KRAS Mutation-Responsive miR-139-5p inhibits Colorectal Cancer Progression and is repressed by Wnt Signaling

Introduction: Colorectal cancer (CRC) frequently harbors KRAS mutations that result in chemoresistance and metastasis. MicroRNAs (miRNAs) are usually dysregulated and play important regulatory roles in tumor progression. However, the KRAS mutation-responsive miRNA profile in CRC remains uninvestigated. Methods: miR-139-5p was identified and evaluated by small RNA sequencing, qRT-PCR and in situ hybridization. The roles of miR-139-5p in CRC cells with and without KRAS mutation were determined by Cell Counting Kit-8 (CCK-8), colony formation, flow cytometry and transwell assays in vitro and by tumorigenesis and metastasis assays in vivo. Microarrays followed by bioinformatic analyses, luciferase reporter assays and Western blotting were applied for mechanistic studies. Results: miR-139-5p was significantly downregulated in KRAS-mutated CRC cells and tissues compared with their wild-type counterparts. Low miR-139-5p expression was associated with aggressive phenotypes and poor prognosis in CRC patients. miR-139-5p overexpression inhibited CRC cell proliferation, migration and invasion in vitro, sensitized tumors to chemotherapy, and impaired tumor growth and metastasis in vivo. Transcriptomic profiling identified multiple modulators in the Ras (JUN and FOS) and Wnt (CTNNB1 and DVL1) signaling pathways and the epithelial-to-mesenchymal transition (EMT) process (ZEB1) as direct targets of miR-139-5p, and inverse correlations were confirmed in CRC clinical tissues. Aberrantly activated Wnt signaling in KRAS-mutant cells was demonstrated to transcriptionally repress miR-139-5p through TCF4, forming a miR-139-5p/Wnt signaling double-negative feedback loop. Conclusions: We identified miR-139-5p as a KRAS-responsive miRNA and demonstrated its involvement in CRC progression. KRAS mutation disrupted the miR-139-5p/Wnt signaling reciprocal negative feedback mechanism, which might cause miR-139-5p downregulation and derepression of oncogenic signaling pathways and EMT. These results reveal a transcriptional regulatory mode of KRAS-driven malignant transformation and highlight miR-139-5p as a novel regulator of crosstalk between the Ras and Wnt signaling pathways in CRC.


Introduction
Colorectal cancer (CRC) remains the second leading cause of cancer-related death worldwide [1]. Although screening and removal of premalignant lesions have reduced the mortality of early CRC, the overall survival of advanced CRC with metastases and chemoresistance remains poor [2]. Approximately 40% of CRC cases exhibit KRAS mutations, most of which occur in codons 12 and 13, arise early during CRC development and are maintained throughout CRC progression [3]. KRAS mutations are GAP-insensitive, which renders the proteins constitutively GTP bound and activated, leading to persistent activation of oncogenic signaling pathways and downstream oncogenes and thus promoting cell proliferation and survival [4]. Clinical investigations have revealed that CRC patients with KRAS mutations have more distant metastases, shorter survival times and poorer prognoses than patients with wild-type (WT) KRAS [5,6]. These observations might be partially explained by findings that KRAS mutation can drive CRC cell invasion and metastasis by inducing the expression of genes related to invasion, epithelial-to-mesenchymal transition (EMT) and stemness [7][8][9]. Although KRAS mutations play important roles in CRC development and progression, the mechanisms underlying the KRASmediated promotion of CRC, especially with regard to transcriptional regulation, are still unclear.
MicroRNAs (miRNAs) are a class of 22-24-nucleotide noncoding RNAs that bind to complementary target mRNAs to degrade or inhibit their translation; this process is an important mode of posttranscriptional regulation [10]. Accumulating evidence has demonstrated that miRNAs are frequently dysregulated and play key roles in cancer development and progression [10]. Recent studies have indicated that genetic mutations in known transcriptional regulators contribute to miRNA dysregulation [11][12][13]. For example, P53 mutation contributes to the chemoresistance of breast cancer cells by reducing the transcriptional activity of miR-223 [14]. APC mutation downregulates miR-155 expression, which reduces the survival and clonogenic capacity of CRC cells [15]. Notably, although KRAS mutations are some of the most frequent genetic mutations in human cancers [3], particularly in CRC, the dysregulated miRNAs and their functions and underlying mechanisms following KRAS mutation have yet to be fully explored.
Dysregulation of miR-139-5p has been reported to occur in various malignancies, including breast cancer, liver cancer, parathyroid cancer, head and neck cancer, and CRC [16][17][18][19][20][21]. Several lines of evidence have elucidated regulatory mechanisms of miR-139-5p expression. For example, P53 has been found to bind to the promoter of miR-139-5p and transactivate its expression, which suppresses PDE4D and cAMP signaling and inhibits cancer cell growth [21]. miR-139-5p is downregulated by adsorption to the long noncoding RNA LINC00152 in gastric cancer (GC), which facilitates cancer cell glycolysis by regulating PRKAA1 [22]. We previously found that CD44 induces deacetylation of histone H3 lysine 9 to suppress miR-139-5p transcription, thereby promoting cell growth and invasion in GC, indicating that epigenetic modifications are involved in miR-139-5p dysregulation [23]. Using small RNA sequencing in a pair of KRAS isogenic CRC cell lines, we have found that miR-139-5p is significantly downregulated in KRAS-mutant CRC cells. However, the cause of miR-139-5p dysregulation associated with KRAS mutation in CRC remains undetermined.
In this study, we verified that miR-139-5p is frequently downregulated in KRAS-mutant CRC cells and tissues compared with their WT counterparts and found that miR-139-5p overexpression inhibits CRC cell proliferation, migration and invasion in vitro and tumorigenesis and metastasis in vivo. miR-139-5p directly targets multiple modulators in the Ras and Wnt signaling pathways and the EMT process. Furthermore, we found a miR-139-5p/Wnt signaling reciprocal negative feedback loop in KRAS-mutant cells. Identification of these new mechanisms underlying KRAS mutation-driven malignancy may promote the development of prognostic biomarkers and potential therapeutics for CRC.

Small RNA sequencing analysis
Total RNA was extracted from each sample, and approximately 1 µg of each RNA sample was used to establish a small RNA library using an NEBNext Small RNA Library Prep Set for Illumina (New England Biolabs). The PCR products were purified using a QIAquick PCR Purification Kit (Qiagen). The products were accurately quantified for sequencing applications using a quantitative real-time PCR (qRT-PCR)-based KAPA Biosystems Library Quantification Kit. Single-end sequencing (50 bp) was performed on a NextSeq 500 Sequencer (Illumina). Mapped reads were annotated, and miRNA expression was quantified using ncPRO-seq (v1.5.1) 61 based on miRbase v19. Differentially expressed miRNAs between DKO-1 and DKs-8 cells were examined using edgeR 60. Differential expression was determined based on fold change (FC) and P values, with |log 2 (FC)|>1 and P<0.01.

qRT-PCR
An RNeasy Plus Mini Kit (50 reactions) (Qiagen, Hilden, Germany) was used to extract total RNA, which was then reverse transcribed with an Advantage RT-for-PCR Kit (Qiagen). The target sequence was amplified via qRT-PCR with a SYBR Green PCR Kit (Qiagen). The 2 -ΔΔCt method was used to determine the relative fold changes (FCs) in target gene expression in the cell lines normalized to the levels in corresponding control cells (defined as 1.0). U6 small nuclear RNA and GAPDH were used as internal controls for miRNA and mRNA assays, respectively. In the 2 -ΔΔCt equation, ΔCt=ΔCt target-ΔCt U6/GAPDH, and ΔΔCt=ΔCt expression vector-ΔCt control vector. All experiments were performed in duplicate. The PCR primers for miR-139-5p and U6 were purchased from RiBoBio (Guangzhou, China); the other PCR primers are listed in Supplementary  Table 3.

Human tissue collection
A commercial tissue microarray containing samples from 20 primary CRC tissues and 80 pairs of CRC and adjacent normal tissues was purchased from Outdo Biotech Company (Shanghai, China). Sixty paired primary CRC tissues and adjacent nontumor tissues were collected from CRC patients who had undergone surgery at Xijing Hospital. Informed consent was obtained from each patient. All procedures in this study were approved by the Protection of Human Subjects Committee of Xijing Hospital.

Targeted Sanger sequencing of KRAS
Mutant KRAS alleles were sequenced by targeted Sanger sequencing. We isolated genomic DNA using a QIAamp DNA FFPE Tissue Kit (Qiagen), and the PCR products were then sequenced using a BigDye Terminator 3.1 Cycle Sequencing Kit (Applied Biosystems) following the manufacturer's protocols.

In situ hybridization (ISH) and immunohistochemistry (IHC)
Sections of samples that were 4 μm thick were embedded in paraffin and used for ISH and IHC. For IHC, slides were probed with primary antibodies against JUN, FOS, β-catenin, DVL1, and ZEB1 that were purchased from Cell Signaling Technology (CST). An antibody against Ki-67 was purchased from Santa Cruz. For ISH, slides were incubated with 5'and 3'-digoxigenin-labeled locked nucleic acid-based miR-139-5p probes (Exiqon). Then, the slides were incubated with horseradish peroxidase-conjugated secondary antibodies (Dako) and stained with DAB chromogenic substrate (Dako). IHC and ISH staining was scored by two independent observers as previously described. Ki-67 staining was quantified by counting the positively stained nuclei per field. miR-139-5p, JUN, FOS, β-catenin, DVL1, and ZEB1 staining was quantified based on the intensity (0, no staining; 1, weak staining; 2, moderate staining; and 3, strong staining) and extent (0, no positive tumor cells; 1, <10%; 2, 10-50%; and 3, >50%) of staining. The staining index (SI) for each specimen was calculated as the product of the staining intensity and the percentage of positive tumor cells. Samples with an SI≥4 were determined to have high expression, and samples with an SI<4 were determined to have low expression.

Plasmid construction
Expression plasmids for β-catenin, TCF3 and TCF4 were constructed by inserting the corresponding cDNA sequences into the pcDNA3.1 vector; 3'-UTR luciferase reporter plasmids for JUN, FOS, DVL1, CTNNB1, TCF4 and ZEB1 were constructed by inserting the WT 3'-UTR sequences of the corresponding genes into the psiCHECK-2 Luciferase vector. Mutant constructs were created by mutating the seed regions of the miR-139-5p-binding site. Promoter luciferase reporter constructs were constructed by inserting the -5000 to +1 sequence of the 5'-flanking region of the MIR139 from human genomic DNA into the pGL3-Basic vector. A QuikChange II Site-Directed Mutagenesis Kit (Stratagene) was used to mutate the TCF4 binding sites.

Transient transfection and lentiviral infection
A synthetic miR-139-5p agomir, antagomir and corresponding negative controls were purchased from RiBoBio. siRNAs against TCF3, TCF4, and β-catenin and their scrambled controls were purchased from GenePharma (Shanghai, China). Lentiviruses expressing miR-139-5p or short hairpin RNAs against miR-139-5p sequences were obtained from GeneChem (Shanghai, China). To generate stable cell lines, the indicated cells were infected with lentiviruses at a multiplicity of infection of 100:1. Infection efficiency was confirmed by qRT-PCR at 72 h after infection, and the cells were selected with puromycin for 2 weeks.

Cell Counting Kit-8 (CCK-8) assay
Cells were seeded into 96-well plates at a density of 1000 cells in 100 μL of complete medium per well. At each time point, the original medium was replaced with a 1:9 mixture of CCK-8 solution (Transgene) and complete medium, and the cells were then incubated at 37°C for 2 h. The absorbance of each sample at 450 nm was analyzed by a microplate reader (Tecan), and each sample was measured three times.

Colony formation assay
Cells were seeded into 6-well plates at a density of 1000 cells in 2 mL of complete medium per well. After 14-18 d of culture, the cells formed stable colonies. The cell colonies were fixed with 70% ethanol and then stained with a crystal violet solution. Colonies containing more than 50 cells were counted, and each group had three replicates.

Cell cycle and apoptosis assays
Transfected cells were fixed in 75% ethanol and stained with propidium iodide (Sigma-Aldrich) supplemented with RNase A for cell cycle analysis. An Annexin V-FITC Apoptosis Detection Kit (BD Biosciences) was used for apoptosis assays according to the manufacturer's protocol. Cells were sorted using a fluorescence-activated cell sorter (BD).

Transwell migration and invasion assays
For invasion assays, chamber inserts with an 8-µm pore size were first coated with 200 mg/mL Matrigel (Corning), and the uppermost chamber was plated with 1×10 5 cells. For cell migration assays, the upper chamber with a noncoated membrane was plated with 5×10 4 cells. Each assay was repeated three times, and three different inserts were used to obtain the mean number of cells in five fields per membrane.

In vivo tumor growth and metastasis assays
All animal procedures were approved by the Fourth Military Medical University Animal Care Committee. BALB/c nude mice (6-8 weeks old) were used based on the institutional guidelines for animal care. For tumor growth models, luciferase-tagged cells were subcutaneously injected into the flanks of mice (6×10 6 cells/100 µL of PBS per injection site; 10 mice per group). The tumor volume was calculated with the following formula: tumor maximum diameter (L)×diameter along the perpendicular axis (W) 2 /2. When the tumor sizes reached approximately 100 mm 3 , the mice were randomized into treatment and control groups. Fluorouracil (5-FU) was intraperitoneally injected at a dosage of 8 mg/kg/d. For tumor metastasis models, luciferase-tagged cells (4×10 6 cells/100 µL of PBS) were injected into the tail veins or spleens of nude mice (10 mice per group). The survival of all mice was recorded throughout the experiment. Firefly luciferase was used for in vivo tracking of tumor formation and metastasis. The bioluminescence signals in tumor-bearing mice were detected using an in vivo imaging system (PerkinElmer).

Immunoblotting analysis
Protein lysates were separated by SDS-PAGE and transferred onto nitrocellulose membranes. The membranes were incubated for 1 h with a mixture of a Tween 20 solution (0.05% Tween 20, 150 mM NaCl and 120 mM Tris-HCl [pH 7.4]) and 5% milk in Tris-buffered saline (Tris-buffered saline with Tween 20, TBST) at room temperature to inhibit nonspecific binding reactions. After incubation with primary and secondary HRP-conjugated antibodies in 5% nonfat milk, the immunoreactive proteins were detected with Dura SuperSignal Substrate (Pierce). Antibodies against β-actin, JUN, FOS, β-catenin, DVL1, ZEB1, E-cadherin, Fibronectin and Vimentin were purchased from CST.

Immunofluorescence
Cells were plated onto glass coverslips, fixed with 4% paraformaldehyde and permeabilized with 0.1% Triton X-100 in PBS. The cells were incubated with primary antibodies against Vimentin (CST) and E-cadherin (CST) at 4°C overnight and were then incubated with FITC-conjugated goat anti-rabbit and Cy3-conjugated goat anti-mouse secondary antibodies for 2 h at room temperature. The immunostaining signals and DAPI-stained nuclei were visualized at room temperature using a confocal microscope (FV10i; Olympus).

Terminal deoxynucleotidyl transferase dUTP nick-end labeling (TUNEL) assay
A TUNEL assay was performed to assess apoptosis using an In situ Cell Death Detection Kit, POD (Roche), following the manufacturer's protocols. Slides were fixed in 4% paraformaldehyde for 1 h at room temperature, processed for TUNEL staining, washed with PBS, labeled with Hoechst 33258 (10 µg/µL), rinsed with PBS and mounted with a propyl gallate solution. A labeling solution without the enzyme was added as the negative control.

Agilent cDNA microarray analysis
DKO-1 cells were transfected with 20 nM miR-139-5p or control agomir. Twenty-four hours after transfection, the cells were harvested to detect changes at the mRNA level. Four hundred nanograms of total RNA was amplified and labeled using a Low Input Quick Amp Labeling Kit (Agilent Technologies) and hybridized onto Agilent Whole Human Genome Oligonucleotide Microarrays. The expression data were preprocessed and normalized by the quantile algorithm in Gene Spring Software v11.0.

Luciferase reporter assay
Luciferase assays were performed using a Dual-Luciferase Reporter Assay System (Promega) according to the manufacturer's protocol. Firefly luciferase activity was normalized to Renilla luciferase as an internal control. The transfection experiments were performed in triplicate for each plasmid construct.

Chromatin immunoprecipitation (ChIP)
ChIP assays were performed using a Magna ChIP G Assay Kit (EMD Millipore). Cells were crosslinked with 1% formaldehyde for 10 min at room temperature and quenched in glycine. DNA was immunoprecipitated from the sonicated cell lysates using a TCF4 antibody (CST) and subjected to PCR to amplify the TCF4 binding sites. The amplified fragments were then analyzed by agarose gel electrophoresis. A nonspecific antibody against IgG (BD) served as the negative control. The primers used in the ChIP assays are listed in Supplementary Table  3.

Bioinformatics analysis
The Kyoto Encyclopedia of Genes and Genomes (KEGG) database was used to analyze the main biological functions of the differentially expressed genes. The CluGO plugin of Cytoscape was used to determine the enriched Gene Ontology (GO) terms and signaling pathways of the differentially expressed genes. The Benjamini-Hochberg algorithm was used for P value correction. Enrichment was considered significant at a Q value ≤ (corrected P value) ≤0.05. Gene set enrichment analysis (GSEA) was performed using GSEA software v2.07. The enrichment score (ES) was calculated to assess the overrepresentation of members of a predefined gene set appearing at the extremes (top and bottom) of the ranked gene list.
Oncomine and The Cancer Genome Atlas (TCGA) datasets were used to determine the expression of miR-139-5p in human cancer specimens compared with normal tissues. TCGA data were analyzed using R studio (v3.5.1).

Statistical analysis
All statistical analyses were carried out with SPSS software (v19.0). Fisher's exact test was used for categorical data, while Student's t-test was used for intergroup comparisons for quantitative data. Kaplan-Meier survival analysis was used to determine cumulative survival rates. The statistical significance threshold was defined as P<0.05.

miR-139-5p is downregulated in KRAS-mutant CRC cells and tissues
To identify miRNAs affected by KRAS mutation in CRC, we utilized isogenic derivatives of DLD-1 cells that contain one WT (DKs-8 cells) or one G13D mutant KRAS allele (DKO-1 cells) [24]. The DKO-1 cells were genetically engineered to express only the mutant KRAS allele, and the DKs-8 cells expressed only WT KRAS. Via small RNA sequencing, we found that 10 miRNAs were upregulated and 23 miRNAs were downregulated in DKO-1 cells compared to DKs-8 cells (|log2(FC)|>1, P<0.01, Figure 1A). miR-139-5p was selected from the downregulated miRNAs in KRAS-mutant cells for further investigation because it has been reported to be a tumor suppressor in several cancer types [16,17,25], yet its relation to KRAS mutation remains unclear. qRT-PCR confirmed that miR-139-5p expression was reduced in DKO-1 cells compared with DKs-8 cells ( Figure 1B). Furthermore, in a panel of CRC cell lines with WT or mutant KRAS, we found lower miR-139-5p levels in KRAS-mutant cells than in WT cells ( Figure 1C). To evaluate the clinical significance of miR-139-5p, we analyzed TCGA data and found that miR-139-5p was significantly downregulated in CRC tissues relative to normal tissues; a similar trend was observed in paired CRC relative to adjacent normal samples ( Figure 1D). The downregulation of miR-139-5p was further validated in an independent cohort of 60 pairs of CRC and adjacent normal samples from Xijing Hospital of Digestive Diseases (XHDD, Figure 1E and Supplementary Table 1). Notably, among CRC tissues, lower miR-139-5p levels were observed in KRAS-mutant tissues than in WT tissues ( Figure 1E). Moreover, ISH of a tissue microarray (TMA) cohort including 20 cases of primary CRC tissues and 80 pairs of CRC and adjacent normal tissues in KRAS-mutant CRC tissues ( Figure 1F). Further correlation analyses showed that lower miR-139-5p expression was associated with a more aggressive phenotype in CRC (Table 1). Kaplan-Meier analysis revealed poor survival of CRC patients with low miR-139-5p expression ( Figure 1G). Univariate and multivariate analyses indicated that low miR-139-5p expression may serve as an independent risk factor for CRC prognosis (Table 2). These findings suggest that miR-139-5p might be a clinically relevant KRAS-regulated miRNA in CRC.

miR-139-5p inhibits proliferation, migration and invasion in KRAS-mutant CRC cells
To determine the phenotypes resulting from miR-139-5p loss in mutant RAS cells, we transfected KRAS-mutant DKO-1 and SW620 cells with a miR-139-5p agomir and introduced a miR-139-5p antagomir into WT KRAS-expressing DKs-8 and Caco-2 cells. The inhibition or overexpression of miR-139-5p in the corresponding cells upon transfection was confirmed by qRT-PCR (Supplementary Figure  S1A). CCK-8 assays and colony formation assays were performed to examine the effect of miR-139-5p on cell proliferation capacity. Overexpression of miR-139-5p in DKO-1 and SW620 cells significantly reduced proliferation and colony-forming efficiency. In contrast, inhibition of miR-139-5p in DKs-8 and Caco-2 cells increased cell growth and colony formation (Figure 2A and 2B). Flow cytometry analyses showed that miR-139-5p overexpression substantially increased apoptosis and induced G2/M phase cell cycle arrest, while miR-139-5p inhibition with the antagomir suppressed apoptosis and facilitated cell cycle progression ( Figure 2C). Transwell assays showed that miR-139-5p overexpression markedly suppressed migration and invasion in DKO-1 and SW620 cells, while miR-139-5p inhibition increased the migratory and invasive abilities of DKs-8 and Caco-2 cells ( Figure 2D). These results indicate that miR-139-5p plays tumor-suppressive roles in CRC cells, opposing the effects of KRAS mutation.

miR-139-5p increases responsiveness to chemotherapy and suppresses metastasis in KRAS-mutant CRC cells
To extend the above findings to an in vivo setting, stable miR-139-5p-overexpressing DKO-1 and SW620 cells were established by lentivirus infection and then subcutaneously injected into the flanks of nude mice. Increased expression of miR-139-5p upon transfection was confirmed by real-time PCR (Supplementary Figure S1B). Tumor size and tumor weight were significantly lower in mice implanted with miR-139-5p-overexpressing cells than in control mice ( Figure 3A and Supplementary Figure S1C-D). 5-FU, a standard chemotherapeutic agent for CRC treatment, was injected intraperitoneally after the tumor volume reached approximately 100 mm 3 . Treatment with 5-FU impaired tumor growth, and the inhibitory effect was more pronounced when miR-139-5p was overexpressed ( Figure 3A and Supplementary Figure  S1C-D). Furthermore, xenografts derived from miR-139-5p-overexpressing cells exhibited decreased proliferation and increased apoptosis, as evidenced by Ki-67 and TUNEL staining ( Figure 3B and Supplementary Figure S1E). The liver and lung are the most common metastatic sites for CRC [26]. To determine the role of miR-139-5p in metastasis, we established liver and lung metastasis models by injecting miR-139-5p-overexpressing DKO-1 and SW620 cells into the spleens and tail veins of nude mice, respectively. Compared to the control condition, overexpression of miR-139-5p decreased the incidence of liver metastasis and the number of metastatic nodules and improved survival in mice ( Figure 3C). Similarly, miR-139-5p overexpression reduced lung metastasis following tail vein injection ( Figure 3D). These results indicate that miR-139-5p retards tumor progression and reinforces responsiveness to chemotherapy in KRAS-mutant CRC cells.

High-throughput screening and identification of miR-139-5p targets in CRC cells
To identify functional targets of miR-139-5p in CRC, we performed gene expression microarray profiling in DKO-1 cells after miR-139-5p overexpression and identified 3878 differentially expressed genes (|log2(FC)|>1, P<0.01). KEGG analysis revealed that the Ras and Wnt signaling pathways were the most significantly affected pathways ( Figure 4A). Biological process analysis suggested that miR-139-5p inhibited the cell cycle, focal adhesion and cellular adhesion ( Figure 4B), consistent with our experimental results. GSEA revealed that the miR-139-5p-induced gene expression alterations were negatively associated with the KEGG Reactome terms 'RAS pathway' and 'Wnt signaling pathway' (Figure 4C). Among the differentially expressed genes, 370 were associated with RAS or Wnt signaling, and 21 of them contained miR-139-5p binding sites in their 3'-UTRs ( Figure 4D and Supplementary Table 2). Among the 21 candidates, FOS and JUN are transcription factors (TFs) regulated by RAS-activated signaling [27]; DVL1, TCF4 and CTNNB1 are key regulators in Wnt signaling [28]; and ZEB1 is an important EMT regulator involved in Ras and Wnt signaling [29]. We therefore selected these 6 genes for further validation. Using dual luciferase reporter assays, we confirmed that FOS, JUN, DVL1, CTNNB1 and ZEB1, but not TCF4, were direct targets of miR-139-5p in DKs-8 cells ( Figure 4E). The immunoblotting results confirmed the regulation of the targets by miR-139-5p in DKO-1 and DKs-8 cells ( Figure 4F).  EMT plays important roles in regulating metastasis and drug resistance [30]. Given that miR-139-5p was found to directly target ZEB1, we examined whether miR-139-5p regulates EMT in CRC cells. GSEA indicated that miR-139-5p-induced gene expression changes were negatively correlated with EMT core gene expression ( Figure  4G). Immunofluorescence and immunoblotting further showed that overexpression of miR-139-5p in DKO-1 cells significantly upregulated the epithelial marker E-cadherin and downregulated the mesenchymal markers Vimentin and Fibronectin ( Figure 4H). In contrast, E-cadherin was downregulated, while Vimentin and Fibronectin were upregulated, following inhibition of miR-139-5p in DKs-8 cells ( Figure 4H). These results indicate that miR-139-5p plays a tumor-suppressive role by targeting multiple genes in oncogenic pathways and regulating EMT in CRC cells.
TCF3 and TCF4 are correlated with β-catenin and form β-catenin/TCF3 and β-catenin/TCF4 complexes that contribute to CRC progression [34]. We thus next determined whether TCF3 or TCF4 is responsible for Wnt/β-catenin signaling-mediated miR-139-5p repression. We found that ectopic expression of TCF4, but not TCF3, suppressed miR-139-5p expression ( Figure 5E). Knockdown of TCF4 rescued miR-139-5p expression after Wnt3a treatment ( Figure 5F and Supplementary Figure S2B). We further analyzed the promoter region of miR-139-5p with the JASPAR database [35] and identified three putative TCF4 binding sites ( Figure  5G). Sequential deletions and mutations of these binding sites revealed that binding sites 1 and 2 are the major sites for TCF4-mediated repression ( Figure  5G and 5H, Supplementary Figure S2C). ChIP assays also confirmed that TCF4 binds directly to binding sites 1 and 2 in the promoter of MIR139 ( Figure 5I).

Expression patterns of miR-139-5p and its targets in human CRC tissues
To validate the correlations between miR-139-5p and its targets, we examined the expression of these molecules in the XHDD cohort of 60 CRC patients with and without KRAS mutation (Supplementary Table 1). ISH and IHC revealed that JUN, FOS, DVL1, CTNNB1 and ZEB1 levels were higher in most KRASmutant samples in which miR-139-5p expression was reduced or lost than in WT KRAS samples in which miR-139-5p was expressed at relatively high levels ( Figure 6A). Spearman correlation analysis revealed that the levels of JUN, DVL1, CTNNB1 and ZEB1 were negatively correlated with those of miR-139-5p, but the correlation between FOS and miR-139-5p did not reach statistical significance ( Figure 6B). Similar correlations were observed by qRT-PCR analyses using CRC tissues ( Figure 6C). These clinical observations further indicate that miR-139-5p suppresses multiple targets in CRC.

Discussion
Accumulating studies have revealed that driver gene mutations have comprehensive effects on miRNA expression [12,36,37]. KRAS mutation is one of the most common genetic changes in human cancers [4]. Increasing evidence has indicated that KRAS mutations alter miRNA expression and contribute to tumor development and progression. In non-small cell lung cancers, KRAS mutation induces the expression of miR-29b, which represses TET1 expression and thereby transcriptionally silences multiple tumor suppressors [38]. In pancreatic cancer, KRAS mutation represses miR-489 through NF-κB/YY1 signaling, thus inhibiting cell migration and metastasis by targeting ADAM9 and MMP7 [39]. However, in KRAS-mutant CRC, whether miRNAs contribute to KRAS-driven aggressiveness has remained largely unexplored. In the present study, we performed small RNA-Seq using DLD-1 isogenic cell lines to identify KRAS mutation-regulated miRNA signatures in CRC. We ultimately identified a group of KRAS mutation-responsive miRNAs. Among these miRNAs, miR-139-5p was validated to be downregulated in a panel of KRAS-mutant CRC cell lines compared to their WT counterparts. Moreover, we analyzed the expression patterns of miR-139-5p in three different cohorts of CRC patients (the TCGA, XHDD and TMA cohorts) and found similar trends: the levels of miR-139-5p were lower in KRAS-mutant tissues than in WT tissues. These results suggest that miR-139-5p might be a novel effector associated with KRAS mutation in CRC.  miR-139-5p has been previously reported to be downregulated in various types of cancer and to function mainly as a tumor suppressor [16-18, 21, 40]. In liver cancer, downregulation of miR-139-5p contributes to cell proliferation, migration and invasion via targeting of ETS1 [40]. In breast cancer, overexpression of miR-139-5p alleviates radiotherapy resistance by suppressing multiple genes related to DNA repair and ROS defense [16]. However, miR-139-5p levels in blood have been reported to be higher in prostate cancer patients than in healthy controls [41]. Notably, some previous studies have indicated that miR-139-5p is downregulated in CRC tissues [42], but others have reported that miR-139-5p is upregulated in metastatic CRC [43,44]. Given these conflicting findings, the ambiguous roles of miR-139-5p in CRC require further exploration. In the present study, we obtained several lines of evidence that miR-139-5p functions as a potent tumor suppressor in CRC. First, in vitro gain-and loss-of-function experiments showed that miR-139-5p inhibited CRC cell proliferation by inducing apoptosis and cell cycle arrest and repressing cell migration and invasion. Second, in vivo tumorigenesis and metastasis assays demonstrated that miR-139-5p suppressed tumorigenicity and hepatic and pulmonary metastasis in mice. Notably, ectopic expression of miR-139-5p increased 5-FU responsiveness in KRAS-mutant CRC cells, suggesting that miR-139-5p could be developed as a therapeutic agent for chemosensitivity enhancement. Third, we found that multiple genes and signaling pathways associated with the cell cycle, adherens junctions and DNA repair cellular processes were influenced by miR-139-5p. These findings partially explain how miR-139-5p functions as a potent tumor suppressor and suggest that it might be a potential prognostic biomarker and therapeutic target for advanced CRC.
TFs can activate or suppress miRNAs to fine-tune certain signaling pathways [45]. For instance, HNF4α binds to the miR-124 promoter and activates its expression, after which miR-124 targets IL6R and modulates the IL6R/STAT3 pathway during hepatocellular transformation [46]. c-Myc has been shown to transactivate the miR-17-92 cluster, which negatively regulates the expression of E2F1 and controls proliferative signaling in lymphoma cells [47,48]. In the present study, we demonstrated that Wnt/β-catenin signaling suppressed miR-139-5p transcription in a TCF4-dependent manner in KRAS-mutant CRC cells, adding to the body of knowledge regarding miR-139-5p transcriptional regulation.
Given that hyperactivation of Wnt/beta-catenin signaling pathway is a critical step in colorectal tumorigenesis, the reciprocal regulation between miRNAs and Wnt signaling represents an important regulatory pattern that orchestrates gene expression in CRC [45,49]. We found that the Wnt signaling regulators CTNNB1 and DVL1 are direct targets of miR-139-5p and therefore contribute to a double-negative feedback loop between Wnt/β-catenin signaling and miR-139-5p. These results provide new insights into KRAS mutation-driven malignancies: miR-139-5p is continuously repressed by Wnt/β-catenin signaling activation and by the reciprocal feedback loop in KRAS-mutant cells. This cellular context-dependent regulatory mechanism may enhance crosstalk among signaling pathways and participate in an elaborate network supporting cancer progression.

Conclusions
In summary, we found that miR-139-5p is significantly downregulated in KRAS-mutant CRC cells and tissues. miR-139-5p inhibits CRC cell proliferation and metastasis by targeting multiple modulators of the Ras and Wnt signaling pathways and EMT. Transcription of miR-139-5p is suppressed by Wnt/β-catenin signaling in mutant CRC cells. These findings reveal a novel mechanism of miRNA dysregulation and provide promising prognostic biomarkers and therapeutic targets for advanced CRC.