Differential chromatin binding of the lung lineage transcription factor NKX2-1 resolves opposing murine alveolar cell fates in vivo

Abstract

Differential transcription of identical DNA sequences leads to distinct tissue lineages and then multiple cell types within a lineage, an epigenetic process central to progenitor and stem cell biology. The associated genome-wide changes, especially in native tissues, remain insufficiently understood, and are hereby addressed in the mouse lung, where the same lineage transcription factor NKX2-1 promotes the diametrically opposed alveolar type 1 (AT1) and AT2 cell fates. Here, we report that the cell-type-specific function of NKX2-1 is attributed to its differential chromatin binding that is acquired or retained during development in coordination with partner transcriptional factors. Loss of YAP/TAZ redirects NKX2-1 from its AT1-specific to AT2-specific binding sites, leading to transcriptionally exaggerated AT2 cells when deleted in progenitors or AT1-to-AT2 conversion when deleted after fate commitment. Nkx2-1 mutant AT1 and AT2 cells gain distinct chromatin accessible sites, including those specific to the opposite fate while adopting a gastrointestinal fate, suggesting an epigenetic plasticity unexpected from transcriptional changes. Our genomic analysis of single or purified cells, coupled with precision genetics, provides an epigenetic basis for alveolar cell fate and potential, and introduces an experimental benchmark for deciphering the in vivo function of lineage transcription factors.

Fig. 1. NKX2-1 binds chromatin in a cell-type-specific manner.

a Genetic labeling and FACS purification of AT1 and AT2 nuclei using Wnt3a^Cre and Sftpc^CreER, respectively. Confocal images of immunostained lungs showing that both AT1 and AT2 cells express NKX2-1 while AT2 cells are distinguishable by LAMP3 and cuboidal E-Cadherin (ECAD). Rosa^Sun1GFP marks the nuclear envelop (arrowhead; green in the diagram). All nuclei are Sytox Blue positive. 10-wk, 10-week-old. For Sftpc^CreER, 3 mg tamoxifen was administrated 3 days before tissue harvest. Scale: 10 µm. The color scheme (AT1: green; AT2: purple) applies to subsequent figures. b Validation of cell-type-specific ChIP-seq using an active promoter marker H3K4me3. Compared to the whole lung, the site (asterisk; same in subsequent figures) near an AT1 gene Spock2 shows enrichment in purified AT1 nuclei but depletion in purified AT2 nuclei, whereas the site near an AT2 gene Lamp3 shows the opposite pattern. A pan-epithelial gene Cdh1 (also known as E-Cadherin) is enriched in both AT1 and AT2 nuclei, whereas a mesenchymal gene Pdgfra is depleted in both. P, postnatal day. For Sftpc^CreER, 250 µg tamoxifen was administrated 3 days before tissue harvest. c ChIP-seq heatmaps (2.5 kb flanking the peak center; same in subsequent figures) of purified AT1 and AT2 nuclei for NKX2-1, H3K4me3 (active promoter), H3K27ac (putative active enhancer; note its bimodal pattern surrounding the center), and H3K4me1 (putative enhancer and active gene body), grouped by differential NKX2-1 binding into AT1-specific, AT2-specific, and common peaksets and sorted by the H3K4me3 signal. The common peakset is divided into lineage (epithelial) and housekeeping sets, based on scATAC-seq (Supplementary Fig. 2). Compared to the cell-type-specific and lineage sets, the housekeeping set has abundant H3K4me3, which also corresponds to depletion of H3K4me1 (top portion), whereas the bottom portion (curly bracket) is depleted of H3K4me3 and is enriched for the insulator CTCF motif. d Example NKX2-1 binding sites of the AT1-specific (Spock2), AT2-specific (Lamp3), lineage (Cdh1), and housekeeping (Gapdh) sets. The Y-axes are scaled via foreground normalization so that peak heights can be directly compared across samples; quantification and statistics are reported in the corresponding Supplementary Data files with marked peaks highlighted (same in subsequent figures).

Fig. 2. NKX2-1 is required for accessibility at cell-type-specific sites.

AT1-specific (a) and AT2-specific (b) deletion of Nkx2-1 using Rtkn2^CreER and Sftpc^CreER, respectively. Confocal images of immunostained mature lungs, showing efficient loss (filled versus open arrowhead) of NKX2-1 in recombined cells marked by GFP (a and b), as well as an AT1 marker (HOPX in a) and AT2 markers (LAMP3 and SFTPC in b). Note the increased ECAD staining in the AT1-specific Nkx2-1 mutant (a), consistent with our publication. Two doses (3 mg each; 48 h interval) of tamoxifen (Tam) were given and the lungs were harvested at 5 or 7 days after the initial dose. Scale: 10 µm. c ATAC-seq heatmaps of purified AT1 cells from AT1-specific Nkx2-1 mutant (as in a; abbreviated as NKX2-1^Rtkn2) and littermate control (AT1 C) lungs, and purified AT2 cells from AT2-specific Nkx2-1 mutant (as in b; abbreviated as NKX2-1^Sftpc) and littermate control (AT2 C) lungs, at the same NKX2-1 ChIP-seq peaksets organized as in Fig. 1. Both mutants are from mature lungs at 5 days after the initial tamoxifen administration as in a. AT1-specific and AT2-specific NKX2-1 peaksets have reduced accessibility in the corresponding cell type and cell-type-specific mutants; the lineage peakset has reduced accessibility in both mutants; whereas the housekeeping peakset is largely unaffected. d Example chromatin accessibility signals at NKX2-1 binding sites of the AT1-specific (Spock2), AT2-specific (Lamp3), lineage (Cdh1), and housekeeping (Gapdh) sets. Accessibility at sites near Cdh1 in AT2 cells is unaffected by Nkx2-1 deletion, possibly due to AT2-specific redundant mechanisms to maintain accessibility. An NKX2-1-bound lineage site ~30 kb upstream of Sftpb (black asterisk) has reduced accessibility in both Nkx2-1 mutants. Sftpb also has an AT2-specific NKX2-1 site at its promoter (purple asterisk) that has reduced accessibility specifically in the NKX2-1^Sftpc mutant.

Fig. 3. NKX2-1 establishes cell-type-specific binding via selectively acquiring de novo sites and retaining sites bound in progenitors.

NKX2-1 ChIP-seq heatmaps from E14.5 whole lungs, purified AT1 or AT2 nuclei from P7 or 10-week-old lungs, grouped by differential NKX2-1 binding in 10-week-old AT1 versus AT2 cells into AT1-specific (a) and AT2-specific (b) peaksets and sorted by differential NKX2-1 binding between E14.5 lungs and 10-week-old AT1 (a) or AT2 (b) cells. The top (acquired) and bottom (retained) 20% sites (boxed area) are averaged for the profile plots (dash represents the biological replicate) and the corresponding log2 fold changes from the E14.5 lungs. Acquired AT1-specific sites have little binding in progenitors, and increase over time in AT1 but not AT2 cells; whereas retained AT1-specific sites have high binding in progenitors, and remain high in AT1 but not AT2 cells (a). The same kinetics applies to AT2-specific sites (b). The small difference in the number of peaks from Fig. 1 is due to incorporation of E14.5 and P7 NKX2-1 ChIP-seq data. Example NKX2-1 binding sites of acquired (Pdpn) or retained (Hopx) AT1-specific sets (c) and acquired (Sftpb) or retained (Lamp3) AT2-specific sets (d). e NKX2-1 binding signal of the indicated four categories in a and b, showing the kinetics leading to differential binding at cell-type-specific sites as AT1 (left) versus AT2 (right) cells mature.

Fig. 4. Acquired and retained NKX2-1 sites have comparable kinetics in transcriptional divergence of progenitors toward opposing cell fates.

a ScRNA-seq UMAPs (Uniform Manifold Approximation and Projection) of color-coded E14.5 to 15-week-old lungs identify the indicated cell types and numbers in the alveolar region and show the gradual transcriptomic shifts of AT1 and AT2 cells from SOX9 progenitors, as supported by the dot plot for AT1 (Spock2), AT2 (Lamp3), progenitor (Sox9), and proliferative (Mki67) cells. Cells from a similar developmental stage (P4, P6, P7, and P8) are clustered together, suggesting the differences among shifted clusters are largely biological. b Monocle trajectory analysis of cells in a excluding proliferative AT2 cells, showing three branches (circled number) consistent with SOX9 progenitors differentiating into AT1 or AT2 cells, as supported by expression of the respective markers Sox9, Spock2, and Lamp3. The same Monocle trajectory in c split by sample time points and colored as in a. E14.5 and E16.5 cells are largely homogeneous and limited to branch 1 (progenitor branch). d Seurat module scores of gene sets associated with acquired or retained AT1 or AT2-specific NKX2-1 sites as defined in Fig. 3, plotted along the Monocle trajectories as numbered in c. e NKX2-1 binding signal of the four categories as in d in the AT1 or AT2 cell over time, using data from Fig. 3, showing that acquired and retained sites have distinct binding kinetics but comparable transcriptional kinetics.

Fig. 5. AT1-specific partner factors YAP/TAZ establish AT1-specific NKX2-1 binding and cell fate, and antagonize those of AT2 cells.

a Top enriched HOMER de novo motifs (binomial test) for AT1-specific (TEAD), common (FOXA and CTCF), AT2-specific (CEBP) NKX2-1 binding sites. b Nuclear YAP/TAZ and CEBPA are specifically detected (arrowhead) in AT1 and AT2 cells, respectively. Shh^Cre genetically labels AT1 and AT2 nuclei, distinguishable by AT2 cell-specific LAMP3 and cuboidal ECAD. Scale: 10 µm. c Confocal images of immunostained E18.5 Y/T^Sox9 mutant and littermate control lungs received 3 mg tamoxifen (Tam) at E15.5, showing loss of nuclear YAP/TAZ from AT1 cells, as outlined by ECAD (arrowhead). Scale: 10 µm. d NKX2-1 ChIP-seq heatmaps of E18.5 Y/T^Sox9 mutant and littermate control (C) whole lungs, sorted by fold change and cross-referenced with NKX2-1 binding from Fig. 1. e Examples of NKX2-1 peaksets in d. Lamp3 peaks do not reach statistical significance. NKX2-1 binding for Spock2 and Lamp3 in 10-week cells are shown in Fig. 1 (same in Fig. 6c). f ScRNA-seq UMAP comparison of Y/T^Sox9 mutant and littermate control epithelial cells with cell typing supported by the dot plot. The mutant has a decrease in the number of AT1 cells accompanied by an increase in the number of AT2 cells. g Volcano plots (MAST differential expression) comparing AT1 (left) and AT2 (right) cells in f. Differentially expressed, curated lists of AT1 and AT2-specific genes (Supplementary Data 6) are labeled (same in subsequent figures). h Monocle trajectory analysis of AT1 and AT2 cells in f, showing a linear transcriptomic shift from AT1 to AT2 cells in the control but further extending to transcriptionally exaggerated AT2 cells (bracket) in the Y/T^Sox9 mutant. The progenitor marker Sox9 is present at a low level but normally higher in AT2 cells. i Monocle gene clusters with similar dynamics along the transcriptomic shift in h. j Seurat module scores of gene sets associated with the 20% most decreased (top) or increased (bottom) NKX2-1 binding sites in d, plotted along the Monocle trajectory in h, showing concordant changes in NKX2-1 binding and gene expression.

Fig. 6. YAP/TAZ maintain AT1-specific NKX2-1 binding and cell fate, and prevent AT1-to-AT2 conversion.

a Confocal images of immunostained lungs showing loss of YAP/TAZ in recombined AT1 cells (filled versus open arrowhead) in the Y/T^Wnt3a mutant. Scale: 10 µm. b NKX2-1 ChIP-seq heatmaps of purified AT1 nuclei from P15 Y/T^Wnt3a mutant and littermate control lungs, sorted by fold change and cross-referenced with NKX2-1 binding from Fig. 1. c Examples of NKX2-1 peaksets in b. d ScRNA-seq UMAP comparison of Y/T^Wnt3a mutant and littermate control epithelial cells with cell typing supported by the dot plot. AT1 cells in the mutant are higher in number, likely resulting from more efficient cell dissociation due to the phenotype in cell morphology (h). e Volcano plots (MAST differential expression) comparing AT1 (left) and AT2 (right) cells in d. f Monocle trajectory analysis of AT1 and AT2 cells in d, showing a linear transcriptomic shift from AT1 to AT2 cells in the control but via intermediate cells (bracket) in the mutant, which express a subset of AT1 (Spock2) and AT2 (Sftpc) genes as well as Sfn. g Monocle gene clusters with similar dynamics along the transcriptomic shift in f. h Confocal images of immunostained lungs showing AT2 markers (LAMP3 and SFTPC) in recombined cells in the Y/T^Wnt3a mutant (filled versus open arrowhead). Some SFTPC-expressing mutant cells still have extended morphology at P15 (open arrow), and become mostly cuboidal at 10 weeks. Scale: 10 µm. i Seurat module scores of gene sets associated with the 20% most decreased (left) or increased (right) NKX2-1 binding sites in b, plotted along the Monocle trajectory in f, showing concordant changes in NKX2-1 binding and gene expression. j Unsupervised principal component analysis of NKX2-1 binding across indicated color-coded samples. The E18.5 Y/T^Sox9 mutant is right-shifted as far as mature AT2 cells, whereas the P15 Y/T^Wnt3a mutant is between AT1 and AT2 cells, reflecting the exaggerated AT2 cells and intermediate cells in the respective mutants. k A diagram, reminiscent of and color-coded as in j, depicting the normal differentiation of progenitors toward AT1 and AT2 cells, as well as the drift in both NKX2-1 binding and transcriptome observed in Y/T^Sox9 and Y/T^Wnt3a mutants.

Fig. 7. Cell-type-specific Nkx2-1 mutant cells explore distinct epigenetic space including the opposing cell fate.

a ATAC-seq heatmaps of sites with increased accessibility in purified AT1 (top) and AT2 (bottom) cells from the respective NKX2-1^Rtkn2 and NKX2-1^Sftpc mutants as in Fig. 2, cross-compared with and sorted by NKX2-1 binding in 10-week-old AT1 versus AT2 cells from Fig. 1. The sites from the two mutants have limited overlap and are largely not bound by NKX2-1, although 10% sites (boxed area) have NKX2-1 binding (arrowhead) and corresponding increased accessibility (log2 fold change shown) in control cells of the opposing fate. b Top enriched HOMER de novo motifs (binomial test) for sites in a as well as their p values and percentage of sites containing the predict motif. Note limited overlap except for the AP-1 motif and CEBP and TEAD motifs in the opposing cell types. c MA plots of genes associated with the boxed 10% sites in a using published bulk RNA-seq data of AT1-specific and AT2-specific Nkx2-1 mutant developing lungs, showing that increased accessibility of sites of the opposing cell fate does not correspond to gene upregulation. d Confocal images of immunostained mature lungs from AT1-specific (left panels) and AT2-specific (right panels) Nkx2-1 mutants and their respective littermate controls, showing aberrant epithelial cell clusters and expression of gastrointestinal markers PIGR and TFF2 in both mutants. Two doses (3 mg each; 48 h interval) of tamoxifen (Tam) were given and the lungs were harvested at 21 days after the initial dose. Scale: 10 µm. e Example increased accessibility in sites near a gastrointestinal gene (Hnf4a; distinct sites for NKX2-1^Rtkn2 versus NKX2-1^Sftpc mutants and annotated to a nearby gene Ttpal), AT2-specific NKX2-1 sites (Lyz2) in the NKX2-1^Rtkn2 mutant, and AT1-specific NKX2-1 site (Pdpn) in the NKX2-1^Sftpc mutant. The decreased accessibility of Lyz2 and Pdpn sites in concordant mutants are examined in Fig. 2. f A diagram to depict that NKX2-1 promotes the differentiation of progenitors toward AT1 or AT2 fate possibly in partner with YAP/TAZ/TEAD or CEBPA, respectively. Without NKX2-1, AT1 and AT2 cells drift toward the opposing cell fate while adopting the gastrointestinal fate.