Spatiotemporally resolved colorectal oncogenesis in mini-colons ex vivo – Nature

Mice

Apc^fl/fl mice (a gift from T. Petrova) were crossed to Cdx2-cre^ERT2 mice (The Jackson Laboratory). Apc^fl/flCdx2-cre^ERT2 mice (termed A) were then crossed with Kras^LSL-G12D/+Trp53^fl/fl mice (a gift from E. Meylan) to generate Apc^fl/flKras^LSL-G12D/+Trp53^fl/flCdx2-cre^ERT2 mice (termed AKP). AKP mice were then back-crossed with C57BL6/J (The Jackson Laboratory) to generate Apc^fl/flKras^LSL-G12D/+Cdx2-cre^ERT2 mice (termed AK).

To induce tumorigenesis in vivo, Cre^ERT2 recombinase was activated at the age of 8–10 weeks by a single intraperitoneal injection of 18 mg kg^–1 tamoxifen (Sigma-Aldrich, T5648) in sunflower oil. Tumours were allowed to develop for 6 weeks. Mice were then sacrificed for tissue and cell isolation. See also below the specific section for transplantation of organoids in immunocompromised mice.

All animal work was conducted in accordance with Swiss national guidelines, reviewed and approved by the Service Veterinaire Cantonal of Etat de Vaud (VD3035.1 and VD3823). These regulations established 800 mm³ as the maximal subcutaneous tumour volume allowed, which was not exceeded in any of the experiments. In experiments in which tumorigenesis was induced in vivo, the locomotion, appearance, body condition and intestinal function of the mice were monitored twice weekly and assigned numerical scores to allow quantitative decision making in case humane end points were necessary before the predefined end point of the experiment (6 weeks). All of the mice in this study reached the predefined end point. Mice were kept in the animal facility under EPFL animal care regulations. They were housed in individual cages at 23 ± 1 °C and 55 ± 10% humidity under a 12 h–12 h light–dark cycle. All of the animals were supplied with food and water ad libitum.

OptoCre module plasmid generation

The OptoCre module was designed by integrating the following constructs: (1) FUW-M2rtTA, which constitutively expresses the reverse tetracycline transactivator (rtTA); (2) FUW-tetO-GAVPO, which expresses the light-switchable trans-activator GAVPO after rtTA binding in the presence of doxycycline; and (3) FUW-OptoCre, which expresses Cre recombinase after GAVPO binding in the presence of blue light (Extended Data Fig. 1a). FUW-M2rtTA was purchased from Addgene (20342). Vectors containing GAVPO and the GAVPO-binding promoter (UASG)₅-P_min, developed previously⁴¹, were a gift from M. Thomson⁴². For FUW-tetO-GAVPO generation, GAVPO was subcloned into the doxycycline-responsive FUW-TetO backbone (Wernig Lab, Stanford) using the EcoRI and NheI restriction sites (Extended Data Fig. 1a). For FUW-OptoCre generation, (UASG)₅-P_min was inserted into the FUW-TetO backbone from which the TetO promoter had been removed (Wernig Lab, Stanford) using the BstBI and BamHI restriction sites. We then introduced the Cre recombinase (Addgene, 25997) downstream of (UASG)₅-P_min using the Pac1 restriction sites (Extended Data Fig. 1a).

Isolation of colon cells

Healthy colon or tumour pieces were finely chopped using a scalpel and transferred to a gentle-MACS C-tube (Miltenyi, 130-093-237) containing 4 ml of digestion medium (RPMI (Thermo Fisher Scientific, 22400089), 1 mg ml^–1 collagenase type IV (Life Technologies, 9001-12-1), 0.5 mg ml^–1 dispase II (Life Technologies, 17105041) and 10 μg ml^–1 DNase I (Applichem, A3778)). Tissues were then digested using the 37C_m_TDK_1 program on the gentle-MACS Octo Dissociator with heaters (Miltenyi). After the program was complete, the cell suspension was passed through a 70-μm strainer (Corning, 431751) and centrifugated at 400g for 5 min.

Organoid and stromal cell culture

To establish organoids, colon cells were embedded in growth-factor-reduced Matrigel (Corning, 356231) (~2 × 10⁴ cells per 20 μl dome) and cultured in Advanced DMEM/F-12 (Thermo Fisher Scientific, 12634028) supplemented with 1× GlutaMax (Thermo Fisher Scientific, 35050038), 10 mM HEPES (Thermo Fisher Scientific, 15630056), 100 μg ml⁻¹ penicillin–streptomycin (Thermo Fisher Scientific, 15140122), 1× B-27 supplement (Thermo Fisher Scientific, 17504001), 1× N2 supplement (Thermo Fisher Scientific, 17502001), 1 mM N-acetylcysteine (Sigma-Aldrich, A9165), 50 μg ml⁻¹ primocin (InvivoGen, ant-pm-2), 50 ng ml⁻¹ EGF (Peprotech, 315-09), 100 ng ml⁻¹ noggin (produced at EPFL Protein Production and Structure Core Facility), 500 ng ml⁻¹ R-spondin (produced at EPFL Protein Production and Structure Core Facility), 50 ng ml^–1 WNT3A (Time Bioscience, rmW3aL-010), 10 mM nicotinamide (Calbiochem, 481907) and 2.5 μM Thiazovivin (Stemgen, AMS.04-0017). This full medium is termed ‘WENRNi’. The base version of this medium without EGF, noggin, R-spondin, WNT3A and nicotinamide is referred to as BMGF and was used for the expansion of colon tumour organoids since they do not require the additional growth factors. The base version of BMGF without B-27, N2 and N-acetylcysteine is termed BM or basal medium, and was used for growth-factor deprivation experiments. A detailed protocol describing organoid culture can be found elsewhere⁹. Where indicated, organoids were treated with the following compounds or growth factors: regorafenib (8 μM, Selleckchem, S1178), ripretinib (1 μM, Selleckchem, S8757), infigratinib (1 μM, Selleckchem, S2183), SCF (100 ng ml^–1, PeproTech, 250-03), FGF2 (50 ng ml^–1, Thermo Fisher Scientific, PMG0035) and IGF1 (100 ng ml^–1, R&D Systems, 291-G1-200). Stromal cells were derived from cell suspensions from the primary tissue cultured in EGM-2 MV Microvascular Endothelial Cell Growth Medium-2 (Lonza, CC-3202) on conventional cell culture flasks. This medium selection strategy was followed by magnetic-activated cell sorting (MACS) on EPCAM (Miltenyi Biotec, 130-105-958) according to the manufacturer’s instructions to discard epithelial cells. The presence of stromal cells was further confirmed by immunofluorescence analyses of vimentin expression (see below). Cells were tested for mycoplasma before cryopreservation and in randomized routine checks using the MycoAlert PLUS Mycoplasma Detection Kit (Lonza, LT07-705).

Generation of light-inducible cells

Lentiviral particles carrying the three components of the OptoCre module (see above; Extended Data Fig. 1b) and a Cre recombination reporter were produced at the EPFL Gene Therapy Platform by transfecting HEK293 cells with each plasmid of the OptoCre module and pLV-CMV-LoxP-DsRed-LoxP-eGFP (Addgene, 65726) plasmids. Lentivirus-containing supernatants were collected and concentrated by centrifugation (1,500g for 1 h at 4 °C). Lentiviral titration was performed using the p24-antigen ELISA (ZeptoMetrix, 0801111). For transduction, colon organoids (around 2 × 10⁵ cells) were dissociated into single cells by incubating in TrypLE Express Enzyme (Thermo Fisher Scientific, 12605028) at 37 °C for 5 min. Cells were then washed with basal medium supplemented with 10% fetal bovine serum (FBS) (Thermo Fisher Scientific, 10500064) and resuspended in WENRNi medium containing 8 μg ml⁻¹ polybrene (Sigma-Aldrich, TR-1003-G) and the following amounts of viral particles: ~10 ng of p24 FUW-M2rtTA per ml, ~80 ng of p24 FUW-tetO-GAVPO per ml, ~80 ng of p24 FUW-OptoCre per ml and ~1,000 ng of p24 CMV-LoxP-DsRed-LoxP-eGFP per ml. These cells were plated in a 24-well plate, centrifuged at 600g for 60 min at room temperature, and incubated for 6 h at 37 °C. After incubation, the cells were collected, centrifuged, plated in 20 μl Matrigel domes in a 24-well plate and cultured in WENRNi medium. Cells expressing the Cre recombination reporter were selected by supplementing WENRNi medium with 8 μg ml⁻¹ puromycin (InvivoGen, ant-pr-1).

Light-mediated oncogenic recombination

The OptoCre module requires (1) doxycycline to induce rtTA-mediated GAVPO expression and (2) blue light to induce GAVPO-mediated Cre recombinase expression (Extended Data Fig. 1a,b). At the desired time of oncogenic induction, 2 μg ml⁻¹ doxycycline hydrochloride (Sigma-Aldrich, D3072) was added to the culture medium of either the organoids or mini-colons. Light induction was then performed using a custom-made LightBox built by Baur SA and the Instant Lab at EPFL. The LightBox consisted of an Acqua A5 System (Acme Systems) that could be remotely parametrized using a custom-made web-based application. Communication between the Acqua A5 System and the microcontroller (PJRC, Teensy 3.2) was done through Blocky programming, which allowed for control of the LED drivers (Sparkfun, PicoDuck). The LEDs (Cree LEDs, XLamp XP-C Blue LEDs) were placed into a custom multilayer 24-well plate holder made of black anodized aluminium and polyphenylsulfone; the height was optimized for homogeneous light distribution within each well. The entire LightBox, plate-holder, LEDs and cables were made to be placed in the incubator (watertight and heat resistant). Diffusive elements (Luminit, Light Shaping Diffuser 80°) were used to render the illumination more homogeneous inside each well. The intensity of the blue light (450–465 nm, peak at 455 nm) was optimized, set to 100 μW cm⁻² and shined on the cells for 3 h. After blue-light exposure, doxycycline was removed by washing the cultures with fresh medium. In experiments targeting the light to specific regions of the mini-colon, work was carried out in the dark using a near infrared light (Therabulb, NIR-A) to prevent leaky Cre expression. Light-targeting was performed using a photomask that was adapted to the dimensions of the mini-colon and that was created from a photoresist and chrome-coated standard 5 × 5 inch silica plate (Nanofilm) with an automated machine (VPG200 Heidelberg Instrument, 2.0 µm resolution). Once the exposed photoresist was developed, the chrome layer was wet-etched and the remaining photoresist was stripped using a mask processor (Hamatech HMR900)⁹.

Microdevice design, fabrication and loading

The microfluidic device used for mini-colon cultures was designed using Clewin 3.1 (Phoenix Software) and fabricated as previously described⁹. It was composed of three main compartments: (1) a hydrogel chamber for cell growth in the centre; (2) two basal medium reservoirs flanking the hydrogel compartment; and (3) inlet and outlet channels for luminal perfusion⁹. An extracellular matrix containing 80% (v/v) type I collagen (5 mg ml⁻¹, Reprocell, KKN-IAC-50) and 20% (v/v) growth–factor-reduced Matrigel was loaded into the hydrogel compartment. The microchannels constituting the mini-colon architecture within the hydrogel were designed using Adobe Illustrator CC 2019 and Wolfram Mathematica 11.3. They were then read by PALM RoboSoftware 4.6 (Zeiss) and ablated using a nanosecond laser system (1 ns pulses, 100 Hz frequency, 355 nm; PALM Micro-Beam laser microdissection system, Zeiss). The dimensions of the mini-colon architecture were described previously⁹. A detailed description of all the key steps required for the generation and maintenance of mini-guts is available at Protocol Exchange (https://doi.org/10.21203/rs.3.pex-903/v1)³⁸.

Mini-colon culture, development and tumorigenesis

Colon organoids were dissociated into single cells by incubating in TrypLE Express Enzyme for 5 min at 37 °C followed by vigorous pipetting. This cell suspension was washed in 5 volumes of Advanced DMEM/F-12 supplemented with 10% FBS and passed through 40 μm cell strainers (Corning, 431750). After centrifugation at 400g for 5 min, cells were resuspended in WENRNi medium at around 10⁶ cells per ml. The mini-colon luminal microchannel was filled with 10 μl of this cell suspension. Cells were allowed to settle down in the mini-colon crypt-shaped cavities for 5 min, and the leftover unadhered cells were washed out from the microchannel by medium perfusion. The basal medium reservoirs were filled with 100 μl of WENRNi. Unless otherwise indicated, once the healthy colonic epithelium was fully formed (around 2 days after seeding), the medium in the luminal channel was switched to BM, while WENRNi was kept in the basal medium reservoirs. This gradient of growth factor from basal medium reservoirs to luminal space favours colonocyte differentiation across the crypt–lumen axis. For low-differentiation conditions of the differentiation experiments, WENRNi was kept in both the lumen and basal medium reservoirs. Conversely, high-differentiation mini-colons were cultured in WENRNi medium without WNT3A and nicotinamide (termed ENR). Unless otherwise stated, once the colonic epithelium was fully formed, oncogenic induction in the mini-colons was performed as stated above. Where indicated, tiopronin (5 mM, Selleckchem, S2062) or CHIR99021 (3 μM, StemCell Technologies, 100-1042) was added to the basal medium reservoirs after or before oncogenic induction, respectively. For co-culture experiments, ~500 stromal cells were seeded in each hydrogel before the laser-mediated ablation of the mini-colon pattern. The rest of the culture conditions and procedures remained unchanged. To avoid potential unspecific results derived from the small (but non-zero; Extended Data Fig. 1d,e) leakiness of the optogenetic system, each replication across all studies was performed using independent OptoCre organoid lines freshly generated before each experiment. In all cases, the mini-colons were incubated at 37 °C in 5% CO₂ humidified air, with daily luminal perfusions and medium changes every other day.

Mini-colon whole-mount immunofluorescence staining

Mini-colons were rinsed with phosphate-buffered saline (PBS) and fixed in 4% paraformaldehyde (Thermo Fisher Scientific, 15434389) overnight at 4 °C. After rinsing with PBS, the hydrogels were extracted from the PDMS scaffold using a scalpel, placed into a 48-well plate, permeabilized with 0.1% Tween-20 (Sigma-Aldrich, P9416) in PBS (10 min at 4 °C) and blocked in 2 mg ml⁻¹ bovine serum albumin (Sigma-Aldrich, A3059) in PBS containing 0.1% Triton X-100 (Sigma-Aldrich, T8787) (blocking buffer) for at least 45 min at 4 °C. The samples were subsequently incubated overnight at 4 °C in blocking buffer with the corresponding following primary antibodies: CD44 (1:200; Abcam, ab157107), FABP1 (1:100; R&D Systems, AF1565), SOX9 (1:200; Abcam, ab185966), GPX2 (1:200; Bioss Antibodies, BS-13396R), IL-1α (1:200; R&D Systems, AF-400-SP), CDKN2A (1:100; Abcam, ab211542), E-cadherin (1:100; Abcam, ab11512) and vimentin (1:200; Abcam, ab92547). After three washes in blocking buffer for a total of 6 h at room temperature, the samples were incubated overnight at 4 °C in blocking buffer with the following corresponding secondary antibodies: Alexa Fluor 488 anti-goat (1:400, Thermo Fisher Scientific, A-11055), Alexa Fluor 488 anti-rat (1:400, Thermo Fisher Scientific, A-21208) and Alexa Fluor 647 anti-rabbit (1:400, Thermo Fisher Scientific, A-31573). After 3 washes in blocking buffer for a total of 6 h at room temperature, the samples were incubated with DAPI (1 μg ml⁻¹; Tocris Bioscience, 5748) for 10 min at room temperature in blocking buffer. Before imaging, the hydrogels were mounted onto 35 mm glass bottom dishes (Ibidi, 81218-200) in Fluoromount-G (SouthernBiotech, 0100-01).

Mini-colon sectioning and histochemistry

Mini-colons were fixed and extracted from the PDMS scaffold as indicated above and were prepared for cryosectioning by incubating in 30% (w/v) sucrose (Sigma-Aldrich, S1888) in PBS until the sample sank. Subsequently, the samples were incubated for 12 h in a mixture of Cryomatrix (Epredia, 6769006) and 30% sucrose (mixing ratio 50/50) followed by a 12 h incubation in pure Cryomatrix. The samples were then embedded in a tissue mould, frozen on dry ice, and cut into 40-µm-thick sections at −20 °C using the CM3050S cryostat (Leica). Haematoxylin and eosin staining was performed at the EPFL Histology Core Facility using the Ventana Discovery Ultra automated slide preparation system (Roche).

Microscopy and image analysis

Bright-field and fluorescence imaging of living organoids and mini-colons was performed using the Nikon Eclipse Ti2 inverted microscope with ×4/0.13 NA, ×10/0.30 NA and ×40/0.3 NA air objectives and a DS-Qi2 camera (Nikon Corporation). Time lapses were taken in a Nikon Eclipse Ti inverted microscope system equipped with ×4/0.20 NA and ×10/0.30 NA air objectives and DS-Qi2 (Nikon Corporation) and Andor iXon Ultra 888 (Oxford Instruments) cameras. Both systems were controlled using the NIS-Elements AR software (Nikon Corporation). The extended depth of field (EDF) of bright-field images was calculated using a built-in NIS-Elements function. Fluorescence confocal imaging of fixed mini-colons was performed using the Leica SP8 STED 3X inverted microscope system equipped with ×10/0.30 NA air and ×25/0.95 NA water objectives, 405 nm diode and supercontinuum 470–670 nm lasers, and the system was controlled by the Leica LAS-X software (v.3.5.7, Leica microsystems). Histological sections were imaged using a Leica DM5500 upright microscope with ×10/0.30 NA and ×20/0.75 NA air objectives, a ×40/1.0 NA oil objective and a DMC 2900 Color camera, and the system was controlled by the Leica LAS-X software. Image processing was performed using standard contrast- and intensity-level adjustments in ImageJ (NIH). For oncogenic recombination analyses, the GFP-positive area was measured from 16-bit EDF images by subtracting the background, sharpening the images, and applying a signal threshold and a mask. The ratio between GFP-positive area and total organoid area was used for analyses. Recombined cells were segmented using StarDist with the default parameters (https://github.com/stardist) on the GFP channel of mini-colon images. Cell debris was discarded from segmentation analyses by setting an empirically established size threshold. For tumour quantification in the mini-colon, neoplastic structures with at least three times the thickness of the surrounding healthy epithelium were considered to be tumours. Videos of immunostainings were rendered using Imaris (Oxford Instruments).

Mini-colon shedding evaluation

The medium from the luminal compartments of the mini-colons, together with an additional luminal perfusion of 10 μl of basal medium, was collected every day for 4 days after the blue-light-induced oncogenic recombination. The protein content in these extracts was analysed using conventional Bradford assays (Bio-Rad, 5000006) and used as an indicator of cell shedding.

Mini-colon cell line derivation

Mini-colon-containing hydrogels were extracted from their microfluidic devices with a scalpel as indicated above and incubated with 0.1% (w/v) collagenase I (Thermo Fisher Scientific, 17100-017) at 37 °C for 10 min. Once the hydrogel was fully digested, the mini-colon was washed with PBS and digested with TrypLE Express Enzyme for 5 min at 37 °C. The resulting cell suspension was washed with Advanced DMEM/F-12 supplemented with 10% FBS, pelleted, embedded in Matrigel and cultured as indicated above for regular colon organoids.

Transplantation of organoids in immunocompromised mice

Organoid lines were established as indicated above from either in vivo colon tumours (reference AKP) or tumour-bearing mini-colons (mini-colon AKP). These organoids were dissociated into single cells using TrypLE Express Enzyme for 5 min at 37 °C, washed with Advanced DMEM/F-12 supplemented with 10% FBS, pelleted and embedded in Matrigel at 2.5 × 10⁶ cells per ml. A total of 100 μl of this suspension was inoculated by subcutaneous injection into the right flank of NOD.Cd-Prkdz^scidIl2rg^tm1Wjl/Szj (NSG) mice (Jackson laboratories). Tumour growth was monitored using callipers twice per week until the end point at 18 days after inoculation. Length (L) and width (W) were measured and used to approximate the volume (V) of the tumour in mm³ using the modified ellipsoid formula: V = (L × W²)/2. After euthanasia, tumours were resected from the graft location and measured once more with callipers.

Graft sectioning and histochemistry

Tumour samples were fixed overnight in 4% paraformaldehyde at 4 °C, dehydrated in graded ethanol baths, cleared with xylene, embedded in paraffin and cut into 4-µm-thick sections using the HM 325 Rotary Microtome (Thermo Fisher Scientific). These sections were mounted onto Superfrost plus slides (Epredia, J1800AMNZ) and allowed to dry for 2 days at room temperature. Haematoxylin and eosin staining was performed at the EPFL Histology Core Facility using the Ventana Discovery Ultra automated slide preparation system (Roche).

Mutational screening in colon organoids

Genomic DNA was isolated from colon cells using the PureLink Genomic DNA Mini Kit (Thermo Fisher Scientific, K182001) according to the manufacturer’s instructions. Recombination of the LSL (LoxP-Stop-LoxP) cassette controlling Kras^G12D expression was confirmed by PCR using the protocol and oligos described by the Tyler Jacks laboratory (https://jacks-lab.mit.edu/, Kras^G12D Conditional PCR). Apc and Trp53 recombinations were confirmed through exome sequencing performed at BGI Genomics at 100× coverage using DNBSEQ sequencing technology. DNA reads were mapped to the mouse GRCm39 genome assembly using BWA-MEM (v.0.7.17), filtered using samtools (v.1.9) and visualized using IGV (Integrative Genomics Viewer, Broad Institute, v.2.12.3).

Organoid proliferation assays

Single-cell suspensions of colon cells were generated as indicated above and embedded in 10 μl Matrigel domes at around 10⁴ cells per dome in a 48-well plate. For each of the following 4 days, 220 μM resazurin (Sigma-Aldrich, R7017) was added to the culture medium and incubated for 4 h at 37 °C. Next, the resazurin-containing medium was collected and replaced with regular medium. Organoid proliferation was estimated by measuring the reduction of resazurin to fluorescent resorufin in the medium each day using the Tecan Infinite F500 microplate reader (Tecan) with 560 nm excitation and 590 nm emission filters. In the case of colony-formation assays, seeding was performed at around 10³ cells per dome and the resulting colonies were counted after 3 days.

Organoid RNA extraction and bulk transcriptome profiling

Before RNA isolation, organoids were cultured for 3 days as indicated above and starved for 24 h in BM for the evaluation of growth-factor dependence. In the case of the Gpx2-knockdown experiments, 2 timepoints were analysed: 0 and 2 weeks after blue-light-induced activation (before and after oncogenic recombination, respectively). In all cases, cells were collected using TrypLE Express Enzyme as indicated above and lysed in RLT buffer (Qiagen, 74004), and the RNA was extracted using the QIAGEN RNeasy Micro Kit (Qiagen, 74004) according to the manufacturer’s instructions. Purified RNA was quality checked using a TapeStation 4200 (Agilent), and 500 ng was used for QuantSeq 3′ mRNA-seq library construction according to the manufacturer’s instructions (Lexogen, 015.96). Libraries were quality checked using a Fragment Analyzer (Agilent) and were sequenced in the NextSeq 500 (Illumina) system using NextSeq vm2.5 chemistry with Illumina protocol 15048776. Reads were aligned to the mouse genome (GRCm39) using star (v.2.7.0e)⁴³. R (v.4.1.2) was used to perform the differential expression analyses. Count values were imported and processed using edgeR⁴⁴. Expression values were normalized using the trimmed mean of M values (TMM) method⁴⁵ and low-expressed genes (<1 counts per million) were filtered out. Differentially expressed genes were identified using linear models (Limma-Voom)⁴⁶ and P values were adjusted for multiple comparisons using the Benjamini–Hochberg correction method⁴⁷. Volcano plots and heat maps were generated using the EnhancedVolcano (https://github.com/kevinblighe/EnhancedVolcano) and heatmap3 (https://github.com/slzhao/heatmap3) packages, respectively. The in vivo AKP signature was established from the differentially expressed genes between in vivo and organoid AKP lines with a log₂-transformed fold change of at least |2|. To evaluate the enrichment of the in vivo AKP gene expression program across samples, the enrichment scores for both the upregulated and downregulated signatures were calculated using single-sample GSEA (ssGSEA)⁴⁸. The difference between the two normalized enrichment scores yielded the fit score. ssGSEA was also used to analyse the enrichment of the MSigDB curated Hallmark gene set⁴⁹ in Gpx2-knockdown organoids. Functional annotation was performed using DAVID⁵⁰ on the genes with a log₂-transformed fold change of at least |1|. GOplot⁵¹ was used for the integration of expression and functional annotation data. Known functional interactions among relevant genes were obtained through STRING⁵². Cytoscape⁵³ was used to perform network data integration and visualization.

Single-cell transcriptome profiling and lineage tracing

Lineage tracing was performed using the CellTag system²² (V1 pooled barcode library, Addgene, 115643-LVC). In brief, we co-transduced inducible colon organoids with the CellTag barcode library (multiplicity of infection of around 5) and the OptoCre module as indicated above. These cells were then introduced and induced in the mini-colon system as indicated before. After 7 days in the system and when mini-colon tumours were clearly visible, we extracted the cells from mini-colons as indicated above. After pooling and filtering (40 μm) the cell suspensions from two mini-colons, the single-cell sequencing library was constructed using 10x Genomics Chromium 3′ reagents v3.1 according to the manufacturer’s instructions (10x Genomics, PN-1000269, PN-1000127, PN-1000215). Sequencing was performed using NovaSeq 6000 v1.5 reagents (Illumina protocol #1000000106351 v03) for around 100,000 reads per cell. The reads were aligned using Cell Ranger (v.6.1.2)⁵⁴ to the mouse genome (mm10) carrying artificial chromosomes for both GFP and CellTag UTR genes, as recommended by CellTag developers for facilitating barcode identification⁵⁵. Raw count matrices were imported into R and analysed using Seurat (v.4.2.0)⁵⁶. Dead cells were discarded on the basis of the number of detected genes (less than 3,000) and the percentage of mitochondrial genes (more than 20%), leading to 2,429 cells after filtering. The data were log-normalized and scaled, and dimensionality reduction was conducted using UMAP with 10 dimensions. Louvain clustering yielded 17 clusters that were merged and named on the basis of canonical cell type markers. Stem, cycling, progenitor, goblet and enteroendocrine cell scoring was based on published signatures in mini-intestines and in vivo⁹. Gene sets highlighting bottom, middle and top colonocytes were taken from enterocyte zonation studies²³. Cancer stemness was scored based on the expression of Lgr5, Cd44 and Sox9. Intrinsic consensus molecular subtype (iCMS) signatures for colorectal cancer were obtained from published work²⁷. Signature scoring was performed using burgertools (https://github.com/nbroguiere/burgertools). Visual representations of the data were generated using Seurat internal functions. For lineage-tracing analyses, CellTag detection, quantification and clone calling were performed as indicated by CellTag developers⁵⁵, excluding cells expressing fewer than 2 or more than 30 CellTags. After filtering, 83 clonal populations were identified, from which only those with a minimum size of 5 cells were considered for further analyses. To identify clonal populations belonging to tumour cells, we looked for cells expressing transcripts carrying the genetically engineered Apc and Trp53 mutations, that is, deletions of exons 15 and 2–10, respectively (Extended Data Figs. 3g and 6b,c). Note that this approach could not be performed for Kras, as the mutation is also present in the transcripts from WT cells (but not expressed). As scRNA-seq provides low coverage on exon junctions and therefore the presence of mutations can be assessed only in a small fraction of cells, we used both the cell-type composition and size distributions of bona fide mutationally confirmed tumour clonal populations to classify the rest of clones. Those falling within plus or minus 2 s.d. of the mean cell composition and size of bona fide tumours were classified as tumour clonal populations. Healthy clones were defined as those with a clearly distinct (outside the aforementioned range) cell type composition and the same upper limit size as was observed for tumour clones. After filtering and classification, 16 healthy and 18 tumour clonal populations were obtained and used for further analyses (Extended Data Fig. 6d). To define the most robust tumour-clone-specific markers, the gene expression from cells in each clone was compared to that from cells in each other clone using the Wilcoxon rank-sum test. We considered only the positive markers and selected those with adjusted P < 10⁻⁵. The association of these markers with clinical parameters in patients with CRC (survival, lymph node staging) was performed through cBioPortal (https://www.cbioportal.org/) using the 640-sample CRC TCGA dataset (https://www.cancer.gov/tcga) and a differential expression threshold equal or greater than |2 |. Further information is provided in the Data availability and Code availability sections.

shRNA-mediated transcript knockdown

Organoids were transduced as indicated above with lentiviral particles encoding Gpx2 shRNAs obtained from Sigma-Aldrich (TRCN0000076529, TRCN0000076531 and TRCN0000076532; shGpx2 1, shGpx2 2 and shGpx2 3, respectively) or, as a control, shRNA-free counterparts (Addgene, 65726). Transduced cells were selected with puromycin (5 μg ml⁻¹; InvivoGen, ant-pr-1). Proper transcript knockdown was assessed using quantitative PCR with reverse transcription (RT–qPCR) and RNA-seq.

Analysis of mRNA abundance

Organoids were cultured and collected as indicated above. Cells were then lysed in RLT buffer and RNA was extracted using the QIAGEN RNeasy Micro Kit as indicated above. RT–qPCR was performed using the iTaq Universal SYBR Green One-Step Kit (Bio-Rad Laboratories, 1725150) and the QuantStudio 7 Flex Real-Time PCR System (Thermo Fisher Scientific, 4485701). Raw data were analysed using Design & Analysis Software (v.2.6.0, Thermo Fisher Scientific). We used the abundance of the endogenous Gapdh mRNA as internal normalization control. The following primers were used for transcript quantification: 5′-AGTTCGGACATCAGGAGAACTG-3′ (forward, Gpx2), 5′-GATGCTCGTTCTGCCCATTG-3′ (reverse, Gpx2), 5′-ATCCTGCACCACCAACTGCT-3′ (forward, Gapdh) and 5′-GGGCCATCCACAGTCTTCTG-3′ (reverse, Gapdh).

Microbiota and diet modelling

Inducible mini-colons were generated as indicated above. Once the epithelium was formed and before oncogenic induction, mini-colons were subjected to a conditioning period of 2 days in which luminal medium was (1) supplemented with 100 μM deoxycholate (Sigma-Aldrich, D2510), 10 mM butyrate (Sigma-Aldrich, B5887) or 10 mM β-hydroxybutyrate (Sigma-Aldrich, 54965); or (2) replaced with MEMα (calorie-restricted condition, Thermo Fisher Scientific, 22561-021) or Advanced DMEM/F12 supplemented with 30 μM palmitic acid (calorie-enriched condition, Sigma-Aldrich, P0500). The same concentrations were used in organoid control experiments, but these were added to the full culture medium as the luminal compartment is not accessible in organoids. To assess the relevance of luminal exposure to these factors in the mini-colon, the same total amounts were added in the basal medium reservoirs instead of the luminal channel. In all cases, after conditioning, oncogenic recombination was performed and the mini-colon was cultured as indicated above. The different medium compositions were replenished every day during luminal perfusion.

Statistics and reproducibility

The number of biological replicates (n), the type of statistical tests performed and the statistical significance for each experiment are indicated in the corresponding figure legend. For images associated with quantification charts (Fig. 1b,c with Fig. 1e; Fig. 2b with Fig. 2c; Fig. 2d with Extended Data Fig. 4b; Fig. 2f with Fig. 2g; Fig. 4b with Extended Data Fig. 9a; Fig. 4i with Fig. 4j; Extended Data Fig. 2a with Fig. 1e; Extended Data Fig. 3d,e with Extended Data Fig. 3b; Extended Data Fig. 3h with Fig. 1e; Extended Data Fig. 5b with Extended Data Fig. 5c; Extended Data Fig. 5d with Extended Data Fig. 5e; Extended Data Fig. 9b with Extended Data Fig. 9c; Extended Data Fig. 9e with Extended Data Fig. 9f; Extended Data Fig. 9g with Fig. 4c; Extended Data Fig. 9l with Extended Data Fig. 9m; Extended Data Fig. 9n with Fig. 4g; Extended Data Fig. 10c with Extended Data Fig. 10d; Extended Data Fig. 10e with Extended Data Fig. 10f), the number of replicates is the same as for the corresponding chart and is indicated in the figure legend of the latter. For the rest of representative images (Figs. 1d and 3h and Extended Data Figs. 1f, 2b,c,f, 3a, 4a, 7d, 9h and 10a,g–k), three independent experiments were performed. scRNA-seq (Fig. 3a) and exome sequencing with matched PCR (Extended Data Fig. 3f,g) were performed with two independent sets of samples. Bulk RNA-seq was performed with at least three independent sets of samples. Unless otherwise indicated, statistical analyses were performed using GraphPad Prism v.9 (GraphPad). Data normality and equality of variances were analysed with Shapiro–Wilk and Bartlett’s tests, respectively. Parametric distributions were analysed using the Student’s t-test (when comparing two experimental groups) or ANOVA followed by either Dunnett’s test (when comparing more than two experimental groups with a single control group) or Tukey’s HSD test (when comparing more than two experimental groups with every other group). Nonparametric distributions were analysed using either Mann–Whitney U-tests (for comparisons of two experimental groups) or the Kruskal–Wallis followed by Dunn’s test (for comparisons of three or more than three experimental groups) tests. Sidak’s multiple-comparison test was used when comparing different sets of means. χ² tests were used to determine the significance of the differences between expected and observed frequencies. In all cases, values were considered to be significant when P ≤ 0.05. Data obtained are given as the mean ± s.e.m.

Reporting summary

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