Recurrent COPA mutation drives R-spondin-independent Wnt activation in intestinal tumors

Abstract

The majority of intestinal tumors harbor mutations in canonical Wnt pathway genes such as APC, whereas the lack of such alterations in a subset of tumors implies alternative tumorigenic routes. Here we identify recurrent in-frame deletion in COPA, frequently co-occurring with USP9X-truncating mutation, in small intestinal adenoma and adenocarcinoma. Patient-derived and CRISPR-engineered small intestinal organoids carrying COPA in-frame deletions exhibit R-spondin-independent yet Wnt ligand-dependent growth, maintaining LGR5 expression without canonical Wnt drivers. Mechanistically, COPA mutation stabilizes the Frizzled coreceptor LRP6 irrespective of R-spondin, sustaining Wnt pathway activation under growth factor-restricted conditions. USP9X loss further potentiates this phenotype. Unlike canonical Wnt pathway members, COPA encodes the α-subunit of coatomer complex I, which engages in vesicle trafficking with little prior linkage to intestinal tumorigenesis. Our findings establish COPA mutation as a unique and atypical intestinal tumor driver and implicate USP9X loss as a cooperating lesion.

This study implicates recurrent in-frame deletions in COPA exon 13 in small intestinal tumorigenesis through activation of Wnt signaling in an R-spondin-independent but Wnt ligand-dependent manner.

Main

Although the small intestine is the longest segment in the gastrointestinal tract, measuring 3–4 m in length in adult humans¹, small intestinal adenocarcinoma is a rare cancer and represents only 3% of all gastrointestinal cancers². Patients with small intestinal adenocarcinoma often present with late-stage disease and have a poor prognosis, partly due to the anatomically poor accessibility of the small intestine and delay in diagnosis³. The rarity of the disease has also rendered in-depth investigation of small intestinal adenocarcinoma challenging. Few human-relevant models of small intestinal tumors are available, and research on small intestinal tumors has largely relied on mouse models that develop tumors via the introduction of genetic drivers that mostly reflect colorectal cancer genomics⁴.

Tumorigenesis in the human small intestine follows an adenoma–carcinoma sequence, reminiscent of the well-established process in the colon². Both colon adenomas and adenocarcinomas have a high prevalence of APC-truncating mutations, providing a clear genetic basis for the colonic adenoma-to-carcinoma progression⁵. Most non-APC-mutated colon cancers carry alterations in other Wnt pathway members, including mutations in CTNNB1 and RNF43 and R-spondin fusions, and such enrichment of Wnt pathway mutations in colon cancer highlights the importance of Wnt hyperactivation as a master driver of colon tumorigenesis. Our previous sequencing analysis of small intestinal adenomas also captured APC truncations in ~90% of intestinal-type small intestinal adenomas⁶. Despite this high prevalence in adenomas, large-scale sequencing studies on small intestinal adenocarcinomas have consistently highlighted the low frequency of APC mutation (< 30% of cases)^7–12. Other canonical Wnt-activating mutations, including CTNNB1 mutation^10,13, explain only a minority of tumors. This discrepancy suggests that alternative routes drive small intestinal tumorigenesis.

To explore such drivers of human small intestinal tumorigenesis, we here focused on small intestinal tumors histologically distinct from APC-mutated tumors. Their molecular characterization highlighted recurrent in-frame deletions in COPA, encoding α-COP or COPA, and truncating mutations in USP9X. COPA is a subunit of coatomer complex I that coats vesicles responsible for intra-Golgi trafficking as well as retrograde transport from the Golgi to the endoplasmic reticulum¹⁴. USP9X encodes the USP9X deubiquitinase known to affect multiple pathways, including those related to cancer¹⁵, and is recurrently mutated in several cancers, including specific subtypes of head and neck and ovarian carcinomas^16,17. While COPA in-frame deletion has not been captured in past cancer genomics studies, phenotypic characterization using patient-derived organoids and gene engineering revealed Wnt signaling hyperactivation independent of intestinal stem cell niche factors in COPA-mutated or COPA-engineered small intestinal organoids. Our findings reveal an alternative mode of Wnt activation during small intestinal tumorigenesis, which partly accounts for the low prevalence of APC mutations in small intestinal adenocarcinoma.

Results

COPA^exon13 in-frame deletions in small intestinal adenomas

Pathologically, most small intestinal adenomas are flat, APC-mutant lesions with straight glands lined by small intestinal-type epithelium⁶ (Fig. 1a). However, we identified a morphologically distinct subset with protruded endoscopic appearance and histologically characterized by branched glands and ectopic crypt formation while retaining small intestinal-type differentiation (Fig. 1a,b), suggesting distinct molecular alterations. In a pathological review of 254 consecutive small intestinal adenoma cases including ampullary adenomas, such protruded-type adenomas with intestinal differentiation accounted for 6.3% (16/254) of the cohort (Extended Data Fig. 1a). Whole-exome sequencing (WES) of three cases with typical protrusion and intestinal differentiation, which were not included in the consecutive cases, did not capture mutations in known Wnt drivers (Supplementary Table 1). Instead, all three cases harbored an in-frame deletion in COPA, which were confined to exon 13 (COPA^ex13) (Fig. 2a,b and Extended Data Fig. 1b). These mutations co-occurred with copy-neutral loss of heterozygosity (CN-LOH) at chromosome 1q spanning the COPA locus, suggesting a hypomorphic effect (Fig. 2c and Extended Data Fig. 1c). WES also identified a truncating mutation in USP9X in all three cases (Fig. 2d). Immunostaining revealed mosaic loss of USP9X protein in two tumors, which suggests that USP9X truncation in these tumors occurred secondary to COPA^ex13 in-frame deletion (Fig. 2e and Extended Data Fig. 1d).

Fig. 1. Morphological and histological distinctions in small intestinal adenomas.

a, Endoscopic and histological images of flat (top) and protruded (bottom) intestinal-type adenomas of the duodenum. Left: endoscopic images demonstrate distinct flat and protruded appearances, respectively. Middle and right: hematoxylin and eosin (H&E)-stained histological sections show a flat-type adenoma composed of straight, vertically oriented glands, whereas the protruded-type adenomas exhibit a lobulated architecture composed of blanched glands. Tumor IDs are included on the upper left of each image. An APC frameshift mutation in DUA077 was detected by Sanger sequencing of the APC mutation cluster region in our previous study⁶. Representative data from 108 APC-mutated flat intestinal-type and 3 protruded intestinal-type adenomas. White arrowheads show tumors. Scale bars, 500 μm (middle) and 250 μm (right). b, Serial immunohistochemistry images of normal duodenal mucosa, a flat-type adenoma and a protruded-type adenoma. Ki-67, a marker of proliferating cells, highlights altered proliferative zones in the adenomas. The flat-type adenoma shows proliferative cells in the upper layer, whereas the protruded-type adenoma demonstrates ectopic crypt formation (arrowheads). α-defensin 5 (DEFA5) staining depicts altered Paneth cell distribution in the adenomas (arrowheads). Expression of CD10, MUC2 and CDX2, which serve as markers for absorptive cells, goblet cells and intestinal epithelium, respectively, supports small intestinal epithelial differentiation. Representative data from 13 (normal mucosa), 10 (APC-mutated flat intestinal-type adenoma) and 3 (protruded intestinal-type adenoma) samples. Scale bars, 100 μm.

Source data

Extended Data Fig. 1. COPA and USP9X mutations in small intestinal adenomas.

a, Breakdown of anatomical location, macroscopic appearance and histological subtype in a consecutive series of 254 small intestinal adenomas. b, Genomic sequences of COPA^ex13 in WES-analyzed COPA-mutated duodenal adenomas. c, CN-LOH in chromosome 1 detected in COPA-mutated duodenal adenomas. d, USP9X immunostaining in DUA_B and DUA_C tumors. Low (left) and high (right) magnification. DUA_B shows heterogenous USP9X loss, and dotted lines indicate USP9X-lost (pink) and -retained (blue) areas. DUA_C shows homogenous USP9X loss, and dotted lines outline tumor (pink) and normal (green) areas. Scale bar, 1 mm (left), 100 μm (right). e, USP9X immunostaining patterns in COPA wild-type (left) and -mutated (right) small intestinal adenomas.

Source data

Fig. 2. Recurrent COPA^ex13 in-frame deletion in small intestinal adenomas.

a, WES analysis of three protruded-type duodenal adenomas (DUA_A, DUA_B and DUA_C). b, WES detection of COPA^ex13 in-frame deletions in DUA_A, DUA_B and DUA_C. COPA protein domains and amino acid deletions are shown for the sequenced cases. c, WES-based copy number analysis of COPA-mutant duodenal adenomas. All cases are virtually euploid but show CN-LOH on chromosome 1q spanning the COPA locus. d, Lollipop plot of USP9X mutations detected in COPA-mutated duodenal adenomas. UBL, ubiquitin-like domain; USP, ubiquitin-specific domain. e, USP9X immunostaining of the DUA_A tumor. Low (left) and high (right) magnification images show heterogeneous USP9X immunoreactivity. USP9X-retained and USP9X-lost areas are outlined with blue and pink dashed lines, respectively. Scale bars, 1 mm (left) and 100 μm (right). f, Sanger sequencing of COPA^ex13 in a consecutive series of 254 small intestinal adenomas. The number of COPA^ex13-wild-type (gray) and COPA^ex13-mutated (purple) is shown according to anatomical location, macroscopic appearance and histological differentiation (bottom). A total of 10 cases with COPA^ex13 in-frame deletions were identified, with an enrichment in protruded intestinal-type adenomas (7 of 16 cases). g, Amino acid deletion patterns in the ten cases of COPA-mutated small intestinal adenoma.

Source data

To further investigate the prevalence of COPA^ex13 mutation, we sequenced COPA^ex13 in the aforementioned cohort of 254 small intestinal adenomas and identified 10 cases with a COPA^ex13 mutation, which were enriched in protruded adenomas with intestinal differentiation (7/16 cases) (Fig. 2f). All COPA^ex13 mutations were in-frame deletions affecting 6–18 amino acids, particularly residues L393–P397 (Fig. 2g and Supplementary Table 2). USP9X immunostaining of a subset of these cases from the consecutive series revealed preserved USP9X expression in 95 of 96 COPA-wild-type adenomas and its loss in 2 of 10 COPA-mutant adenomas, consistent with the prior observation (Extended Data Fig. 1e and Supplementary Table 2). Together, these data establish COPA^ex13 in-frame deletion and USP9X-truncating mutation as alternative drivers of small intestinal tumorigenesis independent of canonical Wnt driver gene mutations.

COPA^ex13 in-frame deletions in small intestinal adenocarcinomas

The discovery of COPA^ex13 in-frame deletions in small intestinal adenomas prompted us to analyze their prevalence in small intestinal adenocarcinomas. Sequencing of COPA^ex13 in 55 duodenal adenocarcinomas identified 7 cases with a COPA^ex13 in-frame deletion encompassing L393–K398, supporting the recurrence and relevance of this genetic event during small intestinal carcinogenesis (Fig. 3a and Supplementary Table 3). In line with the USP9X immunostaining pattern in adenomas, USP9X expression was retained in COPA-wild-type adenocarcinomas, whereas a fraction of COPA-mutant adenocarcinomas lost USP9X (Fig. 3b,c and Supplementary Table 3).

Fig. 3. COPA^ex13 in-frame deletion in small intestinal adenocarcinomas.

a, Sanger sequencing of COPA^ex13 in 55 duodenal adenocarcinomas identified 7 cases with COPA^ex13 in-frame deletions. b, USP9X immunoreactivity patterns in COPA-wild-type (left) and COPA-mutated (right) duodenal adenocarcinomas. c, USP9X immunostaining in two COPA-mutant duodenal adenocarcinoma cases. Cases with heterogeneous (left) and homogeneous (right) USP9X loss are shown. Pink and blue dashed lines demarcate USP9X-lost and USP9X-retained areas, respectively. The dashed lines in DUC21 indicate the boundary between the adenocarcinoma and non-neoplastic glands (pink, tumor; green, normal). Scale bars, 100 μm. d, Mutation profile of small intestinal adenomas and adenocarcinomas with COPA^ex13 in-frame deletions. Genetic mutations detected by WES in our cases and three COPA-mutant cases found in a previous study³⁴ are shown. e, A case of COPA-mutated duodenal adenocarcinoma with an adenoma component (DUC02). A macroscopic tumor image and histology micrographs of each component are shown. A COPA^ex13 in-frame deletion was detected in both components by Sanger sequencing but was missed by WES, presumably due to the relatively large size of the deletion (60-bp deletion) and mapping failure. Scale bars, 10 mm (macroscopic image) and 200 μm (micrographs).

Source data

We performed WES analysis of COPA-mutated duodenal adenocarcinoma cases that met the DNA quality threshold for WES, together with three adenoma cases in the consecutive series and one adenoma found to have a COPA^ex13 in-frame deletion post hoc (DUA_D), totaling seven adenomas and four adenocarcinomas with COPA^ex13 in-frame deletions. WES confirmed the absence of canonical Wnt driver gene mutations in these COPA-mutant small intestinal tumors, while other drivers, including TP53 and KRAS mutations, were detected mainly in adenocarcinoma (Fig. 3d and Supplementary Table 1). Around half of the cases harbored USP9X-truncating mutations (Fig. 3d). A review of prior WES studies on small intestinal and ampullary cancer^7,9,13,18,19 further identified three cases with a COPA^ex13 in-frame deletion, validating its presence in other cohorts (Fig. 3d).

Notably, one adenocarcinoma case (DUC02) was found to have a coexisting adenoma component (Fig. 3e). COPA^ex13 Sanger sequencing and WES of the adenoma component revealed that the adenoma component had a COPA^ex13 in-frame deletion identical to the one in the adenocarcinoma component but lacked canonical cancer drivers present in the adenocarcinoma component, including KRAS hotspot and ELF3-truncating mutations (Fig. 3e and Supplementary Table 1). Albeit a single case, this example clearly illustrates adenocarcinoma development from a COPA-mutant adenoma. Together, these findings suggest that COPA^ex13 in-frame deletion represents the primary hit during carcinogenesis of COPA-mutant tumors with USP9X-truncating mutation occurring as a second hit, and additional alterations in canonical cancer drivers such as KRAS and TP53, as well as chromosomal instability, may further facilitate adenoma–carcinoma progression.

Activation of Wnt signaling in COPA-mutated adenomas

To understand the biological effect of COPA^ex13 in-frame deletions, we extracted RNA from formalin-fixed paraffin-embedded (FFPE) sections of normal duodenal mucosa, COPA-mutated adenomas and flat intestinal-type adenomas and performed RNA sequencing (RNA-seq) (Supplementary Table 4). Principal component analysis clearly distinguished these three types based on the transcriptome pattern (Fig. 4a and Extended Data Fig. 2a). COPA-mutant adenomas upregulated Wnt target and intestinal stem cell genes, including LGR5, AXIN2 and SMOC2, while downregulating enterocyte markers, albeit to a lesser extent than flat intestinal-type adenomas (Fig. 4b and Extended Data Fig. 2b,c). Inference of transcription factor network activity by regulon analysis confirmed decreased activity of differentiation-related transcription factors such as HNF-1α and CDX2 and enhanced activity of transcription factors associated with Wnt signaling (β-catenin and TCF7L2) and the cell cycle in COPA-mutant adenomas (Fig. 4c). Transcriptome analysis also highlighted unique upregulation of metallothionein genes in COPA-mutant adenomas (Fig. 4b,d and Extended Data Fig. 2c,d). Although the biological significance of elevated levels of metallothioneins in these tumors remains unclear, these genes may serve as potential biomarkers for COPA^ex13 in-frame deletion. Collectively, COPA mutations activate Wnt signaling in the absence of canonical pathway mutations.

Fig. 4. Transcriptomic characterization of small intestinal adenomas with COPA^ex13 in-frame deletion.

a, A transcriptome-based principal component (PC) analysis plot of normal small intestinal mucosa, flat intestinal-type adenomas and COPA-mutated adenomas. PC1 mainly divides normal and tumor tissues, whereas PC2 partitions COPA-mutated adenomas from the others. b, The expression of PC1- and PC2-contributing genes. Genes with an absolute loading above the 75th percentile are shown. c, Transcription factor (TF) activity inference in each sample. The activity of transcription factor networks was calculated using decoupleR. Transcription factors with a standard deviation above the 90th percentile are shown in the heatmap. d, Genes differentially expressed between COPA-mutated adenomas and normal duodenal mucosa. Differentially expressed genes were analyzed using DESeq2 with Benjamin–Hochberg P-value correction.

Source data

Extended Data Fig. 2. Transcriptomic characterization of COPA-mutated duodenal adenomas.

a, Ontology analysis of genes with a high or low loading in principal component (PC) 1 or PC2. Ontology enrichment was analyzed using Metascape with Benjamin-Hochberg P-value correction. b, Enrichment of marker genes for the indicated cell types in the small intestinal epithelium. The gene sets were obtained from a previous single-cell analysis of the human intestine⁴⁸. c, Differentially expressed gene analysis of COPA-mutated adenomas versus flat-type adenomas. The analysis was performed using DESeq2 with Benjamin-Hochberg P-value correction. d, The expression of metallothionein genes in the analyzed samples. No samples expressed MT3 and MT4 genes. TPM; transcripts per million.

Source data

Prospective derivation of COPA-mutated duodenal adenoma organoids

To functionally investigate the impact of COPA^ex13 in-frame deletions in human small intestinal epithelium, we set out to establish patient-derived organoids from COPA-mutant small intestinal tumors. Although COPA-mutant small intestinal adenomas are rare, their unique protruding macroscopic appearance and intestinal differentiation allowed us to prospectively select cases suspected to have a COPA^ex13 in-frame deletion. Consequently, we were able to sample tissues from such cases and successfully established three patient-derived duodenal adenoma organoid lines with COPA^ex13 in-frame deletions (Fig. 5a and Supplementary Table 5). Sanger sequencing and WES confirmed COPA^ex13 in-frame deletions in the established adenoma organoids and their absence in patient-matched normal counterpart organoids (Fig. 5b, Extended Data Fig. 3a and Supplementary Table 6). Consistent with the initial finding in clinical samples, WES detected similar CN-LOH at chromosome 1q spanning COPA in all organoid lines (Fig. 5b and Extended Data Fig. 3b), whereas USP9X mutation was absent (Supplementary Table 6). Testing of the dependence of COPA-mutated organoids on individual growth factors revealed that COPA-mutated organoids were stringently dependent on exogenous Wnt-3A and receptor tyrosine kinase ligands, including EGF, IGF-1 and FGF-2, but were less dependent on R-spondin, which is a ligand for LGR5 and an essential amplifier of Wnt signaling, and bone morphogenetic protein (BMP) antagonist Noggin, compared with the normal counterparts (Fig. 5c and Extended Data Fig. 3c). RNA-seq analysis further demonstrated retained Wnt activation in COPA-mutated organoids despite R-spondin withdrawal, as well as high expression of metallothionein genes, suggesting cell-intrinsic upregulation by COPA mutation (Fig. 5d and Extended Data Fig. 3d,e). These results collectively indicate that COPA^ex13 in-frame deletions drive Wnt activation in a Wnt ligand-dependent but R-spondin- and Noggin-independent manner, thereby conferring a selective growth advantage under environments with limited R-spondin and Noggin availability.

Fig. 5. Establishment of patient-derived COPA-mutated duodenal adenoma organoids.

a, Prospective derivation of COPA-mutated organoids. A total of three lines (DUA_PDO1–DUA_PDO3) were established. Endoscopic, histological and organoids images of each tumor are shown. Histology images show H&E staining of two endoscopically resected tumors (DUAD_PDO1 and DUAD_PDO2) and an endoscopic biopsy specimen (DUAD_PDO3). The organoid images represent at least ten organoids with similar morphology. Scale bars, 500 μm (middle) and 50 μm (right). b, Top: the genotype of COPA^ex13 in each organoid line analyzed by WES and Sanger sequencing. Bottom: WES-based copy number analysis showing the presence of chromosome 1q CN-LOH in all lines. c, The growth of paired counterpart normal and COPA-mutated adenoma organoids in the indicated conditions. Counterpart normal organoids were derived from the normal duodenal mucosa of each patient. Each growth factor was removed from the full medium, including Wnt-3A (W), R-spondin1 (R), Noggin (N), A83-01 (A) and EGF/IGF-1/FGF-2 (E/I/F) following single-cell plating, and organoid growth was analyzed on day 7 using an ATP-based viability assay. Organoid growth is shown as the ratio of ATP luminescence to that in the condition with full medium. Each dot shows a technical replicate. n = 3 replicates for each condition. Data are shown as mean ± s.d. Statistics, two-sided Welch’s t test. d, RNA-seq analysis of COPA-mutated adenoma organoids and counterpart normal organoids cultured in the indicated conditions. R-spondin1 was withdrawn from the medium 2 days before RNA extraction.

Source data

Extended Data Fig. 3. Characterization of COPA-mutated patient-derived adenoma organoids.

a, Genomic sequences of COPA^ex13 in patient-derived normal counterpart and COPA-mutated duodenal adenoma organoids. b, CN-LOH at chromosome 1 in COPA-mutated adenoma organoids. CN; copy number. c, Representative bright-field images of paired normal counterpart and COPA-mutated adenoma organoids cultured in growth factor-removed conditions. Following single-cell dissociation, 10,000 (DUA_PDO3) or 3,000 cells (the others) were plated per well and cultured in the indicated conditions. The images were captured on day 7. Representative images from 3 technical replicates with similar results. Scale bar, 100 μm. d, Differentially expressed gene analysis of normal counterpart and COPA-mutated organoids. RNA-seq data of organoids cultured in the R-spondin-lacking condition were used. Patient ID was used as a covariate in DESeq2 analysis to account for inter-patient difference. P-values were adjusted using Benjamin-Hochberg correction. e, Metascape analysis of genes upregulated or downregulated in COPA-mutated adenoma organoids. P-values were adjusted using Benjamin-Hochberg correction.

Source data

Modeling COPA^ex13 in-frame deletion in normal intestinal organoids

To directly validate whether COPA^ex13 in-frame deletion drives R-spondin- and Noggin-independent growth, we introduced a 6-amino acid deletion into the COPA alleles of human normal small intestinal organoids via CRIPSR-mediated genome editing (Fig. 6a and Extended Data Fig. 4a). Although we devised a knock-in strategy without knock-in-specific selection, genome editing was remarkably efficient in duodenal and ileal organoids, achieving biallelic in-frame deletion in 38 of the 41 randomly selected clones (Extended Data Fig. 4b). Consistent with the results from patient-derived COPA-mutated adenoma organoids, COPA-engineered (COPA^6del) normal organoids remained dependent on Wnt-3A but showed tolerable growth in the absence of R-spondin or Noggin (Fig. 6b,c and Extended Data Fig. 4c). Further evaluation of growth factor dependence using multiple clones with differing COPA mutation patterns validated R-spondin and Noggin independence in clones with biallelic COPA^ex13 in-frame deletion (Fig. 6d and Extended Data Fig. 4d,e). Notably, clones with a monoallelic COPA^ex13 in-frame deletion accompanied by a truncating or intact counter-allele failed to thrive in the R-spondin- or Noggin-deprived conditions, in agreement with the invariable involvement of chromosome 1q CN-LOH in clinical COPA-mutant adenomas (Extended Data Fig. 4d). COPA-engineered human normal colon organoids also showed modest R-spondin independence (Fig. 6d and Extended Data Fig. 4c). Thus, COPA-edited small intestinal organoids phenocopied patient-derived COPA-mutant organoids and showed that COPA^ex13 in-frame deletion alone is sufficient for the acquisition of R-spondin and Noggin independence.

Fig. 6. CRISPR-mediated modeling of COPA^ex13 in-frame deletion in human small intestinal organoids.

a, The strategy to model COPA^ex13 in-frame deletion using CRISPR-mediated knock-in in human normal small intestinal organoids. sgRNA, single guide RNA. b, The growth of normal ileal organoids (Ileum1) and COPA-engineered organoids (Ileum1-COPA^6del) in growth factor-removed conditions. The organoids were plated with a density of 3,000 cells per well and cultured in the indicated conditions for 7 days. Scale bars, 100 μm. At right, organoid growth is shown as the ratio of ATP luminescence to that in the condition with full medium. The dots show technical replicates. n = 3 replicates for each concentration. Data are shown as mean ± s.d. Scale bars, 100 μm. Statistics, two-sided Welch’s t test. P = 5.38 × 10⁻⁵ ( − W), 0.0245 ( − E/I/F). c, The effect of R-spondin1 concentration on the growth of Ileum1 and Ileum1-COPA^6del organoids. ATP abundance was measured on day 7. The dots show technical replicates. n = 3 replicates for each condition. Data are shown as mean ± s.d. Statistics, two-way ANOVA. d, Validation of the effect of COPA deletion in multiple intestinal organoid lines. The dots show technical replicates. n = 3 replicates for each condition. Data are shown as mean ± s.d. Statistics, two-sided Welch’s t test. e, RNA-seq analysis of COPA-engineered organoids. Parental organoids (Ileum1), a clone with a monoallelic 6-amino acid deletion in COPA (Ileum1-COPA^het) and COPA^6del organoids were analyzed. R-spondin was removed 2 days before RNA extraction. f, Phospho- and total LRP6 protein levels in Ileum1 and Ileum1-COPA^6del organoids. R-spondin was removed 2 days before protein collection. The values indicate the protein chemiluminescence area normalized by the β-actin area and are shown as the ratio to the control (Ileum1, +R-spondin1). Representative data from two assays with similar results.

Source data

Extended Data Fig. 4. Genetic modeling of COPA^ex13 in-frame deletion and USP9X truncating mutation in small intestinal organoids.

a, COPA editing and cloning workflow in detail. b, Genotype of all established subclones. The numbers show clone counts per each genotype. WT; wild-type. c, Representative growth of COPA-edited small intestinal and colon organoids in full and R-spondin1-removed conditions. The organoids were plated as single cells with a density of 3,000 (Duo1, Duo2 and Ileum2) or 1,500 (Colon1) cells per well, and the images were captured on day 7. Representative images from 3 technical replicates with similar results. Scale bar, 100 μm. d, The growth of COPA-engineered Ileum1 clones with varying COPA genotypes. The organoids were plated as single cells (3,000 cells/well) and cultured for 7 days in the indicated conditions. Organoids growth was analyzed using ATP luminescence assay. Organoids with one COPA^ex13 in-frame deletion allele and a wild-type or frameshift counter-allele (COPA^het and COPA^{het-2, 3}) do not show R-spondin independence. Additional 2 COPA^6del clones with a distinct GFP fluorescence (COPA^{6del-2, 3}) were also analyzed for validation. Each dot shows a technical replicate. n = 3 replicated for each condition. Data are shown as mean ± SD. e, The growth of a COPA-engineered Ileum2 clone with heterozygous COPA^ex13 in-frame deletion (18- and 21-bp deletion) exhibiting R-spondin independence. Each dot shows a technical replicate. n = 3 replicates for each condition. Data are shown as mean ± SD. f, Transcriptome analysis of COPA-edited organoids and their normal counterparts. R-spondin was withdrawn from the medium 2 days before RNA extraction. g, TF network activity in COPA^6del organoids versus parental organoids. Data from the R-spondin1-removed condition were used. Parental organoid information was set as a covariate in DESeq2 analysis, and the Wald statistics in the DESeq2 output was used for TF activity calculation with decoupleR. P-values were adjusted using Benjamin-Hochberg correction.

Source data

COPA^ex13 in-frame deletion stabilizes LRP6

To probe the mechanisms of R-spondin and Noggin independence by COPA^ex13 in-frame deletion, we performed RNA-seq of COPA-engineered organoids. Consistent with the RNA-seq analysis of patient-derived COPA-mutant adenoma organoids, COPA^6del organoids, but not counterpart normal organoids or a clone with a heterozygous 6-amino acid deletion in COPA, exhibited retention of Wnt activation and expression of intestinal stem cell genes following R-spondin removal (Fig. 6e). Similar trends were observed in multiple small intestinal organoid lines, as well as in a colon organoid line (Extended Data Fig. 4f,g). A COPA-engineered ileum organoid line with an LGR5-tdTomato reporter (Ileum2)²⁰ showed partial retention of endogenous LGR5 expression in COPA^6del organoids in the absence of R-spondin (Extended Data Fig. 5a). Mechanistically, COPA-engineered small intestinal organoids, as well as COPA-mutated patient-derived adenoma organoids, stabilized LRP6, a Wnt coreceptor normally degraded by RNF43 and ZNRF3 in the absence of R-spondin^21,22 (Fig. 6f and Extended Data Fig. 5b,c). COPA-edited colon organoids also showed LRP6 stabilization in the absence of R-spondin (Extended Data Fig. 5d). This LRP6 stabilization by COPA^ex13 in-frame deletion was considered to be regulated at the protein level, as the mRNA levels of LRP6 and RNF43 did not parallel those of LRP6 protein in ileum organoids with or without COPA editing (Extended Data Fig. 5e). The similar expression pattern of LRP6 and RNF43, as well as AXIN2 and LGR5, between COPA-edited ileum and colon organoids suggested that LRP6 stabilization by COPA mutation is conserved between the small and large intestines at least at the molecular level (Extended Data Fig. 5e). The same pattern of LRP6 stabilization uncoupled from mRNA levels in COPA-edited organoids was also evident following Noggin removal (Extended Data Fig. 5f,g). Given the negative regulation of LGR5 by BMP signaling²³, the R-spondin and Noggin independence of COPA-mutant organoids could be attributed to LGR5-independent LRP6 stabilization.

Extended Data Fig. 5. Mechanism of R-spondin/Noggin-independence in COPA^6del organoids.

a, Fluorescence of the LGR5-tdTomato reporter in Ileum2-LGR5^tdTomato and Ileum2-LGR5^tdTomato-COPA^6del organoids cultured with or without R-spondin. In the −R-spondin1 condition, R-spondin was removed 2 days before image capture. Representative data from 2 experiments with similar results. Scale bar, 1 mm. b, Immunodetection of phospho-LRP6 and total-LRP6 in Ileum2 and Ileum2-COPA^6del organoids cultured with or without R-spondin. R-spondin1 was removed from the medium 2 days before protein harvest. The protein chemiluminescence area was normalized by the β-actin area and shown as the ratio to the control (Ileum2, +R-spondin1) value. Data from 1 run. c, Immunodetection of phospho-LRP6 and total-LRP6 in patient-derived paired normal counterpart and COPA-mutated adenoma organoids cultured with or without R-spondin. R-spondin1 was removed from the medium 3 days before protein harvest. The protein chemiluminescence area was normalized by the β-actin area and shown as the ratio to the control (Normal counterpart, +R-spondin1) value. Data from single runs. d, Immunodetection of phospho-LRP6 and total-LRP6 in Colon1 and Colon1-COPA^6del organoids cultured with or without R-spondin. R-spondin1 was removed from the medium for 3 days before protein harvest. The protein chemiluminescence area was normalized by the β-actin area and shown as the ratio to the control (Colon1, +R-spondin1) value. Representative data from 2 assays with similar results. e, Quantitative PCR (qPCR) analysis of AXIN2, LGR5, LRP6 and RNF43 in Ileum1 and Colon1 parental and COPA^6del organoids cultured with or without R-spondin. R-spondin1 was removed from the medium 2 days before RNA extraction. Each dot indicates a technical replicate, and data are shown as mean ± SD. n = 3 replicates for each condition. Statistics, two-sided Welch’s t-test. f, Immunodetection of phospho-LRP6 and total-LRP6 in Ileum1 and Ileum1-COPA^6del organoids cultured with or without Noggin. Noggin was removed from the medium 3 days before protein harvest. The protein chemiluminescence area was normalized by the β-actin area and shown as the ratio to the control (Ileum1, +Noggin) value. g, qPCR analysis of AXIN2, LGR5, LRP6, RNF43 and ID1 in Ileum1 and Ileum1-COPA^6del organoids cultured with or without Noggin. Noggin was removed from the medium 3 days before harvest. ID1 upregulation corroborates the effect of Noggin removal. Each dot indicates a technical replicate, and data are shown as mean ± SD. n = 3 replicates for each condition. Statistics, two-sided Welch’s t-test. h, Immunostaining of COPA (green) and a cis-Golgi protein GM130 (red) in Ileum1 (top) and Ileum1-COPA^6del (bottom) organoids. Areas in white frame are shown in insets. Nuclear counterstaining, Hoechst 33342. Scale bar, 50 μm (low magnification), 2 μm (insets). i, Quantification of COPA and GM130 colocalization. The colocalization score was quantified as the proportion of double-positive pixels in COPA- or GM130-positive regions. Data are shown as mean ± SD. Each dot represents one organoid. n = 6 (Ileum1) and 7 (Ileum1-COPA^6del). Statistics, two-sided Wilcoxon rank sum test. j, RNA-seq-based enrichment score of metallothionein genes and canonical Wnt target genes in COPA-edited and parental organoids. Each dot shows an organoid line. n = 5 organoid lines. Statistics, two-sided paired t-test.

Source data

The involvement of COPA in cis-Golgi and retrograde vesicle transport prompted us to analyze the localization of COPA in parental and COPA^6del ileum organoids. Consistent with a previous study²⁴, COPA largely colocalized with cis-Golgi surface protein GM130 in parental organoids, and the introduction of COPA deletion did not significantly affect this colocalization pattern (Extended Data Fig. 5h,i). COPA^6del organoids also featured upregulation of metallothionein genes, in line with their unique upregulation in patient-derived COPA-mutant organoids (Extended Data Fig. 5j). R-spondin removal did not affect the expression of metallothionein genes in COPA^6del organoids while marginally downregulating canonical Wnt target genes (Extended Data Fig. 5j), suggesting that metallothionein upregulation in COPA-mutant tumors occurs independently of Wnt hyperactivation.

USP9X loss promotes niche independence in COPA-engineered organoids

Finally, we asked whether USP9X loss provides any growth advantage by genetically disrupting USP9X in ileal organoids with or without the COPA^6del deletion (Fig. 7a and Extended Data Fig. 6a). Knockout of USP9X in COPA^6del organoids, but not in wild-type organoids, enhanced organoid expansion in R-spondin- and/or Noggin-lacking conditions (Fig. 7b,c and Extended Data Fig. 6b), consistent with the patchy USP9X loss in clinical COPA-mutated tumors. RNA-seq of USP9X-knockout (USP9X^KO) organoids and COPA^6del/USP9X^KO double-mutant organoids, together with parental and COPA^6del organoids, revealed upregulation of genes related to Wnt signaling and intestinal stem cells in COPA^6del/USP9X^KO double-mutant organoids but not in USP9X^KO organoids (Extended Data Fig. 6c,d). In line with this observation, USP9X knockout in COPA^6del organoids further augmented LRP6 stabilization following R-spondin or Noggin removal (Fig. 7d and Extended Data Fig. 6e). Dishevelled 2 (DVL2) expression mirrored the dynamics of LRP6 expression following COPA/USP9X editing and R-spondin removal, consistent with its interaction with the LRP6–Frizzled receptor complex²⁵, whereas USP9X knockout alone did not markedly alter the expression of DVL2 and COPA (Fig. 7d). These findings confirm that COPA^ex13 in-frame deletion confers R-spondin and Noggin independence, which is further augmented by USP9X loss.

Fig. 7. Genetic disruption of USP9X.

a, USP9X knockout in Ileum1 and Ileum1-COPA^6del organoids. b, The growth of Ileum1-USP9X^KO and Ileum1-COPA^6del/USP9X^KO organoids in the indicated conditions. The organoids were plated as single cells, and growth was analyzed on day 7. Representative images from three technical replicates with similar results. Scale bars, 100 μm. c, ATP luminescence measurement of Ileum1, Ileum1-USP9X^KO, Ileum1-COPA^6del and Ileum1-COPA^6del/USP9X^KO organoids cultured in the indicated conditions. Each dot shows a technical replicate. n = 3 replicates for each condition. Data are shown as mean ± s.d. Statistics, one-way ANOVA followed by Dunnett’s post hoc test. Data from parental organoids were used for the control. d, Phospho-LRP6, total-LRP6, DVL2 and COPA protein levels in Ileum1, Ileum1-USP9X^KO, Ileum1-COPA^6del and Ileum1-COPA^6del/USP9X^KO organoids. R-spondin was removed 2 days before protein collection. The protein chemiluminescence area was normalized by the β-actin area and shown as the ratio to the control (Ileum1, +R-spondin1) value. Representative data from two assays with similar results. e, The design of the doxycycline-inducible RNF43^HiBiT construct. f, Measurement of total and cell-surface RNF43^HiBiT in wild-type, USP9X^KO, COPA^6del and COPA^6del/USP9X^KO Ileum1 organoids. Gene-engineered organoids were newly generated from an Ileum1-RNF43^HiBiT clone to ensure equal induction of RNF43^HiBiT. Each dot shows a technical replicate. n = 6 replicates for each condition. Data are shown as mean ± s.d. Statistics, one-way ANOVA followed by Dunnett’s post hoc test. The parental organoid values were used as the control.

Source data

Extended Data Fig. 6. Impacts of USP9X knockout.

a, Design of USP9X knockout and sequences of knockout lines. b, Replicate analysis of growth factor dependence using additional USP9X^KO Ileum1 clones. The organoids were plated as single cells at a density of 3,000 cells/well, and ATP luminescence was measured at day 7. Each dot shows a technical replicate, and data are shown as mean ± SD. n = 3 replicates for each condition. Statistics, two-sided Welch’s t-test. c, RNA-seq analysis of USP9X- and/or COPA-engineered organoid lines cultured with or without R-spondin. R-spondin1 was withdrawn from the medium 2 days before harvest. Genes with a SD > 98th percentile are used for heatmap visualization. d, Enrichment of intestinal stem cell and Wnt target genes in USP9X- and/or COPA-engineered organoids. e, Immunodetection of phospho- and total-LRP6 protein in Ileum1, Ileum1-USP9X^KO, Ileum1-COPA^6del and Ileum1-COPA^6del/USP9X^KO organoids cultured with or without Noggin. Noggin was removed 3 days before protein harvest. The protein chemiluminescence area was normalized by the β-actin area and shown as the ratio to the control (Ileum1, +Noggin) value. Data from 1 run.

Source data

COPA^ex13 in-frame deletion reduces cell-surface RNF43 levels

To further investigate the mechanism of LRP6 stabilization in COPA mutants, we set our focus on the expression of RNF43, which facilitates cell-surface LRP–Frizzled clearance in an R-spondin- and LGR5-sensitive manner²². To this end, we used the HiBiT system, which enables accurate luciferase-based measurement of total or cell-surface protein tagged with the HiBiT peptide (Fig. 7e). Induced expression of HiBiT-tagged RNF43 (RNF43^HiBiT) downregulated LRP6 and AXIN2 in Ileum1 organoids, validating the expression of functional RNF43 (Extended Data Fig. 7a,b). Using this inducible RNF43^HiBiT-expressing Ileum1 organoid clone (Ileum1-RNF43^HiBiT), we newly generated USP9X^KO, COPA^6del and COPA^6del/USP9X^KO organoid lines (Extended Data Fig. 7c). To avoid LGR5 downregulation and cytostasis by RNF43^HiBiT expression, we applied GSK3 treatment before measurement (Extended Data Fig. 7d,e). Introduction of the COPA^6del deletion downregulated cell-surface levels of RNF43 in both the presence and the absence of R-spondin (Fig. 7f), which suggests that LRP6 stabilization in COPA mutants is partly due to reduced cell-surface RNF43 levels. In contrast, USP9X knockout reduced total RNF43 abundance, and COPA^6del/USP9X^KO organoids exhibited a further composite reduction in both total and cell-surface RNF43 levels (Fig. 7f). Altogether, COPA^ex13 in-frame deletion promotes R-spondin- and LGR5-independent growth at least in part through downregulation of cell-surface RNF43 levels and subsequent LRP6 stabilization.

Extended Data Fig. 7. Inducible expression of RNF43^HiBiT in small intestinal organoids.

a, LRP6 downregulation by induced RNF43^HiBiT expression in Ileum1 organoids. The organoids were cultured in the full medium and treated with the indicated concentrations of doxycycline (Dox) for 2 days. The protein chemiluminescence area was normalized by the β-actin area and shown as the ratio to the control (Ileum1-RNF43^HiBiT, no Dox) value except for RNF43 (HA). N.D., not detected. b, Dose-dependent RNF43 induction and AXIN2 downregulation in Ileum1-RNF43^HiBiT organoids. The organoids were cultured in the full medium and treated with the indicated doses of Dox for 2 days. mRNA expression was analyzed by qPCR. Primers specific to the RNF43 transgene target FLAG and HA tags near the C terminus of RNF43^HiBiT. Each dot shows a technical replicate, and data are shown as mean ± SD. n = 3 replicates for each condition. Statistics, one-way ANOVA followed by Dunnett’s post-hoc test. The Dox-null condition was used as control. c, Genotype of Ileum1-RNF43HiBiT-USP9X^KO and Ileum1-RNF43HiBiT-COPA6^del-USP9X^KO clones. d, Culture protocol for the detection of RNF43^HiBiT. Following single-cell plating, the organoids were cultured in the full medium containing all growth factors including Wnt-3A and R-spondin1. After 5 days, Wnt-3A was removed from the medium, whereas GSK3 inhibitor CHIR99021 (5 μM) and Dox (50 ng/ml) were added. In the −R-spondin condition, R-spondin1 was also removed. The abundance of total and cell-surface RNF43^HiBiT was measured on day 7. e, The expression of the indicated mRNAs in Ileum1-RNF43^HiBiT (parental) and Ileum1-RNF43^HiBiT-COPA^6del organoids cultured with the medium lacking Wnt-3A and supplemented with CHIR99021 for 2 days. Each dot shows a technical replicate, and data are shown as mean ± SD. n = 3 replicates for each condition. Statistics, two-sided Welch’s t-test.

Source data

Discussion

Activation of Wnt signaling is essential for the self-renewal of the intestinal epithelium, as represented by the stringent dependence of human small intestinal and colonic organoids on exogenous Wnt-3A and R-spondin²⁶. Recurrent APC mutation, which confers Wnt-3A and R-spondin independence on human intestinal organoids^27,28, has characterized the inception of the adenoma–carcinoma sequence in the colon, whereas the identification of a similar genetic pathway in the small intestine has remained elusive. Our recognition of the protruded intestinal-type small intestinal adenoma led to the illumination of recurrent COPA^ex13 in-frame deletion, and patient-derived organoids coupled with gene editing further supported its role as an alternative driver of R-spondin-independent growth during small intestinal tumorigenesis. Unlike conventional Wnt regulators, COPA encodes a subunit of coat protein complex I, which mediates intra-Golgi transport and Golgi-to-endoplasmic reticulum retrograde vesicle trafficking¹⁴, defining an alternative mechanistic class of tumor drivers. Although germline COPA missense mutations in specific codons in COPA exons 7–10 are known to cause an autoinflammatory syndrome termed COPA syndrome^29,30, these mutations are distinct from the COPA^ex13 in-frame deletions identified in small intestinal tumors.

Although COPA^ex13 in-frame deletion has not been documented in cancer genomics studies, including those on colorectal cancer, introduction of the COPA^6del deletion in human normal colon organoids provided modest R-spondin independence, as well as LRP6 stabilization and Wnt signaling retention in the absence of R-spondin (Extended Data Fig. 5d,e). The regulation of Wnt activation by COPA^ex13 in-frame deletion thus appears to be preserved across the small intestine, colon and perhaps other tissues at least at the molecular level. Although the reason for the lack of COPA^ex13 in-frame deletion in colorectal cancer is unclear, colorectal carcinogenesis may require higher Wnt activity, which is predominantly driven by alterations in downstream molecules such as APC and β-catenin, compared to the small intestine, thereby favoring ligand-independent constitutive Wnt activation over ligand-driven activation.

COPA^ex13 in-frame deletion also upregulated metallothionein genes, which are known to be induced by external stimuli such as heavy metal exposure, inflammation and oxidative stress³¹. Dysregulation of such signals was not apparent in the transcriptome analysis of patient-derived COPA-mutant adenomas (Extended Data Fig. 2a). Expression of metallothionein genes was also not linked with Wnt activation (Extended Data Figs. 2d and 5j). Metallothionein upregulation in COPA-mutant and COPA-edited organoids cultured in a uniform and antioxidant-rich milieu further suggests endogenous mechanisms. Although the cause of metallothionein induction in COPA-mutant tumors remains elusive, the unique upregulation of these genes may be exploited as a surrogate marker for COPA mutation.

Our analysis further highlighted USP9X loss as a context-dependent growth potentiator that provided R-spondin and Noggin independence only in the presence of COPA^ex13 in-frame deletion. USP9X is a deubiquitinase known to target multiple cancer-related pathways, including the Wnt pathway^15,32,33. Although COPA^ex13 in-frame deletion, as well as additional USP9X knockout, reduced cell-surface RNF43 levels, thereby stabilizing LRP6, further studies are warranted to clarify the mechanisms by which specific COPA mutations, in concert with USP9X loss, downregulate cell-surface RNF43 levels and confer both R-spondin and Noggin independence.

In summary, we identified small intestinal tumors characterized by the COPA^ex13 in-frame deletion. Patient-derived and genetically engineered small intestinal organoids with COPA^ex13 in-frame deletion provide valuable human-relevant models for studying the phenotypic and mechanistic impact of COPA mutations. Our findings not only highlight a unique driver of intestinal tumorigenesis but also shed light on the previously unknown regulation of canonical Wnt signaling by coatomers.

Methods

Human samples

For histological assessment and DNA/RNA analysis of FFPE small intestinal tissues, we used patient samples archived in the National Cancer Center Hospital, Tokyo, Japan, and National Cancer Center Hospital East, Chiba, Japan. Tissue samples for organoid derivation were collected from patients undergoing endoscopy or elective surgery at Keio University Hospital and The University of Tokyo Hospital. The use of archived paraffin-embedded sections was approved by the ethical committee of the National Cancer Center under approval ID 2013-011, and the establishment of patient-derived organoids and their downstream analysis were approved by the ethical committees of Keio University School of Medicine and Graduate School of Medicine and Faculty of Medicine, The University of Tokyo, under approval IDs 20170170 and G3553, respectively. Written informed consent was obtained from all patients. Patient information is provided in Supplementary Tables 2, 3 and 5.

Immunohistochemistry and immunofluorescence

Immunohistochemical staining was performed on FFPE specimens. Antigen retrieval was performed using Target Retrieval Solution pH 9 (Agilent Technologies). Primary antibodies used for immunohistochemistry are listed in Supplementary Table 7. For staining, we used an Autostainer Link 48 (Agilent Technologies) according to the manufacturer’s protocol. An EnVision FLEX kit and Mouse LINKER (Agilent Technologies) were used for detection.

Immunofluorescence staining of organoids was also performed on FFPE sections. Organoids were released from Matrigel using Cell Recovery Solution (Corning), embedded in iPGell (GenoStaff), and fixed with 4% paraformaldehyde in PBS at room temperature overnight. FFPE sections (5-μm thick) were reacted with primary antibodies provided in Supplementary Table 7 and labeled with Alexa Fluor secondary antibodies (Thermo Fisher Scientific). Nuclei were stained with Hoechst 33342 (Thermo Fisher Scientific). Fluorescence images were captured using an FV4000 confocal microscope (Evident). Colocalization of COPA and GM130 signals was analyzed using the ImageJ plugin EzColocalization (version 1.1.4)³⁵.

Macroscopic and histological classification of adenomas

The macroscopic appearance of adenomas was classified as protruded or flat based on endoscopic and/or gross pathological evaluation³⁶. Ampullary adenomas predominantly confined to the common channel or the distal intramural segment of bile and pancreatic ducts were classified as intra-ampullary³⁷. Based on histology, duodenal, ileal and jejunal adenomas were categorized as intestinal-type or gastric-type according to their histologic features³⁸. Ampullary adenomas were classified as intestinal type or gastric/pancreatobiliary type³⁷. Lesions exhibiting both intestinal and gastric or gastric/pancreatobiliary features were designated as mixed-type lesions. Macroscopic and histological subtypes of each adenoma are shown in Supplementary Table 2.

WES

WES of clinical samples was conducted using 200 ng of genomic DNA isolated from snap-frozen or FFPE tumor specimens and noncancerous tissues obtained from patients with COPA-mutated tumors. Tumor regions were microdissected under a microscope using a fine needle, and DNA was extracted using the QIAamp DNA FFPE Tissue Kit (QIAGEN). Exome capture was performed using the Twist Comprehensive Exome Panel or Twist Exome 2.0 plus Comprehensive Exome Spike-in (Twist Bioscience HQ), according to the manufacturer’s instructions. Sequencing was performed on the NovaSeq 6000 platform (Illumina) using 2 × 150-bp paired-end reads, resulting in ~150-fold genome coverage. Basic alignment and sequence quality control were conducted according to the GATK4 best practices pipeline³⁹. The resulting reads from WES were aligned to hg38, using Parabricks v4.3.1.1 (NVIDIA), which delivers the high-speed analysis recommended by the Genome Analysis Toolkit (GATK), with GPU acceleration. Somatic single-nucleotide variants and small insertions/deletions were called using Mutect2 (GATK version 4.1.2.0). All selected variants were manually checked using the Integrative Genomics Viewer (IGV)⁴⁰. Somatic mutations detected in clinical samples are listed in Supplementary Table 1.

Approximately 5 μg of genomic DNA was used for WES analysis of duodenal adenoma organoids. DNA was extracted from organoids using the QIAamp DNA Blood Mini Kit (QIAGEN). WES libraries were generated using the SureSelect Human All Exon V6 kit (Agilent) according to the manufacturer’s instructions and were sequenced on a NovaSeq 6000 device, yielding around 10 Gb of 150-bp paired-end reads. Following adapter trimming with cutadapt (version 1.16), sequencing reads were mapped to the human genome GRCh37 (hg19) using the Burrows-Wheeler Aligner (version 0.17.8). Duplicated reads were tagged using MarkDuplicatesSpark in GATK (version 4.6.1), and base quality score calibration was done using BaseRecalibrator and ApplyBQSR in GATK. Variant calling was performed using Mutect2. Germline variants were removed by filtering out variants with an allele frequency of 0.1% or more in the Database of Single Nucleotide Polymorphisms or the Japanese germline variant database (the Human Genetic Variation Database). Furthermore, variants found in at least two of our panel of normal samples (23 normal organoids and 35 blood samples that were sequenced in our previous studies^20,41,42) were also removed. Germline variants detected in counterpart normal organoids were also filtered out. DNA copy number analysis was performed using FACETS (version 0.6.2)⁴³ with default settings. WES results of patient-derived organoids are provided in Supplementary Table 6.

Sanger sequencing

DNA samples were extracted from FFPE specimens using the QIAamp DNA FFPE Tissue Kit (Qiagen). DNA samples from organoids were collected using the QIAamp DNA Blood Mini Kit (Qiagen). Genomic PCR was performed using the primers listed in Supplementary Table 8. The PCR products were cleaned up using ExoSAP-IT (Thermo Fisher Scientific) and subjected to Sanger sequencing.

RNA-seq

Total RNA was extracted from FFPE tumor specimens using the RNeasy FFPE Kit (Qiagen), and aliquots (10 ng) of the purified material were used to prepare RNA-seq libraries with the SMART-Seq Stranded Kit (Takara Bio). The resultant libraries were subjected to paired-end sequencing of 150-bp reads using the NovaSeq 6000 system (Illumina). The first three bases in the 5′ end of the second sequencing reads were removed as SMART-Seq adapters using cutadapt. The trimmed reads were aligned to human genome hg38 using STAR (version 2.7.11b)⁴⁴, and read counts of the human genes in Ensembl release 81 were calculated using RSEM (version 1.3.3)⁴⁵. Quality metrics of FFPE RNA samples are provided in Supplementary Table 4.

Total RNA for RNA-seq of organoid samples was collected using the Direct-zol RNA Miniprep Kit (Zymo Research). Libraries were prepared using the TruSeq Stranded Total RNA (Illumina) and sequenced on a NovaSeq 6000 device. TruSeq adapters were removed using cutadapt. Mapping and gene-level read calculation were performed using STAR and RSEM as described above.

Read count data were normalized using the varianceStabilizingTransformation function in DESeq2 (version 1.42.1) with the blind = FALSE option, and the normalized data were used for principal component analysis and heatmap construction. Genes with a standard deviation of >98th percentile of all genes were used for principal component analysis. Differentially expressed gene analysis was performed using DESeq in DESeq2. In analyses involving organoids derived from different patients, patient ID was used as a covariate to take into account interpatient differences. Due to the inflation of the fold-change values of metallothionein genes, we adjusted the fold-change values with the lfcShrink function in DESeq2 using the adaptive Student’s t prior shrinkage estimator. Metascape interface⁴⁶ was used for ontology enrichment analysis. Transcription factor networks with weight scores were obtained using get_collectri in decoupleR⁴⁷ (version 2.9.7), and their activities were calculated using run_ulm in decoupleR. Normalized count values or Wald statistics in the DESeq2 output were used for network activity inference. Single-sample gene set enrichment analysis was performed using the gsva function in GSVA (version 1.50.5). Marker genes for cell types in the human small intestine were derived from a recent single-cell analysis of the human intestine⁴⁸. Transit-amplifying cell genes are a union of TA1 and TA2 genes. Previously reported targets of the canonical Wnt signal in human organoids⁴⁹ were used as canonical Wnt target genes. Gene sets used in the study are shown in Supplementary Table 9.

Organoid derivation

Human normal or adenomatous small intestinal tissues were obtained by endoscopic biopsy, by sampling of endoscopically resected tumor or from surgical specimens. The human normal colon organoid line used for COPA engineering (Colon1) was previously described⁵⁰ (normal counterpart of CRC26). The Duo2 (ref. ⁵¹) and Ileum2 (ref. ²⁰) lines were also established previously. Normal intestinal tissues were cut into 1-mm³ fragments using sharp forceps and washed at least ten times with cold PBS by pipetting. The tissues were treated with 2.5 mM EDTA (Thermo Fisher Scientific) for 1 h at 4 °C with gentle rocking, and the epithelium was released by pipetting. The epithelial fragments were embedded into Matrigel (Corning) domes and overlaid with the full medium prepared as follows: Advanced Dulbecco’s Modified Eagle’s Medium/F12 (Thermo Fisher Scientific) supplemented with penicillin/streptomycin (Thermo Fisher Scientific), 10 mM HEPES (Thermo Fisher Scientific), 2 mM GlutaMAX (Thermo Fisher Scientific), 1× B-27 Supplement (Thermo Fisher Scientific), 10 nM gastrin I (Sigma-Aldrich), 1 mM N-acetylcysteine (Sigma-Aldrich), 100 ng ml⁻¹ recombinant mouse Noggin (PeproTech), 50 ng ml⁻¹ recombinant mouse EGF (Thermo Fisher Scientific), 100 ng ml⁻¹ recombinant human IGF-1 (BioLegend), 50 ng ml⁻¹ recombinant human FGF-basic (FGF-2; PeproTech), 1 μg ml⁻¹ recombinant human R-spondin1 (R&D), 500 nM A83-01 (Tocris) and 5% Afamin-Wnt-3A serum-free conditioned medium. We added 50 μM Primocin (InvivoGen) for the first 2 days after organoid establishment and 10 μM Y-27632 (FUJIFILM Wako Pure Chemical Corporation) for the first 2 days after every passaging.

Tumor intestinal tissues were also minced into 1-mm³ fragments and washed with ice-cold PBS. The tumor fragments were digested with 50 μg ml⁻¹ Liberase TH Research Grade (Roche) for 1 h at 37 °C with mechanical disruption by pipetting every 15 min. The supernatant was collected, and the tissue remnants were further treated with TrypLE Express (Thermo Fisher Scientific) for 15 min at 37 °C. The digested tissues were suspended in FBS to inactivate Liberase and washed two times with PBS. Tumor clusters were then embedded in Matrigel and overlaid with the full medium. Primocin and Y-27632 were used as described above. Dependence of organoids on each growth factor was assessed by plating the organoids as single cells with a density of 3,000 cells per well and culturing them in a condition lacking each growth factor. The CellTiter-Glo 3D Cell Viability Assay (Promega) and a GloMax plate reader (Promega) were used for ATP measurement. Clinical information of patient-derived organoids is provided in Supplementary Table 5.

Vector construction and gene engineering of organoids

Genomic targets for CRISPR-mediated gene editing were designed using CRISPick⁵² and cloned into the pX330 vector (42230, Addgene). The single-guide RNA targets are provided in Supplementary Table 8. To construct a donor vector for COPA 6-amino acid deletion, we cloned PCR-amplified 1-kb genomic loci flanking the 5′ and 3′ ends of the 18 bases to be deleted into the pCR2.1-TOPO vector (Thermo Fisher Scientific) using the In-Fusion HD Cloning Kit (Takara Bio).

CRISPR-mediated gene editing in patient-derived organoids was performed via electroporation as previously described⁵³ with a few modifications. Before electroporation, the organoids were treated with 10 μM Y-27632 and 1.25% DMSO for 1 day. The organoids were dissociated into single cells using TrypLE Express and washed two times with Opti-MEM (Thermo Fisher Scientific). The cell pellet was then suspended in 100 μl of BTXpress High Performance Electroporation Solution (BTX Harvard Apparatus). The cell suspension was supplemented with 2.5 μg of each COPA CRISPR vector, 7.5 μg of the donor vector, 3.6 μg of PuroR or NeoR PiggyBAC vector (PB510B-1, PB533-A2, System Biosciences), and 1.4 μg of the transposase vector (PB210PA-1, System Biosciences) for knock-in of COPA 6-amino acid deletion and with 5 μg of each USP9X CRISPR vector, 3.6 μg of CMV-Puro PiggyBAC vector and 1.4 μg of the transposase vector for USP9X knockout. The cell-DNA mixture was applied to a 2-mm-gap cuvette (Nepa Gene) and electroporated using a NEPA21 electroporator (Nepa Gene). The cells were collected in 500 μl BTXpress electroporation solution supplemented with 10 μM Y-27632 and kept at room temperature for 15 min. Following two washes with Advanced DMEM F/12, the cells were plated as Matrigel domes, overlaid with the full medium supplemented with 10 μM Y-27632 and 1.25% DMSO and incubated at 30 °C. The medium was replaced with DMSO-lacking medium the next day, and the organoids were transferred to the 37 °C condition 4–5 days after electroporation. Following a 2-day selection with 2 μg ml⁻¹ puromycin (Thermo Fisher Scientific) or a 1-week selection with 500 μg ml⁻¹ geneticin (Thermo Fisher Scientific), single organoids were manually picked using a micropipette and expanded for genotyping. The indel pattern in each subclone was analyzed using the primers listed in Supplementary Table 8 and the ICE online tool (https://ice.editco.bio/#/).

The vector for doxycycline-inducible expression of RNF43 tagged with HA and HiBiT (RNF43^HiBiT) was constructed by replacing the GPF cassette of the XLone vector⁵⁴ with the RNF43^HiBiT insert. The RNF43^HiBiT construct was recently described⁵⁵ and synthesized via GeneArt Custom Gene Synthesis (Thermo Fisher Scientific). Using the above electroporation method, 7.2 μg of the XLone-RNF43^HiBiT vector was co-introduced with 2.8 μg of Super PiggyBac Transposase (PB210PA-1, System Biosciences) into the Ileum1 organoid line. Transduced organoids were selected by a 1-week treatment with 5 μg ml⁻¹ blasticidin S (FUJIFILM Wako Pure Chemical Corporation). To ensure an equal copy number of the transgene across experiments, we cloned Ileum1-RNF43^HiBiT organoids and engineered one clone with confirmed expression of HA and HiBiT to generate USP9X^KO, COPA^6del and COPA^6del/USP9X^KO organoid lines. Knockout of USP9X and knock-in of COPA^6del in an Ileum1-RNF43^HiBiT clone were done by electroporation as described above.

Real-time quantitative PCR

RNA was collected from organoids using the Direct-zol RNA Miniprep Kit (Zymo Research) and was reverse-transcribed using the Omniscript RT Kit (Qiagen) following the manufacturer’s instructions. Quantitative PCR was performed using TB Green Premix Ex Taq (Tli RNase H Plus) (Takara Bio) on a LightCycler 96 device (Roche Diagnostics). Gene expression was calculated with the ΔCt method using ACTB as a reference housekeeping gene. Primers used for quantitative PCR are listed in Supplementary Table 8.

Capillary immunoassay

Immunodetection of phospho- and total LRP6, DVL2, COPA and HA was performed using Simple Western automated protein detection (ProteinSimple). The organoids were plated as single cells at a density of 3,000 cells per well and harvested at day 7. In the R-spondin1- or Noggin-removed condition, each factor was removed from the medium for 2–3 days before harvest. Otherwise, the organoids were maintained in the full medium. To harvest organoids and organoid lysate, we first dissolved Matrigel using the Cell Recovery Solution (Corning) and lysed organoids using the UltraRIPA Kit for Lipid Raft (BioDynamics Laboratory). Organoids were first lysed in 80 μl of A buffer (RIPA buffer) supplemented with Halt Protease and Phosphatase Inhibitor Cocktail (100×; Thermo Fisher Scientific) via 10 rounds of 30-second on/off sonication cycles using the Bioruptor II sonicator (Sonicbio). After centrifuging, the supernatant was collected and the remnant was further lysed in 40 μl of buffer B (UltraRIPA buffer) by a 10-min sonication. A buffer and B buffer supernatants were combined and used for analysis. Capillary immunoassay was performed on a Jess Simple Western Blot System (ProteinSimple) using the primary antibodies listed in Supplementary Table 7 with a 1:50 dilution and anti-rabbit or anti-mouse secondary antibody (ProteinSimple). The assays were run and analyzed using Compass software (version 6.3.0), and the intensity area was used as a surrogate for protein abundance.

Measurement of HiBiT-tagged RNF43

The abundance of RNF43^HiBiT protein was detected using the HiBiT system (Promega). Organoids transduced with XLone-RNF43^HiBiT were plated on 96-well plates (Corning) with a density of 1,000 cells per well and cultured under full conditions for 5 days. The medium was then replaced with the medium lacking Wnt-3A and supplemented with 5 μM CHIR99021 (Cayman Chemical) and 50 ng ml⁻¹ doxycycline hydrochloride (FUJIFILM Wako Pure Chemical Corporation). R-spondin1 was also removed in the R-spondin-lacking condition. The organoids were cultured for an additional 2 days, and the medium was refreshed 6 h before the assay. To measure total and cell-surface RNF43^HiBiT, we used Nano-Glo HiBiT Lytic and Extracellular Detection Systems (Promega), respectively. Specifically, we added 100 μl per well of Nano-Glo HiBiT Lytic or Extracellular Buffers supplemented with LgBiT protein and substrate according to the manufacturer’s instructions and measured luciferase luminescence on the GloMax plate reader. The background luminescence measured using Dox-untreated organoids was largely constant regardless of the cell number, and the mean background luminescence from three experiments (lytic: 464, extracellular: 4,421) was subtracted from the measured luminescence values. To normalize HiBiT luminescence based on cell number and viability, we measured the ATP abundance of organoids cultured in the same conditions in parallel using the CellTiter-Glo 3D Cell Viability Assay.

Statistics and reproducibility

Statistical testing of two-group or multigroup comparison was performed as described in the legends using R. No data were excluded from the analyses.

Reporting summary

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

Online content

Any methods, additional references, Nature Portfolio reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at 10.1038/s41588-026-02616-9.

Author contributions

M.F., N.A., T.S. and S.S. conceptualized and designed the research. T.S. and S.S. supervised the project. N.A., T.H., N.H. M. Kojima, Y.Y. and S.S. performed histological assessment of clinical specimens. M.F., S.K. and M.M. carried out in vitro experiments. M.F. and K.S. performed bioinformatic analyses. S.T. performed immunofluorescence imaging of organoids. G.C. and B.-K.K. provided the RNF43^HiBiT construct and critical advice on result interpretation. S.N. and M. Kato performed endoscopic assessment and contributed clinical specimens. M.F., T.S. and S.S. wrote the paper with input from all authors.

Peer review

Peer review information

Nature Genetics thanks Calvin Kuo, Douglas Winton and John Mariadason for their contribution to the peer review of this work. Peer reviewer reports are available.

Data availability

The read files from whole-exome and RNA sequencing have been deposited in the NBDC Human Database (https://humandbs.dbcls.jp/) under NBDC Research ID hum0342 with accession ID JGAS000868 (clinical tissue data) and NBDC Research ID hum0201 with accession ID JGAS000881 (organoid data) and are available under controlled access for the protection of patients’ privacy. Data users need to follow the NBDC Guidelines for Human Data Sharing (https://humandbs.biosciencedbc.jp/en/guidelines/data-sharing-guidelines) and the NBDC Security Guidelines for Human Data (for data users; https://humandbs.biosciencedbc.jp/en/guidelines/security-guidelines-for-users). Specifically, data users are required to comply with the following requirements. First, indicate that the head of the institution to which the data users belong has given permission to implement the research plan that includes the dataset the data users plan to use. Second, provide evidence that the data users have engaged in research related to the dataset that the data users plan to use. Third, indicate that the data users have implemented security measures relevant to the access level of the dataset that the data user plans to use. Fourth, obtain approval from the NBDC Human Data Review Board (details at https://humandbs.biosciencedbc.jp/en/data-use). When the application and approval processes go smoothly, the data user can gain access to the data in around 2–4 weeks. Source data are provided with this paper.

Code availability

No custom code or software was used to process or analyze the data. This study used existing packages whose details may be found in the appropriate section of Methods.

Competing interests

T.S. is an inventor on several patents related to organoid culture. The other authors declare no competing interests.

Extended data

is available for this paper at 10.1038/s41588-026-02616-9.

Supplementary information

The online version contains supplementary material available at 10.1038/s41588-026-02616-9.