Ablation of p57⁺ Quiescent Cancer Stem Cells Suppresses Recurrence after Chemotherapy of Intestinal Tumors

Cancer stem cells (CSC) constitute a cell fraction of tumors that is able to generate heterogeneous cell populations and thereby drive cancer propagation (1, 2). Evidence suggests that CSCs are also responsible for posttherapeutic recurrence of multiple cancer types (1–7), and they therefore represent a major obstacle to cancer eradication. In general, CSCs are maintained in a nonproliferative state referred to as quiescence, which renders them resistant to conventional chemotherapy or radiotherapy, both of which preferentially target actively cycling cells (1–8). A recent study suggested that quiescence also renders cancer cells resistant to immunotherapy by promoting the formation of a T cell–suppressive niche (9). The identification and characterization of quiescent CSCs would therefore be expected to inform the development of new anticancer therapies that eradicate tumors by targeting this cell population. Although therapeutic targeting of quiescent CSCs has been successful in models of several cancer types (3, 5–7), its efficacy for intestinal malignancies has remained to be demonstrated, given that the identity of quiescent intestinal CSCs has been unclear (2).

Evidence has suggested that cancer cell populations expressing Lgr5 at a high level (Lgr5^high) serve as CSCs in intestinal tumors (10–12). Lgr5 was initially identified as a marker for crypt base columnar cells (CBC) that serve as normal intestinal stem cells (13), and it was subsequently implicated as a CSC marker for intestinal tumors (14, 15). In the normal intestinal epithelium, Lgr5 expression is restricted to only CBCs (13), with the stemness of these cells being maintained by surrounding niche factors at the crypt base (2, 13, 16). On the other hand, as a result of genomic mutations that allow cell-autonomous hyperactivation of stemness, malignant intestinal tumors are freed from restrictions imposed by this stem cell niche and contain a much higher proportion of Lgr5^high cells compared with the normal intestinal epithelium (10, 11, 17). It has remained unclear, however, whether Lgr5, which manifests such a broad expression pattern in tumors, defines an intestinal CSC fraction with specific biological properties comparable with those of normal CBCs. In addition, although previous studies have shown that normal intestinal crypts harbor both actively and slowly cycling CBC subpopulations (18), the heterogeneity of Lgr5^high intestinal CSCs has been poorly understood.

Genetic ablation of Lgr5^high CSCs has been found to suppress intestinal cancer growth in ectopic transplantation models but not in an orthotopic transplantation model (10, 11, 16). These results have been attributed to a permissive intestinal microenvironment that confers plasticity on relatively differentiated cancer cells, allowing them to replenish the CSC fraction immediately after its depletion (10, 16). The combination of CSC-targeting therapy with conventional chemoradiation therapies that eliminate actively cycling progenitors is therefore thought to be necessary to improve therapeutic efficacy for orthotopic primary tumors. However, previous transplantation models have often relied on NOD.Cg-Prkdc^scidIl2rg^tm1Wjl/SzJ (NSG) mice as recipients (10–12), which are susceptible to gold standard treatments for intestinal tumors such as radiation and 5-fluorouracil (5FU; ref. 19) and therefore not suitable for evaluation of such combination therapeutic approaches in vivo. Furthermore, given that the effects of various anticancer therapies are attributable at least in part to modulation of the immune system (20, 21), the use of immunocompetent recipients is necessary to evaluate the effects of CSC-targeting therapy in the physiologic context.

We have previously shown that the cyclin-dependent kinase inhibitor (CKI) p57 (CDKN1C) is specifically expressed in normal hematopoietic, neural, and intestinal stem cells, and that it is required for maintenance of the quiescence and stemness of these cells (22–24). Furthermore, p57 was recently shown to function as a molecular switch that controls the mitotic activation and quiescence of gastric reserve stem cells, regulating their stem cell activity during tissue homeostasis and regeneration (25). Here we show that p57 is also specifically expressed in a slowly cycling cell population with a high stem cell potential in malignant intestinal tumors. Single-cell transcriptome analysis revealed that Lgr5^high CSCs comprise actively and slowly cycling subpopulations, with the latter cells specifically expressing p57 and showing the highest stem cell potential among all intestinal tumor cell fractions. Tumorigenicity assays, lineage tracing, and ablation of p57⁺ CSCs collectively clarified that this cell fraction contributes minimally to steady-state tumor growth, but is resistant to chemotherapy and substantially drives posttherapeutic recurrence. Our results shed light on the heterogeneity of intestinal tumors and their CSC fractions, and they identify p57⁺ slowly cycling CSCs as a promising target for the development of an eradicative cancer therapy.

The study complies with all relevant ethics regulations. All mouse experiments were approved by the Animal Ethics Committee of Kyushu University (Fukuoka, Japan), and all experiments involving genetic manipulation were approved by the Safety Committee for Recombinant DNA Experiments of Kyushu University (Fukuoka, Japan).

C57BL/6, 129/Sv, NOD.Cg-Prkdc^scidIl2rg^tm1Wjl/SzJ (NSG), and CAG-LSL-tdTomato mice were obtained from The Jackson Laboratory. Apc^Δ716 mice (26) were provided by M. M. Taketo as described previously (24). Athymic nude (Balb/c-nu/nu) mice were obtained from CLEA Japan. Generation of p57-DTR-Venus and p57-CreER^T2 mice was described previously (24). Mice harboring genetic modifications of the p57 gene locus were maintained on the C57BL/6 × 129/Sv background, and the products of F₁ crosses between C57BL/6 and 129/Sv strains were therefore used as recipients for orthotopic transplantation of cancer organoids established from the p57 knock-in mouse lines. For testing of anticancer drug susceptibility, mice were subjected to daily intraperitoneal injection of 5FU (Sigma, catalog no. F6627) at 10 mg/kg. Animals that lost >20% of their body weight relative to before 5FU administration were considered to have reached a humane endpoint and were killed. As a standard chemotherapy for intestinal cancer, mice were injected intraperitoneally with 5FU at 150 mg/kg once or at 50 mg/kg for 3 consecutive days. For lineage tracing experiments, mice were injected intraperitoneally with tamoxifen (Sigma, catalog no. T5648) to activate Cre-induced tdTomato expression. The dose of tamoxifen was 50 mg/kg administered as a single injection in a spontaneous adenoma model or 20 mg/kg for each of five consecutive injections in the transplantation model. Ablation of p57⁺ or Lgr5⁺ cells with the use of p57-DTR-Venus or Lgr5-DTR-tdTomato alleles, respectively, was performed by intraperitoneal injection of diphtheria toxin (DT) at 5 μg/kg (Millipore, catalog no. 322326) on each of 5 consecutive days. Experiments with the cancer organoid–based transplantation model were performed with 7 to 18 weeks old male and female recipient mice. Lineage tracing experiments for spontaneous intestinal adenoma were performed with 12 to 16 weeks old male and female Apc^Δ716/p57-CreER^T2/CAG–loxP-Stop-loxP (LSL)–tdTomato mice.

The establishment of intestinal organoids was performed as described previously (24, 27). In brief, crypts collected from the small intestine (duodenum, jejunum, and ileum) were embedded in 20 μL of Matrigel (BD Biosciences, catalog no. 356231) and maintained in 250 μL of organoid culture medium consisting of Advanced DMEM/F12 (Thermo Fisher Scientific, catalog no. 12634028) supplemented with penicillin-streptomycin, HEPES (Invitrogen, catalog no. 15630106), Glutamax (Invitrogen, catalog no. 35050), B27 (Thermo Fisher Scientific, catalog no. 17504), N-2 (Thermo Fisher Scientific, catalog no. 17502), N-acetylcysteine (Sigma-Aldrich, catalog no. A7250), EGF at 50 ng/mL (Thermo Fisher Scientific, catalog no. PMG8043), R-spondin 1 (R&D, catalog no. 4645-RS) at 1 μg/mL, and Noggin (Pepro Tech, catalog no. 250-38) at 100 ng/mL.

Gene-specific single-guide RNAs (sgRNA) were designed to target coding regions of Lgr5 (5′-CACCGTCTCTAGTGACTATGAGAG-3′ and 5′-AAACCTCTCATAGTCACTAGAGAC-3′), Apc (5′-CACCGTCTGCCATCCCTTCACGTT-3′ and 5′-AAACAACGTGAAGGGATGGCAGAC-3′), Trp53 (5′-CACCGAGGAGCTCCTGACACTCGGA-3′ and 5′-AAACTCCGAGTGTCAGGAGCTCCTC-3′), Kras (5′-CACCGCTGAATTAGCTGTATCGTCA-3′ and 5′-AAACTGACGATACAGCTAATTCAGC-3′), and Smad4 (5′-CACCGATGTGTCATAGACAAGGT-3′ and 5′-AAACACCTTGTCTATGACACATC-3′) as described previously (10, 28, 29), and the corresponding DNA sequences were subcloned into the BbsI site of pX330 (Addgene). The knock-in construct for Lgr5-DTR-tdTomato was designed so that a T2A-DTR-T2A-tdTomato cassette was inserted in-frame into the endogenous Lgr5 coding sequence immediately 5′ to the only TGA codon at this locus. The 5′ and 3′ regions of homology in targeting vectors for Lgr5 and Kras were generated by genomic PCR with appropriate primers, and each final construct was cloned into the multiple cloning site of pBluescript SK II (+). For transfection, intestinal organoids were dissociated into a single-cell suspension for 5 minutes at 37°C with the use of TrypLE Express (Life Technologies, catalog no. 12605). The single cells were resuspended in 1 mL of organoid culture medium supplemented with 100 μL of Opti-MEM (Invitrogen, catalog no. 31985070), 2 μL of Lipofectamine 2000 reagent (Invitrogen, catalog no. 11668019), and 1 μg of appropriate vectors, and were then transfected by spinoculation at 600 × g for 60 minutes at 32°C in a 48-well plate. After transfection, the cells were incubated at 37°C for 5 hours, embedded in Matrigel, and cultured in organoid culture medium. The medium was supplemented with 1 μmol/L Y-27632 (Sigma, catalog no. Y0503) for the first 2 days after plating to prevent anoikis. Mutant organoids were selected by removal of growth factors or addition of compounds [Nutlin-3 (Cayman Chemical, catalog no. 10004372), gefitinib (Selleck Chemicals, catalog no. S1025), and TGFβ (Pepro Tech, catalog no. AF-100-21C)] as shown in Fig. 1B.

Venus and Akaluc (RIKEN BRC, catalog no. RDB15781; ref. 30) cDNAs were cloned into the MSCV vector (3), which contains a puromycin resistance gene, and the resulting constructs were introduced into Plat E packaging cells with the use of the FuGENE HD reagent (Promega, catalog no. E2312). Culture supernatants containing recombinant retroviruses were harvested and passed through a 0.45-μm filter, and the viruses were collected by centrifugation at 6,000 × g for 16 hours at 4°C, resuspended in organoid culture medium, and combined with single organoid cells in a 48-well plate. The plate was centrifuged at 600 × g for 60 minutes at 32°C, after which, the cells were incubated for 8 hours at 37°C, embedded in Matrigel, and cultured in organoid culture medium. Venus-expressing cells were isolated with the use of a FACS Aria II instrument (BD Biosciences), and cells expressing Akaluc and puromycin resistance genes were selected in medium containing puromycin (2 μg/mL).

Transplantation of intestinal cancer organoids into the colonic submucosa was performed as described previously (10, 31), with minor modifications. In brief, mice were anesthetized with a mixture of medetomidine, midazolam, and butorphanol and then placed in a supine position, and their posterior extremities were taped. A blunt-ended hemostat was inserted approximately 1 cm from the anus, closed across a single mucosal fold, and then pulled back out of the body. A 10-μL suspension of organoids (1 × 10⁵ cells) mixed with an equal volume of Matrigel was directly injected into the exposed colonic submucosa with a 29G syringe, and the hemostat was then released so that the exteriorized colon would be reinserted spontaneously.

p57-Venus^– or p57-Venus⁺ cells (2 × 10³) were injected subcutaneously into athymic nude mice. Tumor weight was determined 28 days after cell transplantation.

Tumor tissue and normal small intestine were fixed overnight at 4°C with 4% paraformaldehyde in PBS. For immunofluorescence analysis, the fixed and cryoprotected tissue was embedded in OCT compound (Tissue Tek, catalog no. 4583), frozen, and cut into sections at a thickness of 10 μm with a cryostat. Dried sections were rehydrated at 4°C in PBS and subjected to immunostaining. All sections were incubated overnight at room temperature with primary antibodies to p57 (Santa Cruz Biotechnology, catalog no. sc-1039), to GFP (Abcam, catalog no. ab6673), to Ki67 (Thermo Fisher Scientific, catalog no. MA5-14520), to cytokeratin 19 (Krt19; Abcam, catalog no. ab52625), to Prox1 (Abcam, catalog no. ab101851), or to p27 (Cell Signaling Technology, catalog no. 3686). The sections were washed four times with PBS before incubation for 3 hours at room temperature with species-appropriate Alexa Fluor 647–, Alexa Fluor 546–, or Alexa Fluor 488–conjugated secondary antibodies (Abcam, catalog nos. ab150075 and ab150129; Thermo Fisher Scientific, catalog nos. A10040 and A21206) for detection of unconjugated primary antibodies. The sections were again washed four times and then mounted with the use of Vectashield mounting medium containing DAPI (Vector Laboratories, catalog no. H-1200). Imaging and cell counting were performed with an LSM700 confocal microscopy system (Zeiss). Immunofluorescence intensity for p57 or Prox1 relative to nuclear area was quantified with the use of ImageJ.JS software (https://ij.imjoy.io).

For isolation of single tumor cells, tumors were mechanically dissociated with scissors and digested by incubation for 2 hours at 37°C with collagenase (Wako, catalog no. 034-22363) at 1.5 mg/mL and DNase I (Sigma, catalog no. DN25) at 100 μg/mL in DMEM. Cells were then isolated by vigorous pipetting of the tumor fragments and passage through a 70-μm cell strainer.

For FACS or flow cytometric analysis of live cells, isolated tumor cells were stained for 1 hour at 4°C first with allophycocyanin (APC)–conjugated antibodies to EpCAM (eBiosciences, catalog no. 17-5791-82) and biotin-conjugated antibodies to CD45 (eBiosciences, catalog no. 13-0451-82), both at a 1:500 dilution in FACS buffer (sterile PBS containing 4% FBS), and then with streptavidin conjugated with APC-Cy7 (BD Biosciences, catalog no. 554063). The cells were washed and then resuspended in FACS buffer containing propidium iodide (PI) at 2 μg/mL. Cell sorting and analysis were performed with a FACS Aria II instrument (BD Biosciences) fitted with a 100-μm nozzle. After gating for forward scatter and side scatter and exclusion of doublets, the PI^–EpCAM⁺CD45^– fraction was collected as single live tumor cells.

For assay of apoptosis, tumor organoid cells cultured in vitro were stained for 15 minutes at room temperature with APC-conjugated antibodies to Annexin V (BD Biosciences, catalog no.550475) and PI (5 μg/mL) with the use of 10× Annexin V Binding Buffer (BD Biosciences, catalog no. 51-66121E). Analysis was performed with a FACS Verse instrument (BD Biosciences).

For single-cell RNA sequencing (scRNA-seq) library preparation, single tumor cells were processed and barcoded with the use of the CEL-seq2 technique (24, 32) as performed with minor modifications. In brief, each single cell was directly sorted into a well of a 96-well PCR plate containing primer mix including reverse transcription (RT) primers with unique molecular identifiers (UMI) and cell barcodes, deoxynucleoside triphosphates, RNase inhibitors, and 0.3% Nonidet P-40. Sequences for RT primers are listed in Supplementary Table S1. Each library was labeled with the information for FACS gate (p57-Venus^– or p57-Venus⁺). The cells were lysed by incubation at 65°C for 5 minutes and then subjected to room temperature with SuperScript III Reverse Transcriptase (Invitrogen, catalog no. 18080044). Samples from each plate were pooled into a single library, from which cDNAs were purified with the use of Agencourt AMPure XP Beads (Beckman Coulter, catalog no. A63381) and subjected to second-strand synthesis. The resultant libraries were again bead-purified and then used as templates for in vitro transcription with a MEGAscript T7 Transcription Kit (Invitrogen, catalog no. AM1334). The amplified RNAs were subjected to RT followed by PCR amplification (20 cycles) with Illumina TruSeq Small RNA primers to generate the final cDNA libraries, which were subjected to paired-end sequencing with an Illumina NovaSeq6000 instrument. The sequenced libraries were subjected to bioinformatics analysis as described below.

The FASTQ files obtained from CEL-seq2 analysis were processed with celseq2 software as described previously (24, 32). The reads were thus “demultiplexed” and mapped to the mouse genome (Mm10, Genome Reference Consortium GRCm38) with the use of Bowtie2 software, and gene counting was then performed with HTSeq software. The resultant gene (UMI)-cell matrices for each experiment were concatenated with R software. For calculation of Venus mRNA abundance as shown in Supplementary Fig. S2J and S2K, scRNA-seq reads were mapped on wild-type (WT) p57 and p57-DTR-Venus genomic sequences as described above. The increment in the read number in the latter compared with the former mapping results for each cell was counted as Venus-overlapping reads.

Visualization of scRNA-seq data was performed mostly with the R package Seurat as described previously (24, 33). For analysis of tumor cells with Seurat v4, the gene-cell matrix was filtered on the basis of the number of genes detected per cell; cells with values of <900 or >6,300 or those with values of <200 or >9,000 were discarded for WT or p57-DTR-Venus tumors, respectively. Cell-cycle analysis was performed with the CellCycleScoring() function. The ScaleData() function was run for Z-scoring, with the cell-cycle status and batch effects among libraries being “regressed out” by specifying the vars.to.regress argument. Variable genes were identified with the FindVariableFeatures() function and were subjected to principal component analysis. The first 20 principal components were used for t-distributed stochastic neighbor embedding (t-SNE) and clustering analysis. The FindCluster() function was run at a resolution of 0.5. Genes expressed in every cluster or in specific clusters were identified with the FindAllMarkers() and FindMarkers() functions, respectively. Gene set enrichment analysis (GSEA) for scRNA-seq data was performed with iDEA software (34) with the standard parameters. For analysis of the mean expression level of given gene signatures, the normalized and log-scaled expression values for each gene were min-max scaled so that every gene had an equal dynamic range (Max = 1.0, Min = 0), and the mean expression value of the gene signature for each single cell was then calculated. Given that the gene expression value obtained by Seurat correlates with the absolute quantity of the corresponding mRNA in each single cell, and to reflect this information in heatmaps, the min-max–scaled value of signature expression for each cell was finally multiplied by the average maximal expression value of individual genes, so that the resultant dynamic range corresponds to the average absolute amount of the given signature transcripts. Visualization of heatmaps on the t-SNE projection was performed with the use of the ggplot2 package in R software.

For unsupervised exploration of stem cell populations corresponding to scRNA-seq data, the CytoTRACE algorithm (35) was run according to its reference manual with no modification of parameters. The resultant score for each cell was extracted and plotted on the t-SNE projection or jitter plot graphs with the use of R software.

Cells were directly sorted into a tube containing 800 μL of Isogen (Nippon Gene, catalog no. 319-90211) for isolation of total RNA, and with the use of Gene-Packman Coprecipitant (Nacalai Tesque, catalog no. 12680-30) to facilitate RNA precipitation. RT was performed with the use of ReverTra Ace qPCR RT Master Mix (Toyobo, catalog no. FSQ-301), and the resulting cDNA was subjected to qPCR analysis with Luna Universal qPCR Master Mix (New England BioLabs, catalog no. M3003) and specific primers in a Step One Plus Real-Time PCR System (Applied Biosystems). The abundance of target mRNAs was normalized by that of Hprt1 mRNA. The sequences of the primers were designed for mRNA sequences of Hprt (5′- GCCTAAGATGAGCGCAAGTTG-3′ and 5′- TACTAGGCAGATGGCCACAGG-3′) and p57 (5′- AGCGGACGATGGAAGAACTCT-3′ and 5′- TTGGGACCAGCGTACTCCTT-3′).

Tumor clearing was performed according to the SeeDB protocol (24, 36). In brief, fixed tumors were sequentially immersed in PBS containing 20%, 40%, 60%, and 80% fructose (Sigma, catalog no. F0127) for 4 to 8 hours for each step and with gentle rotation at room temperature. The tumors were then incubated in 100% fructose in water for 24 hours at room temperature and embedded in the same solution. Screening, counting, and imaging of lineage tracing events were performed with a DeltaVision Elite microscopy system (GE Healthcare Life Sciences) and a BZ-X700 microscope (Keyence).

Tumor organoids established from p57-CreER^T2/CAG-LSL-tdTomato mice were exposed to 1 μmol/L 4-hydroxytamoxifen (Sigma, catalog no. H7904) for 2 days and then incubated with or without 5FU (50 μg/mL) for another 2 days. The medium was then replaced with fresh control medium. Imaging of lineage tracing events was performed with a BZ-X700 microscope (Keyence). At 4 and 7 days after the onset of exposure to 4-hydroxytamoxifen, tdTomato⁺ cells were analyzed with a FACS Verse instrument (BD Biosciences).

Mice were anesthetized with a mixture of medetomidine, midazolam, and butorphanol, and injected intraperitoneally with Akalumine-HCl (Fujifilm, catalog no. 012-26701) at a dose of 9.4 μmol/kg and at 20 minutes before image acquisition. Bioluminescence signals were measured with the use of an IVIS imaging system (Perkin Elmer).

Quantitative data were analyzed with the two-tailed Student t test, Welch t test, or the Steel–Dwass test. The analysis was performed with the use of Excel or JMP software. A P value of <0.05 was considered statistically significant.

The scRNA-seq data generated in this study have been deposited in the DDBJ Sequence Read Archive (DRA) under accession codes DRA014608 and DRA014609, and in the Genomic Expression Archive (GEA) under accession codes E-GEAD-522 and E-GEAD-523.

For the purpose of identification and ablation of therapy-resistant cancer cell populations, we set out to establish a transplantation model for mouse intestinal tumor organoids that recapitulates the clinical response to anticancer drugs. The syngeneic transplantation of intestinal tumor organoids in mice has mostly been performed with the use of immunodeficient recipients such as NSG mice (10–12). However, we found that daily administration of 5FU (10 mg/kg), a fundamental drug for the treatment of intestinal tumors (37, 38), resulted in marked weight loss and moribundity that necessitated humane killing within 11 days after the onset of treatment in NSG mice, whereas such an effect was not apparent in immunocompetent WT mice (Fig. 1A). This moribundity was likely attributable to the high sensitivity of immunodeficient mice to various anticancer drugs including 5FU (19), and it precluded our evaluation of chemotherapy in this strain. To circumvent this issue, we sought to establish a syngeneic transplantation model with the use of immunocompetent recipients. We generated normal intestinal organoids from the small intestine (duodenum, jejunum, and ileum) of WT mice and sequentially introduced four oncogenic mutations [Apc knockout (KO), Trp53 KO, Kras^G12D, and Smad4 KO] into these organoids with the use of the clustered regularly interspaced short palindromic repeats (CRISPR)/Cas9 system, as described previously (10, 17). By removing niche factors or adding selection compounds suitable for individual mutations, we were able to select and establish intestinal tumor organoids with different grades of malignancy in a stepwise manner (Fig. 1B). Orthotopic transplantation of the quadruple-mutant organoids into the colonic submucosa of recipient mice (10, 31) resulted in the development of primary tumors in all immunodeficient and immunocompetent recipients (Fig. 1C–E), indicative of the feasibility of transplantation and evaluation of chemotherapy in immunocompetent mice. On the other hand, triple-mutant organoids (Apc KO/Trp53 KO/Kras^G12D) showed minimal engraftment to the colon of immunocompetent mice (Fig. 1C), suggesting that the use of quadruple-mutant organoids is necessary. The tumors formed by quadruple-mutant organoids metastasized to the liver and lung within 2 months in some NSG mice, as described previously (10), whereas such metastasis was not apparent in immunocompetent mice (Fig. 1F–J). This latter finding is consistent with a previous study showing that metastasis is markedly suppressed by immunosurveillance in immunocompetent mice (39), suggesting that our model system recapitulates the microenvironment of spontaneous malignant tumors in vivo. Collectively, these results suggested that the orthotopic transplantation model in immunocompetent mice was suitable for evaluation of chemotherapy against malignant intestinal cancer in the physiologic context, and we indeed adopted this model for subsequent experiments.

Many types of cancer consist of heterogeneous cell populations (1, 2, 40). Although the tumors formed by orthotopically transplanted mutant organoids are generally thought to recapitulate the physiologic properties of spontaneous malignant tumors in vivo (10, 39, 41), detailed profiles of their cellular constituents have remained largely unclear. To investigate the heterogeneity of cancer cells in our model, we first performed scRNA-seq analysis. We orthotopically transplanted quadruple-mutant organoids expressing retrovirally transduced Venus protein into WT mice (Supplementary Fig. S1A). At 2 weeks after transplantation, Venus⁺ cancer cells were sorted from the tumors that developed in the colon of recipient mice (Supplementary Fig. S1B) and were subjected to scRNA-seq analysis with the use of the CEL-seq2 platform (32). Clustering analysis with Seurat software (33) classified the cancer cells into seven unsupervised clusters (Fig. 2A). Clusters 5 and 6 consisted of small numbers of cells that expressed markers either for differentiated epithelial cells including Paneth and goblet cells or for immune cells (Supplementary Fig. S1C–S1G), respectively, suggesting that these clusters represented contamination by these cell types. We therefore considered clusters 0 to 4 as cancer cell populations for further analysis.

Previous studies have shown that Lgr5 is expressed at a high level not only in normal CBCs but also in CSCs of intestinal tumors (14, 15). However, our scRNA-seq data revealed Lgr5 expression in all identified clusters (Fig. 2B and C), suggesting that this gene is expressed widely in malignant intestinal cancer, in contrast to its CBC-restricted expression in the normal epithelium (13). The Lgr5-expressing cells were nevertheless markedly enriched in clusters 4 and 2 (Fig. 2C), suggesting that these clusters correspond to Lgr5^high populations. Analysis by CytoTRACE, an algorithm for unsupervised inference of tissue stem cells on the basis of transcriptomic diversity (35), indeed showed that clusters 4 and 2 had substantially higher stemness scores compared with the other clusters (Fig. 2D and E). These results suggested that clusters 4 and 2 represent CSC populations in intestinal tumors.

We next performed Seurat-based cell-cycle scoring analysis and found that cluster 4 consisted mostly of cells in G₀–G₁ phase, whereas cluster 2 comprised those in S or G₂–M phases (Fig. 2F). Consistent with these findings, expression of the cell proliferation marker Mki67 (Ki67) was almost absent in cluster 4 (Fig. 2G). Collectively, our results suggested that Lgr5^high intestinal CSCs can be classified into slowly cycling (cluster 4) and actively cycling (cluster 2) fractions, as is the case for Lgr5^high CBCs of the normal intestinal epithelium (18). Given that anticancer drugs such as 5FU generally target actively cycling cancer cells, quiescence of cancer or its stem cells has been thought to be a major factor underlying therapeutic resistance (2, 38). We have previously shown that quiescent cells expressing p57 serve as injury-induced intestinal stem cells that are distinct from Lgr5^high CBCs in the normal intestinal epithelium (24). In contrast to the situation in the normal epithelium, our current scRNA-seq data for malignant intestinal tumors revealed that Cdkn1c (p57) is specifically expressed in cluster 4, a slowly cycling subpopulation of Lgr5^high CSCs (Fig. 2H and I; Supplementary Fig. S1H), whereas the genes for the CKIs p21 and p27 are also expressed in the other clusters (Supplementary Fig. S1I–S1L). These results are in analogy with the differential expression patterns of p57 and p27 in crypts and villi of the normal epithelium, respectively (Supplementary Fig. S1M and S1N).

To investigate the relation between p57⁺ cells and Lgr5⁺ cells in intestinal cancer, we generated quadruple-mutant organoids that harbor an in-frame knock-in of coding sequences for the diphtheria toxin receptor (DTR) and tdTomato (each preceded by cDNA for the self-cleaving T2A peptide) linked to the 3′ portion of the endogenous Lgr5 coding sequence (Supplementary Fig. S1O). Immunofluorescence analysis of tumors formed by the organoids revealed that 20% to 30% of total Lgr5⁺ cells expressed p57 (Supplementary Fig. S1P) and that almost all p57⁺ cells were positive for Lgr5-tdTomato (Fig. 2J and K), supporting the notion that p57⁺ cells constitute a subpopulation of Lgr5^high CSCs. Overall, these results revealed heterogeneity of intestinal CSCs and suggested that p57 specifically marks slowly cycling CSCs, which are a promising target for eradication of therapy-resistant cancer cells.

To gain further insight into the gene expression profile of the slowly cycling CSC subpopulation, we analyzed differentially expressed genes (DEG) between cluster 4 and the other clusters (Fig. 3A). The upregulation of p57 expression in cluster 4 was also apparent in this unbiased analysis, suggesting that it is a specific marker for this cell subpopulation. The DEGs upregulated in cluster 4 also included Notum and Nkd1 (Fig. 3A–C), which encode representative negative regulators of Wnt signaling that defines stemness in the normal intestinal epithelium (42). Given that these molecules have been shown to be upregulated as a result of negative feedback of Wnt pathway activation (43–45) and that they block signaling upstream of Apc, which was inactivated in our quadruple-mutant organoids, this finding likely reflects hyperactivation, rather than inactivation, of Wnt signaling in cluster 4. Notum was recently shown to be secreted by intestinal cancer cells and to confer a competitive advantage on them through suppression of Wnt signaling in surrounding normal stem cells (43, 44), with our results now suggesting that cluster 4 might be a major source of such secreted Notum. Expression of the normal CBC marker Smoc2 was also upregulated in cluster 4 (Fig. 3A), whereas that of the differentiated cell markers Krt19 and Ceacam1 was downregulated (Fig. 3A, D, and E), consistent with the stem cell–like transcriptomic characteristics of this cluster (Fig. 2).

To characterize the slowly cycling CSCs at the molecular pathway or gene ontology (GO) level, we next performed GSEA for cluster 4 relative to the other clusters with the use of iDEA software (34). This analysis revealed that the gene set for “positive regulation of canonical Wnt signaling pathway” was significantly upregulated in cluster 4 (Fig. 3F and G), consistent with the predicted activation of Wnt signaling described above. Among the most significantly upregulated gene sets were several GO terms for transcriptional regulation (Fig. 3F and H), consistent with the increased level of transcriptional diversity as scored by the CytoTRACE algorithm (Fig. 2D and E). The term “autophagy” was also significantly enriched in cluster 4 (Fig. 3F and I), consistent with a previous study showing that activation of autophagy is a major hallmark of CSCs and is required for maintenance of their stemness in many types of cancer (46). These results likely reflect the transcriptomic features of cluster 4 as a CSC population. On the other hand, genes for the “translation” term were expressed at a lower level in cluster 4 (Fig. 3F and J), which is reminiscent of the suppression of protein synthesis and degradation in hematopoietic stem cells (47) and likely results from the slowly cycling nature of this cluster. Collectively, our findings suggested that the cells of cluster 4 are characterized by activation of Wnt signaling in spite of an inactive cell-cycle state, as well as by several representative transcriptomic features specific to CSCs.

Posttherapeutic recurrence of disease is generally thought to be driven by therapy-resistant CSCs in many cancer types (1–7), with Lgr5^high CSCs being a potential therapeutic target for intestinal tumors. However, given the pronounced heterogeneity of Lgr5^high CSCs as apparent in our scRNA-seq analysis (Fig. 2), it remains unclear whether Lgr5 expression defines a distinct CSC entity as it does CBCs in normal intestinal crypts (13). In contrast, p57 expression appeared largely specific to the slowly cycling Lgr5^high CSC subpopulation in malignant intestinal tumors (Fig. 2H and I), suggesting that p57 might be a marker for identification of therapy-resistant CSCs. For the purpose of visualization and ablation of p57⁺ cells in malignant intestinal cancer, we established quadruple-mutant intestinal tumor organoids from the small intestine of p57-DTR-Venus mice (24), which harbor an in-frame knock-in of DTR and Venus cDNAs (each preceded by cDNA for the self-cleaving T2A peptide) linked to the 3′ portion of the endogenous p57 coding sequence (Fig. 4A). Given that p57-DTR-Venus mice were maintained on the C57BL/6 × 129/Sv background (24), and that transplantation of cancer organoids established from these mice into C57BL/6 recipients frequently resulted in allogeneic rejection (Supplementary Fig. S2A), we used the products of F₁ crosses between the C57BL/6 and 129/Sv strains as recipients for orthotopic transplantation of p57-DTR-Venus cancer organoids. To examine whether this reporter specifically marks p57⁺ cancer cells, we isolated Venus⁺ and Venus^‒ cells from p57-DTR-Venus tumors that developed in vivo at 2 months after transplantation (Fig. 4B). qRT-PCR analysis showed that endogenous p57 mRNA was more abundant in Venus⁺ cells (upper 0.7% to 1% fraction) than in Venus^‒ cells (Fig. 4C), suggesting that the p57-DTR-Venus reporter recapitulates endogenous p57 expression and allows specific labeling of p57⁺ cells.

To examine whether this p57 reporter specifically marks the slowly cycling CSC subpopulation apparent in the quadruple-mutant organoids derived from WT mice (Figs. 2 and 3), we performed scRNA-seq analysis of tumors that developed in vivo from corresponding p57-DTR-Venus organoids. We sorted almost equal numbers of p57-Venus^‒ and p57-Venus⁺ cells from EpCAM⁺CD45^‒ tumor cells and subjected them to CEL-seq2 analysis. Seurat-based clustering analysis classified the cancer cells into eight unsupervised clusters (Fig. 4D). Clusters 3/7, 5, and 4/6 consisted of small numbers of cells expressing immune cell markers, fibroblast/endothelial cell markers, or differentiated epithelial cell (Paneth, goblet, and tuft cell) markers, respectively (Supplementary Fig. S2B–S2H), leading us to consider clusters 0 to 2 as cancer cell populations for subsequent analysis. We found that cluster 2 consisted mostly of p57-Venus⁺ cells (Fig. 4D and F), suggesting that cluster 2 represents the p57⁺ cancer cell population in our scRNA-seq dataset. Although p57-Venus⁺ cells were detected also in clusters 0 and 1, these cells likely represented p57-Venus^‒ cells that contaminated the Venus⁺ FACS gate as a result of high autofluorescence, given that endogenous p57 mRNA was detected predominantly in cluster 2 (Fig. 4E; Supplementary Fig. S2I). Consistent with this notion, Venus mRNA–positive cells were most enriched in cluster 2 (Supplementary Fig. S2J and S2K). Cluster 2 consisted mostly of cells in G₀–G₁ phase, as shown by Seurat-based cell-cycle scoring analysis (Supplementary Fig. S2L), and it manifested the highest CytoTRACE score among all cell clusters (Fig. 4G and H). Furthermore, the marker genes identified for cluster 4 of tumors derived from WT mouse organoids (Fig. 3A–E) were also correspondingly upregulated (Supplementary Fig. S2M and S2N) or downregulated (Supplementary Fig. S2O and S2P) in cluster 2 of the p57-DTR-Venus tumor dataset. Collectively, both WT and p57-DTR-Venus scRNA-seq datasets suggested that p57 marks a slowly cycling cell population that is endowed with CSC-like transcriptomic features in malignant intestinal cancer.

Consistent with our scRNA-seq data, immunofluorescence analysis showed that p57-Venus⁺ cells were indeed all negative for the proliferation marker Ki67 (Fig. 4I and J), indicative of a quiescent state. Together, these results suggested that p57 is a robust marker for the slowly cycling CSC subpopulation in malignant intestinal tumors, and that the p57-DTR-Venus reporter should allow further detailed analysis of this cell type.

CSCs are capable of reconstituting tumor tissue from a single cell (48, 49), as well as of forming organoids with high efficiency (11, 12). To evaluate the organoid formation capacity of p57⁺ cancer cells, we isolated p57-Venus⁺ cells by FACS from intestinal tumors formed by p57-DTR-Venus quadruple-mutant organoids in vivo, and we then cultured these cells in vitro (Fig. 5A). Although single p57-Venus⁺ cells showed a substantial ability to reconstitute organoids, the efficiency was similar to that of p57-Venus⁻ cells in the steady state (Fig. 5B and C). This finding indicated that the contribution of p57⁺ cells to homeostatic tumor growth is not greater than that of other cancer cells, in spite of the increased stem cell potential of the former cells apparent in our scRNA-seq analysis (Figs. 2 and 4). We postulated that this result was attributable to the quiescence of p57-Venus⁺ cells in the steady state (Fig. 4I and J), and we therefore next examined their organoid formation capacity after treatment with an anticancer drug. Administration of a single dose of 5FU tended to be most therapeutically effective at a dose of 150 mg/kg in our orthotopic transplantation model, leading us to adopt this dose for subsequent experiments (Supplementary Fig. S3A and S3B). Treatment with 5FU at 10 or 50 mg/kg actually increased tumor growth compared with the control at 7 days after drug administration (Supplementary Fig. S3A), consistent with previous studies showing that low doses of chemotherapeutic agents can promote tumor growth through the production of debris from dead cells that stimulates the proliferation of live cells in colorectal cancer (50, 51). At 1 day after injection of 5FU in recipient mice bearing p57-DTR-Venus tumors, the organoid formation efficiency of p57-Venus⁻ cells was equal to or slightly lower than that in the steady state (Fig. 5B–E). In contrast, the organoid formation efficiency of p57-Venus⁺ cells was increased after 5FU treatment (Fig. 5B–E), and it was significantly higher than that of corresponding p57-Venus⁻ cells (Fig. 5D and E). A tumorigenicity assay in vivo also showed a higher reconstitution capacity for p57-Venus⁺ cells than for p57-Venus^– cells after 5FU treatment (Fig. 5A, F, and G). These results suggested that cancer growth in the steady state largely depends on actively cycling CSCs (10), whereas tumor repopulation after chemotherapy is driven by p57⁺ slowly cycling CSCs. Immunofluorescence analysis revealed that the expression of Prox1, a Tcf1-regulated transcriptional activator of p57 in colorectal cancer (52), was significantly decreased in p57⁺ cells of organoid-derived tumors at day 1 (Fig. 5H–K) and that the expression of p57 was downregulated at day 3 (Fig. 5L–N) after 5FU treatment. Together with previous studies showing that p57 plays a key role in maintenance of the nonproliferative state in multiple tissue stem cells (22–25), these results suggested that downregulation of the Tcf1-Prox1-p57 axis after 5FU treatment might underlie the posttherapeutic activation of slowly cycling CSCs.

Given that quiescent cells are resistant to conventional chemotherapy and radiotherapy, which selectively target actively cycling cancer cells (2, 38), we next examined the chemoresistance of p57⁺ cancer cells. Treatment with 5FU induced apoptotic cell death and inhibited cell proliferation in quadruple-mutant organoids maintained in vitro (Supplementary Fig. S3C–S3F). We found that the abundance of p57 mRNA was increased in the live cell fraction that survived 5FU treatment compared with that apparent in the steady state (Supplementary Fig. S3G), suggesting that p57⁺ cells are resistant to chemotherapy compared with p57⁻ cells. Together, these results suggested that, although p57⁺ cancer cells are quiescent and thus contribute in only a limited manner to steady-state tumor maintenance, these cells serve as therapy-resistant CSCs that drive regrowth of intestinal cancer.

To examine the CSC activity of p57⁺ cells in vivo by lineage tracing analysis, we established quadruple-mutant intestinal tumor organoids from p57-CreER^T2/CAG-LSL-tdTomato mice (24) and transplanted them orthotopically into recipient mice (Fig. 6A). At 2 days after a single tamoxifen injection (20 mg/kg) in the recipient mice, tdTomato⁺ cells were found to constitute 0.0026% of EpCAM⁺CD45⁻ tumor cells (Supplementary Fig. S4A and S4C). Given that the frequency of p57-Venus⁺ cells in tumors derived from p57-DTR-Venus organoids was 0.7% to 1% (Fig. 4B), we estimated that only 0.2% to 0.4% of total p57⁺ cells in each tumor was labeled by this dosing regimen. Tamoxifen injection on five consecutive days greatly increased the labeling efficiency (to 0.039% tdTomato⁺ cells among EpCAM⁺CD45⁻ tumor cells; Supplementary Fig. S4B and S4C), leading us to adopt this injection schedule for subsequent experiments (Fig. 6B).

At 1 day after the last tamoxifen injection, we detected the emergence of solitary tdTomato⁺ cells that were negative for Ki67 staining (Fig. 6C and D), suggesting that the initially labeled p57⁺ cells were maintained in a quiescent state. However, lineage tracing events derived from p57⁺ cells at 4 weeks after the first tamoxifen injection contained numerous Ki67⁺ (actively cycling) cancer cells as well as Krt19^high differentiated cells (Fig. 6E and F), suggesting that p57⁺ cells serve as slowly cycling stem cells that give rise to heterogeneous cancer cell populations in vivo. To examine the chemoresistance and posttherapeutic activity of p57⁺ cancer cells, we next performed lineage tracing analysis in combination with 5FU treatment. We found that the frequency of tracing events was significantly increased after the administration of 5FU either as a single high dose (150 mg/kg; Fig. 6G–I) or as three consecutive low doses (50 mg/kg; Supplementary Fig. S4D and S4E) compared with that apparent in the steady state. We obtained similar results for lineage tracing with organoids maintained in vitro (Supplementary Fig. S4F–S4H). These findings suggested that p57⁺ cancer cells present before 5FU treatment are chemoresistant and contribute to subsequent cancer regrowth in vivo, consistent with the results for organoid formation assays with p57-DTR-Venus tumor–derived cells in vitro (Fig. 5).

To investigate the CSC activity and chemoresistance of p57⁺ cells in spontaneous intestinal tumors, we next performed similar lineage tracing experiments for intestinal adenomas of Apc^Δ716/p57-CreER^T2/CAG-LSL-tdTomato mice (Supplementary Fig. S4I; ref. 26). Lineage tracing events in tumors were apparent in the steady state, with their frequency and width being significantly increased after 5FU treatment (Supplementary Fig. S4J–S4M), similar to our findings in the transplantation model. Collectively, these results suggested that p57⁺ cells serve as slowly cycling CSCs in intestinal cancer in vivo, and that they contribute substantially to posttherapeutic tumor recurrence.

Finally, we examined the therapeutic effect of targeting p57⁺ CSCs in intestinal tumors. We infected p57-DTR-Venus cancer organoids with a retrovirus encoding Akaluc luciferase (30) to facilitate live monitoring of tumor volume in vivo, and then transplanted the organoids orthotopically into recipient mice (Fig. 7A and B). Injection of DT (5 μg/kg) for 5 consecutive days resulted in the almost complete loss of the p57-Venus⁺ fraction among EpCAM⁺CD45⁻ tumor cells (Supplementary Fig. S5A), suggesting that p57-Venus⁺ CSCs were ablated specifically and effectively by DT treatment. Although p57⁺ CSC ablation alone slightly inhibited tumor growth compared with untreated controls, the therapeutic effect was not statistically significant (Fig. 7C–E), a result likely attributable to the nondominant contribution of these cells to homeostatic tumor maintenance (Figs. 5 and 6). We obtained similar results for the ablation of Lgr5⁺ CSCs (Supplementary Fig. S5B–S5E), consistent with previous results showing the limited therapeutic effect of targeting only CSCs in an orthotopic transplantation model (10). Given that differentiated cancer cells were recently shown to be able to revert to the CSC state as a result of cellular plasticity (2, 10, 11, 16), these results suggested that the CSC-targeting therapy should be combined with anticancer drugs that target non-CSC populations. In addition, despite the fact that 5FU has been a key drug for the treatment of colorectal cancer (37, 38), administration of a single dose (150 mg/kg) of 5FU alone did not show a significant therapeutic effect in our p57-DTR-Venus organoid transplantation model (Fig. 7C–E), suggestive of the presence of chemoresistant cancer cell populations. In contrast, we found that the combination of p57⁺ cell ablation and 5FU significantly and markedly inhibited intestinal tumor growth in vivo (Fig. 7C–E). Together with our data showing that p57⁺ CSCs are chemoresistant and contribute more actively to cancer growth after treatment (Figs. 5 and 6), these results indicated that p57⁺ CSCs are required for the posttherapeutic recurrence of intestinal cancer. We also examined the potential therapeutic effect of p57⁺ CSC ablation in combination with multiple administrations of 5FU (Supplementary Fig. S5F). Given that treatment with 5FU at 100 mg/kg for 3 consecutive days resulted in pronounced weight loss and moribundity in recipient mice (Supplementary Fig. S3B), we adopted a dose of 50 mg/kg per day as a tolerable dose. Unexpectedly, we found that such low-dose 5FU treatment did not suppress but rather tended to promote tumor growth compared with untreated controls (Supplementary Fig. S5G). This finding was likely attributable to the production of debris that stimulates cancer growth (50, 51) or to the posttherapeutic activation of p57⁺ CSCs (Supplementary Fig. S4D–S4H) by such treatment, in addition to its incomplete elimination of actively cycling cancer cells. Consistent with the latter notion, prior removal of p57⁺ cells tended to suppress the enhanced tumor growth apparent after 5FU treatment to a similar extent as it inhibited tumor growth in untreated controls (Supplementary Fig. S5G), suggesting that ablation of p57⁺ CSCs may be of clinical benefit with regard to preventing such tumor growth–promoting effects of standard chemotherapies. Overall, our study thus suggests that p57⁺ slowly cycling CSCs are a promising target to overcome therapeutic resistance in malignant intestinal cancer, and it highlights the clinical relevance of CSC-targeted therapy combined with conventional anticancer drugs to the development of an eradicative cancer treatment strategy.

We have here established a syngeneic orthotopic transplantation model of malignant intestinal cancer and examined the resistance of cancer cells to an anticancer drug in a physiologic immunocompetent setting. We identified p57 as a specific marker for quiescent intestinal CSCs and subsequently showed that these p57⁺ CSCs are responsible for therapeutic resistance and posttherapeutic tumor relapse. Although the efficacy of CSC-targeting therapy for orthotopic primary intestinal tumors has remained largely unknown, our current results highlight the clinical relevance of this strategy for establishment of an eradicative anticancer treatment.

Despite extensive studies of Lgr5 as a CSC marker in intestinal tumors (10–12), intratumoral heterogeneity of Lgr5^high CSCs has been little understood. Our scRNA-seq analysis revealed that, although Lgr5 was widely expressed in all cell clusters of malignant intestinal tumors, those expressing Lgr5 at the highest level corresponded to CSC fractions as identified in an unbiased manner by CytoTRACE analysis. Cell-cycle scoring and gene expression analysis of these clusters showed that Lgr5^high CSCs can be divided into actively and slowly cycling subpopulations, with the latter specifically expressing p57. The coexistence of proliferative and quiescent stem cells in intestinal cancer is similar to the case of Lgr5^high CBCs in normal intestinal crypts (18), suggesting that such heterogeneity likely underlies resistance to conventional anticancer therapies that preferentially target cycling cancer cells. Of note, although a recent study suggested that p27 preferentially marks Lgr5⁺Ki67⁻ CSCs in human colorectal cancer (53), our study of mouse intestinal tumors suggests that p57 expression is more specific to slowly cycling Lgr5^high CSCs (Fig. 2H and I; Supplementary Fig. S1H–S1L). Further studies are warranted to examine whether the therapeutic targeting of p57⁺ CSCs is effective for the treatment of human colorectal cancer.

The expression pattern of p57 in malignant intestinal cancer contrasts markedly with that in the normal intestinal epithelium, in which p57 expression specifically marks enteroendocrine cell progenitors that serve as injury-induced stem cells and is absent in Lgr5^high CBCs (24). In addition, our current scRNA-seq analysis of malignant intestinal cancer suggested that Wnt signaling is hyperactivated in the quiescent Lgr5^high CSC subpopulation expressing p57, in spite of the mitotically inactive state of these cells. Although the mechanisms underlying such discrepancies or conflicting cellular states remain unclear, our findings suggest that the regulation of p57 expression, in particular by Wnt signaling, might differ between the normal intestinal epithelium and malignant intestinal cancer. A recent study proposed that the Tcf1-Prox1 axis positively regulates p57 expression in Lgr5^high CSCs (52), indicating a possible molecular mechanism for the upregulation of p57 by Wnt signaling in intestinal cancer. Elucidation of the regulatory mechanism of p57 expression in CSCs will be integral to the future development of CSC-targeting anticancer therapy.

CSC-targeted therapy has been considered to be a promising strategy for the curative treatment of cancer. For intestinal tumors, however, such therapy has been shown to be effective only in subcutaneous (10) or renal subcapsular (11) transplantation and liver metastasis (10) models, not for primary tumors in orthotopic transplantation models (10, 16). Consistent with these previous findings, we found that ablation of p57⁺ CSCs attenuated tumor growth only slightly, as was also the case for elimination of Lgr5^high CSCs (Supplementary Fig. S5B–S5E; ref. 10). Given the robust plasticity induced by factors derived from the cancer cell niche (2, 16), these results suggest that CSC ablation alone has a limited therapeutic effect on primary tumors, and that such an approach should be combined with anticancer therapies that target highly proliferative or more differentiated cancer cells. Our data now support this concept and suggest that p57⁺ quiescent CSCs resistant to conventional anticancer drugs are a promising therapeutic target.

p57 is expressed specifically not only in quiescent CSCs but also in quiescent normal tissue stem cells (22–25). This expression pattern and our current data support the CSC hypothesis that cancer hijacks the maintenance system of normal tissues (2). Future studies are warranted to solve the key related question of how to selectively target CSCs without affecting normal tissue stem cells.

No disclosures were reported.

T. Oka: Conceptualization, resources, data curation, software, formal analysis, funding acquisition, investigation, visualization, methodology, writing–original draft. T. Higa: Conceptualization, resources, data curation, software, formal analysis, supervision, funding acquisition, investigation, visualization, methodology, writing–original draft. O. Sugahara: Resources, investigation. D. Koga: Resources, investigation. S. Nakayama: Data curation, software. K.I. Nakayama: Conceptualization, supervision, funding acquisition, writing–original draft, writing–review and editing.

This work was supported in part by KAKENHI grants from Japan Society for the Promotion of Science (JSPS) and the Ministry of Education, Culture, Sports, Science, and Technology of Japan to K.I. Nakayama (JP18H05215) and to T. Higa (JP19K16716) as well as by a grant from the Project for Cancer Research and Therapeutic Evolution (P-CREATE) of the Japan Agency for Medical Research and Development (AMED) to K.I. Nakayama (JP21cm0106105).

The authors thank T. Sato (Keio University, Japan) and N. Sasaki (Gunma University, Japan) for discussion on culture and genetic manipulation of intestinal cancer organoids, M. M. Taketo (Kyoto University, Japan) for providing Apc^Δ716 mice, as well as H. Takayoshi and A. Niihara for technical assistance. Computations were performed in part on the NIG (National Institute of Genetics) supercomputer at Research Organization of Information and Systems (ROIS), Japan.

The publication costs of this article were defrayed in part by the payment of publication fees. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 USC section 1734.