differentially expressed if the adjusted P value was <0.05 and the average log₂ fold change (FC) was >0.25. Kyoto Encyclopedia of Genes and Genomes (KEGG) and Gene Ontology (GO) functional enrichment analyses were performed by using the richR package (https://github.com/hurlab/richR/), and an adjusted P value of <0.05 was chosen as the cutoff value to select significant KEGG or GO terms. The single-cell trajectory analysis was performed using the Monocle package (31) for constructing potential lineage trajectories in pseudotime based on the differentially expressed genes among related single cells. The cell states and the branch points were identified, and the biological functions (GO) were determined based on the specific expressed genes identified for each cell state. The proportions of each cell type for each cell state were calculated for mock and IAV infection.

Cell-cell communication and regulatory network inference.

To understand global communications among cells, CellChat (version 0.0.1) was used to investigate the dynamics of cellular communications during mock and IAV infection (25). Briefly, significant ligand-receptor pairs among 17 cell groups were identified and categorized into signaling pathways. Second, the key incoming and outgoing signals were predicted for specific cell groups as well as global communication patterns by leveraging pattern recognition approaches. Next, the significant signaling pathways were grouped by defining similarity measures and performing manifold learning from topological perspectives. The overall communication probability analysis was performed across the mock and IAV infection data sets.

Single-cell regulatory network inference and clustering (SCENIC) (21) was performed to assess the regulatory network analysis with regard to transcriptional factors and to discover regulons in individual cells. The raw count data were extracted from the Seurat object and filtered using the default parameters, finally resulting in 10,932 mouse genes mapped in the RcisTarget database (21). The coexpressed gene network for each TF was constructed using GENIE3 software (34), and the regulons, groups of target genes regulated by a common transcription factor, were generated using the runSCENIC procedure based on the correlation between the transcriptional factors and the potential targets. Finally, regulon activity was analyzed by using AUCell (area under the curve) software (21) with the default threshold applied to binarize the specific regulons, and transcriptional factor expressions were projected onto UMAP plots.

qPCR.

Total RNA was extracted from lung tissues using the RNeasy Plus minikit (Qiagen, Valencia, CA). cDNA was then synthesized by using a SensiFast cDNA synthesis kit (Bioline, Taunton, MA), and PowerUP SYBR green master mix (Life Technologies) was used for the PCR template. After normalization to the housekeeping gene (β-actin) for each sample, the changes in specific gene expression and fold differences were calculated using the 2^−ΔΔCT method.

Statistics.

Two-tailed Student’s t test was used to assess 2 independent groups, with a P value of <0.05 being considered significant. Genes and pathways were considered significant when the adjusted P values (controlled for multiple comparisons using the Benjamini-Hochberg false discovery rate [FDR] procedure [40]) were <0.05.

Study approval.

Animal care and experimental protocols were performed in accordance with the NIH Guide for the Care and Use of Laboratory Animals (41) and approved by the Institutional Animal Care and Use Committee (IACUC) at the University of North Dakota (protocol number 1808-8).

Data availability.

The data were deposited in the NCBI BioProject database (accession number PRJNA733762).