14-3-3σ Functions as an Intestinal Tumor Suppressor

Colorectal cancer accounts for the second-highest mortality among the 10 most common cancers worldwide. Over 1.8 million new colorectal cancer cases and 881,000 deaths were estimated to occur in 2018, accounting for about 1 in 10 cancer cases and deaths (1). Treatment of advanced colorectal cancer involves intestinal resection and combined chemotherapies. The current 5-year survival rate for local colorectal cancer is around 64%, and 12% for colorectal cancer with distant metastases (2). Thus, a better understanding of the mechanisms underlying colorectal cancer initiation and progression is necessary to improve diagnostics and therapy of colorectal cancer.

We identified 14-3-3σ as a p53-induced gene, which mediates a cell-cycle arrest in the G₂ phase in colorectal cancer cells (3). 14-3-3σ, or stratifin (Sfn), belongs to a gene family, which encodes 7 different 14-3-3 isoforms in mammalian cells (for review see ref. 4). Unlike the other 14-3-3 family members, the 14-3-3σ protein exclusively forms homodimers (5, 6). 14-3-3σ binds to protein ligands via a consensus-binding motif phosphorylated on serine or threonine residues. The binding to 14-3-3 proteins regulate the function of associated proteins through cytoplasmatic sequestration, masking of interaction domains and export/import sequences, prevention of degradation, modulation of enzymatic activity and transactivation, or facilitation of protein modifications (4). The exclusive homodimerization critically contributes to the ability of 14-3-3σ to inhibit cell proliferation in the presence of other 14-3-3 family members, which are often highly expressed (6). More recently, we demonstrated that 14-3-3σ-deficiency in mice leads to an increased frequency and size of DMBA/12-O-tetradecanoylphorbol-13-acetate (TPA)–induced papilloma (7).

14-3-3σ is the only 14-3-3 gene, which is induced after DNA damage (8). The p53-mediated induction of 14-3-3σ was shown to be necessary for the maintenance of a stable G₂–M arrest after DNA damage in colorectal cancer cell lines (9). 14-3-3σ associates with the cyclin-dependent kinase (CDK) CDC2/cyclin B1 complex and prevents its nuclear localization, which is required for progression through mitosis (9). Furthermore, 14-3-3σ was shown to bind to G₁-specific CDKs to suppress cell proliferation (10). We identified ∼100 additional proteins interacting with 14-3-3σ in colorectal cancer cells using a proteomic approach (11). It is therefore likely that 14-3-3σ serves as a nodal regulator of numerous cellular processes.

Here we performed a comprehensive analysis of human colorectal cancer cohorts and found that downregulation of 14-3-3σ is associated with colorectal cancer progression and poor patient survival. Then we used a genetic approach to determine whether these associations also have a causal relationship. The results show that deletion of 14-3-3σ in a mouse model of intestinal cancer enhances tumor initiation and growth. In addition, the expression profiles of 14-3-3σ–negative adenomas revealed a switch towards a mesenchymal state, which is in line with the proposed function of 14-3-3σ as a suppressor of epithelial–mesenchymal transition (EMT) and metastasis.

The generation of 14-3-3σ mice with a C57BL/6J background has been described previously (7). Apc^Min/+ mice (12) were obtained from Marlon Schneider, Ludwig-Maximilians-Universität, Munich, Germany. All animals were housed in the animal facility at the Pathology Institute of the Ludwig-Maximilians-University (Munich, Germany) in individually ventilated cages (IVC). Mice identifications, health controls (scoring) and cage changings were performed under a laminar flow hood. This animal study was approved by the Government of Upper Bavaria, Germany (AZ: 55.2–1-54–2532–4-2014).

Genomic DNA was obtained by overnight digest of ear punches in lysis buffer (50 mmol/L KCl, 1.5 mmol/L MgCl₂, 10 mmol/L Tris pH8.5, 0.01% gelatin, 0.45% NP40, 0.45% Tween 20, 100 μg/mL proteinase K). PCR was performed for 25 cycles with an annealing temperature of 65°C. Primers for detection of 14-3-3σ wild-type and 14-3-3σ knockout (KO) alleles were added in the same ratio to the PCR reaction mix. PCR products (wild-type = 106 and KO = 183 bp) were resolved on a 2% agarose gel. Genotyping of Apc^Min/+ mice was performed as previously described (13), using primers Apc^Min/+ wild-type, Apc^Min/+-common, and Apc^Min/+-mutant (Supplementary Table S1).

For the survival analysis, littermates with the genotypes Apc^Min+/− (control group) and Apc^Min/+/14-3-3σ^−/− (experimental group) were compared. The littermates with the respective genotypes were obtained by crossing Apc^Min/+/14-3-3σ^+/− with 14-3-3σ^+/− mice, both in pure C57BL/6J background. Control group and experimental group mice were housed in separated cages with up to 5 female or up to 4 male mice per cage.

Thirty-six Apc^Min/+ (C57BL/6J; 15 female, 21 male) and 39 Apc^Min/+/14-3-3σ^−/− (C57BL/6J; 19 female, 20 male) mice were maintained and monitored over a period of 282 days. Moribund mice were euthanized, reason of and age at death (days) was documented and adenomas from small intestine and colon were counted, measured (mm²) and analyzed by IHC, in situ hybridization, qPCR, and next generation sequencing (NGS). Average survival probabilities were calculated according to Kaplan–Meier using GraphPad Prism 8.3.0.

Moribund or 120 days-old mice were euthanized, intestinal tracts were isolated, flushed with ice-cold PBS, and opened longitudinally. Small intestines were separated into four equal parts (duodenum, 2x jejunum, ileum) and colon was left as a whole. Single parts were photographed, fixed as Swiss rolls in 4% buffered, formaldehyde and embedded in paraffin. Tumor numbers and sizes were evaluated after photo documentation using ImageJ 1.52a.

Paraffin-embedded tissues were sectioned at 5 μm and used for hematoxylin and eosin (H&E) staining and periodic acid–Schiff (PAS) staining according to standard protocols. Tumors from Apc^Min/+ and Apc^Min/+/14-3-3σ^−/− mice were classified as adenomas with either low- or high-grade dysplasia, based on nuclear-cytoplasmic ratio, nucleus location, prominence of nucleoli, gland architecture, amount of interglandular stroma, and the presence of mucus secretion. IHC was performed using antibodies and reagents listed in Supplementary Table S2. For each IHC detection ≥ 20 adenomas per mouse were examined. Positive staining was evaluated using ImageJ 1.52a.

For detection of intestinal stem cells an Olfm4-specific, DIG-labeled RNA probe was generated using a murine Olfm4 vector (kindly provided by Dr. Hans Clevers, Hubrecht Institute, Utrecht, the Netherlands) in combination with a DIG Northern Starter Kit (Roche Diagnostics). ISH was performed as described (14). For 14-3-3σ–ISH, we developed a novel 14-3-3σ–specific in situ RNA probe. In brief, we amplified murine 14-3-3σ cDNA (NCBI Ref. Seq.: NM_018754.2) from base pairs 14–1605 and reverse inserted the amplified sequence into pBluescriptII-KS(+) (Stratagene) between the restriction sites NotI and XhoI (NotI-1605–14-XhoI). We named the resulting vector pBSII-KS(+)Sfn-UTR. To generate a DIG labeled 14-3-3σ RNA probe, we linearized pBSII-KS(+)Sfn-UTR with ClaI and transcribed the 14-3-3σ RNA probe by using T7 primers and the negative control probe by T3 primers. Transcription using T7 primers results in a 775 bp RNA probe, which includes 459 bp 3′-UTR and 316 bp 3′-CDS of 14-3-3σ.

To visualize intestinal bacteria by fluorescence ISH, the universal eubacteria probe (EUB338) and negative control probe (NON338) were employed. The 5′-FITC–labeled EUB338 DNA (5′-GCTGCCTCCCGTAGGAGT-3′) and 5′-Cy3–labeled NON338 DNA (5′-CGACGGAGGGCATCCTCA-3′) probes were synthesized by Metabion (Planegg) and hybridized to 5-μm paraffin sections as described previously (15).

Total RNA was isolated from tumor samples using the RNeasy Plus Mini Kit (Qiagen). cDNA was generated by Verso cDNA kit (Thermo Scientific) and qPCR was performed by using Fast SYBR Green Master Mix (Applied Biosystems) and a LightCycler 480 II (Roche Diagnostics). Relative gene expression was determined using the 2^–ΔΔC_t method (16). The individual mRNA levels were normalized to cyclophilin. Primers used for qPCR are listed in Supplementary Table S3.

Expression and clinical data of The Cancer Genome Atlas (TCGA) colon adenocarcinoma (COAD) and rectal adenocarcinoma (READ) cohorts was obtained from the MD Anderson standardized data browser (http://bioinformatics.mdanderson.org/TCGA/databrowser/). The RNA sequencing (RNA-seq) by Expectation-Maximization (RSEM) normalized expression values from the Illumina RNASeqV2 (genes) datasets were used. Expression and clinical data of other colorectal cancer patient datasets was downloaded from NCBI Gene Expression Omnibus (GEO; www.ncbi.nlm.nih.gov/geo). Differential expression between tumors and adjacent normal colon tissue was calculated by paired t test. Differential expression between primary tumors from patients with colorectal cancer with and without metastasis was calculated using the limma R-package (17). Differential expression between tumors of different stages was calculated using one-way ANOVA with a posttest for linear trend from stage 1 to stage 4. The statistics for Forest plots and survival curves was calculated by log-rank test. For binary classification of cases (high/low expression), the Survminer R-package (https://CRAN.R-project.org/package=survminer) was used to determine optimal cut-off values. The correlation between differentially expressed genes and 14-3-3σ in public colorectal cancer patient datasets was analyzed using the Pearson correlation coefficient analysis. SNAIL-, c-JUN-, YAP1-, and FOXO1-regulated genes were identified on the basis of datasets from the NCBI GEO. The regulation of the expression of these genes was assessed using gene expression profiling datasets of cell lines/tissues with SNAIL, c-JUN, YAP1, or FOXO1 ectopic expression or knockdown (KD)/KO. The binding of SNAIL, c-JUN, YAP1, or FOXO1 to the regulatory regions of these genes was assessed using GEO chromatin immunoprecipitation–sequencing (ChIP-seq) datasets. Raw ChIP-seq data were downloaded from the NCBI Sequence Read Archive (SRA). Reads were mapped to the human or mouse reference genomes and processed using the ChIP-seq analysis module of the CLC Genomics Workbench software (Qiagen Bioinformatics) with default settings. The identifiers of all analyzed GEO public GEO microarray, RNA-seq, and ChIP-seq datasets are provided in Supplementary Tables S4–S7.

Total RNA from murine adenomas was isolated using the RNeasy Plus Total RNA Isolation Kit (Qiagen) according to manufacturer's instructions. RNA profiling was done in quadruplicates. Random primed complementary DNA libraries were constructed and sequenced using the HiSeq2500 (Illumina) platform by Eurofins/GATC. Each sample was covered by at least 30 million single reads of 50-bp length. The raw reads (FASTQ files) were cleaned by trimming of adaptor sequences and low-quality sequences with average quality scores <20. Trimmed reads were mapped to the GRCm38/mm10 mouse reference genome and processed using the RNA-seq module of the CLC Genomics Workbench v8.0 software (Qiagen Bioinformatics) with default settings. RNA-seq data were filtered to exclude weakly expressed transcripts with less than ten mapped reads in all samples from the analysis and subjected to normalization using the R-package RUVSeq (remove unwanted variation from RNA-seq data) package (18). Differential gene expression analysis was performed using limma (17), edgeR (19), and DESeq2 (20) methods. Gene-set enrichment analyses (GSEA) were performed using GSEA software (Broad Institute, Boston, MA) and data from the molecular signature database (MSigDB). The significance of enrichments is presented by normalized enrichment scores with false discovery rate–adjusted q values. Heatmaps were generated with Morpheus (Broad Institute, Boston, MA).

A two-tailed Student t test was used to compare continuous variables. Categorical variables were compared using the χ² method. Kaplan–Meier calculations were used to display the overall survival time and the results were compared with a log-rank test. The Sidak method was used to adjust P values when multiple comparisons were performed. P values less than 0.05 were considered significant and indicated by asterisks (*, P < 0.05; **, P < 0.01; or ***, P < 0.0001). GraphPad 8.3.0 was used for calculations.

The RNA-seq data generated in this study have been deposited at the NCBI GEO with the accession number GSE163257.

14-3-3σ expression is downregulated in several types of human epithelial cancers, among them prostate, lung, breast, skin, and colorectal cancer (4, 21–23). However, 14-3-3σ expression in colorectal cancer has been so far only described in a small number of human patients (n = 116; ref. 23). In this study, downregulation of 14-3-3σ in colorectal cancers was associated with tumor progression and poor prognosis of patients. Thus, we analyzed 14-3-3σ expression and clinical associations in colorectal cancers from multiple patient cohorts deposited in TCGA and GEO databases (24). The expression of 14-3-3σ mRNA was significantly lower in tumors compared with adjacent normal colonic mucosa in 8 from 15 analyzed cohorts of matched pairs of tumor and normal colon tissue (Fig. 1A; n = 644). The expression of 14-3-3σ was also significantly lower in primary colorectal cancers from patients with metastasis compared with patients without metastasis in a pooled cohort of 3,103 patients (Fig. 1B). In colorectal cancers, 14-3-3σ displayed a gradually decreasing expression with increasing tumor stage in 4 of 5 analyzed patient cohorts (Fig. 1C). In colorectal cancers with mutant p53 the expression of 14-3-3σ was significantly lower than in p53 wild-type colorectal cancers (Fig. 1D). Decreased 14-3-3σ expression showed a significant correlation with poor survival in the combined analysis of nine colorectal cancer patient cohorts (Fig. 1E and F). Further analysis of the colorectal cancers with a known p53 status revealed, that this association was more pronounced in p53 wild-type than in p53-mutant colorectal cancers (Fig. 1G and H). Therefore, p53 mutations are presumably not a hidden cause for the association of decreased 14-3-3σ expression with poor survival. Taken together, the clinical associations linked to decreased 14-3-3σ expression identified in the cohorts of patients with colorectal cancer indicate that 14-3-3σ may function as a tumor suppressor in colorectal cancer.

To obtain genetic evidence for a tumor suppressor function of 14-3-3σ in the intestine we decided to analyze the effect of 14-3-3σ deletion in the Apc^Min/+ mouse model of inherited intestinal cancer/familial adenomatous polyposis (FAP). In Apc^Min/+ mice, which carry a mutant Apc allele, the additional spontaneous loss of the remaining wild-type Apc allele leads to the development of multiple intestinal adenomas (12, 25). To determine whether 14-3-3σ suppresses intestinal adenoma initiation and/or promotion, we generated 14-3-3σ–deficient Apc^Min/+ mice. First, we analyzed the 14-3-3σ mRNA expression pattern in Apc^Min/+ mice by ISH. In normal appearing epithelium of the small intestine, 14-3-3σ was moderately expressed above the crypts in progenitor cells of the transient amplifying cell compartment and strongly in differentiated cells towards the apical villus epithelium (Fig. 2A). In the colorectal epithelium, 14-3-3σ was also expressed above the crypts in progenitor cells of the transient amplifying cell compartment and in the differentiated cells towards the colon lumen (Fig. 2B). As expected, 14-3-3σ expression was not detectable in the intestinal epithelium of 14-3-3σ–deficient Apc^Min/+ mice (Fig. 2A and B, right). 14-3-3σ–deficient Apc^Min/+ mice were born at a normal Mendelian ratio and were phenotypically indistinguishable from Apc^Min/+ mice with wild-type 14-3-3σ. Deletion of 14-3-3σ did not affect the lengths of the small intestines (Supplementary Fig. S1A) and the width of the crypts in Apc^Min/+ mice (Supplementary Fig. S1B). However, the depth of the crypts was slightly increased. In addition, the number of intestinal stem cells was not influenced by 14-3-3σ deletion, as determined by ISH with a stem cell specific Olfm4 in situ probe (Supplementary Fig. S1C). In addition, the numbers of Paneth (Supplementary Fig. S1D) and Goblet cells (Supplementary Fig. S1E) were unchanged.

Next, we performed a Kaplan–Meier analysis to determine the average survival of Apc^Min/+ and 14-3-3σ–deficient Apc^Min/+ mice. For the survival analyses, littermates with the respective mutated alleles were compared. Apc^Min/+ mice showed a median survival of 192 days (Fig. 3A), which is in accordance with previous publications (26, 27). Notably, the median survival of 14-3-3σ deficient Apc^Min/+ mice was decreased by 64 days. This effect was independent of the gender of the mice (Fig. 3B and C). Therefore, loss of 14-3-3σ provided a selective disadvantage for the tumor prone Apc^Min/+ mice, which is consistent with a tumor suppressive function of 14-3-3σ. Consistently, the comparison of adenomas from age-matched, 120 days old Apc^Min mice showed, that 14-3-3σ-deficiency increased the number of adenomas by 2.3-fold, total tumor area by 3.6-fold, and the average tumor area by 1.4-fold (Fig. 3D and E). These effects of 14-3-3σ loss were independent of the gender (Supplementary Fig. S2A and B). When mice, that became moribund due to intestinal tumor burden were analyzed, 14-3-3σ–deficient Apc^Min/+ mice displayed an increased number of adenomas and a larger total tumor area (Fig. 3F). However, the mean tumor area showed no difference. The lack of a difference in the tumor area was presumably due to the longer lifetime of 14-3-3σ–proficient Apc^Min mice, which allowed more cell divisions to occur. The effects of 14-3-3σ–deficiency on tumor incidence and size observed in moribund Apc^Min mice were independent of the gender (Supplementary Fig. S3A and S3B).

The increased size of adenomas in 120 days old 14-3-3σ–deficient Apc^Min/+ mice may be caused by an increase in tumor cell proliferation and/or decrease in apoptosis. Indeed, the number of Ki-67–positive cells in adenomas from 120 days–old 14-3-3σ-deficient Apc^Min/+ mice was significantly larger than in Apc^Min/+ mice (Fig. 4A). Consistent with an increased proliferation in 14-3-3σ–deficient intestinal tumors/adenomas we detected a decreased expression of the mRNAs encoding the cell-cycle inhibitors p16INK4a and p19ARF, as well as increased expression of the mitosis promoting cyclin B1 in 14-3-3σ–deficient Apc^Min/+ adenomas (Fig. 4B). Furthermore, the number of cells positive for cleaved-caspase-3 was significantly decreased in 14-3-3σ–deficient Apc^Min/+ mice (Fig. 4C). Therefore, a decrease in apoptosis presumably also contributes to the increase in tumor size observed after deletion of 14-3-3σ.

Next, we analyzed the potential effect of 14-3-3σ deletion on the epithelial barrier function by evaluating bacterial infiltration. ISH with the pan-bacteria probe EUB338 revealed that bacterial infiltration in 14-3-3σ–deficient Apc^Min/+ adenomas is lower than in Apc^Min/+ adenomas (Fig. 4D). Bacterial infiltration was not detected in villi of the small intestinal epithelium of Apc^Min/+ and 14-3-3σ-deficient Apc^Min/+ (Supplementary Fig. S4A). Furthermore, the frequency of infiltrating immune cells, such as T cells, macrophages, B cells, and neutrophils/granulocytes, was not changed significantly in 14-3-3σ–deficient Apc^Min/+ adenomas (Supplementary Fig. S4B). Subsequently, we evaluated whether deletion of 14-3-3σ influences the degree of dysplasia present in adenomas obtained from moribund Apc^Min/+ mice. As described before (28), low-grade dysplasia in adenomas was associated with enlarged, pencil-like, dark nuclei and partial loss of cell polarity, whereas high-grade dysplasia was characterized by enlarged, round, transparent nuclei with visible nucleoli and loss of cell polarity and dense growth. The evaluation of H&E-stained adenoma sections showed that moribund 14-3-3σ–deficient Apc^Min/+ mice displayed significantly more adenomas with high-grade dysplasia (Fig. 4E). The high-grade dysplastic morphology of cells within 14-3-3σ–deficient adenomas implies that loss of 14-3-3σ promotes the dedifferentiation of tumor cells.

To further illuminate the mechanism by which the loss of 14-3-3σ promotes intestinal tumorigenesis in Apc^Min/+ mice, we compared the mRNA expression profiles of intestinal adenomas from Apc^Min/+ and Apc^Min/+/14-3-3σ^−/− mice. For each genotype, 12 adenomas derived from four mice were analyzed (3 per mouse). Principal component analysis revealed that the expression profiles of adenomas from Apc^Min/+/14-3-3σ^−/− mice showed high similarities and were substantially different from profiles of adenomas from Apc^Min/+ mice (Fig. 5A). Differential gene expression analyses using the Limma, edgeR, and Deseq2 algorithms revealed that 2,953 mRNAs were commonly, differentially expressed (P < 0.05) between adenomas from Apc^Min/+ and Apc^Min/+/14-3-3σ^−/− mice (Fig. 5B). Out of these, 618 mRNAs were downregulated and 816 mRNAs were upregulated in 14-3-3σ-deficient adenomas with a fold change in expression > 1.5 (Fig. 5C and D). The mRNAs upregulated in 14-3-3σ-deficient adenomas generally showed a negative correlation with the expression of 14-3-3σ in colorectal cancer samples from 13 patient cohorts, whereas the downregulated mRNAs showed a positive correlation with the expression of 14-3-3σ in colorectal cancers (Supplementary Fig. S5).

GSEA of upregulated mRNAs showed the most pronounced enrichment within EMT and inflammatory cytokine signaling, as well as cell motility and extracellular matrix MSigDB gene sets (Fig. 5E). Because it has been suggested that 14-3-3σ is involved in the regulation of p53 activity (29), we evaluated the expression of p53 target genes. Although, p53 target gene signatures showed a tendency for enrichment in wild-type compared with 14-3-3σ–deficient tumors, the effects were not statistically significant (Supplementary Fig. S6). Therefore, loss of 14-3-3σ did not affect p53-driven gene expression significantly, which is in line with the preferential binding of other 14-3-3 isoforms to p53 (30).

Next, we defined a 14-3-3σ KO associated signature by subtracting the sum of normalized expressions of 618 downregulated mRNAs from the sum of normalized expressions of 816 upregulated mRNAs. High prevalence of the 14-3-3σ KO signature was significantly associated with poor relapse-free survival in six of nine colorectal cancer patient cohorts (Fig. 5F). Moreover, combined analysis of the pooled data of all nine patient cohorts showed that high expression of the 14-3-3σ KO signature is strongly associated with poor relapse-free survival (Fig. 5G). Finally, the expression of the 14-3-3σ KO signature was highest in the CMS4 (mesenchymal) molecular subtype of colorectal cancer (Fig. 5H), which is associated with metastasis and poor prognosis (31).

As the most pronounced changes in expression associated with the loss of 14-3-3σ were within the EMT and inflammatory cytokine signaling signatures, we inspected these signatures in more detail. The MSigDB Hallmark EMT signature contains 200 mesenchymal state–associated genes, of which, 62 were significantly elevated in 14-3-3σ-deficient adenomas (Fig. 6A). Moreover, the expression of core EMT-TFs SNAI1, TWIST1, VIM, ZEB1, and ZEB2, was also elevated in adenomas derived from Apc^Min/+/14-3-3σ^−/− mice (Fig. 6A). Interestingly, the expression of epithelial state associated mRNAs, such as CDH1, EPCAM, and various keratins was downregulated in 14-3-3σ-deficient adenomas (Fig. 6A), indicating that the loss of 14-3-3σ leads to an EMT. Interestingly, the keratins Krt14 and Krt17 are known to directly interact with 14-3-3σ protein to regulate epithelial cell functions (32, 33). The expression of mesenchymal state associated genes generally showed a negative correlation with the expression of 14-3-3σ in tumor samples from 13 public colorectal cancer patient cohorts, whereas the expression of epithelial state associated genes showed a positive correlation with the expression of 14-3-3σ in tumors from a patient with colorectal cancer (Fig. 6A). qPCR analysis of differentially expressed EMT-TFs Twist1 and Snai1, as well epithelial state–associated mRNAs Krt7, Krt18, and Krt19 validated the differential expressions detected by RNA-seq analysis (Fig. 6B). Finally, 14-3-3σ-deficient adenomas showed elevated expression of several inflammatory cytokines and their receptors, such as the protumorigenic IL6 (34) and hepatocyte growth factor (HGF), which is the ligand of the c-MET proto-oncogene that also stimulates the expression of EMT-transcription factors (TF; Fig. 6C; refs. 35, 36). The expression of inflammatory cytokines and their receptors generally showed a negative correlation with the expression of 14-3-3σ in tumors from patients with colorectal cancer (Fig. 6C). Altogether, our expression profiling results suggest that EMT may be involved in the increased dedifferentiation and dysplasia of 14-3-3σ-deficient adenomas. Moreover, the altered cytokine–receptor signaling suggests a role of 14-3-3σ downregulation in enhancing tumor/stroma interactions.

14-3-3σ is known to associate with the TFs SNAIL, c-JUN, YAP1, and FOXO1, thereby promoting nuclear export and/or degradation, and ultimately suppressing activity of these TFs (37–40). Here we analyzed whether gene regulations mediated by these TFs are involved in the tumor-suppressing functions of 14-3-3σ in the Apc^Min/+ mouse model. First, we established signatures of SNAIL, c-JUN, YAP1, and FOXO1 target genes by retrieving expression data from multiple public GEO datasets representing ectopic expression/KD/KO studies to identify genes that are consistently regulated by these TFs. Moreover, we analyzed multiple GEO ChIP-seq datasets to identify genes directly bound by these TFs. Notably, SNAIL, c-JUN, YAP1, and FOXO1 target gene signatures were significantly enriched in 14-3-3σ-deficient adenomas, implying that these TFs are activated by the deletion of 14-3-3σ in adenomas of Apc^Min/+ mice (Fig. 7A, C, E, and G). Figure 7B, D, F, and H show examples of SNAIL, c-JUN, YAP1, and FOXO1 target genes that are differentially expressed between 14-3-3σ-proficient and -deficient Apc^Min/+ adenomas and thus might be involved in the tumor-suppressing functions of 14-3-3σ (see also Supplementary Table S4–S7 for details of the analyzed public GEO microarray, RNA-seq, and ChIP-seq datasets). For example, connective tissue growth factor (CTGF) was upregulated in 14-3-3σ-deficient adenomas and represents a well-established YAP1 target gene (41), that has been associated with the induction of EMT and tumor progression (42, 43). Analysis of colorectal cancer cohorts revealed that elevated expression of CTGF in colorectal cancers is consistently associated with poor survival of patients, CMS4 subtype, and showed a gradually increased expression with tumor progression (Fig. 7I–K). Furthermore, fibroblast growth factor receptor 1 (FGFR1), which is known to be directly induced by c-JUN (44), was upregulated in 14-3-3σ-deficient adenomas and was also found to be associated with poor survival, CMS4 subtype, and tumor stage (Fig. 7L–N). Taken together, the effects of 14-3-3σ loss on adenoma-specific gene expression detected here indicate that the inactivation of 14-3-3σ during tumor progression may have profound effects on multiple cancer relevant processes, such as EMT and tumor–stroma signaling. Furthermore, our results indicate that tumor suppression by 14-3-3σ is, at least in part, mediated by the inhibition of 14-3-3σ–associated TFs which is lost upon 14-3-3σ inactivation.

14-3-3σ expression is epigenetically silenced in several types of human cancers (21), suggesting that 14-3-3σ may also represent a tumor-suppressor in colorectal cancer. However, the in vivo role of 14-3-3σ in tumor suppression has so far only been analyzed genetically in breast and skin cancers (7, 45, 46) and not in colorectal cancer, where its tumor-suppressive function has initially been suggested (3). Here, we describe a protumorigenic effect of 14-3-3σ deletion in the Apc^Min/+ mouse model of colorectal cancer and thereby demonstrate that 14-3-3σ represents a suppressor of intestinal tumorigenesis.

In the normal epithelium of the small intestine of Apc^Min mice we detected 14-3-3σ expression in transient amplifying (TA) crypt cells and differentiated enterocytes of the villi. 14-3-3σ deficiency caused an increase in the frequency and size of small intestinal and colonic tumors in Apc^Min/+ mice. We suggest that the function of 14-3-3σ in the TA cell compartment is different from its function in the differentiated, apical villi cells. In the highly proliferative TA cell compartment 14-3-3σ might be involved in the negative regulation of TA cell proliferation as part of genome surveillance by p53. However, in differentiated, apical villi cells, 14-3-3σ is presumably involved in the regulation of terminal differentiation. This hypothesis is supported by our finding that 14-3-3σ–deficient adenomas are larger and show an increase in proliferating and a decrease in apoptotic cells. Interestingly, the expression of mRNAs encoding the CDK inhibitor (CDKI) p16INK4a, which inhibits G₁–S-transition, and p19ARF was decreased in 14-3-3σ–deficient adenomas. Because p19ARF and p16INK4a harbor antiproliferative activity, their decreased expression may contribute to the increased proliferation and size of 14-3-3σ–deficient adenomas.

Tumor formation in Apc^Min/+ mice is invariably associated with loss of the remaining wild-type allele (12), which occurs early in colorectal cancer tumorigenesis. The frequency is presumably influenced by the degree of chromosomal instability (CIN; ref. 47). Interestingly, 14-3-3σ–deficient colorectal cancer cells show an increased rate of CIN as evidenced by more frequent chromosomal aberrations, such as telomere-telomere fusions, unbalanced translocations, and chromosomal breaks, upon ionizing radiation (48). Because 14-3-3σ has been implicated as a mediator of a G₂–M cell-cycle arrest after DNA damage (3, 9), loss of the second Apc allele might be more frequent in 14-3-3σ-deficient Apc^Min/+ mice, due to their inability to maintain a stable G₂ arrest, which would prevent the accumulation of unrepaired chromosomal aberrations. Interestingly, Apc loss of heterozygosity (LOH) leads to microadenomas, of which, not all develop into macroadenomas (49). Therefore, a tumor-suppressive arrest exists at the early stage of adenoma formation, which may involve the activation of 14-3-3σ expression. Inactivation/downregulation of 14-3-3σ by CpG methylation or other mechanisms may allow, or at least contribute to, the formation of macroadenomas.

Our NGS analysis of Apc^Min/+ adenomas showed that 14-3-3σ-deficiency in adenomas leads to the upregulation of the EMT-TF Snail, which is known to promote the cellular transition from an epithelial to a mesenchymal state (50). It has been suggested, that 14-3-3σ inhibits the transcriptional activity of Snail by sequestering phosphorylated Snail in the cytoplasm and thereby keeps cells in an epithelial-like state (37, 51). Therefore, upregulation of 14-3-3σ may inhibit Snail-induced EMT, whereas downregulation/inactivation of 14-3-3σ might accelerate Snail-induced EMT. Thus, decreased levels of 14-3-3σ might be necessary to maintain a mesenchymal colorectal cancer phenotype, which is found in those colorectal cancers that belong to the CMS4 subtype. This hypothesis is supported by the findings of Raychaudhuri and colleagues (38): Human colorectal cancer cells (HCT116) deficient for 14-3-3σ display upregulation of EMT-markers, increased migration and invasion. Therefore, 14-3-3σ-deficient HCT116 cells presumably passed through an EMT. Thus, 14-3-3σ is presumably involved in the inhibition of EMT (i.e., MET), whereas downregulation of 14-3-3σ may promote EMT during colorectal cancer progression. In line with this scenario, the expression signature associated with deletion of 14-3-3σ characterized here was associated with poor relapse-free survival of patients with colorectal cancer, which is mainly determined by the formation of distant metastasis.

It has been shown that 14-3-3σ binds to the transcription factors SNAIL, c-JUN, YAP1, and FOXO1, and thereby suppresses their activity (37–40). The target gene signatures of these TFs were significantly enriched in 14-3-3σ-deficient adenomas, suggesting that the transcriptional programs mediated these TFs are activated after the loss of 14-3-3σ. For example, the known YAP1 target gene CTGF (41) and the c-JUN target gene FGFR1 (44) were upregulated in 14-3-3σ-deficient adenomas and presumably contribute to the enhanced tumor formation and growth in these mice. These genes have established oncogenic properties in colorectal cancer and other cancer types (38, 40, 51). Moreover, our analyses showed that elevated expression of CTGF and FGFR1 is consistently associated with poor survival in patients with colorectal cancer, CMS4 colorectal cancer subtype, and showed a gradual increased expression with increasing tumor stage in multiple colorectal cancer patient cohorts.

Furthermore, our results implicate a role of 14-3-3σ in the regulation of cytokine receptor signaling, cell motility, and extracellular matrix organization. Via influencing the tumor–stroma interactions, the deregulation of these processes caused by loss/downregulation of 14-3-3σ may contribute to colorectal cancer progression.

Tumor progression in the Apc^Min mouse model is known to be limited to the development of noninvasive adenomas (52). Additional genetic lesions or treatments may result in further progression of these adenomas to invasive adenomas, although with a low frequency (e.g., loss of p53, ref. 53, or treatment with DSS, ref. 54). However, formation of local or distant metastasis has not been observed in Apc^Min mice even after introduction of additional lesions (52). To determine the effect of 14-3-3σ loss on invasion and metastases formation in the context of additional lesions that occur during colorectal cancer progression it is therefore necessary to inactivate 14-3-3σ in mouse models of metastatic colorectal cancer. The results presented here warrant such efforts.

H. Hermeking reports grants from Deutsche Krebshilfe during the conduct of the study; in addition, H. Hermeking has a patent for “14-3-3sigma arrests the cell cycle” United States Patent 6.740.523 issued. No disclosures were reported by the other authors.

M. Winter: Investigation, writing–original draft. M. Rokavec: Investigation, bioinformatics analyses. H. Hermeking: Conceptualization, resources, supervision, funding acquisition, writing–original draft, writing–review and editing.

We thank Ursula Götz for technical assistance, Marlon Schneider for the Apc^Min/+ mice, and Hans Clevers for the Olfm4 cDNA plasmid. The project was funded by grants of the Deutsche Krebshilfe (#70112441 and #70113975 to H. Hermeking).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.