Article Images (0)
×![]()
featured_play_listSave article to folder
X
Create new folder
Full Text
PDF
Table of Contents
FoxA1/2-Dependent Epigenomic Reprogramming
Graphical Abstract
Highlights
Resource
FoxA1/2-Dependent Epigenomic Reprogramming
Katherine Gillis,1,2 Walter A. Orellana,1,2 Emily Wilson,1,2 Timothy J. Parnell,1 Gabriela Fort,1,2 Pengshu Fang,1,2 Headtlove Essel Dadzie,1,2 Brandon M. Murphy,1,2 Xiaoyang Zhang,1,2 And Eric L. Snyder1,2,3,4,*
SUMMARY
INTRODUCTION
RESULTS
DISCUSSION
Limitations Of The Study
RESOURCE AVAILABILITY
KEY RESOURCES TABLE
Continued
Continued
EXPERIMENTAL MODELS AND STUDY PARTICIPANT DETAILS
Animal Studies
METHOD DETAILS
Histology And Immunohistochemistry
DNA Methylation Sequencing
Chromatin Immunoprecipitation Sequencing
QUANTIFICATION AND STATISTICAL ANALYSIS
Resource
FoxA1/2-dependent epigenomic reprogramming
drives lineage switching in lung adenocarcinoma
Graphical abstract
Highlights
d In vivoNKX2-1 loss alters lineage-specific DNAmethylation in lung adenocarcinoma d Recruitment of TET2/3 to regions demethylated upon NKX2-1 loss requires FoxA1/2 d FoxA1/2-dependent changes in 3D chromatin structure accompany lineage switching d KRASG12D dictates cell-type-specific DNAmethylation in lung adenocarcinoma Gillis et al., 2025, Developmental Cell 60, 1–18 February 3, 2025 ª 2024 Elsevier Inc. All rights are reserved, incl for text and data mining, AI training, and similar technologies. https://doi.org/10.1016/j.devcel.2024.10.009 Authors Katherine Gillis, Walter A. Orellana, Emily Wilson, ..., Brandon M. Murphy, Xiaoyang Zhang, Eric L. Snyder Correspondence eric.snyder@hci.utah.edu In brief Gillis et al. use in vivo epigenomic analyses to investigate mechanisms of lineage switching in lung adenocarcinoma (LUAD), finding that NKX2-1, FoxA1, and FoxA2 control lineage-specific DNAmethylation, TET2/3 localization, and 3D chromatin structure. They also show that oncogenic KRAS dictates cell-type-specific DNA methylation patterns in NKX2-1- negative LUAD. uding those ll ll
Resource
FoxA1/2-dependent epigenomic reprogramming
drives lineage switching in lung adenocarcinoma
Katherine Gillis,1,2 Walter A. Orellana,1,2 Emily Wilson,1,2 Timothy J. Parnell,1 Gabriela Fort,1,2 Pengshu Fang,1,2 Headtlove Essel Dadzie,1,2 Brandon M. Murphy,1,2 Xiaoyang Zhang,1,2 and Eric L. Snyder1,2,3,4,*
1Huntsman Cancer Institute, University of Utah, Salt Lake City, UT, USA 2Department of Oncological Sciences, University of Utah, Salt Lake City, UT, USA 3Department of Pathology, University of Utah, Salt Lake City, UT, USA 4Lead contact *Correspondence: eric.snyder@hci.utah.edu https://doi.org/10.1016/j.devcel.2024.10.009
SUMMARY
The ability of cancer cells to undergo identity changes (i.e., lineage plasticity) plays a key role in tumor progression and response to therapy. Loss of the pulmonary lineage specifier NKX2-1 in KRAS-driven lung adenocarcinoma (LUAD) enhances tumor progression and causes a FoxA1/2-dependent pulmonary-togastric lineage switch. However, the mechanisms by which FoxA1/2 activate a latent gastric identity in the lung remain largely unknown. Here, we show that FoxA1/2 reprogram the epigenetic landscape of gastricspecific genes after NKX2-1 loss in mouse models by facilitating ten-eleven translocation (TET)2/3 recruitment, DNA demethylation, histone 3 lysine 27 acetylation (H3K27ac) deposition, and three-dimensional (3D) chromatin interactions. FoxA1/2-mediated DNA methylation changes are highly conserved in human endodermal development and in progression of human lung and pancreatic neoplasia. Furthermore, oncogenic signaling is required for specific elements of FoxA1/2-dependent epigenetic reprogramming. This work demonstrates the role of FoxA1/2 in rewiring the DNA methylation and 3D chromatin landscape of NKX2-1-negative LUAD to drive cancer cell lineage switching.
INTRODUCTION
Cell identity is a fundamental aspect of normal tissue development and homeostasis, dictating cell function and fate within multicellular organisms. It encompasses a complex interplay of genetic, epigenetic, and environmental factors that shape the unique characteristics of each cell type.1 The ability of cells to alter their identity, a quality known as lineage plasticity, is essential during embryonic development and tissue repair.2 However, in the context of cancer, this highly regulated process can be exploited by tumor cells, allowing them to adopt alternative cell fates that provide a selective advantage.3 There are several contexts under which cancer cell lineage switching or transdifferentiation occurs, including (1) natural selection as a response to microenvironmental constraints or the acquisition of genetic alterations,3,4 (2) acute drug response resulting in the formation of drug-tolerant persister populations,5,6 and (3) secondary drug response as a mechanism of acquired resistance.7,8 Given the clinical importance of cancer cell lineage switching, there is a need to elucidate molecular mechanisms driving this process. Cancer cell lineage switching is a process primarily driven by epigenetic and transcriptional reprogramming.9 Although genetic alterations can promote lineage plasticity, epigenetic and transcriptional rewiring enable tumors to rapidly and reversibly alter their fate. Epigenetic reprogramming, including alterations Develo All rights are reserved, including those to the DNAmethylome, histone landscape, and cis-regulatory interactome, constitute important routes by which tumor cells can flexibly modulate their identity. DNA methylation is a critical regulator of cell identity, with lineage-specific methylation patterns preserving transcriptional programs and reinforcing celltype-specific characteristics.10–12 Tumors often exhibit aberrant DNA methylation, including global hypomethylation and sitespecific hypermethylation.13,14 Altered DNA methylation drives cancer cell lineage switching by modulating expression of genes involved in lineage commitment.15 Histone post-translational modifications (PTMs) and three-dimensional (3D) chromatin structure are also key epigenetic regulators of cell identity.9,16 During embryonic stem-cell differentiation, cis-regulatory interactions and histone PTMs are remodeled to facilitate expression of lineage-specific genes.17,18 Epigenetic dysregulation via altered enhancer-promoter (E-P) interactions or impaired histone deposition directly alters cell identity and drives tumor progression.19 Altogether, epigenetic modifications are responsible for the dynamic regulation of a cell’s transcriptional state, thereby controlling cell identity. One profound example of lineage switching occurs in invasive mucinous adenocarcinoma of the lung (IMA). IMA comprises 5%–10% of lung adenocarcinomas (LUADs) overall and harbors distinct transcriptional and molecular features due to pulmonary-to-gastric lineage switching that occurs during tumor pmental Cell 60, 1–18, February 3, 2025 ª 2024 Elsevier Inc. 1 for text and data mining, AI training, and similar technologies. progression. In LUAD genetically engineered mouse models (GEMMs), loss of the pulmonary-lineage-specifying transcription factor (TF) NKX2-1/TTF1 causes gastric transdifferentiation and recapitulates the morphology and gene expression profile of human IMA.20 Although 75% of human LUADs express NKX2-1, IMAs typically lose NKX2-1 expression due to point mutations or epigenetic silencing.21 Moreover, 75% of IMAs are driven by oncogenic KRAS (vs. 20%–30% of LUADs overall).22 As a result, treatment for IMA has lagged behind other LUAD subtypes driven by targetable oncogenic kinases.22 We have previously shown that the lineage switch caused by NKX2-1 loss in LUAD is mediated by differential chromatin binding of the forkhead box TFs FoxA1 and FoxA2. FoxA1/2 control endodermal development, differentiation, and metabolism.23 FoxA1/2 regulate gene expression by multiple mechanisms, including opening chromatin via pioneer factor activity,23 recruitment of histone-modifying enzymes,24,25 and facilitating DNA demethylation via the ten-eleven translocation (TET) dioxygenases.26 In LUAD, FoxA1/2 colocalize with NKX2-1 at adjacent sites in both human27 and murine tumors,20 including regulatory elements of pulmonary genes. Upon Nkx2-1 deletion, FoxA1/2 lose binding at many of these shared sites and relocate throughout the genome to de novo binding sites at gastric markers,20 many of which are known FoxA1/2 targets in the gastrointestinal (GI) tract.28Nkx2-1 deletion also induces histone PTMs associated with gene activation at de novo FoxA1/2-binding sites, such as histone 3 lysine 27 acetylation (H3K27ac). However, it is unknown whether FoxA1/2 directly mediate these chromatin modifications or whether chromatin modifications occur independently of FoxA1/2 binding. Here, we investigate the epigenetic basis of gastric lineage switching in LUAD using a sequential in vivo recombination system to discern the precise role of NKX2-1 and FoxA1/2 in modifying the LUAD epigenome. We show that FoxA1/2 are responsible for TET2/3 recruitment and demethylation at lineage-specific sites in NKX2-1-negative LUAD. Additionally, we find that FoxA1/2 facilitate H3K27ac deposition and E-P interactions at genes defining gastric identity. Finally, we show that oncogenic KRAS fine-tunes the epigenetic state of LUAD, modifying the precise identity adopted by cells after NKX2-1 loss.
RESULTS
NKX2-1 loss leads to widespread DNA methylation changes in LUAD DNA methylation is a stable epigenetic mark that governs cell identity. Unique cell-type-specific unmethylated regions define cell origin and type,10,29 and manipulation of enzymes that control DNA methylation can block cell differentiation.11,30,31 To determine whether LUAD lineage switching is accompanied by alterations in DNA methylation, we developed a sequential recombination GEMM that enables isolation of tumor cells and nuclei after Nkx2-1 deletion in established KRAS-driven LUAD. We generated KrasFSF-G12D/+; Rosa26FSF-CreERT2/Sun1; Nkx2-1F/F (KN) mice as well as control mice harboring a single conditional allele of Nkx2-1 (K) (Figure S1A). Tumors were initiated via intratracheal delivery of adenovirus expressing codon-optimized Flp recombinase (FlpO) from the surfactant protein C (SPC) promoter,32 which activates KrasG12D and CreERT2 in distal lung 2 Developmental Cell 60, 1–18, February 3, 2025 epithelium. Tamoxifen administration led to efficient (>95% recombination, refer to Figure S2) CreERT2-mediated deletion of Nkx2-1 and expression of a sfGFP-tagged nuclear membrane protein, Sun1. We used fluorescence-activated cell sorting (FACS) to isolate GFP-positive cells and nuclei, from which we extracted RNA (transcriptomics) and DNA (methylation), respectively (Figures S1B and S1C). Bulk RNA sequencing (RNA-seq) on sorted tumor cells from K and KN tumors (n = 3 replicates/genotype) identified 5,463 differentially expressed genes (DEGs; log2FC > 1; padj < 0.05, Table S1). Consistent with our previous findings,20 Nkx2-1 deletion caused a lineage switch (Figure S1D) in which tumor cells shed their pulmonary identity and adopted a state similar to gastric pit cells. Immunostaining for lineage-specific pulmonary and gastric markers validated this identity change at the protein level (Figure S2). We then profiled DNA methylation changes caused by NKX2-1 loss. Whole-genome enzymatic methyl-seq (EM-seq) on sorted nuclei from K and KN tumors (n = 3 replicates/genotype) demonstrated that NKX2-1 loss leads to extensive changes in DNA methylation. We identified 23,942 differentially methylated regions (DMRs) between K and KN tumors using the bsseq33 pipeline with a stringency cutoff of 15% difference in fraction methylation (Figures 1A and S1E; Table S2). Of these regions, 10,676 sites showed a significant gain in methylation with Nkx2-1 deletion (‘‘hyperDMRs’’), whereas 13,266 showed a significant reduction (‘‘hypoDMRs’’). The NKX2-1-binding motif was the most enriched in hyperDMRs, suggesting that NKX2-1 maintains a hypomethylated state at its binding sites in K tumors (Figure 1B). Additionally, hyperDMRs exhibited a strong enrichment for motifs bound by TFs that promote alveolar type II (AT2) and type I (AT1) identity including CCAAT enhancer binding protein (CEBP) and TEA domain (TEAD) families, respectively34. HypoDMRs were enriched for motifs bound by TFs that can control GI identity, including HNF4, GATA, and KLF. FoxA1/2bound motifs were enriched in both hyper- and hypoDMRs, reflecting their ability to promote pulmonary or gastric identity in LUAD.20,32 However, NKX motif was not enriched in hypoDMRs, consistent with our observation that NKX2-1 inhibits gastric marker genes predominantly by controlling FoxA1/2 localization rather than direct binding.20 Annotation of DMRs to their adjacent genes identified 3,362 hyper- and 4,122 hypomethylated genes in KN vs. K tumors. Gene set enrichment analysis (GSEA) of these differentially methylated genes (DMGs) revealed an enrichment for pulmonary and gastric cell signatures in hyper- and hypomethylated genes, respectively (Figure 1C). Gastric pit cell signatures were themost significantly enriched among hypomethylated genes. Enrichr analysis with the Genotype-Tissue expression (GTEx) database (dbGaP: phs000424.v8.p2) substantiated these lineage-specific methylation changes (Table S3). Fraction methylation profiles of DMRs located to pulmonary and gastric gene promoters highlight their contrasting methylation changes following NKX2-1 loss (Figure 1D). DNA methylation significantly increased at numerous AT2 marker genes (e.g., Sftpb and Cxcl15) following Nkx2-1 deletion (Figure 1E). However, a subset of AT2 markers, including Napsa and Sftpd, remained hypomethylated at their promoters even in a gastric state (Table S2). Conversely, markers of gastric pit cells (e.g., Tff1 and Gkn1) and pan-gastric (A) Heatmap of DMRs identified in K and KN tumors showing relative difference in mean methylation at CpGs within each DMR. DNA methylation scores calculated by subtracting the row mean from the average methylation level of each DMR per sample. Samples collected 14 weeks post-tumor initiation. (B) TF motifs enriched in hyper- and hypoDMRs identified between K and KN samples. (C) GSEA for cell-type signatures on DMGs identified between K and KN samples. (D) Heatmap showing the average fraction methylation values at DMRs localized to pulmonary and gastric gene promoters in K and KN samples. (E) Methylation tracks of AT2 (left) or GI (right) genes in K and KN tumors. Lines with asterisk indicate significant DMRs. (F) Intersection of DEGs and DMGs in K and KN tumors. Shown are representative pulmonary (left) or gastric (right) targets where gene expression correlates with methylation. See also Tables S1, S2, and S3. Developmental Cell 60, 1–18, February 3, 2025 3 markers (e.g., Hnf4a and Lgals4) demonstrated a significant decrease in their fraction methylation. Intersection of methylation and transcriptional datasets revealed a substantial overlap: 34.0% of upregulated genes after Nkx2-1 deletion showed a significant decrease in methylation (vs. 11.8% of downregulated genes; Figure 1F). Conversely, 24.4% of downregulated genes had a corresponding increase in methylation (vs. 12.3% of upregulated genes). Many AT2 and gastric-specific genes demonstrated concomitant methylation and transcriptional changes, suggesting that DNA methylation may be one epigenetic mechanism regulating lineagedefining gene expression. To investigate the dynamics of epigenetic changes at lineage-defining sites following NKX2-1 loss, we performed a time-course experiment profiling DNA methylation and protein expression changes at 3 days, 1 week, and 2 weeks post tamoxifen (Figure S3A; Table S2). Although some gastric proteins, such as HNF4a, were detectable on day 3, others were not readily detectable until 1–2 weeks after NKX2-1 loss. EMseq demonstrated a progressively higher number of DMRs over time (7,281 at day 3, 7,814 at 1 week, and 11,789 at 2 weeks post-Nkx2-1 deletion). Moreover, methylation gains preceded demethylation (Figures S3B and S3C), as 6,670 sites gained methylation whereas only 611 sites show reduced methylation on day 3. Interestingly, the pit cell marker Tff1 showed a gradual induction that correlated with methylation changes at an upstream enhancer (Figure S3D) that contacts the Tff1 promoter (Figure 5F). This enhancer undergoes early demethylation, correlating with weak Tff1 activation. By 2 weeks post deletion, Tff1 is more robustly expressed, corresponding with additional demethylation events near the Tff1 promoter and gene body. DNA methylation changes observed with Nkx2-1 deletion in LUAD GEMMs are conserved across human development and cancer pathogenesis We next sought to understand the extent to which DMRs induced by Nkx2-1 deletion in GEMMs correlate with DNA methylation patterns in normal human cells. First, we assessed the magnitude of DMRs between normal human lung and GI epithelium.10 Using the same analysis parameters, we identified 73,977 DMRs between human alveolar and gastric epithelial cells (Table S2). Therefore, loss of a single TFwithin KRAS-driven LUAD induces one-third as many DMRs as two developmentally distinct cell fate trajectories. Next, we intersected the top 1,000 differentially unmethylated regions in human alveolar and gastric epithelial cells with murine DMRs. We found that murine hypoDMRs were significantly more associated with human gastric unmethylated regions, whereas murine hyperDMRs correlated more with human alveolar unmethylated regions (Fisher’s exact test; p value < 0.05, representative pulmonary and gastric loci in Figure S4A). Finally, NKX2-1 and FOX motifs were the most highly enriched in unmethylated regions specific to human lung alveolar cells,10 corresponding with our motif analysis for hyperDMRs. Conversely, motifs enriched in hypoDMRs, including HNF4, GATA, and KLF families, were significantly associated with unmethylated regions in human GI epithelium. These analyses show that DNA methylation changes in our GEMM closely mirror normal human development and suggest that TFs control 4 Developmental Cell 60, 1–18, February 3, 2025 LUAD cell identity via methylation reprogramming that reflects their normal epigenetic function. We then asked whether similar lineage-specific DNA methylation patterns occurred in human LUAD. We binned KRASmutant LUAD samples from The Cancer Genome Atlas (TCGA) Pan-Cancer Atlas (n = 168/566) into two groups: NKX2-1-high, HNF4A-low (i.e., K tumors; n = 64, Table S1) and NKX2-1-low, HNF4A-high (i.e., KN tumors; n = 21, Table S1). GSEA on DEGs (n = 5,158; padj < 0.05; qval < 0.25, Table S1) between the two groups demonstrated significant enrichment for alveolar signatures in human K tumors and gastric pit signatures in human KN tumors (Figure S4B). Evaluation of DMRs (n = 3,152; padj < 0.05; qval < 0.25, Table S2) revealed that pan-gastric and pit cell marker genes had significantly lower DNA methylation levels in human KN vs. K tumors (Figure 2A). One of the top enriched cell-type signatures for KN hypomethylated genes was the gastric immature pit cell (Figure 2B). Pulmonary AT2 markers were heterogeneous, as some genes had higher DNA methylation in K tumors (e.g., SFTPB and NAPSA) whereas others showed no significant differences (e.g., SFTPC). One limitation of analyzing TCGA data is the contribution of stroma to theDNAmethylation signal.Usingpublishedmethylation and transcriptional profiling of patient-derived LUAD organoids,35 we examined DNA methylation in patient-derived organoids (PDOs) exhibiting a pulmonary or gastric transcriptional state. ‘‘Pulmonary’’ PDOs were identified by high expression of NKX2-1 and SFTPA2 (detailed in STAR Methods). ‘‘Gastric’’ PDOs, including potential IMA samples, were identified by high HNF4A and TFF1 expression. Intriguingly, we again found that human LUAD tumors exhibiting a gastric identity have low DNA methylation of pan-gastric and pit cell marker genes compared with tumors with a pulmonary identity (Figure 2C). Analysis of both primary tumors and PDOs suggests that NKX2-1 loss during human LUAD progression and deletion of Nkx2-1 in LUAD GEMMs both cause similar, conserved changes in DNA methylation. Given the conserved nature of these demethylation events, we then asked whether other cancer types exhibiting gastric lineage switching show similar epigenetic alterations. Pancreatic intraepithelial neoplasia (PanIN), a precursor of pancreatic ductal adenocarcinoma (PDAC), exhibits an extrapancreatic foregutlike differentiation state.36 Thus, we asked whether PanINs undergo demethylation of gastric marker genes compared with their presumed cell-of-origin, acinar cells. Using published whole-genome methyl-seq on laser-captured normal pancreas, PanINs, and PDAC,37 we found that the transition from acinarto-PanIN is significantly associated with demethylation of gastric markers. Genes demethylated in PanINs were most strongly associated with gastric pit cell signatures (Figures 2D, 2E, and S4C). Many of these sites remained demethylated in PDAC compared with normal pancreatic cells. Thus, DNA methylation changes that occur with gastric lineage switching in LUAD are broadly relevant to lineage plasticity and malignant progression in other types of human cancer. FoxA1/2 are required to demethylate gastric gene regulatory elements to facilitate the pulmonary-togastric lineage switch in LUAD FoxA1/2 control identity by facilitating DNA demethylation at lineage-defining genes both in vitro and in vivo.26,31 These demethylation events are dependent on the activity of TET dioxygenases, the enzymes that mediate DNA demethylation through progressive oxidative reactions.38 FoxA1/2-mediated demethylation is necessary for enhancer activation and gene expression of lineage-specific targets. We therefore investigated whether FoxA1/2 facilitate DNA methylation changes that occur after NKX2-1 loss. We generated a GEMMenabling concomitant Foxa1/2; Nkx2-1 deletion in established KRAS-driven neoplasia (KNF1F2) and performed bulk RNA-seq on sorted tumor cells (Figure S5A), which identified 4,273 DEGs (log2FC > 1; padj < 0.05, Table S1). Consistent with our previous findings32 (Figure S1), Foxa1/2 deletion in vivo prevented gastric transdifferentiation (Figures S5B and S5C). Instead, KNF1F2 mice harbor both squamous cell carcinomas (SCCs) and microscopic lesions resembling the squamocolumnar junction (SCJ) of the GI tract.32 FoxA1/2 loss also resulted in downregulation of processes involved in cholesterol, bile acid, and fatty acid metabolism (Figure S5D). Of note, Foxa1/2 deletion within established KN tumors did not significantly alter tumor burden as we were able to extract similar numbers of cells and nuclei from Developm both genotypes (approximately 80–100 million per KN or KNF1F2 lung; 10-fold higher than K controls). We next used whole-genome EM-seq on sorted nuclei from KNF1F2 tumors (Figures 3A and S5E; Table S2) to identify FoxA1/2-dependent methylation changes, merging DMRs from distinct pairwise comparisons between the tumor types (described in STAR Methods). We first evaluated regions demethylated after Nkx2-1 deletion. Of the 10,355 hypoDMRs, the vast majority (80%) were significantly higher in KNF1F2 than KN, showing that their demethylation was FoxA1/2 depen- dent (Figure 2A, purple). Motif analysis of these FoxA1/2-dependent hypoDMRs showed an enrichment for GI TFs, including HNF4, FOX, and SPDEF families (Figure 3B). Annotation and Gene Ontology (GO) analysis revealed that genes demethylated in a FoxA1/2-dependent manner were most significantly enriched for gastric pit and isthmus cell signatures (Figure 3C). Fraction methylation profiles of DMRs located at pan-gastric and pit cell marker gene promoters demonstrated a striking dependency on FoxA1/2 for demethylation (Figure 3D; representative methylation in Figures 3E and S5F). These data show that FoxA1/2 are required for demethylation of lineage-defining sites during gastric transdifferentiation in LUAD. In contrast, regions demethylated in a FoxA1/2-independent manner following Nkx2-1 deletion (Figure 2A, green) showed no enrichment for FOX and HNF4 motifs but a significant association with other TFs involved in GI differentiation (e.g., KLF and GATA; Figure 3B). GSEA of FoxA1/2-independent hypoDMRs revealed an enrichment for both basal and GI cell signatures. Enrichr analysis using the GTEx database showed that these regions correlate most significantly with human esophagus ental Cell 60, 1–18, February 3, 2025 5 (A) Heatmap showing relative difference in mean methylation of CpGs at DMRs identified in K, KN, and KNF1F2 tumors. DMRs categorized by FoxA1/2-dependency and methylation status (e.g., methylated vs. demethylated). Samples collected 14 weeks post-tumor initiation. (B) TF motifs enriched in FoxA1/2-dependent vs. -independent methylated and demethylated regions following Nkx2-1 deletion. (C) GSEA for cell-type signatures on DMGs identified from each of the four categories defined in Figure 2B. (D) Heatmap showing average fraction methylation values at DMRs localized to gastric gene promoters in K, KN, and KNF1F2 tumors. (E) Methylation tracks at gastric pit cell gene, Tff1, in K, KN, and KNF1F2 tumors. Lines with asterisk indicate significant DMRs. (F) Methylation tracks at AT2 cell gene, Sftpb, in K, KN, and KNF1F2 tumors. Lines with asterisk indicate significant DMRs. (G) Intersection of KN vs. K DEGs with DMRs defined in Figure 2A. Regions demethylated in KN tumors, regardless of FoxA1/2 dependency, are primarily associated with upregulated genes. Conversely, methylated regions are associated with downregulated genes. See also Tables S1, S2, S3, and S4. 6 Developmental Cell 60, 1–18, February 3, 2025 (A) Heatmap of TET3 occupancy at differential TET3-bound sites in KN and KNF1F2, as determined by Diffbind analysis (significance cutoff of p adjusted < 0.05). (B) TF motifs enriched in KN- and KNF1F2-specific TET3-bound sites. (C) Mean occupancy profile of TET3 at FoxA1/2bound sites in KN and KNF1F2 tumors. (D) GSEA for cell-type signatures on differentially bound TET3 sites identified between KN and KNF1F2 tumors. (E) Fraction methylation profiles at KN- and KNF1F2-specific TET3-bound sites in K, KN, and KNF1F2 tumors. (F) FoxA1, FoxA2, and TET3 ChIP peaks and methylation tracks at gastric pit cell genes, Tff1 and Gcnt3, in KN and KNF1F2 tumors. See also Table S6. (Table S3), which is consistent with the presence of SCCs in KNF1F2 mice. Given that gastric differentiation is dependent on FoxA1/2 at the transcriptomic and morphologic level, we were surprised to find pit cell and duodenal signatures enriched for genes demethylated in a FoxA1/2-independent manner. We therefore asked whether demethylation of these GI genes leads to increased expression, even in KNF1F2 samples. First, we found that all but one of the 27 genes driving the gastric pit cell signatures contained both FoxA1/2-dependent and -independent demethylation sites. FoxA1/2-independent sites were less abundant and located farther from the gene itself. Intersection with bulk RNA-seq showed that these methylation changes correlated strikingly well with gene expression. Genes comprising the gastric pit cell signature demonstrated the highest expression in KN tumors; however, they were induced to a lesser extent in KNF1F2 samples compared with K controls (Figure S5G). Of note, some of these genes (e.g., Klf3/4) play a role in squamous differentiation, suggesting that they may not be as specific to pit cells as some canonical markers, such as Gkn1 Developm and Tff1.39,40 Thus, there are FoxA1/2-independent mechanisms that partially demethylate and transcriptionally activate a subset of genes in the pit cell signature upon Nkx2-1 deletion. However, FoxA1/2 are required for full demethylation and activation of a gastric pit cell program in NKX2-1-negative LUAD, particularly canonical marker genes. We then asked whether hyperDMRs are also FoxA1/2 dependent. We found that 84.5% of the 7,212 hyperDMRs did not require FoxA1/2. Motif analysis of FoxA1/2-independent hyperDMRs revealed an enrichment for pulmonary TFs, including NKX, CEBP, and TEAD families (Figure 3B, beige). These findings correspond with GO analysis showing that genes methylated in a FoxA1/2-indepen- dent manner were significantly enriched for AT2 signatures (Figure 3C). Evaluation of individual genes revealed that most AT2specific sites gained methylation after NKX2-1 loss irrespective of FoxA1/2 activity (Figure 3F; Table S2). FoxA1/2-dependent hyperDMRs comprised the smallest number of DMRs (n = 1,113 sites). Although these regions were enriched for the NKXmotif, the FOXmotif was not enriched, suggesting that their FoxA1/2-dependency is more indirect than FoxA1/2-dependent hypoDMRs (Figure 3B, blue; FoxA1/2 localization analysis in Figure 4). GO analysis on FoxA1/2-dependent hyperDMRs did not show a consistent enrichment for a specific cell type, perhaps due to the small number of sites. Intersection of DEGs with DMRs revealed a strong inverse relationship between methylation and transcription. Specifically, 33.2% of genes upregulated with NKX2-1 loss were demethylated in a FoxA1/2-dependent manner (vs. 13.73% of downregulated genes; Figure 3G), including numerous pan-gastric and pit cell marker genes. Another 13.7% of upregulated genes were demethylated in a FoxA1/2-independent manner (vs. 5.7% of ental Cell 60, 1–18, February 3, 2025 7 downregulated genes), including Krt15, a gene that is highly expressed in basal cells of the lung, esophagus, and skin. Conversely, hyperDMRs demonstrated the opposite trend, correlating more strongly with transcriptional repression. Of the 2,462 genes downregulated with Nkx2-1 deletion, 22.8% and 7.47% gained methylation in a FoxA1/2-independent and -dependent manner, respectively (vs. 11.4% and 4.6% of upregulated genes), including numerous AT2-specific marker genes. These findings demonstrate the strong association between methylation and transcription, suggesting that methylation regulates expression of lineage-defining genes in LUAD. Finally, we evaluated regions that gained or lost methylation specifically in KNF1F2 tumors. We identified a total of 5,262 DMRs, with 77.4% of these sites exhibiting an increase in methylation and 22.6% a decrease (Figures S5H and S5I; Table S2). Interestingly, the FOX motif was the most highly enriched in KNF1F2-specific hypermethylated regions but was not associated with hypomethylated regions (Figure S5J). Thus, FoxA1/2 appear tomaintain a demethylated state at a subset of genomic loci in an NKX2-1-independent manner. These regions also demonstrated enrichment for motifs bound by HNF1a/b, TFs known to co-regulate hepatic and intestinal gene programs with FoxA1/2.41–43 HNF1b was highly expressed across all genotypes, suggesting that FoxA1/2 may be required for its activity or localization in K and KN tumors (Table S1). GO on KNF1F2-specific DMRs identified diverse cell-type signatures enriched for both hyper- and hypomethylated genes (Table S4). Of note, hypomethylated genes were not significantly associated with KNF1F2 tumor cell identity (e.g., squamous or SCJ cell signatures). This may be due to the small number of DMRs within this category (n = 1,191) or the possibility that these genes are not predominantly regulated by DNA methylation. Finally, KNF1F2-specific hypermethylated regions demonstrated enrichment for fatty acid and bile acid metabolic processes, suggesting that FoxA1/2 promote demethylation of metabolism-related genes in K and KN tumors (Table S4). FoxA1/2 recruit TET2/3 to lineage-defining gastric sites following NKX2-1 loss Given that FoxA1/2 are required for 80%of demethylation events in NKX2-1-negative LUAD, we investigated the mechanism by which FoxA1/2mediate thesemethylation changes. First, we performed chromatin immunoprecipitation followed by sequencing (ChIP-seq) for FoxA1 and FoxA2 on sorted nuclei from KN tumors, which identified 17,822 and 14,823 binding sites, respectively. Integration of both datasets revealed significant colocalization with few differentially bound sites (Figure S6A). Based on their genomic colocalization and functional redundancy, we mergedFoxA1/2-bound sites for downstreamanalysis. This identified 9,272 sites bound by both FoxA1/2, which localized primarily to introns and distal regulatory elements (Figure S6B). Motif analysis of FoxA1/2-bound sites showed a strong enrichment for the FOX motif along with GI TFs seen in hypoDMRs (Figure S6C). FoxA1/2-bound sites significantly overlapped with genes downregulated after their deletion (Fisher’s exact test; p value < 10 4). Specifically, FoxA1/2 bind 41.8%of genes downregulated after deletion and 14.7% of genes upregulated, demonstrating their role as predominantly transcriptional activators in KN tumors (Figure S6D). 8 Developmental Cell 60, 1–18, February 3, 2025 Given the association between FoxA1/2 binding and transcriptional activation, we sought to determine whether FoxA1/2 directly regulate demethylation at their binding sites. We found that FoxA1/2-bound sites were significantly more abundant in FoxA1/2-dependent hypoDMRs compared with independent hypoDMRs (18.7% vs. 9.2%, Fisher’s exact test; p value < 10 4; Figure S6E). These observations imply that FoxA1/2 directly facilitate demethylation by recruiting TET dioxygenases to their binding sites after Nkx2-1 deletion. To address this directly, we performed ChIP-seq for TET2 and TET3 in KN and KNF1F2 tumors (Tet2/3 are expressed at similar levels in all 3 genotypes; Figure S6F). Overall TET2 genomic localization was very similar to TET3, albeit the signal was significantly weaker (findings summarized in Figures S6G–S6J). Thus, we focused on results from TET3 ChIP-seq for downstream analysis. Differential peak analysis of TET3 binding identified 1,100 KN-specific and 711 KNF1F2-specific sites (Figure 4A). Motif analysis of differential TET3-bound sites identified FOX, HNF4, and KLF families as the top enriched motifs in KN samples (Figure 4B). Consistent with this, 59.1% of KN-specific TET3 peaks (n = 650) directly overlapped with FoxA1/2-bound sites. In contrast, only 2% of KNF1F2-specific TET3 sites (n = 14) colocalized with FoxA1/2. These findings suggest that FoxA1/2 directly recruit TET3 to lineage-specific sites within KN tumors. This is further supported by the loss of TET3 occupancy at FoxA1/2 peaks after Foxa1/2 deletion (Figures 4C and S6K). Differential TET3-bound sites were then annotated to their adjacent genes for downstream analysis. KN-specific sites were significantly enriched for gastric cell types, including pit and isthmus cell signatures as well as lipid metabolic processes (Figures 4D and S6L). Enrichr cell-type analysis substantiated these findings, revealing that KN-specific TET3 sites were most strongly associated with gastric and intestinal cell types (Figure S6M). Finally, we evaluated methylation signal at KN-specific TET3 sites and found a FoxA1/2-dependent reduction in methylation at TET3 peaks (Figure 4E). Inspection of specific marker genes demonstrated a clear loss of TET3 recruitment with Foxa1/2 deletion at both pit cell and pan-gastric markers (Figure 4F). Thus, FoxA1/2 are required to recruit TET3 to lineage-defining gastric sites to facilitate DNA demethylation. We then evaluated KNF1F2-specific TET3 sites. Motif analysis identified the top enriched TF motifs as squamous lineage specifier p63 and its paralogs44 as well as the AP-1 complex, which is known to promote squamous differentiation in stratified epithelia45 (Figure 4B). GSEA and Enrichr analysis on KNF1F2specific TET3 sites showed a significant association with basal cell and squamous signatures (Figures 4D and S6M). KNF1F2specific TET3 binding was identified at squamous epithelial markers as well as genes highly expressed in the SCJ of the stomach, including Chil4 and Mmp7 (Figure S6N). Intriguingly, when we evaluated methylation signal at KNF1F2-specific TET3 sites, we saw no significant difference in the fraction methylation between tumor types. Instead, these regions were lowly methylated in all tumor samples, suggesting that KNF1F2-specific TET3 sites are demethylated during either lung development or upon KRASG12D activation. This suggests that methylation is not a major mechanism regulating transcription of these genes and raises the possibility that TET3 is either superfluous to their activation or may perform non-canonical functions at KNF1F2-specific sites to regulate the epigenetic state (e.g., recruitment of chromatin-modifying enzymes).46 Overall, these data show that FoxA1/2 loss results in de novo TET3 binding to squamous-specific sites that are stably hypomethylated, regardless of TET3 localization. FoxA1/2 are required for H3K27ac deposition and E-P interactions at gastric-specific sites in NKX2-1negative LUAD In addition to modifying DNAmethylation, FoxA1/2 are known to direct histone PTMs via recruitment of histone-modifying enzymes.25 We have previously shown that de novo FoxA1/2-binding sites exhibit an increase in H3K27ac, a marker of active regulatory elements, following NKX2-1 loss.20 However, it is unknown whether these epigenetic changes are dependent on FoxA1/2 or whether they coincide with lineage-specific TET3 binding and DNA demethylation. We therefore performed ChIP-seq for H3K27ac on sorted nuclei from KN and KNF1F2 tumors. Differential peak analysis identified 4,092 KN-specific and 4,217 KNF1F2-specific H3K27ac sites (Figure 5A). Consistent with TET3 motif analysis, we found that KN-specific H3K27ac sites were significantly enriched for FOX and HNF4 motifs (Figure 5B). Accordingly, 37.5% of KN-specific H3K27ac sites (n = 1,535) directly overlapped with FoxA1/2-bound sites (vs. 2.8% of KNF1F2-specific H3K27ac sites). H3K27ac occupancy also demonstrated significant depletion at FoxA1/2 peaks following Foxa1/2 deletion (Figures S7A and S7B). Cell-type enrichment analysis revealed that KN-specific sites were most significantly associated with gastric pit and goblet cell signatures, showing that FoxA1/2-dependent H3K27ac sites occur at gastric-defining genes (Figure S7C; Table S3). Finally, we found that KN-specific HK27ac sites were enriched for mucin synthesis and metabolic processes (Figure S7D). Thus, FoxA1/ 2 are required not only for TET3 recruitment but also for H3K27ac deposition at genomic sites involved in regulation of gastric identity and cell metabolism. Next, we evaluated KNF1F2-specific H3K27ac sites. Themost highly enriched TF motif for these regions was bound by the AP1 family, which regulates squamous differentiation45 and was also seen in KNF1F2-specific TET3 peaks (Figures 3B and 5B). Although GSEA did not identify enrichment of a specific identity program, Enrichr analysis with the GTEx database showed that KNF1F2-specific H3K27ac sites were most significantly associated with esophagus (Figure S7E). Moreover, we saw significant H3K27ac accumulation at markers of squamous and SCJ identity in KNF1F2 tumors (Figure S7F). TF enrichment analysis using Encode and ChEA ChIP-X datasets showed a significant enrichment for SOX2 and TP63 (Table S3). Thus, a portion of the differential H3K27ac sites occupy lineage-defining squamous and SCJ sites in KNF1F2 tumors. These data show that FoxA1/2 are required for H3K27ac deposition at gastric-specific sites and that, in the absence of FoxA1/2, H3K27ac accumulates at genes controlling squamous differentiation. Given the association between H3K27ac deposition and tumor cell lineage, we next wanted to determine whether differential H3K27ac sites coincided with altered TET3 recruitment and DNA methylation patterns. We evaluated the distribution of TET3 ChIP-seq signal and fraction methylation at genotype-specific and shared H3K27ac peaks in KN and KNF1F2 tumors. Although shared H3K27ac peaks exhibited similar levels of TET3 binding and DNA methylation (Figure 5C), KN-specific H3K27ac sites demonstrated significantly higher TET3 recruitment and lower DNA methylation in KN tumors (Figure 5D). These data highlight the genomic colocalization of FoxA1/2dependent epigenetic changes and show that lineage-specific H3K27ac sites undergo FoxA1/2-mediated TET3 recruitment and demethylation. In contrast, KNF1F2-specific H3K27ac sites exhibited increased TET3 localization with no methylation changes relative to K or KN tumors (Figure 5E). This aligns with the aforementioned finding that KNF1F2-specific TET3 sites do not coincide with differential methylation and imply that alternative epigenetic mechanisms regulate transcription. As many of the affected H3K27ac sites are distant to gene promoters, they likely regulate their target genes through E-P loops. We therefore performed H3K27ac HiChIP on sorted nuclei from KN and KNF1F2 tumors, which identified 473,024 and 712,989 significant interactions between H3K27ac peaks, respectively. Next, we categorized H3K27ac contacts by the presence of only one transcription start site (TSS) at one anchor to identify E-P loops specifically. We then assessed the global changes of FoxA1/2-dependent cis-regulatory E-P interactions caused by Foxa1/2 deletion. As expected, shared H3K27ac sites present in both tumor types exhibited no significant difference in the number of E-P contacts (Figure S7G). However, lineage-specific H3K27ac sites demonstrated an increased number of E-P contacts within each respective tumor type. Intersection of E-P loops with DEGs identified after FoxA1/2 loss show that genes upregulated in KN or KNF1F2 tumors exhibit significantly more E-P loops within each respective tumor type (Figure S7H). These findings suggest that KNF1F2specific genes are activated by histone modifications and cis-regulatory interactions. Lastly, we examined the epigenetic landscape at lineagedefining gastric and SCJ marker genes. The genomic region containing GI genes Tff1, Tff2, and Tff3 contained numerous sites bound by FoxA1/2 in KN tumors (Figure 5F). These FoxA1/2-bound sites directly overlapped with regions demonstrating FoxA1/2-dependent demethylation, TET2/3 recruitment, H3K27ac deposition, and E-P looping. In particular, we see strong physical interactions between an enhancer located approximately 20 kb upstream from the Tff1 TSS and the promoters of both Tff1 and Tff2, which are disrupted after Foxa1/2 deletion. Similar FoxA1/2-dependent epigenetic changes were seen in genomic loci containing other gastric pit cell markers, including Gcnt3 (Figure S7I). Thus, FoxA1/2 are required to comprehensively reprogram the epigenetic landscape of lineage-defining genes to promote a gastric differentiation state in NKX2-1-negative LUAD. Evaluation of SCJ marker genes highly expressed in KNF1F2 tumors revealed contrasting epigenetic patterns. For example, SCJ marker Chil4 does not contain FoxA1-binding sites or exhibit an active chromatin state in KN tumors (Figure 5G). However, following Foxa1/2 deletion, we see significant accumulation of H3K27ac, TET2/3 binding, and local E-P contacts. There is subtle demethylation at the Chil4 promoter that does not meet the significance threshold. This supports our prior conclusions that most genes exhibiting increased TET3 recruitment after Foxa1/2 deletion are not differentially methylated. Examination Developmental Cell 60, 1–18, February 3, 2025 9 10 Developmental Cell 60, 1–18, February 3, 2025 (A) Representative images of KG1A and 22E organoids 2 weeks after treatment with EtOH or 4-OHT. IHC of NKX2-1, FoxA1, FoxA2, and HNF4a shown. Scale bar: 100 mm. (B) GSEA for cell-type signatures on DEGs identified in EtOH vs. 4-OHT for KG1A and 22E organoids. Organoids collected 2 weeks post treatment. (C) Heatmap showing log2-normalized counts of GI genes in EtOH vs. 4-OHT for KG1A and 22E organoids. (D) TF motifs enriched in hyperDMRs identified in EtOH vs. 4-OHT for KG1A and 22E organoids. (legend continued on next page) Developmental Cell 60, 1–18, February 3, 2025 11 of another SCJ marker, Mmp7, demonstrates a similar trend (Figure S7J), although the promoter of Mmp7 contains a small KNF1F2-specific hypoDMR, suggesting that a minor subset of SCJ marker genes may be regulated by DNA methylation. FoxA1/2 are required to maintain gastric identity, but not hypomethylation of gastric regulatory elements, in NKX2-1-negative LUAD DNA methylation is a central epigenetic mark that can be stably inherited by progeny cells in order to maintain cell identity. As such, mechanisms required to establish methylation patterns may not be necessary for their maintenance. For example, FoxA1/2 are required to establish hypomethylated patterns at liver-specific enhancers in vivo but they are dispensable for their maintenance.26 Thus, we investigated whether FoxA1/2 are required to maintain gastric gene expression and hypomethylated patterns at gastric loci following Nkx2-1 deletion by employing an in vitro 3D organoid system that uncouples Foxa1/2 deletion from NKX2-1 loss. After deriving organoids from KNF1F2 tumors, we noticed that Nkx2-1 is often stochastically downregulated. This uncoupling event is likely a consequence of positive selection for NKX2-1 loss in KRASG12Ddriven lung neoplasia.20,47 We derived and screened multiple KNF1F2 organoid lines to identify those that had (1) uniformly downregulated Nkx2-1 expression, (2) retained Foxa1/2 expression, and (3) adopted a gastric state. This led to the selection of two organoid lines, KG1A and 22E, which met all criteria. We then confirmed that treatment with 4-hydroxytamoxifen (4-OHT) in vitro successfully deleted Foxa1/2 and caused downregulation of the FoxA1/2 target gene, Hnf4a (Figure 6A). To determine whether FoxA1/2 are required for maintenance of gastric identity, we performed bulk RNA-seq on KG1A and 22E organoids in the presence and absence of FoxA1/2. We found that deletion of Foxa1/2 in vitro led to extensive transcriptional changes, with 2,144 DEGs identified in KG1A and 1,302 DEGs in 22E (log2FC > 1; padj < 0.05, Figure S8A; Table S5). Intersection of these datasets with in vivo RNA-seq revealed a considerable overlap, whereby 27.7% and 19.5% of the genes downregulated in vivo with FoxA1/2 loss were also downregulated in KG1A and 22E, respectively (Figure S8B). GSEA of cell identity programs closely mirrored results of Foxa1/2 deletion in vivo (Figure 6B). In both organoid lines, gastric identity, particularly pit and goblet cell signatures, was Foxa1/2-dependent. Individual gastric targets, including pit cell markers, Tff1 andGkn1, andmucous neck cell markers, Tff2 andGkn3, were substantially downregulated with FoxA1/2 loss in both lines (Figure 6C). Pathway analysis also identified FoxA1/2-dependent metabolic processes in vitro (Figure S8C) that were similar to those observed in vivo (Figure S8D). On the other hand, Foxa1/2 deleted lines showed enrichment for multiple cell-type signatures and inflammatory processes. Of note, KG1A and 22E were significantly associated with distinct identities after deletion, which partially recapitulates cell fate heterogeneity seen af- (E) Intersection of DEGs and DMRs identified in EtOH vs. 4-OHT for KG1A and 2 increase in methylation. (F) FoxA1 ChIP peaks and methylation tracks at gastric genes, Gcnt3 and Tff2, i See also Tables S4 and S5. 12 Developmental Cell 60, 1–18, February 3, 2025 ter Foxa1/2 deletion in vivo. Overall, these data demonstrate that FoxA1/2 are required for maintaining a gastric identity in NKX21-negative LUAD. We next sought to determine whether FoxA1/2 are required for maintenance of DNA hypomethylation patterns at lineagedefining genes. We performed whole-genome EM-seq on KG1A and 22E in the presence and absence of FoxA1/2 as well as ChIP-seq for FoxA1 in both lines. Interestingly, loss of FoxA1/2 in NKX2-1-negative organoids led to relatively modest DNA methylation changes when compared with the results of concomitant Foxa1/2;Nkx2-1 deletion in vivo (Figure 3). In KG1A and 22E, we identified a total of 3,914 and 2,280 DMRs, respectively (Figure S8D; Table S5). FoxA1 ChIP-seq identified 21,827 peaks in KG1A and 26,091 in 22E, localized primarily to intronic and intergenic regions (i.e., potential enhancer sites, Figure S8E). To determine whether FoxA1/2 directly alter methylation patterns, we intersected FoxA1-bound sites with DMRs. Regions that gainedmethylation after FoxA1/2 loss (hyperDMRs) demonstrated a significant overlap with FoxA1-bound sites (as compared with regions that lost methylation or hypoDMRs). For example, in KG1A, 27.6%of hyperDMRs directly overlapped with FoxA1-bound sites vs. 6.24% of hypoDMRs (Fisher’s exact test; p value < 10 4). These findings were substantiated by motif analysis, which demonstrated enrichment for the FOX motif in hyperDMRs, but not hypoDMRs, for both lines. These data highlight the role of FoxA1/2 in maintaining a demethylated state at a subset of binding sites in KN organoids. Given the association between FoxA1 binding and demethylation, we next investigated whether FoxA1/2 are responsible for maintaining hypomethylated patterns at gastric-specific sites. Cell-type enrichment analysis on hyperDMRs revealed enrichment for various midbrain signatures but no GI cell types (Table S4). Although motif analysis of hyperDMRs showed a strong enrichment for the FOX motif, it lacked other GI TFs, including HNF4 and KLF families found in vivo (Figure 5D). Intersection of DMRs with transcriptional changes revealed a minor overlap: less than 10% of genes downregulated with Foxa1/2 deletion had a corresponding increase inmethylation (Figure 6E). In fact, numerous gastric marker genes, both bound by FoxA1 and downregulated with Foxa1/2 deletion, exhibited no change in methylation, including targets Tff2 and Gcnt3 (Figure 6F). Although most gastric targets did not undergo substantial methylation changes, there were several notable exceptions, including Tff1 and Gkn3, which were significantly methylated upon FoxA1/2 loss (Figure S8F). Altogether, these findings demonstrate the diverging roles of FoxA1/2 in regulation of gastric gene transcription and methylation. Although FoxA1/2 are required to maintain a gastric cell identity in NKX2-1-negative LUAD, they are not required tomaintain hypomethylated patterns at most gastric loci. These findings substantiate previous work10,26 showing that lineage-specific DNA methylation patterns are stable and do not always require the TFs with which they were established. 2E organoids. Less than 10% of downregulated genes show a corresponding n KG1A organoids treated with EtOH or 4-OHT. (A) Representative images of KN tumors and N hyperplasias. IHC of pERK, TFF1, GKN1, and PGC shown. Scale bar: 100 mm. (B) GSEA for cell-type signatures on DEGs identified in KN vs. N samples. (C) Log2 fold change values of genes comprising the gastric pit, gastric chief, or ciliated cell signatures in KN vs. N tumors. (D) Enrichment score determined by ssGSEA for immature gastric pit cell and gastric neck cell signatures in K, KN, KNF1F2, and N samples. ***p value < 10 3 and **p value < 10 2 in KN vs. N cells, t test. (E) Heatmap of mean relative CpGmethylation over DMRs identified in KN vs. N samples. Samples collected 14 weeks post-tumor initiation. (F) TF motifs enriched in KN vs. N DMRs. (G) FoxA1 and FoxA2 ChIP peaks and methylation tracks at gastric pit cell gene, Tff1, in KN, N, and K samples. See also Tables S1, S2, and S3. Oncogenic KRAS promotes FoxA1/2-dependent transcriptional activation and demethylation of gastric pit cell genes in NKX2-1-negative LUAD Aberrant KRAS signaling plays a pivotal role in LUAD pathogenesis, inducing tumor cell survival and proliferation via activation of its canonical target, extracellular-signal-related kinase (ERK). The importance of ERK activity is underscored by the observation that small molecule KRAS inhibitors block ERK activation to a greater extent than other downstream signaling pathways.48,49 In addition to its roles in tumor progression, ERK signaling regulates cell differentiation and development.50 In mouse embryonic stem cells, ERK signaling suppresses pluripotent gene expression to induce an endodermal cell state.51,52 We have previously shown that RAF/MEK/ERK activity modulates the exact gastric identity adopted by NKX2-1-negative LUAD. Specifically, RAF/MEK/ERK signaling in a BRAFV600E IMA model promotes a gastric pit identity while repressing chief and tuft cell states.5 We have also shown that pulmonary Developm Nkx2-1 deletion in the absence of KRASG12D induces small mucinous alveolar hyperplasias that express pangastric markers such as HNF4a but not gastric pit cell marker GKN1.20 These Nkx2-1-deleted hyperplasias (N) exhibit reduced ERK activation compared with KN tumors by immunohistochemistry (IHC) (Figure 7A). Thus, ERK activity appears to alter the precise identity adopted with Nkx2-1 deletion. We therefore performed RNA-seq to determine the full impact of KRASG12D on the transcriptome and identity of Nkx2-1-deleted AT2 cells, which identified 2,465 DEGs between N hyperplasias and KN tumors (log2FC > 1; padj < 0.05; Figure S9A; Table S1). GO analysis showed a significant enrichment for gastric pit cell signatures in KN tumors (Figure 7B), as exemplified by genes such as Tff1, Gkn1, and Gcnt3 (Figures 7A and 7C). Given the fact that both KN tumors andNkx2-1-deleted hyperplasias upregulate pan-gastric markers, we evaluated the extent to which each sample is associated with distinct gastric identities. Using single-sample GSEA (ssGSEA), we found that N hyperplasias are enriched for pit cell signatures relative to K and KNF1F2 tumors but to a much lesser extent than KN tumors (Figures 7D and S9B). Other gastric programs including tuft cell signatures demonstrated similarly high enrichment in KN and N samples compared with K and KNF1F2 (Figure S9C). Consistent with this, pan-gastric markers Hnf4a and Ctse were expressed at similarly high levels in KN and N samples (Figure S9D). Finally, using KNF1F2-derived organoids, we found that MEK inhibition blocks FoxA2-mediated induction of gastric pit cell genes, while increasing activation of chief cell targets (Figure S9E). ental Cell 60, 1–18, February 3, 2025 13 In contrast to KN tumor cells, N cells demonstrated a strong enrichment for fetal ciliated epithelial cell types as well as cilia and microtubule gene sets (Figures 7B and S9F; Table S3). Specific genes involved in cilia regulation and assembly were strongly upregulated in N lesions, including Foxj1, Tmem231, Spag16, and Drc3 (Figure 7C). In addition to cilia-related gene sets, N cells also upregulated gastric chief cell markers, including Pgc, Pga5, and Cckar (Figures 7A and 7C), and exhibited a strong association with gastric neck cell signatures (Figure 7D). Of note, stomach chief cell signatures were associated with N samples but did not meet the significance p value threshold of 5%. Altogether, these results show that Nkx2-1 deletion within AT2 cells is sufficient to activate elements of a gastric differentiation program, but the combination of NKX2-1 loss with KRASG12D (likely due to increased ERK signaling) causes maximum induction of a gastric pit cell identity. In the absence of a mutant oncogene, NKX2-1-negative lung epithelial cells adopt a ciliated epithelial state with co-expression of a subset of gastric chief and neck cell markers. Whole-genome EM methyl-seq on sorted N and KN nuclei identified 10,620 DMRs, with 9,026 sites exhibiting higher methylation in N samples and 1,594 sites exhibiting lower methylation (Figure 7E; Table S2). HOMER analysis identified the FOX motif in both N and KN DMRs (Figure 7F). However, N samples lacked enrichment for additional GI TF motifs present in KN samples, including the KLF and HNF4 families; instead, they were associated with an assorted group of motifs, including RELA, NF1, and TEAD. Intriguingly, KN samples did not exhibit higher expression of Foxa1/2 or Hnf4a/g, suggesting that KRASG12D might alter their function through post-translational mechanisms and/or changes in the composition of TF complexes. Intersection of these data with FoxA1/ 2-dependent DMRs (Figure 3A, purple) showed that 29.1% of all FoxA1/2-dependent hypoDMRs (n = 2,421) directly overlapped with KRASG12D-dependent hypoDMRs. Moreover, 40.2% of these co-dependent sites (n = 974) were directly bound by FoxA1 and/or FoxA2. These data demonstrate a close association between FoxA1 binding and KRASG12D-dependent DMRs and also raise the question of whether RAF/MEK/ERK signaling promotes FoxA1/2 binding to these sites to regulate cell identity. GO analysis showed that the top two cell-type signatures enriched for KN hypomethylated genes were gastric immature and mature pit cells (Figure S9G). Individual pit cell markers Tff1, Gkn1, and Gcnt3 as well as gastric markers Lgals4 and Anxa8 all demonstrated demethylation that was dependent on the combination of KRASG12D activation andNkx2-1 deletion (Figures 7G and S9H; Table S2). Inspection of key KN-specific pit cell markers showed that most genes either directly overlapped with FoxA1/2-bound sites (e.g., Tff1) or contained local binding sites (e.g., Gcnt3 and Gkn1). In contrast, pan-gastric markers, including Hnf4a and Ctse, were demethylated upon NKX2-1 loss independent of KRASG12D activation. Evaluation of N-specific hypomethylated genes did not reveal any significant associations with specific cell types or GO signatures (perhaps due to the small number of sites identified). Moreover, cilia marker genes Foxj1, Tmem231, andDrc3 did not undergo demethylation despite being upregulated in N hyperplasias, suggesting that transcriptional regulation of these genes is methylation 14 Developmental Cell 60, 1–18, February 3, 2025 independent. Altogether, these findings show that the combined action of FoxA1/2 and oncogenic KRAS, likely through increased ERK signaling, leads to demethylation and transcriptional activation of specific gastric loci during the lineage switch that follows the loss of NKX2-1 in LUAD.
DISCUSSION
Cell plasticity is intricately tied to the evolutionary fitness of a cancer cell. The ability of tumor cells to alter their identity in response to selective pressures (natural or therapy-induced) is essential for their continued growth, survival, and malignancy potential. Therefore, it is imperative to define the molecular mechanisms, including transcriptional and epigenetic factors, that control cancer cell identity. Here, we show that pioneer factors FoxA1/2 reprogram the epigenetic landscape of NKX2-1negative LUAD to facilitate cancer cell lineage switching. Following Nkx2-1 deletion, FoxA1/2 localize to regulatory elements of gastric-specific sites where they coordinate DNA demethylation, H3K27ac deposition, and E-P interactions. We also show that FoxA1/2 recruit TET2/3 dioxygenases to lineage-defining genes, providing a mechanism for their demethylation. The propensity of LUADs to undergo systematic rewiring to a gastric epigenetic state upon NKX2-1 loss is likely linked to the developmental origins of the lungs. Foregut endoderm, a portion of the gut tub that forms numerous GI structures, also gives rise to the primitive lung. As endodermal lineage specifiers, FoxA1/2 regulate the development and differentiation of both pulmonary and GI tissues derived from foregut endoderm. Deletion of the pulmonary lineage specifier Nkx2-1 likely releases FoxA1/2 from their roles in maintaining a pulmonary identity, allowing them to revert to transcriptional and epigenetic functions performed during foregut specification. In other endodermal tissues such as the liver, FoxA1/2 are known to mediate similar epigenetic changes, including demethylation of lineage-specific enhancers and maintenance of an active chromatin state.26,53 Although the specific functions performed by FoxA1/2 during stomach differentiation are not as well understood,54 our findings suggest that FoxA1/2 may aid in gastric pit cell specification by facilitating epigenetic alterations at lineage-defining sites. Epigenetic rewiring during gastric lineage switching in LUAD results in DNA methylation changes that are conserved across other types of human cancer. We found that the malignant transformation of pancreatic acinar cells to PanIN lesions is associated with demethylation of genes defining the gastric pit cell identity. FoxA1/2 are known to be expressed in normal pancreatic epithelia, PanINs, and well-differentiated PDACs.55 They are part of a transcriptional network with PDX1 and HNF4⍺ that promote the progenitor or classical subtype of PDAC by specifying a pancreas cell fate.56 Thus, it is likely that FoxA1/2 function similarly in PanIN formation to facilitate demethylation and transcriptional activation of genes defining a gastric pit cell program. Downregulation of acinar-specific TFs such as MIST1 also enhance PanIN formation.57 These transcriptional events may aid in the pit cell lineage switch by modulating activity and function of the FoxA proteins (e.g., altering FoxA1/2 genomic localization). Altogether, our findings hold important implications for understanding the mechanism of lineage switching observed in other types of cancer. Mitogen-activated protein kinase (MAPK) signaling is known to regulate GI epithelial differentiation.58 Here, we show that oncogenic RAS/RAF/MEK signaling alters the precise differentiation state adopted by NKX2-1-negative LUAD tumors cells, promoting a gastric pit cell program while repressing ciliated, neck, and chief cell identities. We also find that regardless of RAS signaling levels, the FoxA motif is associated with genotype-specific unmethylated regions. This poses the fundamental question of how RAS/RAF/MEK signaling alters FoxA1/2 activity to modulate cell identity, which could include regulation of FoxA1/2 genomic localization and/or co-factor interactions. Results from these studies have important implications for understanding the cell identity that LUAD adopts following targeted KRAS inhibition and the role of FoxA1/2 in facilitating this differentiation program.More broadly, we anticipate that future experiments to delineate the temporal order and precise contribution of each epigenetic modification to lineage switching will facilitate the development of therapeutic strategies aimed at preventing or counteracting identity-specific vulnerabilities in LUAD.
Limitations of the study
Although changes in DNAmethylation, histone H3K27ac, and E-P contacts can regulate gene expression in other contexts, we have not directly assessed the relative contribution of each of these epigenetic changes to lineage switching in vivo. The development of FoxA1/2 mutants that selectively abrogate FoxA1/2 recruitment of specific enzyme families (e.g., TET, CBP/p300, etc.) would enable further investigation in this area. Bulk assays used in this study may not identify heterogeneous epigenetic changes that occur in only a subset of tumor cells. Single-cell transcriptional and epigenetic assays would provide deeper insights into intratumoral heterogeneity. Finally, using H3K27ac as bait for HiChIP captures interactions among enhancers and promoters but is suboptimal for calling topologically associating domains (TAD) structures. Additional studies, such as antibody-independent Hi-C assays or CCCTC-binding factor (CTCF)/cohesinbasedHiChIP assays, will be needed to determinewhether a subset of FoxA1/2-bound sites affect CTCF/cohesin binding at TAD boundaries to influence the associated TAD structures.
RESOURCE AVAILABILITY
Lead contact Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Eric L. Snyder (eric. snyder@hci.utah.edu). Materials availability Murine organoid lines are available upon request. Data and code availability d Bulk RNA-seq, methyl-seq, ChIP-seq, and HiChIP data have been deposited at Gene Expression Omnibus GEO: GSE247855 and are available as of the date of publication. d Microscopy data reported in this paper will be shared by the lead con- tact upon request. d This paper does not report original code. d Any additional information required to reanalyze the data reported in this work paper is available from the lead contact upon request. ACKNOWLEDGMENTS We are grateful to Scott Hale, Bryant Perkins, K.-T. Varley, andmembers of the Snyder lab for suggestions and comments. We thank Brian Dalley for sequencing expertise, Jay Gertz for ChIP-seq expertise, James Marvin for FACS expertise, and Ian MacIsaac for assistance with data analysis. The results published here are in part based upon data generated by the TCGA Research Network: https://www.cancer.gov/tcga. E.L.S. was supported by grants from the NIH (R01CA212415, R01CA240317, and R01CA237404), the American Lung Association (LCD-821670), and institutional funds (Department of Pathology and Huntsman Cancer Institute/Huntsman Cancer Foundation, University of Utah). K.G. and G.F. were supported by the NIH/NCI (F31CA275266 and F31CA275328). G.F. was also supported by Genetics Training Grant (T32GM141848). The graphical abstract for this publication was created with Biorender.com. Research reported in this publication utilized shared resources (including High Throughput Genomics, Bioinformatics, Flow Cytometry, and Biorepository and Molecular Pathology) at the University of Utah and was supported by the National Cancer Institute of the National Institutes of Health under award number P30CA042014. Work in the flow cytometry core was also supported by the National Center for Research Resources of the National Institutes of Health under award number 1S20RR026802-1. The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH. AUTHOR CONTRIBUTIONS K.G. and E.L.S. designed experiments. K.G., W.A.O., G.F., H.E.D., E.W., P.F., B.M.M., and X.Z. performed experiments. K.G., G.F., T.J.P., E.W., X.Z., and E.L.S. analyzed data. E.L.S. performed histopathologic review. K.G. and E.L.S. wrote the manuscript. All authors discussed results and reviewed and revised the manuscript. DECLARATION OF INTERESTS The authors declare no competing interests. STAR+METHODS Detailed methods are provided in the online version of this paper and include the following: d KEY RESOURCES TABLE d EXPERIMENTAL MODELS AND STUDY PARTICIPANT DETAILS B Animal studies B Husbandry and housing conditions of experimental animals B Primary 3D organoid cultures d METHOD DETAILS B Tumor initiation and tamoxifen administration in vivo B Histology and immunohistochemistry B Establishing primary murine LUAD organoids B In vitro 4-hydroxytamoxifen (4OHT) treatment B Generating a single cell suspension from organoid cultures B DNA Methylation sequencing B RNA sequencing B Chromatin immunoprecipitation sequencing d QUANTIFICATION AND STATISTICAL ANALYSIS SUPPLEMENTAL INFORMATION Supplemental information can be found online at https://doi.org/10.1016/j. devcel.2024.10.009. Received: November 15, 2023 Revised: June 21, 2024 Accepted: October 15, 2024 Published: November 7, 2024 Developmental Cell 60, 1–18, February 3, 2025 15
KEY RESOURCES TABLE
REAGENT or RESOURCE SOURCE IDENTIFIER Antibodies Rabbit monoclonal to Cytokeratin 5 (IHC) Abcam Ab52635[EP1601Y]; RRID: AB_869890 Rabbit monoclonal to FoxA1 (ChIPseq) Abcam ab170933[EPR10881]; RRID: AB_3065104 Rabbit monoclonal to FoxA1 (IHC) Abcam ab173287 [EPR10881-14] Rabbit monoclonal to FoxA2 (WB; IHC) Abcam ab108422 [EPR4466]; RRID: AB_10891055 Goat polyclonal to Galectin-4 (IHC) R&D Systems Cat#AF2128; RRID: AB_2297050 Mouse monoclonal to GKN1 (IHC) Abnova Cat#H00056287-M01 [2E5]; RRID: AB_1505437 Rabbit monoclonal to GFP (IHC) Cell Signaling Technologies Cat#2956 (D5.1); RRID: AB_1196615 Rabbit monoclonal to HNF4a (IHC) Cell Signaling Technologies Cat#3113 (C11F12); RRID: AB_2295208 Rabbit polyclonal to H3K27ac (ChIPseq) Active Motif Cat#39133; RRID: AB_2561016 Rabbit polyclonal to H3K27ac (HiChIP) Abcam Ab4729; RRID: AB_2118291 Mouse monoclonal to MUC5AC (IHC) Abnova Cat#MAB13117[SPM488] Rabbit monoclonal to NKX2-1 (Murine IHC) Abcam ab76013 [EP1584Y]; RRID_AB: 1310784 Rabbit monoclonal to p63 (IHC) Cell Signaling Technologies Cat#13109 (D2K8X); RRID: AB_2637091 Rabbit monoclonal to pERK (IHC) Cell Signaling Technologies Cat#4370 (D13.14.4E); RRID: AB_2315112 Rabbit polyclonal to PGC (IHC) Sigma Cat#HPA031718 Rabbit monoclonal to ProSP-C (IHC) Abcam Ab211326[EPR19839]; RRID: AB_2927746 Rabbit polyclonal to TET3 (ChIPseq) Sigma Cat#ABE290 Rabbit polyclonal to TFF1 (IHC) Origene Cat#TA382458 Bacterial and virus strains Adeno-mSPC-FlpO University of Iowa Viral Vector Core Snyder-6695 Stbl3 Thermo Fisher Cat#C737303 Chemicals, peptides, and recombinant proteins 4-Hydroxytamoxifen Cayman Chemicals Cat#17308 Advanced DMEM/F12 Thermo Fisher Scientific Cat# 12634028 B-27 Thermo Fisher Scientific Cat#17504044 Collagenase Type I Thermo Fisher Scientific Cat#17100-017 Cell Recovery Solution Corning Cat#10378016 DAPI Sigma Cat#D9542 Dispase Corning Cat#354235 DNase Worthington Cat# LS002006 DMEM/F12 Thermo Fisher Scientific Cat#11320082 Dynabeads Protein A Thermo Fisher Scientific Cat#10002D Gastrin I Sigma Aldrich Cat#G9020 HEPES Thermo Fisher Scientific Cat#15630080 Human FGF-10 PeproTech Cat#100-26 Murine EGF PeproTech Cat#315-09 Murine Noggin PeproTech Cat#250-38 Murine R-Spondin-1 PeproTech Cat#315-32 nAcetylcysteine Sigma Aldrich Cat#A7250 N-2 Thermo Fisher Scientific Cat#17502048 Penicillin-Streptomycin-Glutamine Thermo Fisher Scientific Cat#10378016 Plasmocin InvivoGen Cat#ant-mpp Polybrene Sigma Aldrich Cat#107689 (Continued on next page) Developmental Cell 60, 1–18.e1–e8, February 3, 2025 e1
Continued
REAGENT or RESOURCE SOURCE IDENTIFIER Tamoxifen Sigma Aldrich Cat#T5648 TransIT-293 Mirus Cat# MIR2700 TRIzol Reagent Thermo Fisher Scientific Cat#15596018 TrypLE Express Enzyme Thermo Fisher Scientific Cat#12604021 Critical commercial assays BLOXALL Endogenous Blocking Solution, Peroxidase and Alkaline Phosphatase Vector Labs Cat#SP-6000-100 ImPRESS HRP Horse Anti-Rabbit IgG Polymer Detection Kit, Peroxidase Vector Labs Cat#MP-7401 ImPRESS Goat Anti-Rat IgG (Mouse Adsorbed) Polymer Kit, Peroxidase Vector Labs Cat#MP-7444 Rodent Block M Biocare Medical Cat#RBM961G Mouse-on-Mouse HRP Polymer Biocare Medical Cat#MM620 NEBNext Ultra II DNA Library Prep Kit NEB Cat# E7645 PureLink RNA Mini Kit Thermo Fisher Scientific Cat# 12183018A Deposited data Bulk RNA-seq, Methyl-seq, ChIP-seq, and Hi-ChIP data NCBI Gene Expression Omnibus GSE247855 Experimental models: Cell lines 293T: Homo sapiens (Fetus) ATCC Cat#CRL-3216 22E: Mus musculus (Female) This paper N/A KG1A: Mus musculus (Female) This paper N/A L-WRN: Mus musculus (Male) ATCC Cat#CRL-3276 Experimental models: Organisms/strains Foxa1F/F: Mus musculus mixed C57BL/6J x 129SvJ background Dr. Klaus H. Kaestner (Upenn, Philadelphia, PA) N/A Foxa2F/F: Mus musculus mixed C57BL/6J x 129SvJ background Dr. Klaus H. Kaestner (Upenn, Philadelphia, PA) N/A KrasFSF-G12D/+: Mus musculus mixed C57BL/6J x 129SvJ background Dr. Tyler Jacks (MIT, Cambridge, MA) N/A Nkx2-1F/F: Mus musculus mixed C57BL/6J x 129SvJ background Dr. Shioko Kimura (NCI/NIH, Bethesda, MD) N/A RosaFSF-CreERT2: Mus musculus mixed C57BL/6J x 129SvJ background Dr. Dieter Sau (Technische Universitat Munchen, Munchen, Germany) N/A Trp53FRT/FRT: Mus musculus mixed C57BL/6J x 129SvJ background Dr. David G Kirsch (Duke University, Durham, NC) N/A Recombinant DNA d8.9 (plasmid) Dr. Tyler Jacks (MIT, Cambridge, MSBackspaceA) DuPage et al.59, PMID:19561589 VSV-G (plasmid) Dr. Tyler Jacks (MIT, Cambridge, MA) DuPage et al.59, PMID:19561589 Software and algorithms Adobe Illustrator 25.0 Adobe N/A BEDtools Quinlan and Hall60 https://doi.org/10.1093/bioinformatics/btq033 BioRender BioRender https://www.BioRender.com Clumpify 38.34 Bushnell61 http://sourceforge.net/projects/bbmap/ DESeq2 1.34.0 Love et al.62 https://bioconductor.org/packages/ release/bioc/html/DESeq2.html Ensembl Release 102 https://useast.ensembl.org/index.html ggplot2 https://ggplot2.tidyverse.org/ N/A HOMER v4.11 Heinz et al.63 http://homer.ucsd.edu/homer/ ( on next page) e2 Developmental Cell 60, 1–18.e1–e8, February 3, 2025
Continued
REAGENT or RESOURCE SOURCE IDENTIFIER IGV 2.12.3 Broad Institute https://github.com/igvteam/igv MACS2 Zhang et al.64 https://doi.org/10.1186/gb-2008-9-9-r137 NIS-Elements 4.30.02 Nikon N/A Prism 9.0.1 GraphPad Software https://www.graphpad.com/ Samtools 1.16 Li et al.65 https://www.htslib.org/ STAR 2.7.6a Dobin et al.66 https://github.com/alexdobin/STAR Novocraft novoalign 4.03.01 N/A https://www.novocraft.com/products/novoalign/ Other BD FACSAria University of Utah Flow Cytometry Core Serial #: P69500142
EXPERIMENTAL MODELS AND STUDY PARTICIPANT DETAILS
Animal studies
Mice harboring KrasFSF-G12D (Young et al.67), Rosa-FSF-CreERT2 (Schönhuber et al.68), Nkx2-1flox (Kusakabe et al.69), Foxa1flox (Gao et al.70), Foxa2flox (Sund et al.71), R26-CAG-LSL-Sun1-sfGFP-myc (Mo et al.72), and p53frt (Lee et al.73), have been previously described. All animals were maintained on a mixed 129/B6 background. All experimental mice were between 2 and 6 months of age at intubation. Mice of both sexes were used throughout each study, though the effect of sex on study results was not assessed. were approved by the IACUC of the University of Utah, conducted in compliance with the Animal Welfare Act Regulations and other federal statutes relating to animals and experiments involving animals, and adhered to the principles set forth in the Guide for the Care and Use of Laboratory Animals, National Research Council (PHS assurance registration number A-3031-01). Husbandry and housing conditions of experimental animals Experimental mice were mated or purchased. Appropriate housing conditions, including temperature, humidity, and light cycles, were maintained, and the mice were provided with a consistent diet. Transmission of transgenic and/or knockout alleles was monitored via DNA isolated from ear biopsy. Animals identified as negative for the presence of relevant alleles were used as control littermates or euthanized. Each mouse was marked by ear tagging and assigned a unique number. To preserve the genetic integrity of mouse models, careful handling practices were followed, and detailed records were kept throughout the duration of the study. Primary 3D organoid cultures All primary murine organoid cultures (see key resources table) were established within Matrigel (Corning or Preclinical Research Shared Resource core facility) submerged in recombinant organoid medium for approximately two weeks (Advanced DMEM/F-12 supplemented with 1X B27 (Gibco), 1X N2 (Gibco), 1.25mM nAcetylcysteine (Sigma), 10mM Nicotinamide (Sigma), 10nM Gastrin (Sigma), 100ng/ml EGF (Peprotech), 100ng/ml R-spondin1 (Peprotech), 100ng/ml Noggin (Peprotech), and 100ng/ml FGF10 (Peprotech). After organoids were established, cultures were switched to 50% L-WRN conditioned media (Miyoshi and Stappenbeck, 2013). Organoid lines were tested periodically for mycoplasma contamination. To maintain organoid cultures mycoplasma free, all culture media were supplemented with 2.5 ug/ml Plasmocin.
METHOD DETAILS
Tumor initiation and tamoxifen administration in vivo Autochthonous lung tumors were initiated by administering viruses via intratracheal intubation. Adenoviral mSPC-FlpO was used to initiate all tumors in this publication. Adenoviruses were obtained from University of Iowa Viral Vector Core. Tumor-specific activation of CreERT2 nuclear activity was achieved by intraperitoneal injection of tamoxifen (Sigma) dissolved in corn oil at a dose of 120mg/kg. Mice received 4 injections over the course of 5 days. One day following injections, mice were given pellets supplemented with 500mg/kg (Envigo) tamoxifen for 7 days.
Histology and immunohistochemistry
All tissues were fixed in 10% formalin overnight and when necessary, lungs were perfused with formalin via the trachea. Organoids were first fixed in 10% formalin overnight and then mounted in HistoGel (Thermo Fisher Scientific). Mounted organoids and tissues were transferred to 70% ethanol, embedded in paraffin, and four-micrometer sections were cut. Immunohistochemistry (IHC) was performed manually on Sequenza slide staining racks (Thermo Fisher Scientific). Sections were treated with Bloxall (Vector Labs) followed by Horse serum 536 (Vector Labs) or Rodent Block M (Biocare Medical), primary antibody, and HRP-polymer-conjugated secondary antibody (anti-Rabbit, Goat and Rat from Vector Labs; anti-Mouse from Biocare). The slides were developed with Impact Developmental Cell 60, 1–18.e1–e8, February 3, 2025 e3 DAB (Vector Labs) and counterstained with hematoxylin. Slides were stained with antibodies to NKX2-1 (1:2000, Abcam, EP1584Y), GFP (1:200, CST, 2956S), FoxA1 (1:4000, Abcam, 10881-14), FoxA2 (1:1200, Abcam, 4466), HNF4a (1:500, CST, C11F12), pro-SPC (1:4000, Abcam, EPR19839), Galectin 4 (1:200, R and D Systems, AF2128), Muc5AC (1:100, Abnova, SPM488), p63 (1:500, CST, D2K8X), Cytokeratin 5 (1:200, Abcam, EP1601Y), Gastrokine 1 (1:50, Abnova, 2E5), pERK1/2 (1:500, CST, D13.14.4E), TFF1 (1:100, Origene, TA382458), and Pepsinogen C (1:100, Sigma, HPA031718). Images were taken on a Nikon Eclipse Ni-Umicroscope with a DS- Ri2 camera and NIS-Elements software. Histological analyses were performed on hematoxylin and eosin-stained and IHC-stained slides using NIS-Elements software. All histopathologic analysis was performed by a board-certified anatomic pathologist (E.L.S.). Establishing primary murine LUAD organoids Five months after tumor initiation in KNF1F2 mice (3D organoid lines; Adeno-mSPC-FlpO), tumor bearing mice were euthanized and lungs were isolated. Individual macroscopic tumors were removed from lungs, minced under sterile conditions, and digested at 37 C for 30 min with continuous agitation in a solution of Advanced DMEM/F12 containing the following enzymes: Collagenase Type I (Thermo Fisher Scientific, 450U/ml), Dispase (Corning, 5U/ml), DNaseI (Sigma, 0.25mg/ml). Enzymatic reactions were stopped by addition of cold Advanced DMEM/F-12 with 10% FBS. The digested tissue was repeatedly passed through a 20-gauge syringe needle, sequentially dispersed through 100mm, 70mm, and 40mm cell strainers, and treated with erythrocyte lysis buffer (eBioscience) to obtain a single cell suspension. Organoid cultures were established by seeding 1x105 tumor cells in 50ul of Matrigel (Corning) and plated in 24-well plates. Matrigel droplets were overlaid with recombinant organoid medium as previously described (Cell lines and primary cultures). Two weeks after organoid establishment, cultures were switched to 50% L-WRN conditioned media. Organoid cultures were screened via immunohistochemistry and qPCR, and lines that uniformly downregulated Nkx2-1 but retained expression of Foxa1 and Foxa2were selected for subsequent analysis. Of note, the 22E organoid line harbors one conditional allele of Trp53 but maintains a well-differentiated gastric identity in the NKX2-1-negative state. In vitro 4-hydroxytamoxifen (4OHT) treatment Cells were transiently treated with 2mM 4-OHT (Cayman Chemical Company, dissolved in 100% Ethanol) or vehicle for 72 (organoid culture) hr to activate CreERT2 nuclear activity and generate isogenic pairs. Generating a single cell suspension from organoid cultures Matrigel droplets containing organoid cells were broken down via repeated pipetting in Cell Recovery Solution (Corning, 500 ml per Matrigel dome). Cell Recovery Solution containing organoids was transferred to sterile conical tubes and submerged in ice for 20– 30 min before centrifugation at 4 C (300-500G). Cell Recovery Solution supernatant was removed and the cell pellet was washed in 1X PBS. Cells were then resuspended in pre-warmed TrypLE Express Enzyme (Thermo Fisher Scientific) and incubated for 10 min at 37 C. TrypLE reaction was quenched via dilution with cold Splitting Media (Advanced DMEM/F-12 [Gibco], 10 mM HEPES [Invitrogen], 1X Penicillin-Streptomycin-Glutamine [Invitrogen]). Cells were centrifuged and then resuspended in a pre-warmed DNase solution (L-WRN media supplemented to a final concentration of 200U/ml DNase [Worthington], 2.5 mM MgCl2, 500 mM CaCl2) and incubated for 10 min at 37 C. Cells were centrifuged and washed in PBS before use.
DNA Methylation sequencing
In vitro Methyl-seq DNA was collected from biological replicates of isogenic organoid cultures, KG1A and 22E. Four Matrigel domes per sample were collected two weeks after 4-OHT or ethanol treatment. Matrigel domes were resuspended directly into Cell Recovery Solution (500 ml per dome) and submerged in ice for 30 min before centrifugation at 4 C (300G). Cells were washed twice in cold PBS and cell pellets were frozen at -80 C. DNA was purified from organoid cell pellets using the DNeasy Blood and Tissue kit according to the manufacturer’s instructions (Qiagen). Library preparation was performed using the NEBNext Enzymatic Methyl-seq Kit. Lambda phage DNA was added as a spike-in control for measuring conversion efficiency. Sequencing was performed using the Illumina NovaSeq 6000 (150 x 150 bp paired-end sequencing, 100 million reads per sample). In vivo Methyl-seq 14 weeks after tumor initiation, K, N, KN and KNF1F2 mice were euthanized. For the time course experiment in Figure S3, KN mice were euthanized 3-days, 7-days, and 2-weeks after tumor initiation. Immediately after euthanasia, the rib-cage was dissected to reveal the trachea and heart. Lungs were perfused with cold PBS, removed, and snap frozen in liquid nitrogen. Flash-frozen lungs were minced on ice in 1 ml of ice-cold Tween with salts and Tris (TST) buffer74 (146 mM NaCl, 10 mM TrisHCL pH 7.5, 1 mM CaCl2, 21 mM MgCl2, 0.01% BSA, 0.006% Tween-20 in ultrapure water) for 5 – 8 min depending on tumor burden. Lysis was quenched with 3 ml of 1X salts and Tris (ST) buffer (146 mM NaCl, 10 mM Tris-HCL pH 7.5, 1 mM CaCl2, 21 mM MgCl2 in ultrapure water) and solution was filtered through a 40 mm cell strainer. Sample was transferred to a 15 ml conical and centrifuged for 5 min at 4 C (300G). Nuclei pellet was carefully resuspended in 1 ml of FACS buffer (1X PBS, 1%BSA, 1% serum, 2 mM EDTA) with protease inhibitors and filtered through a 35 mm cell strainer. Samples were stained with DAPI (Sigma) and evaluated on a fluorescence microscope to assess for GFP positivity and nuclear integrity. Nuclei were sorted on the BD FACSAria with the 100 mm nozzle to obtain e4 Developmental Cell 60, 1–18.e1–e8, February 3, 2025 a GFP-positive, DAPI-positive population. Samples were sorted into 1 ml of cold PBS with 10% serum. After sorting, nuclei were centrifuged for 10 min at 4 C (300G) and pellets were frozen at -80 C. DNA was purified and sequenced as previously described (In vitro Methyl-seq). Methyl-seq data processing and analysis Fastq reads were aligned to themouse genome (build mm10) with Novocraft novoalign (version 4.03.01 https://www.novocraft.com/ products/novoalign/) in bisulfite mode (-b 4) with the following options: set penalty for unconverted CHG or CHH cytosines (option -u) to 12, hard clip 3’ bases to quality 20 (option -H), adapter trimming (option -a), and performance tuning set to ‘NOVOSEQ’. Alignments to non-standard chromosomes were ignored. Extreme coverage regions were identified using depth-normalized (Reads Per Million) mean coverage of replicates andMACS2 bdgpeakcall with an absolute depth cutoff of 2 RPM (mean depth was < 0.1), length 200 bp, and gap 500 bp. Alignments overlapping these exclusion intervals were excluded from further analysis. Duplicate alignments were removed using samtools65 (version 1.16) fixmate and markdup (mode -s). Alignments were processed with USeq NovoalignBisulfiteParser (version 9.3.0) to collect converted and non-converted C counts withminimummapping quality score (-q) of 13 andminimumbase quality score (options -b and -c) of 20. Counts in CpG context were extracted with USeq ParsePointDataContexts using the context ‘‘..CG.’’. Fraction methylation tracks were generated using USeq BisStat requiring a minimum coverage of 4 reads. Measured lambda phage DNA conversion efficiency was typically > 99.7%. BisMark compatible coverage files were generated from the Converted and NonConverted.useq files using the useq2bismark_ methylation_extractor application (https://github.com/tjparnell/HCI-Scripts/blob/master/Methylation/useq2bismark_methylation_ extractor.pl). DMRs were identified following the recommended pipeline detailed in the bsseq33 package. CpG counts were smoothed using BSmooth and default parameters of 70 CpGs and sliding window of 1000 bp. CpGs were subsequently filtered for a minimum coverage of at least 5 reads. For each pairwise comparison, t-statistic values (differences in means) were generated at each CpG. Regions with differential CpGs were identified with dmrFinder using reciprocal quantile cutoffs of 2.5% (rather than absolute thresholds), a minimum number of 5 CpGs, a maximum gap of 200 bp, and an absolute minimum mean fraction methylation differential of 15%. DMRs were annotated to the nearest gene using the ChIPseeker package (citation) and custom annotation consisting of protein-coding, Gencode-basic genes from Ensembl annotation release 102. For comparative analysis betweenmultiple pair-wise comparisons, DMRs were intersected to generate a master list of all possible DMR intervals, and thesewere re-scored formeanCpG fractionmethylation values (non-smoothened) with BioToolBox get_datasets (https://github.com/tjparnell/biotoolbox). To identify FoxA1/2 dependent or independent regions from the merged intervals, differential values were re-calculated from replicate mean values and filtered for the minimum change of 15% in the appropriate direction. Relative difference methylation levels were calculated in heat maps by subtracting the mean methylation value across all replicates. Heat maps were generated with pHeatmap package (citation). Motif analysis was performed on DMRs for both known and de novo motifs using the HOMER package.63 GSEA-Preranked was run on the differentially methylated gene lists generated from bsseq and ranked by their log2 fold change in methylation. DMGs were run against the following MSigDB gene sets: c2, c5, c6, c8, and Hallmarks. Gene sets smaller than 15 and larger than 500 were excluded from analysis. RNA sequencing In vitro RNA-seq RNA was collected from biological replicates of isogenic organoid cultures, KG1A and 22E. Samples were collected two weeks after 4-OHT or ethanol treatment. Three Matrigel domes were collected per sample directly into Trizol, then stored at -80 C until purification. RNA was isolated via Trizol-chloroform extraction followed by column-based purification. The aqueous phase was brought to a final concentration of 35% ethanol, and RNA was purified using the PureLink RNA Mini kit according to the manufacturer’s instructions (ThermoFisher Scientific). Library preparation was performed using the NEBNext Ultra II Directional RNA Library Prep with poly(A) mRNA isolation. Sequencing was performed using the Illumina NovaSeq 6000 (150 x 150 bp paired-end sequencing, 25 million reads per sample). In vivo RNA-seq 14weeks after tumor initiation, K, N, KN, and KNF1F2micewere euthanized and the rib-cagewas dissected to reveal the trachea and heart. Cardiac perfusion of the pulmonary vasculature was performed using PBS until the lungs turned pale. Lungs were then removed, digested, and filtered as previously described (Establishing primarymurine LUAD organoids) to obtain a single cell suspension. Samples were resuspended in FACS buffer with DAPI. Cells were sorted on the BD FACSAria with the 85 mm nozzle to obtain a GFP-positive, DAPI-negative population. Samples were sorted into 1ml of cold PBSwith 10% serum. After sorting, cells were centrifuged for 10 min at 4 C (300G) and resuspended in 1 ml of Trizol. RNA was isolated via Trizol-chloroform extraction and the PureLink RNA Mini kit as previously described (In vitro RNA-seq). Library preparation was performed using the NEBNext Ultra II Directional RNA Library Prep with rRNA Depletion Kit for mouse. Sequencing was performed using the Illumina NovaSeq 6000 (150 x 150 bp paired-end sequencing, 25 million reads per sample). RNA-seq data processing and analysis The mouse GRCm38 genome and gene feature files were downloaded from Ensembl release 102 and a reference database was created using STAR version 2.7.6a.66 Optical duplicates were removed from NovaSeq runs via Clumpify v38.3461 (Bushnelle et al.61). Reads were trimmed of adapters and aligned to the reference database using STAR in two-pass mode to output a BAM Developmental Cell 60, 1–18.e1–e8, February 3, 2025 e5 file sorted by coordinates. Mapped reads were assigned to annotated genes using featureCounts version 1.6.3.75 Raw counts were filtered to remove features with zero counts and features with five or fewer reads in every sample. DEGswere identified using the hciR package (https://github.com/HuntsmanCancerInstitute/hciR) with a 5% false discovery rate and DESeq2 version 1.34.0.62 GSEAPreranked was run with the differential gene list generated from DESeq2 and the following MSigDB gene sets: c2, c5, c6, c8, and Hallmarks. Gene sets smaller than 15 and larger than 500 were excluded from analysis. TCGA analysis of human lung adenocarcinoma We first filtered all TCGA PanCancer Atlas lung adenocarcinoma samples by the presence of KRAS mutations and availability of expression data (n=168/566 LUAD). The 168 KRAS-mutant samples were further filtered on NKX2-1 andHNF4A expression. We first selected patients with above average NKX2-1 expression (z-score>0; n=98/168) and low HNF4A expression (z-score<0, n=64/98) to comprise the pulmonary K tumor cohort. We then selected patients with low NKX2-1 expression (z-score<0; n=56/168) to capture tumor samples that have downregulated NKX2-1. We then filtered this group by high HNF4A expression (z-score>3, n=21/56) to comprise the gastric KN cohort. GSEA-Preranked was ran on human K and KN groups using the gene expression log2 ratio values generated by TCGA for the following MSigDB gene sets: c8 and Hallmarks. RNA-seq and methylation analysis of LUAD PDO samples RNA-seq and methylation-micoarray data from 21 LUAD PDOs, as reported in Ebisudani et al.35 were utilized to profile the transcriptional and methylation landscapes of human LUAD. First, LUAD PDO samples were categorized into ‘pulmonary’ or ‘gastric’ transcriptional states based on their expression of lineage-specific marker genes. Raw counts obtained from Supplemental Table 3 of Ebisudani et al.35 were processed as previously described in the ‘‘RNA-seq data processing and analysis’’ section of methods. PDOs with a pulmonary state were identified by filtering for samples whose expression of both NKX2-1 and SFTPA2 fell within the top quartile of log2 normalized counts for all 21 LUAD cases (n=5/21). Gastric PDOs including potential IMA samples were identified as those having the top quartile of log2 normalized counts for both HNF4A andMUC5AC (n=4/21). Gene methylation levels are shown as M-values as reported in Supplemental Table 4 of Ebisudani et al.35 Methylation analysis of human pancreas, PanIN, and PDAC Whole genome bisulfite sequencing of normal pancreas, PanINs, and PDAC derived from a Lo et al.37 was utilized for downstream functional analysis. GSEA-Preranked was ran on the 3500 differentially methylated genes identified between human pancreatic acinar cells vs. PanINs (as reported in Supplemental Table 4 of Lo et al.37) for the following MSigDB gene sets: c8 and Hallmarks. Genes were ranked by the mean difference in methylation between the two sample types. Fraction methylation tracks of pancreatic acinar cells, ductal cells, PanINs, and PDACs were generated using the same analysis pipeline as previously described in the ‘‘methyl-seq data processing and analysis’’ section of methods.
Chromatin immunoprecipitation sequencing
In vitro organoid ChIP-seq For all organoid ChIP-seq experiments, two 24-well plates (approximately 4 – 8 million cells) were collected in Cell Recovery Solution (500 ml per dome). Organoids were submerged in ice for 30 min before centrifugation at 4 C (300G). Cells were washed three times in cold PBS. On the second wash, PBS was supplemented with DNase solution (containing a final concentration of 200U/ml DNase [Worthington], 2.5 mM MgCl2, 500 mM CaCl2). After the third wash, organoids were resuspended in 5 ml of 2 mM DSG buffer (1X PBS, 1 mM MgCl2) and rotated at room temperature for 35 min. Formaldehyde was then added to a final concentration of 1% and cells were crosslinked for 10 min. The cross-linking reaction was stopped with the addition of glycine to a final concentration of 125 mM. Cells were washed with cold PBS, then frozen at -80. Cell pellets were thawed on ice for 5 min then lysed in 1 ml of Farnham lysis buffer (5 mM PIPES pH 8.0, 85 mM KCl, 0.5% NP40) for all TF experiments or 400 ul of Chromatrap Hypotonic buffer (catalog #100007) for all histone experiments. Samples were centrifuged at 4 C (1000G) then resuspended in 1ml of RIPA lysis buffer (1X PBS, 1% NP40, 0.5% sodium deoxycholate, 0.1% sodium dodecyl sulfate) or 100 ml of Chromatrap Lysis buffer (catalog #100005) for TF and histone ChIPs respectively. All lysis buffers were supplemented with protease inhibitors. Chromatin extract was sonicated with QSonica Q800R (pulse: 30 s on/30 s off; sonication time: 20 min; amplitude: 70%). After sonication, chromatin extract of histone samples was brought up to a volume of 1 ml with RIPA lysis buffer. Chromatin from all samples was immunoprecipitated with antibodies premixed with Protein A Dynabeads: FoxA1 (Abcam, ab170933, rabbit monoclonal, 5 mg/ChIP). Library preparation was performed using the ChIP-seq with NEBNext DNA Ultra II library prep kit using Unique Molecular Indexes (UMIs). Sequencing was performed using the Illumina NovaSeq 6000 (150 x 150 bp paired-end sequencing, 25million reads per sample). For ChIP-qPCR assays, primers were designed targeting FoxA1 binding sites identified from ChIP-seq data on Integrative Genomics Viewer (IGV; https://software.broadinstitute.org/software/igv/). All the primers are listed in Table S6. In vivo nuclei ChIP-seq 14 weeks after tumor initiation, KN and KNF1F2 tumor-bearing mice were euthanized and the rib-cage was dissected to reveal the trachea and heart. Lungs were perfused with cold PBS, removed, and snap frozen in liquid nitrogen. Flash-frozen lungs were minced on ice in 2 ml of ice-cold PBS for 3 – 5 min. Minced lungs were dounced 10 times with a large 7 ml homogenizer (Wheaton, item #23ND78) and centrifuged for 5 min at 4 C (300G). Tissue was crosslinked in 2 mM DSG buffer and 1% Formaldehyde as previously described (In vitro organoid ChIP-seq). After fixation, lung samples were resuspended in 2 – 3ml of TST buffer with protease inhibitors for 5 min. During this time, samples were dounced an additional 5 times in a small 2 ml homogenizer (Wheaton, item #23ND70) to extract nuclei. Lysis was quenched with 5 ml of 1X ST buffer plus protease inhibitors. Nuclei were washed in cold PBS and e6 Developmental Cell 60, 1–18.e1–e8, February 3, 2025 resuspended in 5 – 10ml cold PBS supplemented with DAPI and protease inhibitors. Nuclei suspension was then sequentially filtered through a 70 mm and 35 mm cell strainer. Before sorting, nuclei were evaluated on a fluorescence microscope to assess for GFP positivity and nuclear integrity. Nuclei were sorted using BD FACSAria with the 85 mmnozzle for GFP-positive, DAPI-positive nuclei. Samples were sorted into 1ml of cold PBSwith 1%BSA and 10X protease inhibitor. Approximately 3million nuclei were sorted for histone ChIP-seq and 10 million for TF ChIP-seq experiments. After sorting, nuclei were pelleted for 10 min at 4 C (500G). Samples were then resuspended in Chromatrap Hypotonic and Lysis buffers as previously described (In vitro organoid ChIP-seq) and sonicated with QSonica Q800R (pulse: 30 s on/30 s off; sonication time: 20 min; amplitude: 70%). Chromatin was immunoprecipitated with antibodies premixed with Protein A Dynabeads: FoxA1 (Abcam, ab170933, rabbit monoclonal, 5 mg/ChIP), FoxA2 (CST, D56D6, rabbit monoclonal, 5 mg/ChIP), TET2 (CST, D6C7K, rabbit monoclonal, 2.5 mg/ChIP), TET3 (Sigma, ABE290, rabbit polyclonal, 5 mg/ChIP) and H3K27ac (Active Motif, 39133, rabbit polyclonal, 5 mg/ChIP). ChIP-seq libraries were prepped and sequenced as previously described (In vitro organoid ChIP-seq). For ChIP-qPCR assays, primers were designed targeting FoxA1, FoxA2, TET2, TET3, and H3K27ac binding sites identified from ChIP-seq data on IGV. All the primers are listed in Table S6. ChIP-seq data processing and analysis Fastq alignments were pre-processed with the merge_umi_fastq application from the UMIScripts package (https://github.com/ HuntsmanCancerInstitute/UMIScripts) to associated the UMI sequence, provided as a third Fastq file, into the read comment. Reads were aligned using Bowtie276 to the standard chromosomes of the mouse genome (version mm10). Duplicate alignments based on the UMI code were removed using the bam_umi_dedup application (UMIScripts) allowing for 1 mismatch. Peaks were called using MACS264 (v2.2.6 https://github.com/macs3-project/MACS) with a significance of q-value < 0.05. Coverage tracks were generated with MACS2 as Reads Per Million. Input libraries were obtained from all tumor and organoid samples and were used as controls for each ChIP-seq experiment. All ChIP-seq experiments were performed in biological duplicates. Genomic annotation of binding sites was performed using the HOMER package.63 Overlaps between peaks were determined using Bedtools60 (v2.28.0 https://bedtools. readthedocs.io) with a 1-bp minimum overlap, and percentages were calculated by dividing the number of overlapping peaks by the number of peaks in the smaller set (i.e., the percentage of maximal possible overlap). For TET2 ChIP-seq in KN tumors, all called peaks were retained from each replicate for downstream analysis. Motif finding was performed on 100-bp regions surrounding the summit of identified peaks. Both known and de novo motifs were identified using the Homer package.63 Differential ChIP-seq peaks were identified using the Diffbind package (https://bioconductor.org/packages/release/bioc/vignettes/DiffBind/inst/doc/ DiffBind.pdf) with a q-value cutoff < 0.05. Motif finding for differential ChIP-seq peaks was performed by extending the differential region to a length of 1000 bps. HOMER motif analysis was then performed on the extended differential region to identify known and de novo motifs. Heat maps and plots were generated with custom R scripts using pHeatmap (https://cran.r-project.org/ package=pheatmap) and ggplot2 (https://ggplot2.tidyverse.org). Hi-C chromatin immunoprecipitation (HiChIP) Tumor processing, fixation, and sortingwas performed as previously described (‘‘In vivo nuclei ChIP-seq’’). HiChIP was performed as described (https://www.nature.com/articles/nmeth.3999) with minor modifications as described (https://www.nature.com/articles/ s41467-021-27055-4). Cross-linked chromatin was digested with the MboI restriction enzyme followed by end-repair with dNTPs including biotin labeled dATP, ligation using T4 DNA ligase, and sonication to obtain 1kb chromatin fragments using Qsonica (Q800). To enrich for chromatin interactions occurring at active regulatory elements, an anti-H3K27ac antibody (Abcam, ab4729, rabbit polyclonal, 7.5ug/HiChIP) was used for DNA fragment capture. Streptavidin magnetic beads were used to pull down ligated DNA fragments and HiChIP libraries were prepared using Illumina Tagment DNA Enzyme and Buffer Kit. Sequencing was performed with Illumina NextSeq. The paired-end HiChIP sequencing reads were aligned to the mouse genome mm10 with the HiC-Pro pipeline (https:// genomebiology.biomedcentral.com/articles/10.1186/s13059-015-0831-x). HiC-DC+ was used to call chromatin loops by binning the genome into 5kb bins (https://www.nature.com/articles/s41467-021-23749-x#citeas). Chromatin loops were removed if the qvalue < 0.05 or anchors overlapped with ENCODE blacklist regions (https://www.nature.com/articles/s41598-019-45839-z). BEDTools pairToBed was used to identify enhancer to promoter interactions (https://academic.oup.com/bioinformatics/article/ 26/6/841/244688). HiChIP data was visualized at the resolution of 5kb using the R package gTrack (https://github.com/mskilaborg/gTrack). Lentiviral production and transduction HEK293T cells were transfected with doxycycline-inducible TRE-FoxA2 lentiviral vector, d8.9 packaging vector and VSV-G envelope vectormixedwith TransIT-293 (Mirus Bio). Virus-containing supernatant was collected 48, 60, and 72 h after transfection, centrifuged to pellet floating HEK293T cells, and filtered using 0.45mm filters before storing long term at -80 C. For stable transduction of organoids, organoid cultures were first prepared into single cell suspensions by subjecting them to successive incubations with Cell Recovery Solution (Corning) and TrypLE (Gibco). Cells were then resuspended in a 1:1 mixture, by volume, of 50% L-WRN and lentivirus containing supernatant. After supplementation with 8 mg/ml polybrene, cells were incubated for 24 hr. Cells were then pelleted, mixed back with Matrigel, and seeded. 72 hr later, Blasticidin selection for 1 week was performed to achieve stable lines. Developmental Cell 60, 1–18.e1–e8, February 3, 2025 e7
QUANTIFICATION AND STATISTICAL ANALYSIS
All graphing and statistical analysis was performed with PRISM software, with all graphs showing mean and standard deviation. The statistical details can be found in the corresponding figure legend. All NGS statistical analysis was performed according to published pipeline protocols cited, with a statistical significance cutoff of padj<0.05. e8 Developmental Cell 60, 1–18.e1–e8, February 3, 2025
Article Images (0)
×![]()
Bioz is the world’s most comprehensive AI search engine for scientific experimentation. The Bioz search engine offers researchers billions of data-driven product, technique, and protocol recommendations. Bioz taps into the latest advances in Artificial Intelligence (AI) – including Natural Language Processing (NLP), Machine Learning (ML) and Generative AI to mine and structure hundreds of millions of pages of complex and unstructured scientific papers. By harnessing the power of AI, Bioz provides researchers with an unprecedented amount of summarized scientific experimentation knowledge right at their fingertips, ultimately helping to speed up drug discovery and the rate of success in finding cures for diseases. Bioz is used by over 15 Million researchers from over 17,000 different universities and companies in 196 countries.
