CEBPA restricts alveolar type 2 cell plasticity during development and injury-repair – Nature Communications

Postnatal transcriptomic and epigenomic maturation of AT2 cells is separate from their embryonic specification

To explore how cell plasticity can be shaped by development, we first delineated the associated molecular progression of AT2 cells using scRNA-seq and scATAC-seq. Mouse lung epithelial cells profiled by scRNA-seq at 12-time points spanning from embryonic day (E) 14.5 to 15-week adult stages, clustered by time and cell types (Fig. 1A, Supplementary Fig. 1). AT2 cells clustered separately from AT1 cells only by E18.5, indicating measurable specification from their E14.5 and E16.5 SOX9 progenitors (Fig. 1A). Nascent AT2 cells at E18.5 were distinct from their neonatal and adult counterparts, suggesting maturation following specification, which was supported by Monocle trajectory analysis (Fig. 1B). Eighty-two genes specific to 15-week AT2 cells, in comparison to 15-week AT1 cells, were designated as early versus late AT2 genes based on their presence versus absence of expression in >25% of E16.5 SOX9 progenitors7 (Fig. 1C). Ten out of the 41 late AT2 genes, including immune response genes (Lyz2, Lrg1, Chil1, and H2 paralogs), only reached maximal expression (>4-fold increase from E18.5 to 15-week) during postnatal maturation (Fig. 1C).

Fig. 1: AT2 cells undergo transcriptomic and epigenomic maturation postnatally, separate from their embryonic specification.
figure 1

A Aggregated scRNA-seq UMAPs of alveolar epithelial cells of wild type lungs from 12-time points color coded by time (top) or cell type (bottom). Cell numbers are in parenthesis. Each time point consists of at least 2 mice profiled as one sample. B Top: UMAP of AT2 and SOX9 progenitor cells subset from in (A), color-coded by time, and their monocle pseudotime analysis (bottom) showing molecular progression from E14.5 SOX9 progenitors through E18.5 nascent AT2 to 15-wk mature AT2 cells. Cell numbers are in parenthesis. C Expression heatmap of 82 AT2-specific genes, classified as early if present in at least 25% of cells at E16.5. The remaining late genes are considered mature if the fold increase from E18.5 to 15-wk is more than 4. D Principal component analysis (PCA) of scATAC-seq pseudobulk duplicates showing the distinct epigenome of E18.5 nascent AT2 cells. ScATAC-seq heatmaps and profile plots were categorized and color-coded as early lost (E16.5 vs E18.5), late lost (E18.5 vs 9-wk), early gained (E16.5 vs E18.5) and late gained (E18.5 vs 9-wk) and their Homer motif analysis. Peak numbers are in parenthesis. Each time point consists of at least 2 mice profiled as one sample.

Supporting this transcriptional post-specification maturation, scATAC-seq sampling of 5 developmental time points from E16.5 to 9-week adult revealed epigenomic maturation of AT2 cells that occurred after their E18.5 specification (PCA plot in Fig. 1D). Accordingly, we classified differential ATAC peaks into lost and gained groups, each with early and late subgroups separated at the E18.5 time point of specification (Fig. 1D). Although ATAC peaks often correlate with RNA transcripts and may act over long-range and on high-order chromatin structure, they measure individual regulatory regions without averaging over the whole gene, predict regulatory TF motifs, and bypass the confounding issue of perdurance in RNA-seq. As shown in Fig. 1D, the early lost peaks coincided with AT2 specification at E18.5, were near progenitor and AT1 genes (e.g., Adamts18 and Clic5, respectively; see, Supplementary Fig. 2 for representative genomic snapshots; same below), and contained SOX and TEAD motifs, likely reflecting the termination of the SOX9-mediated progenitor program and the low level of YAP/TAZ/TEAD-mediated AT1 program in E16.5 progenitors7,8. The late lost peaks decreased postnatally, were near stem cell genes (e.g., Klf4, Id4, Etv6, and Hif3a)1,9,10,11, and contained FOXA and NKX motifs, correlating with the postnatal decrease in AT2 cell proliferation and potential progenitor-specific functions of FOXA2 and NKX2-1. Conversely, the majority of differential peaks (19,258 out of 30,410; 63%) were in the early gained subgroup and were near AT2 genes (e.g., Lyz1) and enriched for CEBP and NKX motifs, which were examined in detail in this study. Last, the late gained peaks followed the RNA kinetics of AT2 cell maturation (Fig. 1C), were near the corresponding genes (e.g., H2-Aa and Cd74), and contained the AP-1 motif. Supporting the epigenomic distinction between specification and maturation, each of the 4 groups was associated with distinct biological pathways and unsupervised clustering of top 100 variable motifs predicted from scATAC-seq revealed lost, early gained, and late gained groups, dominated by SOX and TEAD, CEBP, and AP-1 motifs, respectively (Supplementary Fig. 3A, B).

We reasoned that, while the gain in chromatin accessibility reflected AT2 cell differentiation, the concurrent loss in accessibility not only indicated the conclusion of prior cell fates, but also predicted available fates when AT2 cells became more plastic. Accordingly, TFs promoting AT2 cell differentiation might also restrict AT2 cell plasticity. Of particular interest was CEBPA, whose motif was enriched among the early gained peaks, as well as our published AT2-specific NKX2-1 ChIP-seq peaks8 (Fig. 1D). Compared to other CEBP family members, Cebpa was specific to AT2 cells and reached maximal expression coinciding with AT2 specification (Supplementary Fig. 3C,   D). Supporting this, CEBPA was absent on the protein level at E14.5 when branch tips consisted of SOX9+ progenitors and had weak diffuse expression in occasional cells at E16.5, likely corresponding to spatially asynchronous onset of alveolar differentiation and consistent with prior reports12,13; from E18.5 on, CEBPA was expressed in a subset of epithelial cells that were cuboidal, SOX9- HOPX- LAMP3 + , and thus nascent AT2 cells (Fig. 2A, Supplementary Fig. 3E F). Taken together, our time-course transcriptomic and epigenomic roadmap of AT2 cell development expand on published datasets8,14,15, highlighting the sequential specification and maturation of AT2 cells and implicating CEBPA in their differentiation and plasticity.

Fig. 2: CEBPA promotes AT2 and suppresses progenitor programs in neonatal AT2 cells.
figure 2

A Confocal images of immunostained wild type lungs show little CEBPA expression at E14.5 and E16.5 when branch tips are dominated by SOX9 progenitors but CEBPA expression in cuboidal cells outlined with E-Cadherin (ECAD) at E18.5 (n = 3 mice each). B Confocal images of immunostained neonatal AT2-specific Cebpa mutant and littermate control lungs showing loss of CEBPA in GFP+ recombined cells (asterisk: escaper), without affecting its expression in alveolar macrophages (AM) in the airspace, and reduced LAMP3. Tam, 250 μg tamoxifen. Images are representative of at least three lungs (same for subsequent immunostainings). C Transmission electron microscopy (TEM) images show a reduction in lamellar bodies in mutant AT2 cells without affecting their apical microvilli (n = 2 mice each). Tam, 250 μg tamoxifen. See Supplementary Fig. 4B for more examples and quantification. D Confocal images showing lineage labeled mutant AT2 cells expressing an AT1 marker HOPX and no longer cuboidal (ECAD outline) (arrowhead). E Confocal images showing lineage labeled mutant AT2 cells ectopically expressing a progenitor marker SOX9 and a proliferation marker KI67. F Quantification of (D) and (E). Each symbol represents one mouse from littermate pairs. P values were calculated using two-tailed Student’s t test. Scale: 10 μm for all except for (C) 1 μm.

CEBPA promotes AT2 and suppresses progenitor programs in neonatal AT2 cells

Although CEBPA had been shown to promote both AT1 and AT2 cell differentiation in embryonic lungs16, its role in subsequent AT2 cell maturation and plasticity was unclear. Accordingly, we generated an inducible AT2-specific knockout model CebpaF/F; SftpcCreER/+; RosaSun1GFP/+. To target AT2 cells shortly after specification, we induced Cre-recombination at a neonatal stage (P2) with an efficiency of 99% and specificity of 96% (1310 GFP+ cells from 3 mice) and deleted CEBPA in AT2 cells with an efficiency of 88% (1951 GFP+ cells from 3 mice), without affecting its normal expression in alveolar macrophages (Fig. 2B). By P9, Cebpa mutant AT2 cells had a drastic decrease in LAMP3 and a loss of IL33, both AT2 markers, compared to adjacent escapers of Cre-recombination or AT2 cells in the littermate control (Fig. 2B, Supplementary Fig. 4A). Transmission electron microscopy showed that cuboidal/columnar epithelial cells in the Cebpa mutant lung often lacked lamellar bodies, a defining feature of AT2 cells, but still had characteristic apical microvilli (Fig. 2C, Supplementary Fig. 4B).

Our previous study of the CEBPA equivalent in AT1 cells, YAP/TAZ/TEAD, showed activation of the alternative alveolar program8 and thus prompted us to examine the AT1 program in Cebpa mutant AT2 cells. To our surprise, only a small fraction of mutant AT2 cells expressed an AT1 marker HOPX (12.5% of 2049 GFP+ cells from 3 mice), lost LAMP3, and were no longer cuboidal as outlined by E-Cadherin, compared to a baseline of 1.2% in the control (1353 GFP+ cells from 3 mice) that possibly resulted from driver non-specificity/leakiness and normal neonatal conversion of AT2 to AT1 cells (Fig. 2D). Instead, we noticed adjoining Cebpa mutant cells reminiscent of SOX9 progenitors at embryonic branch tips (Supplementary Fig. 4C). Remarkably, 80% of GFP+ cells in the mutant lung (1202 GFP+ cells from 3 mice) expressed SOX9, compared to 0.7% in the control (1370 GFP+ cells from 3 mice) (Fig. 2E). Like SOX9 progenitors, Cebpa mutant cells were also much more proliferative (KI67+ in 25% 3414 GFP+ cells from 3 mice, compared to 6% 2737 GFP+ cells from 3 mice in the control), likely contributing to their clustering (Fig. 2E, F). Therefore, without CEBPA, neonatal AT2 cells reduce their AT2 program and gain plasticity toward SOX9 progenitors and, to a much lesser extent, AT1 cells.

Single-cell multiome defines CEBPA-dependent neonatal AT2 cell program and plasticity

To fully characterize the CEBPA-dependent changes, we FACS-purified E-Cadherin+ epithelial cells from our neonatal Cebpa mutant and littermate control lungs and performed single-cell multiome for concurrent profiling of their transcriptomes and epigenomes (Supplementary Fig. 5A). On the combined single-cell UMAPs, club, ciliated, and AT1 cells from control and mutant lungs formed superimposed clusters, suggesting minimal changes, as they were not targeted by SftpcCreER and lacked the GFP transcript from RosaSun1GFP (Fig. 3A, B, Supplementary Fig. 5B). In contrast, while 3.7% (267 out of 7047 cells) AT2 cells from the mutant lung were intermixed with those from the control and still expressed Cebpa, consistent with them being escapers of deletion, the rest formed a separate Cebpa cluster (mutant AT2), as well as a proliferative cluster mainly made of cells from the mutant lung as predicted by KI67 immunostaining (Fig. 3A, B). A marker Lamp3 and a 100-gene signature score for AT2 cells were reduced in mutant AT2 cells, whereas a marker Sox9 and a 119-gene signature score for SOX9 progenitors were increased (Fig. 3C). The mutant AT2 cluster extended toward the AT1 cell cluster, forming a bridge that was GFP+ and thus descendant of recombined AT2 cells (Fig. 3C). This bridging population was specific to the mutant lung and expressed a marker Hopx and a 100-gene signature for AT1 cells, but still clustered separately from normal AT1 cells, possibly due to their AT2 cell origin and limited time for AT1 differentiation after Cebpa deletion. RNA velocity analysis confirmed this predicted trajectory bridging AT2 and AT1 cells specifically in the mutant (Supplementary Fig. 5B). Considering the observed HOPX immunostaining in LAMP3- non-cuboidal cells (Fig. 2D, Supplementary Fig. 5C), we named this bridging population HOPX + AT1-like cells. Differential expression analysis of Cebpa mutant versus control AT2 cells confirmed downregulation of AT2 genes (e.g., Lyz2, Lyz1, Sftpb, and Il33) and upregulation of progenitor (e.g., Sox9, Clu, and Col18a1) and AT1 genes (e.g., Akap5, Fbln5, and Rtkn2), although some surfactant genes including Sftpc were less reduced – unlike their near-absence upon embryonic Cebpa deletion16, possibly due to RNA perdurance or redundant transcriptional activation (Fig. 3D, Supplementary Fig. 5DE). Relatedly, opposite to the phenotypes reported here and given the AT2-restricted expression of CEBPA (Fig. 2A, Supplementary Fig. 3EF), the defective AT1 cell differentiation in the pan-epithelial embryonic Cebpa mutant16 was likely non-cell autonomous or due to potential toxicity associated with the Cre driver17.

Fig. 3: Single-cell multiome defines CEBPA-dependent neonatal AT2 cell program and plasticity.
figure 3

A Sc-multiome UMAPs of purified epithelial cells from Cebpa mutant and littermate control lungs color-coded by cell type (left) and the corresponding percentages (right). Esc, escaper; prolif, proliferating; Tam, 250 μg tamoxifen. See Supplementary Fig. 5A for the sorting strategy. Each sample includes 1 male and 1 female mouse profiled as one sample (same for subsequent sc-multiome experiments). B Dot plot showing the lineage marker (Sun1GFP), Cebpa, and cell type markers. See also feature plots in (C) and Supplementary Fig. 5B. C Sc-multiome UMAP color-coded for genotype (left) and feature plots of metagene scores (top) and representative genes (bottom). The circled population is specific to the mutant, GFP + , and expresses AT1 genes, thus labeled as HOPX + AT1-like cells in (A). See Source Data for metagene lists. D Volcano plot (two-tailed, non-parametric Wilcoxon rank sum test) showing downregulation of AT2 genes and upregulation of progenitor genes in mutant AT2 cells (right) compared to control AT2 cells (left) defined in (A). E Scatter plot correlating changes in the accessibility of scATAC-seq peaks (y-axis) and scRNA-seq expression of their nearest genes (x-axis), color-coded as concordant or discordant as well as the directionality of change. F ScATAC-seq heatmaps and profile plots of decreased and increased peak sets in the mutant and associated log2 fold changes, as well as the corresponding scATAC-seq data in wild type cells and associated Homer motifs. G Feature plots of motif activity scores showing that the mutant has lower CEBP, and higher SOX and TEAD (circle) activities.

To explore the epigenetic mechanism of the transcriptional changes, we performed differential accessibility analysis of Cebpa mutant versus control AT2 cells as pseudobulks and identified 10,621 differential peaks. Assigning each peak to its nearest gene, we found a high concordance (71%) between peaks and gene expression, with increases in both (39%) for progenitor genes including Sox9, Bspry, and Adamts18 and decreases in both (32%) for AT2 genes including Lyz1, Il33, and S100g (Fig. 3E). Moreover, the 5,287 decreased peaks were normally more accessible in AT2 cells, in comparison to AT1 and SOX9 progenitor cells, and were enriched for CEBP and NKX motifs, consistent with Cebpa deletion (Fig. 3F, G). The 5334 increased peaks were normally more accessible in SOX9 progenitor or, to a lesser extent as expected from the small number of HOPX + AT1-like cells, AT1 cells, in comparison to AT2 cells, and were enriched for SOX and TEAD motifs, consistent with activation of progenitor and AT1 programs (Fig. 3F, G). Therefore, besides the more predictable role of CEBPA in promoting the AT2 program, marker and whole-genome analyses unexpectedly show that neonatal AT2 cells have the plasticity to revert to SOX9 progenitors when unconstrained by CEBPA. The 5334 increased peaks represent CEBPA-dependent plasticity of neonatal AT2 cells that will be explored later.

CEBPA recruits NKX2-1 to promote the AT2 program and indirectly restricts the progenitor program

The identification of differential accessibility peaks that were largely concordant with differential gene expression prompted mechanistic analysis to link ATAC peaks to CEBPA chromatin binding. Given the enrichment in CEBP and NKX motifs (Fig. 3F) and the normal expression of NKX2-1 in the Cebpa mutant (Supplementary Fig. 5F), we used our published cell-type-specific ChIP-seq protocol8 to perform CEBPA ChIP-seq on control AT2 cells at P2, the time of Cre-recombination, as well as NKX2-1 ChIP-seq on AT2 cells from control and Cebpa mutant lungs at P8 (Supplementary Fig. 6A, Fig. 4A). The 5287 decreased peaks, as exemplified by a peak near an AT2 gene Il33, were bound by CEBPA and NKX2-1; NKX2-1 binding decreased upon Cebpa deletion, suggesting recruitment of NKX2-1 by CEBPA. NKX2-1 binding at these sites was specific to AT2 cells but not E14.5 SOX9 progenitors despite NKX2-1 expression in both, suggesting that CEBPA was required to acquire AT2-specific NKX2-1 binding (Fig. 4A, B). The predicted CEBP and NKX motifs, but not SOX motif, were concentrated at the center of the decreased peaks (Fig. 4A). Furthermore, the average distance between CEBPA and NKX2-1 binding sites were 54 bp, consistent with proximity or even direct binding between CEBPA and NKX2-1, although such protein-protein interaction needed technically challenging biochemical studies of purified AT2 cells. The parallel changes in both chromatin accessibility and NKX2-1 binding during their developmental gain from progenitors to AT2 cells and loss upon Cebpa deletion (Figs. 3F, 4A diagram) does not establish the sequence of events nor rule out the possibility that despite the presence of adjacent CEBPA and NKX2-1 motifs, decreased NKX2-1 binding to these sites was secondary to chromatin closure due to some impact of CEBPA deletion elsewhere in the genome, calling for future systematic deletion of CEBPA binding sites and/or interference of CEBPA binding to them. Relatedly, the recruitment model (Fig. 4A diagram) highlights the new chromatin binding specificity of NKX2-1 conferred by CEBPA, but does not require sequential binding of CEBPA and then NKX2-1 to the chromatin.

Fig. 4: CEBPA recruits NKX2-1 to promote the AT2 program and indirectly restricts the progenitor program.
figure 4

A Heatmaps and profile plots of CEBPA and NKX2-1 binding for decreased and increased peak sets from Fig. 3F, as well as associated frequency distributions of CEBP, NKX, SOX motifs. CEBPA binds to decreased peaks but not increased peaks. NKX2-1 binding decreases (log2 fold change) for decreased peaks and increases (log2 fold change) for increased peaks in the mutant, corresponding to AT2 and progenitor/AT1-specific binding in wild type lungs. Diagram: a recruitment model, in which CEBPA normally recruits NKX2-1 to activate AT2 genes, whereas without CEBPA, NKX2-1 is released from AT2 genes and possibly relocates to progenitor and AT1 genes. Loss of NKX2-1 binding due to Cebpa deletion is associated with chromatin closure (open vs closed). See Supplementary Fig. 6A for nuclei sorting strategy. B Representative coverage plots of (A) showing a decreased peak near an AT2 gene Il33, and an increased peak near a progenitor gene Acaca. C Venn diagram showing NKX2-1 and CEBPA co-bound and single-bound peak sets in purified P2 AT2 cells (left) and frequency distributions of NKX and CEBP motifs for the co-bound peak set (right). D Heatmaps and profile plots for the 3 peak sets in (C) and associated log2 fold changes showing the largest decreases for the co-bound peak set.

In contrast, the 5334 increased peaks, as exemplified by a peak near a progenitor gene Acaca, had little CEBPA binding and a slight increase in NKX2-1 binding upon Cebpa deletion, possibly attributable to its relocation to progenitor and AT1-specific sites without sequestration by CEBPA (Fig. 4A, B). Interestingly, NKX2-1 bound more at these sites in E14.5 SOX9 progenitors as well as AT1 cells compared to AT2 cells, suggesting that these progenitor-specific NKX2-1 binding sites normally lost NKX2-1 binding and were closed during alveolar differentiation, but were reopened upon Cebpa deletion (Fig. 4A). Alternatively, given the robust SOX9 expression and the prevalent SOX motif for these sites (Figs. 2, 3), CEBPA directly or indirectly repressed Sox9, which in turn initiated the progenitor program. Indeed, we identified a putative regulatory region 3’ to Sox9 that was open in the SOX9 progenitors and reopened in the Cebpa mutant, a profile mirrored by NKX2-1 binding (Supplementary Fig. 7).

The link between CEBPA/NKX2-1 chromatin binding and differential accessibility peaks was also examined in the reverse direction. CEBPA and NKX2-1 binding sites in AT2 cells were categorized as co-bound and single-bound for each TF (Fig. 4C). Compared to NKX2-1 single-bound sites, CEBPA/NKX2-1 co-bound sites, had a greater decrease in NKX2-1 binding and accessibility upon Cebpa deletion, reinforcing the said recruitment model and implicating other regulators of NKX2-1 binding and accessibility at the NKX2-1 single-bound sites (Fig. 4D). The CEBPA single-bound sites had limited accessibility and NKX2-1 binding as well as limited changes, suggesting a minor impact of CEBPA on its own (Fig. 4D). Taken together, in AT2 cells, CEBPA recruits NKX2-1 to promote the AT2 program, but does not bind to and thus indirectly represses sites that remain plastic in neonatal AT2 cells.

CEBPA maintains the AT2 program without affecting the progenitor program in mature AT2 cells

As the transcriptomic and epigenomic landscape of AT2 cells matured postnatally (Fig. 1), we posited that they would reinforce their gene regulatory network and exhibit less cell plasticity. To test this, we induced Cre-recombination in mature AT2 cells in >5-week old lungs and achieved 92% efficiency in deleting Cebpa (1919 GFP+ cells from 3 mice), again without affecting CEBPA expression in alveolar macrophages (Fig. 5A). As in the neonatal deletion model, mature Cebpa mutant AT2 cells downregulated LAMP3, lost IL33, and had fewer lamellar bodies (Fig. 5A, B, Supplementary Fig. 6BC). However, they did not express SOX9 or KI67, suggesting a loss of cell plasticity toward SOX9 progenitors (Fig. 5A).

Fig. 5: CEBPA maintains the AT2 program without affecting the progenitor program in mature AT2 cells.
figure 5

A Confocal images of immunostained adult AT2-specific Cebpa mutant and littermate control lungs showing loss of CEBPA in GFP+ recombined cells (asterisk: escaper), without affecting its expression in alveolar macrophages in the airspace (AM), and reduced LAMP3, but no extra HOPX, SOX9, or KI67. Tam, two doses of 3 mg each tamoxifen at 48 h interval (same for the rest of Fig. 5) (n = 3 mice each). Scale: 10 μm. B TEM images showing a reduction in lamellar bodies in mutant AT2 cells without affecting their apical microvilli. Large granules in mutant AT2 cells lack characteristic lamellae (n = 2 mice each). Scale: 1 μm. See Supplementary Fig. 6C for quantification. C Sc-multiome UMAPs of purified epithelial cells from Cebpa mutant and littermate control lungs color-coded by cell type (left), the corresponding percentages (middle), and metagene scores. Esc, escaper. See Source Data for metagene lists. D Dot plot showing the lineage marker (Sun1GFP), Cebpa, and cell type markers. Rtkn2, but not Spock2, is expressed in HOPXlow AT1-like cells. E Volcano plot (two-tailed, non-parametric Wilcoxon rank sum test) showing downregulation of AT2 genes but minimal upregulation of progenitor/AT1 genes in mutant AT2 cells (left) compared to control AT2 cells(right) defined in (C). Compare with Fig. 3D. F Scatter plot correlating changes in the accessibility of scATAC-seq peaks (y-axis) and scRNA-seq expression of their nearest genes (x-axis), color-coded as concordant or discordant as well as the directionality of change. Compared to Fig. 3E, few concordant pairs are upregulated. See Source Data for the complete list. G Heatmaps and profile plots of decreased and increased scATAC-seq peak sets in the adult mutant and associated log2 fold changes, as well as the corresponding CEBPA and NKX2-1 binding and scATAC-seq data in wild type cells and associated Homer motifs. Decreased peaks have CEBPA binding and decreased NKX2-1 binding, corresponding to ATAC accessibility and NKX2-1 binding in wild type AT2 cells. Increased peaks are many fewer and have no CEBPA binding and increased NKX2-1 binding, corresponding to NKX2-1 binding in wild type AT1 cells.

Single-cell multiome profiling of E-Cadherin+ epithelial cells from mature Cebpa mutant and littermate control lungs showed a transcriptional shift only in targeted GFP + AT2 cells (Fig. 5C, D). Notwithstanding additional heterogeneity including a Lyz1+ population in the control lung and a Meg3+ population in the mutant, possibly related to lung cancer and fibrosis18,19, the most prominent change was downregulation of AT2 genes in Cebpa mutant AT2 cells, without activating progenitor genes or forming a proliferative population as in the neonatal lungs (Fig. 5C). Compared to the control lung, the Cebpa mutant lung had a larger population clustered near AT2 cells and expressing some but not all AT1 gene transcripts (Fig. 5C, D), although few HOPX+ cells were detected by immunostaining (Fig. 5A). Accordingly, we considered this population HOPXlow AT1-like cells to indicate their limited AT1 differentiation (Fig. 5C). The reduction in the AT2 program, no increase in the progenitor program and limited increase in the AT1 program were supported by differential gene expression analysis (Fig. 5E).

Compared to the neonatal model, differential accessibility analysis of Cebpa mutant versus control mature AT2 cells identified 2619 decreased peaks but only 692 increased peaks, suggesting less CEBPA-dependent cell plasticity than neonatal AT2 cells. These differential peaks were still 76% concordant with gene expression (Fig. 5F). The decreased peaks were AT2-specific, enriched for CEBP and NKX motifs, had CEBPA and NKX2-1 binding in control AT2 cells but decreased NKX2-1 binding in Cebpa mutant AT2 cells, and had NKX2-1 binding in normal AT2 but not progenitor nor AT1 cells, supporting the same recruitment model of NKX2-1 by CEBPA in mature AT2 cells (Fig. 5G, Supplementary Figs. 6D, 4A diagram). The few increased peaks had no CEBPA binding, were enriched for NKX and TEAD motifs, but not SOX motif, and had some accessibility enriched for progenitors and AT1 cells but to a much lesser extent than the neonatal increased peaks (Fig. 5G, Supplementary Fig. 6D). NKX2-1 binding in purified control AT2 cells was low but increased in Cebpa mutant AT2 cells, possibly due to its limited redistribution to AT1-specific sites in the considerable number of HOPXlow AT1-like cells (Fig. 5G).

The main difference between the mature versus neonatal Cebpa deletion models was the inability of mature AT2 cells to reactivate the SOX9 progenitor program. This decrease in cell plasticity as AT2 cells matured were molecularly defined as the CEBPA-dependent, increased peaks unique to neonatal AT2 cells (5124 peaks in Supplementary Fig. 6E). The neonatal-specific plasticity was in regions that were accessible in progenitors and closed for 2 days versus 35 days when Cre-recombination was induced in neonatal versus mature lungs, respectively (Supplementary Fig. 6E). The duration of chromatin closure might lead to less reversible changes in histone modifications, DNA methylation, or high-order chromatin structure across the sites of differential plasticity or a few nodal sites of master genes, such as Sox9. Although CEBPA did not bind to these differentially plastic sites, its deletion revealed their presence.

To further define the temporal window of AT2 cell plasticity and potential regulators, we deleted Cebpa at additional neonatal time points and found robust SOX9 activation upon deletion at P4, but weak SOX9 at P7 and little SOX9 at P10 (Supplementary Fig. 6F). Accordingly, comparison of P4 and P10 scRNA-seq of AT2 cells revealed downregulated and upregulated genes that might promote and suppress plasticity, respectively (Supplementary Fig. 6G). Intriguingly and worthy of future investigation, Dlk1 was among the most downregulated at P10 and had been implicated in antagonizing Notch signaling and AT2 self-renewal during injury repair20.

Taken together, in neonatal and mature AT2 cells, CEBPA recruits NKX2-1 to promote and maintain the AT2 program; without CEBPA, neonatal but not mature AT2 cells have the plasticity to reactivate the SOX9 progenitor program.

Viral infection expands CEBPA-dependent plasticity in mature AT2 cells

The temporal restriction in cell plasticity from neonatal to mature AT2 cells reminded us of the doctrine that injury-repair recapitulates development and prompted us to test if respiratory virus infection would reactivate the neonatal plasticity in mature AT2 cells. We infected our mature Cebpa deletion model with Sendai virus, which was known to preferentially injure AT2 cells, forming AT2-less regions, and trigger AT2 cell proliferation 14 days post infection21. Strikingly, while the infected control lung repaired itself with no SOX9 expression and only isolated KI67 expression in AT2 cells, the infected Cebpa mutant lung had SOX9 expression in 8% of AT2 cells (7776 GFP+ cells from 3 mice) and clusters of KI67 + AT2 cells, reminiscent of the neonatal Cebpa mutant (Fig. 6A). SOX9 + AT2 cells often abutted lobe edges or airways and macro-vessels, topologically distal ends of the respiratory tree favoring de novo growth as we described21 (Supplementary Fig. 8A). The regional preference, in conjunction with localized virus delivery, suggested that the percentage of mutant AT2 cells capable of expressing SOX9 could be much higher. SOX9 activation depended on infection because saline treated control and Cebpa mutant lungs did not express SOX9 (Supplementary Fig. 8B).

Fig. 6: Viral infection expands CEBPA-dependent plasticity in mature AT2 cells. Figure 6. Viral infection expands CEBPA-dependent plasticity in mature AT2 cells.
figure 6

A Experimental timeline of tamoxifen injection (Tam, 3 mg), Sendai virus (SeV) or saline (PBS) administration, and lung harvest at 14 dpi (day post-infection). Confocal images of immunostained infected Cebpa mutant and littermate control lungs, showing mutant-specific activation of SOX9 and increase in KI67 near airways (aw) and lobe edges (inset; scale: 10 μm). Scale: 100 μm. B Confocal images of lungs in (A) showing increased KRT8 and CLDN4. Scale: 100 μm (inset: 10 μm). C Confocal images of lungs in (A) showing lineage-labeled HOPX+ cells with little LAMP3 (arrowhead). Scale: 10 μm. D Quantification of (A, B, C). KI67+ cells in the control and Cebpa mutant are stratified by CEBPA and SOX9 expression, respectively. Each symbol represents one mouse from littermate pairs. P values were calculated using two-tailed Student’s t test. E Sc-multiome UMAPs of purified epithelial cells from infected Cebpa mutant and littermate control lungs color-coded by cell type (left) and the corresponding percentages (right). Esc, escaper; prolif, proliferating. F Dot plot showing the lineage marker (Sun1GFP), Cebpa, and cell type markers. G Sc-multiome UMAP color-coded for genotype (left) and feature plots of metagene scores (top) and representative genes (bottom). A published damage-associated transient progenitor (DATP) score marks KRT8/CLDN4+ cells.

E-Cadherin+ epithelial cells from infected control and Cebpa mutant lungs were profiled with single-cell multiome (Fig. 6E, F). As in prior neonatal and mature mutant models, Cebpa– AT2 cells from the infected mutant lung clustered separately from escapers of deletion, as well as AT2 cells in the infected control lung. Progenitor score and genes including Sox9, Dlk1, and Kif4 were higher in the mutant, despite the said spatial restriction (Fig. 6G, Supplementary Fig. 9C). Proliferative AT2 cell cluster was much more prominent in the mutant, corroborating the KI67 immunostaining, and expressed Sox9, suggesting a possible coupling between proliferation and SOX9 activation (Fig. 6F, G). Supporting this, despite their low percentage, SOX9 + AT2 cells were more likely to be KI67+ than SOX9- AT2 cells (45% vs 3.5%; 696 SOX9+ cells out of 8562 GFP+ cells from 3 mice; Fig. 6D). By comparison, KI67 + AT2 cells in the infected control lung were equally likely to be CEBPA+ or CEBPA- (5.9% vs 6.0%; 233 CEBPA- cells out of 2642 GFP+ cells from 3 mice; Fig. 6D), suggesting that transition of control AT2 cells into other CEBPA- populations, as examined further below, was uncoupled from proliferation. Despite the transcriptional resemblance, Cebpa mutant AT2 cells were not bona fide SOX9 progenitors since without CEBPA, they were not expected to meet the functional definition of differentiation to AT2 cells and they existed in an adult, injured niche different from their native embryonic environment.

Sendai virus infection also induced in both control and mutant lungs two other GFP + AT2-derived cell populations: KRT8/CLDN4+ transitional cells and AT1-like cells, marked by respective gene signatures (Fig. 6G). By immunostaining, the former had high KRT8 and ectopic CLDN4, but low LAMP3 and no HOPX; the latter had HOPX but no LAMP3 and were no longer cuboidal (Supplementary Fig. 8CD). Although the two populations could represent sequential steps during AT2 to AT1 differentiation22,23,24,25, their locations on the UMAPs were also compatible with two parallel states with only the AT1-like cells transitioning to AT1 cells and KRT8/CLDN4+ cells being arrested. Regardless, the Cebpa mutant lungs had a dramatic expansion of KRT8/CLDN4+ transitional cells, as confirmed by immunostaining, which were distinct from SOX9-expressing cells (Fig. 6B, Supplementary Fig. 9A). Despite the higher number of AT1-like cells captured by single-cell multiome, they were not reliably detected by HOPX immunostaining, possibly due to the higher sensitivity of single-cell multiome in documenting the gradual AT2-AT1 transition (Fig. 6C, D, E). CEBPA was normally lost in both populations even in the control lung by single-cell profiling and immunostaining (Supplementary Fig. 9B, C), consistent with their decreased/lost LAMP3 expression and the described role of CEBPA in maintaining the AT2 program. Therefore, the population expansion in the mutant was likely because most AT2 cells became eligible for alternative fates as the result of losing CEBPA. Furthermore, as the loss of CEBPA in KRT8/CLDN4+ and AT1-like cells alone was insufficient to activate Sox9 in the infected control lung, SOX9 activation in the infected mutant represented a separate plasticity from injury-induced loss of the AT2 program and adoption of KRT8/CLDN4+ and AT1-like programs. Taken together, Sendai virus infection increases the plasticity of mature AT2 cells, which manifests upon Cebpa deletion as activation of the SOX9 progenitor program and expansion of the KRT8/CLDN4+ program.