Morphogenesis and regeneration share a conserved core transition cell state program that controls lung epithelial cell fate

SUMMARY

Transitional cell states are at the crossroads of crucial developmental and regenerative events, yet little is known about how these states emerge and influence outcomes. The alveolar and airway epithelia arise from distal lung multipotent progenitors, which undergo cell fate transitions to form these distinct compartments. The identification and impact of cell states in the developing lung are poorly understood. Here, we identified a population of Icam1/Nkx2-1 epithelial progenitors harboring a transitional state program remarkably conserved in humans and mice during lung morphogenesis and regeneration. Lineage-tracing and functional analyses reveal their role as progenitors to both airways and alveolar cells and the requirement of this transitional program to make distal lung progenitors competent to undergo airway cell fate specification. The identification of a common progenitor cell state in vastly distinct processes suggests a unified program reiteratively regulating outcomes in development and regeneration.

Graphical abstract

In brief

Ke et al. identify a pool of ICAM1/NKX2-1 epithelial progenitors undergoing a highly conserved transitional cell state program controlling crucial outcomes during morphogenesis and regeneration of the human and mouse lungs.

INTRODUCTION

Cells in transitional states are found in multiple biological systems and are enigmatic for their mixed identities, transient gene expression signature, and for harboring key regulators of cell fate switch, thus being plastic and at the crossroad of important events.^1–3 The understanding of the sequence of events and regulation of these transitional states is particularly challenging in developmental systems, where dynamic tissue interactions lead to morphogenetic changes that rapidly generate new compartments and boundaries.^4–6 The developing murine lung offers a paradigm for studying these states. During lung morphogenesis, distal epithelial progenitors in growing buds undergo dramatic changes in fate and cytoarchitecture as they transition into airway progenitors, and later initiate their differentiation program. Lineage studies in mice show that distal epithelial cells give rise to the future alveolar compartment and the epithelial tubules that form the bronchial tree.^7,8 The establishment of these distinct and well-balanced compartments labeled by Sox9 or Sox2, respectively, is a key event during formation of the bronchial tree, preceding the appearance of differentiated epithelial cell phenotypes.^9,10 Cells undergoing fate transition are at the most plastic state as they are subjected to signals that allow them to progressively initiate a distinct program while repressing their initial fate.^1,5,11

The gene expression signatures associated with the distal and proximal lung phenotypes have been extensively reported.^9,10,12,13 However, it has been unclear what characterizes the epithelial cells that lie in the transition between the Sox9-Sox2 fates, their regulation and impact on the developing lung. This is further complicated by species-specific differences in how these compartments form—in humans, for instance, cell fates segregate from a common SOX9/SOX2+ distal progenitor.¹⁴ Thus, the identity, phenotypic features, and behavior of these transitional cells remained elusive.^15–17

Using an unbiased single-cell genomic approach, we identified a panel of markers that allowed labeling and isolation of cells undergoing Sox9-Sox2 transition during lung morphogenesis. This revealed a previously uncharacterized population of Nkx2-1+/Icam1+ epithelial progenitors bearing a signature distinct from all the other developing cell populations. Surprisingly, this signature showed an extensive overlap with that reported in the adult alveolar epithelium undergoing repair after severe injury. In the developing lung, these cells were found dynamically where nascent airways were forming, at sites of SOX9/SOX2 segregation in human fetal lungs. Functional analysis showed that these cells are plastic, responding to morphogenetic cues by changing their abundance and spatial distribution, reflecting the changes in proximal-distal cell fate. Notably, lineage-tracing analysis in vivo and in vitro showed that they are progenitors to alveolar and airway cell types of the lung. Moreover, using the markers identified in the developing mouse lung, we show that in the adult lung this transitional cell state emerges in both alveolar and airway epithelia undergoing repair after severe injury in vivo.

These observations reveal evidence of a common progenitor cell state in vastly distinct processes acting as a unified program reiteratively regulating outcomes in development and regeneration.

RESULTS

Single-cell analysis identifies Icam1 as a marker of epithelial cells undergoing distal-proximal fate transition during lung morphogenesis

Lineage studies have shown that during lung morphogenesis, distal epithelial progenitors undergo cell fate transition to generate the airway compartment. At E14.5, this process is highly active, providing an opportunity to optimize sampling of these epithelial cells undergoing fate transition.¹³ Thus, embryonic lungs from C57BL6 mice were isolated, sorted using Epcam antibody, and profiled by single-cell RNA sequencing (scRNA-seq) (E14.5, n = 3, 2,688 cells). This revealed Epcam+-Cdh1+Nkx2-1 epithelial cells distributed in two main clusters identified by known markers of distal (e.g., Sox9 and Bmp4) and proximal (e.g., Sox2 and Scgb3a2) cell fate markers, as well as three minor indistinct clusters expressing both epithelial and non-epithelial markers (e.g., Acta2, Col1a2, and Lyve1) (Figure 1A). To identify potential markers of cells undergoing distal-proximal cell fate transition, we leveraged “Protein Activity Inference in Single Cells” (PISCES, see STAR Methods), a network-based pipeline that, akin to a highly multiplexed gene reporter assay, measures the activity of relevant proteins in individual cells based on the expression of their transcriptional targets^18,19 (see STAR Methods). We derived single-cell protein activity profiles from 2,438 Nkx2-1+Cdh1+Epcam+ cells representative of the entire lung epithelium. Analysis of the distal-proximal trajectory recapitulated the expected signatures of the distal, proximal, and the transitional domains. We then focused on proteins uniquely activated in cells representative of the transitional domain, particularly on cell surface markers (Figures S1B and S1C). Candidates were further assessed for their localization in areas of active bud morphogenesis at the distal-proximal transition.²⁰ While several candidates emerged from this analysis, Icam1, an established marker of vascular endothelial cells,²¹ showed a striking pattern of distribution in epithelial tubules of the embryonic lungs.

Figure 1. Identification of a transitional cell state during lung distal-proximal specification.

(A) ScRNA-seq of Epcam+ E14.5 lungs. UMAP: distal (red) and proximal (green) epithelial populations and small mesenchymal-like populations.

(B) Immunofluorescence (IF) E14.5 lung: Icam1 (arrow) identifies epithelial progenitors undergoing distal (Sox9) to proximal (Sox2) cell fate transition (bracket) and endothelial cells (small arrow).

(C) Icam1 labels Sox9-Sox2 transitioning Nkx2-1+ epithelial cells throughout branching morphogenesis.

(D) Mouse embryonic lungs cultured in control, Tgfbr inhibitor (SB431542), or Fgfr inhibitor (PD173074) for 48 h. Brightfield (bottom), and IF of Sox9, Sox2, Icam1 (top, middle); Icam1 domain in brackets or * when absent.

(E) ScRNA-seq of Epcam+Icam1+ E14.5 lungs. UMAP: five distinct subpopulations of transitional cells.

(F and G) Dot plot and heatmap: top differentially expressed genes per cluster.

(H) Highest Icam1 enrichment in cluster 4 (dot plot). IF and quantitation of Icam1, Sox9, and Sox2 in E14.5 lung. Graph: mean fluorescence intensity (MFI, arbitrary units), highest Icam1 signals at the Sox9-Sox2 transition. Yellow line (distance pixels).

Scale bars: 50 μm in (B), (C), and (H) and 100 μm in (D).

Icam1 was expressed at the highest level in a transitional domain at the stalk region of lung buds where neither Sox9 nor Sox2 signals were detected. Epithelial Icam1 induction coincided with the downregulation of Sox9 and extinguished after Sox2 expression became prominent (Figure 1B). This pattern was remarkably reproducible reiteratively throughout branching morphogenesis in vivo, and in lung explant cultures (Figures 1C and 1D). Although their fate was undefined, the lung identity was maintained at all stages, supporting the idea that Icam1+Nkx2-1+ cells represent a pool of lung epithelial progenitors undergoing a dynamic transitional cell state.

We reasoned that this population should be exquisitely sensitive to perturbations in the local microenvironment that alter proximal-distal patterning. Thus, wedisrupted endogenous TGF-β/activin signaling during lung morphogenesis using SB431542 (SB, 48 h) and examined the impact in the transitional domain. The resulting disproportional stimulation of distal budding (Sox9+ compartment) relative to the Sox2 domain^15,22–24 was consistent with the aberrantly expanded Icam1+ domain where the Sox9-Sox2 transition occurs. By contrast, treatment with PD173074 (PD17), a potent fibroblast growth factor receptor (FGFR) inhibitor^22,25,26 that disrupts distal bud formation, prevented both the Sox9+ and the Icam1+ transitional domains from forming resulting in aberrantly expanded Sox2+ domain (Figure 1D).

Thus, our analysis identified a not-yet-characterized population of Nkx2-1/Icam1 transitional cells throughout the pseudoglandular period, which are able to change in abundance and regional distribution in response to modulators of pattern formation.

The Nkx2-1⁺ Icam1⁺ progenitors encompass a heterogeneous population of transitional cells

To gain further insights into these transitional cells as they undergo cell fate changes, we used Icam1 as a cell surface marker to isolate and characterize their diversity and behavior. Epithelial cells from embryonic lungs (E14.5, n = 7) double-labeled with Epcam and Icam1 were isolated by fluorescence-activated cell sorting (FACS) and scRNA-seq profile was generated (10× genomics platform (5,622 cells) (Figure 1E).

Clustering analysis²⁷ identified five distinct subpopulations with signatures suggestive of a distal-proximal distribution of epithelial fates along the transition domain (Figure 1F; Tables S1–S5). Uniform Manifold Approximation and Projection (UMAP) visualization revealed Sox2 expression restricted to a small cell population (6.6%) in a single cluster (cluster 3). By contrast, Sox9 expression was found diffusely in cells from all clusters at different proportions (0: 36%; 1: 63%; 2: 51%; 3: 18%; 4: 37%), suggesting that the loss of Sox9 fate occurred progressively throughout the transition domain in contrast to the acquisition of Sox2. Sox9/Sox2 overlap was minimal in Icam1+ cells (1.5%) (Figure S1D). Analysis of these cell populations showed that cluster 1 was highly enriched in distal lung epithelial markers, and cluster 2 was enriched in regulators of cell proliferation. Cluster 3 was recognizable by Sox2 expression, and for its enrichment in components of the retinoic acid (RA)²⁸ and the Yap pathways²⁹ (Figures 1F and 1G). Cluster 0 showed high enrichment in TGF-β pathway components,^30–33 and markers later associated with alveolar type I cells.^34,35 Among these, cluster 4 was the most intriguing of all as it showed no recognizable markers of distal-proximal cell fate and instead was enriched in genes such as Sprr1a, F3, Hbegf, Areg, Lgals3, and Krt8 widely reported in adult lungs undergoing an alveolar type 2-to-type 1 (AT2-AT1) transition cell state during alveolar repair.^2,3,36–38

We had evidence that the epithelial progenitors identified by Icam1^high and cluster 4 genes were the most transitional of all in that domain. Dot plot representation showed Icam1 preferentially enriched in cluster 4. Mean fluorescence intensity of lung sections stained with Icam1, Sox2, and Sox9 antibodies showed the strongest Icam1 signals in the stalk region of epithelial buds where Sox2 and Sox9 signals were the weakest (Figure 1H). A survey of available gene expression databases (Genepaint; Allen Mouse Brain Atlas^20,39) identified various cluster 4 genes mapped to the distal lung epithelium in areas potentially associated with cell fate transition, a pattern further confirmed by immunofluorescence (IF) for representative markers (Figures 2C–2E). From their distinctive signature, we inferred a distal-to-proximal trajectory of these subpopulations ordered along the transition zone from 2, 1, 0, 4, to 3 as determined by RNA velocity,⁴⁰ where cluster 4 resides between the clusters associated with prospective alveolar (2, 1, and 0) and airway cell fate (3) (Figures 2B and S1D).

Figure 2. Nkx2-1/Icam1^high marks a transitional domain during branching morphogenesis.

(A) Heatmap: top differentially enriched genes in cluster 4.

(B) Trajectory analysis: cluster distribution along the distal-proximal transition (boxed area: cluster 4).

(C) FACS and qPCR analysis (Sox9, Sox2, Mki67, and representative transitional markers) of freshly isolated Icam1^high and Icam1^low epithelial cells from E14.5–E16.5. Graphs: mean ± SEM, *p < 0.05, n = 3, paired Student’s t test.

(D and E) Spatial distribution of cluster 4 markers. Tacstd2 (IF) and Fgfbp1 (in situ hybridization) in E14.5 transitional domain (D, arrowheads). Lgals3, Cdkn1a, and Areg (IF) expression in the Icam1+ transition domain (brackets/arrows, side panels) throughout pseudoglandular stage.

We then examined whether the transitional cells identified in cluster 4 could be isolated based on Icam1 levels at different stages. For this, we isolated Icam1^high and Icam1^low epithelial cells from E14.5–E16.5 lungs (~peak-end of branching morphogenesis) and compared the expression of Sox9, Sox2, and cluster 4 genes in these populations. Real-time quantitative PCR showed higher levels of Sox2 and Sox9 in Icam1^low non-transitional cells, as expected from their origin outside the transition zone. Mki67 and Cdkn1a analysis showed that Icam1^high maintained their ability to proliferate, but to a lesser extent than their Icam1^low counterpart (Figure 2C). Icam1^high cells expressed the highest levels of Sprr1a, Lgals3, and Cdkn1a at all stages (Figure 2C). This was validated by similar findings in Icam1^high vs. Icam1^low epithelial populations using published independent scRNA-seq datasets from embryonic mouse lungs (Figure S1E).⁴¹ Consistent with our qPCR data, immunofluorescence (IF) showed maintained expression of transitional cell markers in the Icam1+ domain throughout the pseudoglandular stage (Figures 2D and 2E). The strong IF signals for Icam1, Lgals3, and Cdkn1a at E15.5-E16.5 correlated with the respective increase in mRNA levels of the Icam1^high population at corresponding stages (Figure 2C). Nevertheless, in our survey of E14.5 lungs, several of these markers detected at the mRNA level showed low-to-undetectable protein signals (e.g., Lgals3 in Figure 2E).

Transitional cells are multipotent lung progenitors of both distal and airway compartments

We tested whether these transitional cells served as a pool of multipotent cells able to maintain their uncommitted state and differentiate in culture. For this, Icam1^high and Icam1^low epithelial populations were isolated as above, and examined for their ability to respond to distinct environmental cues as 3D-organoids.

When cultured in LPM3D, a medium that fosters the expansion of multipotent distal embryonic lung progenitors,⁴² both populations formed a large number of organoids by day 10. We found no significant differences in the number or size of the organoids generated from these populations (Figure 3A). Nevertheless, Icam1^high-derived organoids distinctly maintained higher levels of Icam1 and cluster 4 marker expression (Figures 3B and 3C). This was validated by IF of Icam1^high organoids, as shown by the strong Icam1 signals co-labeling with Lgals3, Sprr1a, Krt8, and Cdkn1a, and nearly no expression of Sox9 (n = 3) (Figures 3D and 3E), thus, confirming their maintained transitional character. Interestingly, despite originating from Icam1^high progenitors, these organoids could contain clusters of non-transitional cells. They were readily discernible by their weaker expression of Icam1 and cluster 4 markers, increased Mki67 and strong Sox9 labeling (Figure 3E). This suggested that when cultured under conditions that allowed distal development (LPM3D), the transitional cells could not only generate Icam1^high cells (self-renew) but were plastic to respond to LPM3D and turn on the distal fate program (see in subsequent sections). Neither Icam1^high nor Icam1^low progenitors gave rise to Sox2-expressing organoids in LPM3D, confirming previous observations that this medium cannot induce airway cell fate⁴² or maintain viability of the proximal non-transitional cell population (Sox2+ Icam1^low, see later).

Figure 3. Transitional cells are progenitors to distinct epithelial cell types in organoid cultures.

(A) Generation of organoids from FACS-isolated E14.5 transitional (Icam1^high) and non-transitional (Icam1^low) progenitors in LPM3D. Day 10: wholemount view and quantitative analysis (ImageJ) of the number of organoids/well (top) and organoid size (bottom, area: μm²). Bar: mean ± SEM, from >291 organoids per condition from n = 3 mice; Student’s t test.

(B) FACS analysis: Icam1 fluorescence intensity in day 10 organoids from transitional/non-transitional cells. Graphs: mean ± SEM, n = 3, p = 0.018, paired Student’s t test.

(C) qPCR of transitional markers depicting expression differences between the two organoid populations at day 10 (mean ± SEM, n ≥ 3, *p < 0.05; **p < 0.01, paired Student’s t test).

(D) Inverse correlation of Icam1/Lgals3 with Sox9 signals in LPM3D-cultured organoids from Icam1^high cells (IF), *low or absent signals.

(E) Icam1^high-derived organoids (Icam1+, Sprr1a+, Krt8+, and Cdkn1a+) sometimes contain subpopulations of non-transitional cells. Boxes enlarged in lower panels depict regional heterogeneity in the same organoid. Preferential Mki67 expression in non-transitional cells (Sox9+, dotted area) compared with transitional (Icam1+ and Sox9−). Single channel displayed in side panels.

(F and G) Organoids from E14.5 Icam1^high epithelium cultured in MTEC. Representative wholemount view of day 11 cultures (see also Figure S6 and text). IF: robust induction of Sox2 and programs of basal (Krt5, p63) and secretory (Scgb3a2) differentiation. Small panels in (G): Icam1 expression in a subpopulation of Sox2+ p63+ cells.

(H and I) IF of LPM3D (H) or MTEC (I) organoids confirming the Nkx2-1 lung identity of Icam1+ transitional cells. Lateral panels: single channels or enlarged views.

Scale bars: 50 μm in (D)–(I) and 500 μm in (A).

However, when cultured in MTEC, known to foster the airway cell program, the transitional cell-derived organoids initiated a robust program of differentiation as seen by expression of Sox2, Scgb3a2, p63, and Krt5, markers of the secretory and basal cells (Figures 3F and 3G). These markers were also found in MTEC-cultured organoids from the Icam1^low non-transitional progenitors (see later). The induction of p63+Krt5+ cells was intriguing since, in mice, basal cells are expected to arise from extrapulmonary but not intrapulmonary progenitors, indicating the de-repression of the basal cell program under these culture conditions.⁷ Notably, in MTEC organoids derived from transitional cells, Icam1 expression was found solely in the population that differentiated into basal cells (Sox2+ p63+ Krt5+ population), suggesting that, in response to proximal cues, Icam1 is induced selectively in airway progenitors destined to become basal cells (Figure 3G, insets). Organoids maintained their Nkx2-1 lung identity regardless being cultured in LPM3D or MTEC (Figures 3H and 3I).

To further investigate the contribution of transitional cells to the lung, we used Fgfbp1 (Fgf binding protein 1) as a surrogate marker of the transitional state in lineage-tracing studies in organoid assays. This was well supported by (1) the identification of Fgfbp1 as a key transitional state marker significantly enriched in the cluster 4 in vivo (p^adj:7.22 × 10⁻¹⁴) (Figures 1G and 2A; Table S5), (2) the differentially higher Fgfbp1 expression in Icam1^high-derived organoids compared with Icam1^low (Figure 3C), (3) clear Fgfbp1-Icam1 epithelial colocalization in the transitional domain (Figures 2D and 4A).

Figure 4. Lineage analysis shows contribution of transitional cells to alveolar and airway compartments.

(A) Icam1 and Fgfbp1 co-expression at the transitional domain (arrow). Right panels: single channels.

(B) Generation of lung organoids from transitional (Icam1^high)/non-transitional (Icam1^low) epithelium from Fgfbp1-CreERT2; R26-tdTom lungs in LPM3D or MTEC.

(C) Lineage tracing of transitional cells after 4-hydroxytamoxifen-induced recombination. Abundant tdTom lineage labeling (arrows) in Icam1^high organoids in LPM3D/MTEC. Minimal to no labeling (*) in organoids from Icam1^low cells.

(D and E) tdTom double-labeled with distal (Sftpc) or proximal (Sox2 and p63) markers in Icam1^high organoids cultured in LPM3D or MTEC, respectively. Right panels: enlarged or single channel views. Diagrams depicting transitional cells responses to distal or proximal cues.

(F–I) Lineage tracing of Fgfbp1-CreERT2; R26-tdTom in vivo. Diagrams: tamoxifen administration for lineage analysis at E18.5 (F) or PN5 (I). tdTom (large and boxed areas) double-labeled with markers for AT1 or AT2 alveolar cells (Hopx and Sftpc) and airway secretory (Scgb1a1 and Scgb3a2) or ciliated (Foxj1, acetylated aTub) cells (F and H). (G and I) Quantitative analysis of lineage-labeled cell types (graph: mean + SEM, n ≥ 18 clones/cell type in 3 mice).

Scale bars: 25 μm in (A), 500 μm in (C), 50 μm in (D) and (E), 100 and 20 μm in (F) and (H).

We sorted Epcam+ Icam1^high or Icam1^low cells from Fgfbp1-CreERT2; R26-tdTomato⁴³ E14.5 lungs and generated organoids under LPM3D or MTEC conditions to foster a distal or an airway program, as before. Recombination was induced by treatment with 4OHT (1 μM, 48 h). Analysis of day 8 cultures showed tdTom abundantly expressed in Icam1^high-derived organoids cultured in LPM3D or MTEC (Figure 4C). IF showed tdTom double-labeling with markers of both compartments, confirming Fgfbp1 lineage-labeled descendants in both distal (Sftpc) and proximal (Sox2, p63) compartments (Figures 4D and 4E). By contrast, Icam1^low-derived organoids were largely tdTom negative or showed only rare scattered tdTom-labeled cells, indicating minimal, if any, contribution to the distal or proximal compartment.

We then asked whether these transitional cells served as multipotent progenitors to the distinct epithelial cell populations of the lung in vivo. We conducted lineage-tracing analysis of transitional cells in Fgfbp1-creERT2; R26-tdTom mice, inducing recombination prenatally and tracing their fates at late gestation or neonatally. Analysis of E18.5 lungs (n = 4) exposed to tamoxifen (Tmx) at E11.5 and E13.5 showed extensive tdTom labeling in both distal (72.3%) and airway (27.7%) compartments. To test for a potential bias in differentiation toward a particular cell type in vivo, we performed quantitative analysis of tdTom+ cells double labeled with markers of differentiation (Figures 4F and 4G). This showed lineage labeling predominantly in the alveolar compartment (AT1: 39.9% ± 3.5%; AT2 32.2% ± 6.5%) (Figure 4G). Labeling of airways was identified in secretory (Scgb3a2: 25.5% ± 6.8%; Scgb1a1: 11.7% ± 7.8%), and ciliated (acetylaTub: 15.35% ± 10.31%; Foxj1: 11.23% ± 7.6%).

Analysis of post-natal day 5 (PN5) lungs in which recombination was induced at E13.5 by a single exposure to Tmx also showed tdTom labeling predominantly in the alveolar compartment compared with airways (Figures 4H and 4I). We cannot rule out overrepresentation of distal areas in our sections. Interestingly, while AT1 and AT2 cells were lineage-labeled at similar proportions in the E18.5 lung, at PN5 tdTom increased disproportionally in AT1 cells compared with AT2 and the other cell types. We also found a small population of Sftpc+ Hopx+ double-labeled cells at E18.5 that was extinguished by PN5 (Figures 4G and 4I).

Altogether, these data indicate that transitional cells not only are able to respond to signaling cues to alter proximal-distal cell fate in vitro but also as part of their normal developmental program in vivo. By bearing an Icam1+ transitional signature, they serve as progenitors to multiple cell populations and represent a transient uncommitted plastic cell state.

P53 and AP-1 activities regulate the balance of distal and transitional fates in lung epithelial progenitors

To gain insights into the nature and cues that allow epithelial progenitors to undergo a transitional cell state, we performed gene ontology (GO)/gene set enrichment analysis (GSEA) using the 50 most significantly enriched genes in cluster 4. This revealed enrichment in pathways associated with inflammatory responses, p53 signaling, hypoxia, and lipid metabolism—categories similarly reported in transitional states of the adult lung post-injury.³⁶ P53 is a multifaceted transcription factor often associated with the cellular response to stress, senescence, and inhibition of proliferation.^44,45 Although in our analysis, p53 itself was not differentially expressed in any of the clusters, several of its targets (Cdkn1a, Sfn, and Lgals3)^3,46,47 were enriched in cluster 4, indicative of p53 activation (Table S5). Functional studies in mice have shown that proper control of p53 activity is crucial for lung morphogenesis.^48,49 Notably, p53 gain of function resulted in lung hypoplasia, suggesting that p53 activation regulates the epithelial progenitor cell pool during early lung development.⁴⁸ Based on these observations, we asked how altering p53 signaling could influence the behavior of the epithelial progenitors undergoing this transitional state.

Nutlin3a has been widely used to disrupt p53-MDM2 binding, preventing p53 degradation and thus maintaining this pathway aberrantly activated. Icam1^high and Icam1^low epithelial progenitors were isolated and cultured with Nutlin3a (2 μM) or DMSO (control) in LPM3D for 10 days. Nutlin allowed efficient generation of organoids and expansion of Icam1-expressing transitional cells from both groups. However, the increase in Icam1+ cells was notably higher in the Icam1^low-derived organoids (Figures S2A and S2B). qPCR of these organoids showed increased expression of Icam1 and cluster 4 marker genes compared with controls. Notably, these changes were accompanied by a significant decrease in Sox9 expression (p < 0.05, n ≥ 3). IF supported these findings. Nutlin-treated Icam1^low-derived organoids showed a dramatic increase in Icam1, Lgals3, and Sprr1a signals in sharp contrast with the low levels of Sox9 in their respective controls (Figures S2C and S2D). Thus, unrestricted activation of p53 in the Icam1^low population fostered epithelial progenitors to lose their distal fate and assume a transitional state phenotype. Nutlin treatment of Icam1^high-derived organoids also resulted in increased expression of Icam1 and transitional state markers, which by day 8, became attenuated (Figures S2E–S2G). These observations were remarkably consistent with the reported role of p53 activation in the transition of AT2-AT1 during alveolar repair.³

Additional genes enriched in cluster 4 included AP-1 components, such as Jund, Junb, Fosl2, and Atf3 (p^adj:1.10 × 10⁻⁴¹, 1.08 × 10⁻²⁰, 1.17 × 10⁻⁸, 4.86 × 10⁻¹⁰, respectively) (Figure S2H). AP-1 comprises a family of dimeric transcriptional regulators (Jun, Fos, and Atf) involved in cell fate, reprogramming, and stress response.^50–52 Evidence that some of these components localize to branching epithelial tubules at the distal-proximal transition²⁰ led us to examine AP-1 activity in our system. Treatment of Icam1^low or Icam1^high cells with AP-1 inhibitors (T-5224, SR11302)⁵³ increased Sox9 expression in both compared with controls, suggesting that AP-1 inhibition fosters a distal character in these populations. This was further supported by the strong Sox9 and weak Lgals3 IF signals in the Icam1^high-derived organoids in which AP-1 was inhibited, contrasting with controls (Figure S2I). Whether Sox9+ epithelial progenitors maintain their distal fate or progress to a transitional state is strongly influenced by p53 and AP-1. Interestingly, several lines of evidence link the p53 pathway and AP-1 with stress responses and the control of cell behavior.^52,54,55 Whether they act independently or cooperatively in our system is unclear.

Distal epithelial progenitors are not competent to initiate an airway cell fate program without undergoing a transitional state

Our studies in organoids from Icam1^high and Icam1^low cells provided relevant insights into their ability to respond to different cues. However, how the distal (Sox9+) and proximal (Sox2+) cellular components of the Icam1^low population responded individually to these cues were limited by the inability to isolate these specific components.

To overcome this limitation, we used embryonic lungs from a Sox9EGFP reporter mouse^42,56 and sorted epithelial cells into three distinct populations based on their expression of EGFP, Icam1 and Epcam. Freshly isolated distal (Icam1^lowSox9EGFP^high), transitional (Icam1^highSox9EGFP^low), and proximal (Icam1^lowSox9EGFP^negative) cells were cultured under LPM3D or MTEC conditions to generate organoids as before (Figures S6A and S6B). Consistently, transitional cells generated abundant organoids in both LPM3D and MTEC media. By contrast, the distal and proximal populations were unable to grow or properly generate organoids in MTEC or LPM3D media, respectively (Figures S6B–S6D).

FACS analysis of the organoids cultured in LPM3D showed a significant increase in Icam1 fluorescence intensity in those derived from transitional cells compared with those from distal cells. qPCR confirmed the significant increase in expression of Icam1 and representative transitional state markers, aligning with our earlier findings in Icam1^high- and Icam1^low-derived cultures (Figure S6C). qPCR could still detect decreased Sox9 expression in the transitional cell-derived organoids; however, FACS analysis showed Sox9EGFP in both populations with no significant difference in EGFP intensity between these organoids. This supports the notion that in the developing lung, transitional cells represent a pool of highly plastic progenitors able to self-renew and respond to signaling cues of the LPM3D environment (e.g., Wnt and Fgfs) converting into a distal fate.

Notably, when cultured in MTEC media, the distal progenitors, which under the current sorting strategy contained none of the proximal components of the former Icam1^low cells, no organoids were generated. This contrasted with the transitional population, which responded to the MTEC cues, generating abundant organoids that expressed airway markers (Sox2, Scgb3a2, and p63) at levels comparable to those of proximal-cell-derived organoids (Figure S6D). Immunofluorescence staining corroborated these findings (Figure S6E). Thus, despite being multipotent,⁸ Sox9+ distal progenitors are not competent to respond to proximalizing signals to initiate an airway proximal program. This competence appears to be acquired as they undergo this transitional cell state.

Lastly, to evaluate the extent of cell fate conversion within the transitional and distal populations in response to LPM3D media, we employed FACS to compare the changes in Icam1/EGFP fluorescence intensity between freshly isolated transitional and distal populations, and their respective derived organoids after 8 days in culture. The “in vivo” features of each of these populations were used as a reference for establishing gates to measure population shifts. This showed that 88.4% of cells derived from the original transitional pool cultured in LPM3D maintained high levels of Icam1, while 95.3% maintained low levels of EGFP. This suggests that transitional cells predominantly exhibit self-renewal characteristics in vitro yet display a degree of plasticity allowing them to adopt a distal fate. Similarly, when quantifying changes in the distal population after culture, 31.7% of cells derived from the original distal pool cultured in LPM3D maintained low levels of Icam1, and only 21.6% upheld high levels of EGFP. These findings suggest that distal cells shifted to a transitional state under these conditions, which would be also consistent with previous evidence that Sox9+ distal buds are progenitors to all lung epithelial lineages during development⁸ (Figures S6F and S6G).

The transitional cell state program elicited during epithelial morphogenesis is induced during lung regeneration-repair

The striking resemblance in gene expression signatures between cells undergoing transitional state in embryonic lungs and in bleomycin-injured lungs led us to further investigate this relationship. We performed a cluster-specific enrichment analysis of several published signatures from lung injury studies that identified an intermediate AT2-AT1 cell state (PATs and DATP) that was enriched at a specific stage during regeneration-repair of the alveolar epithelium^2,3,36,37 (see STAR Methods). We found the highest statistically significant enrichment of these signatures in cluster 4, suggesting that programs similar to those found in injury-repair are at play in development at the transition zone (Figure S3A). Given that Icam1 was also most enriched in this cluster (Figure 1H), we asked whether areas undergoing injury-associated transitional state can be identified by Icam1.

For this, we induced alveolar injury in adult mice with bleomycin³ and compared Icam1 IF signals between control and injured lungs. Analysis of 14 day post-injury (dpi) lungs, when transitional cells are most abundant, showed aberrantly increased Icam1 signals in the epithelium lining the abnormal airspaces of severely damaged areas. These strongly Icam1-expressing cells had different morphologies, sometimes assuming a flat AT1-like shape, but without expression of AT1 markers. By contrast, in uninjured control lungs, Icam1 was largely confined to vascular endothelial cells, with only weak epithelial signals in Hopx-labeled AT1 cells (Figure 5A).

Figure 5. Icam1 identifies transitional cell states in adult multipotent progenitors from airways and alveoli.

(A) Bleomycin treatment of adult lungs and analysis 14 days post-injury (dpi). Strong Icam1 signals (IF) in injured alveolar (alv) epithelium contrasting with control alveolar/endothelial (bv, blood vessel) signals. Hopx labels AT1 in control (arrow) but not Icam1+ transitional cells (*) of injured lungs.

(B) FACS isolation and quantitation of Icam1+Epcam+ cells from control and bleomycin 14 dpi lungs. Graph: mean ± SEM, fluorescence intensity n = 3, p < 0.05, paired Student’s t test.

(C) Significant differences in transitional marker expression between Icam1^high and Icam1^low epithelial cells from 14 dpi bleomycin-treated lungs (qPCR). Mean ± SEM, n = 3, *p < 0.05, **p < 0.01, paired Student t test.

(D and E) Icam1 co-expression with Krt8 and Cdkn1a in bleomycin 14 dpi alveolar epithelium. Graph: percentage of Icam1+ cells coexpressing transitional makers. Mean ± SEM, n = 3.

(F) Airway injury-repair after polidocanol. IF: Icam1 and transitional markers in control and 7 dpi adult tracheas. Control (left): basal cells (Krt5+, arrows) express Icam1 but not transitional markers (*). Major transitional marker induction in regenerating epithelium (right).

(G) Mean fluorescence intensity of transitional markers in control and polidocanol-injured airways. Graph: mean ± SEM of >70 cells per condition (each dot represents one cell) n ≥ 3, ****p ≤ 0.0001, Student’s t test.

Scale bars: 50 μm in (A, top) and (F) and 20 μm in (A, bottom), (D), and (E).

We then tested whether we could isolate injury-induced transitional cells in the adult lung as we did in the developing lung. Uninjured and injured adult lungs were labeled with Epcam and Icam1, and Icam1 mean fluorescence intensity was compared. Bleomycin-injured lungs were readily distinguished from controls by their significantly higher Icam1 signals (Figure 5B). Next, we examined whether we could isolate transitional cells from other epithelial populations of bleomycin-injured lungs (14 dpi) based on Icam1 levels. Indeed, qPCR analysis of sorted Icam1^high and Icam1^low cells showed significantly higher levels of Icam1, Lgals3, Cdkn1a, and Sprr1a in the Icam1^high cells, confirming that these represented the population of transitional cells in the injured lungs (p < 0.05 and p < 0.01) (Figure 5C). We performed a similar analysis of the most significantly enriched markers of cluster 4 in Icam1^high and Icam1^low epithelial populations using a scRNA-seq dataset from control and bleomycin-treated Sftpc-CreERT2; R26-tdTomato adult mice.³⁶ This revealed enrichment of these markers in Icam1^high population consistent with our results (Figures S3B and S3C). IF of 14 dpi bleomycin-injured lungs showed that cells undergoing intermediate transitional states identified by strong Icam1 co-express cluster 4 markers (Figures 5D, 5E, S3D, and S3E). Quantitative analysis showed Icam1 co-labeling at different proportions, as exemplified by Krt8 (60%–82%), Cdkn1a (47%–62%), Lgals3 (29%–43%), and Sprr1a (15%–29%). These results further emphasize the dynamic nature of this transitional state, presumably activating selected components of this program at specific phases. The regulation of this process is likely to be multifactorial and include dynamic changes in factors, such as cell shape.³ Indeed, we observed increased levels of Icam1 and nuclear localization of Cdkn1a in cells with different morphologies in injured lungs (Figures S3D and S3E).

Importantly, transitional states have been reported in both bleomycin-induced lung fibrosis in mouse models as well as in human idiopathic pulmonary fibrosis (IPF).^3,37 We performed IF analysis of KRT8 and LGALS3, transitional markers we identified in mouse embryonic lungs that were aberrantly expressed in human lungs from IPF subjects. In these lungs, ICAM1 was co-expressed with these markers in the injured alveolar epithelium (Figures S3F and S3G).

We tested the possibility that the transitional program identified during morphogenesis and alveolar repair could emerge in other regionally distinct compartments of the respiratory system undergoing repair. We examined the regenerative response of the airway epithelium after massive injury to the luminal cells in established mouse models of injury-repair in vivo. Intratracheal instillation of polidocanol (POL), a detergent/sclerosing agent, is known to result in extensive epithelial sloughing followed by a regenerative response that repopulates the airway epithelium within 2–3 weeks.⁵⁷ At 7 dpi, POL-injured mice showed the typical regenerative response with expansion of the basal and luminal cell populations. IF of tracheal sections stained for Icam1 and the markers of transitional state showed a striking increase in expression both in the basal and newly formed luminal cells (Figure 5F, 5G, and S4A). Quantitative assessment of mean fluorescence intensity confirmed the significant increase in Icam1, Lgals3, Sfn, Cdkn1a, Tacstd2, Sprr1a, F3, and Hbegf in POL-treated compared with controls (p < 0.05; n = 3) (Figure 5G).

The widespread increased expression of transitional cell markers in the regenerating injured airways led us to investigate whether they were associated with the emergence of any specific cellular phenotype. 7 dpi tracheal sections were co-stained with transitional markers (Icam1 and Lgals3) and a panel of markers for secretory, ciliated, basal, and hillock cells. This showed expression of these transitional markers in all the cell types listed above. Quantitative analysis revealed Icam1 and Lgals3 co-labeling with p63 and Krt13+ hillock cells at nearly equal proportions and less labeling in secretory Scgb1a1 cells (Figures S5A and S5B). Transitional state markers were similarly induced during repopulation of the airway epithelium post-naphthalene injury. Interestingly, at later stages, once regeneration was nearly completed, expression of these markers declined (Figure S4B).

Intriguingly, in our IF analysis of the adult tracheal epithelium under homeostatic conditions, basal cells were clearly labeled by Icam1 but not for transitional state markers. We have recently reported a comprehensive analysis of scRNA-seq of adult murine basal cells. We revisited the scRNA-seq database from adult murine basal cells we recently reported to check whether transitional markers were indeed absent in these cells.⁵⁸ This, instead, showed Icam1 and cluster 4 transcripts widely expressed in basal cells. Some of these transcripts (i.e., Sprr1a, Hbegf, Areg, and Tnfrsf12a) were preferentially enriched in specific basal cell subpopulations, potentially reflecting in the differences in cell behavior in response to injury (Figure S4C). Together, these observations argue for the appearance of a transitional cell state in the lung in a much broader context, both under normal and altered conditions.

Epithelial progenitors of the human fetal lung undergo a highly conserved transitionalcell program during distal-proximal cell fate specification and early differentiation

Lung development in humans and mice undergoes overall similar morphogenetic events. However, unlike in mice, the human lung bud tips maintain a population of double SOX9+SOX2+ epithelial progenitors throughout branching morphogenesis. These fates segregate progressively as new generations of airways form until most of the branching is completed by the end of the pseudoglandular period.¹⁴ We asked whether this distinct mechanism of distal-proximal specification also involved the appearance of a transitional cell state and how this related to our findings in the murine lung.

First, we tested whether ICAM1 could identify sites of distal-proximal transition throughout the pseudoglandular stage as it did in mice. IF analysis of 10- to 18-week gestational human fetal lungs showed a remarkably conserved distribution of ICAM1 at these sites (Figure 6A compare with Figure 1C). Prior to 11 weeks, epithelial ICAM1 was nearly undetectable compared with the adjacent endothelium, increasing significantly later to occupy a large domain later when the SOX9 and SOX2 fates segregated in the 18-week lung (Figures 6A and 6B).

Figure 6. Human fetal lung harbors a highly conserved transitional cell program during morphogenesis.

(A) IF: human fetal lung (gestation weeks 10, 11, 13, 15, and 18 post-coitum). ICAM1 signals in epithelium (arrows/bracket) at the SOX9-SOX2 transition and in endothelium.

(B) ICAM1 mean fluorescence intensity in epithelial over endothelial signals (10–18 weeks lungs). Graph: mean ± SEM, 3–6 measurements per stage.

(C) UMAP: EPCAM+ cells from scRNA-seq of 11-, 15-, and 18-week human distal lungs. Dot plot: representative genes differentially enriched in each cluster. Cluster 1 (box) expresses a highly conserved transitional state signature (see Figure 2).

(D) UMAP: ICAM1, SOX9, and SOX2 are enriched in distinct but partially overlapping clusters at different stages. Graph: ICAM1 percent overlap with SOX9/SOX2 quantified from reported scRNA-seq dataset.⁵⁹

(E) Violin plot of ICAM1 expression in the developing human lung epithelium.

(F and G) Representative IF images of 11.5–13-week human lungs showing ICAM1, LGALS3, TACSTD2, KRT8 expressed at the transitional domain where SOX9-SOX2 cell fates segregate (arrows).

Scale bars: 20 μm in (F) and 50 μm in (A) and (G).

Next, we combined information from publicly available scRNA-seq databases with high resolution microscopy and quantitative analysis of human fetal lungs to gain information on how these dramatic changes in ICAM1 expression related to the SOX9-SOX2 fates and define the signature of the transitional state in these cells. Single-cell transcriptomic datasets from 11, 15, and 18 weeks human fetal distal lungs were selected for their representative stages of branching morphogenesis.⁵⁹ To identify cells potentially at the transitional state, epithelial cells (EPCAM+) were computationally extracted, integrated, and analyzed (Figures 6C and S7A). This revealed five distinct clusters. The cluster of cells exhibiting a transitional state signature was readily recognized for its remarkable similarity with that we found in the murine lung with differential enrichment in ICAM1, LGALS3, TACSTD2, F3, ATF3, and others. The remaining clusters were recognized by expression of known distal genes (cluster 0 and 2: SFTPC, CPM, and SOX9), proximal genes (cluster 3: SOX2 and SCGB3A2) and proliferation markers (cluster 4: MKI67 and PCNA). UMAP showed SOX2-, SOX9-, and ICAM1-expressing epithelial populations distributed in partially overlapping but distinct clusters (Figures 6C, 6D, and S7A). Analysis of different stages showed the ICAM1+ cells increased in numbers and expression over time (Figure 6E). Notably, the differential enrichment of the transitional markers in ICAM1-expressing epithelial cells was evident at all stages analyzed (Figures 6C and S7A). However, ICAM1 co-labeling with SOX2 and/or SOX9 varied considerably at each of these stages.

At 11 week, ICAM1 was expressed in the SOX2+ SOX9+ double-labeled cells (17.7%) and in a single-labeled SOX9 (23.5%) or SOX2 (52.9%) epithelial population. ICAM1 was also found in a small population of SOX2/SOX9 double-negative epithelial cells (5.9%) (Figure 6D). IF of 11–13 week lungs confirmed ICAM1 signals along the distal-proximal transition in a pattern remarkably consistent with the sites of SOX9-SOX2 segregation, coinciding with the local expression of transitional markers, such as TACSTD2, LGALS3, and KRT8 (Figures 6F and 6G). By 15 weeks, once several generations of airways have formed, a double SOX9/SOX2-negative domain became evident where these fates had segregated. The strong expression of ICAM1, TACSTD2, and LGALS3 confirmed their transitional state character (Figures S7B and S7C). The presence of NKX2-1 at all times indicated that these transitional progenitors maintained their lung identity (Figure S7D). The ICAM1+ SOX9−/SOX2− domain continued to lengthen toward the canalicular stage. Quantitative analysis showed an increase in this cell population, from 5.9% (11 weeks) to 46.2 % (15 weeks) and later to nearly 80% at 18 weeks (Figure 6D). By then, the ICAM1+ SOX9+/SOX2+ population was nearly extinguished, and co-labeling with SOX9 or SOX2 individually greatly reduced. Similar analysis of another scRNA-seq database from human fetal lungs confirmed these trends (Figure S7E). The strong ICAM1 and transitional marker expression at 18 weeks suggested that this cell population maintained its transitional state but was no longer associated with SOX9/SOX2 segregation. Interestingly, scRNA-seq datasets from 11–18 weeks lungs showed representative markers of AT1 (HOPX, AGER, and PDPN) and basal-like/basaloid (KRT17)⁶⁰ differentially enriched in cluster 1 (transitional; Dotplot Figures 6C, S8A, and S8B). The continued AGER/KRT17 enrichment in the ICAM1-high-expressing cells (Figures S8A and S8B) led us to examine their distribution along the expanded ICAM1 domain as SOX2/SOX9 fates segregated. IF confirmed AGER/KRT17 double- or single-labeled cells in the ICAM1 domain (Figures 7A–7C and S8C), indicating that after SOX9-SOX2 segregation, ICAM1 continued to identify cells undergoing cell fate transition, presumably regions initiating an early program of basal-like or AT1 cell fate commitment.

Figure 7. The transitional program is associated with initiation of differentiation and restricted proliferation in the human fetal lung.

(A and B) Strong KRT17 Immunofluorescence signals at the site of SOX2/SOX9 segregation (A) overlapping with AGER (B, arrows) in 13–15 weeks human fetal lungs. Boxed areas enlarged in side panels.

(C) KRT17/AGER double-labeling (arrow) in the 18-week transitional SOX9/SOX2 double-negative domain (1: brackets; see also Figure 6A) but not in the SOX2+ proximal epithelium (2).

(D) Distinct distribution of SCGB3A2+ in 18-week lungs in (1) ICAM1+ transitional domain (brackets) as small cell clusters SOX2^low and, in (2) SOX2^high proximal domain (airway). Boxed areas enlarged in lower panels.

(E and F) MKI67, ICAM1, and SOX9 immunostaining in human fetal lungs. No increased MKI67 labeling in ICAM1+ cells (yellow or blue arrowheads) prior to or during expansion of the transitional domain (bracket, F).

(G) Proportion of ICAM1^high and ICAM1^low proliferating epithelial cells in 9–18-week human lungs from analysis of scRNA-seq databases.⁶¹ Graph: percentage of G2M cells at the stages analyzed.

Scale bars: 50 μm (A)–(F).

Studies in IPF and human 3D-lung organoids have shown that transdifferentiation of AT2 toward a basal cell phenotype involves the emergence of alveolar-basal (AB) intermediate states, ABI1 and then ABI2, identified by a discrete gene expression signature.^37,62,63 KRT17 (KRT5−) is a distinctive feature of the ABI2 signature. We inquired whether the KRT17 found in the transitional domain could represent an AB intermediate state toward a distal program of basal-like cells. We compared the signatures of the fetal human lungs (11, 15, and 18 weeks) and the 3D organoids undergoing ABI1 to ABI2 states⁶³ (Figures 6C and S9A).

We selected 25 genes that most significantly distinguished ABI1 from ABI2 and examined their enrichment in the epithelial populations of human fetal lungs (clusters 0–4 in Figures S9A and S9B). This showed a greater similarly of fetal human transitional cells (cluster 1) with ABI1 rather than ABI2. Based on this, KRT17 is more likely to be a bona fide marker of transitional state of the human pseudoglandular stage rather than a marker of the basal/basaloid program in the transitional domain. Indeed, AGER/KRT17 cells could be identified in 11-week lungs in the transitional domain undergoing SOX2/SOX9 segregation, and no other basal cell marker was expressed locally (Figure S8D). Moreover, ICAM1 co-labeled mostly with AGER and minimally with SFTPC (AT2 marker) (Figures S8E and S8F). Notably, analysis of 18-week lung showed SCGB3A2, a marker of airway secretory cells, in the expected SOX2+ compartment but also in a small population of ICAM1+ transitional cells (Figure 7D). This subset of SCGB3A2+ cells did not distinctly express SOX2 but was intriguingly associated with scattered SOX2+ cells, suggesting this to be a not-yet characterized program of cell fate emerging in this extended ICAM1+ domain.

Lastly, to gain insights into how the transitional domain expanded, we mapped the sites of epithelial cell proliferation during branching morphogenesis. Surprisingly, although MKI67 labeled SOX9+ distal cells, at this stage, labeling was largely associated with the stalk regions (Figures 7E, 7F, and S10A). This was further confirmed by PCNA staining and altogether differed sharply from the pattern reported in the developing mouse lung⁶⁴ (Figure S10B). Analysis of another transcriptome database from human fetal lungs⁶¹ and IF quantification confirmed that most proliferative cells were not in the SOX9 high compartment (Figures S10C–S10E). We found little to no evidence of significant MKI67/ICAM1 overlap. Rather, MKI67 was found most frequently in ICAM1− neighbor cells (Figure 7E). Although ICAM1+ epithelial cells increased substantially from 9 to 18 weeks, the majority of G2M cells were ICAM1− (Figure 7G). Overall proliferation declined later in the human fetal epithelium (Figure S10C). By 18 weeks, when ICAM1+ SOX9−/SOX2− cells occupied a large domain, MKI67 expression was found only in a few cells (Figure 7F). The data indicated that the expansion of the SOX9−/SOX2− transitional domain is driven by proliferation of a largely ICAM1− population or by a currently undetermined mechanism.

DISCUSSION

During lung morphogenesis multipotent Sox9/Id2/Nkx2-1 distal buds generate two major pools of epithelial progenitors that later form the alveolar and airway compartments.^7,8 Here, we provide evidence of an additional pool of epithelial progenitors collectively identified by their (1) high Icam1 levels and a signature strikingly similar to that of adult injured alveolar and airway epithelium undergoing repair, (2) consistent localization in a transitional domain adjacent to newly formed distal buds throughout the pseudoglandular stage in both mouse and human fetal lungs, and (3) ability to generate distal and proximal epithelial phenotypes in organoid assays and in vivo, as demonstrated by lineage-tracing analysis.

Although we identified Icam1 as a key marker of the transitional cell state, there is no evidence of lung developmental abnormalities in Icam1-null mice.⁶⁵ Furthermore, so far, no obvious lung phenotype has been reported in mice deficient in Tacstd2, Fgfbp1 and other transitional markers.^66–68 It is interesting that in the human 18 week lung an extended ICAM1 domain expresses canonical markers of AT1, basal-like and secretory cells after SOX9-SOX2 segregation, as if initiating distinct differentiation programs. The identification of SCGB3A2+ cell clusters in the transitional domain, suggested the appearance of a local distal secretory cell program independent of that initiated in the proximal airway epithelium. Indeed, markers of neuroendocrine differentiation have also been identified closer to the bud tips of 11 week human fetal lungs, where we found transitional state markers.⁶¹ This raises the possibility that the SCGB3A2+ cells identified in the transitional domain could be part of an early differentiation program in emerging respiratory bronchioles, providing a possible developmental origin of a poorly understood architectural niche in the human lungs.^69,70

Altogether these findings provide first evidence of a common transitional cell state program identified by NKX2-1/ICAM1 early in epithelial progenitors of the embryonic lung remarkably conserved in the adult lung stem cell pool regardless if in facultative (AT2) or dedicated (basal cells) progenitors.⁷¹ In spite of the major overlap with markers originally described in injured and IPF lungs, our data suggest that the transitional program identified by this signature is not pathological but rather represents an integral part of the normal developmental and regenerative processes. Our observations broaden the understanding of these transitional cell states and highlight their potential relevance in the development of regenerative strategies and management of pulmonary conditions.

Limitations of the study

Although we provide accumulated evidence that transitional cells serve as epithelial progenitors to airway and alveolar compartments, additional work is required to determine the impact of their ablation in the developing lung and their ability to repopulate the adult injured lung. Additional work is also needed to further understand the differentiation program emerging from the extended transitional domain in human fetal lungs where we identified SCGB3A2+ in presumptive future respiratory bronchioles. Lastly, the epithelial-mesenchymal cross talks associated with the transitional cell state have not been investigated here and are currently being studied.

RESOURCE AVAILABILITY

Lead contact

Further information and requests for resources and reagents should be directed to and fulfilled by the lead contact, Wellington V. Cardoso (wvc2104@cumc.columbia.edu).

Materials availability

This study did not generate new unique reagents.

Data and code availability

Single-cell data are publicly available (GEO accession numbers GEO: GSE254356). Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

STAR★METHODS

EXPERIMENTAL MODEL AND STUDY PARTICIPANT DETAILS

Mice

We used timed-pregnant and adult mice C57BL/6J (JAX Strain #000664), Sox9-IRES-EGFP (JAX Strain #030137) obtained from Jackson Laboratory and Fgfbp1-CreERT2; R26-tdTomato⁴³ mice. TMX (Sigma, T5648) was dissolved in sunflower seed oil (Sigma, S5007) at a stock concentration of 20 mg/ml. Timed-pregnant mice were exposed to 240 μg/g body weight TM via oral gavage. TMX-administration regimens for all experiments including lineage tracing in vivo and in vitro are described in the results section. Detection of a vaginal plug was considered to be E0.5. All mice were housed in standard ventilated polysulfone microisolator cages, up to 5 mice per cage, on irradiated cob bedding. Mice had access to reverse osmosis water and an irradiated LabDiet 5053 pelleted diet ad libitum. Mice were free of Sendai virus, pneumonia virus of mice, murine hepatitis virus, minute virus of mice, mouse parvovirus, Theiler’s mouse encephalomyelitis virus, reovirus type 3, epizootic diarrhea of infant mice virus, lymphocytic choriomeningitis virus, ectromelia virus, mouse adenovirus, K virus, polyoma virus, Mycoplasma pulmonis, ectoparasites, and endoparasites. Animals were maintained in accordance with the Guide for the Care and Use of Laboratory Animals in an AAALAC-accredited facility. All procedures outlined in the study were approved by Columbia University’s IACUC (WVC IACUC #: AC-AABF2567).

Human Lung Tissue

Fetal (8–18 week gestation) human lungs from terminated pregnancies were obtained from the Birth Defects Research Laboratory at the University of Washington Seattle with ethics board approval. Maternal written consent was given, and the donors were aware that tissue acquired would be used in research. This study was performed in accordance with ethical and legal guidelines of the Columbia University institutional review board.

METHOD DETAILS

Naphthalene/Polidocanol/ Bleomycin injury mouse models

Wild-type C57BL/6 6–12-week-old mice were used for induction of Naphthalene or Polidocanol injury-repair. Only females were used for naphthalene injury. Both male and females were used for polidocanol injury. The studies were approved by Columbia University Institutional Animal Care and Use committees (IACUC). For Naphthalene injury, a single dose of Naphthalene dissolved in sunflower oil was administered to female wild-type mice by IP injection at 275 mg/kg body weight to induce injury. Freshly prepared naphthalene was administered before noon.⁷³ For Polidocanol injury, wild-type mice were anesthetized and administered 20μl of 2% polidocanol (poli, freshly prepared in PBS) by oropharyngeal aspiration delivery following previously published protocols.⁵⁷ Animals were sacrificed on day 3, 5, and 30(Nap), or day 7 (Poli) post injury. For bleomycin-induced lung injury, 2 U/kg bleomycin was administered by oropharyngeal aspiration, The mice were monitored daily and were sacrificed on day 14 post injury.

Embryonic lung explant cultures

E12.5 embryos were collected, and lungs were dissected out and placed on a 6-well transwell insert. BGJb media, supplemented with 1% FBS, 1% Penicillin-Streptomycin and 0.25mg/ml ascorbic acid, was added in the bottom chamber. Media was sterilized with a 0.22μm filter prior to use and stored for no more than a week. Explants were incubated at 37°C with 5% CO2 and imaged every 24 hours. Treatments to modulate Fgf and Tgfb signaling were initiated at culture start by adding 1μM PD173074 or 10μM SB431542, respectively. Control cultures were treated with vehicle, excluding the active compound. After 48 hours, cultures were washed with PBS and fixed in 4% PFA overnight at 4°C. The next day cultures were washed in PBS and processed for Wholemount imaging.

Generation and analysis of mouse 3D organoid cultures

For distal culture conditions, sorted cells were resuspended in Matrigel droplets at 100 cells/μl and seeded in 12-plate well supplemented with LPM3D media. For proximal culture, 4×10³ sorted cells were resuspended in MTEC/Plus medium, mixed 1:1 with growth factor reduced Matrigel, and seeded into Transwell inserts. MTEC/Plus was added to the lower chamber and media was changed every two days until collection. For Nutlin treatment, 2μM of Nutlin was added to 4-day cultures for 4d or 6d before harvested. For T-5224 treatment, 40μM was added for 2d and for SR11302 treatment, 10μM was added 4d at day 4. For lineage labeling of Fgfbp1-CreERT2; R26-tdTom organoids, 1μM of 4OHT was added directly to either LPM3D or MTEC media for 48h, and then with fresh media. For subsequent FACS analysis, organoids dissociated from Matrigel with 2mg/ml Dispase for 1hr at 37°C were then resuspended in TryplE for 5min. The single cells were then suspended in FACS buffer (2% FBS, 2mM EDTA, 5μM Y27632, 1M HEPES in HBSS) and stained as described below.

Preparation of single-cell suspensions of adult lungs for FACS

Mice were euthanized and perfused through injection of PBS into the right ventricle. These lungs were incubated in PBS on ice before dissecting off lobes and placing them into 4ml of digestion buffer (Dispase, 5 U/ml; DNase I, 0.33 U/ml; and collagenase type I, 450 U/ml). Lungs were incubated at 37°C on a rocker for 40 minutes before being dissociated with frequent pipetting using a 1000ul pipette. Cells were passed through 70μm and then 40μm cell strainers. Red cell lysis buffer was used to remove red blood cells. Cells were then resuspended in FACS buffer (2% FBS, 2mM EDTA, 5μM Y27632, 1M HEPES in HBSS) and stained as described below.

Preparation of single-cell suspensions of embryonic lungs for FACS and scRNAseq

To obtain a single cell suspension for scRNAseq, two C57/B6 females were euthanized and E14.5 embryos were collected. Lungs were dissected and kept in ice-cold HBSS until the start of dissociation. Lungs were pooled and minced manually with a blade into pieces of ~1mm in diameter in 4ml dissociation buffer (0.8ml HBSS (1X), 2ml Dispase II, Stemcell Technologies #7913, 2.5 U/ml, 400μl Collagenase A, Roche #10103578001, 800ul DNase, 200ug/ml, Sigma, Cat#DN25), followed by incubation at 37°C for 1 hour on a rotator with gentle P1000 pipette trituration every 20 minutes. Then, ice-cold FACS buffer (2% FBS, 2mM EDTA, 5μM Y27632, 1M HEPES in HBSS) was added. The suspension was filtered through a 70μm filter, spun down at 5 minutes at 400g, and incubated 5 minutes in red blood cell lysis buffer (sigma). The buffer was diluted with FACS buffer, spun down for 5 minutes at 400g and resuspended in 1ml FACS buffer. The suspension was then passed through a 40μm filter, cells were counted and incubated 40 minutes with Epcam-APC antibody (1:150) and Icam1-PE antibody (1:150) on ice in the dark. After staining, the suspension was three times washed in FACS buffer and spun down 5 minutes at 400g. Finally, DAPI was added at a final concentration of 1μM in FACS buffer.

Single cell RNA sequencing and Computational analysis

Samples for single cell RNA sequencing were stored on ice immediately after cell sorting and processed at the Columbia University Genome Center for 103 Chromium single cell sample preparation. Single Cell 3’ libraries were prepared using the – 3’ v3 Protocol according to the manufacturer’s manual (103 Genomics). The pooled, 3’- end libraries were sequenced using Illumina NovaSeq 6000. Cellranger v2.1.1 was used for demultiplexing, alignment and mapping using Mus musculus reference genome v10 (mm10). Downstream analysis and quality controls were performed on Seurat v 4.3.0. We excluded from analysis cell doublets, cells containing more than 8% of mitochondrial RNA reads and cells with less than 200 genes detected (indicative of dying cells). Genes Gm42418 and AY036118 were also removed, as they overlap the rRNA element Rn45s and represent rRNA contamination. Potential effects of the mitochondrial gene expression percentage were regressed out by feeding this variable to ‘ScaleData’ through the ‘vars.to.regress’ argument. Using the ‘CellCycleScoring’ command, cell cycle phase was estimated per cell, and these estimations were used to regress out the effect of cell cycle by re-running ‘ScaleData’ with this variable fed to the ‘vars.to.regress’ argument. Principal components (PCs) were selected by running ‘RunPCA’ followed by diagnostic Jackstraw, heatmap and elbow plots to aid in PC selection. The first 10 PCs were selected and used to cluster the cells by Louvain community detection (‘FindNeighbours’ and ‘FindClusters’), followed by UMAP-based dimensionality reduction and visualization (‘RunUMAP’). UMAP and Louvain clustering were used to identify and remove non-epithelial cells. Louvain community detection was run iteratively on the epithelial only cells with increasing resolution parameter (0.1–3) and optimal clustering result was selected (0.5). Top marker genes for each cluster were calculated using ‘FindMarkers’ and plotted in a heatmap for visual inspection. Markers for each cluster (Tables S1, S2, S3, S4, and S5), obtained using the FindAllMarkers command in Seurat, were utilized for identifying specific signaling pathways and gene ontology through Enrichr.

For comparison to public single cell data from GEO: GSE149563⁹, GEO: GSE145031³⁶ , EMBL-EBI: E-MTAB-8221⁵⁹, and EMBL-EBI: E-MTAB-11278⁶¹, the indicated conditions/timepoints were integrated into a single UMAP using SCTransform and analyzed using Seurat package. To generate heatmaps, differential gene expression tests were run and DEG heatmaps of selected marker genes were plotted. Icam1^high populations are defined through careful examination of Icam1 normalized expression value through gene expression visualization plots to determine the true transitional population. The threshold used are as below: Icam1>=1.5 in Frank et al.⁹; Icam1>=3 in Choi et al.³⁶; and ICAM1>=0.5 in Miller et al.⁵⁹ For SOX9/SOX2 cell fate segregation quantitative analysis in Figures 5 and S6 using data from Miller et al.,⁵⁹ quantification was done by defining the SOX9/SOX2 negative population as SOX9<=0/ SOX2<=0; ICAM1/SOX9/SOX2 positive population as SOX9>=0.1/SOX2>=0.1/ICAM1>=0.1. For proliferating cells quantification in Figures 6 and S8, proliferating cells were defined as G2M phase cells unbiasedly determined by Seurat in the epithelial single cell data set from He et al.⁶¹ ICAM1^high proliferating cells defined as ICAM1>1 in G2M phase.

PISCES analysis

For the analysis of the scRNAseq signatures generated from Epcam+ sorted E14.5 epithelial cells we also used a computational approach to infer protein activity as a metric of the enrichment in transcriptional targets of the differentially expressed genes compared to a baseline control state. PISCES (Protein Activity Inference in Single Cells; bioRxiv 445002) comprises the sequential use of the algorithms ARACNe and VIPER¹⁹ (see also Lachmann et al.⁷⁴; Margolin et al.⁷⁵). ARACNe analyzes single cell transcriptional data in conjunction with a list of putative regulator-target interactions to reverse engineer scRNASeq-derived transcriptional profiles and build a network of regulator target interactions, termed a ‘regulon’ specific to the analyzed dataset. Using Epcam+ scRNA dataset as input, this produced a regulon encompassing 2267 regulators derived from 2438 Nkx2-1+Cdh1+Epcam+ cells. VIPER then combines regulon and gene expression signatures derived from the scRNAseq transcriptional profiles to infer cell-specific protein activity profiles. To analyze the protein activity profile in the Epcam+ E14.5 single cell dataset, we constructed a trajectory between the distal and the proximal cluster using the Multiway K-Means algorithm. k=2 was chosen to represent the distal and proximal clusters. The output of the multiway k-means clustering algorithm was an ordered vector phi of epithelial cell protein activity profiles for each cell that putatively represented the changes in their protein activity that occur as they move through the proximal-distal transition zone. When cells are sorted based on their phi value, they represent a gradual transition trajectory from the Distal cluster to the proximal cluster. To infer gene expression/protein activity that are upregulated only in the transition between the distal and proximal cluster, the PISCES-inferred Interactome was correlated with the hyperbolic transformation of phi: phi’ (phi.prime). Through this method, proteins and genes with high correlation with the hyperbolic activity patterns would be ranked first with high activity in the transition zone, thereby generating the heatmaps in Figure S1 and allowing the discovery of putative relevant markers or regulators of the transitional state.

Velocity analyses

The cleaned Seurat object after clustering was converted to a.h5ad file for downstream Python analysis. Cellranger output files were used as input for Velocyto to derive the counts of unspliced and spliced reads in loom format. Next, the sample-wise loom files were combined, normalized and log transformed using scvelos (https://github.com/theislab/scvelo). To infer future states of individual cells we made use of spliced and unspliced information. We employed scVelo (https://github.com/theislab/scvelo). The latent time was calculated using scVelo based on dynamical model. The cluster marker genes expression trends along latent time were calculated using dynamical model with the scv.tl.latent_time(adata)function. Heatmaps of the selected marker genes expression trends were plotted using the scv.pl.heatmap() function.

Quantitative PCR analysis

RNA was extracted using QIAGEN RNeasy Mini Kit and cDNA was synthesized using the SuperScript IV First-Strand synthesis system (Thermo Fisher). Reactions were performed using Taq-Man Advanced Master Mix (Thermo Fisher #4444556) using b-Actin (for mouse samples) as internal control and a Step-One Plus Instrument (Applied Biosystems). DDCT method was used to calculate changes in expression levels.

Immunofluorescence staining, RNA in situ hybridization and confocal analysis

Embryonic and adult lungs and tracheas (human and mouse) were fixed in 4% paraformaldehyde in PBS at 4°C overnight. After washing in PBS, samples were processed for frozen or paraffin-embedding. Immunofluorescence (IF) was performed in tissue sections (6–8μm). Slides were subjected to heat-based antigen retrieval using a citrate-based buffer (Vector Labs, H-3300) for 8:30 min and blocked with 1% bovine serum albumin (Sigma) and 0.5% TritonX-100 (Sigma) for 1 hr at room temperature. Primary antibodies were incubated in 1% bovine serum albumin (Sigma) and 0.5% TritonX-100 at 4°C overnight. Sections were then washed with PBS and incubated with Alexa Fluor-conjugated secondary antibodies (1:300) and NucBlue Live Cell ReadyProbes Reagent (DAPI) (Life Technology) for 1 hr. After washing, samples were mounted with ProLong Gold antifade reagent (Life Technology). For IF staining in Figure 2E, Tyramide Signal Amplification was performed on frozen sections. Briefly, after antigen retrieval, slides were treated with 3% H₂O₂ at room temperature for 10 minutes, followed by primary antibody staining. The next day, corresponding HRP secondary antibodies were applied. The slides were then incubated with TSA-fluorescein (1:100, NEL701A001KT, Akoya Biosciences) in the provided amplification buffer.

The following secondary antibodies were used: donkey anti-rabbit (conjugated with Alexa Fluor 488, 568, 647); donkey anti-chicken (conjugated with Alexa Fluor 488); donkey anti-mouse (conjugated with Alexa Fluor 488, 568, 647); donkey anti-rat (conjugated with Alexa Fluor 488, 647); donkey anti-goat (conjugated with Alexa Fluor 488, 568, 647); HRP-conjugated donkey anti-mouse; HRP-conjugated donkey anti-rat; HRP-conjugated donkey anti-rabbit; All secondary antibodies were purchased from Thermo Fisher Scientific or Jackson ImmunoReseach.

For wholemount lung explant culture IF staining, lung explants were fixed on the transwell by removal of the media in the bottom chamber, followed by two PBS washes and addition of 1.5ml 4% PFA to the bottom chamber and left at 4°C overnight. The day after, lung explants were transferred to PCR tubes (one lung each), washed 3 times 5 minutes in PBS, then permeabilized and blocked, first in PBS + 0.5% TritonX100 (PBST) 3 times 10 minutes, then in 5% donkey serum + 0.5% TritonX100 in PBS, 2 times 1 hour. Explants were then incubated overnight at 4°C in primary antibodies on a rotator for 2 nights. The following day, explants were washed in 0.5% PBST and incubated in secondary antibody overnight at 4°C on a rotator. Explants were washed in PBS and mounted on a glass slide using ProLong Gold antifade reagent (Invitrogen).

Organoid cultures were washed 3 times for 5 minutes each to remove media, and then dissociated from Matrigel as above. They were fixed in 4% paraformaldehyde in PBS at room temperature for 15 min and then processed whole-mount staining. For IF staining, cultures were permeabilized and blocked by incubation in 5% donkey serum + 0.5% TritonX100 in PBS for 2 hours at room temperature. Cultures were then incubated overnight at 4°C on a rotator in primary antibodies, and the next day the cultures were washed in PBS 3 three times for 1 hour each. Cultures were incubated with secondary antibodies overnight in the dark at 4°C, washed 3 times 1 hour in PBS, and kept in PBS for storage and during imaging.

In situ hybridization was performed in tissue sections using a RNAscope probe designed for Fgfbp1 mouse gene (ACDBio) and experiments are performed according to the manufacturer’s instructions. Confocal microscopy was performed using a Zeiss LSM 710 and Leica Stellaris confocal microscope through 20×, or 40× lens.

QUANTIFICATION AND STATISTICAL ANALYSES

Quantification of transitional marker colocalization with Icam1+ cells in Bleomycin-injured lungs

Day 14 bleomycin-injured lung sections were co-stained with Icam1 and representative transitional state markers (Krt8, Sprr1a, Cdkn1a, and Lgals3). 15 to 30 Icam1+ cells per biological replicate were randomly sampled, and transitional state marker-positive cells double-labeled with Icam1 were manually counted in images (Zen 2.3 lite) obtained from three different animals (n=3 biological replicates). The percentage of transitional marker positive cells was calculated as the number of transitional markers+ Icam1+ cells per total number of Icam1+ cells and plotted as mean+SEM.

Fluorescence intensity profile analysis of transitional cell markers in the regenerating tracheal epithelium of injured mice

For quantification of fluorescence intensity, paired IF images from both uninjured and polidocanol-injured tracheas at 7dpi were taken using the same digital settings to quantify the fluorescence intensity of epithelial cells expressing transitional state markers. For each set of paired comparisons, ≥70 tracheal epithelial cells were randomly selected and precisely circled around the labeled cellular shape. The average fluorescence intensity for each cell was measured using Zen 2.3 lite software and analyzed for statistical significance. At least three different biological replicates per group of animals were analyzed for each marker.

Quantitative assessment of ICAM1 mean fluorescence intensity (MFI) at different stage of human fetal lungs

Confocal images of human fetal lung sections stained with SOX9, SOX2 and ICAM1 at 10-, 11.5-, 13-, 15- and 18-week gestation encompassing both the epithelial and mesenchymal compartments were taken using the same digital settings for comparisons. To quantify the changes in ICAM1 fluorescence intensity in the epithelium, ICAM1+ epithelial cells and ICAM1+ endothelial cells were randomly selected and carefully circled around the labeled cellular shape. The mean fluorescence intensity for each cell was measured using Zen 2.3 lite software. Fold change of ICAM1 fluorescence intensity was calculated as the average epithelial MFI over the average endothelial MFI at each stage. Values obtained for each timepoint were plotted and compared over time to reflect the relative changes in ICAM1 epithelial expression. At least 3 sections from human fetal lungs at each stage were used for the analysis.

Analysis of transitional marker colocalization with airway cell types in regenerating tracheal epithelium

To count the number of Icam1+ or Lgals3+ tracheal epithelium cells expressing specific airway markers, tracheal sections were stained with representative markers including p63 (basal), Krt13 (hillock), Foxj1 (ciliated), and Scgb1a1 (secretory). Between 14 to 33 tracheal epithelium cells per section that were either Icam1 or Lgals3+ were randomly selected across three biological replicates. The percentage of cells expressing each airway marker was calculated as the number of cell type marker+ cells per total Icam1 or Lgals3+ cells.

Morphometric analysis of Fgfbp1-CreERT2; R26-tdTomato lineage-labeled cells in embryonic and neonatal lungs

To quantify the lineage fate of tdTom-labeled cells giving rise to either alveolar or airway cell types, sections were co-stained with tdTomato/RFP along with representative markers of AT1 or AT2 alveolar cell types (Hopx, Sftpc) and airway secretory (Scgb1a1, Scgb3a2) or ciliated (Foxj1, acetylated aTub) cells. Cells expressing any of the representative cell type markers that were lineage-labeled were counted manually, with 18–93 clones counted per cell type marker. Percentages were calculated as the number of cell type marker+ cells per total tdTom cells. Cells positive for alveolar markers (Sftpc, Hopx) are categorized as alveolar-fated cells, and cells positive for airway markers (Scgb1a1, Scgb3a2, Foxj1, acetylated aTub) were categorized as airway-fated. The percentage of either alveolar or airway fate was calculated per total tdTom cells. This quantitative analysis was performed in at least five different lung sections from one PN5 neonate that was exposed to TMX at E13.5 and four different E18.5 embryos that were exposed to TMX at E11.5 and E13.5. Data were expressed as mean±SEM percentage tdTom per compartment (alveolar or airway) or tdTom doubled labeled with each marker.

Quantification of organoid colony-forming efficiency

To quantify organoid colony-forming efficiency in LPM3D, the total number of organoids per well from independent cultures isolated from three biological replicates was counted. Brightfield images of organoids were captured using an EVOS M5000 microscope at 4× magnification. The Cellpose package⁷² on ImageJ was used to quantify organoid size (area covered) and the number of organoids. For MTEC culture organoids, brightfield images were recorded in the same manner, with organoids manually counted and the diameter of all organoids in a random field per well measured using Fiji ImageJ. Individual organoid sizes were calculated based on their diameters as circular areas. All experiments were conducted using independent cultures from three biological replicates

For all experiments, statistical analyses were performed in Microsoft Excel or GraphPad Prism software using Student’s t test for comparisons between groups, as indicated in the legends and main text.

KEY RESOURCES TABLE

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
goat anti-Icam1	R&D	AF796;RRID: AB_2248703
rabbit anti-ICAM1	Atlas	HPA002126;RRID: AB_1078470
rabbit anti-Trp63a	CST	13109s;RRID: AB_2637091
chicken anti-Krt5	Biolegend	905901;RRID: AB_2565054
rat Epcam-APC antibody	eBioscience	#17-5791-82;RRID: AB_2716944
rat anti-CD54-PE	BioLegend	116108;RRID: AB_313699;
rabbit anti-Hopx	Santa Cruz	SC-30216;RRID: AB_2120833
rabbit anti-Nkx2.1	Seven Hills	WRAB-TTF1;RRID: AB_451727
rat anti-Scgb3a2	R&D Systems	MAB3465;RRID: AB_2183548
goat anti-Scgb1a1	Santa Cruz	SC-9772;RRID: AB_2238819
rabbit anti-Ace-tubulin	CST	5335S;RRID: AB_10544694
mouse anti-Foxj1	Invitrogen	14-9965-82;RRID: AB_1548835
rabbit anti-Areg	Proteintech	66433-1;RRID: AB_2881803
rat anti-Sox2	eBioscience	14-9811-82;RRID: AB_11219471
goat anti-Sox9	R&D systems	AF3075;RRID: AB_2194160
rabbit anti-Pro-SPC	Seven Hills	WRAB-76694;RRID: AB_2938817
rabbit anti-Sprr1a	Gift from Dr. Anton Jetten, NIEHS	None
rat anti-Lgals3	Cedarlane	CL8942AP;RRID: AB_10060357
goat anti-Trop-2/Tacstd2	R&D	AF1122;RRID: AB_2205662
rabbit anti-TROP2	Atlas	HPA055067;RRID: AB_2682687
mouse anti-p21 (Cdkn1a)	Santa Cruz	sc-6246;RRID: AB_628073
rabbit anti-HB-EGF	Bioss	bs-3576R;RRID: AB_10858006
rabbit anti-Sox9 Antibody	Fisher Millipore	AB5535;RRID: AB_2239761
rat anti-Krt8	DSHB	Troma-1;RRID: AB_2891089
rabbit anti-Krt17 antibody	Abcam	ab53707;RRID: AB_869865
goat anti-AGER	R&D	AF1145;RRID: AB_354628
goat anti-SCGB3A2	R&D	AF3545;RRID: AB_2183543
goat anti-F3	R&D	AF3178;RRID: AB_2278143
goat Anti-14-3-3 sigma/SFN antibody	Abcam	ab77187;RRID: AB_1523039
rat anti-MKI67	eBioscience	14-5698-82;RRID: AB_10854564
mouse anti-MKI67	BD Biosciences	550609;RRID: AB_393778
mouse anti-PCNA	Novusbio	NB500-106SS;RRID: AB_10003640
rabbit anti-RFP Pre-adsorbed	Rockland	600-401-379;RRID: AB_2209751
goat anti-Tdtomato	Biorbyt	orb182397;RRID: AB_2687917
Biological samples
Primary human lung samples	Birth Defects Research Laboratory at the University of Washington Seattle	N/A
Primary human lung samples	Gift from Dr. Josh Sonett, Columbia University Medical Campus	N/A
Chemicals, peptides, and recombinant proteins
Tamoxifen	Sigma	Cat#T5648;
Sunflower seed oil	Sigma	Cat#S5007
4-OHT	Sigma	Cat#H7904
Nutlin3a	R&D	3984/10
SR11302	MCE	HY-15870
T-5224	MCE	HY-12270
BGJb	Gibco	12591038
PD173074	Tocris	3044
SB431542	Tocris	1614
Advanced DMEM/F12	Thermofisher	12634010
Fgf10	R&D	6224-FG-025
Fgf9	R&D	7399-F9-025
Egf	R&D	2028-EG-200
CHIR99021	Tocris	4423
BIRB796	Tocris	5989
Y27632	Tocris	1254
A8301	Tocris	2939
Heparin	Sigma	H3149
Insulin	Roche	11376497001
Transferrin	Roche	10652202001
Pen/Strep	Thermofisher	15140122
Glutamine	Thermofisher	25030081
Anti-Anti	Thermofisher	15240062
Collagenase A	Roche	10103578001
Dispase II	Stemcell Technologies	7913
DNase I	Sigma	DN25
Hanks’ Balanced Salt Solution (HBSS)	Gibco	14025-092
Red Blood Cell Lysis Buffer	Sigma	R7757
Cholera toxin	Sigma	C8052
Bovine pituitary extract	Gibco	13-028-014
Fetal bovine serum	Gibco	26140079
Retinoic acid	Sigma	R2625
Dispase	Corning	354235
Collagenase type I	Gibco	17100-017
TritonX-100	Sigma	T8787
32% Paraformaldehyde	Electron Microscopy Sciences	50-980-495
Critical commercial assays
SuperScript IV First-Strand synthesis system	Thermo Fisher	Cat# 18091050
RNAScope ® Multiplex Fluorescent Reagent Kit v2	Advanced Cell Diagnostics	Cat#323100
RNA-Protein Co-Detection Ancillary Kit	Advanced Cell Diagnostics	Cat#323180
Deposited data
scRNAseq data: E14.5 Icam1+ Epcam+ transition zone clel	This paper	GEO: GSE254356
scRNAseq data: E12, E15, E17 mouse embryonic lung	Frank et al.9	GSE149563
scRNAseq data: Bleomycin injured lung	Choi et al.41	GSE145031
scRNAseq data: 11.5w, 15w and 18w human fetal distal lung	Miller et al.64	E-MTAB-8221
scRNAseq data: 5-22w human fetal lung	He et al.72	E-MTAB-11278
Experimental models: Organisms/strains
Sox9-IRES-EGFP	Jackson lab	JAX Strain #030137
Fgfbp1-CreERT2; R26-tdTomato	Capdevila et al.43	N/A
Oligonucleotides
RNAscope ® Mm Fgfbp1 probe	Advanced Cell Diagnostics	Cat#508831-C4
Software and algorithms
Prism	GraphPad	https://www.graphpad.com
FlowJo	Becton Dickinson & Company	https://flowjo.com/solutions/flowjo
ImageJ	National Institutes of Health	https://imagej.nih.gov/ij/
PISCES (Protein activity Inference in Single Cells)	bioRxiv 445002	N/A
Leica Application Suite X (LAS X)	Leica Microsystems	https://www.leica-microsystems.com/products/microscope-software/p/leica-las-x-ls/
Zen blue 3.5 (ZEISS ZEN lite)	Carl Zeiss Microscopy	https://www.zeiss.com/microscopy/en/products/software/zeiss-zen-lite.html
R v4.0.3	R Core Team	https://www.r-project.org/
Seurat v4.2.0	Satija Lab	https://satijalab.org/seurat/
ggplot2 v3.3.6	N/A	https://github.com/tidyverse/ggplot2
scVelo v0.2.2	Bergen et al.35	https://scvelo.readthedocs.io/
velocyto v.0.17	Kharchenko Lab	https://velocyto.org/velocyto.py/
Python v3.7.7	Python Software Foundation	https://python.org/
Other
mouse Icam1	Thermofisher	Mm00516023_m1
mouse Sox9	Thermofisher	Mm00448840_m1
mouse Sox2	Thermofisher	Mm03053810_s1
mouse Lgals3	Thermofisher	Mm00802901_m1
mouse Sprr1a	Thermofisher	Mm01962902_s1
mouse Cdkn1a	Thermofisher	Mm04205640_g1
mouse Mki67	Thermofisher	Mm01278617_m1
mouse Sftpc	Thermofisher	Mm00488144_m1
mouse Hopx	Thermofisher	Mm00558630_m1
mouse Scgb3a2	Thermofisher	Mm00504412_m1
mouse Krt8	Thermofisher	Mm04209403_g1
mouse F3	Thermofisher	Mm00438853_m1
mouse Hbegf	Thermofisher	Mm00439306_m1
mouse Fgfbp1	Thermofisher	Mm00456064_s1
mouse Plaur	Thermofisher	Mm01149438_m1

Highlights.

• ICAM1/NKX2-1 identifies a transitional state in fetal and regenerating lungs

• ICAM1/NKX2-1 are plastic multipotent progenitors for alveolar and airway cell types

• The transitional cell program provides competency to respond to morphogenetic cues

• Human and mouse developing lungs share a highly conserved transitional state program