key: cord-0273389-x2c1vy74
authors: Brownfield, Douglas G.; de Arce, Alex Diaz; Ghelfi, Elisa; Gillich, Astrid; Desai, Tushar J.; Krasnow, Mark A.
title: Alveolar cell fate selection and lifelong maintenance of AT2 cells by FGF signaling
date: 2022-01-20
journal: bioRxiv
DOI: 10.1101/2022.01.17.476560
sha: 706953d8aa306e5dfcabda6b427cfc2dabb5a87d
doc_id: 273389
cord_uid: x2c1vy74

The lung’s gas exchange surface comprises thin alveolar type 1 (AT1) cells and cuboidal surfactant-secreting AT2 cells that are corrupted in some of the most common and deadly diseases including adenocarcinoma, emphysema, and SARS/Covid-19. These cells arise from an embryonic progenitor whose development into an AT1 or AT2 cell is thought to be dictated by differential mechanical forces. Here we show the critical determinant is FGF signaling. FGF Receptor 2 (Fgfr2) is expressed in mouse progenitors then restricts to nascent AT2 cells and remains on throughout life. Its ligands are expressed in surrounding mesenchyme and can, in the absence of differential mechanical cues, induce purified, uncommitted E16.5 progenitors to form alveolus-like structures with intermingled AT2 and AT1 cells. FGF signaling directly and cell autonomously specifies AT2 fate; progenitors lacking Fgfr2 in vitro and in vivo exclusively acquire AT1 fate. Fgfr2 loss in AT2 cells perinatally results in reprogramming to AT1 fate, whereas loss or inhibition later in life immediately triggers AT2 apoptosis followed by a compensatory regenerative response. We propose Fgfr2 signaling directly selects AT2 fate during development, induces a cell non-autonomous secondary signal for AT1 fate, and stays on throughout life to continuously maintain healthy AT2 cells. One Sentence Summary FGF signaling induces and distinguishes the two cell types of the lung’s gas exchange surface, and the pathway remains on throughout life to maintain one that can be transformed into lung cancer or targeted in the deadly form of SARS/Covid-19.

Gas exchange occurs in alveoli, tiny terminal air sacs of the lung lined by two intermingled epithelial cell types: exquisitely thin alveolar type 1 (AT1) cells that provide the gas-exchange surface, and cuboidal AT2 cells that secrete surfactant to prevent alveolar collapse and acute respiratory failure 1 . Alveoli are also the site of some of the most significant but poorly understood and difficult to treat human diseases, including bronchopulmonary dysplasia in infants and COPD/emphysema 2 , pulmonary fibrosis 3 , lung adenocarcinoma 4, 5 , and viral infections including Severe Acute Respiratory Syndrome (SARS) 6, 7 and the current pandemic of COVID-19 8 in adults. Understanding how these key alveolar cell types arise during development and are maintained throughout life is critical for understanding and treating these diseases, and for guiding developmental, regenerative and tissue engineering approaches to create healthy alveoli and the surfactant that lines them.

Despite the extreme differences in structure and function of AT1 and AT2 cells, marker expression, lineage tracing, and clonal analysis in mice indicate that each arises directly from a common progenitor during embryonic development 5, 9 . This model is supported by single cell RNA sequencing studies of developmental intermediates that reconstructed the full, bifurcating gene expression program from bipotent progenitors at embryonic day 16.5 (e16.5) to either AT1 or AT2 cells over the next several days of fetal development 10, 11, 12 , although a recent genetic labeling study claims an earlier fate commitment 13 . Multiple developmental signaling pathways 14, 15, 16, 17, 18 and transcription factors 19 can influence alveolar structure, maturation or the balance of the two cell types 20, 21, 22 , but the key driver of alveolar fate selection and differentiation is thought to be mechanical forces 23 . The build up and increased movement of luminal fluid in late gestation is proposed to stretch and flatten progenitors into AT1 cells 18, 24, 25 , as stretch does to cultured AT2 cells to cause AT1 transdifferentiation in vitro 26 . Live imaging indicates that alveolar progenitors protected from the luminal mechanical forces during budding default to the AT2 cell program 18 .

Here we demonstrate that alveolar cell fate specification is in fact dictated by a classic growth factor signal, the same FGF signaling pathway that controls the earlier steps of airway budding and branching 22, 27, 28, 29, 30 and also initiates subsequent budding 18 . After branching, the receptor Fgfr2 remains on in alveolar progenitors and dynamically restricts to the AT2 lineage shortly before cellular differentiation, while its ligands Fgf7 and Fgf10 continue to be expressed by surrounding stromal cells. We show by the effects of the ligands on isolated e16.5 progenitors in cultures lacking budding and differential mechanical cues, and by genetic mosaic and pharmacological studies in culture and in vivo, that Fgfr2 signaling directly and cell autonomously specifies AT2 fate while inducing a secondary, non-autonomous signal that promotes AT1 fate of neighboring progenitors. This developmental FGF pathway remains on throughout life where it serves continuously and ubiquitously to maintain healthy AT2 cells 31, 32 , which if deprived of the FGF signal early in life reprogram to AT1 fate and later in life immediately undergo apoptosis.

To identify signaling pathways that might control alveolar cell fate selection, we searched the single cell transcriptional program of mouse alveolar development 11 for receptor genes enriched in either the AT1 or AT2 lineages. Of the five receptor genes showing at least two-foldenrichment in the AT2 lineage ( Fig. 1A; Fgfr2 , Fzd8, Cd36, Cd74, and Ngfrap1), only Fgfr2 is known to be required for embryonic lung development 22, 30 . The other four either have no consequence (Fzd8), only adult lung phenotypes (Cd36, Cd74), or have not yet been examined (Ngfrap1). Fgfr2 is expressed early and throughout the developing lung epithelium and its deletion abrogates airway branching 22, 29, 30 , so its later function has been more challenging to investigate 14, 31, 32, 33 . The single-cell RNA sequencing (scRNAseq) profiles demonstrated that the iiib isoform of Fgfr2 is expressed in bipotent progenitors and maintained in the AT2 lineage but downregulated in the AT1 lineage (Fig. 1A, Supplementary Fig. 1C ). Fgfr2 immunostaining during mouse alveolar development confirmed that Fgfr2 is expressed in bipotent progenitors ( Fig. 1C) , downregulated in nascent AT1 cells, and maintained in nascent AT2 cells as they bud into the surrounding mesenchyme and become histologically and molecularly distinct from AT1 cells ( Fig. 1C; Supplementary Fig. 1A) . Indeed, Fgfr2 is among the earliest known markers of AT2 fate selection.

To identify the relevant Fgfr2 ligands, we performed scRNAseq on the adjacent mesenchymal cells during alveolar differentiation. Two Fgfr2 (isoform iiib) ligands, Fgf7 and 

To explore the function of Fgfr2 signaling in alveolar development, we first examined the effect of Fgf7 and Fgf10 on purified epithelial progenitors isolated from the tips of e16.5 lungs, before AT1 and AT2 cells are detected. When the progenitors were cultured for up to 8 days in Matrigel in the absence of exogenous FGFs, they failed to develop into either AT1 or AT2 cells, remaining cuboidal and organizing into multicellular clusters around a lumen ( Fig. 3H ). Thus, Fgfr2 signaling drives differentiation of alveolar progenitors and formation and growth of alveolus-like structures ("alveolospheres") with intermingled AT1 and AT2 cells, and it does so independent of extrinsic mechanical forces and without the early budding and associated structural features proposed to differentially protect progenitors from such forces. However, it is possible that extrinsic mechanical forces and budding in vivo might regulate some aspect of Fgf signaling or the developmental process that is dispensable in the culture system. along with a fluorescent reporter (GFP) to mark infected cells (Fig. 3A) . We found that nearly every progenitor cell (95%) infected with virus expressing wild type Fgfr2 differentiated into an AT2 cell, and conversely infected cells expressing Fgfr2 DN almost always (88%) acquired AT1 fate (Fig. 3B,C) ; the rare cells that did not acquire the predominant fate remained undifferentiated (Fig. 3C, Supplementary Fig. 4A ). Experiments with a control virus (AAV-nGFP) that expresses GFP but no Fgfr showed that randomly-infected progenitors have a roughly equal chance of acquiring AT1 or AT2 fate during culturing ( Supplementary Fig. 4B ).

These results, together with the above result showing e16.5 distal progenitors cultured for over a week in the absence of FGF fail to differentiate into AT1 or AT2 cells ( Fig. 2 ; Supplementary   Fig. 3C ), indicate that the fate of alveolar progenitors is not committed at e16.5, countering claims of an earlier fate commitment 13 but consistent with prior lineage tracing 9 , clonal analysis 5 , and scRNAseq results 10, 11, 12 . We conclude that distal epithelial progenitors at e16.5 are indeed bipotent, having the capacity to differentiate into either AT1 or AT2 cells, and that AT2 fate is selected directly and cell autonomously by Fgfr2 signaling. The results also imply that the observed induction of AT1 fate by FGF addition to the cultures must be indirect (cell nonautonomous), presumably via a secondary signal produced by maturing AT2 cells and received by neighboring progenitors (see Discussion). AT1 differentiation cannot simply be the default fate because it too required addition of FGF ligands to the culture.

To determine if Fgfr2 signaling plays a similar role in alveolar cell fate selection in vivo, we used an Nkx2.1-Cre transgene with low Cre activity to mosaically delete Fgfr2 in the developing lung epithelium, combined with a fluorescent Cre reporter allele (Rosa26 mTmG ) to mark the cells in which Cre was active (Fig. 4, Supplementary Fig. 4C ). In control mice carrying a wild type Fgfr2 allele in trans to the conditional Fgfr2 allele (Tg Nkx2.1-Cre ;Fgfr2 fl/+ ; Rosa26 mTmG/+ ), one third of distal cells with Cre reporter activity (36%, n> 970 GFP+ cells scored in 3 mice) acquired AT1 fate by the end of fetal life (PN0) and the rest acquired AT2 fate (64%) (Fig. 4A,B) . By contrast, in animals with an Fgfr2 deletion in trans to the conditional allele (Tg Nkx2.1Cre ;Fgfr2 fl/del ;Rosa26 mTmG/+ ), the targeted distal cells almost exclusively (91%, n>460 GFP+ cells scored in 3 mice) acquired AT1 fate (Fig. 4A,B) ; the few cells that did not acquire AT1 fate had not yet lost Fgfr2 expression (Fig. 4C ). We conclude that Fgfr2 is the critical determinant of AT2 fate and is required cell autonomously for AT2 fate selection in vivo as well as in vitro, and that alveolar progenitors lacking Fgfr2 signaling acquire the alternative, AT1 fate.

Curiously, this developmental signaling pathway remains on later in life, for months or even years after alveoli have acquired their canonical structure and function. Fgfr2 continues to be selectively expressed in AT2 cells (Fig. 5A ,B,D; Supplementary Fig. 5 ), Fgf7 and Fgf10 continue to be expressed in Wnt2-expressing alveolar fibroblasts and at lower levels in myofibroblasts ( Fig. 5E-I, Supplementary Fig. 2B,C) , and the Fgfr2 signaling pathway remains active in mature AT2 cells as detected by phosphorylated MAP kinase (Fig. 5C,D) and expression of Fgfr2-induced genes such as Spry2 (Fig. 5A) . To investigate the role of this selective and persistent Fgfr2 signaling in AT2 cells, we first used a Cre knock-in allele at the endogenous Lysozyme2 (Lyz2) locus, which turns on in some AT2 cells early in postnatal life and is progressively activated later in other AT2 cells 5 (see below), to conditionally delete Fgfr2 in AT2 cells at different ages. There was a dramatic effect of Fgfr2 deletion from AT2 cells during juvenile life, and an equally significant but entirely different effect in adults.

Lyz2-Cre activity is first detectable in rare AT2 cells at PN5, as shown by the Cre reporter (Fig. 5J) , and then gradually becomes active in additional AT2 cells (Fig. 5P, 6A Fig. 3D ,E). We conclude that Fgfr2 signaling is necessary to maintain AT2 fate in newly formed AT2 cells during juvenile life, and that loss of Fgfr2 signaling during this period results in rapid and robust reprogramming to AT1 fate.

After the juvenile period (PN17) in the above experiment, there was no further increase in AT2lineage-labeled cells with AT1 fate (Fig. 6A , right panel; Supplementary Fig. 6 ), indicating that AT2 cells that lose Fgfr2 in adult life do not reprogram to AT1 fate. There was also no further accumulation of lineage-labeled AT2 cells during this period (Fig. 6A , right panel), suggesting that Fgfr2 deletion in adult AT2 cells results in their loss. Indeed, we occasionally detected expression of cleaved Caspase-3 in such cells, indicating they had initiated apoptosis (Fig. 6B) .

Recently, others have noted a variable reduction in AT2 cell markers or abundance and impaired recovery from severe lung injury when one or more Fgfr genes were conditionally deleted 31, 32, 33 , attributing the deficits to a reduction in stem cells 31 or their proliferation 32, 33 . However, another recent study reported Fgfr2 in AT2 cells is entirely dispensable during homeostasis 36 . To more precisely determine the cellular and molecular consequences of acute abrogation of FGF signaling in adult AT2 cells, we first tried a tamoxifen-inducible Cre recombinase (Sftpc CreER/+ ; Fgfr2 del/fl ; Rosa26 mTmG/+ ) but found that removal of Fgfr2 occurred but was inefficient under these conditions (Fig. 4C ) (see below and Methods). We therefore developed three additional approaches, which revealed that Fgfr2 signaling is ubiquitously and continuously required for AT2 cell maintenance throughout adult life, and even brief deprivation of signaling immediately initiates apoptosis.

Two important limitations of the conditional Fgfr2 deletion experiment noted above were that the gene was inefficiently deleted, and even when the gene was deleted the protein persisted (perdured), limitations that could have contributed to the modest or entire lack of effect observed in other studies 32,36 . To overcome the former limitation, we designed a combinatorial genetic approach using the MASTR transgene 37 (Fig. 6C ), which following Flp-mediated recombination of the transgene provides constitutive and high level Cre expression to ensure recombination of all Cre-dependent alleles in the cell, such as Fgfr2 fl and Rosa26 mTmG . We used the MASTR allele in conjunction with an AAV virus we engineered to express Flp recombinase specifically in AT2-cells using an Sftpc promoter element (Fig. 6C,D) ; when the virus was instilled intratracheally into adult control mice (Fgfr2 +/fl ; Rosa26 mTmG/MASTR ), specific expression of Credependent reporter genes was observed in AT2 cells (>98% of GFP + cells, n = 3 lungs, >168 GFP + cells counted per lung (data not shown) at 2 weeks (Fig. 6E ), similar to the AT2 specificity observed in prior studies with this approach 38, 39, 40 . When the virus was instead instilled into Fgfr2 fl/fl ; Rosa26 mTmG/MASTR mice, 60% fewer GFP + AT2 cells were observed (Fig. 6E ,F); of the GFP + AT2 cells that remained, nearly all (96%) were found to have retained Fgfr2 protein due to perdurance (Fig. 6H) , and the rare GFP + Fgfr2 -AT2 cells detected (4 of 96 scored GFP + AT2 cells from 3 mice) had been extruded into the alveolar space and had a rounded morphology ( Fig. 6G ). This result indicates that adult AT2 cells broadly and uniformly undergo apoptosis following removal of Fgfr2. Apoptosis may even initiate with only partial reduction in Fgfr2 levels because one-third (36%) of the remaining GFP + AT2 cells in this experiment, nearly all of which expressed detectable levels of Fgfr2 (Fig. 6H) , were positive for cleaved-Caspase-3 ( Fig.   6I ,J).

To overcome the problem of Fgfr2 perdurance in AT2 cells and corroborate these results, we employed two other approaches. In a classical pharmacologic approach (Fig. 6K) , we instilled Fgfr inhibitor FIIN-1 into the lungs of adult (PN60) Sftpc CreER/+ ; Rosa26 mTmG/+ mice treated with tamoxifen two weeks earlier to label mature AT2 cells. A fluorescent marker (WGA-405) was included in the instillation to identify regions of FIIN-1 exposure (Fig. 6L ). As early as two hours after the instillation, inhibition of Fgfr signaling triggered widespread apoptosis of AT2 cells in the FIIN-1 exposed (WGA-405 labeled) regions (Fig. 6M,N) , with no conversion to AT1 identity ( Supplementary Fig. 7) . To ensure this effect was attributable to

Fgfr2 inhibition, we also tested a soluble recombinant Fgfr2b-Fc protein that binds its cognate ligands (see Fig. 1 ), preventing their engagement and activation of the receptor (Fig. 6O ). With this inhibitor too, there was substantial induction of AT2 cell apoptosis in the exposed regions shortly (6 hours) after instillation ( 

Despite the profound effects of Fgfr2 loss or inhibition on AT2 cells in juvenile and adult life, curiously alveolar structure and the total number and proportion of AT2 cells were not grossly altered (Fig. 7A, B) , obscuring its ubiquitous and constant requirement for AT2 cell maintenance throughout life. This is because apoptosis (or reprogramming) of targeted AT2 cells is almost completely compensated by proliferation of untargeted cells. This was evident in the Lyz2-Cre mice (Lyz2 cre/+ Fgfr2 fl/fl ;Rosa26 mTmG/+ ), where the population of untargeted AT2 cells (lacking the Cre-dependent lineage label) in PN60 mice was increased by 52% relative to that in the Fgfr2+ control (Lyz2 Cre/+ ;Fgfr2 fl/+ ;Rosa26 mTmG/+ ), maintaining the normal density of AT2 cells (~5 per 100 um 3 ) and alveolar morphology despite the substantial loss of lineage-labeled AT2 cells (reduced from ~36% of AT2 cells in control Lyz2 Cre/+ ;Fgfr2 fl/+ ;Rosa26 mTmG/+ mice to 2.8% in Supplementary Fig. 9 ). The proliferating AT2 cells carrying out the restorative response (with Fgfr2 signaling intact because they were not from a FIIN-1-exposed area or because FIIN-1 had since cleared) are executing the canonical stem cell function of AT2 cells 5, 44, 45 , replacing lost AT2 cells and any AT1 cells lost along with them in the FIIN-1-exposed areas (Fig. 7H ).

The results identify FGF signaling, not differential mechanical forces, as the critical factor that induces and controls alveolar epithelial fate selection. Fgfr2 is expressed in bipotent alveolar progenitors but rapidly restricts to nascent AT2 cells and remains on exclusively in the AT2 lineage, while its ligands Fgf7 and Fgf10 are expressed in surrounding mesenchyme. Addition of either ligand to cultured progenitors induced formation of alveolus-like structures with intermingled AT2 and AT1 cells, in the absence of any extrinsic forces or cell budding that might protect some progenitors from such forces. In mosaic cultures, randomly selected progenitors with constitutive Fgfr2 expression exclusively acquired AT2 fate whereas those with the Fgfr2 pathway inhibited exclusively acquired AT1 fate, and a similar result obtained for 7J ). Why would a developmental signaling pathway remain on for months or even years after its developmental role has been completed, simply to keep the selected cell alive? The considerable energy expenditure involved in lifelong signaling suggests some substantial benefit of active AT2 maintenance. Perhaps it allows rapid conversion to a progenitor state and other fates following alveolar injury to quickly restore gas exchange function, akin to the rapid transdifferentiation of club into ciliated airway cells when Notch signaling is blocked 47 . Or perhaps it prevents initiation of lung adenocarcinoma, the leading cancer killer, when an AT2 cell divides and a daughter loses contact with the underlying stromal cells that express the Fgf7 and Fgf10 survival signals (Fig. 7K ). Whatever the reason, the pathway plays an important role in human lung health because genetic studies identify Fgf7 as a susceptibility locus for the common and devastating disease COPD (chronic obstructive lung disease) 56 .

Our results show that loss or inhibition of Fgfr2 signaling has dire consequences for the alveolus and gas exchange, though they may not become apparent immediately because of the robust regenerative response activated by AT2 cell death. However, widespread targeting or loss of AT2 cells, as may occur in SARS, MERS, and COVID-19 coronavirus infections 6, 7, 8 , causes acute and severe alveolar injury and can be rapidly fatal. Pharmacologic modulation of Fgfr2 signaling could be used to support alveolar health during acute injuries like these, helping sustain AT2 cells as a virus or toxin destroys them 6, 7, 8 , and perhaps similarly in chronic diseases such as COPD/emphysema and pulmonary fibrosis 57 . Modulators could also be used to create alveolar cells for in vitro studies or cell therapies for these diseases.

The authors thank members of the Krasnow laboratory for helpful discussions and critical reading of the manuscript. This work was supported by R00HL127267 

The authors have no competing interests to declare.

Timed-pregnant C57BL/6J females (abbreviated B6; Jackson Laboratories) were used for all embryonic time points, with gestational age verified by crown-rump length. For studies of adult wild type lungs, B6 males and females were used. Mosaic labeling and deletion studies were conducted by Cre recombinase expression using gene targeted alleles BAC-Nkx2.1-Cre 61 , Lyz2 

Immunohistochemistry was performed as previously described 66 using primary antibodies against the following epitopes (used at 1:500 dilution unless otherwise noted): pro-SftpC (rabbit, Chemicon AB3786), RAGE (rat, R&D MAB1179), E-cadherin (rat, Life Technologies ECCD-2), Podoplanin (hamster, DSHB 8.1.1), Mucin 1 (hamster, Thermo Scientific HM1630 and rabbit, Novus NB120-15481), Ki67 (rat, DAKO M7249), Fgfr2 (rabbit, SCBT SC-122), cleaved Caspase 3 (rabbit, Novus NB100-56708), GFP (chicken, Abcam ab13970), Fgfr2iiib-Fc (human, R&D Systems) and Phospho-p44/42 MAPK (rabbit, CST D13.14.4E). Primary antibodies were subsequently detected using Alexa Fluor-conjugated secondary antibodies (Life Technologies) unless noted otherwise and incubated in Vectashield with DAPI (5 µg/ml, Vector labs). EdU was detected using the standard Click-iT reaction (ThermoFisher). Images were acquired using a laser-scanning confocal microscope (Zeiss LSM 780) and subsequently processed using ImageJ.

Two methods of multiplexed single molecule fluorescence in situ hybridization (smFISH) of mRNAs was performed. To simultaneously detect Sftpc, Fgf7, and Fgf10 RNAs, RNAscope multiplex assay V1 kit (ACD) was used as described 11 . To visualize Sftpc and Fgfr2 RNAs, proximity ligation in situ hybridization (PLISH) was used 67 . Briefly, 20µm thick sections were cut from OCT-embedded, cryopreserved tissue and hybridized with multiple anti-sense probe pairs that hybridize to the target transcript. Probe pairs that targeted each gene share a common barcode that is unique and complementary to circle and bridge constructs. The circle and bridge constructs undergo proximity ligation to form a closed circle that undergoes rolling circle amplification. Detection oligonucleotides conjugated to fluorophores anneal to the rolling circle amplification product, generating discrete puncta for each transcript. The following sets of primer pairs were used to detect transcripts of the indicated genes:

Expression was detected by confocal fluorescence microscopy as ~0.5 µm fluorescent puncta.

Adult AT2 cells and embryonic (e16.5) bipotent alveolar epithelial progenitors were purified as previously described 11 . Adult 2-month-old mice were euthanized by administration of CO2. For e16.5, embryos were removed from the mother and their lungs were isolated en bloc without perfusion and pooled by litter (five to seven embryos) for further processing. Lungs were microdissected to remove the proximal lung tissue, leaving only the distal (alveolar region) tissue. Distal e16.5 lung cells were dissociated in dispase (BD Biosciences) and triturated with glass Pasteur pipettes until a single-cell suspension was attained. For adult lungs, the vasculature was perfused through the right ventricle with 37°C media (DMEM/F12, Life Technologies). The trachea was punctured and lungs inflated with digestion buffer (DMEM/F12 containing elastase (1U/ml, Worthington) and dextran (10%, Sigma)) for 20 minutes at 37°C. Digested lungs were minced with a razor blade into 1mm 3 fragments, suspended in 5ml of digestion buffer containing DNase I (0.33U/ml; Roche), incubated with frequent agitation at 37°C for 45 minutes, and triturated briefly with a 5-ml pipette.

To deplete red blood cells, an equal volume of DMEM/F12 supplemented with 10% FBS and penicillin-streptomycin (1U/ml; Thermo Scientific) was added to the lung single-cell suspensions before filtering through 100µm mesh (Fisher) and centrifuging at 400g for 10 min.

Pelleted cells were resuspended in red blood cell lysis buffer (BD Biosciences), incubated for 2 minutes, passed through a 40µm mesh filter (Fisher), centrifuged at 400g for 10 minutes and then resuspended in MACS buffer (2mM EDTA, 0.5% BSA in PBS, filtered and degassed) for purification.

From the resultant single cell suspensions of distal e16.5 lung, bipotent progenitors were isolated by MACS using MS columns (Miltenyi Biotec) according to the vendor protocol. Before column loading, suspensions were passed through a 35µm cell strainer (BD Biosciences). Other cell types were first depleted with antibodies against CD31, CD45 and Pdgfrα (Miltenyi Biotec), then bipotent progenitors were positively selected using a biotinylated EpCAM antibody (clone G8.8, eBiosciences) and streptavidin-conjugated magnetic beads (Miltenyi Biotec). This generates a >95% pure preparation of e16.5 bipotent alveolar progenitors (Sftpc + Rage + cells).

AT2 cells were isolated from the single cell suspensions of adult lungs in the same way, generating a >90% pure preparation of AT2 cells. To enrich for adult fibroblasts, MACS depletion was conducted as above using antibodies against CD31, CD45, and EpCAM.

For culturing bipotent progenitors, cell density was calculated using a hemocytometer. A density of 100,000 cells per well was used for culture in 8-well #1 coverglass chambers (Labtek) precoated with growth factor reduced Matrigel (80µl, BD Biosciences) for 30 minutes at 37°C.

Cells were supplemented with Fgf7 (50ng/ml, R&D Systems), Fgf10 (100ng/ml, R&D Systems), FIIN-1 (20nM, Tocris), Fgfr2iiib-Fc (1µg/ml, R&D Systems), CK666 (40µM, R&D), and HSPG (100ng/ml, Sigma) to the indicated concentrations in DMEM/F12. Cells were maintained at 37°C in 400µl of DMEM/F12 with media changes every other day in a 5% CO2/air incubator typically for four days, except where indicated otherwise.

Time lapse microscopy of cultured alveolospheres was conducted using an inverted LSM 880 equipped with environmental control to maintain 37°C and 5% CO2. Isolated bipotent progenitors were plated and cultured for ~12 hours before confocal microscopy. Brightfield zstacks (20x multi-immersion objective, NA 1.33) were acquired every 15 minutes for at least 3 days. Lumenal area, measured at the center z-position of each organoid, was used as an indicator of stretch, as commonly implemented in forskolin-induced swelling experiments 68 . Budding status of a cell was determined using the published criteria 18 of basal extrusion with substantial reduction of lumenal cell surface.

Processing, scRNAseq, and analysis of e18.5 and adult mouse lung cells was conducted as previously described with minor alterations to enrich for the respective populations 11 . Briefly, single-cell suspensions sorted by MACS as CD45 -CD31 -EpCAM -(for mesenchymal cells) or Subsequent analysis including hierarchical clustering was performed as described 11 . To determine isoform expression of Fgfr2, STAR was used to align and count reads specific to either exon IIIb or IIIc To determine receptor genes restrictively expressed along either AT1 or AT2 lineage, receptor genes were first screened for ones detectably expressed in at least 6 cells, then Welch's t-test was used to identify receptor genes with at least two-fold higher expression in each respective lineage (p ≤ 0.01).

For the analysis of cell type specific gene expression in published scRNAseq datasets, the processed mRNA counts for each cell were used from developing (SE119228) 10 and adult mouse lung (GSE109774) 60 datasets. Expression matrices were analyzed via R using the Seurat package. Clusters of cells with similar expression profiles were identified via shared nearest neighbor analysis using the Louvain algorithm, clusters were visualized in UMAP plots, and expression of specific genes was represented in heatmaps, violin plots, and dot plots generated in Seurat.

To mosaically modulate Fgfr2 activity in culture, a lentiviral approach was developed using the Lenti-X Tet-On 3G Inducible Expression System (Clontech). The ORF encoding Fgfr2iiib (Origene; MC221076), the IRES sequence from pLVX-IRES-Puro (Clontech; 632186), and sfGFP (superfolder GFP) coding sequence from pBAD-sfGFP (Addgene; 54519) were PCRamplified with adaptor sequences (IDT) and assembled (NEB; E5520) into the lentiviral backbone pTetOne (Takara; 631018) linearized by EcoRI/BamHI digestion (NEB) to generate pTetOne-Fgfr2-IRES-sfGFP (Fig. 3A) . To generate a dominant negative form of Fgfr2iiib, a truncated version of Fgfr2iiib lacking the tyrosine kinase domain 69 was also PCR-amplified with adapter sequences (as above) to assemble pTetOne-Fgfr2 DN -IRES-sfGFP. Lentiviruses ("Lenti-Fgfr2", "Lenti-Fgfr2 DN ") were produced from these plasmids following the manufacturer's instructions for transfection, lentiviral concentration, and titer estimation (Lenti-X Expression System, Clontech). An AAV virus that constitutively expresses eGFP from CMV promoter (Addgene, 105545-AAV9) was used as a cell labeling control (Supplementary Fig. 4B To mosaically but efficiently delete Fgfr2 in adult AT2 cells, we used a recombinant AAV described below to infect and express Flp recombinase mosaically in AT2 cells, and the Flp-dependent Rosa26-MASTR allele to constituitively express Cre recombinase 37 to delete the conditional Fgfr2 fl/fl allele. The pAAV-EF1α-Flpo plasmid 70 was modified using the assembly scheme described above (NEB; E5520) to replace the EF1α promoter (removed by MluI/BamHI restriction digestion) with a previously defined 38 320 bp minimal promoter element of Sftpc that was chemically synthesized with added adapter sequences (IDT) (Fig. 6C) . Recombinant AAV-Sftpc-Flpo virus was prepared by the HHMI Janelia Farm Viral Tools facility (2x10 14 GC/ml) and endotracheally instilled (1µl of viral solution diluted in 50µl PBS) into the lungs of mice harboring Rosa26 MASTR/mTmG and either heterozygous or homozygous for the floxed Fgfr2 allele.

Efficiency of Fgfr2 deletion was assessed by immunostaining for Fgfr2. For adult time points, the density of AT2 cells was calculated and expressed as the number of AT2 cells per 100µm x 100µm x 100µm region of alveolar tissue. To determine AT2 proliferation, EdU (1mg/ml, Cedar Lane) was administered in the drinking water for the indicated period.

To rapidly and persistently inhibit Fgfr2 signaling, the irreversible Fgfr2 inhibitor FIIN-1 (14.75 mg/kg, Tocris) 71, 72, 73 or recombinant Fgfr2iiib-Fc (100µg/30g mouse, R&D Systems) was used.

Mice at least 2 months (and up to over 1 year) of age carrying the Sftpc Cre-ERT2-rtTA ;

Rosa26 mTmG/mTmG alleles were administered tamoxifen (3mg/30g mouse, Sigma) via intraperitoneal injection to label >95% of AT2 cells. After 1 week, the mice were anesthetized with isoflurane and administered by endotracheal instillation either vehicle alone (PBS with biotinylated WGA), or vehicle with the inhibitor. To test the effect of Akt signaling, the small molecule Akt activator SC79 (Tocris) was administered (0.04 mg drug per g mouse) via intraperitoneal injection 30 minutes prior to Fgfr2 inhibition. Lungs were isolated at the indicated time points and areas of high exposure to inhibitor were identified by WGA immunostaining.

Data analysis and statistical tests were performed with R or GraphPad Prism software. Replicate experiments were all biological replicates with different animals, and quantitative values are presented as mean ±S.D. unless indicated otherwise. Student's t-tests were two-sided. No statistical method was used to predetermine sample size, and data distribution was tested for normality prior to statistical analysis and plotting. Both male and female animals were used in experiments, and subjects were age-and gender-matched in biological replicates and in comparisons of different groups. Post-hoc power analysis was done (Supplemental Table 1 to GFP label AT2 cells, then 1 week later instilled with Fgfr inhibitor FIIN-1 as in Figure 6C -F and allowed to recover for one week (upper panel) or two weeks (lower panel). Note disorganization of alveolar structure at one week that resolves by two weeks. Scale bar, 50 µm.

(B) FIIN-1-exposed alveolar region (upper box) and control unexposed region (lower box)

following FIIN-1 instillation as above and stained two weeks later for AT2 lineage marker mGFP, mTomato (marking other cell types), and collagen a1 (Cola1). Note that the FIIN-1 exposed region has recovered and appears normal histologically with normal Col1a1 staining.

Scale bars, 50µm. Table 1 

Structure of respiratory tissue

Chronic obstructive pulmonary disease

Idiopathic pulmonary fibrosis

Cancer statistics

Alveolar progenitor and stem cells in lung development, renewal and cancer

Lung pathology of severe acute respiratory syndrome (SARS): a study of 8 autopsy cases from Singapore

Post-hoc power analysis of t-tests was done with standard values (alpha = 0.5) for a continuous endpoint, two independent sample study. Note power for each comparison was >0.99, indicating adequate sample size to determine a difference of this magnitude between the comparison groups. Power of goodness of fit test was done for chi-squared analysis (Fig. 3C ). Power analysis was not done for Fig. 5I , which showed a non-normal data distribution; p-value shown was for Mann Whitney U test.