Main

Colorectal cancer is a leading cause of cancer-related death1 and progresses owing to the sequential acquisition of specific genetic alterations, including functional loss of tumour suppressors such as APC, TP53 and SMAD4, as well as activating mutations in oncogenes such as KRAS2. Only a subset of cells within these tumours has tumorigenic potential3,4, suggesting a hierarchical organization. These cells, referred to as CSCs, are thought to represent the clonogenic core of the tumour and can be identified by specific markers, including leucine-rich repeat-containing G-protein-coupled receptor 5 (LGR5)5,6,7,8. Targeting CSCs is predicted to have widespread clinical implications, since it is assumed that these cells are involved in tumour progression and metastatic dissemination9. Although CSCs have been identified in various tumour types10,11, their clinical relevance remains unclear and still awaits the development of relevant pre-clinical experimental systems in which the selective depletion of CSCs can be achieved12,13.

Modelling colorectal cancer progression

To investigate the functional role of CSCs in colorectal cancer initiation, growth and metastasis, we combined the Apcmin/+;KrasLSL-G12D/+;Vil1Cre (AKV) model of intestinal tumorigenesis6,14 with the Lgr5DTR/eGFP allele15. The resulting Apcmin/+;KrasLSL-G12D/+;Vil1Cre;Lgr5DTR/eGFP (AKVL) animals (Fig. 1a) carry two of the most frequently mutated genes found in human colorectal cancer and, in addition, express a diphtheria toxin receptor (DTR) fused to an enhanced green fluorescent protein (eGFP) under the endogenous regulatory region of Lgr5. As a result, Lgr5+ cells can be visualized and ablated selectively after administration of diphtheria toxin (DT). Although these animals form polyps that contain Lgr5+ cells, the animals are not suitable to study the contribution of these cells to tumour growth because long-term DT treatment results in toxicity linked to expression of Lgr5 in other tissues, including the liver15,16. Therefore, we established primary AKVL organoid tumour cultures amenable for transplantation into recipient animals that express wild-type (WT) Lgr5 (Lgr5WT). As previously described, removal of R-spondin and epidermal growth factor (EGF) from the intestinal crypt culture medium allows the selective in vitro expansion of cystic Apcmin and KrasG12D mutant cells over time17,18,19 (Fig. 1b). To generate organoid models that recapitulate the colorectal cancer progression paradigm2, we sequentially introduced Trp53 (AKVPL) and Smad4 (AKVPSL) gene mutations using CRISPR–Cas9 genome editing (Fig. 1c, d and Methods). The resulting organoids were sequenced to validate the presence of frameshift-inducing insertions and deletions (indels) at the targeted loci (Extended Data Fig. 1a, b), and to confirm the absence of any alterations at predicted off-target sites for the single-guide RNAs (sgRNAs) that were used (Extended Data Fig. 1c, d and Supplementary Table 1). In addition, we functionally validated Trp53 and Smad4 pathway inactivation by analysing the expression levels of canonical target genes (Extended Data Fig. 1e, f). Finally, we determined the level of chromosomal instability, a hallmark of colorectal cancer progression, and observed moderate aneuploidy in both the AKVL and AKVPL mutants, whereas extensive aneuploidy was evident in AKVPSL organoids (Fig. 1e, f).

Figure 1: Engineering, characterization and engraftment of mouse-derived intestinal tumour organoids.
figure 1

a, Genotype used to study the role of Lgr5+ cells in a mouse model of colorectal cancer. AKVL, A: Apcmin/+; K: KrasLSL-G12D/+; V: Vil1Cre; L: Lgr5DTR/eGFP. b, AKVL organoids were established in ENR (E: EGF; N: noggin; R: R-spondin) media before Apcmin/KrasG12D mutant clones were selected in noggin-only media. c, CRISPR–Cas9 strategy used to generate mutant organoid lines. P: Trp53; S: Smad4; green depicts Lgr5-GFP+ cells among single cells and within organoids. d, Culture conditions for selection of mutant organoids. e, Copy-number alterations in each organoid line (red, gain; blue, loss; Chr: chromosome). f, Per cent genome alteration for each organoid line. g, Growth rate of subcutaneous organoid allografts. Mean tumour volumes ± s.e.m. are shown for AKVPSL (n = 8), AKVPL (n = 5) and AKVL (n = 4). h, Representative images of Ki67 staining of subcutaneous tumours (n = 4 per organoid line). Scale bars, 125 μm (b), 500 μm (d), 20 μm (h). *P < 0.05, **P < 0.01; n represents the number of independent animals.

PowerPoint slide

Source data

Full size image

To test the effect of the increasing number of driver mutations and chromosomal instability on tumour growth in vivo, we subcutaneously transplanted organoids from each genotype in mice and followed their progression longitudinally. While AKVL- and AKVPL-derived tumours exhibited indolent and moderate growth, respectively, AKVPSL-derived tumours grew exponentially (Fig. 1g). In line with these distinct growth properties, histopathological evaluation showed a low proliferative index for AKVL tumours, and progressively increasing proliferation in AKVPL and AKVPSL grafts (Fig. 1h). To compare the tumorigenic capacity of AKVPL and AKVPSL organoids directly, we transplanted limiting dilutions of bulk tumour cells from each line in vivo. The tumour-initiating capacity, an attribute closely associated with CSC potential20, was significantly higher in AKVPSL compared to AKVPL (Extended Data Fig. 2a). Therefore, it seems that the growth characteristics of our engineered mouse organoid allografts mimic the gradual disease progression in patients as they acquire both these specific mutations2 and chromosomal instability21,22. These data are also consistent with previous reports demonstrating that the number of driver mutations in human-derived intestinal organoids is proportional to their tumorigenic capacity and the extent of chromosomal instability17,18.

Lgr5 marks tumour-initiating cells

We next evaluated the effect of genotype on phenotypic attributes of CSCs. Immunostaining and fluorescence-activated cell sorting (FACS) of tumours collected five weeks after engraftment revealed an increase in the fraction of Lgr5-GFP+ cells in AKVPSL compared to AKVPL grafts (Extended Data Fig. 2b, c), suggesting that the increased tumorigenic potential correlates with an increased CSC content. This appears to be at least in part a cell-intrinsic property, as this correlation was also observed in cultured organoids that lack stroma (Extended Data Fig. 2d, e). Previous analysis of Lgr5+ cells in mouse models of intestinal tumorigenesis had mainly been restricted to early-adenomatous lesions8, which are typically benign tumours and probably follow the growth dynamics of AKVL allografts. To define the molecular properties of these cells in more aggressive tumours, we transcriptionally profiled FACS-isolated Lgr5-GFP+ cells from AKVPL and AKVPSL tumours (Extended Data Fig. 3a). These cells were enriched for genes previously shown to be expressed by normal intestinal stem cells23 and also showed increased expression of canonical Wnt target genes24, when compared to Lgr5-GFP− tumour cells (Extended Data Fig. 3b, c). To further define the tumorigenic potential of Lgr5-GFP+ cells within each genotype, we performed limiting-dilution transplants of FACS-sorted tumour cells. Tumour-engraftment analysis revealed that tumour-initiating capacity was enriched in the Lgr5-GFP+ tumour-cell fractions (Extended Data Fig. 3d, e). These grafts contained both Lgr5-GFP+ and Lgr5-GFP− cells, confirming their ability to self-renew and to give rise to more differentiated progeny. Taken together, our results demonstrate that Lgr5+ cells exhibit important biological properties closely associated with cancer stemness. Notably, tumours that eventually grew after transplantation of Lgr5− cells also contained a sub-population of Lgr5+ cells, which is consistent with the concept of stemness being a dynamic rather than a fixed trait in colon cancer9,25 (Extended Data Fig. 3f).

Lgr5+ CSC depletion does not lead to tumour regression

Next, we took advantage of the Lgr5DTR/eGFP allele to study the requirement of Lgr5+ CSCs during tumour growth. AKVPL organoids that had been transplanted into immune-compromised Lgr5WT animals allow tumour-specific Lgr5+ cell ablation for extended periods of time (Fig. 2a, b). Administration of DT resulted in tumour stasis for as long as DT treatment was continued (Fig. 2b, c). Notably, treatment discontinuation quickly led to re-initiation of tumour growth (Fig. 2c), indicating that sustained Lgr5+ cell depletion is required to prevent tumour progression. The rapid regrowth of tumours even after a prolonged (three-week) period of DT treatment directly challenges the idea that selectively killing colon CSCs may lead to colorectal tumour regression26.

Figure 2: Selective depletion of Lgr5+ cells reveals a dependency on CSCs for subcutaneous tumour growth.
figure 2

a, Strategy for tumour-specific Lgr5-lineage depletion. s.c., subcutaneous. b, Diphtheria toxin (DT) dosing schedule (applied every other day during the ‘DT on’ phase). c, Growth rate of subcutaneous AKVPL tumours in response to treatment. Mean tumour volumes ± s.e.m. are shown for saline (n = 11) and DT (n = 10). d–g, Representative images of Lgr5-GFP+ staining of subcutaneous AKVPL tumours collected either during or one week after DT treatment (n = 4 per condition). h, Per cent Lgr5-GFP+ cells in subcutaneous AKVPL tumours collected at indicated time points (mean ± s.d., n = 4 tumours per condition). i, Immunostaining (top) and FACS analysis (bottom) of Lgr5-GFP+ cells in cultured AKVPL organoids either during or 4 days after DT treatment. Representative images and FACS plots are shown from two independent experiments. j, Quantitative set analysis for gene expression (QuSAGE) of indicated gene signatures in DT- versus saline-treated subcutaneous AKVPL tumours 24 h after treatment. Scale bars, 20 μm (d–g), 10 μm (i). ***P < 0.001; NS, not significant; n represents the number of independent animals.

PowerPoint slide

Source data

Full size image

The tumour stasis observed during continuous depletion of CSCs prompted us to further explore the dynamics of Lgr5+ cell ablation. We performed immunostaining and FACS analysis of tumour tissues at various time points during treatment (Fig. 2d–h). A single dose of DT resulted in a complete loss of all Lgr5-GFP+ cells (Fig. 2d, e and Extended Data Fig. 4a, b), and these cells remained absent from the tumour mass upon prolonged DT treatment (Fig. 2f). However, tumours that were collected one week after DT treatment had regained their full complement of Lgr5-GFP+ cells when compared to vehicle-treated controls (Fig. 2g, h). These newly recovered Lgr5-GFP+ cells functionally contributed to tumour regrowth, as DT re-dosing affected tumour growth again (Extended Data Fig. 4c, d). Notably, the dynamics of tumour growth in response to DT-mediated depletion of Lgr5-GFP+ cells was highly similar between genotypes (Extended Data Fig. 4e, f) and was fully recapitulated in organoid cultures (Fig. 2i), indicating that Lgr5+ cell recovery can take place in the absence of tumour stroma.

To gain better insights into the molecular mechanisms that contribute to tumour regrowth, we performed a time-course analysis of transcriptional changes that occurred in tumour cells after a single dose of DT in vivo. Differential expression analysis showed that the molecular profiles of DT-treated tumours were highly divergent from vehicle-treated tumours 24 h after DT treatment, but became concordant again following a 96-h recovery (Extended Data Fig. 5a, b and Supplementary Table 2), consistent with the idea that the Lgr5+ cell state is reversible. Gene ontology analysis indicated that the transcriptome of DT-treated tumours at 24 h after DT treatment displayed an interferon-α response and activation of Myc signalling (Fig. 2j and Supplementary Table 3). While the former may reflect the response of tumour cells to tissue injury and cell death, the latter is reminiscent of normal intestinal tissue regeneration after damage27. We also found an overrepresentation of genes involved in cell-cycle progression (Extended Data Fig. 5c). Accordingly, proliferation was readily detected in tumours during the course of DT treatment (Extended Data Fig. 5d). This observation is in line with the rapid tumour regrowth observed after cessation of DT treatment, and is also consistent with a recent study where Lgr5+ cells were pharmacologically depleted using an antibody drug conjugate6. Furthermore, the molecular profile of Lgr5− cells did not reveal an enrichment of one particular intestinal cell lineage after DT treatment (Extended Data Fig. 6a and Supplementary Table 4). Therefore, the reappearance of Lgr5-GFP+ cells after DT treatment is reminiscent of the stem-cell plasticity observed in the normal intestine upon Lgr5+ cell depletion or injury15,28, where various types of Lgr5− cells can revert back to a Lgr5+ stem-cell state15,29,30. Notably, we did not observe an obvious increase in differentiation upon Lgr5+ cell ablation (Extended Data Fig. 6b). This is in contrast to the situation we have previously described for RSPO3-fusion-positive colorectal cancer31. In this tumour subset, interference with RSPO3 signalling led to a rapid loss of CSC function, which resulted in reduced tumorigenicity with a concomitant increase in differentiation. Thus, depleting CSCs from the tumour cell population is not the same as blocking oncogenic signals that are essential to acquire and/or maintain cancer stemness. Collectively, our data show that tumours are maintained by proliferating Lgr5− cells in the absence of Lgr5+ CSCs. However, these Lgr5− cells are not functionally equivalent to Lgr5+ CSCs, as they cannot fuel tumour growth as efficiently at least in the context of the AKVPL and AKVPSL genotypes analysed here.

Lgr5+ cells are essential for metastasis

Although hypothesized by many, evidence supporting the causal relationship between CSCs and metastasis is still lacking. Unfortunately, our subcutaneously transplanted tumours never led to the development of metastasis. To circumvent this issue, we developed an orthotopic model by injecting organoids directly into the colonic mucosa (Fig. 3a), which yielded tumours that grew at a single location within the colon (Fig. 3b). We found that mice with an AKVPSL tumour in the colon had reduced survival rates compared to animals with an AKVPL graft at the same orthotopic site (Fig. 3c), in line with clinical evidence suggesting that the level of chromosomal instability is an important predictor of patient overall survival21. We next transduced organoids with a lentivirus co-expressing red fluorescent protein (RFP) and luciferase (Luc) to facilitate FACS identification of all tumour cells and to monitor tumour growth by bioluminescence imaging (Fig. 3d). AKVPL tumours only grew at the site of injection and showed no evidence of tumour cell dissemination to other organs, when they were evaluated six weeks after tumour implantation in the colon (Fig. 3d). By contrast, AKVPSL tumours disseminated as early as three weeks after injection, primarily to the liver (Fig. 3d–f). Furthermore, both the fraction of AKVPSL-injected animals that displayed disseminated disease as well as the extent of metastatic burden per animal increased over time (Fig. 3e, f). Notably, FACS analysis revealed enrichment of Lgr5-GFP+ cells in micro-metastasis (5 weeks after injection) compared to both primary tumours and macro-metastasis (6 weeks after injection) (Fig. 3g), suggesting that dissemination and/or colonization of distant sites may originate from Lgr5+ CSCs.

Figure 3: Lgr5+ CSCs are required for the formation of liver metastasis.
figure 3

a, Schematic for orthotopic implantation of organoids. b, Growth of an orthotopic AKVPL tumour monitored by endoscopy at the indicated time points. c, Kaplan–Meier survival curve of mice with orthotopic AKVPL (n = 9) or AKVPSL (n = 13) tumours (***P < 0.001 by log-rank test). d, Representative bioluminescent images of mice with an orthotopic AKVPL (n = 5) or AKVPSL (n = 6) tumour (L.M.+, positive for liver metastasis). e, Quantification of the bioluminescence signal in the liver of mice with an orthotopic AKVPSL tumour at 3 (n = 8), 4 (n = 10), 5 (n = 12) and 6 (n = 7) weeks after implantation. c.p.m., counts per minute. f, Representative images of liver sections stained with haematoxylin and eosin from animals with orthotopic AKVPSL tumours collected at indicated time points (n = 6 per time point). Boxed areas show tumour cells at a higher magnification. g, Per cent Lgr5-GFP+ cells in primary tumours and corresponding liver metastasis determined by FACS at 5 (n = 11) and 6 (n = 6) weeks after implantation. h, DT dosing schedule for Lgr5-lineage ablation in orthotopic tumours. i, Representative bioluminescent images of mice with an orthotopic AKVPSL tumour after the indicated time and treatment (n = 12 for saline and n = 8 for DT). j, Representative images of liver sections stained with haematoxylin and eosin from saline- and DT-treated animals with an orthotopic AKVPSL tumour (n ≥ 6 per treatment group; asterisks indicate metastatic foci). k, Number of RFP+ tumour cells in enzymatically dissociated livers of mice with an orthotopic AKVPSL tumour after the indicated treatment (FACS data are represented as means (lines) and individual tumours (circles or squares); n ≥ 8 per treatment group). w.p.i., weeks post-injection. Scale bars, 500 μm (f, j). **P < 0.01, ***P < 0.001; n represents the number of independent animals.

PowerPoint slide

Source data

Full size image

To further investigate the importance of Lgr5+ CSCs during the metastatic process, we initiated DT treatment two weeks after implantation of AKVPSL organoids in the colonic mucosa (Fig. 3h), well before metastasis can be detected by either bioluminescence imaging or histological evaluation of the entire liver (Fig. 3f). After three weeks of DT treatment, we analysed liver tissues from animals with equivalent tumour burden at the orthotopic site (Extended Data Fig. 7a). Notably, even though DT treatment did not impair the ability of tumour cells to invade the submucosa (Extended Data Fig. 7b), both bioluminescence imaging and histopathological analysis showed a substantial decrease in liver metastatic burden (Fig. 3i, j and Extended Data Fig. 7c, d). In addition, FACS analysis of enzymatically dissociated livers demonstrated a marked decrease in the absolute number of RFP+ tumour cells following treatment with DT (Fig. 3k), confirming that CSCs play a noteworthy role in the initiation of metastasis.

Lastly, we evaluated the requirement of Lgr5+ CSCs for the maintenance of established liver metastasis, as patients with metastatic disease currently have a poor prognosis and represent an important unmet medical need. Injection of organoids directly into the portal vein led to the formation of liver metastasis in the absence of a primary tumour (Fig. 4a and Extended Data Fig. 8a, b). Remarkably, DT treatment profoundly reduced tumour burden in the liver (Extended Data Fig. 8c) with little evidence of metastasis up to two weeks after the last dose of DT (Fig. 4b–g and Extended Data Fig. 8d–f). Livers from control mice displayed extensive metastatic disease, whereas residual disease in DT-treated animals was limited to occasional metastatic foci (Fig. 4h, i). Collectively these data are consistent with a notable role for Lgr5+ CSCs in supporting liver metastasis and highlight that their functional contribution during colorectal cancer maintenance is influenced by tissue location and tumour microenvironment.

Figure 4: Lgr5+ CSCs are essential for the maintenance of colon-cancer-derived liver metastasis.
figure 4

a, Schematic for injection of portal veins with organoids to establish liver metastasis without a primary tumour. b, DT dosing schedule of mice with liver metastasis. c–e, Representative bioluminescent images (c) and corresponding livers (d, e) from either saline- or DT-treated animals with liver metastasis that were collected one week after treatment (n = 6 per treatment group). f, g, Quantification of the bioluminescence signal in saline- and DT-treated animals with liver metastasis either 1 (f; n = 6 per group) or 2 (g; n = 7 per group) weeks after treatment. h, Representative images of liver sections stained with haematoxylin and eosin from saline- and DT-treated animals with liver metastasis (n = 4 per time point and treatment group). Note the extensive metastatic involvement in control livers, whereas only micro-metastases are present in livers from DT-treated animals. Scale bars, 300 μm (h). **P < 0.01; n represents the number of independent animals.

PowerPoint slide

Source data

Full size image

Discussion

Together, our data provide a comprehensive analysis of Lgr5+ CSCs in colon tumorigenesis, including their role in tumour initiation, growth and metastasis. Contrary to expectations, depletion of CSCs in primary tumours does neither lead to exhaustion of the proliferative transit-amplifying cell compartment nor to tumour regression. Instead, plasticity appears to enable tumour maintenance once Lgr5+ cells have been ablated, although it is not able to fuel tumour growth as efficiently. Indeed, in the absence of Lgr5+ CSCs, tumours are maintained by proliferative Lgr5− cancer cells, which constantly attempt to replenish the CSC pool. Lgr5+ CSCs reappear upon treatment discontinuation, leading to rapid tumour regrowth. This is reminiscent of what has been observed in the normal intestine, where homeostasis is maintained by alternative cell compartments that compensate for Lgr5+ cell loss32, but these are unable to drive regeneration following injury caused by irradiation or chemotherapy33. While we surmise that CSCs are functionally relevant for primary tumour growth, our work suggests that blocking the biological cues mediating CSC plasticity will be required to provide a durable benefit9.

Additionally, our findings underscore a critical role for Lgr5+ CSCs during the development and maintenance of colorectal-cancer-derived metastasis and imply that patients with unresectable liver metastasis potentially may benefit from targeting CSCs. While the exact mechanism by which distinct environmental cues influence tumour regrowth after CSC depletion remain to be fully established, our data suggest that certain tissue-specific milieux may be more permissive (that is, colon) than others (that is, liver) for plasticity of epithelial tumour cells. Indeed, a similar requirement for CD36+ cells during dissemination, but not primary tumour growth, was recently observed in oral carcinomas34. The models of orthotopic and metastatic colorectal cancer that we have developed here should permit characterization of these biological responses in greater detail.

Methods

Animal studies

All studies involving animals were approved by Genentech’s Institutional Animal Care and Use Committee and adhere to the NRC Guidelines for the Care and Use of Laboratory Animals. Female NOD.Cg-PrkdcscidIl2rgtm1Wjl/SzJ (NSG) mice were purchased from the Jackson Laboratory (8–12 weeks old; stock 005557). Apcmin/+ (multiple intestinal neoplasia (min));KrasLSL-G12D/+ (loxP-Stop-loxP KrasG12D/+);Vil1Cre (AKV) mice were described previously14 and were crossed with Lgr5DTR/eGFP mice15 to generate AKVL animals.

Subcutaneous transplantation

Organoids were collected from Matrigel plugs using Cell Recovery Solution (Corning) and mechanically dissociated to generate small clusters of cells. Approximately 105 cells were resuspended in organoid basal media, admixed with 50% Matrigel (Corning) to a final volume of 200 μl, and injected subcutaneously in the left flank of NSG mice. Tumour dimensions were measured using calipers and tumour volume was calculated as 0.523 × length × width × width. Animals were humanely euthanized according to the following criteria: clinical signs of persistent distress or pain, significant body-weight loss (>20%), tumour size exceeding 2,500 mm3, or when tumours ulcerated. Maximum tumour size permitted by the Institutional Animal Care and Use Committee (IACUC) is 3,000 mm3 and in none of the experiments was this limit exceeded.

Tumour dissociation and limiting dilution

Tumours were collected and dissociated enzymatically for 30 min with a mixture of collagenase (1.5 mg ml−1; Roche) and hyaluronidase (20 μg ml−1; Sigma-Aldrich) at 37 °C. The cells were then filtered (70-μm pore size), washed and debris was removed by either lympholyte centrifugation (M; Cedarlane) or ACK lysis. The cell suspensions were stained with the following antibodies for flow cytometry analysis: anti-EpCAM (1:1,000, APC-conjugated, Clone G8.8, Biolegends), anti-CD45 (1:1,000, APC-Cy7-conjugated, Clone 30-F11, Biolegends). Dead cells were excluded with Sytox blue (1:1,000, Invitrogen). Gating for Lin−, EpCAM+ singlet, live cells was used to determine the percentage of Lgr5+ cells. Normal colonic cells from an Lgr5WT host were used as control, and GFP+ analysis gates were established so that zero GFP+ events were detectable in Lgr5WT tissue specimens.

For limiting dilution analysis 100, 500 and 2,500 cells from either the 10% lowest or the 10% highest Lgr5-GFP intensities were deposited by FACS in a 96-well plate containing Organoid basal medium, admixed with Matrigel (50%), and injected subcutaneously as described above. Clonal frequency and statistical significance were evaluated with the Extreme Limiting Dilution Analysis (ELDA) limdil function (http://bioinf.wehi.edu.au/software/elda/index.html). Similarly, in vitro clonogenicity was evaluated by limiting dilution deposition of FACS-sorted, live, bulk tumour cells at 1, 2, 4, 6, 8, 12 and 16 cells per well in a flat-bottom 96-well plate precoated with Matrigel.

Orthotopic injection of organoids

The lumen implantation procedure has previously been described14. Briefly, mice were anaesthetized by isoflurane inhalation and injected intraperitoneally (i.p.) with buprenorphine at 0.05–0.1 mg kg−1. Animals were then placed in a supine position, and the extremities secured to a gauze-covered platform with tape. A blunt-ended haemostat (Micro-Mosquito, 13010-12, Fine Science Tools) was inserted approximately 1 cm into the anus. The haemostat was angled towards the mucosa and opened slightly so that a single mucosal fold could be clasped by closing the haemostat to the first notch. The haemostat was retracted from the anus, exposing the clasped exteriorized mucosa. A 10-μl cell suspension of organoids admixed with 50% Matrigel (Corning) was directly injected into the colonic mucosae. After rehydrating the mucosa with phosphate-buffered saline (PBS), the haemostat was released and a blunt gavage needle was used to re-insert the exteriorized colon thereby reversing the rectal prolapse.

Portal-vein injection of organoids

Portal-venous tumour cell injections were performed using a modified experimental mouse model described elsewhere35. Mice were anaesthetized by isoflurane inhalation and an abdominal incision site was shaved and prepped with betadine and alcohol. Mice were placed in a supine position, and the extremities secured on a heating pad with tape to maintain body temperature at around 36 °C. Body temperature was checked using a rectal probe. A median laparotomy extending to the xyphoid cartilage was performed and the duodenum mobilized to expose the portal vein. We injected 100 μl of bulk AKVPSL organoids equivalent to around 50,000 cells into the portal vein of each animal. Tumour cells were slowly injected over the course of 1 min using a 33-gauge needle. After careful removal of the needle, bleeding was stopped by applying gentle pressure on the injection site using cotton-tipped swabs and haemostyptic calcium-alginate fleece (Sorbalgon, Hartmann Inc.). Subsequently, the intestine was placed back, and the muscle layer and the skin were sewn shut using 6-0 vicryl, and ethilon, respectively, in a two-layer technique.

Endoscopy

Endoscopic imaging was performed exactly as described14. Before imaging, mice were anaesthetized by isoflurane inhalation, placed in a supine position, and their colons evacuated of stool using a gavage needle. Endoscopic imaging equipment consisted of a Hopkins II 0° straight-forward 1.9-mm outer diameter telescope encompassed by an examination and protection sheath, an Image-1 high-definition three-chip digital camera attached to a Mikata Point Setter telescope-holding system, a fibre-optic-light guide cable connected to a D Light System xenon light source, an electronic CO2 insufflator to maintain colon insufflation during imaging, and an AIDA Connect high-definition documentation system connected to a high-definition colour monitor (Karl Storz). Endoscopic videos were reviewed using VLC Media Player (VideoLAN Team) and still images were captured from these videos.

Overall survival

Animal status was recorded up to 60 days after orthotopic implantation of tumour cells. Morbidity criteria during that period of time included: animals found dead, moribund animals that had to be euthanized owing to severe weight loss (>20%), rectal prolapse that could not be reversed and animals that showed clinical signs of persistent fatigue or distress. Survival data were analysed using the standard Kaplan–Meier analysis and survival curves were compared using a log-rank test (Prism).

Bioluminescence imaging

Mice were anaesthetized using isoflurane, injected i.p. with 200 μl of 25 mg ml−1 d-luciferin (Invitrogen, L2912) using a tuberculin syringe, and imaged on the Photon Imager (BioSpace Laboratory). During image acquisition, animals continued to receive anaesthesia from a nose-cone delivery system, while their body temperatures were maintained on a thermostatically controlled platform. Photon counts per min per cm2 of observational area were calculated and compared using M3 Vision software (BioSpace Laboratory).

Diphtheria toxin and BrdU treatments

DT (Millipore, # 322326) was administered via i.p. injection at a dose of 50 μg per kg every other day for the time indicated in each experimental setup. Proliferative cells were labelled by a single i.p. injection of BrdU (30 mg per kg, BD Bioscience) two hours before tumour collection. For in vitro treatment of organoids, DT was added to the culture medium at a final concentration of 100 ng ml−1.

Culture of mouse intestinal AKVL organoids

AKVL mouse organoids were established from isolated crypts collected from the small intestine and maintained as previously described19. Organoids were first established in basal culture medium containing advanced DMEM/F12 medium (Invitrogen), B27 (Invitrogen), N-acetylcysteine (Sigma-Aldrich), noggin (Peprotech), R-spondin 3 (Genentech), and EGF (Peprotech). One week after seeding, EGF and R-spondin 3 were omitted from the basal media to select for KrasG12D and Apcmin mutant organoids.

Genome editing

Gene-specific sgRNAs were designed to target coding regions at the 5′ end of Trp53 and Smad4 to maximize disruption. Guides with the best off-target score based on the number and location of mismatches were cloned into pLKO-SHC201, for co-expression with Cas9. At least 3 independent gRNAs were tested per gene. Organoids were dissociated into a single-cell suspension using accutase (Sigma-Aldrich) for 5 min at 37 °C and then transfected with 1 μg of DNA using Lipofectamine 2000 (Invitrogen). Cells were washed twice with basal media 6 h after transfection, embedded in Matrigel and cultured in basal media supplemented with 1 μM of Y-27632 (Sigma-Aldrich). Trp53-mutant organoids were selected by growth in the presence of 10 μM Nutlin-3 (Sigma-Aldrich) 48 h after transfection, whereas Smad4-mutant organoids were cultured in the absence of Noggin and presence of TGF-β (Peprotech, 10 ng ml−1).

Lentivirus production

Viral particles were produced as previously described36. In brief, HEK293T cells were plated on 15-cm dishes in high glucose DMEM with 10% heat inactivated FBS 24 h before transfection. Lentiviral supernatants were prepared by co-transfection using 6 μg of pGEIL-tRFP (EF1α-tRFP-IRES-Luc2), 12 μg of the packaging vector ∆8.9, 3 μg of the envelope vector pVSV-G (Clontech) and the transfection reagent GeneJuice (Novagen). The culture medium was replaced 12 h after transfection and viral supernatant was collected 24 h later, filtered through a 0.45-μm PES filter (Nalgene) and stored at 4 °C until further processing. Viral supernatants were concentrated 200-fold by ultracentrifugation at 100,000g for 1 h and 30 min. Viral pellets were resuspended in organoid basal media and stored at −80 °C.

Immunostaining

Samples were formalin-fixed and paraffin-embedded using standard procedures. For Ki67 antibody stains, rehydrated sections were pressure-cooked for 15 min in antigen unmasking buffer (DAKO), blocked in serum-free protein block (DAKO) and incubated in anti-Ki67 (1:400, Sigma-Aldrich) overnight. Sections were then incubated in HRP-conjugated anti-rabbit antibody (DAKO) and detected with DAB reaction (DAKO). For immunofluorescence stains, sections were incubated overnight with either anti-CC3 (1:400, rabbit, Cell Signaling), anti-BrdU (1:400, rat, Rockland) or anti-GFP (1:100, Cell Signaling clone D5.1, XP) antibodies. The signal was further amplified by TSA (Molecular Probes, T20932) according to the manufacturer’s instructions. Anti-rabbit and anti-rat conjugated to either Alexa Fluor-488 or -555 were used as secondary antibodies (1:500; Molecular Probes). Pictures were acquired with an NIKON A1R using N3 elements software.

qPCR

RNA was isolated from samples using either the RNeasy Micro or Mini kit (74004 and 74104; Qiagen). qPCR was performed in 10-μl reactions with 50 ng total RNA using the One-step Real-time RT–PCR mastermix (4392938; Life Technologies) according to the manufacturer’s instructions. Taqman probes Actb (Mm01205647_g1) and p21 (Mm04205640_g1) were from Life Technologies. qPCR reactions were run on a 7900HT Fast Real-Time PCR system (ABI) at the following thermal cycling conditions: holding step of 30 min at 48 °C followed by a holding step of 10 min at 95 °C and 40 cycles of 10 s at 95 °C and 1 min at 60 °C.

RNA-sequencing

RNA-sequencing (RNA-seq) libraries were prepared with the TruSeq RNA Sample Preparation kit (Illumina, CA). The libraries were sequenced on Illumina HiSeq 2500 sequencers to obtain on average 34 million 50-bp single-end reads per sample. RNA-seq reads were first aligned to ribosomal RNA sequences to remove ribosomal reads. The remaining reads were aligned to the mouse reference genome (NCBI Build 38) using GSNAP37 version 2013-10-10, allowing a maximum of two mismatches per 50-base sequence (parameters: -M 2 -n 10 -B 2 -i 1 -N 1 -w 200000 -E 1–pairmax-rna = 200000–clip-overlap). Transcript annotation was based on the RefSeq database (NCBI Annotation Release 104). To quantify gene expression levels, the number of reads mapped to the exons of each RefSeq gene was calculated. Read counts were scaled by library size, quantile normalized and precision weights were calculated using the voom R package38. Subsequently, differential expression analysis on the normalized count data was performed using the limma R package39. In addition, gene expression was obtained in the form of normalized reads per kilobase gene model per million total reads (nRPKM) as described in ref. 40.

Whole-exome and low-coverage whole-genome sequencing

Exome capture was performed using Agilent’s SureSelect Mouse 50Mb baits and processed using the SureSelect version 1.5.1 protocol. Both whole-exome sequencing (WES) and whole-genome sequencing (WGS) were performed on Illumina HiSeq 2500 sequencers to obtain 75 bp paired-end reads with an average of 32 million and 111 million fragments per sample, respectively. Reads were aligned to the mouse reference genome (NCBI Build 38) using GSNAP37 version 2013-10-10, allowing a maximum of two mismatches per 75-base sequence (parameters: -M 2 -n 10 -B 2 -i 1–pairmax-dna = 1000–terminal-threshold = 1000–gmap-mode = none–clip-overlap).

Genotyping

We used RNA-seq data to validate the CRISPR–Cas9 modification of the Trp53 and SMAD4 loci. For this purpose, reads in the targeted regions were extracted from the original RNA-seq mapping data with Samtools (version 1.2)41. Subsequently, GATK (version 3.5-0) HaplotypeCaller42 was used to assemble haplotypes and assign reads to them (parameters: -bamWriterType ALL_POSSIBLE_HAPLOTYPES -U ALLOW_N_CIGAR_READS). Haplotype frequencies were determined using a custom in-house script.

Identification of CRISPR off-target effects

We used WES data from organoids and matching spleen to identify potential mutations in coding regions that may have been introduced by off-target binding of the CRISPR–Cas9 complex. More specifically, we used loFreq43 to call single-nucleotide variants (SNVs) and indels by comparing WES from organoids of the three genotypes to WES from the spleen. At the same time, we predicted putative off-target binding regions of the gRNAs using an in-house algorithm. The predicted regions were queried for the presence of SNVs and indels, and were visually inspected using IGV44.

Identification of copy-number alterations

To identify chromosomal aberrations we called copy-number changes in each of the three genotypes by comparing their WGS data to the sequencing data from the spleen of the donor using Control-FREEC version 9.9 with a window size of 50 kb45.

Gene-set analysis

We performed quantitative set analysis for gene expression (QuSAGE)46 to identify relevant biological processes associated with Lgr5+ cell identity and depletion. For that purpose, we contrasted Lgr5+ with Lgr5− samples or DT-treated with saline-treated samples, respectively. For each contrast we then calculated the gene-set activity (that is, the mean difference in log expression of the individual genes that compose the set) for each of the 12,984 curated gene sets in the collection from MSigDB47 and for selected sets from ref. 23.

Statistics

Data are represented as mean ± s.e.m., unless stated otherwise. None of the experiments were blinded and no statistical methods were used to pre-determine sample size for in vivo experiments. The number of animals used for each experiment is specified in the relevant figures and legends, and was estimated based on the variability in tumour take rate and growth observed for each model. With regards to randomization, for subcutaneous transplantation, animals were distributed among treatment groups when tumours reached a mean volume of approximately 200–400 mm3. For orthotopic and portal-vein injections, mice with comparable bioluminescence signal intensity were randomized into treatment groups. Unless stated otherwise, statistical significance between two groups was determined by the Mann–Whitney U-test (Prism version 6.0, GraphPad Software, Inc.). All experiments were reproduced at least twice. To determine statistical significance of gene set activity, false discovery rate (FDR)-adjusted P values were obtained by comparing the probability distribution function of log fold changes in a given gene set (highlighted in grey) to a baseline value of zero using a one-sided test.

Data availability

All sequencing data from this study have been deposited in the Gene Expression Omnibus under accession number GSE92966. Source Data files underlying the graphical representation of all in vivo experiments can be accessed in the online version of this paper. All other data are available from the corresponding author upon reasonable request.