Crypt density and recruited enhancers underlie intestinal tumour initiation

Abstract

Oncogenic mutations that drive colorectal cancer can be present in healthy intestines for long periods without overt consequence^1,2. Mutation of Adenomatous polyposis coli (Apc), the most common initiating event in conventional adenomas³, activates Wnt signalling, hence conferring fitness on mutant intestinal stem cells (ISCs)^4,5. Apc mutations may occur in ISCs that arose by routine self-renewal or by dedifferentiation of their progeny. Although ISCs of these different origins are fundamentally similar^6,7, it is unclear if both generate tumours equally well in uninjured intestines. Also unknown is whether cis-regulatory elements are substantively modulated upon Wnt hyperactivation or as a feature of subsequent tumours. Here, we show in two mouse models that adenomas are not an obligatory outcome of Apc deletion in either ISC source but require proximity of mutant intestinal crypts. Reduced crypt density abrogates, and aggregation of mutant colonic crypts augments, adenoma formation. Moreover, adenoma-resident ISCs open chromatin at thousands of enhancers that are inaccessible in Apc-null ISCs not associated with adenomas. These cis-elements explain adenoma-selective gene activity and persist, with little further expansion of the repertoire, as other oncogenic mutations accumulate. Thus, cooperativity between neighbouring mutant crypts and new accessibility at specific enhancers are key steps early in intestinal tumourigenesis.

Colorectal cancer (CRC) is a paradigm for the role of mutations in human tumours³ and mouse intestinal epithelium has revealed how key mutations subvert physiologic control of ISCs, which depend on regulated Wnt/RSPO signalling⁸. Inactivating mutations in APC make Wnt signalling constitutive and underlie ~80% of sporadic conventional CRCs (small bowel adenomas in mice)^3,9, distinct from those associated with microsatellite instability^10,11. Although constitutive Wnt signalling is believed sufficient for adenoma formation^8,12, not every Apc-null mouse crypt spawns an adenoma¹³ and human APC-null crypts take years to do so¹. One constraint is that mutant ISCs must first compete with their siblings to become clonally ‘fixed’¹⁴. Apc loss promotes clonal fixation^4,5 by expressing Wnt antagonist NOTUM, which suppresses nearby wild-type (WT) ISCs^15,16; similarly, mutant Kras^G12D induces suppressive bone morphogenetic proteins (BMPs) in sub-epithelial cells¹⁷. It is not known if the absence of additional mutations alone constrains tumourigenesis or if non-genetic factors contribute.

Monoclonal Lgr5-expressing ISCs at the bottoms of intestinal crypts¹⁸ compete neutrally with each other^19,20 for niche signals. ISCs that leave the crypt base cease Lgr5 expression and differentiate into enterocytes or secretory (Sec) cells²¹. Lineage tracing in mouse adenomas and RNA trajectory inference in human adenomas reveal ISC origins^11,22. Indeed, ISC-targeted deletion of Apc in Lgr5^eGFP-CreERT2 mice (ISC-Cre) induces adenomas, but the same perturbation in ISC progeny does not²³, even when combined with the Kras^G12D mutation^24,25. ISCs therefore appear uniquely capable of generating conventional adenomas. One reason is that other crypt cells are short-lived. However, crypt cells routinely reverse differentiation in homeostasis^21,26 and more so after ISC attrition^24,27. Accordingly, mice with Cre recombinase activity targeted to differentiated cells, e.g., Neurog3^Cre, Dll1^Cre, Alpi^Cre or Atoh1^Cre, invariably mark some Lgr5⁺ cells^7,24,27–31. In mice with a Wnt-pathway mutation targeted to differentiated cells, adenomas may therefore originate in occasional Lgr5⁺ ISCs that arose by dedifferentiation and expanded after tissue injury^32–34. We investigated factors that might drive tumourigenesis only in certain Apc-mutant cells and whether epigenome alterations play a part.

Non-deterministic adenoma formation

Mouse strains that express Cre recombinase in enterocytes, the predominant ISC derivative, typically also mark some resting ISCs²⁴. In contrast, Atoh1 is tightly restricted to the Sec lineage³⁵ and Atoh1^Cre(ER-T2) mice (Sec-Cre), which mark only that lineage³⁶, are useful to study dedifferentiated ISCs. One day after tamoxifen (Tam) activated Cre recombinase in Sec-Cre mice crossed with inducible Rosa26^L-S-L-TdTom (R26R^Tom) reporter mice³⁷, fluorescence was restricted to Sec (mainly goblet and Paneth) cells. Ten days later, few labelled goblet cells remained and tdTom⁺ enterocytes in most villi represented the progeny of dedifferentiated ISCs competing neutrally^19,20 with their unlabelled siblings (Extended Data Fig. 1a). Three weeks post-Cre induction (PCI), fluorescence persisted only in long-lived Paneth cells and in rare whole crypt-villus units, reflecting clonal fixation of Atoh1^Cre-marked ISCs²⁸; labelled ‘ribbons’ persisted at least 100 days, indicating that Atoh1^Cre derived ISCs are long-lived (Extended Data Fig. 1a–b). Further verifying homeostatic dedifferentiation of Atoh1⁺ Sec progenitors, single-cell RNA analysis of ileal Sec-Cre;R26R^Tom crypt cells 0.5 day and 1.5 days PCI captured substantial numbers of tdTom⁺ cells expressing ISC markers (Extended Data Fig. 1c–d). Three weeks PCI, 24.6 ±4.8% of ileal crypts were fluorescent in Lgr5^eGFP-CreERT2 (ISC-Cre);R26R^Tom mice, while 0.9 ±0.4% were fluorescent in Sec-Cre;R26R^Tom mice (Extended Data Fig. 1e). Thus, most ISCs arise by replication of native Lgr5⁺ cells –or enterocyte dedifferentiation– but a small fraction arises by routine Sec-cell dedifferentiation. Because ISCs of either origin are indistinguishable by gene expression or open chromatin⁷, ISC-Cre and Sec-Cre mice can track tumourigenesis from native (ISC-Cre: direct labelling) or dedifferentiated (Sec-Cre: “indirect” labelling) ISCs. To that end, we inactivated Apc in each mouse strain and, to follow recombined Apc-null clones, all crosses included the R26R^Tom allele (Fig. 1a).

Fig. 1 |. Consequences of direct or indirect Wnt activation in ISCs.

a, Apc-null ISC clones (crypts), labelled after Cre expression from the Lgr5 (ISC, direct) or Atoh1 (Sec, indirect by dedifferentiation) locus, engender distinct clonal outcomes: adenomas, small cysts, or crypt-villus ‘ribbons.’ Images represent hundreds of examples (Extended Data Fig. 2a, scale bars 50 μm. b, Cross-sectional and gross images of representative ISC-Cre (21 days PCI, N=8 mice) and Sec-Cre (6 weeks PCI, N=15 mice) Apc^Fl/Fl ileum (center: proximal, periphery: distal). Labelled Paneth cells persist in Sec-Cre, alongside fewer clonal features, including ‘ribbons’ (insets) than in ISC-Cre. Adenomas, abundant in ISC-Cre (especially in distal ileum), were rare in Sec-Cre. Scale bar 1 mm. c, Fractions of ilea with ≥1 adenoma (left, Fisher Exact test, two-tailed) and fractions of Apc-null (Tom⁺) clones (right, Mann-Whitney test, one tailed) in ISC-Cre <3 weeks PCI and in Sec-Cre 6 weeks and 16 weeks PCI. N, mouse numbers. d, Clonal labelling rates in ISC-Cre and Sec-Cre Apc^+/+ and Apc^Fl/Fl ileum. Sec-Cre labelling, increased owing to Apc^−/− fitness, remains lower than direct ISC-Cre labelling. Unpaired t test with Welch’s correction, one-tailed. e, Kras^G12D or Smad4^null mutations did not increase Sec-Cre;Apc^Fl/Fl ileal adenomas (Mann-Whitney test, one-tailed). f, Pearson correlations of transcriptomes in ISCs purified from ISC-Cre and Sec-Cre Apc^+/+ and Apc^Fl/Fl ileum. Tumour-bearing ISC-Cre samples cluster apart from others. g, Differential gene expression (log₂ fold-change >2, q <0.01, Wald test, two-tailed, corrected for multiple comparisons) in ISCs from ISC-Cre and Sec-Cre Apc^+/+ and Apc^Fl/Fl intestines (N=2 replicates each). Duod, duodenum. h, PyGenome tracks for Notum and Sox17 mRNA levels in purified ISCs of indicated genotypes and sources. Scale for Notum in ISC-Cre is 30 times larger than others. i, Comparable Notum expression (RNA in situ hybridization) in ISC-Cre and Sec-Cre ileal clones N=4 mice each, scale bars 50 μm. Bars in all graphs: mean ±SD.

Recombinase activity in ISC-Cre is highest in the duodenum and declines caudally¹⁸, while activity in Sec-Cre increases from duodenum to colon, commensurate with Sec-cell abundance³⁵. In ISC-Cre;Apc^Fl/Fl mice, obstructive Apc-null duodenal tumours (Extended Data Fig. 1f) warranted euthanasia by 21 days. We studied tumourigenesis in the ileum, where labelling frequencies in the two mouse lines are the closest and where Apc disruption elicited three distinct clonal outcomes: (i) “adenomas,” florid dysmorphic epithelial outgrowths into the gut lumen, (ii) small cryptal cysts previously observed to remain stable for months after Wnt activation²³, and (iii) morphologically normal crypt-villus ‘ribbons’ (Fig. 1a). As a fraction of all Tom⁺ crypts, ribbons –which signify clonally fixed Apc^−/− ISCs– and cystic structures were almost equally represented in ISC-Cre and Sec-Cre ilea (Extended Data Fig. 2a). However, while adenomas were abundant 3 weeks PCI in every ISC-Cre ileum, scattered tiny growths were present only in 20% of Sec-Cre ilea at 6 weeks and in 33% at 16 weeks (Fig. 1b–c, Extended Data Fig. 2b). Adenomas, notably rooted in multiple crypts, showed prominent luminal ingrowth, effaced crypt borders, and cell proliferation throughout the structure (Figs. 1a–b, Extended Data Figs. 2b–c), indicating that they represent dysmorphic tumours and not mere accretion of crypts. Accounting for different rates of basal ISC labelling (Extended Data Fig. 1e), they represented 9% to 32% of Apc^−/− ileal clones in ISC-Cre but <0.5% of Sec-Cre clones (Fig. 1c).

One trivial reason could be that vulnerable Apc^−/− ISCs are fewer in Sec-Cre than in ISC-Cre. Five days PCI, however, when adenomas do not interfere with clonal enumeration, ISC-Cre labelled 23 ±3.7% of ileal Apc^−/− crypts (Fig. 1d), similar to the 24.6% fraction labelled in Apc^+/+ mice and as expected from direct ISC labelling. In Sec-Cre mice, where ISC marking is indirect but Apc loss confers selective advantage^4,5, marked Apc^−/− crypts were more frequent (5.9 ±4%) than marked WT (average 0.8%) crypts (Fig. 1d). The ~3.5-fold fewer Apc-null crypts in Sec-Cre ilea cannot alone account for near absence of adenomas. Another trivial possibility is that biallelic Apc excision was inefficient in Sec-Cre. However, some aberrant structures lacked Tom signal (Extended Data Fig. 2d), implying that Apc is no less recombinogenic than Rosa26; Tom⁺ cells isolated by FACS from ISC-Cre or Sec-Cre intestines showed excision of floxed Apc exon 14 (Extended Data Fig. 2e); and Apc-null Sec-Cre clones expressed more of the Wnt target gene Axin2 than WT crypts did (Extended Data Fig. 2f). Third, although additional mutations are unlikely to account for multiple adenomas arising in all ISC-Cre mice <3 weeks PCI, we crossed conditional oncogenic Kras^G12D or Smad4^Fl alleles onto the Sec-Cre;Apc^Fl/Fl background and confirmed recombination of those alleles (Extended Data Fig. 2g and Supplementary Fig. 1). After 6 weeks, none of 8 Apc^−/−;Smad4^−/− ilea and only 1 of 8 Apc^−/−;Kras^G12D ilea showed small adenomas (Fig. 1e). Thus, the above possibilities cannot explain vastly different rates of adenoma formation by functionally similar ISCs in ISC-Cre and Sec-Cre.

We used flow cytometry to purify ISCs (Tom⁺ GFP⁺ cells) from ISC-Cre;Apc^Fl/Fl mice. To isolate their Sec-Cre counterparts, we introduced a Lgr5^DTR-GFP allele³⁸, which marks ISCs with green fluorescence (Extended Data Fig. 3a and Supplementary Fig. 2). We then profiled transcripts in Apc-null and WT (Apc^+/+) ISCs from ISC-Cre duodenum, where adenomas replete with Lgr5⁺ cells are abundant (Extended Data Figs. 1f and 3b), and from Sec-Cre ileum, where labelled ISCs outnumber those in the duodenum. RNA profiles of Apc^−/− ISCs from Sec-Cre, which lacked adenomas, resembled WT duodenal or ileal ISCs more than they resembled Apc-null cells from tumour-bearing ISC-Cre (Fig. 1f). Expression of 2,253 genes was elevated in Apc^−/− ISCs from ISC-Cre (Extended Data Fig. 3c), while those from Sec-Cre showed higher levels of 847 genes (Fig. 1g). Only 323 genes were common, e.g., Notum, which helps Apc^−/− ISCs assert clonal dominance^15,16 and further confirms Apc deletion in Sec-Cre; in contrast, Sox17, a classically adenoma-associated gene³⁹, was elevated >3,000-fold in mutant ISCs from ISC-Cre but <1.5-fold in those from Sec-Cre (Fig. 1h). In situ hybridization confirmed Notum mRNA abundance in scattered Sec-Cre ileal clones, distinct from adjacent WT crypts (Fig. 1i and Extended Data Fig. 3d). Thus, even when APC loss is clonally fixed, only part of the adenoma-associated transcriptome is activated, many Apc-mutant crypts do not spawn adenomas (even in ISC-Cre), and additional oncogenic mutations do not materially overcome tumourigenic constraints in Sec-Cre.

Cell-extrinsic barrier to tumourigenesis

To examine the different paths of Wnt-activated ISCs, we cultured crypts from each mouse model as organoids⁴⁰. Apc^−/− crypts isolated from ISC-Cre or Sec-Cre ilea 10 days PCI yielded large spheroids in the absence of supplemental RSPO (Fig. 2a), distinct from budding RSPO-dependent organoids that grew from WT crypts (Extended Data Fig. 3e). Like Apc^−/− organoids from ISC-Cre, those from Sec-Cre thrived and remained spheroidal in RSPO-free medium over 6 passages, contrasting with the failure of in vivo tumourigenesis in Sec-Cre. Transcriptionally, Apc-null organoids from ISC-Cre or Sec-Cre were disturbed to the same degree and two-thirds of the 732 activated genes overlap with those activated in ISC-Cre-derived Apc^−/− ISCs in vivo, e.g., Dkk2, Notum, Spock2, and Clu (Fig. 2b and Extended Data Fig. 3f). Transcripts increased after Wnt activation in vitro and in adenoma-resident ISCs were enriched for Wnt targets, as were those activated exclusively in vivo, but no theme was apparent among genes activated only in organoids. Notably, Sox17 mRNA, absent in Sec-Cre ISCs in vivo, was plentiful in vitro (Fig. 2c). Notum, Spock2, and elevated Axin2 localized in ISC-Cre adenomas and were absent or low in WT crypts and clonal ribbons (Fig. 2d and Extended Data Fig. 3g), indicating that their presence in Apc^−/− ISCs from ISC-Cre reflects dominance of adenoma-resident cells. Because tumour-associated Apc-null ISCs in vivo and organoids cultured from non-tumourigenic Sec-Cre ISCs display a robust transcriptional response to Wnt activation, but adenomas and most associated RNA changes are not inevitable in vivo (Extended Data Fig. 3h), some contribution external to Apc^−/− ISCs must promote the full oncogenic effect of Wnt hyperactivity in vivo.

Fig. 2 |. Environmental and spatial contexts for tumourigenesis.

a, Crypts from ISC-Cre or Sec-Cre Apc^Fl/Fl ileum 10 days PCI yielded abundant spheroidal organoids in RSPO-free cultures. Scale bars 1 mm. b, In contrast to transcriptional differences in vivo, Apc^−/− organoid transcriptomes from ISC-Cre and Sec-Cre were similar (Pearson correlations); both expressed 732 genes differently than WT organoids (log₂ fold-change >2, q <0.01, DEseq2, Wald test, corrected for multiple comparisons). c, PyGenome tracks for Notum and Sox17 mRNAs (same scale for all samples) in ISC-Cre or Sec-Cre Apc^+/+ (WT) or Apc^Fl/Fl ileal organoids. d, Abundant Axin2 in a representative ISC-Cre adenoma (in situ hybridization, N=3 mice); WT crypt levels are lower. Dashed line demarcates crypt bottoms; scale bar 50 μm. e, Five days PCI, Sec-Cre showed predominantly solitary Apc^−/− clones, whereas ISC-Cre labelling was clustered, as expected from patchy Lgr5^Cre variegation. TdTom quantified in 3 cross-sections per ISC-Cre (N=6 mice) or Sec-Cre (N=15 mice) Apc^Fl/Fl ileum. ISC-Cre vs. Sec-Cre, p <0.0001, Kolmogorov-Smirnov (K-S) test, two-tailed. f, Cumulative frequencies of clustered crypt distributions. Quantitation in ileal whole mounts confirmed Apc^Fl/Fl crypt clustering in ISC-Cre 5 days PCI (N=3 mice), in contrast to predominance of solitary crypts in Sec-Cre 98 days PCI (N=4 mice). ISC-Cre vs. Sec-Cre, p <0.0001 in 2D tissue sections and 3D whole mounts (note different x-axis ranges for cluster size), two-tailed K-S test. g, Fractions of clustered (≥2 consecutive Tom⁺) crypts in ISC-Cre and Sec-Cre ileal whole mounts (circled dots) relative to simulations (violin plots). Note different y-axis ranges. h, Hypothesis: adenomas arise from sizable patches of mutant crypts in ISC-Cre, not from isolated Apc^−/− crypts or small clusters in Sec-Cre or ISC-Cre.

X-inactivation profiles initially suggested that human colorectal adenomas are monoclonal⁴¹, but human^1,42 and mouse^43,44 adenomas were subsequently found to be oligoclonal, suggesting that multiple crypts contribute to early tumours¹³. This possibility is significant in the light of a notable difference between our mouse models: variegated Lgr5^eGfp-Cre marking of ISC-Cre crypt patches¹⁸, in contrast to Atoh1^Cre labelling of isolated Sec-Cre crypts following dedifferentiation (Extended Data Fig. 1b). Indeed, 97% of labelled crypts in Sec-Cre ileal tissue sections lacked adjacent marked crypts (singlets), ~3% were doublets, and <1% had ≥2 consecutive labelled crypts; in contrast, 41% of labelled ISC-Cre crypts presented as singlets and 30% appeared in clusters of ≥3 crypts (Fig. 2e and Extended Data Fig. 4a). To account for disparate basal labelling rates (Fig. 1d) and a non-linear relationship between labelling rate and cluster size (Extended Data Fig. 4b), we developed a Monte Carlo simulation to predict cluster frequencies from measured labelling rates (Extended Data Fig. 4c). Crypt distributions in Sec-Cre aligned with the low clustering simulations, whereas clustering in ISC-Cre was far greater than expected from random labelling (Extended Data Fig. 4d, cumulative frequencies compared in Fig. 2f), in line with the patchy variegation of Lgr5^Cre crypts. To overcome inherent underestimation of clusters in 2D sections, we examined Tom⁺ (Apc^−/−) fluorescence in ileal tissue whole-mounts, where differences in cluster distribution were even more pronounced (Fig. 2f–g). The largest patch in ISC-Cre contained 193 crypts 3 weeks PCI, compared to a maximum of ~18 labelled crypts in Sec-Cre 14 weeks PCI. In both cases (common in ISC-Cre, rare in Sec-Cre), adenomas displaying extra-planar growth, effaced borders, and fused dysmorphic crypts were interspersed among isolated crypts or variably large patches of crypts with intact borders, near-normal morphology, and Notum expression (Extended Data Fig. 4e and see Fig. 1i). The multitude of tumours when Apc^−/− crypts aggregate (ISC-Cre) and the near-absence in non-variegated Sec-Cre mice suggest that tumours initiate in adjacent mutant crypts (Fig. 2h).

Clonal aggregation in adenoma formation

Along the ISC-Cre ileal rostro-caudal axis, adenoma frequency and size were indeed highest where Tom⁺ clusters were the largest, in the terminal ileum (Extended Data Fig. 5a). In ileal whole-mounts 16 days PCI, the smallest structure protruding into the lumen with indistinct borders had 37 crypts, but this is not a deterministic threshold because many non-adenomatous patches (mean 14.75 per ISC-Cre ileum) with ≥37 labelled crypts were present (Extended Data Figs. 4e and 5a). To ask if crypt adjacency influences epithelial cell replication, we assessed MKI67 expression in WT (Tom⁻) and mutant (Tom⁺) crypts present in small or large ISC-Cre clusters. Five days PCI, when Wnt signalling is dysregulated⁴⁵ but Apc^−/− crypt morphology is normal (Extended Data Fig. 4a), MKI67⁺ nuclei in clusters of ≤3 Tom⁺ crypts (mean 14.1 ±2.4) were increased ~22% over WT crypts in the same tissue sections and the counts in clusters with ≥5 Tom⁺ crypts (mean 16.6 ±2.7 cells) were further increased ~15% (Fig. 3a). This augmented cell replication in clustered crypts likely contributes to adenoma formation.

Fig. 3 |. Adenomas arise mainly from Apc-null crypt aggregates.

a, MKI67⁺ cell numbers per Apc-null (Tom⁺) crypt are increased in large ISC-Cre clusters (≥5 adjacent crypts in 2D sections) compared to small clusters (≤3 contiguous crypts) or Apc^+/+ (WT) crypts. N=3 mice per group, 40 Tom⁺ crypts per mouse, unpaired t test with Welch’s correction, one-tailed. Right: representative immunostains, scale bars 50 µm. b, Increased rates of clonal crypt marking in Sec-Cre;Apc^Fl/Fl mice with 12 injections of Tam or by ablating Lgr5^DTR-Gfp ISCs with Diphtheria toxin (DT, Extended Data Fig. 5b) elicited small increases in cluster size. K-S test, two-tailed. c, These small increases in crypt clustering did not increase the incidence of adenomas. N, mouse numbers; Mann-Whitney test, one-tailed. d, Sec-Cre labels 4.91% of Apc^+/+ (WT) colonic, compared to 0.88% of WT ileal, crypts (N=6 per intestinal region). Unpaired t test with Welch’s correction, one-tailed. e, Tumour incidence in Sec-Cre;Apc^Fl/Fl ileum and colon after 2X Tam (standard), ISC ablation, or 12X Tam. N, mouse numbers; Fisher Exact test, two-tailed. f, Clonal labelling in ISC-Cre ilea treated with 0.1 mg Tam (N=8 mice) is reduced by 67.9% from mice treated with 1 mg Tam (N=6 mice), to approximately the same rate as Sec-Cre (N=13 mice, both compared to Fig. 1d). Unpaired t test with Welch’s correction, one-tailed. g, Reduced labelling decreased crypt clustering, as measured in ISC-Cre ileal 2D tissue sections and whole mounts (note different x-axis scales). N, mouse numbers; P <0.0001 for 1 mg vs 0.1 mg Tam in both measures (K-S test, two-tailed). h, Lack of adenomas 3 weeks and 8 weeks PCI after 0.1 mg Tam in ISC-Cre mice. N, mouse numbers; Fisher Exact test, two-tailed. Representative micrograph (scale bar 100 μm) shows predominance of clonal ribbons 8 weeks PCI. Bars in all graphs: mean ±SD.

To test whether mutant crypt density influences adenoma formation, we sought first to increase clustering of Apc-null crypts in Sec-Cre. Instead of inducing Cre activity with 2 doses of Tam, we administered Tam twice weekly for 6 weeks; separately, we delivered Diphtheria toxin (DT) to Atoh1^Cre;Apc^Fl/Fl;R26R^Tom;Lgr5^DTR mice, where DT ablation of native ISCs would trigger Sec dedifferentiation at rates higher than the background²⁸. Each intervention doubled the rate of ileal ISC labelling, from a baseline of 6.0 ±3.8% to 12.7 ±3.2% and 12.2 ±4.5% (Extended Data Fig. 5b). Crypt clustering consequently exceeded doublet numbers predicted by Monte Carlo simulation (Extended Data Fig. 5c) but neither perturbation achieved labelling, clustering, or adenoma frequencies seen in ISC- Cre (Fig. 3b–c). However, basal crypt labelling in Sec-Cre colon was >5-fold higher than in the ileum (Fig. 3d and Extended Data Fig. 5d); 60% of mice had colonic adenomas at 16 weeks and increased ISC labelling after 12 doses of Tam elicited adenomas in all Sec-Cre colons within 6 weeks (Fig. 3e and Extended Data Fig. 5e).

Conversely, we reduced crypt clustering in ISC-Cre. Diminished Tam exposure attenuates tumourigenesis^13,22, but it is uncertain whether that is because mutant ISCs are fewer or, as we postulate, crypt clustering is reduced. Lowering Tam dosage from 1 mg to 0.1 mg (2 injections) reduced the labelling rate from 22.9 ±3.7% of ileal crypts to 8.2 ±3.6%, approaching the basal rate in Sec-Cre (Fig. 3f). Apc was efficiently disrupted (Extended Data Fig. 6a) and, consistent with Wnt activation, mutant crypts expressed Notum (Extended Data Fig. 6b). The ~68% drop in labelling could produce fewer patches similar in size to those observed with 1 mg Tam or scattered labelling, with fewer and smaller clusters. In keeping with patchy Cre expression, 0.1 mg Tam induced more clusters than expected from random labelling, but clusters were smaller and sparser (Fig. 3g and Extended Data Fig. 6c–d). If tumours follow simply from Apc loss in an ISC, then adenoma frequency would decline ~68%, commensurate with the labelling rate, but if tumours initiate only within dense Apc^−/− crypt clusters, the decline would be greater. Three weeks PCI, when 1 mg Tam elicited multiple ileal adenomas, 0.1 mg yielded none; clonal ribbons and small cysts predominated, and adenomas remained absent for at least 5 additional weeks (Fig. 3h and extended Data Fig. 6e). One of 2 mice we followed for 23 weeks after giving 0.1 mg Tam showed small ileal adenomas (Extended Data Fig. 6e), consistent with stochasticity. Thus, cooperation among adjacent mutant crypts appears necessary to initiate tumours.

Distinct enhancers in Apc^−/− tumour cells

To identify molecular distinctions between Apc^−/− ISCs that do and those that do not generate adenomas, we asked whether cis-regulatory differences underlie the finding that adenoma and organoid ISCs dysregulate many more genes than those in isolated crypts (Figs. 1g and 2b). We treated ISC-Cre;Apc^Fl/Fl;R26R^{T om} mice with 1 mg Tam on 2 consecutive days and, to capture ISCs from adenoma-replete tissue (Extended Data Figs. 1f and 2i), we isolated duodenal GFP⁺ Tom⁺ cells 14 days later by FACS (Fig. 4a). Assays for transposase-accessible chromatin with sequencing (ATAC-seq)⁴⁶ on WT and mutant ISCs revealed gains and losses in open chromatin at 7,298 candidate enhancers (>-1 kb or >2 kb from transcription start sites, Extended Data Fig. 7a). The 5,704 enhancers newly accessible in Apc^−/− ISCs lacked activation-associated H3K27ac or H3K4me2 histone marks in WT ISCs (Fig. 4b), signifying their inactivity when Wnt signalling is not constitutive. Demethylation of CpG dinucleotides, another hallmark of active enhancers⁴⁷, was also absent at these sites in WT but evident in adenoma-associated Apc^−/− ISCs (Fig. 4b). Other H3K27ac-marked enhancers were hypomethylated as expected, and by contrast, neuronal enhancers⁴⁸, for example, were largely methylated (Extended Data Fig. 7b). DNA sequence motifs enriched among newly accessible sites included those for TCF/LEF effectors of Wnt signalling⁴⁹ (Extended Data Fig. 7c) and newly accessible enhancers were highly correlated with genes aberrantly expressed in tumourigenic Apc-null ISCs (Fig. 4c and Extended Data Fig. 7d).

Fig. 4 |. Distinct complement of accessible enhancers in tumour-resident Apc-null ISCs.

a, Two weeks post-Tam, all ISC-Cre;Apc^Fl/Fl;R26R^Tom intestines carried adenomas (N=8 mice). Images: representative outcropping of fused, dysmorphic crypts, scale bar 100 μm, and effaced Tom⁺ crypt architecture, scale bar 200 μm. b, Compared to WT Lgr5⁺ ISCs, Apc-null ISCs 14 days PCI showed increased chromatin access (called ATAC-seq peaks ≥log₂ 2.5-fold over WT, FDR <0.05, N=2 duodeni each) at 5,704 enhancers (>−1 kb and >2 kb from transcription start sites). WT duodenal ISCs lack H3K27ac or H3K4me2 at these sites (ChIP-seq) and CpG dinucleotides are methylated (whole-genome bisulfite sequencing, WGBS); public data sources listed in the Online Methods; WGBS data deposited in GEO series GSE241384. Enhancers accessible in both conditions are shown for comparison. c, Significant correlation (p = 4.69*10⁻⁴, one-tailed K-S test) between newly accessible enhancers and Apc^−/− ISC expression of genes within 100 kb; genes reduced in expression are not correlated. Below, representative PyGenome tracks (Pi3kgc) illustrate newly accessible enhancers (asterisks) and increased RNA expression in Apc^−/− ISCs. d, Only 419 of 5,704 enhancers (7.4%) identified in duodenal Apc^−/− ISCs from ISC-Cre are accessible in Apc^−/− ISCs from Sec-Cre (N=4 mice) or from adenoma-free ISC-Cre;Apc^−/− (N=2 mice) ilea after 0.1 mg Tam (≥log₂ 2.5-fold compared to WT ileal ISCs, N=2 mice); the other 92.6% are enriched only in duodenal and ileal ISCs from adenoma-bearing mice. Columns in grey repeated from panel B for reference. Signals at 5,000 Control regions (accessible in all conditions) are shown as background. e, Illustrative PyGenome tracks (ATAC-seq) from samples represented in panel d. Enhancers selectively accessible in adenoma-resident ISCs are highlighted. f, Study conclusion: ISCs in clustered Apc^−/− crypts generate adenomas with new chromatin accessibility at >5,000 enhancers. ISCs in dispersed Apc^−/− crypts neither form adenomas nor access those enhancers.

To ask if these enhancers mark the adenomatous ISC state or merely denote constitutive Wnt activity, we assayed enhancer states in the ileum, where we had examined tumourigenesis. We isolated Apc^−/− ISCs (GFP⁺ Tom⁺ crypt cells) by FACS from ISC-Cre and Sec-Cre ilea 2 weeks after administering 2 doses of 1 mg Tam and from ISC-Cre ilea after 2 injections of 0.1 mg Tam. In ISC-Cre;Apc^−/− ileum, 1,800 chromatin regions were less accessible than in WT ISCs or in those exposed to low-dose Tam and 4,616 regions showed new accessibility (Extended Data Fig. 7e); modest reduction from the 5,704 sites gained in ISC-Cre duodenum likely reflects presence of non-adenomatous ileal ISCs compared to the abundance of duodenal adenomas. Regional differences were also evident in WT and Apc^−/− ISCs, as expected, but ISCs –ileal or duodenal– from tumour-bearing mice were the most alike and the most distinct from others (Extended Data Fig. 7f). Even in the absence of adenomas, Apc^−/− ISCs differed from WT, with highly correlated profiles of open chromatin in Sec-Cre ileum and in ISC-Cre ileum after low-dose Tam (Extended Data Fig. 7f–g; trivial differences in pairwise analysis of ATAC-seq peaks are likely spurious).

Only 419 of the 5,704 newly accessible duodenal enhancers (7.35%) were also accessible, at least weakly, in ISCs from adenoma-free animals (Fig. 4d); these sites hence reflect the basal cis-regulatory response to Wnt hyperactivity. The remaining sites (92.65%) were accessible only in the adenoma setting, i.e., in ISC-Cre mice treated with 1 mg Tam (Fig. 4d–e). Notably, the ~1,600 enhancers showing reduced access in adenoma-resident ISCs (Extended Data Fig. 7a) remained open when ISCs did not generate adenomas (Fig. 4d), indicating that both gains and losses associate with the adenomatous state. Beyond the minority of sites that acquired access in both crypt- and adenoma-resident ISCs and the small fraction specific to an intestinal region, most sites showed new access only in ISCs from adenoma-dominant ISC-Cre ileum (Fig. 4d–e and Extended Data Fig. 7h). Thus, enhancer access is modestly altered in sparse Wnt-activated crypts that fail to form adenomas in ISC-Cre or Sec-Cre; additional chromatin changes occur near induced genes when aggregated crypts progress into adenomas (Fig. 4f).

Apc^−/− mouse intestinal tumours regress when Apc function is restored, even when Kras^G12D and Trp53 mutations are present⁵⁰, implying that any epigenetic changes induced by the latter mutations are inconsequential or Wnt-dependent. To examine the spectrum of adenoma-associated enhancer states, we introduced inducible Kras^G12D and Trp53-null mutations into the ISC-Cre;Apc^Fl/Fl strain. Mice became moribund within 2 weeks after we induced the three mutations with Tam, so we isolated duodenal Tom⁺GFP⁺ ISCs 10 days PCI, when mice were healthy. FACS-purified ISCs confirmed that, like Apc and Kras (Extended Data Fig. 2), the Trp53 locus had also recombined (Extended Data Fig. 8a). Only 519 additional enhancers became accessible after Kras^G12D and none after Trp53 deletion (Extended Data Fig. 8b), indicating that the predominant effect on the epigenome occurs upon Apc loss. Because additional loss of Smad4 increases organoid engraftment as tumours in mice^51,52, but animals with all four mutations were unhealthy even without Cre activation, we used CRISPR editing to introduce Kras^G12D and biallelic Trp53-null and Smad4-null mutations into Apc-null organoids. ATAC-seq analysis of the full mutant series showed the largest difference in enhancer accessibility between WT organoids on one hand and mutants on the other (Extended Data Fig. 8c). Differences among mutant organoids were limited; even sites with superficially higher access after Smad4 loss had simply failed to pass significance cut-offs after Apc mutation alone (Extended Data Fig. 8d). Sites newly accessible in vivo coincided with 2 of the 3 k-means clusters of enhancers activated after Apc loss in organoids, e.g., near aberrantly active Sox17 (Extended Data Fig. 8e–f). Thus, enhancer recruitment in vivo or in vitro occurs predominantly after Apc loss and stability of that repertoire after additional mutations explains why Apc-null cells can reverse tumourigenesis in vivo even when additional mutations are present⁵⁰.

Discussion

These findings provide a unifying explanation and epigenomic correlates for longstanding and recent observations about early intestinal tumourigenesis. Overtly normal colonic crypts carry single ‘driver’ oncogenic mutations in 1% of middle-aged people and only one adenoma is estimated to develop from every 375,000 such crypts in the remaining lifetime². In the prevailing view, loss of heterozygosity, clonal fixation, and additional oncogenic mutations are the principal constraints on tumour formation. We show that a ‘critical mass’ of Wnt-activated crypts, associated with altered chromatin accessibility at thousands of enhancers, is also necessary. Cooperativity among crypts was proposed previously^13,44,53, but is masked when broad Cre expression yields large fields of vulnerable crypts. Moreover, multi-ancestral origins for mouse and human adenomas align with historic^1,42–44 and fresh^54–56 evidence of their oligoclonality and with observations from lineage tracing in vivo²². When Wnt signalling is activated in non-ISC progenitors, tumours typically develop only after tissue injury^32–34. Our findings imply that this is not because ISCs derived by dedifferentiation resist Wnt activation but because the density of mutant crypts is insufficient unless injury triggers substantial localized dedifferentiation. Constitutive Wnt activity induces DNA access at ~400 enhancers and increases cell replication, but most mutant crypts, including those in clusters of modest size, appear normal. In large crypt aggregates, cell division is further augmented and altered accessibility at thousands of enhancers accompanies adenomatous progression. Apc^−/− ISCs in unaggregated crypts fail to render chromatin accessible at these sites and, conversely, retain access at hundreds of enhancers that close in adenoma-resident ISCs.

Crypts in ISC-Cre intestines aggregate naturally because variegated Cre expression is patchy. Otherwise, the probability of Wnt-activating mutations occurring in neighboring crypts is infinitesimal and, because crypt mobility is inherently constrained, mutant crypts likely do not coalesce by migration⁴⁴. Moreover, adenomas in ISC-Cre developed within days, in which time Apc^−/− crypts could not induce gatekeeper mutations that become fixed in nearby WT crypts. The best guess for an aggregation mechanism that leads to sporadic conventional polyps is by crypt fission, a well-known physiologic process^14,57 that increases with constitutive Wnt activity⁵⁸. Over the years that human colonic crypts duplicate by fission^14,58, cells would undergo innumerable divisions and genetic drift, so that a crypt cluster might appear monoclonal with respect to Apc but polyclonal with respect to other loci. However, recent detection of multiple independent Apc mutations in mouse and human adenomas^54–56 suggests that Apc mutations might derive and confer additional fitness if they arise fortuitously near existing Apc-mutant crypts. Because ISC-Cre mice survive only 3 weeks PCI and adenoma formation is not a linear function of crypt density, we cannot estimate the minimum number of adjacent mutant crypts –likely a range, not an absolute threshold– for tumourigenesis. To transition from an inert to an adenomatous state, Apc-null crypts in dense aggregates may signal to each other directly or through the mesenchymal cells that separate adjoining crypts and regulate physiologic ISC self-renewal⁵⁹. Thus, cancer-associated fibroblasts, the targets of extensive study in advanced CRC⁶⁰, may also promote tumour initiation.

METHODS

Mouse strains and husbandry.

Apc^Fl (gift from C. Perret)⁶¹, Lgr5^{eGfp-Cre(ER-T2)} (JAX strain 008875)¹⁸, Atoh1^Cre(ER-T2) (Ref. ³⁶), Rosa26^L-S-L-TdTom (JAX strain 007909)³⁷, Kras^G12D (JAX strain 008179)⁶², Lgr5^DTR-GFP (Ref. ³⁸), Smad4^Fl (JAX strain 017462)⁶³, and Trp53^Fl (JAX strain 008462)⁶⁴ mice were described previously. All animal procedures were approved by the Animal Care and Use Committee at Dana-Farber Cancer Institute. Mice were housed under specific pathogen-free conditions in 12 h light/dark cycles at constant temperature (23 ±1°C) and 55 ±15% humidity, with free access to food and water. Strains were maintained on a predominant C57BL/6 background. Mice were weaned 21 to 28 days after birth; males and females were studied as they became available and were at least 6 weeks old; the parameters we studied showed no difference between the sexes. Unless specified otherwise, mice received 2 doses of tamoxifen (Tam, Sigma-Aldrich T5648), 1 mg in ethanol (Sigma-Aldrich E7023) and sunflower oil (Sigma-Aldrich, S5007), by intra-peritoneal injections 24 h apart. Littermate controls lacking a Cre^ER-T2 allele (designated as wild-type, WT) were treated similarly. Mice were euthanized by CO₂ inhalation. Sample sizes were dictated by availability of mice and statistical considerations. The experimental design and marked differences between cohorts did not require animals to be randomized or investigators to be blinded. No animals developed externally visible tumours. To comply with permitted endpoints, ISC-Cre mice treated with 1 mg Tam were euthanized when they had lost >10% body weight, which inevitably occurred by 22 days PCI and likely reflected absorptive malnutrition. Permitted endpoints were not exceeded in any experiment. Mice of other genotypes did not become moribund and were euthanized humanely on days specified in the text.

Whole-mount inspection, histology, and immunofluorescence.

Intestines were flushed with cold PBS and divided into thirds, nominally representing duodenum, jejunum, and ileum. Each segment was washed with 4% paraformaldehyde (PFA, Electron Microscopy Sciences 15714S) in phosphate-buffered saline (PBS, Gibco 70011–044) and prepared as “Swiss rolls.” Samples for RNA in situ hybridization were passed through a sucrose gradient (10% to 30%) or incubated overnight in 30% sucrose (Sigma S9378) in PBS at 4°C on a rocker, transferred to plastic molds, frozen in optimal cutting temperature compound (OCT, Sakura 4583), and stored at −80°C en bloc or as 8–10 μm tissue sections on charged glass slides. To examine fluorescence, slides were rehydrated in 3 washes of ice-cold PBS, then mounted in DAPI-containing Vectashield medium (LSBio LS-J1033), covered (Fisherbrand 12544D), sealed (Electron Microscopy Sciences 72180), and examined for fluorescence using a Nikon Ti-Eclipse-E microscope.

For whole-mount inspection, dissected ilea were placed on wet filter paper, opened along the vertical axis, washed with 4% PFA, and pinned overnight on agarose (Azura Genomics A1705Z) in a dish containing 4% PFA, with the luminal side up and glass slides arranged to hold samples flat. PFA was then replaced with FocusClear (CelExplorer FC-101) and the ~10-cm ileum was examined on a Nikon Ti-Eclipse-E fluorescence microscope using the large image scan function.

For MKI67 immunofluorescence, frozen OCT tissue sections were thawed, rehydrated at room temperature, boiled in 10 mM citrate buffer for 1 min for antigen retrieval, treated with 0.3% hydrogen peroxide for 60 min, and incubated for 60 min in buffer containing 150 mM NaCl, 100 mM Tris (Invitrogen 15567-027), 0.05% Tween-20 (Biorad 1706531), and 0.5% blocking reagent (Perkin Elmer FP1020). Slides were incubated overnight in the same buffer containing MKI67 (Thermo Scientific RM-9106-50, 1:300) and tdTomato (LSBio LS-C340696, 1:200) antibodies (Ab), followed by biotin-conjugated sheep anti-rabbit IgG (Boehringer Mannheim 1214659), streptavidin-conjugated horseradish peroxidase (BD Pharmingen 554066), and tyramide amplification with a FITC fluorophore (Akoya Biosciences NEL701A001KT). TdTomato Ab staining was revealed using Alexa Fluor 555-conjugated donkey anti-goat IgG (Invitrogen A21432). Slides were mounted in Vectashield medium containing DAPI.

For EdU staining, mice were injected intraperitoneally with 10 μM 5-ethynyl-2′-deoxyuridine (EdU, Invitrogen C10640) 1 h before euthanasia. Intestinal regions prepared as “Swiss rolls” were embedded in OCT compound and 10 μm sections on charged glass slides were stored at −80°C. Frozen tissue sections were thawed, fixed with 4% paraformaldehyde at room temperature for 10 min, and stained using Click-iT EdU Cell Proliferation kit (Invitrogen C10640) according to the manufacturer’s protocol. EdU incorporated in replicating cells was detected by the copper-catalysed “click” reaction, which enables labelling of EdU with Alexa Fluor 647 dye. Slides were mounted with Vectashield medium containing DAPI and imaged on a Zeiss LSM 980 confocal microscope. Images were processed using Fiji software⁶⁵.

Counting of labelled crypts and clonal outcomes.

Clonal labelling rates were assessed by manual counting in 2 non-consecutive 2D tissue spiral sections per mouse, sampled along the thickness of an OCT block. The fraction of Tom⁺ crypts was determined relative to the total of DAPI-stained crypts. To evaluate clonal outcomes after Apc loss, 3 non-consecutive tissue spiral sections from each mouse were counted, ISC-Cre ilea at the terminal time (between 2 weeks and 3 weeks PCI) and Sec-Cre ilea ≥6 weeks PCI. Crypt-villus units with TdTom⁺ cells extending to the villus tip were designated ‘clonal ribbons.’ Isolated Tom⁺ crypt-like structures lying higher than the normal crypt position, wedged in the sub-epithelium, were classified as “cysts.” Adenomas were identified as distorted crypt-like aggregates extending over ≥4 crypts, lacking horizontal alignment with neighboring normal crypts, and absence or severe truncation of villi.

Crypt clusters in different Cre-driver mice were assessed in 3 non-consecutive tissue spiral sections from each animal, Sec-Cre ilea 6 weeks PCI and ISC-Cre ilea 5 days (1 mg Tam) or 21 days (0.1 mg Tam) PCI. Single and ≥2 adjacent Tom⁺ crypts were enumerated, excluding the small cystic structures because their positions varied with respect to other crypts.

In whole-mount fluorescence images, crypt clustering examined at 5 days PCI was defined as the number of adjacent Tom⁺ crypts. In ISC-Cre mice treated with 1 mg Tam, the first 5,000 crypts from the ileal terminus were counted; in ISC-Cre mice treated with 0.1 mg Tam, the whole intestine was counted. Adenomas in ISC-Cre ilea were recognized by their extension beyond the focal plane, distortion of crypt boundaries, and altered shapes of Tom⁺ cells at their edges (criteria validated by correlating histologic features with gross pathology). Adenomas were quantified by using ImageJ to measure their areas and centroid positions. TIFF image files were imported into MatLab in gray scale and crypt fluorescence intensity in isolated tissue areas was measured from the ileocecal boundary to the proximal ileum.

PCR analysis of gene recombination in vivo.

Cre-mediated recombination of mutant alleles was verified on genomic DNA extracted from FACS-sorted ISCs and, as controls, from Cre⁻ toe tissue. To verify APC^Fl recombination, forward and reverse primers were, respectively, 5′-CTAGTACTTTTCAGACGTCATG-3′ and 5′-CAATATAATGAGCTCTGGGCC-3′, and PCR (30 sec at 95°C, 30 sec at 59°C, 1 min at 72°C for 35 cycles) amplified a 240-bp product that corresponds to the recombined allele⁶¹. For Kras^G12D recombination, Dr. Nuné Markosyan (University of Pennsylvania) kindly provided a method using forward and reverse primers 5′-GTCTTTCCCCAGCACAGTGC-3′ and 5′-CTCTTGCCTACGCCACCAGCTC-3′; PCR (30 sec at 95°C, 30 sec at 69°C, 45 sec at 72°C for 45 cycles) amplified a 650-bp product corresponding to the recombined allele. To verify Smad4^Fl recombination, touchdown PCR with primers 5′-AAGAGCCACAGGTCAAGCAG-3′ and 5′-GACCCAAACGTCACCTTCAG-3′ yielded a ~500-bp product from the recombined allele⁶³. For Trp53^Fl recombination, PCR (30 sec at 94°C, 30 sec at 58°C, 50 sec at 72°C for 30 cycles) with forward and reverse primers 5′-CACAAAAACAGGTTAAACCCA-3′ and 5′-GAAGACAGAAAAGGGGAGGG-3′ amplified a 612-bp product that corresponds to the recombined allele,⁶⁴ as confirmed by DNA sequencing.

Cell Isolation.

After flushing intestines with cold PBS, 2-cm duodenal or ileal segments were incubated in 5 mM EDTA (Invitrogen AM9260G) solution in PBS at 4°C on a high-speed rocker with frequent manual shaking of the suspension to separate villi from crypts, followed by passage through a 100 μm filter to exclude villi and centrifugation at 340 g at 4°C. To dissociate crypts into single cells, the pellet was resuspended in TrypLE Express (Gibco 12604–021) in Dulbecco’s Modified Eagle Medium (DMEM, Gibco 10569–010) or Type II Collagenase (3 mg/mL, Worthington LS004177) in Ca⁺⁺ and Mg⁺⁺ free Hank’s Balanced Salt Solution (HBSS, Cytiva SH30268.02) supplemented with 2% fetal bovine serum (FBS, Sigma-Aldrich F2442) and 10 mM HEPES, and rotated at 37°C at 30 rpm for up to 30 min until cells disaggregated. Cells were passed through a 40 μm filter to exclude clumps, pelleted at 340 g at 4°C, and resuspended in FACS buffer (PBS containing 2% FBS, 2 mM EDTA, and 10 mM HEPES) for flow cytometry or in Trizol (Thermo Fisher 15596026) for RNA extraction.

RNA-seq.

RNA was extracted from primary cells purified by FACS or from organoids using Trizol reagent and purified over RNAeasy Plus Micro (Qiagen 74034) or PureLink miniRNA (Invitrogen 12183018A) columns. Libraries were prepared using NEBnext Ultra II non-directional library kits (New England Biolabs E7770S) and sequenced on the NovaSeq 6000 platform. For single-cell RNA sequencing, 10⁴ viable Atoh1^Cre-labelled Tom⁺ cells sorted by flow cytometry were loaded onto a Chromium controller, followed by the Chromium Next GEM Single Cell 3′ V3.1 assay (10X Genomics PN-1000121). Libraries were constructed according to the manufacturer’s recommendations and sequenced on the NovaSeq platform (Illumina).

ATAC-seq.

Omni-ATAC was performed as described previously⁶⁶. FACS-isolated cells were centrifuged and resuspended in ATAC-resuspension buffer (10 mM Tris-HCl, 10 mM NaCl, 3 mM MgCl₂ in nuclease-free water) with 0.1% IGEPAL (Sigma-Aldrich 542334), 0.1% Tween-20 (Sigma-Aldrich P9416), and 0.01% Digitonin (Sigma-Aldrich D141) for 5 min on ice. Ice-cold ATAC-resuspension buffer with 0.1% Tween-20 was added to the sample and inverted 3 times. Pelleted nuclei were resuspended in a mixture of 1x Tagment DNA buffer and 100 nM transposase (Illumina FC-131-1024), 0.01% Digitonin, and 0.1% Tween-20 in nuclease-free water and incubated at 37°C for 30 min on a thermomixer at 1000 rpm. Transposed fragments were isolated using the Qiagen MinElute PCR Purification kit (Qiagen 28004) and amplified using indexed primers and the NEBNext High-Fidelity 2x PCR master mix (New England Biolabs M0541). The number of PCR cycles required for each library was determined after the first 5 cycles. Libraries were size selected using AMPure beads (Beckman Coulter A63881) and sequenced on a NovaSeq 6000 instrument (Illumina).

In situ RNA hybridization.

RNAs were localized using the RNAscope Multiplex Fluorescent Reagent Kit v2 (Advanced Cell Diagnostics, ACD), the manufacturer’s recommended protocol⁶⁷, including 5 min of target retrieval as described⁶⁸, and probes synthesized by the manufacturer to detect Notum (ACD 428981), Axin2 (ACD 400331), Sox17 (ACD 493151), and Spock2 (ACD 438901). Because the protocol quenches tdTomato fluorescence, that signal was recovered by immunostaining. After probe hybridization, slides were washed for 5 min in PBS containing 0.1% Tween-20, blocked for 1 h in PBS containing 1% bovine serum albumin, and incubated overnight at 4°C with tdTomato Ab (LSBio LS-C340696, 1:200), washed in PBS, then incubated with Alexa Fluor 555-conjugated donkey anti-goat IgG (Invitrogen A21432, 1:250) for 90 min at room temperature. DAPI was added to the slides, which were mounted in Vectashield medium without DAPI (LSBio LS-J1032). Images were captured on a Leica SP5X laser scanning or a Zeiss LSM 980 confocal microscope and processed using Fiji software⁶⁵.

Organoids with defined oncogenic mutations.

Crypts were isolated from WT, ISC-Cre or Sec-Cre mice after flushing intestines and washing 2-cm segments, both with cold PBS. Samples were transferred to 2.5 mM EDTA solution, placed on a rotator at 30 rpm for 30 min, and shaken 10 to 20 times until crypts were released. Crypt suspensions were passed over a 70 μm filter to remove villi and centrifuged at 100 g for 10 min at 4°C. Crypts were cultured in 48-well plates in growth factor-reduced (GFR) Matrigel (Corning 354230) submerged in Advanced DMEM (Gibco 12634–010) supplemented with 1x Glutamax (Gibco 35050–061), 1 mg/mL HEPES (Gibco 15630–080), 1x PenStrep (Gibco 15140–122), 1x Antibiotic-Antimycotic (Gibco 15240–062), 7.5 mM N-acetyl cysteine (NAC, Sigma A8199), 1x B27 (Gibco 12587010), 1x N2 (Gibco 17502–048), 50 ng/mL EGF (Peprotech 315–09), 100 ng/mL Noggin (Peprotech 250–38), and 10% RSPO1 conditioned medium (Harvard Digestive Disease Center). Medium was changed every other day. RSPO1 was excluded from cultures of Wnt-independent Apc-null crypts.

We generated the AK, AKP, and AKPS organoid series by subjecting Apc^−/− organoids to one or more sequential rounds of CRISPR/Cas9 editing. Apc^−/− organoids were disaggregated and plated as single cells in GFR Matrigel plugs diluted 50% with Advanced DMEM/F12 (Thermo Fisher 12634010) supplemented as above, without RSPO conditioned medium and with 1x B-27 from a different vendor (Thermo Fisher 17504044). Organoids were expanded over 5 to 7 days with daily medium changes. To harvest cells, Matrigel was dissociated using ice cold PBS with 2.5 mM EDTA for 10 min. Organoids were dissociated into single cells in Accutase (Stem Cell Technologies 07920) for 5 min at 37°C and resuspended at 5x10⁶/mL in Lonza Nucleofector P1 Primary Cell electroporation buffer (Lonza 197188). Cells were electroporated using the CA-137 program on the Lonza 4D Nucleofector X Unit.

sgRNAs targeting Kras, Trp53 or Smad4 were generated by hybridizing an Alt-R CRISPR-Cas9 tracrRNA (Integrated DNA Technologies, IDT 072534) with a gene-specific Alt-R CRISPR-Cas9 crRNA XT (IDT) at 1:1 molar ratio according to the manufacturer’s instructions. sgRNAs were then complexed with TruCut Cas9 Protein v2 (Thermo Fisher A36498) at 3:1 molar ratio for 10 min at room temperature to generate gene-specific CRISPR ribonucleoprotein (RNP) complexes. To generate AK (Apc^−/− Kras^G12D) organoids, we electroporated Apc^−/− organoids with 1 μM CRISPR RNP targeting Kras and 4 μM single-strand oligodeoxynucleotide template specifying the G12D mutation for homology-directed repair. Cells were cultured in complete medium excluding EGF for 14 days to select Kras^G12D mutants, which were verified by Sanger sequencing before subsequent editing. AKP (Apc^−/− Kras^G12D Trp53^−/−) organoids were generated by electroporating AK organoids with 1 μM CRISPR RNP targeting Trp53 at its 5’ end and culturing cells for 7 days in EGF-free medium supplemented with 10 μM Nutilin-3 (Biotechne 3984/10). AKPS (Apc^−/− Kras^G12D Trp53^−/− Smad4^−/−) organoids were generated by electroporating AKP organoids with 1 μM CRISPR RNP targeting the 5’ end of Smad4 and selection in EGF- and Noggin-free medium for 14 days. Trp53 and Smad4 indels were confirmed by Sanger sequencing.

Statistical Analysis.

Statistical analyses were performed using R version 3.5.1 (https://www.r-project.org/) and GraphPad Prism version 9.5.0. Specific tests and significance values are noted in each figure and summarized in Supplementary Table 1.

Computational Analysis.

All high-throughput sequencing reads were mapped to the mouse genome mm10. RNA-seq data were aligned and mapped using Viper⁶⁹, which implements the STAR aligner⁷⁰. Differential gene expression was determined using DESeq2 (version 1.22.2)⁷¹ with parameters as specified in the text. Heatmaps were generated using Pheatmaps⁷² (version 1.0.12). Fastq files from scRNA-seq data were aligned to mouse genome version mm10, unique molecular identifiers (UMIs) were counted using Cell Ranger v3.0.2 (10X Genomics) with default parameters, and quality control, normalization, and clustering were performed in Seurat package v4.3.0.1 in R version 4.2.0 (https://www.r-project.org/). Epithelial (EPCAM⁺ CD45⁻) cells with >600 and <3,000 features and <7% mitochondrial genes in duplicate ileal libraries were retained and the FindAllMarkers function was used with default parameters to assign cell identities and identify genes enriched in each cluster.

ATAC-seq data were aligned using Bowtie⁷³ (version 2.3.4.3) and peaks were identified using MACS2 (version 2.1.2, q <0.05)⁷⁴. Enrichment of chromatin accessibility in peaks was collected using Bedtools⁷⁵ (version 2.27.1) and UNIX commands. Differential analysis was performed using DESeq2, heatmaps were generated using DeepTools⁷⁶ (version 2.2.2), and genome tracks for RNA- and ATAC-seq signals were visualized using pyGenomeTracks⁷⁷ (version 2.1). ChIP-seq data were aligned using Bowtie and plots were created using DeepTools. Whole genome bisulfite sequencing (WGBS) data were aligned and CpG methylation was called using Bismark⁷⁸ (version 0.13.1). Correlations between newly accessible enhancers and gene expression in Apc-null ISCs were computed using BETA analysis⁷⁹. ChIP-seq data are from GEO series GSE131265 (H3K4me2, Ref. 80) and GSE131265 (H3K27ac, Ref. 81). WGBS data on purified Apc^−/− ISCs from this study (GSE241383) were compared with published WGBS data from purified WT ISCs (Ref. 82, NCBI Sequence Read Archive series ERR454965).

Monte Carlo simulations for crypt clustering were performed using functions and commands written in R. Labelling rates for the various simulations were derived from normal distributions fit to the in vivo observations. The function generated a vector of length 1200 as a benchmark for the median number of crypts per ileal cross-section and clustering rates were determined from the prescribed rate of random labelling. Each simulation contained as many biological replicates, and technical replicates per biological specimen, as the in vivo data and was iterated 10⁵ times. To compare simulated with in vivo data, clustered crypts were represented as a fraction of all labelled crypts. Distributions of simulated and observed clustering were visualized using ggplot2 (https://ggplot2.tidyverse.org).

Extended Data

Extended Data Fig. 1 |. Atoh1^Cre(ER) (Sec-Cre) mice capture routine Sec dedifferentiation into conventional homeostatic Lgr5⁺ ISCs.

a, Crypt-villus units from a representative Sec-Cre;R26R^Tom (Apc^+/+) ileum 1 day PCI (N=8 mice), showing labelled villus Sec and crypt base Paneth cells. Ten days PCI, nearly every villus carried scattered goblet cells and strips of consecutive tdTom⁺ enterocytes, which represent the progeny of dedifferentiated Atoh1^Cre labelled ISCs competing neutrally with unlabelled ISCs (N=3 mice). Three weeks PCI, tdTom label persisted only in long-lived Paneth cells and occasional whole crypt-villus units (N=5 mice), reflecting clonal fixation of dedifferentiated Atoh1^Cre-marked ISCs. Scale bars 50 μm. b, Labelled clonal ribbons were present 100 days PCI, indicating that dedifferentiated and clonally fixed ISCs are long-lived. N=3 mice. Scale bars: left 1 mm, right 50 μm. c-d, Single-cell RNA analysis of ileal Sec-Cre;R26R^Tom crypt cells captured by tdTom flow cytometry 0.5 and 1.5 days (combined) PCI, showing a substantial fraction of tdTom⁺ cells that uniquely express classic ISC markers: Lgr5, Axin2, Olfm4, and others shared with replicating Atoh1⁺ Sec progenitors (Sec-pro): Top2a and Mki67. The findings confirm that Sec-cell dedifferentiation is a routine homeostatic event. Ent-pro, enterocyte progenitors; Gob/Pan (GP), goblet/Paneth precursors; EE, enteroendocrine. e, Higher labelling of whole crypts in ISC-Cre (N=4 mice) than in Sec-Cre (N=6 mice) Apc^+/+ (WT) ilea. Bars, mean ±SD; unpaired t test with Welch’s correction, one-tailed). f, Adenomas carpeted ISC-Cre duodenum 3 weeks PCI, necessitating euthanasia (N=12 mice). Scale bar 1 mm.

Extended Data Fig. 2 |. Different tumour loads in ISC-Cre and Sec-Cre Apc^Fl/Fl ileum.

a, Distributions of clonal events in ISC-Cre and Sec-Cre Apc^+/+ and Apc^Fl/Fl ilea. N, mouse numbers. P values from comparisons between Apc^Fl/Fl ilea in ISC-Cre vs. Sec-Cre derived using unpaired t test with Welch’s correction (one-tailed). b, Photomicrograph of an Apc^−/− Sec-Cre ileum 16 weeks PI, showing a single diminutive adenoma (dashed rectangle, magnified in inset). Scale bar 1 mm, inset 100 µm. c, ISC-Cre Apc^−/− ileal adenoma showing that S-phase (EdU⁺) cells are not confined to crypts, as in neighboring Apc^+/+ tissue, but present throughout the tumour. N=4 mice, scale bar 50 μm. d, Illustrative tdTomato⁻ microadenoma (dashed oval, 1 example from N=2 mice) in Sec-Cre ileum, reflecting Cre-induced recombination at the Apc but occasionally not at the Rosa26 locus. Bar 50 μm. e, PCR genotyping of genomic DNA for Apc exon 14 shows the excised (floxed, 240 bp) product in FACS-purified ISCs (Tom⁺ cells) from ISC-Cre or Sec-Cre mice, but not in Tom⁻ cells. A larger non-specific PCR product is amplified in all samples. N=2 independent isolates. Whole gel is shown. f, Axin2 RNA in situ hybridization shows expression as a marker of active Wnt signaling in proliferative normal crypt cells and elevated levels in scattered cystic structures (arrows) in Sec-Cre;Apc^fl/fl ileum. N=3 intestines. Scale bar 1 mm; boxed area is magnified below, scale bar 50 μm. g, PCR genotyping of genomic DNA for Kras^G12D (top, N=3 mice) and null Smad4 (right, N=2 mice) alleles verifies Cre-mediated recombination in FACS-purified ISCs (Tom⁺ cells) but not in DNA extracted from toes (top N=3 mice, bottom N=2 mice), which lack Cre recombinase. Source gels shown in Supplementary Figure 1.

Extended Data Fig. 3 |. Gene deregulation in Apc-null ISCs and organoids.

a, Representative FACS plots for purification of Apc^−/− ISCs (GFP⁺ Tom⁺) from ISC-Cre;R26R^Tom mice (left, n >10) and from Sec-Cre;R26R^Tom mice that also carry the Lgr5^Dtr-Gfp allele⁴⁴ for fluorescent ISC marking (right, n >10). ISCs reflecting dedifferentiation of Atoh1^Cre labelled Sec cells represent 1.7% of viable cells in Sec-Cre mice 6 weeks PCI. Supplementary Fig. 2 shows the FACS gating strategy. b, Apc-null adenomas in ISC-Cre contain a high proportion of GFP⁺ (Lgr5⁺) ISCs, extending well beyond the crypt base. N=8 mice. Left, merged fluorescent signals; right, isolated GFP (Lgr5) signal. Scale bar 50 μm. c, Aberrant expression of 3,286 genes (RNA-seq, DESeq2 analysis, two-tailed Wald test corrected for multiple comparisons, q <0.01, log₂ fold-change >2, base mean >20 counts) in Apc^−/− ISCs isolated from ISC-Cre mice compared to Apc^+/+ (WT, from mice lacking Cre) ISCs, both isolated 14 days PCI. d, Additional example of Notum expression (RNA in situ hybridization) in Sec-Cre ileal clones (N=4 mice, scale bar 50 μm). e, Wild-type (Apc^+/+) organoids are budding structures, distinct from the spheroidal morphology observed in Apc-null organoids. N=5 independent cultures, scale bar 1 mm. f, Aberrant gene expression in Apc^−/− organoids in vitro overlaps with genes dysregulated in Apc^−/− ISCs isolated from ISC-Cre mice. Table shows examples of genes dysregulated (average DESeq read counts from N=2 mice each) in ISCs from ISC-Cre mice and their relative expression in ISCs isolated from WT mice, Apc^−/− ISCs isolated from Sec-Cre mice (in vivo), and organoids cultured from each source of small intestine crypts (N=2 independent cultures from each). g, RNA in situ hybridization shows Notum (top, N=2 mice) and Spock2 (bottom, N=2 mice) expression predominantly in ISC-Cre adenomas. Scale bars 50 μm; dashed lines demarcate crypt bottoms. h, Context-dependent consequences of Apc loss. In gene activity and absence of adenomatous features, ISCs from Sec-Cre behave like WT ISCs in vivo but their expansion in organoid medium lacking RSPO phenocopies their ISC-Cre counterparts.

Extended Data Fig. 4 |. Monte Carlo simulations of crypt clustering.

a, Variegated Lgr5^Cre expression in ISC-Cre mice and sporadic dedifferentiation in ISC-Cre mice result in areas with variably dense (top, 5 days PCI, N=6 mice) and sparse (bottom, N=15 mice) tdTom labelling, respectively. Scale bars 50 μm. b, Top: the relation between crypt clustering (y-axis, values ±SD) and basal rates of crypt labelling (x-axis) is non-linear. Bottom: derivatives of the clustering rates. The largest change in clustering rate occurs between the labelling rates observed in ISC-Cre and Sec-Cre ileum. c, Simulation of crypt clustering in silico. Using normal distributions fit to the observed rates of crypt labelling, sequential crypts with randomly selected cells were represented by vectors, given the prescribed rate of labelling, and simulated in silico. Distributions of clustered crypts were generated from 100,000 iterations for each dataset. d, The cluster distribution of Sec-Cre crypts (top) fell at the low end of the range expected from random labelling. In contrast, ISC-Cre mouse ilea (bottom) showed more clustering than predicted from random basal crypt labelling, as expected from the known patchiness of variegated Lgr5^Cre expression. Dark grey: 5^th to 95^th percentile range, light grey: maximum range for the simulated data. Cumulative frequencies displayed together for comparison in Fig. 2f. e, Whole mount fluorescence micrograph of the largest crypt cluster detected in a Sec-Cre ileum (14 weeks PCI, example from N=4) and showing adenomatous features: extension beyond the tissue plane and effaced crypt boundaries. Scattered isolated crypts appear on the periphery. Whole-mount fluorescence image from Apc^Fl/Fl ISC-Cre ileum 16 days PCI shows an adenoma (dashed yellow structure: extra-planar growth, fused dysmorphic crypts) among variably sized patches of untransformed Apc-null crypts with distinct borders. N=4 mice, scale bars 1 mm.

Extended Data Fig. 5 |. Perturbations of crypt clustering in Sec-Cre mice.

a, Whole-mount images of representative ISC-Cre ilea 5 days and 15 days PCI demonstrate dense, patchy labelling correlated with adenomas (15 days PCI), compared to diffusely scattered singletons observed in Sec-Cre as late as 14 weeks PCI. Bars 1 cm. Below, tdTom fluorescence signals in ISC-Cre whole mounts (5 days PCI, N=4 ilea) are greatest in the distal 2 cm, correlating with tumour size and abundance at euthanasia (2–3 weeks PCI, N=4 ilea). b, Increased rates of clonal crypt marking in Sec-Cre;Apc^Fl/Fl mice with 12 injections of Tam or by ablating Lgr5^DTR-Gfp ISCs with Diphtheria toxin (DT), hence increasing Sec cell dedifferentiation. N, mouse numbers; first two columns repeated from Fig. 1d; unpaired t test with Welch’s correction, one-tailed. c, Upon elevation of basal crypt labelling with 12 doses of Tam (top) or DT treatment (bottom), Monte Carlo simulations showed more doublets than expected from random crypt distributions (dark dots in right graph), but larger clusters remained rare. Cumulative frequencies displayed together for comparisons in Fig. 3b; dark grey shading: 5^th to 95^th percentile range, light grey: maximum range for simulated data. d, High TdTom labelling in Sec-Cre colon, owing to large number of native Sec cells available to dedifferentiate. N=6 colons, scale bar 1 mm. e, High adenoma burden after 12X Tam in a representative Sec-Cre;Apc^Fl/Fl colon from N=8 mice. Scale bar 1 mm, boxed area magnified below (scale bar 150 μm).

Extended Data Fig. 6 |. Perturbation of crypt clustering in ISC-Cre mice.

a, PCR genotyping of genomic DNA for Apc exon 14 shows the excised product (240 bp) in FACS-purified ISCs (Tom⁺ cells) from ISC-Cre mice treated with 0.1 mg Tam (N=2 mice) but not in ISCs purified from animals that did not receive Tam (N=2 mice). Purified Tom⁻ and Tom⁺ crypt cell fractions from mice treated with 1 mg Tam (Extended Data Fig. 2g) serve as negative and positive controls. Whole gel is shown; a larger non-specific PCR product is amplified in all samples. b, In situ hybridization revealed Notum expression in clonal structures in ISC-Cre;Apc^fl/fl ileum after 2 doses of 0.1 mg Tam (N=2 mice). Top: low magnification, bottom (2 images): high magnification; all scale bars 50 μm. c, Monte Carlo simulations revealed high clustering after ISC-Cre mice were treated with 0.1 mg Tam, as expected from the underlying patchy variegation, but cluster sizes were substantially reduced compared to mice treated with 1 mg Tam. Cumulative frequencies compared with others in Fig. 3g; dark grey: 5^th to 95^th percentile range, light grey: maximum range for simulated data. Note different y-axis ranges for 1 mg and 0.1 mg treatments; circled dots: measured clustering frequencies. d, Sparse crypt labelling detected in ISC-Cre Apc^Fl/Fl ileum whole mounts 5 days PCI with 0.1 mg Tam. N=3 mice, scale bar 1 cm. e, Whole ISC-Cre;Apc^Fl/Fl ileum 8 weeks (N=4 mice) and 23 weeks (N=2 mice) PCI after 0.1 mg Tam, showing small discrete adenomas (dashed ovals) in one of the latter ilea. Scale bars 1 cm.

Extended Data Fig. 7 |. Altered enhancer accessibility in adenoma-resident Apc-null ISCs.

a, Aberrant accessibility at 7,298 enhancers (DESeq2 analysis, log₂ fold-change ≥2, two-tailed Wald test corrected for multiple comparisons, q <0.01) in Apc^−/− duodenal ISCs from ISC-Cre mice, compared to Apc^+/+ (WT, from mice lacking Cre) ISCs, both isolated 14 days PCI. N=2 mice each. b, DNA at enhancers newly accessible in ISCs from adenoma-bearing ISC-Cre mice are hypomethylated (range 0% to 100% methylation in sites containing ≥5 CpGs) relative to the same sites in Apc^+/+ ISCs. Signals representing the full spectrum of DNA methylation are shown for comparison: at 5,000 arbitrary active H3K27ac⁺ enhancers (unmethylated in ISCs), at all CpGs (largely methylated), and at 5,000 neuronal enhancers³⁷ (unmethylated in neuronal cells but methylated in WT and Apc^−/− ISCs). c, DNA sequence motifs appreciably enriched among enhancers newly accessible in Apc^−/− ISCs match transcription factors from the AP-1, TCF/LEF and FOX0 families. P values derived using the Fisher Exact test (one-tailed, corrected for multiple comparisons). d, Additional representative PyGenome tracks illustrate newly accessible enhancers (asterisks) in a locus (Spock2) with elevated RNA expression in Apc^−/− ISCs. e, Top: differential analysis of enhancers in Apc^−/− ileal ISCs from ISC-Cre mice treated with 1 mg or 0.1 mg Tam, showing differences similar to those across ~100,000 called ATAC-seq peaks (DEseq2, two-tailed Wald test corrected for multiple comparisons, q <0.01, log₂ fold-change ≥2) between WT and adenoma-resident duodenal ISCs. Bottom, alternative display of volcano plot shown in panel a. f, Pearson correlations among called peaks in ATAC-seq analysis of WT or adenoma-resident Apc^−/− duodenal ISCs and of WT or Apc^−/− ileal ISCs from Sec-Cre mice and from ISC-Cre mice treated with 1 mg or 0.1 mg Tam. Black boxes demarcate like samples: WT ISCs (bottom left, showing regional differences); adenoma-resident ISCs (top right, duodenal and ileal); ISCs not associated with adenomas (center, Sec-Cre mice treated with 1 mg Tam and ISC-Cre mice treated with 0.1 mg Tam). g, Differential analysis of enhancers in Apc^−/− ISCs from Sec-Cre mice treated with 1 mg Tam and from ISC-Cre mice treated with 0.1 mg Tam, showing few differences across ~100,000 called ATAC-seq peaks, in contrast to the modulated chromatin access in adenoma-resident Apc^−/− ISCs from ISC-Cre mice (1 mg Tam, panel e). h, Additional PyGenome tracks from ATAC-seq samples represented in Fig. 4d, showing sites selectively accessible (shaded boxes) in adenoma-resident ISCs.

Extended Data Fig. 8 |. Lack of appreciable additional enhancer accessibility when oncogenic mutations are superimposed on Apc-null ISCs.

a, PCR genotyping of genomic DNA for Trp53 deletion reveals the 612-bp product (verified by DNA sequence) expected from Cre-mediated recombination in FACS-purified ISCs (Tom⁺ cells, N=3 independent isolates) but not in DNA extracted from toes (N=3 mice), which lack Cre recombinase. A non-specific PCR product (uninterpretable DNA sequence) was amplified in all samples. Source gel shown in Supplementary Figure 1. b, Minimal expansion of the enhancer repertoire (519 additional accessible sites, log₂ fold-change ≥2, DESeq2, two-tailed Wald test corrected for multiple comparisons, q <0.01,) in adenoma-associated Apc^−/− ISCs upon addition of Kras^G12D mutation (N=4 mice) and no further expansion upon Trp53 loss (N=2 mice) in vivo. Two samples shown for each genotype. c, Principal component (PC) analysis of open chromatin (MACS2 called peaks) in WT (N=4 independent organoid cultures), Apc^−/− (A, N=3 cultures), Apc^−/−;Kras^G12D (AK, N=4 cultures), Apc^−/−;Kras^G12D;Trp53^−/− (AKP, N=2 cultures), and Apc^−/−;Kras^G12D;Trp53^−/−;Smad4^−/− (AKPS, N=3 cultures) intestinal organoids. The largest distinction (PC1) is between WT and all Apc-null organoids. d, K-means clustering (k=3) of 7,574 sites with objectively differential chromatin access in the mutational series compared to WT ISCs. All mutant organoids gave consistent signals across strong (cluster 1: n=701) and moderate (cluster 2: n=2,376) sites. The volcano plot shows that 4,497 sites (cluster 3) with ostensibly enriched access in AKPS organoids fail stringent thresholds for significance (two-tailed Wald test corrected for multiple comparisons). Control sites are 4,000 arbitrary enhancers accessible in all organoids, including WT. e, Significant overlap between enhancers differentially accessible in Apc-null ISCs in vivo and Apc^−/− organoid cultures. Sites are clustered by whether peaks were called only in vivo, only in organoids, or in both. At sites called only in one, signals are evident in the other, but not in WT ISCs or organoids. f, PyGenome tracks for ATAC-seq signals at the Sox17 locus in Apc^−/− ISCs isolated from ISC-Cre crypts in vivo and in organoids cultured from WT or Apc^−/− intestinal crypts. Boxed areas are magnified below to highlight chromatin access in those locations.

Supplementary Material

Supplementary Information is available for this paper.

Data availability.

Sequencing data, deposited in the Gene Expression Omnibus (GEO) under accession numbers GSE241384 and GSE281056, are available without restriction. Sources of published data for epigenome comparisons (Fig. 4) are listed in the Online Methods. Source data for graphs are available through links in the figure legends.