RHBDD1 promotes colorectal cancer metastasis through the Wnt signaling pathway and its downstream target ZEB1

HCT-116, HCT-8, CACO-2, and DLD1 cells were all obtained from and authenticated by the Cell Resource Center at Peking Union Medical College. All these cell lines were passaged fewer than 6 months after purchase for all the experiments. HCT-116, and DLD1 cells were cultured in IMDM (HyClone, SH30228.01) supplemented with 10% fetal bovine serum (Gibco, 10099141); HCT-8 cells were cultured in RPMI-1640 medium (Cell Resource Center, IBMS, CAMS/PUMC, CCCM019) supplemented with 10% fetal bovine serum; and CACO-2 cells were cultured in MEM-EBSS medium (HyClone, SH30024.01) supplemented with 10% fetal bovine serum and nonessential amino acids (HyClone, SH30238.01). Expression plasmids for RHBDD1, S552D-β-catenin and S675D-β-catenin were cloned into a pcDNA6.0 plasmid with a C-terminal myc-tag. S552D-β-catenin and S675D-β-catenin point mutation constructs were generated using the overlap extension PCR method. The primers used are shown in the Additional file 1: Table S1. The sequences of the two RNAi oligonucleotides targeting RHBDD1 are shown in the Additional file 1: Table S1 and were generated by GenePharma. CHIR99021 was purchased from Selleck Chemicals (S2924) and was dissolved in DMSO. Puromycin was purchased from MP Biomedicals (194539). The transfection regent Lipo3000 was purchased from Invitrogen (L3000008).

A total of 8 histopathologically confirmed CRC tissues at the Cancer Hospital, Chinese Academy of Medical Sciences were used for western blot analysis. Prior patient consent and approval from the Institutional Research Ethics Committee were obtained for the use of these patient specimens for research purposes. The study was conformed to the ethical guidelines of the Declaration of Helsinki. Gray scale scanning was conducted using ImageJ software. The tissue microarray comprising tumor tissues and their corresponding adjacent normal tissues from 71 patients with CRC were obtained from Shanghai Biochip (HCol-Ade150CS-01). The protein expression levels were evaluated using an immunostaining score, which was calculated as the sum of the proportion and intensity of the stain. Briefly, a proportion score was assigned first, which represented the estimated percentage of positively stained tumor cells (0, < 1%; 1, 1–24%; 2, 25–49%; 3, 50–74%; and 4, 75–100%). Next, an intensity score was assigned, which represented the average intensity of the positively stained tumor cells (0, none; 1, weak, 2, intermediate; and 3, strong). The proportion and intensity scores were then added to obtain a final score, which ranged from 0 to 7.

The anti-RHBDD1 mouse monoclonal antibody was prepared in-house. Antibodies against β-catenin (8480), phospho-33/37/41-β-catenin (9561), phospho-552-β-catenin (5651), phospho-675-β-catenin (4176), ZEB1 (3396), and β-actin (4970) were purchased from Cell Signaling Technology. Antibodies against Lamin A/C (2966-1) were purchased from Epitomics. Antibodies against the myc-tag (M5546) and tubulin (T6199) were purchased from Sigma-Aldrich. Antibodies against the GAPDH (TA-08) was purchased from ZSGB-BIO.

The RHBDD1 mutant stable cell line was established in our previous work [16]. The RHBDD1 rescue HCT-116 stable cell line was derived from the HCT-116 mutant stable cell line. The mutant cell line was infected with either lentivirus-RHBDD1 or lentivirus-control. Double dilutions of the cells were seeded in 96-well cell culture plates to generate single cell-derived clones. The positive clones were selected by immunoblotting for RHBDD1. The selected clones were expanded, cultured and preserved for subsequent experiments.

An sgRNA (TAGCAACTTTGGCCCTCAAC) was designed using the Zhang lab website (http://​www.​genome-engineering.​org). Off-target effects were predicted using the Cas-OFFinder website (http://​www.​rgenome.​net/​cas-offinder). The primers used were as follows: Forward: ATGCAACGGAGATCAAGAGG, Reverse: TGTCTCCCTTACCTGAGAAACC. The experiment was conducted according to methods described by Cong et al. [23]. The plasmid used was purchased from Addgene (42230).

The wild-type HCT-116 cell line was transduced with lentivirus-luciferase to generate an HCT-116-luciferase stable cell line. Then, this cell line was used to establish an HCT-116 RHBDD1 knockdown stable cell line. The HCT-116-luciferase stable cell line was transduced with lentivirus-Negative control, lentivirus-Si-RHBDD1-1#, or lentivirus-Si-RHBDD1-2#. Double dilutions of the cells were seeded in 96-well cell culture plates to generate single cell-derived clones. The positive clones were selected based on GFP expression followed by expansion, culturing and preservation for subsequent experiments.

For the cell migration assay, 1 × 10⁵ cells in serum-free IMDM were seeded in the upper chamber of the transwell insert (Millipore, MCEP24H48), and the lower chamber contained complete IMDM with 10% FBS. At 24 h after inoculation, the transwell chamber was collected and stained with crystal violet. For the cell invasion assay, the upper chamber of the transwell insert was pre-coated with Matrigel (BD, 356234), and incubated at 37 °C for 3 h to form the Matrigel layer in the chamber. A total of 2 × 10⁵ cells in serum-free IMDM were seeded in the upper chamber of the transwell insert, and the lower chamber contained complete IMDM with 20% FBS. At 24 h after inoculation, the transwell chamber was collected and stained with crystal violet. The experiment was performed in triplicate.

Animal experiments were performed with the approval of the Peking Union Medical College Animal Care and Use Committees. Male and female NOD/SCID mice (6 weeks old) were used to perform the metastasis analysis in vivo. Three stable cell lines (HCT-116-nc-luciferase, HCT-116-Si-RHBDD1-1#-luciferase and HCT-116-Si-RHBDD1-2#-luciferase; n = 4 per cell type) were inoculated via tail vein injection (5 × 10⁵ cells per mouse). At 28 days after injection, the luciferase activity and the signal distribution in each mouse was detected using an in vivo imaging system.

The TCF/LEF binding regions were used for the canonical Wnt signaling pathway. HCT-116 cells were seeded in a 24-well cell culture plate and co-transfected with the pGL3-Basic plasmid containing the specific promoter (200 ng/well) and the pRL-TK plasmid (10 ng/well). At 36–48 h later, the cells were analyzed for fluorescence intensity using a Dual-Luciferase Reporter Assay System (Promega, E1910). The cells were washed twice with pre-chilled PBS, lysed with 100 μl of PLB per well for 15 min at room temperature, and transferred to a 96-well plate (Corning, 3917) (15 μl lysate/well) for luminescence detection. The results are shown as the ratio of firefly luciferase intensity and renilla luciferase intensity. The experiment was performed in triplicate.

HCT-116 cells were seeded in a 24-well cell culture plate and co-transfected with the pRL-TK plasmid (10 ng/well) and either TOP flash plasmid or FOP flash plasmid (200 ng/well). At 36–48 h later, the cells were analyzed using Dual-Luciferase Reporter Assay System (Promega, E1910) to assess the luminescence intensity. The exact procedures performed in this experiment were the same as those for the dual-luciferase reporter assay. The results are shown as the ratio of TOP Flash activity and FOP Flash activity. The experiment was performed in triplicate.

Total RNA was isolated from the different cell lines using TRIzol Reagent (Invitrogen, 15596018) according to the manufacturer’s instructions. Equal amounts of RNA were reverse transcribed into cDNA using a Transcriptor First Strand cDNA Synthesis Kit (Roche, 04896866001) as instructed by the manufacturer. Quantitative PCR was performed using a ABI Step-One Plus system. PCR reactions were carried out in 10-μl reactions using TransStart Top Green qPCR SuperMix (TransGen Biotech, AQ131–02) and 0.5 mM specific primers. The primers used for PCR are shown in the Additional file 1: Table S1.

Cytosolic and nuclear protein extraction was performed using a Nuclear/Cytosol Fractionation Kit (BioVision, K266-25) according to the manufacturer’s instructions. Each fraction was tested for the presence of the cytosolic marker GAPDH and the nuclear marker Lamin A/C by western blotting as appropriate.

The data are reported as the mean and SEM. All statistical analyses were performed using Prism version 5.0 for Windows. The data were analyzed for the normality of distribution using the Kolmogorov–Smirnov normality test (n > 50) or the Shapiro–Wilk normality test (n ≤ 50). P > 0.05 indicated that the data were normal. The Pearson correlation was performed on the normality distribution and the variance similar data, and the Spearman correlation was performed on the stratified or the variance heterogeneity data. Fisher’s exact test was used for analyzing the proportions between two groups. An unpaired t test was used to compare between two groups. One-way ANOVA was performed for comparisons among more than two groups. P < 0.05 (two-tailed) was considered statistically significant. P < 0.05 was marked as ‘*’, P < 0.01 was marked as ‘**’ and P < 0.001 was marked as ‘***’.

A previous study by our group performed a tissue microarray analysis of 539 colorectal tumor tissues [16]. We reanalyzed the correlation between RHBDD1 and metastatic parameters in the same cohort recently. We assigned intensity score 0 or 1 as negative, and 2 or 3 as positive. In 539 patients, the percentage of patients with positive RHBDD1 expression is higher (P < 0.001) among those with lymphatic metastasis than patients without lymphatic metastasis (Table 1), and the intensity score of RHBDD1 is also significantly higher (P < 0.001) in patients with lymphatic metastasis (Fig. 1a). A total of 146 patients in the cohort have records regarding the presence of local recurrence, implantation metastasis or distal metastasis. Among this patient subset, the percentage of individuals positive for RHBDD1 expression is higher (P < 0.05, one-tailed, other P values in the article are all two-tailed) in patients with distal metastasis than in patients with local recurrence and implantation metastasis (Table 1), and the intensity score of RHBDD1 is also significantly higher (P < 0.05) in patients with distal metastasis (Fig. 1b). Therefore, RHBDD1 expression positively correlates with CRC metastatic parameters.

We analyzed the protein expression profile of RHBDD1 in 4 CRC cell lines. HCT-116, DLD1, and HCT-8 cells have higher levels of RHBDD1 protein expression (Additional file 2: Figure S1A), and CACO-2 cells have lower levels of RHBDD1 protein expression (Additional file 2: Figure S1A); thus, these cell lines were employed for further study. HCT-116, HCT-8, and DLD1 cells showed decreased migration (P < 0.05, P < 0.01, P < 0.001) (Fig. 2a, b) and invasion (P < 0.01, P < 0.05, P < 0.05) (Fig. 2a, b) upon RHBDD1 knockdown. CACO-2 cells showed increased migration (P < 0.001) and invasion (P < 0.01) when RHBDD1 was ectopically over-expressed (Fig. 2c, d). We used an RHBDD1 mutant stable cell line, which degraded endogenous RHBDD1 via the proteasome [16], to analyze its metastatic ability. RHBDD1-mutant HCT-116 cells showed decreased migration (P < 0.01) and invasion (P < 0.05) abilities (Fig. 2e, f). The RHBDD1 mutant cell line was separately infected with either RHBDD1 lentivirus or control lentivirus to obtain the RHBDD1 rescued stable cell line. This stable rescue clone showed increased migration (P < 0.05) and invasion (P < 0.001) abilities compared to the control clone (Fig. 2g, h). These data indicate that RHBDD1 can promote CRC cell metastasis in vitro. Then, we employed an in vivo imaging system to analyze whether RHBDD1 can promote CRC metastasis in vivo. After injection of HCT-116 cells with stable RHBDD1 knockdown via the caudal vein, we observed that pleural metastasis was significantly less severe (P < 0.01) compared to that observed upon injection with control cells (Fig. 2i, j).

We conducted RNA-Seq analyses to find the possible signaling pathway by which RHBDD1 promotes CRC metastasis. RNA from either control HCT-116 cells or HCT-116 cells with RHBDD1 knockdown were isolated and analyzed. Differential expression genes in the two groups were analyzed and enrichened in KEGG pathway database. To elucidate the regulatory mechanism of the signaling transduction of metastasis, we focused on the signal transduction pathway sub-categories from the KEGG pathway database. The pathways in this category were ranked, and the top 10 are shown in Fig. 3a. As shown in the figure, the MAPK signaling pathway had the largest differential gene counts, which was already proven to be influenced by RHBDD1 to regulate CRC growth [16]. These data showed the reliability of the RNA-Seq results. The Wnt signaling pathway had the second largest differential gene counts, which showed the possible role of this pathway in regulating CRC metastasis. We used a dual-luciferase reporter assay to further verify the KEGG enrichment results. The TCF/LEF binding regions were used. The results showed that the canonical Wnt signaling pathway exhibited significant decreases (P < 0.001) in activity when RHBDD1 is knocked down (Fig. 3b). To further validate the Wnt signaling pathway, we conducted the TOP/FOP flash reporter assay. The TOP flash plasmid has 7 TCF/LEF binding sites and the FOP flash plasmid has 6 mutated TCF/LEF binding sites. The ratio of TOP and FOP luciferase activity is analyzed. The results showed that RHBDD1 knockdown can decrease (P < 0.05) pathway activity (Fig. 3c), stable mutant RHBDD1 can decrease (P < 0.05) pathway activity (Fig. 3c), whereas rescuing wild-type RHBDD1 expression can restore (P < 0.01) pathway activity (Fig. 3d).

The core protein in the Wnt signaling pathway is β-catenin. In addition, the primary method of regulating β-catenin is phosphorylation. Because p33/37/41/45-β-catenin, p552-β-catenin and p675-β-catenin are directly associated with Wnt-signaling pathway activity, we analyzed the total and phosphorylated protein levels of β-catenin. The results (Fig. 3e) showed that the total-, p552- and p675-β-catenin, which were proved to activate β-catenin-mediated TCF signaling, decreased in HCT-116 cells with knockdown of RHBDD1 than the control group; while p33/37/41-β-catenin, which led to β-catenin degradation by the proteasome, increased in HCT-116 cells with knockdown of RHBDD1 than the control group. In CACO-2 cells, the expression pattern of all β-catenin was reversed upon over-expressed RHBDD1. The β-catenin expression pattern influenced by RHBDD1 is coincident with the TOP/FOP analysis. Thus, we confirmed that RHBDD1 can regulate Wnt signaling pathway activity and the protein levels of β-catenin.

A critical step of Wnt signaling pathway activation is the translocation of β-catenin from the cytosol to the nucleus. In the nucleus, it can form a complex with TCF/LEF to initiate transcription of downstream genes [11]. We performed cytosolic/nuclear protein isolation and analyzed the distribution of total and phosphorylated β-catenin between the cytosol and the nucleus. In HCT-116 cells with knockdown of RHBDD1, the total and phosphorylated β-catenin expression levels were all decreased in either the cytosol or the nucleus, while the percentages of the cytosol and the nucleus remained relatively stable (Fig. 4a). It seemed that RHBDD1 can influence the overall protein expression of total-, p552- and p675-β-catenin but cannot influence the translocation process. CHIR99021 can potently inhibit GSK3α/β and stimulation with this molecule can protect β-catenin from ubiquitin-associated degradation. We stimulated HCT-116 cells with CHIR99021 after knocking down RHBDD1, and the TOP/FOP flash reporter assay results showed that Wnt signaling pathway activity was still decreased significantly (P < 0.01) compared with stimulated control (Fig. 4b), with a slightly recovery of 7.2% compared with the DMSO treated group. We stimulated HCT-116 cells with CHIR99021 after knocking down RHBDD1, and conducted western blot analyses of total and phosphorylated β-catenin expression after cytosolic and nuclear isolation (Fig. 4c). The results showed that CHIR99021 can rescue β-catenin expression in either the cytosol or the nucleus than the control group. Fig. 4b and Fig. 4c indicate that total-β-catenin is not the key effector of RHBDD1-mediated regulation of the Wnt signaling pathway activity. At the meantime, the levels of ser552 and ser675 phosphor-β-catenin kept decreasing (Fig. 4c) after knocking down RHBDD1 compared with the control group, which may result in the decrease of pathway activity.

We established β-catenin constructs with point mutations of serine 552 and 675 to the phosphomimetic aspartic acid [24], positive mutation of phosphor-β-catenin at residues 552 and 675. HCT-116 cells with either mutant or wild-type RHBDD1 were transfected with control, S552D, or S675D plasmids (Fig. 4d). The TOP/FOP flash reporter assay showed that upon transfection with control plasmid, the pathway activity decreased significantly (P < 0.05) in HCT-116 cells with mutant RHBDD1 compared with in wild-type HCT-116 cells, which was the repetition of Fig. 3c. However, upon transfection with either S552D or S675D constructs, the pathway activity increased significantly in either HCT-116 cells with mutant RHBDD1 (P < 0.001, P < 0.01) or in wild-type HCT-116 cells (P < 0.05, P < 0.01). The results (Fig. 4d) showed that in HCT-116 cells with mutant RHBDD1, upon transfection with either S552D or S675D constructs, the pathway activity can be recovered to levels comparable to those in wild-type HCT-116. Additionally, In HCT-116 cells with mutant RHBDD1, the migration and invasion assays (Fig. 4e) showed that the metastatic ability can be recovered significantly by over-expressing the S552D (P < 0.001, P < 0.01) and S675D (P < 0.001, P < 0.05) constructs. Collectively, these data proved that the pathway activity is directly associated with p552- and p675-β-catenin and that RHBDD1 regulates the pathway activity mainly through these two phosphorylated residues on β-catenin.

The Wnt signaling pathway regulates EMT and ISCs homeostasis. Lauren Averett Byers’ group reported a pan-cancer EMT signature derived from 11 cancer types in 2016 [25]. They used four established EMT markers—CDH1 (epithelial marker, E type), CDH2 (mesenchymal marker, M type), VIM (M type), and FN1 (M type)—as seeds to develop a pan-cancer EMT signature on the basis of TCGA pan-cancer RNA-Seq data. The derived 77 signature genes were identified in our RNA-Seq results, and a heatmap analysis was constructed (Fig. 5a). Seventy-four genes in the signature with calculated FPKM as indicated in the Methods section were shown in the heatmap. The RNA sequencing was performed in triplicate in HCT-116 cells transfected with RHBDD1 Si-RNA or control scramble Si-RNA. We assigned the adjusted P value (padj) < 0.05 found by DESeq as differentially expressed. Then differentially expression genes in the EMT signature were identified and analyzed (Fig. 5b). In mesenchymal genes, 13 genes decreased significantly upon knocking down RHBDD1, with 4 genes have log2FoldChange < − 1; 7 genes increased significantly, with 1 gene have log2FoldChange > 1. In epithelial genes, 12 genes increased significantly upon knocking down RHBDD1, and 5 genes decreased significantly. The proportions of up and down regulated differential expression genes in mesenchymal genes and in epithelial genes were significantly different (P < 0.05) (Fig. 5c). The results showed that knockdown of RHBDD1 led to a decreasing trend in the mesenchymal gene expression pattern. In addition, the epithelial gene expression pattern tended to increase. This result indicated that RHBDD1 was positively correlated with EMT and could maintain the mesenchymal phenotype in colorectal cell lines; this mesenchymal phenotype endowed the cells with metastatic potential. Eduard Batlle reported an ISC signature in 2011 [26]. They isolated ISCs using EphB2, which is expressed at high levels in the ISC membrane. Then, they derived an ISC signature based on EphB2 expression level, thus creating a list of 71 probes (54 annotated genes). We identified the 54 genes in our RNA-Seq results and constructed a heatmap analysis (Fig. 5d). Forty-seven genes in the signature with calculated FPKM were shown in the heatmap. Differential expression genes in the ISC signature were also identified and analyzed (Fig. 5e). Nine genes decreased significantly upon knocking down RHBDD1, with 2 genes have log2FoldChange < − 1; 7 genes increased significantly, but none of them met log2FoldChange > 1. The result showed that RHBDD1 was positively correlated with a stem-like phenotype, which may be a result of CRC cell dedifferentiation via EMT.

We already proved that RHBDD1 can influence gene signatures downstream of the Wnt signaling pathway; thus, we further explored the exact target gene of the Wnt signaling pathway that is regulated by RHBDD1. ZEB1 is reported to be a Wnt target gene [8]. It is a potent EMT activator [27] and can regulate cell stemness [28]. Based on the results in Fig. 5, we discovered the role of ZEB1 in RHBDD1-mediated regulation of colorectal metastasis. In HCT-116 cells, Knockdown RHBDD1 can decrease (P < 0.01) the mRNA (Fig. 6a) level of ZEB1; knockdown or over-expression of RHBDD1 can decrease or increase, respectively, the protein (Fig. 6b) levels of ZEB1. RHBDD1 Cas9-KO HCT-116 cells showed a decreased protein level of ZEB1 relative to the level in wild-type cells (Fig. 6b). HCT-116 cells with rescued RHBDD1 expression showed recovery of ZEB1 protein expression relative to the level in control cells (Fig. 6b). We transfected the empty vector, S552D, or S675D constructs in HCT-116 cells with RHBDD1 knockdown and observed that the protein levels of ZEB1 were rescued relative to the levels in wild-type HCT-116 cells upon transfection with either S552D or S675D constructs (Fig. 6c). The same result was obtained in HCT-116 cells with stable expression of mutant RHBDD1 (Fig. 6d). These results showed that RHBDD1 can regulate ZEB1 at the protein level via p552 and p675 β-catenin.

We validated the results obtained from cell models in primary tissues. CRC tumor tissues from 8 patients were analyzed using western blotting (Additional file 2: Figure S1B). The result was scanned for gray scale, and the statistical analysis showed the protein expression levels of RHBDD1 and ZEB1 have a high degree of positive correlation (Spearman r: 0.8571, P = 0.0107) (Fig. 6e). Furthermore, we conducted a tissue microarray to explore the correlation between these two proteins in 71 colon cancer tissues. Fig. 6f shows that the protein expression levels of ZEB1 are coincident with RHBDD1 expression in different tissues. The statistical analysis showed that RHBDD1 and ZEB1 exhibited a moderately positive correlation in 71 colon cancer tissues (Spearman r: 0.5312, P < 0.0001) (Fig. 6g).