Gain-of-Function p53 Mutation Acts as a Genetic Switch for TGFβ Signaling–Induced Epithelial-to-Mesenchymal Transition in Intestinal Tumors

TGFβ signaling plays a tumor-suppressive role, while it also functions as a tumor promoter; thus, TGFβ has been considered a double-edged sword for cancer development (1). TGFβ and its family member activin bind their type II receptors, resulting in type I receptor activation, which further phosphorylates Smad2/3. Phospho-Smad2/3 then binds Smad4 to form a complex and translocates into nucleus, inducing the expression of TGFβ target genes (2, 3). The induction of cyclin-dependent kinase inhibitors, including p21 and p27, is one of the major mechanisms of tumor suppression by TGFβ family signaling (1, 4–7). Consistently, a genome-wide analysis showed frequent mutations in the TGFβ type II receptor gene (TGFBR2), activin receptor IIA (ACVR2A), and SMAD4 in human colorectal cancer (8). Moreover, we previously showed that the disruption of Smad4 or Tgfbr2 in the Apc^Δ716 mouse model caused the submucosal invasion of intestinal polyps (9, 10). Taken together, these results indicate that TGFβ and activin signaling play a role in tumor suppression.

In contrast, TGFβ and activin promote the malignant progression of cancers by inducing epithelial–mesenchymal transition (EMT; refs. 11–13). Phosphorylated Smad2/3 and the Smad4 complex induce the expression of EMT–transcription factors (EMT-TF), such as Snail and Twist. Recently, it has become apparent that malignant cancer cells rarely reach a fully mesenchymal state through EMT, instead usually proceeding to a partially mesenchymal and partially epithelial state, which is called partial EMT (14). Such a partial EMT state is thought to be critical for malignant phenotypes, such as invasion and metastasis. Furthermore, the concept of a polyclonal metastasis mechanism has been proposed, in which cancer cell clusters break off from primary tumors and move toward distant organs as clusters that develop into metastatic tumors (15–17). Such cluster migration from the primary site is possibly promoted by partial EMT.

It has been shown that Ras activation induces RREB1 phosphorylation, leading to collaboration with the Smad complex to induce EMT-TFs, indicating a role of RAS activation in TGFβ-induced EMT (13, 18). Moreover, TGFβ signaling itself promotes EMT by inducing the expression of HMGA2, which leads to the induction of EMT-TFs (19–21). Furthermore, it has been shown that TGFβ signaling generates two mesenchymal populations in collaboration with HMGA2 or KRTAP2–3, which orchestrates cancer cell proliferation and EMT-induced migration (22). However, it has not yet been clearly understood how cancer cells are allowed to survive and their partial EMTs are induced in the presence of TGFβ or activin signaling.

We have established mouse intestinal tumor models and organoids carrying Apc^Δ716, Kras^G12D, Tgfbr2^−/−, Trp53^R270H, and Fbxw7^−/− mutations in various combinations and showed a link between mutation combinations and malignant phenotypes (10, 23, 24). In this study, we examined the relationship between genetic alterations and activin A–induced phenotypes using the established organoids. Notably, Kras activation mutations protected organoids from activin A–induced growth suppression. Furthermore, a gain-of-function (GOF) Trp53 mutation together with loss of wild-type Trp53 by loss of heterozygosity (LOH) was found to be a key genetic event for activin A–induced partial EMT. Accordingly, these results suggest that the inhibition of activin A signaling is an effective therapeutic strategy against the malignant progression of colorectal cancer when cells carry both Kras and p53 GOF/LOH mutations.

The intestinal tumor-derived organoids used in this study have been previously established (24). In brief, the organoids were established from intestinal tumors of compound mutant mouse models (16 weeks of age), Apc^tm1Mmt (Apc^Δ716; A), B6.129-Kras^tm4Tyj/Nci (Kras^G12D; K), B6.129S6-Tgfbr2^tm1Hlm/Nci (Tgfbr2^−/−; T), 129S4-Trp53^tm3Tyj/Nci (Trp53^R270H; P), and Fbxw7^tm1Kei (Fbxw7^−/−; F) in various combinations. Each capital letter in the organoid name (A, K, T, P, or F) indicates a mutation introduced in the organoid. In this study, Trp53 genotypes are described as M/LOH, M/+, or -/- for the R270H mutation with loss of wild-type Trp53 by LOH, heterozygous R270H/+ mutation, and null mutation, respectively. Namely, AKTP^M/+ organoids carry the Trp53^R270H/+ heterozygous mutation, whereas AKTP^M/LOH organoids carry the Trp53^R270H mutation with LOH.

Aliquots of organoid cells were used within 3 months of thawing. The organoids were cultured in Growth Factor Reduced Matrigel (Corning) with Advanced DMEM/F-12 medium (Gibco) supplemented with 50 ng/mL mEGF (Wako), 10 mmol/L HEPES, 2 mmol/L Glutamax (Gibco), 1×B27, 1×N2 (Invitrogen, Carlsbad), 100 ng/mL murine Noggin (PeproTech) and 1 μmol/L N-acetylcysteine (Sigma). Cells were monitored monthly for Mycoplasma contamination, and cell lines were authenticated using STR analysis (Labcorp). The organoid cell lines are available upon request.

For the activin A treatment experiments, organoids were collected from Matrigel using Cell Recovery Solution (Corning), mechanically dissociated by pipetting, and seeded at 1.5 × 10⁵ cells into collagen gel (Cellmatrix type I-A; Nitta Gelatin). Noggin was removed from the medium, and 20 ng/mL activin A (R&D, #338-AC) or 10 ng/mL TGFβ (R&D Systems, #240-B) was added 12 hours after seeding. The number of organoids was counted at 0 and 48 hours after activin or TGFβ treatment, and the relative efficiency of organoid development to the no-treatment control was calculated. The size of the organoids (Φ >60 μm) was measured at 48 hours after activin treatment, and the mean size relative to the no-treatment control was calculated (n = 3 wells for each genotype).

For the inhibitor treatment experiments, organoids were treated with 3 nmol/L or 5 nmol/L trametinib (ChemScene, #CS-0060) or 0.5 μmol/L or 1 μmol/L selumetinib (Selleck Chemicals, #S1008) for 48 hours, and then the number and size of organoids were examined.

Photographs of organoids were taken under a dissecting microscope at 48 hours after activin A treatment, and the efficiency of morphological change was calculated by counting the number of organoids with protrusions. The ratios of protrusion areas per organoid were also calculated. Morphologic examinations were performed by blind testing without genotype notification. Two independently established organoid lines for ATP^M/LOH, AKP^M/LOH, AKTP^M/LOH, AKTFP^M/LOH and AKTP^−/− were used for the analyses.

Organoids in the collagen gel were fixed in 4% paraformaldehyde and embedded in paraffin, and 4 μm–thick sections were prepared for hematoxylin and eosin (H&E) staining for histologic analysis. Antibody against E-cadherin (R&D Systems, #AF748) was used as the primary antibody, and Alexa Fluor 594–conjugated antibody (Thermo Fisher Scientific, #A21207) was used as the secondary antibody. DAPI was used for nuclear staining.

Organoid cell proliferation was examined using 5-ethynyl-2′-deoxyuridine (EdU) labeling. Organoids were cultured with 10 μmol/L EdU (Thermo Fisher Scientific) for 90 minutes and fixed in 4% paraformaldehyde. Organoids were then stained using a Click-iT EdU Alexa Fluor 488 Imaging Kit (Molecular Probes). DAPI was used for nuclear staining. The number of EdU-positive cells was measured in three independent fields, and the EdU index was calculated. Cell growth rate was examined using the CellTiter Glo Cell Viability Assay (Promega).

For the cell invasion analysis, 25 μL of collagen gel (Cellmatrix Type I-A, Nitta Gelatin) or 30 μL of Matrigel was poured into FluoroBlok inserts with 8-μm pores (Corning). A total of 3 × 10⁵ organoid cells in collagen gel or 2 × 10⁴ SW620 cells (ATCC, CCL-227) were then seeded onto the bottom gel and incubated. The cells were cultured in the presence of activin A (20 ng/mL) for 5 days or TGFβ inhibitor A83–01 (500 nmol/L, TOCRIS Bioscience, #2939) for 3 days. The cells that invaded through the pores of the inserts were stained with calcein AM (eBioscience), and the number of invading cells was counted (n = 5–6 wells for each experiment). Aliquots of SW620 cells were thawed within 3 months. The cells were monitored monthly for Mycoplasma contamination and authenticated using STR analysis (Labcorp).

The pSpCas9 (BB)-2A-Puro (PX459) plasmid (Addgene, #62988) was used to package guide oligonucleotides targeting Acvr2a and Hmga2 in the organoids. The p53 CRISPR/Cas9 KO plasmid (m2 or h; Santa Cruz Biotechnology) was used for Trp53/TP53 gene knockout in organoids and SW620 cells, respectively. Briefly, cells were transfected with the constructs and selected with 1 μg/mL puromycin for 3 days or based on the GFP expression. The available single-guide RNA (sgRNA) sequences are provided in Supplementary Data.

Wild-type (wt)-Trp53 and GOF mutant Trp53^R270H cDNAs were amplified and subcloned from wild-type C57BL/6NCrl (Jackson Laboratory) and 129S4-Trp53^tm3Tyj/Nci mouse liver cDNAs, respectively. These were subcloned into the pPB-CAG-IB PiggyBac transposon expression vector (a gift from Dr. Hitoshi Niwa) and cotransfected with the transposase expression vector using Lipofectamine LTX (Thermo Fisher Scientific). Two cell lines were established for each expression vector by independent transfection experiments.

Organoid cells were lysed in TNE buffer with complete Mini protease inhibitor cocktail (Roche), and 10 μg of the protein sample was separated using a 10%–15% SDS-polyacrylamide gel. Antibodies against phosphorylated-p44/42 Erk1/2 (Cell Signaling Technology, #9101), p53 (1C12; Cell Signaling Technology, #2524), HMGA2 (Cell Signaling Technology, #8197S), and HMGA1 (Abcam, #ab129153) were used as primary antibodies. Anti-GAPDH antibody (FUJIFILM Wako, #016–25523, RRID: AB_2814991) was used as the control. An ECL detection system (GE Healthcare) was used to detect signals.

Female immunodeficient mice, NOD.Cg-Prkdc^scidIl2rg^tm1Wjl/SzJ (NSG mice) were purchased at 6 weeks old (Jackson Laboratory), and transplantation experiments were performed after randomization. To analyze lung metastasis, organoid cells were labeled with Venus using the pPB-CAG-IB PiggyBac transposon expression vector. Organoids were then pretreated with 20 ng/mL activin A for 48 hours and dissociated into single cells using trypsin, and 2.5 × 10⁵ cells in 100 μL PBS were injected i.v. into the tail vein. No pretreated cells were used as the control. At 1, 24, and 48 hours after injection, lung tissues were examined under a fluorescence dissecting microscope, and the number of Venus-expressing cell clusters was counted as individual metastasis lesions or micrometastasis lesions. Tissues were fixed in 4% paraformaldehyde for histologic analysis.

All animal experiments were conducted according to the protocol approved by the institutional review board of Committee on Animal Experimentation of Kanazawa University.

Lung tissue specimens were embedded in paraffin, and 4 μm–thick sections were prepared and stained with H&E. The number and size of tumor foci on the histology sections were measured in all lung lobules of each mouse. To examine cell proliferation, IHC was performed using an antibody against Ki-67 (SP6; Abcam, # 16667). Staining signals were visualized using the Vectastain Elite Kit (Vector Laboratories). The mean Ki-67-labeling index was calculated by counting Ki-67–positive cells in 15–20 independent lesions (n = 6 for each experimental condition).

Total RNA was extracted from organoids using ISOGEN (Nippon Gene, Toyama), reverse-transcribed using the Prime Script RT Reagent Kit (Takara Bio), and amplified using ExTaqII SYBR Premix (Takara Bio) on an Mx3000P real-time thermocycler (Agilent Technologies). Relative mRNA levels of p21 (Cdkn1a), p27 (Cdkn1b), Twist1, Snai1, Ccna2, Ccnd1 to Gapdh were calculated. The primer sequences are provided in the Supplementary Data.

Total RNA was extracted from activin A-treated or untreated AKTFP^M/+ and AKTFP^M/LOH organoids using an RNeasy plus Micro Kit (Qiagen), and RNA-seq libraries were prepared using a SureSelect Strand Specific RNA Reagent Kit (Agilent Technologies). Single-end sequencing (36 bp) was performed using an Illumina NovaSeq6000 (Illumina). Sequencing data were deposited in the DNA Data Bank of Japan (DDBJ; bioproject, PRJDB15570). The quality of raw reads was assessed using FastQC (RRID: SCR_014583; http://www.bioinformatics.babraham.ac.uk/projects/fastqc/). Clean reads with quality scores of >Q30 were processed using the Tuxedo protocol with TopHat2 (RRID: SCR_013035) and Cufflinks (RRID: SCR_014597; refs. 25–27). The reads were aligned to the mouse reference GRCm38.p4/mm10 using Tophat2 with the default parameters.

After sequence alignment, the duplication rate of the reads was assessed by sequence- and mapping-based approaches using RSeQC (28). Gene expression quantification was performed using the Cufflinks. Reads per kilobase of transcript per million mapped reads (RPKM) were calculated as the expression. A principal component analysis (PCA) and clustering analysis were performed using iDEP v0.96 (http://bioinformatics.sdstate.edu/idep/) with fold change values of >2.

The differentially expressed genes were identified by Cuffdiff with the cutoff set at adjusted P < 0.05 and > absolute 1.5-fold change. P values were adjusted for multiple testing using Benjamini–Hochberg method. Significantly upregulated genes (>2 fold) in activin A–treated AKTFP^M/LOH cells compared with control AKTFP^M/LOH and activin A–treated AKTFP^M/+ cells were extracted. The selected genes were analyzed using the Ingenuity Pathway Analysis (IPA) software package (Ingenuity Pathway Analysis; RRID: SCR_008653). Upstream pathways with z-scores > 3 (P < 0.05) and z-scores < −2.5 (P < 0.05) were considered significantly activated and inactivated, respectively.

Using the dataset, we examined the frequency of distant metastasis (M1) and lymph node metastasis (N1-N3) in the TP53 wild-type, TP53 heterozygous missense mutation (diploid), and TP53 missense mutation with LOH (heterozygous deletion) colorectal cancer samples.

Data are presented as the mean ± SD. Statistical analyses were performed using two-sided unpaired t tests or a one-way ANOVA, followed by Tukey post hoc test. P < 0.05 were considered as statistically significant.

The raw mRNA sequencing data were deposited in the DDBJ with bioproject, PRJDB15570 (https://ddbj.nig.ac.jp/resource/bioproject/PRJDB15570). Human colorectal cancer data were analyzed using the TCGA dataset, including COAD and READ (https://portal.gdc.cancer.gov/). All other raw data are available upon request from the corresponding author.

It has been reported that serum activin A levels are significantly increased in patients with colon cancer, particularly in those with metastatic disease (5, 29). Using the Cancer Genome Atlas database (8), we confirmed that the INHBA encoding inhibin βA, a subunit of activin A, was significantly upregulated in colorectal cancer tissues (Supplementary Fig. S1A). An analysis using another database indicated that the expression of INHBA in the metastasis-associated primary colorectal cancer was significantly higher than that in primary colorectal cancer without metastasis (Supplementary Fig. S1B). Moreover, increased INHBA was associated with decreased overall survival and disease-free survival of patients with colorectal cancer (Supplementary Fig. S1C and S1D). These results suggest that activin A plays a role in the malignant progression of colorectal cancer.

Mouse intestinal tumor-derived organoids carrying mutations in Apc^Δ716 (A), Kras^G12D (K), Tgfbr2^−/− (T), Trp53^R270H (P), or Fbxw7^−/− (F) in various combinations showed multistep tumor phenotypes according to the accumulation of mutations (Supplementary Fig. S2; refs. 10, 23, 24). We treated these organoids with activin A to examine the link between genotypes and phenotypes upon activin stimulation (Fig. 1A). Activin A treatment significantly decreased the number and size of Kras wild-type organoids to the level of normal crypts (Fig. 1B; Supplementary Fig. S3). In contrast, organoids carrying the Kras^G12D mutation developed similar numbers and sizes of organoids in the presence of activin A to those of the no-treatment control. These results suggest that Kras activation protects the organoids from activin-induced growth suppression.

We confirmed that Kras^G12D-mutant Tgfbr2 wild-type organoids were also resistant to TGFβ-induced growth suppression (Supplementary Fig. S4). Consistently, the EdU labeling indices of Kras wild-type organoids were significantly decreased by activin A treatment, whereas they were not changed in AK organoids that carried the Kras^G12D mutation (Fig. 1C). Moreover, the growth of AK organoids was suppressed when treated with activin A together with the MEK inhibitor trametinib or selumetinib (Fig. 1D). Similar results were obtained when AKTFP^M/LOH organoids were treated with activin A and MEK inhibitor (Supplementary Fig. S5A and S5B).

It has been shown that activin signaling upregulates the cell-cycle inhibitors p21 and p27 (7, 30). We confirmed that activin A treatment significantly increased the expression of Cdkn1a and Cdkn1b, encoding p21 and p27, respectively, in Kras wild-type organoids; however, no such increase was seen in Kras^G12D organoids (Fig. 1E). Taken together, these results indicate that activation of the Kras/MEK pathway protects tumor cells from activin A–induced growth suppression, possibly through inhibition of the expression of cell-cycle inhibitors.

Given previous findings that activin A signaling induces EMT (5, 31, 32), we examined the morphologic changes in activin A-treated organoids. Notably, the organoid shape dramatically changed from a round cystic structure to a complicated form with multiple protrusions on the surface in the AKP^M/LOH, AKTP^M/LOH and AKTFP^M/LOH organoids (Fig. 2A and B; Supplementary Fig. S3). Such morphologic changes were not found in AK, AKT, or AKTP^−/− organoids, indicating that protrusions do not depend on Kras mutations. We confirmed that activin A induced protrusion of the AKTFP^M/LOH organoid surface by time-lapse analysis (Supplementary Videos S1). Such morphologic changes were significantly suppressed by disruption of the activin receptor gene Acvr2a in AKTFP^M/LOH organoids (Fig. 2C; Supplementary Fig. S6). Histologically, tumor cells at the protrusions showed the loss of the apical–basal cell polarity and glandular structures, although they still expressed E-cadherin (Fig. 2D and E). These results indicated that activin A induces partial EMT in these organoids. Notably, such morphologic changes were associated with the increased invasive ability of AKTFP^M/LOH cells, which was not found in AKTFP^M/+ cells (Fig. 2F). These results suggested the involvement of Trp53 mutation status in partial EMT phenotypes.

AP^M/LOH and ATP^M/LOH organoids also showed partial EMT phenotypes after activin A treatment, although organoid growth was significantly suppressed (Figs. 1B, 2A and B). Accordingly, it is possible that activin A–induced growth suppression and partial EMT were independently regulated. Importantly, all organoids carrying Trp53 GOF mutations with loss of wild-type Trp53 by LOH showed partial EMT in the presence of activin A (Fig. 2A, red), while organoids carrying Trp53 GOF mutations without LOH or Trp53 null mutation did not (Fig. 2A, orange and blue, respectively). We and other groups have previously indicated that the loss of wild-type Trp53/TP53 by LOH is a prerequisite for the efficient stabilization of mutant p53 (33, 34). Accordingly, the results suggest that stabilized mutant p53 plays an important role in activin A–induced partial EMT.

To assess this possibility, we constructed mutant Trp53^R270H-expressing AKTP^−/− organoid cells (Fig. 3A). Notably, overexpression of Trp53^R270H resulted in a significant increase in protrusion formation upon activin A treatment, which was rarely found in the parental control AKTP^−/− organoids (Fig. 3B and C). Next, we constructed wild-type Trp53-expressing AKTFP^M/LOH cells to expect decrease of nuclear mutant p53 levels, as previously described (Fig. 3D; ref. 23). Interestingly, the wild-type Trp53 expression significantly suppressed activin A–induced morphologic changes compared with the control AKTFP^M/LOH organoids (Fig. 3E and F). These results indicate that the Trp53 GOF mutation together with Trp53 LOH (hereafter, Trp53 GOF/LOH) is an important genotype for activin A–induced partial EMT, possibly through the increased stabilization of nuclear mutant p53.

Intriguingly, AF and AKF organoids showed partial EMT at low levels upon activin A treatment, although they did not harbor Trp53 mutations (Fig. 2A). While further study is needed, it is possible that the Trp53 GOF/LOH state and loss of the Fbxw7 function share an EMT induction mechanism.

Next, we examined the interaction between the TP53 GOF mutation and TGFβ signaling in human colorectal cancer cells. The TGFβ inhibitor significantly suppressed the invasion phenotype of SW620 cells carrying TP53^R273H mutation (Fig. 3G–I). Notably, disruption of the TP53 gene in SW620 cells also suppressed their invasion ability to a level similar to that of TGFβ inhibitor treatment. These results suggest that GOF mutant p53 plays a role in TGFβ-induced invasion, possibly through partial EMT induction in human colorectal cancer.

To examine whether activin A–induced partial EMT promotes metastasis, we pretreated AKTFP^M/LOH organoids with activin A and injected cells into the tail vein intravenously to examine lung metastasis (Fig. 4A). Importantly, the number and size of lung metastases increased significantly in activin A–pretreated organoid cells (Fig. 4B and C). In contrast, cell proliferation in the metastatic foci examined by Ki-67 immunostaining was similar in activin A-pretreated and no-treated control AKTFP^M/LOH cells (Fig. 4D). These results suggest that activin A treatment promotes the colonization efficiency of disseminated cells rather than their proliferation.

Next, we examined metastatic lesions at an early stage after intravenously injection of Venus-expressing AKTFP^M/LOH organoids. Under a fluorescence dissection microscope, numerous cell clusters were observed in the lungs of both control and activin A-pretreated organoid-injected mice at 1 hour after injection, and approximately 90% of cell clusters disappeared within 24 hours, possibly through blood stream perfusion (Fig. 4E and F). Importantly, the number of metastatic colonies at 24 and 48 hours after injection was significantly higher in activin A-pretreated cells than no-treated control. These results suggest that activin A increases the survival and colonization of cells after dissemination in the lung. The number and size of metastases were also increased in AKTP^M/LOH cells pretreated with activin A (Fig. 4G–I). In contrast, metastasis was not increased in AKTP^−/− (Trp53 null mutant) organoid cells after treatment with activin A. Taken together, these findings suggest that activin A-induced partial EMT is important for increased Trp53 GOF/LOH cell metastasis in the lung.

We next examined the relationship between TP53 mutations and metastasis frequency in human colorectal cancer by TCGA database analysis. Metastasis frequency was higher in TP53 missense-type mutant tumors than in TP53 wild-type tumors (Supplementary Fig. S7A and S7B). Moreover, the frequency of lymph node metastasis (N1-N3) was the highest in TP53 LOH tumors. Among the several TP53 hotspot mutations, R273/LOH tumors showed a higher frequency in both distant (M1) and lymph node metastasis (N1-N3) than R273/Het tumors (Supplementary Fig. S7C and S7D). These results suggest that the TP53 mutation type is important for promotion of metastasis through partial EMT, although further investigation is required.

To examine the mechanism by which the cooperation between the Trp53 GOF/LOH mutation and activin A signaling promotes malignant progression, we performed RNA-seq of AKTFP^M/+ and AKTFP^M/LOH organoid cells in the presence or absence of activin A stimulation. A PCA showed that the expression profile of the control AKTFP^M/+ organoids was significantly changed toward distinct directions by Trp53 LOH or activin A treatment (Fig. 5A). Moreover, the expression profile of activin A-treated AKTFP^M/LOH organoids was distinct from that of control AKTFP^M/LOH and activin A-treated AKTFP^M/+ organoids. These results indicate that Trp53 mutation and activin A treatment induce different sets of transcriptomic changes.

Using the sequence data, we extracted differentially expressed genes that were significantly upregulated >2-fold in AKTFP^M/LOH organoids by activin A treatment (genes induced by activin A) or Trp53 LOH (genes induced by Trp53 GOF/LOH mutation; Fig. 5B; Supplementary Table S1). By comparison of these gene sets, 115 genes were identified as upregulated by both the Trp53 GOF/LOH mutation and activin A signaling, possibly including genes responsible for the induction of partial EMT.

Using the selected gene set, we next examined the significantly activated and inactivated upstream regulators by an IPA (Fig. 5C). Notably, TGFβ, Wnt/β-catenin (CTNNB1) signaling, inflammatory pathways, and growth factor signaling were activated, whereas the inhibitor treatment–associated pathways for PI3K, JNK, MAPK, and MEK were significantly inactivated (Fig. 5C). These results indicate that the cooperation between the Trp53 GOF/LOH mutation and activin A stimulation activates growth factor signaling and the inflammatory pathway, which may contribute to the malignant progression of intestinal tumors. Consistently, a disease annotation analysis by IPA showed a relationship with metastatic gastrointestinal cancers as well as the activation of fibrosis, interstitial disease, and mesenchymal tumor pathways (Fig. 5D). Accordingly, these results suggest that the EMT program is induced by the cooperation of Trp53 GOF/LOH mutation and activin A signaling.

GOF mutant p53 has been shown to induce a wide range of transcriptomic changes through epigenetic modifications (23, 35). We therefore examined whether the expression of EMT-TFs or their cofactors was upregulated by the Trp53 GOF/LOH mutation. Notably, the expression of Hmga1 and Hmga2 was significantly upregulated in both the control and activin-treated AKTFP^M/LOH organoids (Supplementary Table S2). HMGA1 and HMGA2 play a role in TGFβ-induced EMT and metastasis by inducing the expression of Snai1 and Twist (19–21, 36, 37). We confirmed the induction of HMGA1 and HMGA2 expression in AKTFP^M/LOH cells by Western blotting (Fig. 6A). Thus, we disrupted Hmga2 in AKTFP^M/LOH cells (Supplementary Fig. S8A). Consistent with a previous report (38), Hmga2 disruption resulted in downregulation of Ccna2 and Ccnd1, which encode cyclin A2 and D1, respectively (Supplementary Fig. S8B). However, the cell proliferation rate was not affected (Supplementary Fig. 8C). Importantly, Hmga2 gene disruption significantly suppressed activin A–induced protrusion in AKTFP^M/LOH cells, indicating a role of HMGA2 in the induction of partial EMT upon activin A treatment (Fig. 6B). Moreover, induction of Twist1 expression by activin A treatment was significantly suppressed in Hmga2-disrupted AKTFP^M/LOH organoids, and Snai1 expression was slightly decreased (Fig. 6C). The partial and biphasic downregulation of Twist1 expression in Hmga2 KO cells may be caused by remained Hmga1 expression. Consistent with the attenuated protrusion phenotype, Hmga2 disruption suppressed the activin A–induced invasion (Fig. 6D).

Finally, we examined the metastatic ability of Hmga2-disrupted AKTFP^M/LOH organoids. Importantly, the number of metastatic tumors in the lung at 1 week after intravenous injection was significantly decreased in Hmga2-disrupted cells (Fig. 6E). Taken together, these results indicate that the induction of Hmga2 by Trp53 GOF/LOH is an important mechanism of activin A–induced partial EMT and the promotion of metastasis.

In this study, we observed partial EMT of organoids in activin-treated cancer cells carrying Kras and Trp53 mutations rather than complete full EMT with single-cell dissociation. It has been reported that partial EMT and the collective migration of cell clusters play critical roles in tumor aggressiveness and metastasis (12, 14, 15). Consistently, we found that partial EMT phenotypes are associated with an increased incidence of metastasis. Thus, understanding the mechanism underlying partial EMT is important for future development of therapeutic strategies against malignant cancer progression.

Mutations in the activin receptors ACVR2A and ACVR1B are frequently detected in hypermutation-type colorectal cancer; thus, activin signaling is considered to function as a tumor suppressor, similar to the TGFβ pathway (8). In contrast, activin signaling induces EMT in cancer cells through activation of the Smad2/3/4 complex (5, 29, 30, 32). Therefore, similar to TGFβ signaling, the activin pathway has opposing functions in cancer development, namely, tumor suppressor and tumor promoter functions. However, the mechanism underlying the regulation of such opposing functions of activin is not yet understood.

In this study, we showed that these two activin functions were independently regulated by distinct genetic alterations. Kras activation protects tumor cells from activin-induced growth suppression. On the other hand, Trp53 GOF/LOH mutations are required for activin-induced partial EMT of cancer cells. Interestingly, even in Kras wild-type tumor cells, activin stimulation induced partial EMT when cells carried the Trp53 GOF/LOH mutation. These results confirmed that activin-induced growth suppression and partial EMT induction are independently regulated by Kras and Trp53 mutations, respectively, and that simultaneous mutations are important for the survival and partial EMT of cancer cells during malignant progression and metastasis. Therefore, mutant KRAS or mutant p53 could be possible therapeutic target against malignant colorectal cancer.

In cancer cells, approximately 75% of TP53 mutations are missense mutations, and mutant p53 gains new oncogenic functions through amino acid substitutions (39). The R270H mutation used in this study (corresponding to the R273H mutation in humans) is one of the hotspot GOF mutation. In human cancer cells, TP53 GOF mutation is frequently associated with loss of wild-type TP53 by LOH (40). In addition to the loss of wild-type p53 function, TP53 LOH causes nuclear stabilization of mutant p53, which is required for its oncogenic role of mutant p53. Indeed, we observed the clear nuclear accumulation of p53 in homozygous Trp53 GOF mutant intestinal tumor cells, including AKTP^M/LOH, but not in heterozygous Trp53 mutant cells, such as AKTP^M/+ cells, suggesting that wild-type p53 interferes the nuclear accumulation of mutant p53 (23, 41). Another group also reported that loss of wild-type p53 by LOH is a prerequisite for the expression of the GOF mutant p53 function (33). Accordingly, it is possible that Trp53 LOH is required for partial EMT by stabilizing nuclear GOF mutant p53.

GOF mutant p53 has been shown to modify epigenetic regulation, resulting in a significant change in the transcriptome (23, 35). In this study, we showed that the expression of Hmga1 and Hmga2 was significantly upregulated in Trp53 GOF/LOH-mutant tumor cells compared with Trp53 GOF/+ heterozygous cells. The induction of EMT-TFs by the Smad complex in TGFβ-stimulated cells requires context-dependent cofactors (13), and HMGA2 is an important cofactor that induces the expression of EMT-TFs with active Smad2/3/4 (19–21). Importantly, this study indicates that GOF mutant p53 is required for activin-induced partial EMT by inducing the expression of HMGA2. Although we did not examine the role of HMGA1, it is possible that HMGA1 also plays a role in activin A–induced partial EMT in cooperation with HMGA2 in Trp53/TP53 GOF/LOH cells.

In this study, we found that activin-pretreated AKTP^M/LOH and AKTFP^M/LOH cells showed significantly increased metastasis incidence, although the underlying mechanism remains unclear. However, we found that activin-pretreated tumor cells formed an increased number of micrometastatic foci at 24 and 48 hours after intravenous injection, suggesting that tumor cells that underwent partial EMT extravasated and interacted with residential stromal cells more efficiently than non-EMT cells, potentially promoting their survival and colonization; however, further studies are required to confirm this hypothesis.

In conclusion, we examined gene mutation patterns related to activin-induced growth suppression and partial EMT induction using mouse intestinal tumor-derived organoids. We found that Kras activation confers resistance to activin-induced growth inhibition, and the Trp53 GOF/LOH mutation promotes activin-induced partial EMT through the induction of HMGA2 expression. On the basis of these results, the control of activin signaling appears to be an important preventive and therapeutic strategy against the malignant progression of colorectal cancer carrying TP53 GOF/LOH mutations.

D. Wang reports grants from the Ministry of Education, Culture, Sports, Science, and Technology of Japan during the conduct of the study. M. Oshima reports grants from Japan Agency for Medical Research and Development and grants from Ministry of Education, Culture, Sports, Science, and Technology of Japan during the conduct of the study. No disclosures were reported by the other authors.

D. Wang: Data curation, funding acquisition, investigation. M. Nakayama: Data curation, investigation, methodology. C.P. Hong: Software, validation, investigation, methodology. H. Oshima: Data curation, investigation, methodology. M. Oshima: Conceptualization, supervision, funding acquisition, project administration, writing–review and editing.

We thank Manami Watanabe and Ayako Tsuda for their technical assistance. This work was supported by AMED (22ck0106541h0003 and 22gm4010012h0002 to M. Oshima) from the Japan Agency for Medical Research and Development, Japan; and Grants-in-Aid for Scientific Research (A; 22H00454 to M. Oshima) and Early-Career Scientists (21K15502 to D. Wang) from the Ministry of Education, Culture, Sports, Science and Technology of Japan.

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