An injury-induced mesenchymal-epithelial cell niche coordinates regenerative responses in the lung

Abstract

Severe lung injury causes airway-derived basal stem cells to migrate and outcompete alveolar stem cells resulting in dysplastic repair and a loss of gas exchange function. This “stem cell collision” is part of a multistep process that is now revealed to generate an injury-induced tissue niche containing Keratin 5+ epithelial cells and plastic Pdgfra+ mesenchymal cells. Temporal and spatial single cell analysis reveals that the injury-induced niche is governed by mesenchymal proliferation and Notch signaling, which suppresses Wnt/Fgf signaling in the injured niche. Conversely, loss of Notch signaling rewires alveolar signaling patterns to promote euplastic regeneration and gas exchange. Signaling patterns of injury-induced niches can differentially phenotype fibrotic from degenerative human lung diseases, through apposing flows of WNT/FGF signaling. These data reveal the emergence of an injury and disease associated niche in the lung and the ability of using injury-induced niche specific signaling patterns to discriminate human lung disease phenotypes.

Introduction

The lung is generally quiescent during normal life but upon severe injury or in disease states, specific cell types in the lung alveolus re-enter the cell cycle, differentiate, and replenish lost cells to re-establish respiratory function. How this process occurs and the cellular constituents that are involved remain unclear. Disruption of this multi-step regenerative process can lead to permanent loss of respiratory function and chronic lung diseases.

Respiratory viral infections, such as influenza A, are a leading global cause of infectious morbidity and mortality(1, 2). Respiratory illness caused by influenza follows a stereotypical course with infection of lung epithelium, including alveolar type 2 (AT2) cells, cytotoxic destruction of alveolar tissue, dramatic inflammatory infiltration, and early immune-responsive fibroblast activation(3). This tissue destruction is followed by proliferation of multiple alveolar cell types including AT2 cells, cellular differentiation including AT2 cells differentiating into alveolar type 1 (AT1) cells, and re-establishment of normal alveolar architecture. In mild to moderate lung injury, most of the destroyed alveolar tissue can be regenerated through what has been referred to as the euplastic regenerative response, resulting in the re-establishment of alveolar structure and restoration of gas-exchange function(4, 5). However, following severe injury, a dysplastic response can also occur which entails the egression of intrapulmonary basal stem cells out of the airways into the alveolar parenchyma. The dysplastic basal cells migrate, proliferate, and rapidly expand into regions of severely damaged alveolar regions of the lung. These dysplastic epithelial stem cells express Keratin 5 (Krt5) and form regions referred to as “pods”, which persist and can be found in patients who survived severe influenza infections(6–8). Furthermore, Krt5+ dysplastic epithelium in the alveolar compartment is also found in some chronic and acute lung diseases including COVID-19-induced fibrosis(8). Understanding how the lung balances euplastic versus dysplastic tissue reactions is central to promoting improved lung function post-injury as well as developing new therapies for chronic lung diseases.

The contribution of non-epithelial cells in balancing euplastic versus dysplastic tissue responses to acute lung injury remains unclear. Lineage tracing shows the emergence of Trp63+ intrapulmonary, non-alveolar, basal cells to generate the dysplastic epithelial response post-viral injury. These dysplastic Krt5+ basal cells do not arise from alveolar epithelium, rather these cells exit their airway niche and migrate into the alveolar niche, which is occupied by different cell lineages and signaling networks(7–9). How the interactions between the alveolar and airway cell types within these two niches facilitate the emergence of this keratinized dysplasia remains unknown.

Here, we show that the dysplastic epithelial response after acute injury to the lung leads to the formation of a multi-cellular injury-induced tissue niche. This emergent niche is comprised of both Krt5+ dysplastic epithelium and alveolar Pdgfrb+ mesenchymal cells (also called alveolar fibroblast 2 or AF2), which are derived from alveolar Pdgfra+ cells (also called alveolar fibroblast 1 or AF1). This mesenchymal plasticity is unidirectional, with Pdgfra+ AF1 cells proliferating and differentiating into Pdgfrb+ AF2 cells. Pdgfrb+ AF2 cells do not proliferate or generate Pdgfra+ AF1 cells. Pdgfra-derived Pdgfrb+ cells closely integrate into the injured niche and form a unique Notch mediated niche with the Krt5+ dysplastic epithelium. Suppression of cell proliferation or Notch signaling in Pdgfra+ cells prior to injury results in rewiring of the cell-cell communication axis in the alveolus, leading to a loss of injury-induced niche with a commensurate increase in Wnt and Fgf signaling flow resulting in increased alveolar euplastic regeneration. Using newly generated and existing multi-disease single cell transcriptomic datasets, we show that injury-induced niches and NOTCH mediated dysplastic cell-cell communication pathways are observed in human fibrotic diseases including post-COVID-19 fibrosis and bleomycin-induced human lung injury. In contrast, emphysematous degenerative lung diseases such as chronic obstructive pulmonary disease (COPD) and alpha-1 anti-trypsin (AAT) disease lack this injured niche and exhibit a reversal of cell-cell communication flows, demonstrating an attempt at a euplastic regenerative response in the alveolar niche. These studies delineate an emergent niche that arises after injury and in specific human lung diseases, that dictates whether the tissue engages a dysplastic repair or euplastic regeneration response.

Results

Pdgfra expressing cells are the primary reactive mesenchyme after lung injury and exhibit progenitor cell function

While previous work has shown that the lung mesenchyme reacts to acute injury(10, 11), its contribution towards alveolar regeneration is poorly understood. We analyzed our previously described mouse scRNA-seq data from adult animals(12) and combined this with additional analysis of spatial localization of mesenchymal subsets. This analysis revealed two distinct populations of alveolar fibroblasts called alveolar fibroblast 1 (AF1) and alveolar fibroblast 2 (AF2) (also commonly referred to as alveolar fibroblasts and pericytes, respectively) (Fig. S1a) (10, 13). AF1s expressed Pdgfra, Limch1, and Tcf21 while AF2s express Pdgfrb, Cox4i2, and Notch3 (Fig. S1a,b). GO analysis revealed AF2s are enriched for vascular niche support genes (Fig. S1c). While this suggests AF2 cells likely serve a “pericyte”-like function with the capillary plexus of the alveolus, we identified that AF1s and AF2s exist equidistant to alveolar Erg+ endothelial cells (Fig. S1d, e). We next used Pdgfra^CreERT2 and Pdgfrb^CreERT2 mouse lines to isolate AF1s and AF2s, respectively, and performed scRNA-seq to characterize additional heterogeneity in these mesenchymal lineages (Fig. 1a). These data revealed that Pdgfra-traced cells included AF1s, adventitial, and peribronchial mesenchyme while Pdgfrb-traced cells included AF2s, adventitial, peribronchial, and vascular smooth muscle cells (VSMC) (Fig. 1b, c, Data S1, S2). Using immunohistochemistry (IHC) paired with RNA in situ hybridization we confirmed high efficiency of lineage-labeling of the Pdgfra/Pdgfrb CreER lines within each mesenchymal subtype (Fig. S2a–f) with no recombination occurring in endothelial cells (Fig. S1d, f). We then generated Pdgfra^H2B-GFP:Pdgfrb^CreERT2:Rosa26^LSL-tdTomato mice (Fig. S2g) and used IHC analysis to show that Pdgfra and Pdgfrb expression marked and segregated AF1s and AF2s, respectively, with a dual Pdgfra+/Pdgfrb+ labeled population only observed around the airways and in the adventitial region of the lung (Fig. S2h–j, Data S3, S4).

Figure 1: Reactivity and plasticity of the pulmonary Pdgfra+ mesenchymal cells in the lung.

(a) experimental schematic of the approach to dissect Pdgfra+ and Pdgfrb+ cell heterogeneity in the adult mouse lung at homeostasis. (b) UMAP representation of the scRNA-seq data of the Pdgfra and Pdgfrb-lineage mesenchymal cells. (c) IHC and RNA in situ hybridization to visualize spatial location of AF1 and AF2 cells. (d) schematic approach to study Pdgfra+ and Pdgfrb+ lineage mesenchymal cells after respiratory influenza and separately bleomycin-induced lung injuries. (e) UMAP representation of the scRNA-seq data of the Pdgfra and Pdgfrb-lineage mesenchymal cells after influenza infection. Data represents integrated libraries (sham, day 14, and day 28) for each lineage. (f) UMAPs and FeaturePlots showing division of Pdgfra and Pdgfrb-lineage mesenchymal cells at each timepoint, and the expression of Pdgfra and Pdgfrb of these cells over time. (g) Slingshot trajectory analysis showing trajectory of AF1 cells differentiating into AF2 cells. (h) IHC for Pdgfra-derived AF2 cells (tdTomato+ Notch3-ICD+) showing that these cells are enriched in close proximity to Krt5+ cells, as compared to Krt5− cells. Each dot represents data acquired from a single mouse (i.e. biological replicate). **P < 0.01, evaluated by a paired t-test. (i) summary schematic generated using Biorender.com.

To determine the response of Pdgfra+ and Pdgfrb+ mesenchymal cells to lung injury, we performed lineage tracing with the Pdgfra^CreERT2 and Pdgfrb^CreERT2 mouse lines to fluorescently tag and trace Pdgfra/b+ cells and their future progeny. We coupled this with a temporal series of scRNA-seq experiments across two pathologically relevant lung injury models: H1N1 influenza infection and bleomycin-induced lung injury (Fig. 1d). Integration of Pdgfrb-lineage traced cells from all timepoints revealed these cells do not proliferate nor do these cells exhibit Pdgfrb-to-Pdgfra plasticity (Fig. 1e, S3a–d). In contrast, different subsets of Pdgfra-lineage cells emerged including a subset that exhibited a proliferative signature, another subset that exhibited a myofibroblast gene signature, and another subset that appeared to be AF2s differentiated from AF1s (Fig. 1e, S3e–h). Since Pdgfrb+ cells do not proliferate nor generate Pdgfra+ cells, these data indicate that the Pdgfra+ AF1 lineage, and not the adventitial or peribronchiolar lineages which express both Pdgfra and Pdgfrb, contribute to the Pdgfrb+ AF2 lineage. Indeed, transcriptional trajectory analysis suggested that the Pdgfra+ AF1s transit through a proliferative state before passing through the myofibroblast state and terminating in the AF2 state (Fig. 1g, S3i, j). These AF1-derived AF2s are apparent at day 28 after injury, express high levels of Pdgfrb and Notch3, and lose expression of Pdgfra (Fig. 1f, S3f–h , Data S5). We observed AF1-derived AF2 cells, identified by co-expression of Notch3-ICD and Pdgfra-lineage reporter (tdTomato), spatially enriched within the severely damaged regions of the lung which contain dysplastic Krt5+ epithelium (Fig. 1h).

To determine in an unbiased manner whether Pdgfra-derived AF2s are transcriptionally similar to endogenous AF2s, we merged the transcriptional data from these cells with transcriptional data from AF1s and AF2s isolated at homeostasis and performed Spearman correlation analysis. This analysis confirmed that Pdgfra-derived AF2s exhibit a very high degree of similarity with endogenous AF2s (Fig. S3j,k).

To examine whether this AF1/AF2 lineage relationship is unique to respiratory influenza infection or a general regenerative relationship in the lung, we performed parallel analyses using the bleomycin acute lung injury model. Consistent with respiratory influenza, we found no evidence of Pdgfrb+ cell activation after bleomycin injury, as measured by Pdgfrb+ cell proliferation (via Mki67 expression) or Pdgfrb-to-Pdgfra cell plasticity (Fig. S4a–d). We found Pdgfra-labeled AF1s differentiating into AF2s, with similar transient Pdgfra-lineage+ cell states arising throughout the time course of bleomycin-induced injury (Fig. S4e–j). Interestingly, we observed a population of Pdgfra+ derived cells which expressed Lcn2 and Saa3, genes which are involved in innate immunity and immune cell chemotaxis(Fig. S4e–g). These data suggested that these cells serve an inflammatory function in the injured lung (14). This inflammatory AF1 population also arose after influenza infection, however, since these cells did not cluster out in an unbiased manner from the proliferative AF1 cells, they were labeled as “proliferative” in the influenza dataset (Fig. S3l). Together, these results indicate that after acute lung injury, AF1 cells act as mesenchyme progenitors, transiently proliferate, cross through a myofibroblast cell state, and then differentiate into Pdgfra-derived AF2s (Fig. 1i).

Injury-induced lung myofibroblasts are a transient state that arise from AF1 progenitors

Dynamic gene expression analysis across the trajectory of AF1s as they differentiate into AF2s suggests an increase in Pdgfrb+ expression that initiates during the transient myofibroblast state (Fig. 1g, 2a). To experimentally determine whether myofibroblasts are a transient state of AF1s while differentiating into AF2s, we examined whether myofibroblasts arise uniquely from the Pdgfra-lineage. Lineage tracing shows that AF1s, not AF2s, generated the vast majority of the Acta2+ cells after H1N1 (Fig 2b, c, f)(15). Lineage tracing was then performed at temporally defined times using both Pdgfrb^CreERT2 and Acta2^CreERT2 mice to tag and track mesenchymal cells as they transited through the various cell states before ending in AF2s. Specifically, we administered tamoxifen daily from days 10–13 (4 total doses of tamoxifen) after influenza infection (Fig. 2d, g). These studies revealed that lineage labeling of the Pdgfrb or Acta2 expressing cells during the time course of injury and regeneration captured the active myofibroblasts post-injury (Fig. 2d–i). Since labeling of Pdgfrb+ cells prior to injury demonstrated these cells do not produce Acta2+ myofibroblasts (Fig. 2b, c, f), but labeling Pdgfrb+ after injury did capture the Acta2+ myofibroblasts (Fig. 2d–f), we inferred that we captured AF1s transiting to AF2s through an intermediate myofibroblast cell state. Importantly, we found this transient AF1/myofibroblast cell state in the bleomycin model as well, demonstrating consistency across various modes of lung injury (Fig. S5). By 90 days after injury, these previously labeled myofibroblasts extinguished Acta2 expression (Fig. 2h, i), and persisted in the severely damaged regions of the lung, directly adjacent to the Krt5-expressing dysplastic basal epithelium (Fig. 2j, k). These data demonstrate that the transient myofibroblast state induced by acute injury to the lung results in AF2 cells in persistently damaged regions of the lung.

Figure 2: Source and fate of Acta2 expressing cells in the lung during injury and regeneration.

(a) UMAP of scRNA-seq from Pdgfra-lineage cells after influenza (sham, day 14, and day 28) highlighting myofibroblasts and AF2 cells along the predicted trajectory of AF1 cells differentiating into AF2 cells. (b) Experimental schematic depicting approach to dissect whether Pdgfra+ or Pdgfrb+ cells are the primary producers of myofibroblasts in the lung. (c) IHC of the lineage reporter (tdTomato) and Acta2 at 14 days after influenza infection showing that Pdgfra+ cells are the major producer of Acta2+ myofibroblasts in the lung, relative to Pdgfrb+ cells. (d) Experimental schematic depicting approach to test whether lineage labeling Pdgfrb+ cells during injury labels Acta2+ myofibroblasts. Mice were given daily tamoxifen from days 10–13 and collected for analysis at day 14. (e) IHC of the lineage reporter (tdTomato) and Acta2 at 14 days after influenza infection showing that labeling actively expressing Pdgfrb+ cells during injury captures the Pdgfra-derived Acta2 cells on their way to become AF2 cells. (f) Box plots showing quantification of the percentage of lineage labeled Acta2+ cells out of the total Acta2+ cells after influenza infection. (g) Experimental schematic depicting approach to capture and trace the Acta2 expressing cells during injury and after resolution. (h) IHC of the lineage reporter (tdTomato) and Acta2 showing that administration of tamoxifen after influenza infection labels the actively expressing Acta2+ myofibroblasts during injury at day 14 and that these previously labeled Acta2+ cells extinguish Acta2 expression at day 90 after injury. (i) quantification of IHC images in (h). (j) IHC showing that at later timepoints after influenza infection, labeled Acta2+ cells express the AF2 marker Notch3 and are embedded in injury-induced tissue niches directly adjacent to the dysplastic Krt5-expressing epithelium. All scale bars represent 50 μm. Each data point in (f) and (i) represents data obtained from an individual mouse (i.e. biological replicate). ****P<0.0001, evaluated by one-way ANOVA with Tukey’s adjustment for multiple comparisons (panel f) and unpaired t-test (panel i). Error bars in panel K represent SEM.

Blockade of AF1 cell division suppresses dysplastic repair and promotes euplastic epithelial regeneration

The localization of the Pdgfra-derived AF2s in the severely damaged regions of the lung containing dysplastic keratinized epithelium suggested that the Pdgfra-lineage cells may play a role in the injury-induced emergence or persistence of the Krt5+ basal epithelium. We examined proliferation of AF1s given the scRNA-seq data suggesting these cells enter a transient proliferative phase during lung injury. We observed a robust and transient increase of proliferative Pdgfra-lineage labeled cells within the alveolar region starting at day 7 post-injury but little to no proliferation of the alveolar Pdgfrb-lineage cells (Fig. 3a–d). Since Pdgfrb+ cells do not proliferate in significant numbers, we restricted our assessment of proliferation to damaged alveolar regions of the lung and not the peribronchial/adventitial regions. These data also showed an overall increase in the number of alveolar Pdgfra-lineage cells which was stable to at least 28 days post-injury (Fig. 3d). To determine the importance of the expansion of AF1 cells in alveolar regeneration, we genetically deleted Ect2, a Rho-GEF protein required for cytokinesis, in Pdgfra expressing cells to block cell division (herein termed Ect2^Pdgfra-KO) (Fig. 3e, f). Cells lacking Ect2 can still undergo karyokinesis (nuclear division) but cannot undergo cytokinesis resulting in binucleation and a halt in cell division (Fig. 3f)(16–18). As expected, Ect2^Pdgfra-KO mutant cells became binucleated at 14 days post-injury, and these binucleated cells persisted until at least 28 days after injury (Fig. 3g, h). The majority of bi-nucleated Pdgfra-lineage cells resided in the alveolar region, with a minor contribution from peribronchial fibroblasts (Fig. S6a,b). Surprisingly, blockade of Pdgfra+ cell division improved overall structure and reduced injury severity at 28 days after influenza infection (Fig. 3i–k, S6c). These data suggest that Pdgfra+ cell division and expansion does not influence acute injury severity after influenza infection (i.e. 14 days after infection), but blocking this proliferative response allows the lung to better deploy euplastic regeneration after injury (i.e. 28 days after infection).

Figure 3: Selective inhibition of mesenchymal cell cytokinesis inhibits injury-induced tissue niche formation and promotes euplastic alveolar repair.

(a) UMAP of scRNA-seq from Pdgfra-lineage cells after influenza (sham, day 14, and day 28) highlighting proliferating cells along the predicted trajectory of AF1 cells differentiating into AF2 cells. (b) Experimental schematic depicting approach to evaluate and quantify proliferation of Pdgfra+ and Pdgfrb+ mesenchymal cells over time after influenza infection. (c) IHC of the lineage-reporter (tdTomato) and Ki67 showing Pdgfra+ cell pool proliferates and expands after injury while Pdgfrb+ cells do not proliferate. (d) Box plots showing quantification of percentage of tdTomato+ Ki67+ cells out of the total tdTomato+ cells for both Pdgfra and Pdgfrb-lineage over time after influenza infection and also total number of lineage-traced cells per image field. (e) Experimental schematic depicting approach to evaluate the function of Pdgfra+ cell pool expansion after influenza injury via deletion of Ect2. (f) Cartoon schematic depicting function of Ect2. Schematic made with Biorender.com. (g) IHC showing bi-nucleated Pdgfra-lineage trace cells after Ect2 deletion following influenza infection at days 14 and 28. (h) quantification of bi-nucleated cells in control vs knockout at days 14 and 28 after influenza. (i) representative H&E sections of the left lobes after Ect2 deletion following influenza infection. (j) injury-severity algorithm showing regions of normal (blue), moderately damaged (green), and severely damaged (red) lung tissue after injury. (k) quantification of injury severities as a percentage of total lung area. (l) IHC of the Pdgfra-lineage reporter (tdTomato), SP-C, and Ki67 showing increased AT2 cells proliferation after Ect2 deletion in Pdgfra+ cells. (m) quantification of AT2 cell proliferation from IHC images. (n) IHC of Pdgfra-lineage (tdTomato), SMA, and Keratin5 showing a reduction of Krt5+ cells in the alveolar space after Ect2 deletion in Pdgfra+ cells. (o) quantification of Krt5+ regions in control vs knockout mice. All scale bars represent 50 μm. Each data point in panels d, h, m, and o represents data obtained from a single mouse (i.e. biological replicate). n = 4–6 mice/group in panel k. All error bars represent SEM. *P<0.05, **P<0.01, and ***P<0.001, evaluated by un-paired t-test.

Since our histological analysis indicated that loss of AF1 cell expansion improved the overall lung architecture after injury, we assessed whether AT2 cell proliferation, a hallmark of alveolar regeneration, was affected in this model. Interestingly, blockade of AF1 cell expansion in Ect2^Pdgfra-KO mutants resulted in a significant increase in proliferation of AT2 cells at 14 days after injury compared to wildtype mice (Fig. 3l, m). Moreover, loss of AF1 cell expansion nearly eliminated the dysplastic Krt5+ cell response at 14 and 28 days after injury (Fig. 3n, o), demonstrating that expansion of AF1 cells is essential for dysplastic epithelial repair and suppresses the normal AT2 cell proliferative euplastic response after lung injury.

AF1-derived AF2s form an injury induced tissue niche with dysplastic Krt5+ epithelium

To identify putative factors regulating the function of AF1-derived AF2s and how these cells may signal to Krt5+ dysplastic epithelium, transcription factor activity was assessed in these cells using DoRothEA, a computational tool which identifies putative transcription factors regulating cell population specific transcriptional networks (Fig. 4a, S7a,b, Data S6)(19). One of the top ranked transcription factors in the AF1-derived AF2s was Rbpj, the downstream transcription factor in the Notch signaling pathway (Fig. 4b). GSEA analysis demonstrated AF1-derived AF2s are strongly enriched in expression of Notch signaling pathway components (Fig. 4c), including Notch receptors (e.g. Notch1, 2, 3) and downstream target genes (Heyl, Hey2, Ccnd1, Postn, Tbx2) (Fig. S7c). Spatial analysis of the Notch “active” Pdgfra cells assessed by intracellular Notch3-ICD immunostaining(20), revealed these cells are located in the dysplastic alveolar niche after injury, directly adjacent to the Krt5+ epithelium (Fig. 4d).

Figure 4: Pdgfra-derived AF2 cells are embedded within the injury-induced niche and are enriched for Notch signaling.

(a) UMAP of scRNA-seq data from Pdgfra-lineage cells at 28 days after influenza infection. (b) DotPlot showing activity of Rbpj, the main transcription factor downstream of Notch signaling, in each cell cluster shown in panel (a) demonstrating that Pdgfra-derived AF2 cells have the highest Rbpj activity relative to other Pdgfra-lineage labeled cell types at 28 days after influenza infection. (c) GSEA analysis showing Pdgfra-derived AF2 cells are enriched for downstream target genes of Notch signaling. (d) IHC at 90 days after influenza infection showing Notch+ Pdgfra-lineage cells are embedded within injury-induced niches adjacent to the dysplastic Krt5+ epithelium. (e) Experimental schematic depicting approach to temporally profile immune, endothelial, and epithelial cells using scRNA-seq after influenza infection and separately bleomycin. (f) Notch signaling network at 28 days after influenza infection showing a unique niche between the Krt5+ dysplastic basal cells, the primary sender of the Notch signals, and Pdgfra-derived AF2 cells, the primary receiver of Notch signals. (g) DotPlot showing Krt5+ basal cells express high levels of the Notch ligands Dll1 and Jag2 at 28 days after influenza infection. (h) IHC and quantification showing Jag2+ Krt5+ dysplastic epithelial cells exist directly adjacent to the Pdgfra-lineage mesenchymal cells at day 28 after influenza infection. n=3 mice. Error bars represent SEM. (i) Cartoon summary schematic made with Biorender.com. All scale bars represent 50 μm.

To identify which cell is producing the Notch ligands, we performed scRNA-seq on epithelial, endothelial, and immune cells after both H1N1 influenza infection and bleomycin induced injury at days 14 and 28 (Fig. 4e, S7d, e). We then performed CellChat analysis to identify proper sending-receiving cell patterns induced upon acute lung injury (Fig. 4f)(21). These data showed that Krt5+ dysplastic epithelial cells express high levels of Notch ligands, with the AF1-derived AF2s as the primary Notch receiver (Fig. 4f, g, S8a–c). Further analyses revealed that AF1-derived AF2 cells send BMP, TGFb, EGF, and Eph-Ephrin signaling back to the Krt5+ epithelium (Fig. S8b). Within the bleomycin dataset we similarly observed a strong Notch relationship between Krt5+ basal cells and AF1-derived AF2s along with minor differences such as a small contribution of AT2 and ciliated cells to the Notch network (Fig. S8a–c). IHC revealed that Jag2-expressing Krt5+ dysplastic basal cells are directly adjacent to the AF1-derived AF2s 28 days after influenza injury (Fig. 4h). These analyses demonstrate that AF1-derived AF2s exhibit active Notch signaling and generate a unique injury-induced niche with the Krt5+ dysplastic epithelium (Fig. 4i).

Lineage-selective inhibition of mesenchymal Notch signaling blocks dysplastic repair and promotes euplastic alveolar regeneration

To determine whether Notch signaling in AF1-derived AF2s is critical for formation of the injury-induced niche, Pdgfra^CreERT2:Rosa26^dnMAML mice (Notch^Pdgfra-KD) were used to block intracellular Notch signaling in AF1 cells. Using a geographical imaging approach to identify alveolar, peribronchial, and adventitial Pdgfra+ cells, we found robust Cre recombination of Rosa26^dnMAML in all three Pdgfra+ cell subtypes (Fig. S9a–c). Notch^Pdgfra-KD mice were subjected to influenza injury and examined at 14 days post-injury (Fig. 5a). Blocking intracellular Notch signaling within AF1s improved overall survival after influenza infection (Fig. 5b). Quantitative histological analyses from the lungs of Notch^Pdgfra-KD mice revealed a dramatic reduction in tissue damage and Krt5+ cells relative to wildtype mice post-influenza injury (Fig. 5c–e, S9c). While we still observed CD45+ immune cell infiltrate in the severely damaged regions of the lung at 14 days post injury (Fig. 5e), we identified no significant change in the total CD45+ cells in the infected lungs or of any other measured immune cell population (Fig. S10). The reduction in Krt5+ cells correlated with an increase in AT2 cell proliferation in the remaining damaged regions of the lung, indicating that suppression of mesenchymal intracellular Notch signaling rewires cell-cell communication in the alveolus to favor euplastic regeneration and reduce injury-induced niche formation (Fig. 5f). This was further correlated with a reduction in Pdgfra+ cell proliferation (Fig. 5g) and acquisition of a myofibroblast phenotype (Fig. 5h) consistent with prior reports demonstrating the role of Notch signaling in fibroblast to myofibroblast transition and fibroblast proliferation (22, 23). Importantly, we observed no change in overall tissue damage or Krt5+ cell egression after influenza when blocking Notch signaling in the Pdgfrb+ lineage (Fig. S11). Given the overlap of the Pdgfra+ and Pdgfrb+ expression in peribronchial and adventitial fibroblasts but not alveolar fibroblasts (Fig. 1a–c), these data indicate that activation of Notch signaling in AF1s, relative to other mesenchymal cell types, is necessary for the formation of the injury-induced niche.

Figure 5: Inhibition of mesenchymal Notch signaling blunts dysplastic alveolar remodeling.

(a) Experimental schematic showing approach to understand the function of intracellular Notch signaling in Pdgfra+ cells. (b) Survival curve of wildtype and Notch^Pdgfra-KD mice after influenza infection. P-values calculated from Log-rank (Mantel-Cox) tests. (c) H&E and injury severity zones of the left lobes from wildtype and Notch^Pdgfra-KD mice after influenza infection. (d) quantification of injury severity zones from wildtype and Notch^Pdgfra-KD mice. (e) IHC and quantification showing distribution of Krt5+ and CD45+ cell localization in Notch^Pdgfra-KD mice compared to wildtype mice. (f) IHC and quantification showing an increase in AT2 cell proliferation, as measured by Ki67 staining in Notch^Pdgfra-KD mice compared to wildtype mice. (g) IHC and quantification showing a reduction in Pdgfra+ cell proliferation in Notch^Pdgfra-KD mice compared to wildtype mice. (h) IHC and quantification showing a reduction in Pdgfra-lineage derived myofibroblasts in Notch^Pdgfra-KD mice compared to wildtype mice. Experimental data in panels f-h were generated with mutant mice containing both the Rosa26 tdTomato lineage trace allele and dnMAML mutant allele (as in Fig. S9a). (i) μCT scans of sham, wildtype (day 90 after influenza infection), and Notch^Pdgfra-KD (day 90 after influenza infection) showing a reduction in gross lung scar tissue formation in Notch^Pdgfra-KD mice compared to wildtype mice at 90 days after influenza infection. (j) Box plot of μCT analysis showing a rescue in total lung air volume in Notch^Pdgfra-KD (day 90 after influenza infection) relative to wildtype (day 90 after influenza infection). All scale bars represent 50 μm. All error bars represent SEM. Each data point in panels e-h represent data obtained from an individual mouse (i.e. biological replicate). n = 6 wildtype and 5 Notch^Pdgfra-KD mice/group in panels d and e. n=6 mice/group (for both conditions) in panels f-h. n=12 wildtype sham, n=10 wildtype H1N1 day 90, and n=15 Notch^Pdgfra-KD mice/group in panel i and j. Data are representative of at least 2 independent experiment except panel h which represents a pool from two independent experiments. *P<0.05, **P<0.01, ***P<0.001, and ****P<0.0001, evaluated by unpaired t-test (panels e-h) and one-way ANOVA with Tukey’s correction for multiple comparisons (panel j).

To further examine whether mesenchymal intracellular Notch signaling regulates persistence of the injury induced niche after influenza, micro-computed tomography (μCT) was performed on mice 90 days after influenza infection. Locations of the injury-induced niche were based on aerated lung volume analysis. Control influenza infected wildtype mice exhibited a significant reduction in aerated lung volume at 90 days compared to uninfected mice (Fig 5i, j), consistent with prior work demonstrating a reduction in pulmonary gas-exchange function in mice one year after a single influenza infection(8). Blocking Notch signaling in the AF1 lineage inhibited the reduction of aerated lung volume, further indicating that Notch activity in these mesenchymal cells is essential for promoting dysplastic repair over euplastic regeneration in the lung (Fig. 5i, j).

Since Notch signaling has been reported to be important in the dysplastic Krt5+ epithelium for the formation of the injury-induced niche, we inactivated Notch signaling directly in Krt5+ basal cells and examined the response to influenza induced lung regeneration using Krt5^CreERT2:Rosa26^dnMAML mice (here after called Notch^Krt5-KD) (Fig. S12a). In contrast to inactivating intracellular Notch signaling in Pdgfra+ cells, inactivating intracellular Notch in the Krt5+ dysplastic epithelium (Notch^Krt5-KD) did not improve survival after influenza injury (Fig. S12b). IHC analysis showed Krt5+ dysplastic epithelium continued to migrate into the severely damaged alveolar regions upon basal cell loss of Notch signaling (Fig. S12c). We next employed in vitro assays to assess Notch^Krt5-KD basal cell behavior. We first confirmed that expression of dnMAML in these cells in vitro strongly inhibited Notch signaling (Fig. S12d). However, genetic inactivation of Notch signaling did not have any effect on cell proliferation or organoid growth, in stark contrast to gamma-secretase inhibition with DBZ, which was previously shown to inhibit basal cell growth(7) (Fig. S12e–i). These data suggest that the phenotype upon gamma secretase inhibition in basal cells is likely due to off-target affects rather than Notch inhibition in these assays. Collectively, these data demonstrates that intracellular mesenchymal Notch signaling is the dominant mechanism which controls the generation of the dysplastic repair response and injury-induced niche formation after acute lung injury.

Blockade of injury-induced niche formation via Notch inhibition increases flow and direction of Wnt and Fgf signaling in the alveolar niche

To identify the mechanism by which blocking Notch signaling within mesenchymal cells rewires the alveolar niche to support euplastic over dysplastic responses, we performed scRNA-seq on whole lungs from wildtype and Notch^Pdgfra-KD mice at 14 days post-influenza infection (Fig. 6a, S13a, b). These data revealed three distinct AF1 cell populations which we named AF1a, b, and c (Fig. 6b). While AF1a was composed of relatively equal percentages of cells derived from wildtype and Notch^Pdgfra-KD lungs, AF1b was strongly enriched for AF1 cells derived from wildtype mice, while AF1c was strongly enriched for AF1s derived from Notch^Pdgfra-KD mice (Fig. 6c). None of the remaining mesenchymal cell clusters were dominated by either wildtype or Notch^Pdgfra-KD cells, suggesting that inactivating intracellular Notch within the Pdgfra-lineage preferentially regulates the AF1 cell response after influenza injury (Fig. 6c). We next performed pathway analysis using the genes uniquely enriched in either the AF1b or AF1c cell clusters. We found that AF1b was enriched in gene expression for ECM-receptor interactions, proteoglycans, and focal adhesions (Fig. 6d). The AF1c cluster dominated by Notch^Pdgfra-KD cells was enriched in gene expression for the MAPK, TNF, and PI3K-Akt signaling pathways and genes such as Fgf7, Vegfa, and Wnt2, important regulators of alveolar regeneration (Fig. 6d). These data indicate that inactivation of Notch signaling in the Pdgfra-lineage blunts the fibrotic and myofibroblast response and upregulates euplastic regenerative signaling patterns in support of AT2 cell proliferation and survival including Wnt and Fgf (11, 17, 24–26). Consistent with our IHC analyses in Figure 5e–f, the scRNA-seq analysis of the lung epithelium in Notch^Pdgfra-KD animals revealed a stark reduction of Krt5+ basal cells in the Notch^Pdgfra-KD dataset, with an increase in total AT1 and AT2 cells (Fig. 6e, f).

Figure 6: Mesenchymal Notch signaling rewires the alveolar signaling niche.

(a) Experimental schematic showing approach to identify the mechanism by which inhibiting intracellular mesenchymal Notch signaling rescues euplastic alveolar regeneration. (b) UMAP of mesenchymal cells showing three distinct clusters of AF1 mesenchymal cells. (c) Contribution of each condition (wildtype and Notch^Pdgfra-KD) to the cell clusters shown in panel (b) showing AF1b is enriched for wildtype AF1 cells and AF1c is enriched for Notch^Pdgfra-KD AF1 cells. (d) KEGG pathway analysis of the gene expression programs enriched in AF1b and AF1c clusters and violin plots of relevant genes from the highlighted pathways. (e) UMAP of epithelial cells. (f) Contribution of each condition (wildtype and Notch^Pdgfra-KD) to each cell cluster shown in panel (e) showing strong reduction in Krt5+ basal cells after Notch inactivation in the Pdgfra+ cells. (g) Differential pathway information flow for Fgf and Wnt signaling showing an overall system wide increase in Fgf and Wnt signaling in Notch^Pdgfra-KD mice compared to wildtype mice. (h) Wnt signaling network in wildtype and Notch^Pdgfra-KD conditions showing a re-wiring of the Wnt signaling niche after Notch inactivation in the Pdgfra+ cells. (i) Violin plots showing an upregulation of Wnt ligands (Wnt2) and receptors (Fzd5) in AF1 and AT2 cells respectively in Notch^Pdgfra-KD mice compared to wildtype. (j) cartoon schematic, generated with Biorender, depicting in vitro organoid assay approach. (k) live fluorescent images and quantification showing representative organoid sizes after 14 days in culture. Scale bar represents 1mm. (l) Cartoon summary schematic generated with Biorender. Each dot in panel (k) represents a biological replicate. *P<0.05 evaluated by unpaired t-test.

We next examined differential signaling networks between wildtype and Notch^Pdgfra-KD cells that could explain the rewiring of the injury-induced niche upon loss of mesenchymal Notch signaling. This analysis showed an overall increase in outgoing signals emanating from AF1s and an increase in both outgoing and incoming signals into AT2s in Notch^Pdgfra-KD mutants (Fig. S13c). Cell type specific analysis revealed an increase in AF1-to-AT2 communication after Notch inactivation (Fig. S14a, b). Pathway specific analysis revealed an increase in Fgf and Wnt signaling in Notch^Pdgfra-KD mutants compared to wildtype (Fig. 6g, S14c). In control mice, the dysplastic Krt5+ epithelium was the dominant sender of Wnt signaling. However, in Notch^Pdgfra-KD mutants the Wnt signaling network is rewired with the new dominant Wnt signaling relationship being between AF1 and AT2 cells (Fig. 6h). Consistent with this new signaling relationship, we observed an increase in Wnt ligand expression in AF1s including Wnt2, and an increase in Wnt receptor expression in AT2s including the pro-regenerative receptor Fzd5 in Notch^Pdgfra-KD mutants (Fig. 6i)(27). Furthermore, in Notch^Pdgfra-KD mutant lungs we observed a reduction in BMP, FGF, and Gas signaling from AF1 to Krt5+ dysplastic basal cells after injury demonstrating that intracellular Notch signaling in the mesenchyme directs the signaling network within the alveolus and can be targeted to promote epithelial euplastic repair (Fig. S14d). These data reveal a change in Wnt and Fgf signaling flow and direction in the alveolus depending on the Notch signaling status in AF1 cells. While prior literature has already examined the importance of fibroblast derived Wnt and Fgf signaling in maintaining both the mouse and human alveolar AT2 cell niche(11, 17, 25, 26), we further examined whether inhibition of mesenchymal Notch directly regulates AT2 cell self-renewal and expansion. To test this, we performed in vitro organoid co-culture assays, combining wildtype AT2 cells with Notch mutated fibroblasts (Fig. 6j, k). These organoid assays showed that inhibition of mesenchymal Notch signaling directly increases AT2 cell organoid size demonstrating that blocking Notch in the Pdgfra+ cells directly supports AT2 cell proliferation and expansion (Fig. 6k).

To determine whether blocking intracellular Notch signaling in the Pdgfra-lineage prevents AF1-to-AF2 differentiation, we performed trajectory analysis on the lung mesenchyme in control and Notch^Pdgfra-KD mutant mice resulting in two primary trajectories (Fig. S13i). Trajectory 1 traveled from AF1a through AF1b, the AF1 subcluster dominated by wildtype cells (Fig. S13i) and exhibited a loss of AF1 markers (Pdgfra and Wnt2), with a concordant increase of AF2 and myofibroblast markers (Pdgfrb and Acta2) suggesting these cells are transiting to becoming AF2s (Fig. S13i, j). In contrast, trajectory 2 traveled from AF1a through AF1c, the AF1 subcluster dominated by Notch^Pdgfra-KD mutant cells (Fig. S13j). Dynamic gene expression through trajectory 2 revealed that these cells maintained high expression of AF1 markers (Pdgfra and Wnt2) suggesting that after influenza injury, AF1 cells transit to become AF2 cells and that this decision is dependent on Notch resulting in opposing signaling by Wnt and Fgf pathways (Fig. S13j). Together, these data show that loss of Notch signaling in AF1s leads to their reprogramming and shunting into a pro-euplastic regenerative phenotype, which inhibits dysplastic lung repair (Fig. 6l).

Segregation of human lung disease phenotypes by the presence of the injury-induced niche and rewiring of alveolar signaling flow

Human acute lung diseases that can result in dysplastic and fibrotic responses include post-COVID-19 lung disease and bleomycin-induced lung injury (28, 29). Phenotypes in both of these diseases are initiated by acute lung injury. In contrast, chronic lung diseases such as chronic obstructive pulmonary disease (COPD) are characterized by a degenerative emphysematous phenotype leading to a loss of alveolar structure, which is also exhibited by Alpha 1 anti-Trypsin deficiency (AAT), a genetic form of COPD(30). The role of dysplastic epithelial responses in these diseases is not well established, although the presence of KRT5+ epithelium in the alveolar compartment has been reported in a variety of fibrotic lung diseases(8).

To determine whether the rewiring of the alveolar signaling crosstalk by the injury-induced niche could be used to segregate different phenotypes of human lung disease, a scRNA-seq dataset from bleomycin-induced human lung disease and Alpha-1 anti-Trypsin human lung disease were newly generated and combined with previously published post-COVID-19 fibrotic lung disease datasets(31), to determine whether pathologic alveolar signaling networks driven by the injured niche are present in these diseases. Consistent with the mouse mesenchyme data, analysis of human mesenchymal cells from healthy control lungs revealed 5 distinct mesenchymal cell populations including the two distinct populations of alveolar mesenchymal fibroblasts seen in mice: AF1s (PDGFRA+ TCF21+) and AF2s (PDGFRB+ NOTCH3+) (Fig. S15a–e, Data S7). At homeostasis, AF1s and AF2s exist in nearly equal ratios in the human lung (Fig. 7a, b). However, we observed a stark increase in both AF1s and AF2s in the diseased COVID-19 lung, with an elevated level of AF2s compared to AF1s, suggesting AF2s could be an important cell type in human dysplastic lung repair (Fig. 7b). To determine whether AF2s and dysplastic basal cells communicate to form an injury-induced NOTCH niche in human post-COVID-19 and bleomycin-induced human lung disease, single cell data was analyzed to examine ligand-receptor interactions between the epithelial and mesenchymal cell populations. These data revealed a robust NOTCH niche in the post-COVID-19 and bleomycin human lungs between the AF2s and the KRT5+ basal cells (Fig. 7c). This is supported by IHC showing AF2s, but not AF1s, are present adjacent to KRT5+ dysplastic epithelial cells in post-COVID-19 fibrosis and acute bleomycin injury (Fig. 7d–f, S15f–i).

Figure 7: Phenocopying chronic human lung diseases using NOTCH, WNT, and FGF mesenchymal-epithelial signaling networks.

(a) representative IHC and RNAscope image showing distribution of PDGFRA/AF1 and PDGFRB/AF2 cells in the healthy/control human lung. (b) quantification of destruction of PDGFRA/AF1 and PDGFRB/AF2 cells in healthy vs COVID-19 diseased lungs. (c) Notch signaling network between mesenchymal and epithelial cells from a scRNA-seq multi-disease dataset of a variety of human lung diseases along with representative H&E tile scans of the distal lung isolated from each disease. (d) IHC and RNA in situ hybridization showing PDGFRB+ mesenchymal cells exist directly adjacent to KRT5+ dysplastic epithelium in COVID-19 induced human lung injury. (e) distribution of PDGFRA/AF1 and PDGFRB/AF2 cells to the nearest KRT5+ cell in COVID-19 diseased human lungs (n=4 biological replicates/donors). (f) quantification showing the % of PDGFRA/AF1 and PDGFRB/AF2 cells within 10μm to the nearest KRT5+ cell. Each data point represents a single biological replicate/donor. (g) WNT ligand module score in AF1 cells from fibrotic (COVID-19 & bleomycin) human lungs and degenerative (AAT & COPD). (h) DotPlots showing relative expression level of FGF and WNT ligands and receptors within AF1 and AT2 cells from control, COVID-19, and COPD human lungs. *P<0.05, **P<0.01 evaluated by paired t-test.

In contrast to the fibrotic lung diseases, we observed an increase in WNT ligand expression in AF1 cells from both COPD and AAT lungs compared to normal donor lungs, suggesting a reversal of a NOTCH/WNT axis in degenerative versus fibrotic alveolar diseases (Fig. 7g). This is reflected by a reduction in overall WNT signaling with decreased expression of FGF and WNT ligands (i.e. FGF2 and WNT2) in AF1s and receptors (i.e. FGFR2 and FZD5) in AT2s in fibrotic diseases (Fig. 7h). This is in stark contrast to COPD and AAT lung disease, where a rewiring of the NOTCH mediated signaling niche occurs and KRT5+ cells are no longer the dominant sender of the NOTCH ligands (Fig. 7c). Moreover, COPD and AAT do not exhibit dysplastic injury-induced niche formation (Fig. 7c). Remarkably in COPD and AAT, there is an increase in FGF and WNT ligands and receptors in AF1 and AT2 cells, mimicking the euplastic regenerative response in the mouse lung upon loss of Notch signaling in AF1 derived AF2s and suggesting that in these degenerative diseases, a euplastic regeneration response is being engaged albeit inefficiently (Fig. 7g,h). These findings demonstrate a unique injury-induced tissue niche is formed between the mesenchyme and dysplastic epithelium in human fibrotic but not degenerative lung diseases. While these data suggest that the presence of such a niche, which balances the dysplastic versus euplastic response, is conserved in the mouse and human lung, further experiments are required to delineate this process in human model systems.

Discussion

The cellular and architectural complexity of the respiratory system has made it difficult to unravel the specific cell lineages responsible for promoting functional lung regeneration. Our study reveals that the decision between euplastic regeneration and dysplastic repair is governed by Notch signaling in a specific mesenchymal lineage named AF1 cells, which form an injury-induced niche with the airway-derived keratinized basal epithelium. This new niche results in a maladaptive structure which does not effectively participate in gas exchange. The AF1 response to acute injury involves transiting across multiple cell states, including proliferation and myofibroblast, before assuming an AF2 phenotype, which is characterized by high Notch signaling activity. This AF1-AF2 plasticity and expansion is hijacked by airway-derived Krt5+ dysplastic epithelial cells invading the alveolar niche to establish the injury-induced niche. Signaling patterns from this new niche can be used to phenotype human lung diseases into either acute fibrotic responses such as COVID-19 or bleomycin-induced acute injury, or degenerative phenotypes such as COPD and AAT deficiency. This phenotypic switch is characterized by altered signaling flows of hallmark regenerative pathways including WNT and FGF in the alveolar niche. This study reveals an emergent signaling niche in injured and diseased lungs that defines the decision between euplastic regeneration or dysplastic repair, providing critical insight into phenotyping human lung disease responses.

The role of mesenchymal-epithelial signaling in organ development has been well described(32–34). In the lung, changes in phenotypes within the mesenchymal lineage occurs throughout development with the early lung mesenchyme being derived from a Wnt2+/Pdgfra+ progenitor cell(32). While there are likely transcriptional differences from the developmental Pdgfra+/Wnt2+ mesenchymal progenitor cell and the adult AF1 cell (which also expresses Pdgfra/Wnt2), our study suggest that Pdgfra+ Wnt2+ alveolar mesenchymal cells in the adult lung called AF1 cells retain their developmental capacity to regenerate and replenish the non-proliferative Pdgfrb+ AF2 cell pool after injury, possibly to regenerate the damaged capillary plexus (12). Recent work has further highlighted a similar lineage relationship during normal intestinal development and turnover(35). Our current study shows that the dysplastic Krt5+ epithelium hijacks this AF1 cell plasticity, using a paracrine signaling network instigated by intracellular mesenchymal Notch signaling to form an injury-induced niche. While previous work has suggested Notch signaling within the basal cell lineage as a driving factor in formation of Krt5+ dysplastic epithelium(7), our current study demonstrates that Notch signaling in AF1 cells is the dominant driver of this phenomena to control injury-induced niche formation. A previous study also suggests that Gli1+ mesenchyme can regulate the dysplastic epithelial response due to acute lung injury(36).

The Krt5+ cells that invade the alveolar space of the lung after acute injury are derived from intrapulmonary airway-derived basal stem cells in the mouse(7, 9). Their ability to rapidly expand and migrate into a completely different compartment of the lung and establish an injury-induced niche in partnership with mesenchymal cells, demonstrates the potent ability of basal stem cells to respond to an acute insult and the necessity of these cells to organize as a Notch dependent niche. This invasion competes with the euplastic regenerative response which is governed by AT2 cell proliferation and AT2-AT1 differentiation after acute injury(4, 5). The balance between the airway and alveolar reactions to injury and certain lung diseases is an example of a “stem cell collision” that leads to the rewiring of the alveolar niche and a long-term reduction of gas exchange function.

The ability of AF1 cells to differentiate into AF2 cells likely plays a critical role in replacing lost AF2s upon injury, or during homeostatic turn-over, since AF2s do not proliferate in response to injury. Severe injury can cause an imbalance in multiple processes that may not occur in more moderate injuries, including significant loss of AT1 and AT2 cells due to cytotoxic cell death in influenza infection or bleomycin-induced acute pulmonary toxicity. This may lead to a larger degree of AF1 proliferation and AF1-AF2 differentiation than in less severe injuries which do not result in a dramatic loss of alveolar epithelium. Denuding the alveolar epithelium would provide the invading Krt5+ basal cells an open environment to establish themselves, driving injury-induced niche formation. The high level of injury-induced niches in both COVID-19 and bleomycin acute injured lungs suggests that this process is conserved in acute injury in the human lung.

Repetitive acute injuries may underlie the emergence or progression of chronic lung disease(37, 38). Most of the respiratory system is quiescent at homeostasis but reacts rapidly to injury, leading to the concept that it is a facultatively regenerative organ. While AT2 cells are well characterized as the resident alveolar epithelial stem/progenitor cell, they have critical roles in normal homeostatic lung function including surfactant production and recycling as well as an important immune sensing role(1, 39). AT2 progenitor activity is active at only a very low levels during homeostasis where they slowly self-renew and differentiate into AT1 cells(17). In contrast, basal cells are a dedicated stem cell lineage whose primary purpose is to harbor the cellular and genetic information for regenerating and replacing all airway epithelial lineages during homeostatic turn over and after injury(4). Given the dedicated nature of basal stem cells, it is not surprising that they can outcompete AT2 cells for repopulating severely damaged regions of the lung. Targeted approaches which suppress overactive basal cell behavior may allow for the euplastic regenerative process to proceed more efficiently, improving long term health of the respiratory system, as evidenced by recent work demonstrating that deletion of the core basal-cell transcription factor Trp63 prevents Krt5+ dysplasia and improves lung function(8). Future studies in this direction are warranted in addition to directly promoting the euplastic regenerative response such as recently demonstrated for Wnt signaling(27).

Our studies highlight the importance of defining the cellular and molecular differences in acute and chronic lung disease. The presence of injury-induced niches and their signaling activity in acutely injured human lungs from COVID-19 patients and bleomycin-induced injury suggests a conserved mechanism of dysplastic repair in these disease states. Conversely, the lack of injury-induced niches and a reversal of signaling information flows, in particular increased WNT and FGF signaling in degenerative emphysematous diseases, suggests that in these disease states the lung has activated pathways known to promote regeneration. However, given the degenerative phenotype in these diseases, increased WNT and FGF activity may either be insufficient or is blocked downstream. This finding points to the need for increased focus on why the alveolar niche does not respond efficiently to increased WNT and FGF and indicates the importance of other possible mechanisms driving these diseases including destruction of the supportive matrix and epigenetic regulation of transcriptional networks regulating cell survival and proliferation.

Methods

Human subjects

The normal human samples used in this study were from de-identified non-transplanted lungs initially donated for organ transplantation following an established protocol (Prospective Registry of Outcomes in Patients Electing Lung Transplantation (PROPEL), approved by University of Pennsylvania Institutional Review Board) with informed consent in accordance with institutional and NIH procedures. Consent was provided by next of kin or healthcare proxy. Diseased tissue was obtained from subjects enrolled at the University of Pennsylvania as part of the PROPEL (Penn cohort)(40) (IRB # 849795). All selected subjects had diseases listed as classified by a multidisciplinary clinical team. For subjects with COPD (both typical and secondary to alpha-1 antitrypsin deficiency), none had evidence of other parenchymal disease, and none were on systemic steroids. The institutional review board of the University of Pennsylvania approved this study, and all patient information was de-identified before use. While this tissue collection protocol does not meet the current NIH definition of human subject research, all institutional procedures required for human subject research at the University of Pennsylvania were followed throughout the reported experiments. The sample age and pertinent deidentified clinical information are listed in Table S1.

Mouse lines

All mouse experiments were performed under the protocols approved by the guidance of the University of Pennsylvania Institutional Animal Care and Use Committee (IACUC # 806345). Pdgfra^CreERT2, Pdgfrb^CreERT2, Acta2^CreERT2, Krt5^CreERT2, Sftpc^CreERT2, Pdgfra^H2B-eGFP, Rosa26^tdTomato, Rosa26^YFP, Ect2^flox, and Rosa26^dnMAML-GFP mouse lines have been previously described(18, 41–48). All experiments were performed on 8–16 week old mice that were maintained on a mixed C57BL/6 and CD1 background. Both male and female mice were used in all experimental groups.

Tamoxifen delivery

For Cre recombinase induction, tamoxifen (Millipore Sigma) was dissolved in corn oil and ethanol mixture (90%/10%, v/v) (Millipore Sigma) to produce a stock solution with a concentration of 20mg/mL. All mice received tamoxifen through intraperitoneal injections. Pdgfra^CreERT2 , Pdgfrb^CreERT2, and Sftpc^CreERT2 mice were given tamoxifen at a dose of 200mg/kg for 3 consecutive days. Acta2^CreERT2 mice were given tamoxifen at a dose of 100mg/kg for 4 consecutive days at the timepoints indicated in Figure 2. Krt5^CreERT2 mice were given a dose of tamoxifen (250mg/kg) at days 5 and 8 after influenza, as indicated in Figure S12.

Influenza infection

PR8-GP33 H1N1 influenza virus was provided by Dr. John Wherry at the University of Pennsylvania. Two weeks after tamoxifen induction, mice were intranasally administered a dose of 1 LD₅₀ (determined empirically in our laboratory) diluted in 50 μL sterile saline as previously described(26, 49).

Bleomycin administration

Two weeks after tamoxifen induction, 3.5U/kg bleomycin (Teva), diluted in PBS (Thermo Fisher Scientific), was intratracheally administered to anesthetized mice. Control mice received sterile PBS intratracheally.

Histology, IHC, and RNAscope

Mice were euthanized by CO₂ inhalation and the lungs were perfused with ice-cold PBS through the right ventricle. The lungs were then inflated with 2% PFA (Thermo Fisher Scientific) at a constant pressure of 25 cm H₂O and were fixed overnight at 4 °C. For human tissue, sections of the peripheral parenchyma approximately 1cm³ were dissected from the distal lung and placed in 2% PFA or 4% PFA depending on future use (IHC and RNAscope, respectively) overnight at 4 °C. Tissue was then dehydrated in a series of ethanol concentration gradients, paraffin embedded, and 6μm thick sections were cut. Hematoxylin and eosin staining was performed as previously described(17).

For IHC, the following antibodies were used on paraffin sections: RFP (goat, Origene, cat# AB8181–200), RFP (rabbit, Rockland, cat# 600–401-379), GFP (chicken, Aves, cat# GFP-1020), GFP (goat, Abcam, cat# ab6673), SMA (goat, Novus Biologicals, cat# NB300–978), SMA (mouse, Sigma, cat# A5528), SMA (rabbit, Abcam, cat# ab5694), Ki67 (mouse, BD Biosciences, cat# 550609), SP-C (rabbit, Millipore-Sigma, cat# AB3786), Keratin5 (rabbit, Abcam, cat# ab52635), Keratin5 (mouse, Sigma, AMAb91549), Notch3-ICD (mouse, Santa Cruz, cat# sc-515825), Jag2 (rabbit, LSBio, cat# LS‑C100395), CD45 (rat, Novus Biologicals, cat# NB100–77417), Pdgfra (goat, R&D, cat# AF1062), Pdgfrb (goat, R&D, cat # AF1062), and Erg (mouse, Abcam, cat# ab214341).

For RNAscope, lung tissue was prepared, fixed, embedded in paraffin, and sectioned as described above. RNAscope was performed using the Fluorescent Multiplex Reagent Kit v2 (ACDBio) according to manufacturer’s instructions. All RNAscope probes were purchased through ACDBio. The following RNAscope probes were used: Tcf21 (mouse, cat# 508661-C3), Notch3 (mouse, cat# 425171-C2), Pi16 (mouse, 451311-C2), Aspn (mouse, 502051), PDGFRA (human, cat# 604481-C3), PDGFRB (human, cat# 548991-C2), TCF21 (human, cat# 473071), and NOTCH3 (human, cat# 558991).

Imaging and image analysis

Fluorescent images were acquired at 20x and 40x using z-stacks on an LSM 710 laser scanning confocal microscope (Zeiss) and Stellaris 5 laser scanning confocal microscope (Leica). Cell counting analysis was completed within FIJI. H&E stained lung sections were tile-scanned using a 4x objective on the Eclipse Ni series upright microscope (Nikon). Whole lobe immunofluorescence tile scans in Figure 5e were acquired using a 4x objective on the Nikon Eclipse Ni series upright microscope. Organoid images in Figure 6k were captured using an EVOS FL Auto 2 imaging system using a 1.25x objective. Organoid sizes were calculated in FIJI as previously described (17).

Tissue clearing, sectioning, and imaging

Pdgfra^H2B-GFP:Pdgfrb^CreERT2:Rosa26^LSL-tdTomato mice were given a single dose of tamoxifen as described above. One week later, mice were euthanized, and lungs were perfused with PBS, and inflated with 2% PFA as described above. Lungs were fixed overnight at 4 °C with gentle agitation. Lungs were then cleared using the EZclear protocol(50). Briefly, the next morning, lungs were washed 4× 30 minutes each in PBS then incubated with 50% (v/v) THF (Sigma-Aldrich) in sterile milli-q water for 16 hours. Lungs were then washed 4× 1 hour each in sterile milli-q water with gentle agitation. Lungs were then washed 4× 1-hour each in PBS then once more overnight at room temperature with gentle agitation. Lungs were then immersed in a three-step sucrose gradient (Fisher Scientific) (10%, 20%, and 30%, prepared in 1x PBS). For each step, lungs were incubated in sucrose overnight at 4 °C with gentle agitation. After incubation with 30% sucrose, lungs were embed in OCT medium (Sakura) and snap-frozen on a bed of crushed dry ice, then stored at −80 °C until sectioned. OCT embedded samples were sectioned (200μm) on a cryostat (Epredia, HMH525 NX) and then placed in ice-cold PBS overnight. The next day, sections were placed in EZ view sample mounting and imaging solution (80% Nycodenz (Accurate Chemical & Scientific), 7M urea (Sigma-Aldrich), 0.05% sodium azide (Sigma-Aldrich), in 0.02M sodium phosphate buffer) overnight at room temperature. The next day, cleared sections were mounted on a glass slide in EZ view mounting media and imaged using a LSM 710 laser scanning confocal microscope using a 25x oil immersion objective. Images were analyzed in ImageJ.

H&E assessment program

To assess injury severity based on H&E staining, we used a modified version of a previously published algorithm(17). Using the EBImage package with R(51), we imported images and converted RGB pixel intensity to a matrix, we then applied K-means clustering to segment out, in an unbiased manner, 4 different regions of injury severity: normal, moderate, severe, and background. These regions where then color coded as follows: blue, green, red, and black. Using the area of the lobe outlined in H&E, the percent of each injured zone (normal, moderate, and severe) where then calculated as the percent of total lung lobe area. This R script has been uploaded to the Morrisey Lab Github page (https://github.com/Morriseylab).

Lung tissue digestion, flow cytometry, and FACS

Lungs were harvested and digested into single cell suspensions using collagenase I, dispase, and DNase I, as previously described(11). Red blood cells were removed with ACK lysis buffer and then the cell suspension was stained with antibodies diluted in ice-cold FACS buffer (PBS, 25mM HEPES (Thermo Fisher Scientific), 2mM EDTA (Invitrogen), and 2% Fetal Bovine Serum (FBS) (Deville)). The following antibodies were used for flow cytometry and cell sorting: CD45-PeCy7 (1:200, ThermoFisher, clone: 30-F11, cat# 25–0451-82), CD45-APC (1:200, ThermoFisher, clone: 30-F11, cat# 17–0451-83), CD31-APC (1:100, ThermoFisher, clone: 390, cat# 17–0311-82), and CD326(aka EpCAM)-APC (1:300, ThermoFisher, clone: G8.8, cat# 17–5791-82); DAPI was used to gate out dead cells. All cells were sorted using a 100μm sized nozzle. All cells were sorted into ice-cold FACS buffer. Flow cytometry and cell sorting experiments were performed using the cell analyzer LSR Fortessa (BD Biosciences), the cell sorter Aurora CS (Cytek), and the cell sorter FACSJazz (BD Biosciences).

For flow cytometry data in Figure S10, lungs were harvested from mice, digested using collagenase II and DNase I, and single-cell suspensions were prepared as previously described(52). After 1 hour at 37 °C, lung digestion solution was mechanically dissociated by pipetting in sorting buffer (DMEM (Gibco, cat # 11965092) + 2% FBS (Deville) + 1% pen/strep (Gibco, cat # 15140122)) (herein referred to as “SB”). Next, cell suspensions were filtered through a 70-μm filter and treated with ACK lysis buffer (containing DNase I) for 5 min at room temperature. Single-cell suspensions were then blocked in SB containing 1:50 mouse TruStain FcX (BioLegend, cat # 101320) for 5–10 min at RT. The cell suspension was then stained with the following: cell viability dye (1:1000, eBioscience^™, cat # 65–0865-18), Brilliant Violet 785 anti-mouse CD45 (1:200; BioLegend, clone: 30-F11, cat # 103149), Brilliant Violet 711 anti-mouse/human CD11b (1:200; BioLegend, clone: M1/70, cat # 101241), Brilliant Violet 605 anti-mouse Ly-6C (1:100; BioLegend, clone: HK1.4, cat # 128036), Alexa Fluor 700 anti-mouse Ly-6G (1:200; BioLegend, clone: 1A8, cat # 127622), FITC anti-mouse CD11c (1:100; BioLegend, clone: N418, cat # 117306), PE/Cyanine5 anti-mouse I-A/I-E (1:1000; BioLegend, clone: M5/114.15.2, cat # 107612), PE anti-mouse Siglec-F (1:100; BD Bioscience, clone: E50–2440, cat # 552126), Brilliant Violet 421 anti-mouse F4/80 (1:100; BioLegend, clone: BM8, cat # 123132), PE/Cyanine7 anti-mouse CD64 (FcγRI) (1:200, BioLegend, clone: X54–5/7.1, cat # 139314); PE/Cyanine5 anti-mouse CD3e (1:200; BioLegend, clone: 145–2C11, cat # 100310), Alexa Fluor 700 anti-mouse/human CD45R/B220 (1:50; BioLegend, clone: RA3–6B2, cat # 103232), PE anti-mouse Ly-6G (1:200; BioLegend, clone: 1A8, cat # 127608), BUV395 anti-mouse CD11b (1:200; BD Biosciences, clone: M1/70, cat # 565976), Brilliant Violet 421^™ anti-mouse NK-1.1 (1:100; BioLegend, clone: PK136, cat # 108731), PE-Cyanine7 CD127 Monoclonal (1:100; Invitrogen, clone: A7R34, cat # 25–1271-82) for 30 min at 4 °C. “Fluorescence minus one” (FMO) controls were used for each antibody. Stained cells and FMO controls were then resuspended in sorting buffer. Flow cytometry was performed on a BD FACSymphony A3 Cell Analyzer (BD Biosciences) and analyzed within FlowJo (BD Biosciences).

scRNA-seq and analysis

Lineage-traced Pdgfra/Pdgfrb cells were sorted by gating DAPI⁻, CD45-PeCy7⁻, CD31/326-APC⁻, tdTomato⁺ populations. The tdTomato+ cells were then loaded onto the 10x Chromium controller (10X Genomics). Epithelial and endothelial cells were sorted by gating DAPI⁻, CD45-PeCy7⁻, CD31/326-APC⁺. Immune cells were sorted by gating DAPI-, CD45-PeCy7⁺. Immune cells were then spiked back in with the epithelial and endothelial cells at a concentration of 25%. The resultant epithelial, endothelial, and immune cell mixture was then loaded onto the 10x Chromium controller (10x Genomics).

For the whole-lung scRNAseq presented in Figure 6, mouse lungs were digested and single cell suspensions were prepared as described above. Cell suspensions were then incubated with magnetic CD45 microbeads (Miltenyi Biotec, 130–052-301) for 30 minutes on ice. Cells were then passed through a magnetic column (LS column, Miltenyi Biotech) to deplete the cell suspension of CD45+ immune cells. CD45+ immune cells were then later recovered and spiked back into the CD45-depleted cell suspension at a concentration of 20%.

For human scRNAseq experiments, lungs were processed and CD45 cells were depleted using the MACS LS columns as previously described(53). CD45 positive cells were not spiked back for human samples. The resultant cell mixture as then loaded onto the 10x Chromium controller.

All samples were loaded to aim for a recovery of 10,000 cells, and libraries were prepped according to the manufacturer’s protocol using the Chromium Single Cell 3’ v3.1 chemistry. Libraries were then sequenced on an Illumina Novaseq 6000 instrument. The sequenced data was processed by aligning reads and obtaining unique molecular identifiers (UMIs) using STARsolo (v2.7.9a)(54). For scRNA-seq from lineage-traced samples, libraries were aligned to a modified mm39 mouse genome to include the tdTomato reference sequence. The scRNA-seq data was further processed and analyzed using the Seurat v4 package(55). Cells which had less than 200, or greater than 5000, detected genes were removed from the analysis. Cells were also removed if the percent mitochondrial reads were greater than 10%. For the human bleomycin lung single cell data in Figure 7, a percent mitochondrial read cutoff of 30% was used. Feature (gene) data was scaled in order to removed unwanted sources of variation using the Seurat SCTransform function based on percent mitochondrial reads, and the number of genes, and total reads. Integrated libraries were integrated using the RPCA method within Seurat. Non-linear dimension reduction was performed using uniform manifold approximation and projections (UMAPs) (reduction = ‘pca’ and n.neighbors = 15) and Louvain graph-based clustering algorithms. Previously published healthy, COPD, and COVID-19 human lung scRNA-seq data were downloaded from GSE168191 and GSE159585(31, 53).

Marker genes for each cell type were identified using the FindAllMarkers command using the RNA assay within Seurat. Mesenchymal cells were annotated based on their anatomical location in the lung (i.e. alveolar fibroblast 1 and 2). Peribronchial, adventitial, and vascular smooth muscle cells were annotated using previously published canonical marker genes(10). Epithelial, endothelial, and immune cells were annotated using marker genes and LungMAP labels(13).

Spearman correlation analysis was performed in R. Slingshot trajectory analysis was performed using the SCP package within R using default settings anchoring the trajectory at the cluster enriched for AF1 cells from sham lungs(56). Transcription factor activity scores were assigned to cells using the DoRothEA and decoupleR database within R(19). GSEA analysis for Notch signaling (GO:0007219) was performed using the SCP package within R using default settings. Ligand-receptor analysis was performed using CellChat v1(21). Prior to ligand-receptor analysis, each cell type was randomly down sampled to 300 cells per cell type to minimize artifacts arising due to discrepancy in cell numbers. After down sampling, CellChat was performed with default settings. We performed ligand-receptor analysis using the ‘Secreted Signaling’ and ‘Cell-Cell Contact’ ligand-receptor databases, excluding ‘ECM-Receptor’ signaling. For the Notch signaling network in Figure 7, both immune and endothelial cells were removed from the ligand-receptor analysis to allow focus on mesenchymal-epithelial niche signaling in disease. Differential ligand-receptor analysis in Figure 6 was performed using CellChat with default settings. GO and KEGG pathway analysis was performed using the clusterProfiler R package using default settings(57). WNT ligand module scores in Figure 7 were generated using the AddModuleScore function within Seurat using all human canonical WNT ligands. Normalized module scores were then converted into a heatmap using the heatmap function within base R. Visualization of corrplot, Slingshot, transcription factor activity scores, GSEA, and CellChat were generated using functions within each software package, base R, or ggplot2. UMAPs, violin plots, and dot plots were generated using both Seurat and scCustomize using the RNA assay.

micro-computed tomography (μCT) and analysis

μCT images were generated using a X-Cube CT scanner (Molecubes) located at the Small Animal Imaging Facility at Penn Medicine. Live mice were anesthetized with isoflurane and images were acquired using a respiratory gating approach. The settings for image acquisition were as follows: reconstructed voxel size: 100 μm, X-Ray tube voltage: 50 kVp, X-Ray tube current: 440 μA, exposure time: 32 ms, type of filter used: 0.8 mm aluminum filter, degree of rotation stepping: 0.75 degrees (480 total steps in a 360 degree scan). Reconstructed μCT images were analyzed within Horos.

In vitro mouse alveolar organoids

Mouse alveolar organoid assays were performed as described previously with modification(11, 17). In brief, single cell suspensions were made from the lungs of Sftpc^CreERT2; Rosa26^tdTomato, Pdgfra^CreERT2; Rosa26^YFP, and Pdgfra^CreERT2; Rosa26^YFP/dnMAML mice. Single cell suspensions were prepared as described above. Mouse AT2 cells (tdTomato+) and mouse Pdgfra+ cells (YFP+) were sorted as described above. Live cell numbers were assessed by Trypan Blue solution (0.4%, Gibco) with a hemocytometer. Mouse AT2 cells were combined with mouse fibroblasts in a 1:20 ratio (5,000/100,000). Cells were resuspended in 50% Matrigel (Corning) and 50% base media. Base media consisted of DMEM-F12 (Gibco) mixed with selected components of Small Airway Growth Mediums (SAGM) supplements: insulin/transferrin, bovine pituitary extract, gentamycin, and retinoic acid as well as 0.1 mg/mL Cholera Toxin (Millipore Sigma, catalog # C9903), 25ng/mL EGF (Peprotech, catalog # AF-100–15), 1% antibiotics/antimycotics (Gibco, catalog # 15240062), and 5% FBS. Cell/media/Matrigel mixtures were placed in 24-well cell culture transwell inserts (Thermo Fisher). Cell/media/Matrigel mixtures were then allowed to solidify by placing in incubator for 15 minutes before base media as added under the transwell. ROCK inhibitor (10μm Y-27632, Caymen Chemicals) was added for the first three days of culture. An ALK5 inhibitor (10μm SB431542, Caymen Chemicals) was added for the first seven days of culture. Images were acquired and analyzed as described above using a EVOS FL Auto 2 imaging system using a 1.25x objective.

Culture of primary post-influenza basal cells

Primary basal cells were isolated from the distal lung of Krt5^CreERT2/+; Rosa26^dnMAML-GFP/+ mice at day 22 post-influenza (no tamoxifen was administered prior to injury), using a previously described lung digestion protocol(8). Cells were initially plated on 10% Matrigel coated plates and expanded in PneumaCult^™-Ex Plus Medium (Stemcell Technologies) + 1x PneumaCult^™-Ex Plus Supplement (Stemcell Technologies) + 1:1000 hydrocortisone stock solution (Stemcell Technologies) + 1μM A8301 (TGFb inhibitor, Millipore Sigma) + 10μM Y-27632 (ROCK inhibitor, Cayman Chemical) + 1:5000 Primocin^® (Invivogen), which specifically facilitates the outgrowth of basal cells. Cells were treated with DBZ (HY-13526, MedChemExpress) or 4- hydroxytamoxifen (3347905, Cayman) as described in Figure S12.

Post-influenza Basal Cells Monolayer Colony formation

Cells isolated as described above were seeded at 500 cells per well in growth medium in 6-well plates coated with 10% Matrigel in PBS. The medium was changed to the indicated growth medium supplemented with DMSO, 500 nM 4- hydroxytamoxifen (4OHT), 10 μM DBZ, or 20 μM DBZ after 24 hours. Cells were fixed 6 days later in 3.2% PFA and washed twice with PBS, followed by staining with crystal violet (0.5% crystal violet in 20% methanol) for approximately 10 min to visualize colony formation. The number of colonies formed were counted within FIJI (ImageJ).

Post-influenza Basal Cells EdU Proliferation analysis

Cells were seeded at a density of 5000 cells/mL on glass cover slips coated with 10% Matrigel in PBS in a 24-well plate. Cell culture media was changed to the indicated growth medium with the DMSO, 500 nM 4OHT, 2 μM DBZ, 5 μM DBZ, 10 μM DBZ 24 hours after plating. After 48 hours, the Click-iT^™ EdU Imaging Kit (Thermo-Scientific, C10086) was used according to manufacturer’s protocol to detect the EdU+ cell numbers and the DAPI cell number. Images were acquired using LAS X software with a 20x objective (Leica). The cell numbers were counted within FIJI (ImageJ).

Post-influenza Basal Cells Organoid Size

Cells were seeded inside 200 μL Matrigel on a 48 well plate at a density of 50 cells/μL. Pneumacult media with supplements was added after the Matrigel solidified at 37 °C. 24 hours after plating, fresh media containing DMSO, 500 nM 4OHT, or 10 μM DBZ was added. Brightfield images were acquired using LAS X software with a 10x objective. Organoid diameters were measured at day 7 using FIJI (ImageJ).

RNA isolation, cDNA synthesis, and qRT-PCR

For qRT-PCR, RNA was isolated using the Direct-zol RNA Miniprep Plus Kit (Zymo). cDNA synthesis was performed using iScript reverse transcriptase supermix (Biorad). Gene expression was calculated relative to Rpl19 (“L19”) within that sample and expressed as fold change over the average expression. qPCR was run on an Applied Biosystems QuantStudio 6 Real-Time PCR System (Thermo Fisher Scientific) with PowerUp SYBR Green Master Mix (Applied Biosystems). All primer sets are as listed in Table S2.

Statistical analysis

An unpaired two-tailed t test was used to compare two groups and a one-way ANOVA with Tukey’s adjustment for multiple comparisons was used when comparing multiple groups. Statistical significance was considered when calculated p-values were less than 0.05. All statistical analyses were completed using Graphpad Prism 9.

Supplementary Material

Figure S1

Supplemental Figure 1: Identification of subtypes of alveolar mesenchyme in the adult mouse lung

(a) UMAP representation of scRNA-seq data from adult mouse lung mesenchyme at homeostasis. Data downloaded from GSE149563(12). (b) DotPlot showing relative gene expression of unique marker genes of each mesenchymal subtype in panel (a). (c) GO analysis from the genes uniquely expressed in each cluster in (a). Terms are ranked by adjusted p-values. (d) representative IHC image showing Pdgfra/AF1 and Pdgfrb/AF2 cells exist equidistant to the nearest Erg+ endothelial cell. (e) quantification of panel (d). (f) % of Pdgfra+ and Pdgfrb+ cells which co-express Erg assessed by IHC analysis. Each dot in panel (e) and (f) represent data obtained from a single mouse. All error bars represent SEM.

Figure S2

Supplemental Figure 2: Specificity of Pdgfra-CreER and Pdgfrb-CreER in mouse lung mesenchymal cells.

(a) schematic depicting approach to assess labeling efficiency of the Pdgfra-CreER mouse line. (b) representative IHC and RNAscope images depicting overlap of the lineage reporter (tdTomato) with representative RNA markers of each mesenchymal cell type. (c) quantification of labeling efficiency. (d) schematic depicting approach to assess labeling efficiency of the Pdgfrb-CreER mouse line. (e) representative IHC and RNAscope images depicting overlap of the lineage reporter (tdTomato) with representative RNA markers of each mesenchymal cell type. (f) quantification of labeling efficiency. (g) experimental schematic of the approach to dissect overlap of Pdgfra+ and Pdgfrb+ cells in the mouse adult lung. (h) flow-cytometry showing overlap and separation of Pdgfra+ (eGFP+) and Pdgfrb+ (tdTomato+) cells in the adult mouse lung. (i) IHC confirmation of flow-cytometry using Pdgfra and Pdgfrb as unique markers of AF1s and AF2s respectively. (j) quantification of the % overlap of GFP+ vs GFP+ tdTomato+ from images in panel (i). n=3 biological replicates/group. Error bars represent SEM.

Figure S4

Supplemental Figure 4: Reactivity of pulmonary mesenchymal cell lineages after bleomycin-induced lung injury

(a) UMAP representation of scRNA-seq data from Pdgfrb+ lineage trace cells after bleomycin. Data shown represent integration of sham, day 14, and day 28 timepoints. (b) UMAP shown in (a) separated by timepoint. (c) Contribution of each timepoint to cell populations shown in panel (a). (d) FeaturePlots showing Pdgfrb+ cells do not proliferate (by Mki67 gene expression) after influenza. Distribution of Pdgfrb and Pdgfra gene expression also shown. (e) UMAP representation of scRNA-seq data from Pdgfra+ lineage trace cells after bleomycin. Data shown represent integration of sham, day 14, and day 28 timepoints. (f) DotPlot showing relative expression of unique marker genes expressed in cell populations shown in panel (e). (g) Contribution of each timepoint to cell populations shown in panel (e). (h) Slingshot trajectory analysis of Pdgfra-lineage trace cells showing a potential trajectory (Lineage3) from AF1 cells to AF2 cells after bleomycin. (i) representative IHC image showing localization of AF1-derived AF2s. (j) quantification of panel (h). Each dot represents data obtained from a single mouse.

Figure S3

Supplemental Figure 3: Reactivity of pulmonary mesenchymal cell lineages after respiratory influenza infection

(a) UMAP representation of scRNA-seq data from Pdgfrb+ lineage trace cells after influenza. Data shown represent integration of sham, day 14, and day 28 timepoints. (b) FeaturePlots showing Pdgfrb+ cells do not proliferate (by Mki67 gene expression) after influenza. Distribution of Pdgfrb and Pdgfra gene expression also shown. (c) Contribution of each timepoint to cell populations shown in panel (a). (d) UMAP separated by timepoint showing that no new cell cluster arises in UMAP space within the Pdgfrb+ cell lineage after influenza infection. (e) UMAP representation of scRNA-seq data from Pdgfra+ lineage trace cells after influenza. Data shown represent integration of sham, day 14, and day 28 timepoints. (f) DotPlot showing relative expression of unique marker genes expressed in cell populations shown in panel (e). (g) Contribution of each timepoint to cell populations shown in panel (e). (h) All Slingshot trajectories generated when AF1s were anchored as starting trajectory point. (i) Dynamic gene expression through pseudotime of genes enriched in each cell state within the trajectory. (j) Correlation plot showing Spearman correlation coefficients of Pdgfra-derived AF2 cells compared to endogenous AF1 and AF2 cells. (k) Spearman correlation plot and calculation of R correlation coefficient and p-value revealing the high degree of similarity between wildtype AF2 and Pdgfra-derived AF2 cells. (l) FeaturePlots generated using the RNA assay showing overlap of inflammatory cells (Saa3+ Lcn2+) with proliferating cells (Mki67+ Top2a+).

Figure S5

Supplemental Figure 5: Pdgfra+ cells are primary producers of myofibroblasts after bleomycin-induced lung injury.

(a) Experimental schematic showing approach. Pdgfra and Pdgfrb lineage trace mice were given tamoxifen daily for 3 days. 14 days later, mice were given bleomycin (3.5U/kg) and analyzed 14 days later. (b) IHC for tdTomato (lineage-trace) and SMA showing Pdgfra+ lineage trace cells become SMA positive after bleomycin-induced lung injury. (c) IHC for tdTomato (lineage-trace) and SMA showing Pdgfrb+ lineage trace cells do not become SMA positive after bleomycin-induced lung injury. (d) Experimental schematic showing approach. Pdgfrb lineage trace mice were given bleomycin and tamoxifen was delivered 4 days during injury (daily, days 10–13) and mice were collected 1 day later at day 14. (e) IHC for tdTomato (lineage-trace) and SMA showing labeling Pdgfrb+ cells after bleomycin-induced lung injury captures the actively expressing SMA myofibroblasts. (f) quantification of IHC images from panels (b), (c), and (e). All scale bars represent 50 μm. ****P<0.0001, evaluated by one-way ANOVA with Tukey’s adjustment for multiple comparisons.

Figure S7

Supplemental Figure 7: Examining transcriptional profile of epithelial/endothelial/immune cells by scRNA-seq after respiratory influenza.

(a) DotPlot showing unique gene expression markers of each cluster shown in Figure 4a. (b) Heatmap depicting the top 25 active transcription factors unique to each cell cluster. (c) DotPlot showing expression level of Notch receptors and downstream target genes in cell clusters shown in Fig. 4a. (d) UMAP representation of the scRNA-seq data showing endothelial, epithelial, and immune cells. (e) UMAP representation of the scRNA-seq data showing aggregation of cells by time point (sham, day 14, and day 28).

Figure S6

Supplemental Figure 6: Deletion of Ect2 in Pdgfra+ cells prior to influenza infection

(a) representative IHC images showing majority of bi-nucleated Pdgfra+ cells at 28 days after infection are localized within the alveolar region, as compared to peribronchial and adventitial region. Scale bar represents 50 μm. (b) quantification of bi-nucleated Pdgfra+ cell subtypes at 28 days after influenza infection. Each point represents data obtained from a single mouse. **P<0.01, ***P<0.001, evaluated by one-way ANOVA with Tukey’s adjustment for multiple comparisons. (c) H&E and injury-severity images of all images used for analysis in Figure 3i–k.

Figure S8

Supplemental Figure 8: Epithelial and mesenchymal cell crosstalk in the lung after injury

(a) Outgoing cell signaling from the Krt5+ dysplastic basal cells at 28 days after influenza infection showing that Krt5+ epithelium sends the Notch ligands Jag1, Jag2, and Dll1 to the Pdgfra-derived AF2 cells. (b) Outgoing cell signaling from the Pdgfra-derived AF2 cells at 28 days after influenza infection. (c) Notch signaling network between the epithelium and mesenchyme at 28 days after bleomycin-induced lung injury showing Krt5+ dysplastic epithelium and Pdgfra-derived AF2 cells form a Notch niche.

Figure S10

Supplemental Figure 10: Assessing immune cell population changes in Notch mutant lungs after respiratory influenza.

(a) experimental approach to examine changes in immune cell populations after influenza infection. (b) quantification of flow cytometry data showing no changes in immune cell populations after influenza infection. p-values quantified using unpaired t-test. (c,d) gating strategy to identify and quantify immune cell populations using flow cytometry.

Figure S9

Supplemental Figure 9: Blocking Notch signaling in Pdgfra+ cells prior to injury

(a) schematic depicting approach to assess efficiency of Cre mediated dnMAML expression in all 3 Pdgfra+ cell types (alveolar, adventitial, and peribronchial). (b) representative IHC images showing overlap of dnMAML expression with tdTomato (lineage-trace). Scale bar represents 50 μm. (c) quantification of IHC images shown in panel (b). Error bars represent SEM, n=3 mice/group. (d) H&E and injury-severity images of all images used for analysis in Figure 5c, d.

Figure S12

Supplemental Figure 12: Inhibiting intracellular Notch signaling in the basal cell lineage after flu does not inhibit injury-induced niche formation and basal cell function.

(a) Experimental schematic showing approach. Mice were given tamoxifen at days 5 and 8 after influenza and analyzed at day 12. (b) Survival curve showing inhibition of Notch signaling in Krt5+ cells does not improve survival after flu infection. (c) IHC showing Notch^Krt5-KD Krt5+ dysplastic epithelial cells establishing a dysplastic niche after influenza. Scale bar represents 20μm. (d) qRT-PCR of Notch downstream target genes from cultured in vitro basal cells after exposure to DMSO (vehicle control, n=5), 4-OHT (to genetically activate the dnMAML construct, n=4) or DBZ (small-molecule inhibitor of gamma secretase/Notch signaling, n=5). (e) Representative organoid images of basal cells isolated from Notch^Krt5-KD mice after flu showing administration of 4OHT (resulting in intracellular inactivation of Notch) has no effect on organoid size relative to DMSO control. Scale bar represents 100 μm. (f) Crystal violet assay showing administration of 4OHT (resulting in intracellular inactivation of Notch) has no effect on colony formation relative to DMSO control. (g) Quantification of organoid sizes shown in panel (d). (h) Quantification of colony formation shown in panel (e). (i) EdU proliferation analysis showing administration of 4OHT (resulting in intracellular inactivation of Notch) has no effect on basal cell proliferation relative to DMSO control. Data points in (f) represent individual organoids, (g) technical replicates in wells. Error bars represent standard deviation. *P<0.05, **P<0.01, ***P<0.001, ****P<0.0001, evaluated by one-way ANOVA with Tukey’s adjustment for multiple comparisons.

Figure S13

Supplemental Figure 13: Examining mechanism of how Notch inhibition improves regenerative capacity of the lung after influenza.

(a) UMAP of scRNA-seq data from wildtype and Notch^Pdgfra-KD Pdgfra-lineage cells at 14 days after influenza infection. (b) UMAP highlighting spatial location of cells from each condition (wildtype and Notch^Pdgfra-KD). (c) Incoming/outgoing signaling strength of all identified cell types from panel (a) in both conditions (wildtype and Notch^Pdgfra-KD) showing an increase in the incoming signaling from AT2 cells and an increase in outgoing signaling in AF1 cells in Notch^Pdgfra-KD mice compared to wildtype mice. (d) UMAP representation of scRNA-seq data from endothelial cells at 14 days after influenza infection. Data shown represent merged wildtype and Notch^Pdgfra-KD. (e) UMAP representation of scRNA-seq data from immune cells at 14 days after influenza infection. Data shown represent merged wildtype and Notch^Pdgfra-KD. (f) Contribution of genotype/sample to each endothelial cell population shown in panel (d). (g) GO analysis from the genes uniquely enriched in CAP1_a and CAP1_b cell populations. (h) Contribution of genotype/sample to each immune cell population shown in panel (e). (i) Slingshot trajectory analysis depicting potential trajectory of AF1a cells, the common AF1 subcluster between wildtype and Notch^Pdgfra-KD mice, transiting into AF1b (dominated by wildtype) or AF1c (dominated by Notch^Pdgfra-KD) subclusters. (j) Dynamic gene expression from each trajectory shown in panel (i) of AF1 markers (Pdgfra, Wnt2) and transient AF1 markers on their way to becoming AF2 cells (Pdgfrb, Acta2).

Figure S11

Supplemental Figure 11: Inactivating Notch signaling in Pdgfrb+ cells prior to injury does not prevent formation of the injury-induced niche.

(a) schematic depicting approach to examine the effect of inactivating Notch signaling in Pdgfrb+ cells prior to influenza infection. (b) All H&E and analyzed injury-severity images of all images used for analysis shown in panel (c). (c) quantification of injured regions in panel (b). (d) representative IHC image showing Krt5+ cell expansion in both wildtype and Notch mutated mice. (e) quantification of Krt5+ cell area at 14 days after influenza infection. Each dot represents data obtained from a single mouse. Statistical significance assessed by unpaired t-test.

Figure S14

Supplemental Figure 14: Differential ligand-receptor analysis between wildtype and Notch^Pdgfra-KD cell types

(a) Differential number of predicted ligand-receptor interactions between the sending cell (y-axis) and receiving cell (x-axis). Data shown are interaction values relative to wildtype. (b) Differential interaction strength between the sending cell (y-axis) and receiving cell (x-axis). Data shown are interaction strengths relative to wildtype. (c) System wide information flow for the pathways listed. (d) Specific differential ligand-receptor interactions with the AF1 cell as the sending cell. Wildtype = wildtype animals. Maml = Notch^Pdgfra-KD.

Figure S15

Supplemental Figure 15: Conservation of alveolar mesenchymal cell types and signaling relationships in the healthy and diseased human lung

(a) UMAP representation of scRNA-seq data of mesenchymal cells from healthy human lungs (n=13 donors). Data shown represent an integrated dataset of all donors. (b) DotPlot showing relative expression of unique marker genes identified in each cell population in panel (a). (c) IHC and RNA in situ hybridization showing localization of AF1 and AF2 in the human lung. (d) GO analysis from the genes uniquely expressed in each cell type listed in panel (a). GO terms ordered by adjusted p-value. (e) IHC and RNA in situ hybridization showing AF1s (marked by PDGFRA RNA expression) express PDGFRA protein and AF2s (marked by PDGFRB expression) express PDGFRB protein. Scale bars represent 50μm. (f) IHC image from Figure 7d split by channel to show distribution of PDGFRA/AF1 cells (white) and PDGFRB/AF2 cells (red) in the COVID-19 diseased human lung. (g) quantification from main Figure 7d split by donor. (h) IHC and RNA in situ hybridization showing PDGFRB+ mesenchymal cells exist directly adjacent to KRT5+ dysplastic epithelium in bleomycin-induced human lung injury. (i) distribution of PDGFRA/AF1 and PDGFRB/AF2 cells to the nearest KRT5+ cell in the bleomycin diseased human lung. All scale bars represent 50μm.

Data S1

Supplemental Data 1: Differential gene expression for Pdgfra-lineage mesenchymal cell types

Data S5

Supplemental Data 5: Differential gene expression for integrated scRNA-seq dataset from Pdgfra-lineage trace after influenza infection (sham, day 14, and day 28).

Data S2

Supplemental Data 2: Differential gene expression for Pdgfrb-lineage mesenchymal cell types

Data S6

Supplemental Data 6: Predicted transcription factor activity within Pdgfra-lineage mesenchymal cell types at day 28 after influenza infection.

Data S7

Supplemental Data 7: Differential gene expression for human mesenchymal cell types.

Data S3

Supplemental Data 3: Cleared 200um thick lung section showing separation of Pdgfrb (red) and Pdgfra (green) in the alveolar space.

Each frame is 1 μm is in the z-direction. Video is played at 10 frames/second (i.e. 10 μm/second).

Data S4

Supplemental Data 4: Cleared 200um thick lung section showing separation of Pdgfrb (red) and Pdgfra (green) in the alveolar space.

Each frame is 1 μm is in the z-direction. Video is played at 10 frames/second (i.e. 10 μm/second).

Table S2

Supplemental Table 2: Primer sequences.

Primer sequences used to generate the qRT-PCR data in this study.

Table S1

Supplemental Table 1: Patient characteristics.

A list of all of the patients who were included in the analyses presented herein (other than previously published data when mentioned). Age, gender, self-identified race, and cause of death or disease at time of transplantation are provided. Where available, smoking history, FEV1, and/or arterial partial pressure of O2 to fraction of inspired O2 (P/F ratio) is also reported. The use of tissue is indicated by the type of experiment.

Data and code availability

All newly generated mouse scRNA-seq data have been deposited into GEO (GSE249931). Any additional information required to reanalyze the data reported in this paper is available from the lead author upon request.