Tropism for ciliated cells is the dominant driver of influenza viral burst size in the human airway

Significance

Influenza virus causes severe damage to the airway and exhibits a broad cellular tropism. It is currently unknown how each of these cell types contributes to the production of new virions. Here, we used primary human airway cells differentiated under a battery of culture conditions to demonstrate that infected ciliated cells correlate with multiround viral burst. These data demonstrate the critical differences in cell-type responses that may contribute to virus success or failure within hosts.

Abstract

Influenza viruses pose a significant burden on global human health. Influenza has a broad cellular tropism in the airway, but how infection of different epithelial cell types impacts replication kinetics and burden in the airways is not fully understood. Using primary human airway cultures, which recapitulate the diverse epithelial cell landscape of the human airways, we investigated the impact of cell type composition on virus tropism and replication kinetics. Cultures were highly diverse across multiple donors and 30 independent differentiation conditions and supported a range of influenza replication. Although many cell types were susceptible to influenza, ciliated and secretory cells were predominantly infected. Despite the strong tropism preference for secretory and ciliated cells, which consistently make up 75% or more of infected cells, only ciliated cells were associated with increased virus production. Surprisingly, infected secretory cells were associated with overall reduced virus output. The disparate response and contribution to influenza virus production could be due to different pro- and antiviral interferon-stimulated gene signatures between ciliated and secretory populations, which were interrogated with single-cell RNA sequencing. These data highlight the heterogeneous outcomes of influenza virus infections in the complex cellular environment of the human airway and the disparate impacts of infected cell identity on multiround burst size, even among preferentially infected cell types.

Influenza viruses cause acute respiratory infection in humans and can drive significant pulmonary pathology, leading to respiratory distress and potentially death. Influenza A virus (IAV) has a broad cellular tropism within the lungs, primarily infecting epithelial cells including ciliated cells, secretory cells, and type 1 and 2 alveolar cells (1). Pulmonary epithelial cells also act as early immune sentinels and produce type I and III interferon as well as other cytokines and chemokines to blunt virus replication and aid in the innate and adaptive immune responses (2–4). Virus tropism can dramatically impact IAV pathogenesis. Highly pathogenic avian IAVs display broader cellular tropism, more frequently infecting alveolar cells, macrophages, and endothelial cells compared to seasonal human strains, which correlates with increased disease severity (5). Experimentally restricting tropism from endothelial cells and immune cells can reduce pathogenicity and decrease proinflammatory cytokine secretion (6, 7). In contrast, a shift in IAV tropism from airway epithelial cells to alveolar epithelial cells in mice by overexpressing fibroblast growth factor-9 results in increased morbidity and mortality (8). However, how individual epithelial cell types shape virus replication kinetics and pathogenesis remains to be elucidated.

Primary human cells from upper and lower airways can be differentiated in culture and serve as a valuable model system to study human IAV tropism, replication dynamics, and innate immunity. Differentiation of primary normal human bronchial/tracheal epithelial (NHBE) cells and normal human nasal epithelial (NHNE) cells at air–liquid interface (ALI) results in a polarized, pseudostratified epithelium with a heterogeneous mixture of ciliated epithelial cells, secretory cells, and basal cells that recapitulates the environment of the upper airways (9, 10). The cellular composition of the cultures can be altered by manipulating the differentiation conditions. Adding IL-13, a Th2 cytokine involved in antiparasitic and allergic responses, drives increased goblet cell differentiation (11). Adding an inhibitor of Notch signaling promotes ciliated epithelial cell differentiation and restricts differentiation into other cell types (i.e., secretory cells) (12–15). Media composition varies between commercial sources and can also drive differences in cell composition (16–18). Finally, there is potential donor-to-donor variability in biological features of the culture system such as mucus production (19) which may be driven by differences in cell types present in culture, though this remains to be fully characterized.

IAV-infected primary human airway cultures can support varying levels of replication, but the basis for these differences is not fully understood (20). While many cell types are susceptible to IAV infection, it is currently unknown how each contributes to the overall virus burden within the airways. We hypothesize that variation in cell type frequencies within a culture system impacts IAV replication and multiround viral burst size. Here, we rigorously test this hypothesis by measuring virus tropism and replication in over 30 separate differentiations of primary NHBE or NHNE cells using several donors, varying culture conditions to expand certain cellular subsets, and changing the duration of differentiation. Each differentiation exhibited distinct frequencies of cell types, allowing us to investigate whether even small changes in cell populations can impact IAV replication. We found that although many cell types are susceptible to infection and replication, ciliated cells are the dominant producer of new virions in upper airway cells. This finding has significant implications for understanding virus transmission biology and future therapeutic interventions.

Results

Airway Epithelial Cell Type Heterogeneity across Differentiation Conditions and Anatomical Location.

NHBE cells cultured with Pneumacult media at ALI for 3 wk to differentiate them into a pseudostratified layer of several epithelial cell types, which we evaluated using flow cytometry and immunofluorescence microscopy (Fig. 1A and SI Appendix, Fig. S1 A–F). We identified ciliated cells (SiR-tubulin⁺CD271⁻), secretory cells (SiR-tubulin⁻CD271⁻CD66c⁺TSPAN8^-), basal cells (SiR-tubulin⁻CD271⁺CD66c⁻), and goblet cells (SiR-tubulin⁻CD271⁻CD66c⁺TSPAN8⁺). We also identified several populations that are not as well defined. Cells that were negative for all the lineage-defining markers for ciliated cells, basal cells, and secretory cells (SiR-tubulin, CD271, and CD66c, respectively) were termed triple-negative cells. We also observed cells that were negative for SiR-tubulin but had intermediate levels of CD271 and variable levels of CD66c. We classified these as CD66c+ other or CD66c− other according to their CD66c levels. A Uniform Manifold Approximation and Projection (UMAP) of our flow cytometry markers showed several distinct clusters of cells, many of which overlapped with the major lineage-defining markers and populations defined by our gating strategy (Fig. 1B). We also evaluated the diversity of cell types based on anatomical location. NHNE cells cultured at ALI tended to have more ciliated and secretory cells and fewer basal cells than NHBE cultures (Fig. 1A and SI Appendix, Fig. S1A). We used RNA sequencing (RNAseq) of sorted cell type populations to examine the transcriptional signatures of ciliated, secretory, basal, triple negative, and CD66c+ other cells. Undifferentiated, unsorted cells were also included. Sorted cell populations showed distinct transcriptional profiles in a Principal Components Analysis (PCA) and clustered separately from undifferentiated cells, further illustrating these cultures were differentiating into unique cell types (Fig. 1C). We compared lung cell-type specific markers (21) to the top differentially expressed genes (DEGs) in each sorted cell type and found that triple-negative cells express ionocyte-specific markers along with pulmonary tuft cell, pulmonary neuroendocrine cell, FOXN4+ cell, and SLC16A7+ cell-specific markers at lower amounts (Fig. 1D). CD66c+ other cells expressed only basal cell-specific markers (Fig. 1D). In all, these data demonstrate our approach robustly identified phenotypic differences among cell types within a heterogeneous population of differentiated airway epithelial cells.

Fig. 1.

(A) Gating strategy (Left) for identifying epithelial cell types in NHBE and NHNE cell cultures (gated from live, singlet, Live/Dead⁻ cells). Depicts NHBE cells in Pneumacult media and analyzed 3 wk post-ALI. Stacked bar plot (Right) of average cell composition from three individual wells per differentiation/treatment at 3 wk post-ALI. Error bars indicate the SEM. Data representative of n = 7 NHBE Pneumacult differentiations and n = 2 NHNE differentiations, each with 3 to 5 wells per differentiation/treatment. (B) UMAP analysis of NHBE cells from A. Data combined from three wells. (C) PCA plot of RNAseq gene expression from sorted cell populations and undifferentiated, unsorted cells. NHBE cells (n = 1 donor) were cultured in Pneumacult media. Data are from n = 3 sorting replicates. (D) Dot plot of Top DEGs within sorted cell populations (x axis labels) and their overlap with cell-type specific markers (y axis labels) from ref. 21. Dot size indicates the number of DEGs that belonged to markers from each cell type. (E) NHBE cells grown in Pneumacult media phenotyped via flow cytometry at 0, 1, 2, and 3 wk post-ALI. Heat map depicts log10 fold change in cell types compared to week 0. n = 3 wells per time point. Data representative of two independent differentiations. (F) Heat map of log10 fold changes in cell type frequencies in NHBE or NHNE cells cultured under various conditions at 3 wk post-ALI compared to a baseline of NHBE cells cultured with Pneumacult media. Data representative of n = 7 NHBE Pneumacult, n = 6 NHBE Pneumacult + Low (1 ng/mL) IL-13, n = 6 NHBE Pneumacult + High (10 ng/mL) IL-13, n = 3 NHBE Pneumacult + 10 μM DAPT, n = 3 NHBE Lonza, and n = 2 NHNE Pneumacult treatments, from seven independent rounds of differentiations. (G) PCA biplot of NHBE and NHNE cell type frequencies 3 wk post-ALI grown in the conditions described in F. (H) MDS plot of cell type frequencies from cultures grown in Pneumacult media from seven independent rounds of differentiations with cells from five different donors.

To determine the heterogeneity of cell types present during the differentiation process, we evaluated NHBE and NHNE cultures at 0, 1, 2, and 3 wk post-air lifting (Fig. 1E and SI Appendix, Figs. S1B and S2A). At day 0 (prior to air lifting), we found SiR-tubulin⁺ ciliated cells that were also CD271⁺ (named CD271+ ciliated) and CD271⁺ basal cells that were also CD66c⁺ (named CD66c+ basal cells) in both culture types. The high frequency of basal-like cells shrank over time, with differentiated ciliated and secretory cells present by 1 wk post-air lifting. By 3 wk post-air lifting, there was a complex mixture of ciliated, secretory, basal, and other cell types. Alpha diversity of cultures increased after 1 wk of differentiation, reflecting greater cellular complexity (SI Appendix, Fig. S1G).

We next evaluated cell types present when we altered the culture conditions. We treated cells with the Th2 cytokine IL-13 or the Notch inhibitor DAPT during differentiation to drive increased mucus-producing cells or prevent the differentiation of nonciliated cells, respectively (11, 14). We also used media from another commercial source (Lonza). Low (1 ng/mL) and high (10 ng/mL) concentrations of IL-13 treatment increased goblet cells and DAPT treatment skewed the cultures toward ciliated epithelial cell differentiation (Fig. 1F and SI Appendix, Fig. S1 C and E). Culturing cells with media from Lonza increased basal cells, CD66c+ basal cells, and, to a lesser degree, secretory cells. It also slightly decreased the prevalence of ciliated cells (Fig. 1F and SI Appendix, Fig. S1D). Diversity was reduced under DAPT treatment compared to the remaining treatment groups (P = 0.0003, ANOVA with the Tukey post hoc test, SI Appendix, Fig. S1H). We used PCA to cluster all week 3 differentiations based on their underlying cell type frequencies, with arrowheads depicting the cell type frequency responsible for driving sample points in a given direction (Fig. 1G). The largest drivers of heterogeneity among culture replicates were basal, secretory, triple negative, and CD66c- other cells (SI Appendix, Fig. S2B). NHBE cells treated with IL-13 generally clustered closer with the NHNE cells, driven by the presence of secretory and goblet cells (Fig. 1G; dark green and yellow circles, “secretory” and “goblet” arrowhead). The DAPT-treated cultures grouped together and consistently demonstrated higher levels of basal cells and lower levels of nonciliated epithelial cells (Fig. 1G; dark blue points; “basal” arrowhead and opposing “secretory,” “goblet,” and “triple negative” arrow ends). Interestingly, replicates that underwent the same treatment did not always cluster similarly, highlighting the differential response of individual donors to treatments and the inherent heterogeneity of the system. Finally, to visualize how donor variability affected the overall makeup of the cultures, we performed multidimensional scaling (MDS) analysis of Pneumacult-treated culture systems. This revealed that individual donors are separate from each other and even independent differentiations from the same donor do not cluster together, though they generally were closer together than to some of the other individual donors (Fig. 1H). Altogether, these data demonstrate the heterogeneity in this culture system and the cell type plasticity potential of primary human airway cultures.

IAV Tropism Preference Is Narrow Compared to Overall Cellular Diversity.

We used the high cell type diversity in NHBE and NHNE cultures to determine how IAV tropism differs across upper airway epithelial cells and the relative contributions of these various cell types to overall IAV replication. NHBE and NHNE cells were infected with Cal/07/09 and stained for HA and NP at 36 hours post-infection (hpi) to identify infected cells. Ciliated and secretory cells were the dominantly infected cell types (Fig. 2A), which was true across all culture conditions despite a highly diverse cellular environment and variable rates of infection (Fig. 2B). These data are consistent with the observed Cal/09 tropism preference for ciliated cells in the mouse lung (4) and the high numbers of PR8-infected ciliated cells in the mouse upper airway (22–24). To test whether this tropism bias was due to the infection dose, we doubled the amount of virus in the NHBE inoculum (MOI = 1 vs. 0.5) but found no changes in overall tropism (Fig. 2C). Cell type diversity was consistently lower among infected cells compared to the uninfected cultures, further highlighting the narrow tropism for IAV in the airway cultures (Fig. 2B). These data suggest that despite the extremely variable culture environment into which IAV is introduced, ciliated cells are the dominant tropism.

Fig. 2.

(A) UMAP of flow cytometry phenotyping data from three combined wells of NHBE cells cultured with Pneumacult and infected with 0.5 MOI of A/Cal/07/09. Left: UMAP colored by cell subset. Right: UMAP colored by A/Cal/07/09 HA (Center) or NP FI (far right). (B) Alluvial plots depicting culture cell type frequencies before infection (Left), frequency of total infected cells (HA⁺ and/or NP⁺) (Center), and proportion of infected cell types out of total infected cells (Right) after A/Cal/07/09 infection (0.5 MOI, 36 hpi) and analyzed by flow cytometry. Cultures were 3 wk post-ALI NHBE cells cultured under the indicated conditions and NHNE cells cultured with Pneumacult media. Numbers above the Left and Right columns depict Simpson’s alpha diversity index of uninfected and infected wells, respectively, for each condition. Data representative of n = 17 independent differentiations and infections: n = 4 NHBE Pneumacult, n = 4 NHBE Pneumacult + Low (1 ng/mL) IL-13, n = 4 NHBE Pneumacult + High (10 ng/mL) IL-13, n = 2 NHBE Pneumacult + 10 μM DAPT, n = 2 NHBE Lonza, and n = 1 NHNE Pneumacult treatments. (C) NHBE cells grown in Pneumacult media infected with A/Cal/07/09 at MOI 0.5 or 1 and infected (HA⁺ and/or NP⁺) cell types compared by flow cytometry. n = 1 independent differentiation, error bars show SEM of n = 2 wells per condition. (D) NHBE cells (n = 1 independent differentiation, n = 3 wells per condition) grown in Pneumacult media analyzed for surface expression of α2,6 and α2,3 sialic acids using a biotin-conjugated S. nigra lectin (SNE, EBL) and M. Amurensis lectin II (MALII), respectively, and fluorescently conjugated Streptavidin secondary. Sialic acid⁺ frequency (Top) and log10 geometric mean FI (gMFI) (Bottom) by cell type. Error bars show SEM of n = 3 wells per stain. Letters above each indicate statistical groups determined by ANOVA with the Tukey HSD post-hoc test. (E) NHBE cells (n = 1 independent differentiation, n = 1 well per condition) cultured in Pneumacult or Lonza were infected apically or basolaterally with A/Cal/04/09 at the indicated MOI and analyzed for HA and NP expression by flow cytometry at 36 hpi. (F) NHBE cells grown in Pneumacult media (n = 1 independent differentiation, n = 3 wells per condition) were infected with 0.5 MOI of A/Cal/07/09 prior to air lifting (week 0) and HA and NP expression analyzed at 36 hpi by flow cytometry. Alluvial plot generated as in B.

Basal cells are highly prevalent in cultures but were infected at very low frequencies. This was especially striking in the DAPT-treated cultures and remained true even at a higher MOI (Fig. 2 B and C). To determine whether this was due to differences in receptor distribution, we evaluated the level of α2,6 and α2,3 sialic acids on the cell surface of all cell types. α2,6 and α2,3 expression was highest on ciliated, CD271+ ciliated, and CD66c+ basal cells, as measured by percent of cells and geometric mean fluorescence intensity (gMFI) (P = 2.14 e-11 for α2,6; P = 3.46e-16 for α2,3; ANOVA with the Tukey post hoc test; Fig. 2D). Surprisingly, most basal cells had both α2,6 and α2,3 sialic acids on their surfaces, indicating that the lack of infection was not due to differences in receptor expression (Fig. 2D). Once differentiated, cultures form a three-dimensional pseudostratified epithelium with basal cells at the bottom, potentially physically protecting them from the virus. To determine whether basal cells were resistant to infection due to a lack of access, we took two approaches. First, we added virus to the basal media to allow direct access to the bottom of the transwell, which did not result in basal cell infection (Fig. 2E). To evaluate whether basal cells could be infected on the apical side prior to the formation of a pseudostratified epithelium, we infected cultures at week 0, 1, and 3 post-air lifting, which includes time points before full pseudostratification and when basal and basal-like cell numbers are higher (Fig. 1A and SI Appendix, Fig. S1B). At week 0 and 1, we again observed limited infection of basal cells, though infected basal-like cells were present (Fig. 2F and SI Appendix, Fig. S2C). These data suggest that basal cells are semiresistant to IAV infection irrespective of infection dose, receptor expression, and their location in the culture.

Cell Type Identity Affects Replication Levels, but Not Dynamics, of IAV.

IAV replication levels can be impacted by several factors, including cell type and innate immune activation status (22). To determine the replication potential in cultures with diverse cellular compositions, we infected NHBE and NHNE cells under various differentiation conditions with Cal/07/09 and measured multicycle replication. We found varying levels of replication across culture conditions, even when controlling for differences in total cell numbers across cultures (Fig. 3A). Importantly, most primary culture conditions produced greater levels of virus than A549s despite not all cell types being susceptible to infection (Fig. 3A). Given the heterogeneity of cell types present and production of extracellular factors such as mucus, we evaluated the amount of virus remaining in the inoculum after the hour infection. We did not observe any significant differences in virus absorption across conditions and there was also no correlation between virus absorption and multicycle replication (P > 0.05, ANOVA, Fig. 3B).

Fig. 3.

(A) Multicycle growth curves of NHBE and NHNE cells cultured under the indicated conditions and A549 cells infected with A/Cal/07/09 at an MOI of 0.1. Data representative of n = 16 NHBE differentiations (n = 4 Pneumacult, n = 4 Pneumacult + Low (1 ng/mL) IL-13, n = 4 Pneumacult + High (10 ng/mL) IL-13, n = 2 Pneumacult + 10 μM DAPT, and n = 2 Lonza), n = 1 NHNE differentiation (Pneumacult), and A549s, all with five wells per differentiation/treatment. Error bars show SEM. (B) NHBE and NHNE cells were grown under the indicated conditions and infected as in A. After the 1 hpi, the inoculum was removed and titered to determine the amount of virus remaining. Graphs depict the (Top) percent remaining in the inoculum for each condition and (Bottom) the relationship between virus remaining and AUC of the multicycle growth curve as in A. Data from n = 2 independent NHBE differentiations per condition, n = 1 NHNE differentiation (Pneumacult), and n = 2 A549s all with five wells per differentiation/treatment. (C) Log10 FI of HA and NP from individual infected cells in Fig. 2B. Data depict one independent differentiation, n = 3 wells per treatment. Horizontal lines in violin plots show median and interquartile range of FI data. Data are representative of n = 4 separate flow cytometry experiments and a total of n = 4 NHBE Pneumacult, n = 4 NHBE Pneumacult + Low (1 ng/mL) IL-13, n = 4 NHBE Pneumacult + High (10 ng/mL) IL-13, n = 2 NHBE Pneumacult + 10 μM DAPT, n = 2 NHBE Lonza, and n = 1 NHNE Pneumacult treatments. (D) NHBE cells grown in Pneumacult were infected with A/Cal/07/09 at an MOI of 0.5. At 6 and 12 hpi, ciliated and secretory cells were sorted and sequenced by stranded RNAseq. The InVERT pipeline was used to computationally identify IAV m/c/vRNAs from each segment and mRNA for splice products. Graphs depict the (Left) proportion of each RNA species in ciliated vs. secretory cells and (Right) total IAV transcripts per million quantified in ciliated and secretory cells. Two points per segment or splice product reflect the two timepoints. InVERT data are representative of two independent differentiations with n = 3 to 4 wells per timepoint per group.

To determine replication levels in susceptible cell types we evaluated the HA and NP FI of each infected cell and compared between cell types in each condition (Fig. 3C). FI was used in place of gMFI to account for differences in the number of infected cells within a cell type. Ciliated and secretory cells had the highest number of infected cells, as depicted by the width of the violin plots, further emphasizing that these two are the dominant infected cell types. Although FI values cannot be compared between experiments, we consistently saw within an experiment that ciliated and CD271+ ciliated cells displayed higher mean log FI than most other cell types for nearly all culture conditions (ANOVA with the Tukey post hoc test; Fig. 3C and Dataset S1).

We also examined replication kinetics at the RNA level in our NHBE cultures. IAV generates several distinct RNA species during replication which relate to infection stage. The incoming polymerase drives primary transcription that generates messenger RNA (mRNA). With the assistance of newly translated polymerase protein, replication begins with positive sense complementary RNA (cRNA) production that is used as a template to make more negative sense viral RNA (vRNA). To evaluate each stage of replication in the two most susceptible cell types, we sorted ciliated and secretory cells from infected NHBE cultures and sequenced the positive and negative sense RNA. We then evaluated mRNA, cRNA, and vRNA levels using the Influenza Virus Enumerator of RNA Transcripts (InVERT) computational pipeline (25). Despite differences in HA and NP protein abundances between these two cell types (Fig. 3C), they had similar ratios of IAV mRNA, cRNA, and vRNA for all virus segments (R² = 0.99, Fig. 3D). Secretory cells produced higher levels of viral transcripts at 6 hpi (t test, P = 0.0006). Ciliated cells became the dominant producer of viral transcripts at 12 hpi though the difference was not significant. Together, these data demonstrate that while the total abundance of IAV varies across cell types, the dynamics of replication do not vary in the two most abundantly infected cell types.

Cell Type Composition’s Impact on IAV Multiround Viral Burst Size.

The heterogeneous nature of NHBE and NHNE culture makeup offers a unique opportunity to determine what biological features most impact overall IAV replication. Replication varied substantially between cultures (Fig. 3A) and we hypothesized that certain cell types may be responsible for driving these differences. First, we plotted the percent infection (determined by flow cytometry) with multiround viral burst size [measured by the area under the curve (AUC) of the multicycle growth curve at 0 to 36 hpi] for our NHBE differentiations and only observed a weak positive correlation (R² = 0.22) (Fig. 4A). Univariate analyses of frequencies of infected cell types vs. multiround viral burst size showed only the frequency of infected ciliated and CD271+ ciliated cells were positively correlated with flu production (SI Appendix, Fig. S2D).

Fig. 4.

(A) Correlation between total percent influenza-infected (IAV+) cells determined by flow cytometry (MOI 0.5, 36 hpi) and multicycle growth curve area under the cure (AUC) (MOI 0.1, 0 to 36 hpi) from NHBE cultures differentiated for 1 or 3 wk. Data from n = 18 NHBE differentiations and infections (n = 1 Pneumacult, n = 1 Lonza at 1 wk post-ALI; n = 4 Pneumacult, n = 4 Pneumacult + Low (1 ng/mL) IL-13, n = 4 Pneumacult + High (10 ng/mL) IL-13, n = 2 Pneumacult + 10 μM DAPT, and n = 2 Lonza at 3 wk post-ALI) with n = 3 to 5 wells per differentiation/treatment. (B) PCA biplot of NHBE and NHNE cultures differentiated for 1 or 3 wk and phenotyped by flow cytometry after infection with A/Cal/07/09 (MOI 0.5, 36 hpi). PCA was performed on the frequencies of each infected cell type. Dots are colored by AUC of the 0 to 36 hpi multicycle growth curve (MOI 0.1). Letters denote the specific differentiation condition of each dot, denoted in Dataset S2. n = 18 independent NHBE differentiations and infections [n = 1 Pneumacult, n = 1 Lonza at 1 wk post-ALI; n = 4 Pneumacult, n = 4 Pneumacult + Low (1 ng/mL) IL-13, n = 4 Pneumacult + High (10 ng/mL) IL-13, n = 2 Pneumacult + 10 μM DAPT, n = 2 Lonza at 3 wk post-ALI] and n = 1 NHNE differentiation and infection (Pneumacult at 3 wk post-ALI). (C) PCA biplot of NHBE and NHNE cultures described in B. PCA performed on cell type frequencies of uninfected wells paired with wells described in B. Dots are colored by AUC of the 0 to 36 hpi multicycle growth curve (MOI 0.1). Letters denote the specific differentiation condition of each dot, denoted in Dataset S2.

To better understand the overall contributions of infected cell types relative to one another, we used PCA to group differentiations by the frequencies of infected cell types and analyzed multiround viral burst size (0 to 36 hpi AUC) as a supplementary variable. Differentiations with low or moderate multiround burst sizes were associated with low levels of all infected cell types (Fig. 4 B, Bottom Left Quadrant). Thus, not surprisingly, low overall infection yielded a lower multiround burst size. PCA showed that moderate or high producing differentiations tended to have higher frequencies of infected ciliated, CD271+ ciliated, and triple-negative cells compared to other cell types (Fig. 4B), trends which are obscured by univariate analyses. Notably, infected secretory cells were not strongly associated with higher multiround burst size (Fig. 4B and SI Appendix, Fig. S2D) despite being among the most frequently infected cell type (Figs. 2B and 3C). Next, we did the same PCA but with the cell type frequencies prior to infection (Fig. 4C) to determine whether the presence of certain cell types in the culture was associated with high or low IAV output. This showed that the highest producing differentiations began with higher frequencies of CD271+ ciliated cells and moderate levels of CD66c+ basal and CD66c+ other cells. Differentiations that produced moderate multiround viral burst sizes exhibited a range of preinfection composition, including high frequencies of basal cells and ciliated cells. Differentiations that yielded the lowest multiround viral burst sizes had high frequencies of triple negative, secretory, and goblet cells (Fig. 4C). Altogether, these data demonstrate that ciliated cells and ciliated-like cells, when infected, contribute to larger multiround viral burst sizes. Additionally, cell type frequencies prior to infection could potentially be useful for anticipating which types of culture systems will best control IAV replication.

Cell Type-Specific Antiviral Gene Expression.

To determine transcriptional features that may be impacting influenza infection and multiround viral burst size disparities across cell types we profiled uninfected and infected NHBE cultures by single-cell RNAseq. Using UMAP and cell type annotation with ScType, we found clusters of cells that broadly recapitulated the cell types identified in our flow cytometry panel (Fig. 5A). Secretory cells express CEACAM6, which encodes CD66c. Club cells also express CEACAM6 and in our flow cytometry panel may be captured in either the secretory or CD66c+ other populations. NGFR, which encodes CD271, was predominantly expressed in the basal cell cluster. TSPAN8 was expressed in a subset of secretory/goblet cluster likely representing the goblet cell lineage. Finally, ciliated cells cluster in multiple places, including close to basal cells which likely represent the CD271+ ciliated population identified by flow cytometry. We found infected cells where influenza reads represent >0.07% of the total transcriptome in multiple cell types, particularly in club cells, ciliated epithelial cells, and secretory cells (Fig. 5A and SI Appendix, Fig. S3). Differences in the expression of host dependency factors which can enhance IAV infection may account for certain cell types supporting higher IAV replication (26); however, we observed even expression of host factors across all cell types (Fig. 5B). We also examined the expression of interferon-stimulated genes (ISGs), as many of the products from these genes can potently restrict influenza virus replication. We observed cell type-specific expression of some ISGs at baseline (Fig. 5C). Interestingly, we found that basal cells, which resist high levels of influenza infection, express the potent anti-influenza genes IFITM3 and IFITM1 (27). After infection, we also observed ISG expression in ciliated and secretory cells, including many potent antiviral influenza genes such as MX1, IFITM3, and IFITM1, indicating a robust response to infection (Fig. 5D). Secretory cells expressed a greater number of ISGs in the infected condition compared to uninfected and had higher expression of ISGs shared with ciliated cells. We found LY6E, an enhancer of influenza infection (28), was expressed in ciliated cells but not secretory cells after infection. We also compared infected (i.e., % flu > 0.07%) ciliated to infected secretory populations to identify differentially expressed ISGs and again determined a much larger number of ISGs unique to secretory cells (Fig. 5E). Together, these data suggest potential mechanisms underlying the resistance of basal cells to infection and the greater contribution of ciliated cells over secretory cells to multiround viral burst size.

Fig. 5.

(A) (Left) Integrated UMAP of uninfected and Cal/04/09-infected (MOI 0.5, 10 hpi) NHBE cells, colored by cell types assigned with ScType. (Right) Feature plots of percent IAV gene expression per cell (in Cal/04/09-infected condition only), and CEACAM6, NGFR, and TSPAN8 gene expression (in integrated dataset). (B) Heatmap of expression of 101 IAV host factors across NHBE cell types. Color represents log-transformed expression values. (C) Heat map of the top 50 most differentially expressed ISGs per cell type in the uninfected condition (Padj < 0.05). Color represents scaled expression of top variable genes. (D) Dot plot of differentially expressed ISGs in ciliated (blue) or secretory/goblet (purple) cells in IAV-infected condition vs. uninfected condition. (E) Dot plot of differentially expressed ISGs in infected (i.e., flu read content > 0.07%) secretory/goblet or ciliated cells. Blue represents infected ciliated vs. infected secretory/goblet cells; purple dots represent infected secretory/goblet vs. infected ciliated. Data from n = 1 NHBE differentiation.

Discussion

Studies have utilized single-cell approaches to help understand viral replication and multiround viral burst sizes in individual cells (29, 30), but monocultures cannot recapitulate complex cell–cell interactions present within tissues. Though others have looked at cell type heterogeneity present in primary culture systems (31), our study interrogates causal relationships between cell type frequencies and IAV output. Additionally, we show how pushing these cultures to highly skewed cellular makeups can uncover unique phenotypes. For example, cultures treated with DAPT, which are comprised almost exclusively of ciliated cells and basal cells (Fig. 1 E and F and SI Appendix, Fig. S1C), produce equivalent levels of virus as untreated cultures (ANOVA, P > 0.999, Fig. 3A), further highlighting how significantly ciliated cell infection drives high levels of virus production.

This work focused on cell type prevalence as a driver of viral output, though other biological variables are likely at play as well. Transcriptional response of infected cells impacts polymerase activity in vivo (22) and differential innate immune signatures among infected cell populations may affect IAV burst size. This could be due to cell type-specific antiviral gene signatures, or variation in innate immune response from donor to donor. We previously showed in NHBE cultures infected with SARS-CoV-2 that higher levels of viral transcripts were found in cells with lower expression of ISGs (32). Here, we observed cell type-specific ISG expression at baseline which may impact replication success (Fig. 5B). This would be consistent with a recent report demonstrating varied baseline expression of antiviral genes across different cell types that could contribute to tropism preference and higher virus output from certain cell types (33). Additionally, coinfection of cells with multiple IAV particles can increase viral production (34, 35). Future work could investigate whether certain cell types in NHBE or NHNE cultures are more susceptible to viral coinfection and could lead to a better understanding of multiround viral burst size heterogeneity. There is likely a severe bottleneck during influenza transmission resulting in very few virions to establish an infection in a new host. The cell types that these first virions land in may impact the success or failure of that transmission event.

Limitations

In addition to the variables discussed above that could also be affecting IAV production, there are a few limitations of our experiments. To understand the relationship between infected cell type frequencies and multiround viral burst size, we used flow cytometry to measure IAV tropism and multicycle growth curves to measure multiround viral burst size. Due to technical limitations, cultures used for flow cytometry were infected with an MOI of 0.5 while multicycle growth curves utilized an MOI of 0.1. These MOIs were applied consistently across all replicates, and we assume that the relationship between multiround viral burst size and infected cell type frequencies would not change with a difference in MOI. We indeed demonstrated that changing MOI from 0.5 to 1.0 does not yield a difference in infected cell phenotyping (Fig. 2C). There are also some disparities in the cell types identified by flow cytometry and scRNAseq. Interestingly, club cells—which are important targets in mice in vivo (36)—are identified in our scRNAseq but are not well defined by our flow cytometry panel. They express CEACAM6, which encodes for CD66c, and may be captured in our secretory or CD66c+ other gates. These data highlight the need for additional flow markers. Additionally, primary human airway cultures produce mucus (10) which could contribute to differences in IAV infection. Although we did not formally measure mucus production, viral absorption was not significantly changed among IL-13 treated cultures (which have higher mucus-producing goblet cells) compared to remaining conditions. Finally, NHBE and NHNE cultures lack innate immune cells and endothelial cells present in the airways. Without innate immune cells to help clear newly released virions, the rate of reinfections and spread is likely higher than in vivo. Thus, our findings are most relevant for understanding initial virus infection and early spread, rather than later stages of infection.

Despite these limitations, this work underscores the complex interplay between IAV infection and the heterogeneity of primary human airway epithelial cell cultures. By examining the diversity of cell types within the cultures and their impact on IAV replication dynamics, tropism, and multiround viral burst size, we found that ciliated and ciliated-like cells consistently display a dominant tropism for IAV across culture conditions. Secretory cells may reduce the overall multiround viral burst size of infected cultures despite having strong tropism for IAV. This finding holds significant implications for understanding IAV transmission and the potential development of therapeutic strategies at a cellular level.

Materials and Methods

A549 and MDCK Cell Culture.

Human lung adenocarcinoma A549 cells and MDCK cells were maintained in 1× Dulbecco’s Modified Eagle’s Medium (DMEM) supplemented with 10% fetal bovine serum and 1% penicillin-streptomycin.

NHBE and NHNE Donor Information.

NHBE cells from a total of six individual donors were utilized in our experiments and analyses. Donor A is a Caucasian female, aged 78. Donor B is a Caucasian male, aged 54. Donor C is a Caucasian male, aged 21. Donor D is a Caucasian male, aged 55. Donor E is a Hispanic male, aged 66. Image data were generated from a Hispanic male, age 43. NHNE cells from one donor (Caucasian male, aged 48) were utilized in our experiments and analyses. All primary human samples are deidentified and IRB except.

NHBE and NHNE Cell Culture.

NHBE (Lonza Bioscience, CC-2540S) and NHNE (PromoCell, C-12620) cells, harvested from healthy individuals, were expanded using either Pneumacult-Ex Plus Media (STEMCELL Technologies) or B-ALI Growth Basal Media (Lonza Bioscience) and seeded on 6.5 mm Transwell 0.4 μm pore polyester inserts (STEMCELL Technologies). Once confluent 2 to 4 d post-seeding, the ALI cultures were airlifted: Apical media was removed to expose cells apically to air and basal compartment media was replaced with either Pneumacult-ALI Basal Media (STEMCELL Technologies) or B-ALI Differentiation Basal Media (Lonza Bioscience). Cells were maintained at air–liquid interface (ALI) at 37 °C and 5% CO₂ in a humidified incubator. For IL-13 treated wells, Pneumacult-ALI Media was supplemented with 1 ng/mL (“Low”) or 10 ng/mL (“High”) recombinant human IL-13 (PeproTech) beginning at day 8 post-ALI and continued through the remainder of the culture time. For DAPT-treated wells, Pneumacult-ALI Media was supplemented with 10 µM DAPT (Tocris) beginning at day 10 post-ALI and continued through the remainder of the culture time. Cells were used at 3+ wk post-ALI unless otherwise noted.

Virus Infection.

The apical surface of cells was washed with 1× PBS and Influenza A/California/07/2009 (kindly provided by Chris Brooke, University of Illinois) in infection media (PBS with Ca/Mg, 1% penicillin-streptomycin, 5% BSA) was added directly to the apical surface and incubated for 1 h at 37 °C. A549 cells were infected alongside the ALI cultures. Unless otherwise noted, flow cytometry phenotyping experiments were performed with MOI 0.5, and multicycle growth experiments were done with MOI 0.1. After 1 h, the virus inoculum was removed, stored at −80 °C, and the apical surface washed with 1× PBS. Cells were incubated at 37 °C until harvest.

For comparing apical and basolateral infection, Influenza A/California/04/2009 was used. Apical infections were performed as described above at MOI 0.5 in infection media. For basolateral infection, the normal basolateral media was removed and the basal compartment washed once with 1× PBS. Virus in infection media was added to the basal compartment at MOI 0.5 or MOI 2.0 and incubated for 1 h at 37 °C before being removed. Basal compartment was washed with 1× PBS before replacing with appropriate basal media.

Flow Cytometry Analysis.

While still on the transwell, the apical side of cells were incubated with SiR-tubulin (Spirochrome/Cytoskeleton Inc.) in appropriate basal media at 37 °C for 1 h and then washed and trypsinized using the ReagentPack Subculture Reagents (Lonza Bioscience). Single-cell suspensions were washed with 1× PBS and stained with Ghost Dye Violet 450 (Tonbo) for 30 min covered on ice. Cells were washed once with FACS buffer (cold HBSS supplemented with 2% bovine serum) and blocked with 5% goat serum (Millipore Sigma) and Human TruStain FcX (Biolegend) for 15 min covered at room temperature. After washing once with FACS buffer, cells were stained with surface antibodies for 30 min covered on ice. Samples were washed with FACS buffer once more and fixed for 30 min covered at room temperature with the eBioscience Foxp3/transcription factor staining buffer kit (ThermoFisher). After fixing, samples were washed with the 1× Perm/Wash buffer and stained with intracellular antibodies in 1× Perm/Wash for 30 min covered on ice. Cells were washed twice with FACS buffer prior to running on the flow cytometer.

For sialic acid staining, single-cell suspensions were prepared as described above until the fixation step. Samples were then fixed with 4% PFA (ThermoFisher) in a 37 °C water bath for 10 min, covered. Cells were washed with FACS buffer and stained with Biotin-conjugated lectin for 45 min covered at room temperature. After washing once with FACS buffer, cells were stained with a streptavidin-conjugated AF488 secondary for 30 min covered on ice. Cells were washed twice with FACS buffer prior to running on the flow cytometer.

Samples were run on a BD LSR Fortessa (Becton Dickinson) and the data analyzed in FlowJo (Version 10.9.0; Becton Dickinson). Antibodies used include the following: Alexa Fluor 488-Streptavidin (ThermoFisher); BV421-conjugated CD66c (B6.2/CD66) (BD Biosciences); PE-Cy7-conjugated CD271 (ME20.4) (Biolegend); PE-conjugated TSPAN8 (458811) (R&D Systems); mouse IgG2a anti-IAV NP (HB65) (BioXCell); APC/Fire 750 anti-mouse IgG2a (Biolegend); and AF488-conjugated mouse anti-HA (EM4 CO4) (gift from Chris Brooke, University of Illinois, Champaign-Urbana, IL). Biotin-Maackia Amurensis lectin II (MALII) (Vector) and Biotin-Sambucus nigra lectin (SNE, EBL) (Vector) were used for α2,3 and α2,6 sialic acid staining, respectively.

Multicycle Growth Curve Collection and Plaque Assay.

A549 and ALI cells were infected with Influenza A/California/07/2009 as described above at an MOI of 0.1. Supernatant was collected at 0 (prior to infection), 12, 24, and 36 hpi as described below. For A549 cells, 200 μL of viral growth media (1× DMEM with 25 mM HEPES buffer, 2.5% BSA Fraction V (7.5% solution), and 1% penicillin-streptomycin) supplemented with 1 μg/mL TPCK-trypsin was added after infection in place of normal media. At each time point, media was collected and replaced with 200 μL fresh viral growth media plus 1 μg/mL TPCK-trypsin. For ALI cells, 200 μL of 1× PBS was incubated on the apical surface for 10 min at 37 °C and collected. Virus collections were frozen at −80 °C until they were titered by plaque assay.

10-fold serial dilutions of virus collections were prepared and Madin-Darbycanine kidney (MDCK) cells were infected in infection media at 37 °C. After 1 h, infection media was replaced with an agar overlay (1× MEM, 1 μg/mL TPCK-trypsin, 0.01% w/v DEAE-dextran, 0.1% w/v NaHCO₃, and 0.6% w/v Oxoid agar) and incubated for 40 to 42 h at 37 °C. Plates were fixed with 4% formaldehyde for 30 min prior to removal of the overlay. Blocking and immunostaining were performed rocking at room temperature for 1 h in 5% milk in PBS. Plaques were stained with an in-house polyclonal mouse anti-Cal/04/09 (1:5,000), generated using serum harvested at 30 d post-infection of mice inoculated with 5,000 PFU of Cal/04/09. Secondary staining was done using sheep anti-mouse-HRP (1:5,000; Cytiva, NA931). Virus plaques were resolved using KPL TruBlue Peroxidase Substrate (Seracare Life Sciences, 5510-0030) and counted.

ALI Culture Sectioning and Imaging.

NHBE cultures were washed and fixed with 4% PFA at 3 wk post-ALI. After washing in a 30% w/v sucrose PBS solution, the membranes were extracted from the transwell scaffolds and frozen in Tissue-Tek O.T.C. compound (Sakura Finetek USA) using a dry ice bath. Sections were cut from the OTC block at 12 µ at −20 °C on a Leica CM1860 cryostat and transferred onto a Fisherbrand MicroProbe glass slide (22230900). To stain, slides were blocked using 3% BSA in PBST for 1 h and treated with a conjugated antibody panel overnight: alpha-tubulin Alexa Fluor™ 488 (1:500; ThermoFisher, 322588), MUC5AC Alexa Fluor™ 555 (1:50; Abcam, ab218714), and Cytokeratin-5 Alexa Fluor 647™ (1:100; Abcam, ab193895). After washing with PBST, the slide was imaged using a Leica Stellaris confocal microscope.

Cell Sorting and RNA Extraction for Sequencing.

Cells were grown in Pneumacult media for 3 wk at ALI as described above. Cells were SiR-tubulin stained, trypsinized, washed, blocked, and surface antibody stained as described above except PBS supplemented with 5% bovine serum was used in place of FACS buffer. A minimum of 25,000 cells of each type were sorted directly into RLT Plus (Qiagen) supplemented with 2-mercaptoethanol (10 μL/mL) and Reagent DX (0.5% v/v; Qiagen). RNA was extracted using the RNeasy Plus Micro Kit (Qiagen) and prepared for sequencing as described below.

Bulk RNAseq and Analysis.

For analysis of RNA replication kinetics, cDNA libraries were prepared using the Stranded Total RNA v2 PicoMammalian kit (Takara). Samples were sequenced as 150 base pair paired-end reads using NovaSeq (Illumina). For gene expression analysis of sorted cell populations, cDNA libraries were prepared using the SMART-Seq mRNA LP kit (Takara) and sequenced as 150 bp paired-end reads using a NovaSeq X plus 10B flowcell. For both experiments, reads were trimmed of adapter sequences and low-quality bases with Trimmomatic (37) and mapped to a hybrid reference genome of human and Influenza A/California/07/2009 or Influenza A/California/04/2009 (accession nos. CY121680-CY121687 or FJ966079-FJ966086) with STAR (38).

Quantification of mRNA, cRNA, and vRNA levels was performed using the InVERT pipeline (25). For each of the eight segments of the influenza genome, read counts to the positive (mRNA and cRNA) and negative strand (vRNA) were quantified using Cuffdiff (39). To determine the proportion of forward strand read counts belonging to mRNA and cRNA, the ratio of mRNA to cRNA reads was determined by calculating the depth of coverage on the forward strand at either the 3′ polyA tail sequence (for the mRNA) or a segment-specific 3′ cRNA sequence following 5 to 6 adenosines (for the cRNA) using SAMtools (40). FPKM values from Cuffdiff were converted to TPM for graphing. Code for InVERT is available at https://github.com/langloislab/invert. The proportion of mRNA, cRNA, and vRNA out of the total RNA TPM values per transcript was calculated for the 6 and 12 h infection time points and compared between ciliated and secretory cells to determine whether proportions of RNA species varied between cell types. Gene expression analysis was performed using HTSeq (41) and DESeq2 (42). DEGs were retrieved for each cell population with a contrast of a given cell type compared to the remaining 4 cell types. DEGs for each cell population were compared to previously determined cell-type specific markers to infer cell type identity of unknown populations (21).

Single-Cell RNAseq and Analysis.

NHBE cells were infected with A/California/04/2009 at an MOI of 0.5. 10 hpi cells were washed and trypsinized to generate single-cell suspensions. Single-cell suspensions were counted, and a target of 10,000 cells per condition were used to generate Gel Bead-in Emulsions (GEMs). The 10× Chromium 3′ GEX Capture kit and the 10× Chromium controller were used to generate GEMs and barcoding per the manufacturer’s protocol (10× Genomics, Pleasanton, CA). GEMs generated were used for cDNA synthesis and library preparation using the Chromium Single Cell 3′ Library Kit and run on a NovaSeq S2. 2×100 PE Run. cDNA libraries were sequenced in duplicate to produce two sets of FASTQ files for each sample.

FASTQ files from single-cell RNAseq runs were mapped to a hybrid reference genome of human (GRCh38 release 110) and Influenza A/California/04/2009 (accession numbers FJ966079-FJ966086) using 10× Genomics Cellranger version 7.2.0. Mapped filtered feature barcode matrices were analyzed in Seurat version 4.3.0 (43). Briefly, uninfected- and IAV-infected conditions were assessed for sequencing quality, integrated, and clustered. Conserved feature markers per cluster were identified and cell type identities were annotated with scType (44) using the included lung cell marker database. Influenza-infected cells were defined as cells with greater than 0.07% influenza transcript content, a cutoff which was drawn from violin plots of percent influenza reads per cell (SI Appendix, Fig. S3). A heatmap of host dependency factor expression was made by visualizing log2-transformed expression values within the uninfected condition. A list of 101 host dependency factors identified across multiple studies was used to filter genes for visualization (26). Heatmaps of cell type–specific interferon stimulated genes (ISGs) were made with the uninfected condition dataset using FindAllMarkers and filtering using an adjusted P-value of < 0.05 and log2 fold change (Log2FC) of > 0. ISGs were then identified using a curated list of human ISGs (45, 46). Genes up-regulated during infection within ciliated and secretory/goblet cells were found using FindMarkers of either the ciliated or secretory/goblet cell subsets of the IAV-infected condition vs. the uninfected condition. FindMarkers was also used to compare infected ciliated vs. infected secretory/goblet cell subsets of the IAV-infected condition. ISGs were identified by filtering the gene list using the curated human ISG list, and cutoffs for adjusted P-value cutoff < 0.05 and log2FC > 0. Full analysis details, including reference ISG list, are available at https://github.com/langloislab/roach_shepherd_mickelson_2023.

Statistics and Data Visualization.

UMAPs were made with flow cytometry data from one differentiation of NHBE cells cultured in Pneumacult media for 3 wk at ALI. Uninfected and infected samples were phenotyped with manual flow cytometry gating as described above (n = 3 wells per group). Channel values for markers from each manually gated population were exported from FlowJo and imported to R for dimensionality reduction and UMAP creation using the Spectre package (47). Uninfected and infected replicates were merged to perform dimensionality reduction using channel values from SiR-tubulin, CD66c, CD271, and TSPAN8 markers. A UMAP was then created using a subsampled population of 20,000 cells from each of the uninfected and infected conditions. UMAPs were colored by marker staining or according to the manually identified population using ggplot2 (48).

The vegan package in R (49) was used to calculate Simpson’s alpha diversity index and perform MDS. Total PFU as determined by plaque assay was divided by the cell counts used for the MOI calculations to determine PFU/cell for each culture condition. AUC of the PFU/cell multicycle growth curves was calculated in GraphPad Prism (Version 10.0.0). PCA of uninfected phenotyping data was performed in R using the stats package (50) with proportions of cell type frequencies as active variables and PFU/cell AUC as a supplementary variable. PCA of infected phenotyping data was performed using the stats package, using the proportion of cell types within all infected cells as the active variable and the PFU/cell AUC as a supplementary variable. Alluvial plots of infection phenotyping data were created in R using ggplot2 (48). Sialic acid gMFI values were log transformed and statistical calculations performed by ordinary one-way ANOVA with Tukey’s multiple-comparison test. Ordinary two-way ANOVA with Tukey’s multiple-comparison test was used for statistical calculations of the percent PFU remaining in virus inoculums. Simple linear regressions were used to calculate correlations between PFU/cell AUC and percent PFU remaining, PFU/cell AUC and total percent infection, and PFU/cell AUC and cell type frequency and infected cell type frequency. Violin plots of HA and NP log10 FI were created using ggplot2 (48). Mean log10 FI values were statistically compared using ANOVA with Tukey’s multiple-comparison test in R.

Supplementary Material

Appendix 01 (PDF)

Dataset S01 (XLSX)

Dataset S02 (XLSX)

Data, Materials, and Software Availability

Sequence data were deposited and are available as FASTQ files in the NCBI sequence read archive under BioProject no. PRJNA1010235 (51). Single-cell RNAseq and Bulk RNAseq count data are available under GEO accession numbers GSE247979 (52) and GSE263356 (53), respectively. Code for the manuscript figures and analyses is available at https://github.com/langloislab/roach_shepherd_mickelson_2023 (54). All other data are included in the manuscript and/or SI Appendix.