Characterization of RNF43 frameshift mutations that drive Wnt ligand‐ and R‐spondin‐dependent colon cancer

Abstract Loss‐of‐function mutations in RNF43 induce activation of Wnt ligand‐dependent Wnt/β‐catenin signaling through stabilization of the Frizzled receptor, which is often found in microsatellite instability (MSI)‐type colorectal cancer (CRC) that develops from sessile serrated adenomas. However, the mechanism underlying how RNF43 mutations promote tumorigenesis remains poorly understood. In this study, we established nine human CRC‐derived organoids and found that three organoid lines carried RNF43 frameshift mutations associated with MSI‐high and BRAF V600E mutations, suggesting that these CRCs developed through the serrated pathway. RNF43 frameshift mutant organoids required both Wnt ligands and R‐spondin for proliferation, indicating that suppression of ZNRF3 and retained RNF43 function by R‐spondin are required to achieve an indispensable level of Wnt activation for tumorigenesis. However, active β‐catenin levels in RNF43‐mutant organoids were lower than those in APC two‐hit mutant CRC, suggesting a lower threshold for Wnt activation in CRC that developed through the serrated pathway. Interestingly, transplantation of RNF43‐mutant organoids with intestinal myofibroblasts accelerated the β‐catenin nuclear accumulation and proliferation of xenograft tumors, indicating a key role of stromal cells in the promotion of the malignant phenotype of RNF43‐mutant CRC cells. Sequencing of subcloned organoid cell‐expressed transcripts revealed that two organoid lines carried monoallelic RNF43 cis‐mutations, with two RNF43 frameshift mutations introduced in the same allele and the wild‐type RNF43 allele remaining, while the other organoid line carried two‐hit biallelic RNF43 trans‐mutations. These results suggest that heterozygous RNF43 frameshift mutations contribute to CRC development via the serrated pathway; however, a second‐hit RNF43 mutation may be advantageous in tumorigenesis compared with a single‐hit mutation through further activation of Wnt signaling. Finally, treatment with the PORCN inhibitor significantly suppressed RNF43‐mutant cell‐derived PDX tumor development. These results suggest a novel mechanism underlying RNF43 mutation‐associated CRC development and the therapeutic potential of Wnt ligand inhibition against RNF43‐mutant CRC. © 2022 The Authors. The Journal of Pathology published by John Wiley & Sons Ltd on behalf of The Pathological Society of Great Britain and Ireland.


Introduction
The Wnt signaling pathway plays a role in the maintenance of the stem cells of adult tissues, which is regulated by the destruction complex, in which GSK3 phosphorylates β-catenin, leading to the degradation of β-catenin by the ubiquitin-proteasome pathway [1]. Frizzled (FZD), leading to its turnover. Mouse genetic studies have shown that the simultaneous disruption of Rnf43 and Znrf3 in the intestine results in Wnt ligandindependent growth and development of intestinal tumors through Wnt signaling activation [2][3][4].
Consistent with this, genetic alterations in RNF43 are found in approximately 18% of colorectal cancer (CRC) cases [5]. RNF43 mutations were also identified in cancers in the ovary, stomach, and pancreas [6][7][8]. In hereditary serrated polyposis, germline RNF43 mutations were identified [9], and second-hit inactivation via loss of heterozygosity (LOH) was also found in tumors [10]. In addition, loss of wild-type RNF43 was also reported in ovarian cancer [6]. These results suggest a role for two-hit RNF43 mutations in the promotion of tumorigenesis.
Serrated intestinal adenomas are considered precursors of CRC [11]. Notably, RNF43 mutations in sessile adenomas and serrated pathway-type CRC were associated with the BRAF V600E mutation and CpG island methylator phenotype (CIMP), leading to the methylation of the MLH1 promoter, which further causes microsatellite instability (MSI) without DNA mismatch repair gene mutations [4,[10][11][12][13][14][15]. In addition, RNF43 mutations were detected in MSI cancers in the stomach, esophagus, and uterine endometrium [16], suggesting that RNF43 mutations are major drivers of MSI-type cancer. Furthermore, colitis-associated colon tumor development was accelerated in Rnf43 À/À mice [17], and in vivo screening using the Sleeping Beauty transposon identified Rnf43 as a candidate driver of intestinal tumorigenesis [18]. These results, taken together, indicate that an increased level of FZD receptors caused by RNF43 dysfunction drives intestinal tumorigenesis via Wnt signaling activation, particularly in MSI-CRC through the serrated pathway. Wnt is palmitoleated by porcupine O-acyltransferase (PORCN), which is required for Wnt secretion [19], and a PORCN inhibitor suppressed RNF43-mutant CRC cell growth [12,20]. Consistent with this, translational research indicates that PORCN inhibition is an effective therapeutic strategy against Wnt liganddependent cancer [19,21].
Recent studies have identified a link between RNF43 mutation types and their tumorigenic potential [22][23][24]. However, the genetic information for RNF43 mutations is still not sufficient to understand the mechanism of CRC development. Furthermore, it has been shown that 37% of Wnt ligand-dependent pancreatic cancer cells require exogenous Wnt ligands, while others utilize endogenous ones [25]. Such cell-type differences have not been well characterized in CRC.
In the present study, we established human CRC-derived organoids and found RNF43 frameshift mutations in CIMPhigh/MSI-high CRC. Importantly, the RNF43-mutant organoids were dependent on endogenous or exogenous Wnt ligands, and PORCN inhibitors significantly suppressed the organoid proliferation and xenograft tumorigenesis. Furthermore, sequencing of subcloned cell-derived transcripts suggested that heterozygous monoallelic RNF43 mutations drive the development of CRC and that second-hit mutations may be advantageous for tumorigenesis.

Establishment of patient tumor-derived organoids
Human primary CRC samples were obtained from 50 patients who underwent surgical resection at Ishikawa Prefectural Central Hospital, Japan. All specimens were used for the establishment of organoids and histological analyses. All experiments using human samples were approved by the Human Genome/Gene Analysis Research Ethics Committee of Kanazawa University (2016-086-433), and written informed consent was obtained from the patients.
Immortalized intestinal myofibroblasts (LmcMF) have been described previously [27]. LmcMF cells were cultured in 8 ml of DMEM for 48 h, and their conditioned medium was collected. Organoids were cultured with 66.7% LmcMF-CM. To label organoid cells, Venus and tdTomato cDNAs were subcloned into a pPB-CAG-IP PiggyBac transposon vector (a gift from Dr Hitoshi Niwa) and co-transfected with transposase expression vector using Lipofectamine (Thermo Fisher Scientific, Rockford, IL, USA). Transfected clones were selected by puromycin screening.

Patient-derived xenograft (PDX) mice
All animal experiments were performed using the protocol approved by the Committee on Animal Experimentation of Kanazawa University.

Histology and immunohistochemistry
Tissues were fixed in 4% paraformaldehyde, embedded in paraffin, and cut into 4-μm-thick sections. The sections were used for H&E or immunohistochemistry. The antibodies used for immunohistochemistry are described in Supplementary materials and methods. Immunostaining signals were visualized using an ImmPACT DAB Substrate Kit (Vector Laboratories, Burlingame, CA, USA).

Immunofluorescence
Cell proliferation was examined by EdU labeling. Organoids were cultured with 10 μM EdU for 1 h and fixed in 4% PBS-formaldehyde and then permeabilized with 0.1% Triton X-100. Antibodies are described in Supplementary materials and methods. Fluorescence images were obtained using a TCS SP8 confocal microscope (Leica Microsystems, Wetzlar, Germany).

Immunoblotting
Organoids were lysed in TNE buffer with Complete Mini protease inhibitor cocktail (Roche Diagnostics, Mannheim, Germany), and 10 μg of the protein samples was separated using 10% SDS-polyacrylamide gel. Antibodies for immunoblotting are described in Supplementary materials and methods. Relative band intensities were measured using ImageJ software (National Institutes of Health, Bethesda, MD, USA).

MSI analyses
Genomic DNA was extracted, and the MSI analysis was performed according to the Revised Bethesda Guidelines for two mononucleotides (BAT25 and BAT26) and two dinucleotide (D2S123 and D5S346) microsatellite markers (System Biotics, Sagamihara, Japan) [30]. If two or more of these markers showed instability, the sample was diagnosed as MSI-high.

Organoid subcloning and RNF43 genotyping
Organoids were dissociated to single cells by trypsin treatment and single cells were seeded into 96-well plates at 1 cell per well to obtain subclones. Total RNA was prepared from the subclones, and mRNA was reverse-transcribed (RT) by SuperScript III (Thermo Fisher Scientific), amplified by polymerase chain reaction (PCR), and subcloned into the plasmid vector with Mighty Cloning Reagent Set (#6027; Takara Bio). PCR primer sequences are provided in Supplementary materials and methods. Subcloned cDNAs were sequenced using an Applied Biosystems 3500 genetic analyzer (Thermo Fisher Scientific).

Genotyping of RNF43 codon 370
Genotypes of RNF43 codon 370 were examined using the TaqMan SNP Genotyping Assay (Applied Biosystems, Waltham, MA, USA). The sequences of TaqMan primers and probes for wild-type RNF43 codon 370 (VIC) and p.Rpo370fs (FAM) are provided in Supplementary materials and methods. The PCR results were analyzed by an allele discrimination/SNP system in Mx3000P (Stratagene, La Jolla, CA, USA). The custom RNF43 p.Pro370fs cDNAs were synthesized to obtain homozygous mutant controls.

Human gene expression database analyses
Human CRC expression data were downloaded from the Cancer Genome Atlas (TCGA) database (https:// www.cancer.gov/about-nci/organization/ccg/research/ structural-genomics/tcga) [31]. The expression data of CRC carrying APC-deep deletions (n = 19) or RNF43 frameshift mutations at an N-ter of codon 659 (n = 12) were extracted, and upstream regulators were examined using Ingenuity Pathway Analysis (IPA) (Ingenuity Systems, Qiagen; www.ingenuity.com). A Z-score of ≥2 and a P value of <0.05 were considered to indicate activation with statistical significance.
Evolutional RNF43 mutations in colon cancer via the serrated pathway 41

Transposon insertion site analyses
Transposon mutagenesis screening was previously performed in mice carrying sensitizing driver mutations of CRC [18]. Using the screening data, we analyzed the mutation frequency of Rnf43, Znrf3, and Apc in Trp53 R172H mouse intestinal tumors, as described in Supplementary materials and methods.

Statistical analyses
The data were analyzed using two-sided unpaired t-tests and are presented as the mean AE SD for in vitro experiments and the mean AE SEM for in vivo analyses. p < 0.05 was considered statistically significant.

Establishment of human CRC-derived organoids and PDXs
Primary CRC tissues were collected from 50 patients, and nine CRC-derived organoids were successfully established. The clinicopathological characteristics of the nine patients are indicated in supplementary material, Table S1. Histologically, all CRCs, with the exception of C31, were diagnosed as tubular adenocarcinoma; C31 was diagnosed as mucinous adenocarcinoma. Ki67 immunostaining confirmed the increased proliferation of cancer cells in all tissues ( Figure 1A). Nuclear accumulation of β-catenin was detected in C3, C15, and C46 cancer cells, suggesting transcriptional activation of Wnt/β-catenin signaling. All established organoid lines showed tubular structures ( Figure 1B). When these organoids were s.c. transplanted into SHO mice, all organoids developed tumors with continuous proliferation detected by Ki67 immunostaining ( Figure 1C). Histologically, the tubular morphology of the PDX tumors resembled the original primary tumors. Furthermore, β-catenin nuclear accumulation was detected in the C3, C15, and C46 PDX tumors, indicating that the morphology and Wnt signaling activity of the primary tumors remained in the organoids and PDX tumors.
Interestingly, the immunoblotting results showed significantly higher active β-catenin levels in CRC organoids with APC two-hit mutations (C15, C46, and C48) than in those with RNF43 frameshift mutations (C14, C24, and C45) ( Figure 2D and supplementary material, Table S2), which is consistent with the nuclear β-catenin accumulation noted in APC-mutant cancer cells ( Figure 1A). Furthermore, an IPA using a public database [29] indicated that Wnt/β-catenin pathways, such as CTNNB1, WNT1, and lithium chloride, are significantly activated in CRC with APC homozygous deletions compared with CRC with RNF43 truncation mutations (supplementary material, Table S3). These results suggest that the Wnt activation level is lower in serrated pathway CRC cells than in conventional-type CRC.

Wnt ligand and R-spondin dependency of RNF43 frameshift mutant organoids
We next examined the Wnt ligand dependency of RNF43 frameshift mutant organoids by reducing the WRN-conditioned medium (WRN-CM), which included Wnt3a, R-spondin, and Noggin. As expected, the number of proliferating cells was significantly decreased in C14, C24, and C45 organoids in the absence of WRN-CM (Noggin was added to the medium), while other organoids continued proliferating 42 D Yamamoto et al ( Figure 3A). Consistent with this, the EdU labeling efficiency was dramatically decreased in RNF43 frameshift mutant organoids in the WRN-0% + Noggin condition (supplementary material, Figure S2). Supplementation of WRN-0% cultures with both recombinant Wnt3a and R-spondin rescued the proliferation of C14, C24, and C45 organoids to a similar degree as culture with WRN-50% ( Figure 3B,C). Notably, R-spondin alone also rescued the C14 and C24 organoid proliferation in WRN-0% + Noggin, while Wnt3a alone did not. Thus, C14 and C24 cells required exogenous R-spondin but utilized endogenously expressed Wnt ligands ( Figure 3D). In contrast, neither Wnt ligand nor R-spondin rescued C45 cell proliferation, indicating that C45 cells required both exogenous Wnt3a and R-spondin.
To confirm the Wnt ligand dependency, we treated organoids with the PORCN inhibitors C59 and IWP2 in the presence of R-spondin under WRN-0% + Noggin conditions. As expected, both PORCN inhibitors significantly suppressed C14 and C24 cell growth, confirming the need for endogenously expressed Wnt ligand ( Figure 3E). In this assay, C45 cells were unable to proliferate because of the absence of exogenous Wnt ligand.
Proliferation of RNF43-mutant cells with support from the microenvironment Lung cancer cells have two distinct subpopulations: one forms a niche providing a Wnt ligand to the other, activating Wnt signaling [35]. It is thus possible that Wnt Evolutional RNF43 mutations in colon cancer via the serrated pathway 43 ligand-dependent CRC cells utilize the Wnt ligands secreted by other subpopulations. To assess this possibility, C14 and C45 organoid cells were differentially labeled with tdTomato and Venus, respectively, and cultured separately or mixed. We confirmed that R-spondin rescued the proliferation of C14, but not C45, cells under WRN-0% + Noggin conditions ( Figure 4A). However, C45 cells proliferated under the same culture conditions when co-cultured with C14 organoids, suggesting that C14-secreting Wnt ligands activate the Wnt signaling of C45 cells to promote their proliferation.
Stromal cells are also an important source of Wnt ligands and R-spondin [36]. We therefore cultured C45 organoids in the presence of CM from intestinal myofibroblasts (LmcMF cells) [27] and found that the C45 cells proliferated under WRN-0% + Noggin conditions, suggesting that LmcMF cells provided sufficient Wnt ligand and R-spondin to C45 cells ( Figure 4B,C). We next co-transplanted C45 organoids and LmcMF cells into SHO mice s.c. (Figure 4D). Histological analyses indicated that a fibroblast-rich microenvironment was generated by tdTomato-labeled LmcMF cells in

D Yamamoto et al
co-transplanted tumors ( Figure 4D). Ki67-positive proliferating cells were found only in the marginal zone of C45-alone tumors ( Figure 4D, arrowheads), while proliferation was found in the whole tumor area of cotransplanted tumors. Furthermore, clear nuclear accumulation of β-catenin was detected in C45 tumor cells when co-transplanted with LmcMF cells, which was rarely found in C45-alone tumors. These results suggest that cancer-associated fibroblasts contribute to malignant phenotypes of RNF43-mutant CRC by expressing Wnt ligands and R-spondin.
To examine whether these mutations were heterozygous (monoallelic) or biallelici.e. cis or trans mutationswe performed subcloning of the parental organoid lines and examined the mutation types in subclones via the amplification of RNF43 mRNA by RT-PCR followed by cDNA sequencing ( Figure 5A). PCR primers were designed to amplify cDNA fragments, including two mutation sites at both ends ( Figure 5B). cDNA sequencing indicated a 1-bp deletion in (C) 5 for p.Arg225fs and p.Pro370fs and in (G) 7 for p.Gly659fs, resulting in amino acid changes ( Figure 5C). Before sequencing, we confirmed the Wnt ligand and R-spondin dependency of the organoid subclones (supplementary material, Figure S3A). Furthermore, AXIN2 and ZNRF3 expression was significantly decreased in C14-, C24-, and C45-derived subclones compared with APC-mutated conventional-type CRC organoids (supplementary material, Figure S3B,C), which was a characteristic of CIMP-high/MSI-high serrated CRC [12,22,37].
Importantly, allele-specific genomic PCR indicated that C14 and C24 subclones possessed both RNF43 codon 370 wild-type and mutant genes with the same copy number, while conventional pathway-type CRC organoids had only wild-type RNF43 ( Figure 5F). These results exclude the possibility of loss of wild-type RNF43 by LOH in C14 and C24 cells. Interestingly, the mutant allele frequency increased to approximately 90% in C14-1 and C24-1 subclones. The wild-type RNF43 gene may have been lost by LOH during the subcloning process, allowing LOH cells to become dominant because of a possible growth advantage.
We further examined the data previously obtained by transposon mutagenesis screening in mouse intestine with the Trp53 R172H genetic background [18]. The frequencies of Rnf43, Znrf3, and Apc mutations by transposon insertions in intestinal tumors are indicated in Figure 5G (top). Single insertions in Rnf43 were found in 14.5% of tumors, and two cases did not carry mutations in Apc and Znrf3. Most transposon insertions were found in introns 2 and 3, resulting in truncation at the N-terminal of RNF43 ( Figure 5G, bottom). These results support the idea that a single-hit RNF43 mutation contributes to intestinal tumorigenesis.

Suppression of RNF43-mutant PDX tumors by a PORCN inhibitor
Finally, we examined whether or not a PORCN inhibitor suppresses the tumor development of RNF43 frameshift mutant cells in PDX models. C14, C24, and C45 parental and subclone organoids were transplanted s.c. into SHO mice, and mice were treated with the PORCN inhibitor ETC-159 ( Figure 6A). Importantly, ETC-159 treatment nearly completely suppressed the tumor development in all PDX model mice ( Figure 6B,C and supplementary material, Figure S4A). Histological analyses indicated that ETC-159 treatment resulted in decreased Evolutional RNF43 mutations in colon cancer via the serrated pathway 45 (E) Relative cell proliferation, examined by the luciferase activity of the respective organoids cultured in the presence or absence of PORCN inhibitors (C59 or IWP2), is indicated by line graphs (mean AE SD). A two-sided t-test was used to calculate statistical significance. Asterisks (for R-spo + C59) and daggers (for R-spo + IWP2), p < 0.05 versus R-spondin only (no inhibitor) control (blue lines).

D Yamamoto et al
branching and enlarged glands with fewer dysplastic features and significantly decreased proliferation ( Figure 6D and supplementary material, Figure S4B).
It is possible that cancer cells were differentiated by PORCN inhibition-induced suppression of Wnt signaling, as previously reported [21]. Evolutional RNF43 mutations in colon cancer via the serrated pathway 47  Evolutional RNF43 mutations in colon cancer via the serrated pathway 49

Discussion
Serrated polyps with MSI are reportedly associated with a predisposition to progress to CRC [10,11]. Moreover, RNF43 mutations were frequently found in sessile serrated lesions (SSLs) and MSI-type CRC, suggesting that RNF43 mutation drives CRC development through the serrated pathway ( Figure 6E). Consistently, we found RNF43 frameshift mutations in CIMP-high and MSIhigh CRC-derived organoids, which were associated with BRAF V600E mutations. Interestingly, the level of active β-catenin in RNF43mutant CRC was significantly lower than that in APC two-hit CRC, and nuclear accumulation of β-catenin was found only in APC-mutant CRC. These results suggest that the threshold of Wnt activation for serrated pathway tumorigenesis is lower than that via the conventional pathway. Consistently, it has been reported that Wnt activation is greatly reduced in MSI-high hypermutated CRCs relative to non-hypermutated CRCs [37]. It is possible that a high level of Wnt activation is needed for the expansion of intestinal stem/progenitor cells to form adenomatous polyps as an initial event of the conventional pathway, as found in Apc Δ716 knockout mice [38]. Although adenomatous proliferation was found in Rnf43 À/À Znrf3 À/À compound mouse intestine [2,3], it was not observed in Rnf43 À/À simple mutant mice [17]. These results suggest that the degree of Wnt activation induced by RNF43 mutation is not sufficient for the initiation of conventional-type CRC but can contribute to serrated pathway tumorigenesis.
In the present study, all RNF43 frameshift mutant CRC cells proliferated in an R-spondin-dependent manner. It is possible that suppression of ZNRF3 function and retention of RNF43 function by R-spondin is required to achieve the level of Wnt signaling necessary for tumorigenesis in RNF43-mutant cells.
Loss of wild-type RNF43 by LOH or second-hit somatic mutations has been reported in ovarian tumors and serrated colon tumors [6,10], suggesting the need for two-hit inactivation of RNF43 for tumorigenesis. However, the present results showed heterozygous monoallelic RNF43 mutations in C14 and C24 cells, indicating that biallelic inactivation of RNF43 is not necessarily required for serrated pathway tumor development. In contrast, we also found two-hit biallelic trans-mutations in RNF43 in C45 cells, suggesting that two-hit mutations of RNF43 are advantageous for CRC development, possibly through an increase in the Wnt activation level ( Figure 6E). Furthermore, cis mutations in RNF43 for p.Pro370fs and p.Gly659fs were found in the same allele. A recent analysis indicated that the most common mutation of RNF43, p.Gly659fs, modestly compromises the RNF43 function [24], and N-terminal truncation mutations are more effective at increasing Wnt signaling activity than C-terminal mutations [23]. Therefore, the p.Gly659fs mutation may have been introduced at an early stage of tumorigenesis in C14 and C24 with a slight increase in Wnt signaling. An additional N-terminal p.Pro370fs mutation was then introduced at the same allele, which may have facilitated tumor promotion by increasing the Wnt activity ( Figure 6E). Importantly, co-transplantation of RNF43 frameshift mutant CRC cells with myofibroblasts increased the β-catenin nuclear accumulation and cell proliferation of PDX tumors, indicating that the malignant phenotype of RNF43-mutant cells is highly dependent on stromal cells. Several studies have demonstrated the role of Wnt signaling in metastasis; namely, Wnt activation promotes the epithelial-mesenchymal transition program by increasing the nuclear Smai2 expression [39], and liver metastasis of colon cancer cells is accelerated by the nuclear accumulation of β-catenin and FOXO3 [40]. Accordingly, it is possible that metastatic niche cells play a key role in the proliferation of disseminated RNF43-mutant CRC cells through activation of Wnt signaling. In this study, we demonstrated the significant suppression of RNF43-mutant PDX tumors by ETC-159 treatment. Thus, it is important to further examine whether or not the development of metastatic foci by RNF43-mutant cells can be suppressed by PORCN inhibitors as a future clinical strategy.
In conclusion, we established RNF43 frameshift CRC-derived organoids. The Wnt/β-catenin activation level of RNF43-mutant serrated pathway CRC was lower than that of conventional pathway-type CRC. Furthermore, heterozygous RNF43 frameshift mutation may contribute to tumorigenesis through the serrated pathway. However, second-hit RNF43 mutations may facilitate CRC development by increasing the Wnt activation level. Finally, PORCN inhibition significantly suppressed PDX tumor development of RNF43-mutant CRC cells, suggesting that PORCN inhibitors are an effective therapeutic strategy against RNF43-mutant CRC development and metastasis.
analysis. HaT examined the transposon data. TO provided cancer-associated fibroblasts. HiT, NI and MO supervised the project. All of the authors were involved in writing the paper and had final approval of the submitted version.

SUPPLEMENTARY MATERIAL ONLINE
Supplementary materials and methods Figure S1. The results of CIMP analysis for RNF43 frameshift mutant organoid cells Figure S2. Wnt ligand and R-spondin dependency of RNF43 frameshift mutant organoid cells Figure S3. Biological characteristics of RNF43 frameshift mutant CRC cells Figure S4. Suppression of RNF43 frameshift mutant tumor development by PORCN inhibitor Table S1. Clinicopathological characteristics of CRC patients Table S2. Mutation variants and characteristics of CRC organoids Table S3. Activated upstream regulators in APC-deleted CRCs compared with RNF43 truncation mutation CRCs (IPA)

52
D Yamamoto et al