Notch1 mutations drive clonal expansion in normal esophageal epithelium but impair tumor growth

NOTCH1 mutant clones occupy the majority of normal human esophagus by middle age but are comparatively rare in esophageal cancers, suggesting NOTCH1 mutations drive clonal expansion but impede carcinogenesis. Here we test this hypothesis. Sequencing NOTCH1 mutant clones in aging human esophagus reveals frequent biallelic mutations that block NOTCH1 signaling. In mouse esophagus, heterozygous Notch1 mutation confers a competitive advantage over wild-type cells, an effect enhanced by loss of the second allele. Widespread Notch1 loss alters transcription but has minimal effects on the epithelial structure and cell dynamics. In a carcinogenesis model, Notch1 mutations were less prevalent in tumors than normal epithelium. Deletion of Notch1 reduced tumor growth, an effect recapitulated by anti-NOTCH1 antibody treatment. Notch1 null tumors showed reduced proliferation. We conclude that Notch1 mutations in normal epithelium are beneficial as wild-type Notch1 favors tumor expansion. NOTCH1 blockade may have therapeutic potential in preventing esophageal squamous cancer.

NOTCH1 mutant clones occupy the majority of normal human esophagus by middle age but are comparatively rare in esophageal cancers, suggesting NOTCH1 mutations drive clonal expansion but impede carcinogenesis. Here we test this hypothesis. Sequencing NOTCH1 mutant clones in aging human esophagus reveals frequent biallelic mutations that block NOTCH1 signaling. In mouse esophagus, heterozygous Notch1 mutation confers a competitive advantage over wild-type cells, an effect enhanced by loss of the s ec ond a llel e. W id es pr ead Notch1 loss alters transcription but has minimal effects on the epithelial structure and cell dynamics. In a carcinogenesis model, Notch1 mutations were less prevalent in tumors than normal epithelium. Deletion of Notch1 reduced tumor growth, an effect recapitulated by anti-NOTCH1 antibody treatment. Notch1 null tumors showed reduced proliferation. We conclude that Notch1 mutations in normal epithelium are beneficial as wild-type Notch1 favors tumor expansion. NOTCH1 blockade may have therapeutic potential in preventing esophageal squamous cancer.
Aging tissues accumulate somatic mutations [1][2][3][4] . Some mutations confer a competitive advantage on progenitor cells, which may form mutant clones that colonize normal tissue. These clonal expansions are often associated with mutations linked to cancer and may represent the first step in malignant transformation 4 . However, the under-representation of NOTCH1 mutants in esophageal cancer compared with normal aging epithelium suggests NOTCH1 mutations may inhibit malignant transformation 2,5 .
NOTCH1 is a cell surface receptor composed of an extracellular domain (NEC) and a transmembrane and cytoplasmic subunit (NTM), interacting noncovalently through the negative regulatory region (NRR; Extended data Fig. 1a) 6,7 . The NRR comprises three Lin12-Notch repeats (LNR) and a heterodimerization domain (HD) that inhibits NOTCH1 activation in the absence of ligand 8 . Ligands bind to conserved epidermal growth factor (EGF) repeats in the NEC. This results in proteolytic cleavage events releasing the intracellular domain (NICD), which translocates to the nucleus and alters target gene transcription 8 . In the esophagus, NOTCH1 protein is expressed in proliferating cells and regulates both development and adult tissue maintenance (Extended data Fig. 1a,b) 9 .
Different studies have suggested that NOTCH1 is a tumor suppressor or conversely may promote esophageal carcinogenesis [10][11][12] . Here we Article https://doi.org/10.1038/s41588-022-01280-z cell from the basal layer, ensuring basal cell density is kept constant 14 . Dividing progenitors generate either two progenitor daughters, two differentiating daughters or one cell of each type. In wild-type tissue, the probabilities of each progenitor outcome are balanced, generating equal proportions of progenitor and differentiated cells, maintaining cellular homeostasis (Extended data Fig. 2a) 15,16 . Mutations that alter progenitor fate leading to excessive production of progenitors drive mutant clone growth 17,18 .
For lineage tracing, we generated AhCre ERT Rosa26 floxedYFP Notch1 flox triple transgenic (YFPCreNotch1) mice. These animals carry a conditional Notch1 allele and a genetic labeling system. An inducible Cre recombinase (AhCre ERT ) was used to delete one or both conditional Notch1 alleles in Notch1 wt/flox or Notch1 flox/flox animals and induce a separate conditional yellow fluorescent protein (YFP) reporter allele (Rosa-26 floxedYFP ) 15,19 . YFP was expressed in recombined epithelial cells and their progeny (Extended data Fig. 2b,c). This model was induced at low dose to recombine scattered single basal cells (clonal induction) or at a higher level to recombine a large proportion of basal cells (high induction) (Extended data Fig. 2c,d).
Excision of the Notch1 allele and expression of the YFP reporter at the Rosa26 locus can occur in combination or separately, resulting in Notch1 mutant or wild-type cells expressing YFP or not (Extended data Fig. 2c,d). We confirmed the recombination status of exon 1 of Notch1 of wild type and fully recombined Notch1 +/− and Notch1 −/− esophageal epithelium. Notch1 mRNA and protein expression was halved in Notch1 +/− and abolished in Notch1 −/− cells compared with wild-type keratinocytes (Extended data   Table 6). We then performed genetic lineage tracing by inducing recombination in scattered single progenitors in YFPCreNotch1 +/+ , YFPCreNotch1 +/flox or YFPCreNotch1 flox/flox mice. YFP-expressing clones were detected by imaging sheets of epithelium stained for YFP and NOTCH1 (Fig. 2a). YFP + Notch1 +/− or YFP + Notch1 −/− clones were identified from reduced intensity or absence of NOTCH1 immunostaining, respectively, a method validated by detecting Notch1 recombination in microdissected clones (Fig. 2b,. The number and location of cells in YFP-expressing clones of each genotype were determined by 3D confocal imaging. The size of YFP + Notch1 +/− clones was substantially increased compared to wild-type YFP + Notch1 +/+ clones at all time points. YFP + Notch1 −/− clones were larger still Extended data Fig. 3i,j and Supplementary Table 7). To examine the cellular mechanisms underlying mutant clonal expansion, we used short-term cell tracking by labeling cycling cells with the S phase probe 5-ethynyl-2′-deoxyuridine (EdU).
We first counted the proportion of basal cells positive for EdU at 1 h after labeling, which measures the fraction of cells in S phase (Fig. 2e,f). This value was similar for cells within Notch1 +/− clones and wild-type cells distant from clones (Fig. 2g and Supplementary Table 8). Within Notch1 −/− mutant clones, the proportion of EdU + basal cells was investigate how NOTCH1 mutants colonize the epithelium, their impact on tissue maintenance and their effect on esophageal carcinogenesis 2,4 .

NOTCH1 mutant clones in human esophagus
Deep targeted sequencing studies have revealed numerous NOTCH1 mutants in human esophagus but have not visualized clones and resolved which NOTCH1 mutation(s) or copy number alterations they carry 2,4 . To achieve this, histological sections of normal epithelium from elderly donors were immunostained for NOTCH1 (Fig. 1a). Positive and negative staining areas were microdissected and targeted sequencing for 322 genes associated with cancer was performed (Fig. 1b). A total of 247 protein-altering somatic variants were identified across 86 samples from six donors aged . The predominant mutant genes were NOTCH1, TP53 and NOTCH2 (refs. 2,4 ; Supplementary Tables 1-3 and Supplementary Note). Near clonal NOTCH1 mutations with an average variant allele frequency (VAF) of 0.36 were detected in 81% (70/86) of samples (Fig. 1c,d). Ninety-three percent (25/27) of negative staining areas carried nonsense, essential splice mutations or indels in NOTCH1 with copy neutral loss of heterozygosity (CNLOH) of the NOTCH1 locus (human 388,440,238) or a further mutation, likely to disrupt the second NOTCH1 allele (Fig. 1d,e). Fifty-nine percent (35/59) of positively stained samples carried a missense NOTCH1 mutation and most of these had either CNLOH or a second mutation Extended data Fig. 1c,d and Supplementary Table 4). Overall, most samples (73%, 51/70) had likely biallelic NOTCH1 alterations (Fig. 1f). To test if the mutations disrupted NOTCH1 function, we stained consecutive sections from additional donors for NOTCH1 protein and NICD1, which is detectable in the nucleus during active signaling (Fig. 1g,h,Extended data Fig. 1a and Supplementary Table 5) 13 . The proportion of epithelium with active NOTCH1 decreased with age (Kendall's tau-b = −0.67, P = 0.014). In older donors, in whom NOTCH1 mutations are common, NOTCH1 − areas were associated with NICD1 loss. We also found occasional NOTCH1 + NICD1 − areas, consistent with the presence of missense mutant proteins that reach the cell membrane but lack signaling activity (Fig. 1g,h). NICD1 + and NICD1 − areas were histologically undistinguishable, with no significant differences in tissue thickness, cell density or the expression of the proliferation marker Ki67 (Extended data . We conclude that many NOTCH1 mutant clones in aging human esophagus carry biallelic alterations that disrupt signaling.

Notch1 mutations increase clonal fitness
To investigate how NOTCH1 mutant clones colonize normal epithelium, we tracked the fate of Notch1 mutant clones in transgenic mice using lineage tracing. Mouse esophageal epithelium consists of layers of keratinocytes. Proliferation is restricted to progenitor cells in the basal layer (Extended data Fig. 2a). Differentiating cells cease dividing, leave the basal layer and migrate toward the epithelial surface where they are shed. Progenitor division is linked to the exit of a nearby differentiating Article https://doi.org/10.1038/s41588-022-01280-z marginally lower than in wild-type cells (Fig. 2h). We conclude neither Notch1 +/− nor Notch1 −/− clonal expansion results from an increase in mutant cell division rate compared with wild-type cells.
A 48 h EdU experiment labeled S phase cells and tracked the fate of the two cells generated by the subsequent mitosis over the following 48 h. The pair of labeled cells may remain in the basal layer, or   Article https://doi.org/10.1038/s41588-022-01280-z one or both may differentiate and exit the basal layer (Fig. 2i,j). The ratio of EdU-labeled suprabasal cells to the total EdU-labeled cells reflects the rate of production of differentiating cells in the basal layer and their stratification into the suprabasal layers. In Notch1 +/− and Notch1 −/− clones, this ratio is decreased, consistent with a tilt in mutant progenitor cell fate, so that more progenitors and fewer differentiating daughters are produced per average cell division (Fig. 2k). Strikingly, adjacent to Notch1 −/− clones, there was an increase in the suprabasal EdU + :total EdU + cell ratio in the wild-type cells at the clone margin compared with wild-type cells further from the mutant clone (Fig. 2j,l). This, along with a small decrease in the proportion of wild-type S phase cells at the clone edge, indicates that wild-type cells adjacent to the clone exit the cell cycle, differentiate and exit the basal layer at an increased rate, a phenomenon also reported in previous studies of Notch inhibited keratinocytes interacting with wild type cells (Fig. 2h) 18,20 . These observations explain the increased fitness of Notch1 −/− over Notch1 +/− clones. Cell density was similar in both mutant genotypes and wild-type areas, suggesting that the linkage between cell division and the exit of a nearby differentiating cell from the basal layer is maintained (Fig. 2m,n). Within this constraint, the driving of wild-type cell differentiation and stratification permits Notch1 −/− cell division at the clone edge, accelerating clonal expansion (Fig. 2o).
These observations were integrated into a Wright-Fisher style quantitative model in which fit mutant clones expand until they collide with other mutant clones of similar fitness, at which point they revert to neutral competition 21 . We fitted this model to the clone size data. The inferred fitness for Notch1 +/− clones was higher than that of wild-type cells and the inferred fitness of Notch1 −/− clones markedly greater than that of heterozygous clones (Extended data Fig. 4a-d, Video 1 and Supplementary Note).

Notch1 haploinsufficiency enables epithelial colonization
Clones generated by the transgenic deletion of Notch1 alleles may not reflect the behavior of Notch1 mutants that appear during aging. We therefore investigated spontaneous Notch1 mutant clones in control YFPCreNotch1 +/+ mice, and the heterozygous epithelium of highly induced YFPCreNotch1 +/flox animals. Both strains were aged before immunostaining the epithelium for NOTCH1 (Fig. 3a). The area of epithelium stained negative for NOTCH1 increased progressively to 12% of Notch1 +/+ and 78% of Notch1 +/− epithelium by 65 weeks (Fig. 3b,c and Supplementary Table 9). Widespread loss of NICD1 staining was seen in aged Notch1 +/− tissue (Extended data Fig. 5a,b). These observations suggest that, as in humans, Notch1 mutants colonize the aging mouse esophagus and that selection is enhanced in Notch1 +/− epithelium.
To localize potential clones, we stained for NOTCH1 and the YFP reporter. Aging Notch1 +/− epithelium contained multiple ovoid areas of homogenous NOTCH1 staining, positive or negative for YFP but far larger than most YFP labeled clones (Fig. 3d,e). These were suggestive of clonal expansion. A total of 246 such 'expanded' areas along with typical 'nonexpanded' regions were dissected and underwent targeted sequencing for 73 Notch pathway and cancer-related genes (Supplementary Tables 1, 10 and 11). We analyzed for CNLOH and mutations with VAF ≥ 0.2, as below this threshold mutations were considered unlikely to drive clonal expansion. Nintey-seven percent (180/185) of the 'expanded' areas had either Notch1 protein-altering mutations with VAF ≥ 0.2 or CNLOH involving the Notch1 locus (GRCm38-chr2: 26,457,503,822). In contrast, only 2 of 61 nonexpanded areas carried Notch1 mutations and none had Notch1 CNLOH (Fig. 3f,Extended data Fig. 5c,d and Supplementary Table 10). Only a few mutations in other genes were found, some may have been passengers within a Notch1 mutant clone. Ninety-four percent (169/180) of expanded areas with Notch1 altering events carried only a single event (about 50% one Notch1 protein-altering mutation and the remainder CNLOH) with an average VAF 0.44, consistent with them being clones carrying spontaneous changes affecting the nonrecombined Notch1 allele Supplementary Tables 10,11 and Extended data Fig. 5c,d). Among clones carrying a Notch1 mutation, 85% of those stained positive for NOTCH1 + harbored missense mutations while NOTCH1 negatively stained clones carried mainly indel/splicing (51%) or nonsense mutations (46%) (Fig. 3g). Overall, these results were consistent with findings in aging human esophagus (Fig. 1).
Collectively these observations reveal that haploinsufficiency is key for the normal esophagus to be colonized so effectively by Notch1 mutants. Neutral mutants do not colonize the tissue 15,28 . Loss of one allele biases mutant progenitor cell fate toward the production of progenitors, increasing the likelihood that mutant clones will expand and persist in the epithelium (Extended data Fig. 4e,f and Video 2). Notch1 inactivated cells have a further increased fitness so that subclonal loss of the second allele within a persisting heterozygous clone will generate cells that outcompete both Notch1 +/+ and Notch1 +/− neighbors (Fig. 3h). This model explains the high prevalence of clones with NOTCH1 mutation and CNLOH in aging human esophagus.

Notch1 −/− epithelium has minimal phenotype
Epithelium lacking functional NOTCH1 might be expected to have a cellular phenotype. To explore the effects of Notch1 loss in the mouse esophagus, we first performed bulk RNA sequencing (RNA-seq) on  showing the spontaneous appearance of expanding NOTCH1 − cells (black) with aging, possibly caused by genetic events affecting the Notch1 locus. e, Highly induced YFPCreNotch1 +/flox mice were aged 54-78 weeks old, when esophageal epithelium was collected and stained for NOTCH1 (magenta), YFP (green) and DNA (blue). Expanding areas devoid or fully stained with YFP appeared distinct from normal-appearing areas marked with a patchwork of small YFP + clones. Expanded NOTCH1 − (yellow) and NOTCH1 + (orange) areas and normalappearing areas (blue) were isolated for targeted sequencing (n = 246 biopsies from ten mice    YFPCreNotch1 flox/flox mice were treated with DEN and SOR. Tissue was collected 28 weeks after treatment. Tumors were dissected from underlying submucosa and normal epithelium was cut into a gridded array of 2 mm 2 samples before targeted sequencing. Scale bar, 1 mm. b, Number of Notch1 mutations per amino acid is plotted by NOTCH1 protein domains in normal gridded biopsies (upper) and tumors (lower) from Notch1 wild type mice (normal, n = 115 biopsies from six mice; tumors, n = 17 biopsies from seven mice). Domains: EGF-like repeats, LNR, HD, TM, transmembrane, RAM, RBP-J-associated module, ANK, ankyrin repeats, TAD, trans-activation domain, PEST, rich in proline, glutamate, serine and threonine. c, dN/dS ratio for Notch1 mutations (top plot) and proportion of Notch1 mutant tissue in normal epithelium (purple bars) (n = 115 biopsies from six mice) and tumors (n = 17 biopsies from seven mice). Two-tailed P value, likelihood ratio test of dN/dS ratios 2 . d, Representative NOTCH1 (magenta) and KRT14 staining (green) in tumors and surrounding tissue, DNA is blue. Image typical of 10 tumors from six animals. White dashed lines delineate tumor from adjacent normal tissue. Scale bars, 250 µm. e, Proportion of NOTCH1 + staining area in normal epithelium and tumors from the same control animals (each dot represents a mouse, n = 40 tumors from four mice). Two-tailed paired Student's t test. f, Representative images showing nuclear NICD1 (magenta) in keratinocytes (KRT14, green) inside a tumor in comparison to the normal adjacent tissue. DNA is blue. Image typical of 10 tumors from six animals. Scale bars, 25 µm. g, Proportion of KRT14 + keratinocytes with nuclear NICD1 staining in tumors and surrounding epithelium in the same sections (each dot represents a tumor, n = 10 tumors from six mice). Two-tailed paired Student's t test. See Supplementary  Article https://doi.org/10.1038/s41588-022-01280-z To phenotype fully colonized Notch1 −/− epithelium, we performed single-cell RNA-seq (scRNA-seq) on highly induced YFPCreNotch1 flox/flox and uninduced control mouse esophagus (Fig. 4a-i, Extended data 7a-k and Supplementary Table 16). After filtering out poor-quality cells, a total of 13,111 cells remained for analysis, from two biological replicates per genotype (Fig. 4b and Supplementary Note). The proportions of keratinocytes, fibroblasts, immune and endothelial cells were similar in both genotypes, confirmed by staining esophageal sections (Fig. 4c and Extended data Fig. 7b-d) 31 . Keratinocytes showed no significant difference in density in Uniform Manifold Approximation and Projection (UMAP) space between the two genotypes ( Fig.4d and Supplementary Note). The analysis revealed a continuum of keratinocyte cell states, from progenitors expressing Krt14 to differentiating cells expressing Krt4 or Tgm3 to cornified cells expressing Lor (Extended data Fig. 7h-k).
We used these markers to discriminate basal and suprabasal cells in UMAP space, finding similar proportions of both populations in control and Notch1 −/− epithelium (Fig. 4e and Supplementary Note). In a further analysis, we assigned keratinocytes to cycling basal, resting basal or differentiating cells, finding no substantial differences between genotypes 32 .
To validate the scRNA-seq findings, we performed a cell-tracking assay. Mice with Notch1 −/− esophageal epithelium and littermate controls were injected EdU and 5-bromo-2′-deoxyuridine (BrdU) at 48 h and 1 h, respectively, before collection (Extended data Fig. 7l). Staining for EdU revealed the fate of S phase cells over the following 48 h, BrdU + cells were currently in S phase. Cells were also stained for phospho-Histone H3 (pHH3), a G2/M phase marker (Extended data Fig. 7m). The ratio of suprabasal EdU + :total EdU + cells reflecting the Article https://doi.org/10.1038/s41588-022-01280-z generation of differentiating cells and their stratification, the proportion of BrdU + basal cells and the percentage of pHH3 + , BrdU − basal cells were all similar in wild type and Notch1 −/− epithelium, consistent with the scRNA-seq findings (Extended data Fig. 7n- We also examined the epithelium in induced YFPCreNotch1 flox/flox mice and control littermates that were aged 52 weeks. Tissue thickness, basal cell density and expression of the differentiation markers KRT14, KRT4 and LOR and the proliferation marker Ki67 were similar in both genotypes, . Pulse labeling and short-term lineage tracing for 48 h with EdU confirmed no significant difference in the proportion of S phase cells or in the stratification of differentiating cells, respectively, between Notch1 −/− and wild-type esophagus (Fig. 5e-h).
We conclude that once Notch1 −/− cells have occupied the epithelium, their behavior reverts toward that of wild-type cells so that tissue integrity is maintained.

Notch1 loss slows tumor growth
Next, we explored the role of Notch1 in esophageal carcinogenesis. We began by treating YFPCreNotch1 wild-type mice with the mutagen diethylnitrosamine (DEN), and sorafenib (SOR), a protocol that generates high-grade dysplastic lesions 33 . Tissue was collected after aging 28 weeks (Fig. 6a). Deep targeted sequencing of 73 cancer-associated and Notch pathway genes was performed on macroscopic tumors and a gridded array of normal epithelium .
The normal epithelium contained a high density of clones carrying protein-altering mutations. To determine which genes conferred a clonal advantage, we calculated the ratio of silent to protein-altering mutations in each gene, dN/dS 3,34 . Mutant genes under positive selection with a dN/dS ratio substantially above 1 (q < 0.05) were the Notch pathway genes Notch1, Notch2 and Adam10, plus Fat1, Trp53 and Arid1a, all of which are selected in normal human esophagus along with Ripk4 and Chuk (Supplementary Table 21) 2,21 .
In tumors, the most prevalent mutant gene was the known mouse esophageal tumor driver Atp2a2, which is not selected in normal epithelium (Extended data Fig. 8a, 35,36 . Protein-altering Notch1 mutations were under weaker selection and less prevalent in tumors than in the adjacent epithelium (Fig. 6b,c,Extended data Fig. 8a,. Immunostaining confirmed more cells stained positive for NOTCH1 and NICD1 in tumors than in normal tissue . These findings parallel observations in humans and indicate Notch1 wild-type cells are more likely to contribute to tumors than those carrying Notch1 mutations 2,5 . Next, we used a high induction protocol to delete one or both alleles in the entire esophageal epithelium of YFPCreNotch1 flox/flox and YFPCreNotch1 +/flox mice before DEN and SOR treatment. Uninduced littermates were used as controls (Fig. 7a). The density of tumors was similar in all three genotypes, arguing Notch1 is not required for tumor initiation (Fig. 7b,c and Supplementary Table 24). However, tumors were significantly smaller in Notch1 −/− epithelium, in which immunostaining confirmed the loss of Notch1 expression and function . Immunostaining for markers of differentiation (LOR, ITGA6 and KRT14) showed multiple layers of undifferentiated keratinocytes in lesions of both genotypes. Markers of apoptosis (cleaved caspase 3), endothelial cells (CD31) and immune cells (CD45) were also similar in tumors from Notch1 −/− and Notch1 +/+ epithelium (Fig. 7e,f and Extended data Fig. 8d,e). CDH1 loss contributes to tumorigenesis 37 . Tumors from Notch1 +/+ , but not Notch1 −/− , esophagus displayed focal loss of CDH1 expression (Extended data Fig. 8f-h and Supplementary Table 24).
These observations argue that Notch1 favors tumor growth. To test this hypothesis, we treated wild-type mice with a NOTCH1 function blocking antibody (anti-NRR1.1E3) 38 . The antibody reduced levels of cleaved NOTCH1 in esophageal epithelium, abolished nuclear NICD1 immunostaining and altered levels of multiple transcripts encoding  Table 25). Wild-type mice were given DEN and SOR, tumors allowed to develop for 9 weeks and anti-NRR1.1E3 or control antibody given for 6 weeks (Fig. 7g). Anti-NRR1.1E3 significantly reduced tumor size compared with control, indicating NOTCH1 signaling favors the growth of established lesions (Fig. 7h,i and Supplementary Table 24).

Discussion
These results shed light on the disparity in the prevalence of NOTCH1 mutations in normal esophageal epithelium and tumors 2,5 . Mutations reducing the function of one Notch1 allele confer a competitive advantage on mutant progenitors, making it likely they will form persistent, expanding clones. As the heterozygous mutant population grows, the probability that the remaining allele will be lost increases. When this happens, it a b g Tumors vs adjacent normal Article https://doi.org/10.1038/s41588-022-01280-z confers a further increase in fitness (Fig. 3h). By driving wild-type cell differentiation, Notch1 null cells at the clone margins can divide, resulting in extensive colonization of the epithelium (Fig. 2o). This mechanism explains how clones with biallelic NOTCH1 disruption dominate normal human esophagus. Such 'supercompetition' also occurs in the intestine where Apc mutant intestinal stem cells drive the differentiation of their wild-type neighbors to colonize the intestinal crypt 41 .
Once an area has been colonized by biallelic Notch1 mutants, the phenotype of mutant cells reverts toward that of wild-type cells. This reversion toward a near-normal cell state explains the normal appearance of aged human esophageal epithelium despite NOTCH1 signaling being disrupted in most of the tissue.
In Atp2a2 mutant tumors, the constraint that links cell division to the exit of differentiating cells from the basal cell layer to maintain cellular homeostasis does not operate. In this context, the faster cells divide, the faster the lesion will expand. As loss of Notch1 slows the cell division rate, Notch1 −/− lesions are smaller than wild-type tumors (Fig. 8k).
Might these findings be relevant to humans? Over 90% of human esophageal squamous cell carcinoma (ESCC) retain one or more wild-type copies of NOTCH1 but develop from epithelium where a high proportion of cells have biallelic NOTCH1 disruption, arguing wild-type NOTCH1 favors ESCC development. What of the subset of ESCC that does have biallelic NOTCH1 disruption? 5 One possibility is that NOTCH1 loss, in association with multiple other genomic alterations, promotes transformation in these cases. Alternatively, it is plausible that the NOTCH1 alterations in these tumors are 'passengers', carried over from normal tissue with the requirement for wild-type NOTCH1 in carcinogenesis bypassed by other genome changes.
Notch1 illustrates how inactivating mutations in the same gene can drive clonal expansion in normal tissue but impair tumor growth. This is due to the differences in cell dynamics between wild-type normal tissue and a mutated tumor. Our results highlight the potential of NOTCH1 blockade in reducing the growth of premalignant tumors. NOTCH1 inhibitors are in clinical development, and investigation of their potential in esophageal neoplasia seems warranted.

Online content
Any methods, additional references, Nature Portfolio reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at https://doi.org/10.1038/s41588-022-01280-z.  Fig. 2b-d) 15,19,42 . C57BL/6J wild-type mice were also used as indicated.

qPCR recombination assay
Design of the assay. Specific primer sets were designed to analyze excision of the floxed exon 1 of Notch1 by Cre recombinase (Extended data Fig. 3c). Primer set A allows intragenic normalization using the nonfloxed Notch1 exon 3; primer set B measures the disappearance floxed exon 1 with recombination; primer set C specifically detects exon 1 recombination (primer sequences are provided in Supplementary Table 32). Quantitative PCR on genomic DNA was carried out using specific primers and SYBR Green master mix (Thermo Fisher Scientific, 4309155) according to the manufacturer's instructions in a StepOnePlus Real-Time PCR System (Thermo Fisher Scientific, 4376600). Relative qPCR expression was calculated using delta-delta Ct method, a wild type or Notch1 −/− reference sample was used within the same assay for set B or set C, respectively. Validation of the linearity of the recombination assay was performed against a standard curve reproducing different recombination rates with Exon 1/Exon 3 ratios of 1, 0.75, 0.5, 0.25 and 0. The standard curve was made using diluted genomic DNA from the esophagus of highly induced and fully recombined Notch1 −/− mice (as verified by qPCR, staining and protein assay) and from Notch1 wild-type tissue.
Recombination status in highly induced tissues. Genomic DNA was extracted from large pieces of freshly peeled epithelium using either AllPrep DNA/RNA mini kit (Qiagen) or QIAamp DNA micro kit (Qiagen, 56304) and qPCR assay was performed using set B. Full recombination of the esophageal epithelium will reduce the Exon 1/Exon 3 ratio to zero in induced Notch1 −/− mice and halve it in induced Notch1 +/− mice compared to wild-type mice.

Detection of the recombined allele in microdissected fixed tissue.
Clonally induced tissues were fixed and stained for NOTCH1 and YFP at 4 weeks postinduction for YFPCreNotch1 flox/flox mice and 13 weeks postinduction for YFPCreNotch1 +/flox mice. NOTCH1 detection and intensity measurement were used to resolve Notch1 −/− and Notch1 +/− clones, respectively (Supplementary Note). Putative clonal and control areas were then microdissected from the esophageal epithelium. Clonal microdissection was carried out under a Fluorescent Stereo Microscope Leica M165 FC (Leica) using 0.25 mm diameter punch (Stoelting, 57391) as shown in Extended data Fig. 3k-n. gDNA from the microbiopsies was extracted using Arcturus PicoPure DNA extraction kit (Applied Biosystems, 11815-00) following the manufacturer's instructions. gDNA extracted from fixed tissue is fragmented, altering the linearity of the qPCR assay. Therefore, set C rather than Set B was used to determine the recombination status of the microbiopsies as specific detection of the recombined allele above background noise was sufficient to conclude on a reliable discrimination of mutant clones. Nonetheless, on average recombined exon 1 detection increased two folds in Notch1 −/− clones compared to Notch1 +/− clones.

RT-qPCR assay
RNA extractions were performed on peeled mouse esophageal epithelium as described in the RNA-seq method section (Supplementary Note). Total RNA was measured using Qubit RNA BR Assay Kit (Thermo Fisher Scientific, Q10211). cDNA synthesis of 500 ng total RNA was performed using QuantiTect Reverse Transcription Kit (Qiagen, 205313). RT-qPCR was performed with Taqman Fast Advanced Master Mix (Thermo Fisher Scientific, 4444557) on StepOnePlus Real-Time PCR System (Thermo Fisher Scientific, 4376600) and analyzed using StepOne Software v2.3. Relative qPCR expression to Gapdh housekeeping gene was calculated using delta-delta Ct method. The Taqman assays used for quantification are shown in Supplementary Table 32.

Immune capillary electrophoresis
RLT Plus lysates with Complete Protease Inhibitor (Roche, 11836170001) homogenized as described in the 'RNA-seq' section (Supplementary Note) were passed through the RNA binding column from the AllPrep DNA/RNA Mini kit (Qiagen) and the flow through was collected for protein precipitation. For precipitation, nine volumes of ice-cold pure Ethanol were mixed with the lysates before storage overnight at −80 °C. Precipitates were spun for 30 min at 20,000g at 4 °C, pellets were dried and solubilized progressively with 5% Sodium dodecyl sulfate in 100 mM TEAB solution (Sigma-Aldrich, T7408). Total protein quantification was performed using Pierce BCA Protein Assay Kit (Thermo Fisher Scientific, 10678484). Immune capillary electrophoresis was performed using Wes Simple (ProteinSimple) following manufacturer's instructions and analyzed using Compass for SW version 4.1.0. Primary antibodies were the following: anti-NOTCH1 targeting C terminus of the protein (Cell signaling, 3608); anti-NOTCH2 targeting C terminus of the protein (Cell signaling, 5732); anti-α-Tubulin (Cell signaling, 2125).

Whole-mount preparation of mouse esophagus
Tissue preparation. Mouse esophagus was opened longitudinally and the muscle layer was removed with forceps. For lineage tracing and EdU/BrdU experiments, tissue was incubated for 15 min in Dispase I (Roche, 04942086001), diluted at 1 mg ml −1 in PBS before separating the epithelium with fine forceps. For all other immunostaining experiments (including long-term antibody treatment), tissue was incubated for 2 h 15 min to 3 h in 5 mM EDTA at 37 °C before peeling the epithelium. The epithelium was then flattened and fixed in 4% paraformaldehyde Article https://doi.org/10.1038/s41588-022-01280-z for 1 h 15 min at room temperature under agitation, washed in PBS and stored in PBS at 4 °C 21 .
Whole-mount immunostaining. Whole-mount tissues were stained as previously described 43 . Tissues were incubated for 1 h in staining buffer (0.5% BSA, 0.25% fish skin gelatin, 0.5% Triton X-100 and 10% donkey serum in PHEM). This blockage step was followed by incubation with primary antibodies (Supplementary Table 31) in staining buffer overnight at room temperature, three washes of 30 min with 0.2% Tween-20 in PHEM and incubation with secondary antibodies (Supplementary  Table 31) in staining buffer for 3 h at room temperature. After further washes, tissues were incubated for an hour at room temperature with 1 µg ml −1 DAPI or 0.5 µM Sytox Blue solution (Biolegend, 425305) to stain cell nuclei and mounted using Vectashield mounting media (Vector Laboratories, H-1000).

Histology
Hematoxylin and eosin staining (H&E). H&E was either performed on 10 µm cryosectioned tissue processed as described below or on 5 µm paraffin-embedded tissue sections. Before paraffin embedding, esophageal tissue was collected and fixed in 4% paraformaldehyde for at least 2 h before undergoing progressive dehydration in Tissue-Tek VIP 6 AI tissue processor (Sakura). Slides were then scanned at objective ×20 using NanoZoomer S60 Digital slide scanner (Hamamatsu).
Immunostaining on esophageal sections. Esophageal tissue was flash frozen in tissue freezing medium (Leica, 14020108926). Ten micrometer transverse sections were fixed with 4% paraformaldehyde for 10 min, blocked in staining buffer (0.5% BSA, 0.25% fish skin gelatin, 0.5% Triton X-100 and 10% donkey serum in PHEM) and stained with primary and secondary antibodies for 3 h to overnight at room temperature (Supplementary Table 31). PHEM washes were performed between incubations. Before NICD1 staining, sections were incubated 20 min in 50 mM Glycine/PBS solution. Finally, tissues were incubated for an hour at room temperature with 1 µg ml −1 DAPI or 0.5 µM Sytox Blue solution (Biolegend, 425305) to stain cell nuclei and mounted in Vectashield mounting media (Vector Laboratories, H-1000). For Extended data Fig. 3b, freshly collected esophagus was fixed in 4% PFA for 2 h and embedded in 4% low-melting agarose. Hundred micrometer thick Vibratome (Leica) sections were cut permeabilized for 1 h and stained as for whole mounts.

Confocal microscopy
Immunofluorescence images were acquired on a Leica TCS SP8 confocal microscope using ×10, ×20 or ×40 objectives. Typical settings for acquisition were optimal pinhole, line average 3 and 4, and scan speed 400-600 Hz and a resolution of 1024 × 1024 pixels. Visualization and image analysis were performed using Volocity 6 Image Analysis software (PerkinElmer).

EdU lineage tracing
EdU incorporates during replication in the proliferating cells located in the basal layer of the esophageal epithelium. EdU i.p. injection at 10 µg was performed either at 1 h or 48 h before tissue collection. Tissue was processed and EdU was detected in whole mount using Click-iT EdU imaging kit (Life Technologies, C10338 or C10340).

EdU/BrdU lineage tracing
Mice were injected with EdU i.p. injection at 10 µg 48 h before tissue and with BrdU i.p. at 1 mg 1 h before collection to label cells in S phase. Tissue was processed as in 'Whole-mount sample preparation of mouse esophagus'. For immunostaining, tissue was first incubated for 30 min in permeabilization buffer (0.5% BSA, 0.25% fish skin gelatin, 0.5% Triton X-100 in PHEM) followed by 20 min at 37 °C in DNAse buffer containing 500 units DNAse under 500 rpm agitation (NEB, M0303L). Tissue was washed three times in PBS for 20 min. Samples were then processed as described in 'whole-mount immunostaining'. EdU was detected in whole mount using Click-iT EdU imaging kit (Life Technologies, C10338). BrdU was detected using primary antibody anti-BrdU (Abcam, ab6326). PHH3 was detected using conjugated Alexa Fluor 647 Anti-Histone H3 (phospho S10) antibody (Abcam, ab196698) (Supplementary Table 31).

Aging experiments
YFPCreNotch1 mice between 10 and 16 weeks of age were injected i.p. with BNF at 80 mg kg −1 and TAM at 1 mg for Notch1 +/+ and Notch1 +/floxand at 0.25 mg for Notch1 flox/flox and aged up to 78 weeks old. Notch1 +/+ or noninduced mice were used as wild-type controls as indicated. A lower dose of Tamoxifen was used for the YFPCreNotch1 flox/flox mice to minimize the recombination of the Notch1 allele in the corneal epithelium, possibly leading to corneal opacification and keratinization 44 .

Projected NOTCH1 stained area quantification
To quantify the percentage of NOTCH1 + or NOTCH1 − area in the entire esophageal epithelium or the projected surface of NOTCH1 − clones, whole-mount esophageal epithelia were prepared and stained for NOTCH1 and counterstained with DAPI or Sytox Blue as described in the dedicated sections. The entire epithelium was imaged using a high-precision motorized stage coupled to a Leica TCS SP8 confocal microscope. Typical settings for the acquisition of multiple z stacks were optimized 2.41 µm z step size, zoom ×1, optimal pinhole, line average 4, scan speed 400-600 Hz and a resolution of 1024 × 1024 pixels using an ×10 HC PL Apo CS Dry objective with a 0.4 numerical aperture. Images were processed using Volocity 6 software. To measure their projected surface area, we used the 'extended focus' visualization mode on the Volocity software. Regions of interest (ROI) were defined with ROI tool allowing surface area measurement. NOTCH1 staining was automatically detected based on the defined intensity and minimum object size.

Carcinogen treatment
Mice were induced with BNF/TAM at a dose that allowed full coverage of the tissue with the mutant Notch1 heterozygous or homozygous cells within 3 months. YFPCreNotch1 +/flox mice were injected i.p. on two consecutive days with BNF at 80 mg kg −1 and TAM at 1 mg and YFPCreNotch1 flox/flox were injected once with BNF at 80 mg kg −1 and TAM at 0.25 mg. Noninduced YFPCreNotch1 mice were used as wild-type controls. Mice were then treated with DEN (Sigma-Aldrich, N0756) in sweetened drinking water (40 mg l −1 ) for 24 h, 3 d a week for 8 weeks 18,21 .
Article https://doi.org/10.1038/s41588-022-01280-z SOR (LC Chemicals, S8502) was then administered at 50 mg kg −1 (5 µl of 10 mg ml −1 solution per gram bodyweight) by i.p. injection on alternate days during 6 weeks, for a total of 21 doses 33 . Mice were aged for 28 weeks after the last dose of SOR and esophageal tissue was collected. Macroscopic images of unpeeled tissue were obtained under Leica M80 zoom stereomicroscope with Leica Plan ×1.0 Objective M-Series 10450167 coupled with Leica DFC295 Camera (Leica Microsystems). Macroscopic tumors were removed and flash frozen. Esophageal tissue was whole-mount immunostained for KRT6 and DNA. The projected area of lesions was determined using ROI tool in Volocity 6 software 45 .

Immunotherapeutic treatment
To analyze the effect of NOTCH1 neutralizing antibody on tumor growth, uninduced YFPCreNotch1 flox (wild type) mice were first treated with DEN and SOR and aged for 9 weeks to allow the tumors to start growing before starting a treatment with anti-NRR1.1E3 (Genentech) at 10 mg kg −1 or with Ragweed control (Genentech) at 10 mg kg −1 (n = 4 mice per group), once a week for 6 weeks. One week after the last dosage, tissue was collected and processed for macroscopic and microscopic quantification of the projected areas of the tumors using Volocity 6 software as described in 'carcinogen treatment' section.

Cell density in mouse esophagus
Density of the basal cells was measured on whole-mount stained tissue, imaged at ×40 objective using Leica TCD SP8 confocal microscope (see 'confocal microscopy'). DAPI + or Sytox Blue + basal nuclei were quantified per area. For Notch1 mutant clones and control areas, analysis was performed in seven to nine clones and paired areas from three YFPCreNotch1 +/flox mice at 13 weeks postinduction and in three to seven clones and paired areas from seven YFPCreNotch1 flox/flox mice at 4 weeks postinduction. For aged mouse tissue, analysis was performed at three to six random positions of the tissue, n = 4 mice per genotype.

Cell counting and epithelial thickness
Epithelial thickness was quantified in cryosections stained with H&E with NanoZoomer Digital Pathology software (NDP.view2, Hamamatsu). Measurements were performed at 18-23 positions and averaged for each mouse (n = 3 mice). For cell counting, stained sections were imaged by confocal microscopy and analyzed with Volocity 6 software (Perkin Elmer). Ki67 + basal cells were counted at three different positions per animal and averaged for each mouse (n = 3-4 mice). For tumor cells in G2/M phase, cryosections were stained for pHH3, CCNB1, KRT14 and DNA (n = 8 tumors from four wild-type mice; n = 9 tumors from seven mutant mice). For NICD1, sections of tumors and adjacent normal tissue were stained for NICD1, KRT14 and DNA. The proportion of KRT14 + keratinocytes expressing nuclear NICD1 inside the tumor mass and in the adjacent normal epithelium was quantified (n = 10 tumors from six mice).

Fluorescence intensity quantification in tumors and normal tissue
Esophageal sections carrying tumors and adjacent normal tissue stained for CDH1 (E-cadherin), KRT14, and counterstained with DAPI were imaged using a Leica TCD SP8 confocal microscope. Mean intensity was quantified in ROI with the ROI tool in Volocity 6 software. Mean intensity of CDH1 was normalized to the mean intensity of DAPI at each ROI (Supplementary Table 24). For phospho-ERK1/2 (p-ERK) and total ERK1/2 (total ERK), sections were costained for KRT14 and DAPI, and were analyzed as above with the following modifications: within ROI defined on KRT14 + cells in adjacent normal or inside the tumors, p-ERK staining was automatically detected using the 'find objects' function of Volocity 6 software, using 12-255 intensity threshold and a minimum object size of 0.5 µm 2 and a restrictive radius of 2 µm. p-ERK staining was performed in all tumor sections simultaneously. For CDH1, p-ERK and Total ERK quantifications, we verified that staining was not affected in normal tissue of DEN/SOR treated Notch1 −/− mice compared to wild-type mice on tissues stained together on the same slide (n = 3 mice, Supplementary Table 24).

DNA, RNA and scRNA-seq
Methods for sample processing and analysis of sequencing data (DNA sequencing, RNA bulk sequencing and scRNA-seq) are detailed in the Supplementary Note.

Modeling
Stochastic simulations of clonal dynamics are explained in the Supplementary Note.

Statistical analysis
Data are expressed as mean values ± s.e.m. unless otherwise stated. P values <0.05 were considered significant. Each experiment was performed using several biological replicates, with the exception of technical replicates only for primer validation using standard curves. The numbers of replicates are stated in the legends and in the Supplementary tables. Statistical tests are indicated in figure legends and were performed using GraphPad Prism software 8.3.1 and Python package Scipy 1.7.3 (https://scipy.org/citing-scipy/). No statistical method was used to predetermine sample size. Animals of the correct genotype were randomly assigned to experimental groups.

Reporting summary
Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Data availability
Accession numbers for the datasets are as follows: targeted sequencing of Human esophageal epithelium microbiopsies data is deposited in the European Genome-Phenome Archive under the accession code EGAD00001006969. All other sequences are deposited in the European Nucleotide Archive under the following accession codes: targeted sequencing of aged Notch1 +/− mouse esophageal epithelium Article https://doi.org/10.1038/s41588-022-01280-z microbiopsies; ERP126992, targeted sequencing of mouse normal esophageal epithelium 28 weeks after DEN SOR treatment; ERP126993, targeted sequencing of mouse esophageal tumors 28 weeks after DEN SOR treatment; ERP126994, transcriptomic analysis of Notch1 mutant esophageal epithelium; ERP126995, single-cell transcriptional analysis of Notch1 mutant esophageal epithelium; ERP126996, transcriptomic analysis of Notch1 mutant esophageal tumors and adjacent normal tissue 28 weeks after DEN SOR treatment; ERP137375. All numerical data displayed in the figures are shown in Supplementary  Tables 2-30. Mouse strains are available from the Jax repository (https://www.jax. org), except the Ahcre ERT line, which may be obtained by contacting the corresponding author. Source data are provided with this paper.

Code availability
The codes developed in this study has been made publicly available and can be found at https://github.com/PHJonesGroup/Abby_etal_SI_code. Article https://doi.org/10.1038/s41588-022-01280-z Extended Data Fig. 1 | Aging human esophageal epithelium is colonized by NOTCH1 mutant clones. a. NOTCH1 is composed of an extracellular domain (NEC) and a transmembrane and cytoplasmic unit (NTM). Domains of NOTCH1 are indicated, arrows show epitopes recognized by anti-NOTCH1 (blue) and anti-NICD1 (orange) antibodies. Ligand binding results in proteolytic cleavages, after which the intracellular domain (NICD) migrates to the nucleus and activates transcription. Domains: EGF, epidermal growth factor like repeats, LNR, Lin12/Notch repeats, HD, heterodimerization, TM, transmembrane, RAM, RBP-J associated module, ANK, ankyrin repeats, TAD, trans-activation domain, PEST, rich in proline, glutamate, serine, and threonine, NRR, negative regulatory region. b. Human esophageal epithelium. Proliferation is confined to the lower layers. Differentiating cells migrate to tissue surface. Pa, papillae. Protein expression shown on right. c. Representative section stained for NOTCH1 (red) and DNA (blue) showing subset of results in Fig. 1b  Article https://doi.org/10.1038/s41588-022-01280-z Extended Data Fig. 2 | Lineage tracing of Notch1 mutant cells in mouse esophageal epithelium. a. Structure and cellular homeostasis in mouse esophageal epithelium. The basal layer contains progenitor cells that divide to generate progenitor and differentiating daughter cells. Differentiating basal layer cells exit the cell cycle and migrate into the suprabasal layers, moving towards the surface of the epithelium from which they are shed. The division of a progenitor cell (green) produces two progenitors, two differentiating cells or one cell of each type. In homeostatic tissue, the likelihood of each division outcome is balanced and gives on average 50% of progenitors and 50% of differentiating cells across the progenitor population. b. YFPCreNotch1 conditional knock-out mouse strain. LoxP sites (gray arrows) flank exon1 of the Notch1 gene. Notch1 flox animals were crossed with Rosa26 floxedYFP mice carrying a conditional yellow fluorescent protein (YFP) reporter targeted to the Rosa26 locus and with AhCre ERT mice carrying an inducible Cre recombinase. c. For lineage tracing, triple mutant mice were treated with inducing drugs at a dose that resulted in recombination of Notch1 (blue), expression of YFP (green) or both (orange) in scattered individual esophageal basal cells (clonal induction). The recombined cells may expand into clones detected by the reduced intensity (+/−) or absence of NOTCH1 (−/−) and expression of YFP detected by immunostaining. Samples were collected at different time points after induction and the number and location of cells in each clone determined by 3D confocal imaging of sheets of epithelium. d. Triple mutant mice were induced with a high dose of drugs, allowing recombination of cells at high density in the tissue. In the case of mutant clones with a competitive advantage over wild type cells, this protocol allowed the coverage of the tissue by mutant clones relatively shortly after induction. Corresponding author(s): Philip H Jones Last updated by author(s): Nov 27, 2022 Reporting Summary Nature Research wishes to improve the reproducibility of the work that we publish. This form provides structure for consistency and transparency in reporting. For further information on Nature Research policies, see our Editorial Policies and the Editorial Policy Checklist.

Statistics
For all statistical analyses, confirm that the following items are present in the figure legend, table legend, main text, or Methods section.

n/a Confirmed
The exact sample size (n) for each experimental group/condition, given as a discrete number and unit of measurement A statement on whether measurements were taken from distinct samples or whether the same sample was measured repeatedly The statistical test(s) used AND whether they are one-or two-sided Only common tests should be described solely by name; describe more complex techniques in the Methods section.
A description of all covariates tested A description of any assumptions or corrections, such as tests of normality and adjustment for multiple comparisons A full description of the statistical parameters including central tendency (e.g. means) or other basic estimates (e.g. regression coefficient) AND variation (e.g. standard deviation) or associated estimates of uncertainty (e.g. confidence intervals) For null hypothesis testing, the test statistic (e.g. F, t, r) with confidence intervals, effect sizes, degrees of freedom and P value noted Give P values as exact values whenever suitable.

For Bayesian analysis, information on the choice of priors and Markov chain Monte Carlo settings
For hierarchical and complex designs, identification of the appropriate level for tests and full reporting of outcomes Estimates of effect sizes (e.g. Cohen's d, Pearson's r), indicating how they were calculated Our web collection on statistics for biologists contains articles on many of the points above.

Software and code
Policy information about availability of computer code Data collection Confocal images were obtained using acquisition software Leica Application Suite X (LAS X). Confocal Z stack images were rendered and nature research | reporting summary April 2020 (Huang da et al., 2009). For sc-RNAseq, alignment of the sequencing reads and expression quantification was performed for each library individually using the CellRanger pipeline version 3.0.2 (10xGenomics). We subsequently used EmptyDrops version 1.2.2 to detect empty droplets in the raw feature count matrix output from CellRanger and discarded any barcode identified as an empty droplet. All the subsequent analysis described was performed in R version 4.1.3 (https://www.R-project.org/ ) using the Seurat software package version 4.0.3. Custom codes for clone simulations, for copy number analysis and sc-RNAseq analyses are available at https://github.com/PHJonesGroup/ Abby_etal_SI_code For manuscripts utilizing custom algorithms or software that are central to the research but not yet described in published literature, software must be made available to editors and reviewers. We strongly encourage code deposition in a community repository (e.g. GitHub). See the Nature Research guidelines for submitting code & software for further information.

Data
Policy information about availability of data All manuscripts must include a data availability statement. This statement should provide the following information, where applicable: -Accession codes, unique identifiers, or web links for publicly available datasets -A list of figures that have associated raw data -A description of any restrictions on data availability Accession numbers for the datasets are as follows:- Targeted  The codes developed in this study has been made publicly available and can be found at https://github.com/PHJonesGroup/Abby_etal_SI_code Field-specific reporting Please select the one below that is the best fit for your research. If you are not sure, read the appropriate sections before making your selection.

Life sciences Behavioural & social sciences Ecological, evolutionary & environmental sciences
For a reference copy of the document with all sections, see nature.com/documents/nr-reporting-summary-flat.pdf

Life sciences study design
All studies must disclose on these points even when the disclosure is negative.

Sample size
Sample size was not predetermined by statistical methods. Sample size was determined by pilot studies for lineage tracing and clonal sequencing studies and by previously published studies for highly mutagenized sequencing and carcinogenesis studies.
Data exclusions Data exclusion was only made in RNA sequencing studies to improve the quality of the datasets.
Quality control of RNA-seq study: 'Transcriptomic analysis of Notch1 mutant esophageal tumors and adjacent normal tissue 28 weeks after DEN SOR treatment ' revealed one outlier sample on PCA plot out of the initial 36 samples dataset. The outlier sample was removed from the analysis. Quality control of RNA-seq study: 'Transcriptomic analysis of Notch1 mutant esophageal epithelium' revealed an outlier control sample on PCA plot but deep analysis of the noise identified 151 genes with aberrant signal within the control group. Thorough checks with complete analysis were performed with and without these genes and/ or the sample, leading to the conclusion that none of these actions modified the conclusions of the analysis but excluding these genes and not the affected control sample for final analysis preserved the most data and resulted in removing 7 false positive hits. For 'Single cell transcriptional analysis of Notch1 mutant esophageal epithelium', poor quality cells were filtered out as is considered best practice. Details are provided in respective Methods and Supplementary Note.

Replication
For Human histological study, three donors were analyzed per age group (young, middle-aged and elderly). For Human sequencing, we analyzed multiple biopsies from 6 distinct donors. For mouse studies, experiments were performed with at least 3 mice per time point constituting 3 independent biological replicates except on two occasions. scRNA-seq was performed on two biological replicates per genotype as each tissue yielded sequencing data from thousands of cells. Findings were further verified in separate experiments (Immunostaining and EdU/BrdU lineage tracing in young mice; Immunostaining and EdU in aged mice) involving 3 to 4 biological replicates. Pilot neutralizing antibody titration involved some dosing that were not repeated but the final dosage was confirmed with 3 biological replicates. All attempts at replicating the findings from the study were successful.
Randomization For mouse experiments, mice of relevant genotype were randomly assigned to each experimental protocol. For Human study, donors were randomly assigned for sequencing/ histological analysis based on their age at tissue collection.

Blinding
Blinding was performed but its feasibility was sometimes limited. Technicians and investigators were blinded to group allocation during mice treatments, except when performing treatments that required such information (high inductions, antibody treatments). Samples were systematically given identification numbers so that investigators were blinded to genotype or treatment during processing and analysis when this was applicable. Blinding was not applicable or effective when the information of material were required for analysis or when the experiments required sampling using immunostaining that reflected the genotype.