Distinct antiviral signatures revealed by the magnitude and round of influenza virus replication in vivo

Significance

Influenza A virus has a broad cellular tropism in the respiratory tract. Infected epithelial cells sense the infection and initiate an antiviral response. Here, we used single-cycle replication reporter viruses to analyze the early cellular response to influenza infection in vivo. This approach revealed distinct tiers of antiviral responses that were associated with the magnitude of virus replication. We also unveiled disparate protection of epithelial cell types mediated by interferon during virus spread. These results demonstrate the early landscape of virus–host interactions in vivo with the magnitude and round of replication revealing distinct antiviral signatures and responses.

Abstract

Influenza virus has a broad cellular tropism in the respiratory tract. Infected epithelial cells sense the infection and initiate an antiviral response. To define the antiviral response at the earliest stages of infection we used a series of single-cycle reporter viruses. These viral probes demonstrated cells in vivo harbor a range in magnitude of virus replication. Transcriptional profiling of cells supporting different levels of replication revealed tiers of IFN-stimulated gene expression. Uninfected cells and cells with blunted replication expressed a distinct and potentially protective antiviral signature, while cells with high replication expressed a unique reserve set of antiviral genes. Finally, we used these single-cycle reporter viruses to determine the antiviral landscape during virus spread, which unveiled disparate protection of epithelial cell subsets mediated by IFN in vivo. Together these results highlight the complexity of virus–host interactions within the infected lung and suggest that magnitude and round of replication tune the antiviral response.

Influenza A virus (IAV) drives significant morbidity and mortality worldwide each year. IAV has a broad cellular tropism in the respiratory tract with the ability to infect many epithelial cell types (1). Rig-I–like receptors detect virus in epithelial cells, resulting in the production of type I and III interferons (IFNs) and other proinflammatory cytokines (2). IFNs act through autocrine and paracrine signaling pathways to induce the production of IFN-stimulated genes (ISGs), which promote a general antiviral state. Several individual ISGs have been identified that perturb IAV at multiple stages of the viral life cycle. For example, IFITM3 blocks entry, Mx disrupts the IAV polymerase, and PKR inhibits viral protein synthesis (3–7). The induction of an antiviral state can also be driven directly by virus replication, independent of IFN signaling (8–10). It is unknown how the level of virus replication within a single cell affects the induction of global cellular responses. Even in the presence of a robust antiviral response, some infected cells continue manufacturing new viruses and naive cells still become infected. How the antiviral response alters tropism during virus spread has not been determined.

Studies aimed at determining the cellular response to IAV infection have been performed by exploiting powerful genetic systems (CRISPR, RNAi, yeast two hybrid, etc.) in vitro or by assessing bulk infected tissue in vivo (7, 11–13). While these analyses have been critical for evaluating host factors that support or inhibit IAV, the understanding of the complex interplay between different cell types, anatomical locations, and immune responses in the context of virus infection in vivo is still incomplete. Single-cell analyses can help bridge this gap and have demonstrated the heterogeneity in IAV replication and the antiviral response in vitro (14–16). Unfortunately, current single-cell mRNA-seq strategies using WT virus cannot distinguish between newly infected cells and cells in which replication has been controlled in vivo.

To overcome these limitations, we engineered a reporter virus to specifically label cells in the first round of replication. This virus cannot spread; therefore, any differences in viral abundance will be a direct result of replication intensity. Infection of mice revealed uninfected cells and cells with both low and high levels of virus replication. These populations exhibited unique ISG signatures, and this finding was corroborated through the use of a reporter virus capable of specifically detecting active replication. This suggests that the antiviral response is tuned to the level of virus replication to generate a response appropriate to the level of threat. To understand how the antiviral response and tropism change from the first to second wave of replication we sequentially infected mice with viruses incapable of spreading. This strategy uncovered differential protection of ciliated epithelial cells mediated by IFN. These data demonstrate that epithelial cells supporting high or low levels of replication in vivo display tailored antiviral responses and that protection afforded by IFN is not equal among all cell types during virus spread. Together these findings demonstrate the complexity of virus–host interactions in vivo and illustrate how the cellular response is tuned to the level and round of replication.

Results

Single-Cycle Infection Reveals IAV Replication Heterogeneity in Vivo.

To determine the infected cell landscape during the first round of IAV replication in vivo we engineered a reporter virus incapable of disseminating. The hemagglutinin (HA) ORF of IAV was replaced by mCherry while preserving complete HA vRNA 3′ and 5′ packaging signals and virus grown in a HA-complementing cell line (17, 18). The virus cannot produce de novo HA protein and assemble new virions that can infect other cells [thus termed single-cycle IAV (scIAV)]. Infection of mice revealed three distinct populations of lung epithelial cells: those with no, low, or high mCherry fluorescence (Fig. 1A). The heterogeneity in mCherry expression suggests that virus polymerase activity during the early stages of scIAV infection varies from cell to cell. Both low and high mCherry populations were observed at similar ratios at both 12 and 24 hours postinfection (hpi) (Fig. 1A) and required de novo virus polymerase activity (SI Appendix, Fig. S1 A and B). Both mCherry low and high CD24^high and podoplanin⁺ (pdpn) [ciliated epithelial cells and type I alveolar (TIA) cells, respectively] were identified, suggesting cell type does not drive replication heterogeneity (SI Appendix, Fig. S1C).

Fig. 1.

Heterogeneity in replication levels of IAV in epithelial cells in vivo. (A) Mice were infected with 10⁵ pfu of scIAV-ctrl (Left) or scIAV-mCherry, and live CD45⁻CD31⁻ cells were analyzed for mCherry expression at 12 (Middle) or 24 (Right) hpi. Data are representative of 2 (12 hpi) or 10 (24 hpi) experiments with three to four mice per group. (B) Naive mice (Left) or mice infected with scIAV-mCherry (Middle and Right) and analyzed for mCherry low and high cells by histocytometry. Images are representative of two to three experiments with three mice per group and two sections per lung. (C) Mice were coinfected with 10⁵ pfu of scIAV-mCherry and 10⁵ pfu of scIAV-GFP. Live CD45⁻CD31⁻ cells were analyzed for GFP and mCherry expression 24 hpi (Left). Total mCherry⁺, mCherry-high, mCherry-low, and GFP⁺mCherry⁺ cells were quantified (Right). Data are representative of three experiments with n = 3–4 mice per group. (D) Mice were infected with 10⁵ pfu of scIAV-mCherry and virus from lungs was titered on Madin-Darby canine kidney-HA cells at the indicated hpi.

Replication disparity could be driven by anatomical location and/or proximity to other infected cells. To address this, mice were infected with scIAV-mCherry and lungs were analyzed by fluorescence microscopy. We detected mCherry low and high cells in both large and small airways and did not observe any restriction based on proximity to other infected cells (Fig. 1B and SI Appendix, Fig. S1D). To further define tropism during the first round of infection, we determined the number of mCherry low and high club cells (CC10) and type II alveolar cells (SPC). We found similar ratios of mCherry low to high cells in both cell types (SI Appendix, Fig. S1 E–H), suggesting that heterogeneity in early virus replication levels is not driven by cell type.

Multiple IAV particles can infect a single cell, which could drive differences in fluorescence intensity between cells. To address this, mice were infected with a mixture of scIAV-mCherry and scIAV-GFP. At 24 hpi there was only a small percentage of mCherry-GFP double positive cells (Fig. 1C). The difference in fluorescence intensity between the low and high populations is ∼25-fold (SI Appendix, Fig. S1I), further suggesting that infection with multiple particles is not driving the disparity. To determine whether the range of replication is dependent on virus dose, mice were infected with a 10-fold lower inoculum of scIAV-mCherry and cells supporting low and high replication were still observed (SI Appendix, Fig. S1J).

Prolonged presence of infectious particles in the lungs could result in varied time of infection and lead to replication disparity. This is unlikely, given the speed of IAV entry in vitro and that virus dose and duration of infection did not impact heterogeneity in single-cell analyses (15, 19, 20). However, the half-life of an infectious particle in vivo has never been experimentally determined. We exploited the scIAV system to address this question. Because scIAVs cannot spread, only viruses that have not yet entered a cell can be detected. Mice were infected with scIAV-mCherry and virus titer from the lungs was determined at 6, 12, and 24 hpi. By 6 hpi, only ∼6% of the virions delivered in the initial dose were detectable in the extracellular lung environment with a half-life of ∼1.7 h (Fig. 1D). This is consistent with mathematical modeling experiments, which predicted a half-life between 0.6 and 3 h (21, 22). Therefore, the vast majority of virions will have entered cells before the induction of the early innate immune responses.

High Levels of Replication Reveal a Distinct Antiviral Gene Signature.

To investigate the intracellular responses to variable levels of IAV replication, we profiled the transcriptomes of CD45⁻CD31⁻ mCherry negative, low, and high cells by mRNA-seq. Reads that mapped to the IAV genome were markedly higher in the mCherry high population compared with low or negative (Fig. 2A), validating the use of mCherry expression level as a surrogate for virus replication. Failure to package or express all eight segments or internal truncations of segments can result in attenuated replication (23–25). Normalized read counts between segments and read abundances across segments were similar between mCherry low and high cells (SI Appendix, Fig. S2 A and B), suggesting that we are detecting bona fide infected cells. Multidimensional scaling (MDS) of mouse transcripts demonstrated significant differences between mCherry low and high cells across the first two dimensions (Fig. 2B). Importantly, the negative population was derived from the same lung as mCherry⁺ cells and was subjected to the same inflammatory environment, making it an effective control for determining virus replication-specific gene signatures. Global changes in the transcriptional response of mCherry low and high compared with mCherry negative were analyzed by gene ontology (GO) enrichment. mCherry low and high cells down-regulated many of the same pathways, primarily those involved in cell adhesion, extracellular matrix, and development (SI Appendix, Fig. S2C). Cell death pathways were increased in mCherry high cells, while DNA replication and cell cycle pathways were up-regulated in mCherry low cells (SI Appendix, Fig. S2C). To determine whether cell cycle was associated with lower amounts of replication, we pulsed mice with BrdU, infected them with scIAV-mCherry, and analyzed them at 24 hpi. There was no increase in BrdU⁺ cells in the infected cell population, and BrdU⁺ cells made up only 0.6% of the mCherry low population (SI Appendix, Fig. S2D), suggesting that entry into the cell cycle is not driving restriction of viral replication.

Fig. 2.

Unique transcriptional signatures in cells supporting low and high levels of replication. Mice were infected with 10⁵ pfu of scIAV-mCherry and live CD45⁻CD31⁻ mCherry negative, low, and high cells were profiled by mRNA-seq. (A) IAV cpm. (B) MDS of naive and mCherry negative, low, and high cells based on host mRNA reads. (C) Heatmap of 221 ISGs differentially expressed in the indicated populations. Cutoff of false discovery rate (FDR) is <0.05.

Cells with low or high levels of virus replication may differentially activate antiviral pathways. To determine the antiviral response, we analyzed the 221 differentially expressed ISGs and revealed several distinct groups of ISGs that varied with the level of virus replication (Fig. 2C). Cluster 2 ISGs were highly induced in mCherry negative and low cells but not in mCherry high cells. This cluster included several genes with known antiviral activity against IAV, including Eif2ak2 (PKR), Trim56, and Pml (3, 26, 27). Cluster 4 was induced only in mCherry high cells (Fig. 2C), suggesting that high levels of virus replication may produce a distinct antiviral response. Levels of Ifnar1 were similar in mCherry low and high cells (SI Appendix, Fig. S2E). Additionally, Mx1, which is IFN signaling dependent (8), was induced to a similar degree in mCherry low and high cells (SI Appendix, Fig. S2E), suggesting that differential extracellular IFN signaling alone may not drive the disparity in ISG expression. Expression of Irf7 was similar in all infected cells while Irf3 was decreased in mCherry high cells, suggesting that induction of these critical sensors is not driving the ISG disparity (SI Appendix, Fig. S2E). However, Ifnb was higher in the mCherry high cells than in other cell populations (SI Appendix, Fig. S2E). Overall, our data show distinct ISG signatures within cells with low or high virus replication.

Active Virus Replication Imparts Specific Antiviral Responses.

We hypothesized that the ISGs uniquely expressed in mCherry high cells are specific to conditions of unchecked viral replication and may represent a “last resort” antiviral effort by the host. To identify cells harboring actively replicating virus, we developed a scIAV encoding destabilized GFP (destGFP). The half-life of this protein is only 2 h compared with over 24 h for the standard enhanced GFP (28). Due to this rapid degradation, any GFP⁺ cells detected in vivo must be the result of active virus replication (Fig. 3A). Mice were infected with scIAV-GFP or scIAV-destGFP, and CD45⁻CD31⁻ lung epithelial cells were analyzed for GFP expression. Less than 15% of the total scIAV-GFP–infected epithelial cells were detected by scIAV-destGFP, suggesting that many WT GFP⁺ cells are no longer supporting active replication (Fig. 3B and SI Appendix, Fig. S3A). GFP⁺ lung epithelial cells from mice infected with WT GFP- or destGFP-expressing viruses were isolated and their transcriptomes profiled by mRNA-seq. As expected, destGFP⁺ cells contained more IAV mRNA than WT GFP⁺ cells (Fig. 3C). Cellular transcripts were analyzed by MDS and revealed significant differences in destGFP⁺ compared with WT GFP⁺ cells, primarily across the second dimension (Fig. 3D). GO analysis of down-regulated pathways in the destGFP⁺ and WT GFP⁺ populations revealed the same cell adhesion, extracellular matrix, and development pathways that were down-regulated in mCherry low and high cells (SI Appendix, Fig. S3B). This concordance suggests that cells specifically decrease certain sets of genes in response to IAV infection.

Fig. 3.

Virus expressing destabilized GFP labels actively infected cells revealing distinct transcriptional responses. (A) Model for the use of scIAV-destGFP to identify cells with actively replicating virus. (B) Mice infected with 10⁵ pfu of scIAV-ctrl, -GFP, or -destGFP and live CD45⁻CD31⁻ cells analyzed for GFP expression 24 hpi. Data are representative of three experiments with n = 3–4 mice per group. (C and D) Live CD45⁻CD31⁻ GFP⁺ and GFP⁻ cells were sorted and mRNA was mapped to the mouse and IAV genome. (C) Normalized IAV reads in each of the sorted cell populations. (D) MDS plot of the indicated populations based on mRNA reads. (E) Heatmap of the 197 differentially expressed ISGs in the indicated populations. FDR < 0.05 was used as a cutoff. (F) Venn diagram of genes from mCherry cluster 2 in Fig. 2 and GFP cluster 1 in Fig. 3. (G) Venn diagram of genes from mCherry cluster 4 in Fig. 2 and GFP cluster 4b in Fig. 3. Only genes induced to >10 cpm in at least one condition are shown.

To assess the antiviral response generated in cells supporting active replication, we analyzed ISG expression in destGFP⁺ cells compared with destGFP⁻, WT GFP⁻, and naive cells (Fig. 3E). Cluster 1 ISGs were specifically induced in WT GFP⁻ and destGFP⁻ cells. This expression pattern was similar to cluster 2 in the mCherry expression analyses (Fig. 2). We compared these two clusters and found several overlapping genes (Fig. 3F), many of which have been demonstrated to have direct antiviral activity against IAV (3, 26, 27, 29). There was also a cluster of ISGs that were only expressed in actively replicating cells (Fig. 3E, cluster 4b), analogous to cluster 4 in Fig. 3. We compared cluster 4b to the unique ISGs expressed in mCherry high cells and found a high level of concordance (Fig. 3G). These data demonstrate that cells supporting high levels of active virus replication express a distinct set of ISGs that is not expressed in other infected cells. Overall, our analyses using scIAV-destGFP recapitulated results obtained using scIAV-mCherry revealing distinct antiviral signatures in cells supporting active virus replication.

Tropism Is Altered by IFN During Virus Spread.

During an infection, IAV spreads and new cells are infected despite a local antiviral and proinflammatory response. To label cells infected after the initiation of inflammation, we employed a sequential infection strategy using scIAVs with distinct fluorophores to model virus spread in vivo (Fig. 4A). Lung epithelial cells were analyzed for mCherry expression 24 h after the second infection. There was a significant decrease in the overall number of mCherry⁺ cells in the sequential infection, indicating that the first infection induced an immune response that conferred some protection (Fig. 4 B and C). Interestingly, mCherry low and high cells were observed at similar proportions in single and sequential infection. To determine whether there is a change in tropism during the second round of replication, mCherry⁺ cells were analyzed for markers of ciliated epithelial cells and TIA cells (CD24^high and pdpn⁺, respectively). The frequency of mCherry⁺pdpn⁺ cells was similar in single and sequential infection. However, there was a significant decrease in the frequency of mCherry⁺CD24^high cells in the sequentially infected animals compared with mice infected with scIAV-mCherry alone (Fig. 4D). To elucidate the mechanisms of protection during the second round of infection, we sorted CD45⁻CD31⁻ mCherry negative, low, and high cells from sequentially infected mice for mRNA-seq analysis. These data were compared with the single-infection data shown in Fig. 2. MDS analysis of the host transcripts demonstrated significant differences between single and sequentially infected cells along the second dimension (Fig. 4E). GO analysis of the genes significantly down-regulated in both mCherry low and high sequentially infected cells demonstrated enrichment in several pathways involved in ciliated epithelial cell maintenance and development (Fig. 4F), further supporting the data in Fig. 4D. No pathways were significantly up-regulated between single and sequentially infected cells. We hypothesized that the first round of replication drives an IFN response that enhances the protection of ciliated epithelial cells, but not TIA cells. To test this, mice were treated intranasally (i.n.) with IFNβ or IFNλ and 16 h later challenged with scIAV-mCherry. TIA cells and ciliated epithelial cells were analyzed for mCherry expression. Both IFNβ and IFNλ treatment led to a significant decrease in the frequency of infected ciliated epithelial cells but did not protect TIA cells (Fig. 4G and SI Appendix, Fig. S4). These data suggest that tropism is altered during virus spread through differential IFN-mediated protection in vivo.

Fig. 4.

Tropism is altered by IFN during virus spread. (A) Model demonstrating sequential infection strategy. (B) Representative flow plot of A. (C) Numbers of total mCherry⁺, mCherry low, and mCherry high cells in single and sequential infection. (D) Frequency of total and mCherry⁺ CD24^high ciliated cells and pdpn⁺ type I alveolar cells. (E) MDS plot of single infected mCherry negative, low, and high cells from Fig. 3 and mCherry negative, low, and high cells from sequential infection. (F) Gene ontology enrichment analysis (DAVID, biological process) of the down-regulated mCherry low and high genes that were significantly different (FDR < 0.05) from single to sequential infection. Red indicates pathways overlapping in mCherry low and high cells. (G) CD45⁻CD31⁻ cells from mice treated with IFNβ and infected with scIAV-mCherry analyzed for total (black) or infected (red) pdpn⁺ and CD24^high cells. (B–D) Representative of five experiments with n = 3–4 mice per group. (G) Representative of three experiments with n = 3–4 mice per group.

Discussion

Multiple studies using single-cell sequencing have revealed heterogeneity in IAV replication and the antiviral response in vitro (14, 15). These data are consistent with previous reports demonstrating stochasticity in single-cell responses to infection (16, 30–32). Importantly, single-cell sequencing following synchronized and unsynchronized IAV infection revealed that duration and infectious dose are not responsible for the heterogeneity (15). Single-cell in vivo analysis also demonstrated heterogeneity in virus replication and antiviral responses, although the number of epithelial cells analyzed was low (33). Consistent with previous findings, our data reveal heterogeneity in virus replication levels and the cellular response to IAV during the early stages of infection in vivo. While we cannot determine whether these cellular changes are a cause or a consequence of replication disparity, they reveal that the antiviral signatures are not equal in all infected cells.

IFN-I and -III drive the expression of hundreds of ISGs, which can vary by cell type (34). A landmark series of studies demonstrated a paucity of individual ISGs that inhibit a diverse array of viruses (35, 36). IRF1, Mx, and IFITM2/3 were among the few ISGs that were able to significantly blunt IAV infection and replication (35). Additionally, IAV grown in absence of ISGs tolerates more mutations, suggesting that multiple ISGs normally constrain the virus (37). We found a constellation of known anti-IAV ISGs expressed in mCherry low and negative cells that were absent in cells supporting active or high levels of replication. While only correlative, it is interesting to speculate that these may be protective combinations of ISGs, and failure to induce these genes permits high levels of replication. In addition to host factors, replication disparity could be in part due to mutations in the viral genome that enhance resistance or susceptibility to IFN. As such, Du et al. (38) recently identified several IAV mutations that lead to IFN sensitivity. Moreover, the basal expression of one or more ISGs alone, or in combination with induced ISGs, could be driving the disparate levels of virus replication. mCherry high cells also have increased levels of Ifnb expression compared with other populations of cells in the infected lung (SI Appendix, Fig. S2E). An alternative hypothesis is that autocrine IFN signaling alone, in combination with high levels of replication or basal ISG levels, could drive the distinct ISG responses.

Our data reveal a distinct set of ISGs expressed in cells with high or active virus replication. These ISGs may represent a reserve set of genes that are only turned on in cells that fail to blunt replication. Interestingly, some of these genes are chemokines and other inflammatory mediators, which may help to orchestrate an inflammatory response to control virus spread from these cells. It may be important to only express these genes when control of replication has failed to prevent immunopathology. In addition, pattern recognition receptor usage may be an underlying mechanism of this ISG signature. While RIG-I is thought to be the primary sensor for IAV in epithelial cells, recent evidence demonstrates that MDA5 is important for the cellular response (39). High levels of replication might be needed to activate both RIG-I and MDA5 and could act as an additional level of regulation to prevent aberrant activation of proinflammatory ISGs.

Both immune and epithelial cells sense IAV in vivo (40), and responses can be dependent on the cell type. We have previously demonstrated that club cells, which survive virus replication, exhibit prolonged ISG signatures (41, 42). Our data demonstrate that IAV tropism changes over the course of infection. Ciliated cells were afforded greater protection from secondary infection compared with TIA cells. Importantly, this could only be discovered through an in vivo scIAV sequential infection strategy. While IFNβ and IFNλ can drive this selective protection, it is unknown whether IFN-mediated protection is direct or indirect through other epithelial or innate immune cells.

Here we demonstrate heterogeneity in the levels of virus replication in vivo. Using two different virus reporters we reveal that distinct sets of antiviral genes are expressed in cells harboring low and high levels of virus replication. Through a sequential infection strategy we demonstrate that virus tropism is altered during virus spread where ciliated epithelial cells have augmented protection from infection. These results demonstrate a dynamic environment within tissues that is driven by both virus replication levels and the infected cell type.

Materials and Methods

Mice and Virus Infection.

C57BL/6J mice were purchased from The Jackson Laboratory. Mice were infected i.n. with the indicated doses of scIAV. Experiments involving mice were performed as dictated by the University of Minnesota Institutional Animal Care and Use Committee.

Virus Rescue.

Viruses were rescued in 293T cells by plasmid-based transfection with IAV PR8 in the pDZ vector (43, 44). The 5′ (106 bp) and 3′ (156 bp) packaging signals were preserved, except AUG codons in the 5′ packaging signal which were mutated. The designed gene of interest [mCherry, GFP, destGFP (pCAG-GFPd2) was a gift from Connie Cepko, Harvard University, Boston, Addgene plasmid no. 14760 (45), or Cre] was cloned in between the HA packaging signals. Further details are provided in SI Appendix, SI Materials and Methods.

Flow Cytometry.

Lungs were processed as described in SI Appendix, SI Materials and Methods. Single-cell suspensions were stained with a viability stain and with the indicated directly conjugated antibodies. Further details are provided in SI Appendix, SI Materials and Methods.

Next-Generation mRNA Sequencing Analysis.

RNA was obtained and sequenced as described in SI Appendix, SI Materials and Methods. Sequencing reads were mapped to the mouse (mm10) and influenza A/PR/8/34(H1N1) genomes using Bowtie aligner (bowtie2 version 2.2.4) with local mode, -L 22 and -N 1 parameters (46). To obtain significant differentially expressed genes, the experimental groups by design were compared with control group (naive or negative) and the edgeR bioconductor package was used for statistical analysis (47, 48). Sequencing data were deposited under GEO series accession no. GSE112794. Further details are provided in SI Appendix, SI Materials and Methods.

Microscopy and Histocytometry.

Lungs were sectioned and stained with the indicated antibodies. Images were obtained on a Leica DM6000B EPI fluorescent microscope and analyzed using Imaris software. Further details are provided in SI Appendix, SI Materials and Methods.

Statistics.

Statistical analysis was executed using GraphPad Prism 7 software. Comparisons between two groups were performed using a two-tailed Student t test, and P < 0.05 was considered statistically significant. Error bars are calculated using SEM.