RNA‐binding protein RBM24 represses colorectal tumourigenesis by stabilising PTEN mRNA

Abstract Background RNA‐binding motif protein 24 (RBM24) functions as a splicing regulator, which is critical for organ development and is dysregulated in human cancers. Here, we aim to uncover the biological function of RBM24 in colorectal tumourigenesis. Methods Xenograft tumour model, Rbm24 knockout and Apcmin/+ mouse models were utilised. Colorectal cancer cells overexpressing or silencing RBM24 were established. RNA immunoprecipitation (RIP) assay was conducted to detect protein‐RNA associations. Gene expression was measured by immunohistochemistry, western blotting, or quantitative PCR (qPCR). Results Rbm24‐knockout mice developed spontaneous colorectal adenomas with lower expression of phosphatase and tensin homolog (PTEN). Immunohistochemical staining for the proliferation markers Ki‐67 and pHH3 and BrdU assay showed intestinal hyperplasia in Rbm24‐knockout mice compared to wild‐type mice. RBM24 expression in colorectal adenoma tissues of Apcmin/+ mouse was downregulated compared with adjacent normal samples and was positively correlated with PTEN expression. In vitro, RBM24 overexpression suppressed cell proliferation, migration, invasion and increased sensitivity to 5‐FU or cisplatin in CRC cells. Mechanistically, RBM24 maintained PTEN mRNA stability by directly binding to the GT‐rich region at positions 8101–8251 in the 3′‐UTR of PTEN mRNA, prolonging the half‐life of PTEN mRNA, thereby increasing PTEN expression. Hence, low expression of RBM24 downregulated PTEN mRNA, causing the activation of PI3K‐Akt signalling in CRC cells. Furthermore, RBM24 expression in CRC tissues was lower than adjacent normal samples. RBM24 expression was positively correlated with PTEN expression and negatively correlated with Ki‐67 level. CRC patients with high RBM24 expression had a favourable outcome. Conclusions Taken together, RBM24 expression is markedly lower in colorectal tumours than in para‐carcinoma tissues. Rbm24‐knockout mice develop spontaneous colorectal adenomas. RBM24 directly binds and stabilises PTEN mRNA, which could cause the suppression of CRC cell proliferation, migration and invasion, thereby repressing colorectal tumourigenesis. These findings support the tumour‐suppressive role of RBM24. Targeting RBM24 holds strong promise for the diagnosis and treatment of CRC.

3. Decreased PTEN expression mediated by RBM24 contributes to the activation of the PI3K/Akt signalling pathway, promoting colorectal cancer (CRC) development.
which could cause the suppression of CRC cell proliferation, migration and invasion, thereby repressing colorectal tumourigenesis. These findings support the tumour-suppressive role of RBM24. Targeting RBM24 holds strong promise for the diagnosis and treatment of CRC.

K E Y W O R D S
colorectal cancer, mRNA stability, PTEN, RBM24, tumourigenesis

BACKGROUND
Colorectal cancer (CRC) causes approximately 900 000 deaths globally each year. 1 Mounting evidence suggests that genetic mutations, epigenetic modifications, chronic inflammation, diet, lifestyle and intestinal flora are the main risk factors for CRC. 2,3 CRC patients diagnosed at an early stage have a 90% probability of being cured by surgical resection. 4 However, due to insidious onset and unrecognised clinical symptoms, most patients with CRC are diagnosed when they are already in the advanced stage, resulting in poor treatment outcomes. 4,5 Although the pathogenesis of CRC has been extensively studied, the underlying mechanisms of CRC initiation and progression have not been fully elucidated. Tumour suppressor gene mutation or inactivation contributes to colorectal carcinogenesis, 6,7 for example, mutations in APC, 8 P53, 9 and PTEN. 10 These mutations show significant association with the development of CRC. Although previous studies have shown that P53 is mutated in approximately half of human tumours and PTEN mutation occurs in more than 20%, a substantial proportion of individuals without these mutations will also develop CRC. 11,12 Hence, a better understanding of the precise pathogenic mechanisms involved in colorectal tumourigenesis is warranted for the prevention, diagnosis and treatment of CRC.
RNA-binding protein (RBP) contains mRNA-binding domain(s), which can directly bind to intracellular mRNA and further regulate mRNA maturation, transport, localisation and translation. 13,14 RBPs play important role in individual development, homeostasis, pathological or physiological processes of cancer and other diseases. [15][16][17] Via specific RNA-binding domains, RBPs can recognise specific RNA sequences. 16 RBM24, a member of the RBP family, has a highly conserved RNA-recognition motif (RRM) and four exons that encode a 236 amino-acids protein (24.7 kDa). RBM24 gene is located on human chromosome 6 and mouse chromosome 13. RBM24 has been reported to be significantly upregulated during myocardial differentiation of embryonic stem cells. 18,19 Previous studies have demonstrated that RBM24 regulates alterna-tive splicing switches 18,20 and also directly binds to the GU(A)GUGU site on RNA, thereby affecting mRNA maturation or translation. 20 RBM24 regulates the mRNA stability of P53, P21 and P63, which in turn influence protein translation. [21][22][23][24] Recently, it has been revealed that RBM24 has an effect on eye development by adjusting the poly(A) tail length and translation efficiency in the lens. 25 These studies indicate that RBM24 exhibits extensive regulatory functions. Therefore, its biological activity in other organs deserves further investigation.
During the construction process of our Rbm24 knockout mouse model, we unexpectedly detected a certain number of adenomas within the intestine of tamoxifeninduced Rbm24-knockout mice. Herein, we performed experiments both in vitro and in vivo to understand the mechanisms underlying the involvement of RBM24 in colorectal tumourigenesis. Our findings showed that Rbm24 knockout promoted colorectal tumourigenesis. RBM24 expression in human colorectal specimens and mouse colorectal adenoma tissues was significantly lower than in controls. RBM24 stabilised PTEN, a tumour suppressor gene, mRNA and controlled the proliferative and metastatic capabilities of CRC cells. RBM24 exerts a tumour suppressive role in CRC and targeting RBM24 holds strong promise for CRC diagnosis and treatment.

Patients and colorectal tissues
This study was approved by the Ethics Committee of Xiamen University and written informed consent was obtained from all patients (Approval no. XDYX2021008). A total of 36 CRC samples and paired adjacent normal colorectal tissues samples were collected. All patients enrolled in this study haven't received drug treatment, chemo-or radiotherapy prior to surgical resection. Samples were prepared for western blotting, qPCR and immunohistochemistry (IHC) assays. The patient details are listed in Table 1.

Animal experiments
Apc min/+ mice were obtained from Shanghai Model Organisms Center (N = 18). C57BL/6 mice were used as the wild-type (WT) group (N = 18). UBC-Cre/ERT2 mice were generously donated by Professor Chen Yongxiong (Ophthalmology Center, School of Medicine, Xiamen University). Rbm24 loxp/loxp mice were generated by inserting two LoxP sites on the flank of second and third exons of the Rbm24 gene (Gene ID: 666794). 20 Then, Rbm24 loxp/loxp mice were bred with UBC-Cre/ERT2 mice to generate CreERT2-Rbm24 loxp/loxp mice. 26 To generate Rbm24knockout mice, CreERT2-Rbm24 loxp/loxp mice were treated with tamoxifen (100 mg/kg/d, ip) for 5 consecutive days (2month-old; N = 32). C57BL/6 mice given the same dose of tamoxifen were used as the WT control group (N = 32). Knockout efficiency was validated by qPCR and Western blotting analysis. For xenograft tumour models, female BALB/c nude mice (6-week-old) were obtained from Shanghai Model Organisms Center. The nude mice (N = 6 mice per group) were injected subcutaneously with 2 × 10 6 shRBM24-HCT116 cells or 2 × 10 6 OERBM24-HCT116 cells or 2 × 10 6 control cells (suspended in 100 μl of sterile PBS) at the right flanks. The long (L) and short diameter (W) of tumours were detected with callipers every week and analysed by the formula: volume = 0.5 × L × W 2 . Mice were killed on day 42 under isoflurane anaesthesia. Next, the transplanted tumours were stripped and photographed. For lung metastasis model, 1 × 10 5 shRBM24-HCT116 cells or 1 × 10 5 OERBM24-HCT116 cells or 1 × 10 5 control cells (diluted in 50 μl of sterile PBS) were injected via tail vein injection. Metastatic lung nodules were observed after 6 weeks.
All mice were kept in specific pathogen-free (SPF) conditions with 12-h light/12-h dark cycle and free access to food and water. All the experimental procedures were operated in accordance with the guidelines and regulations of animal experimentation approved by the Experimental Animal Center of Xiamen University. The Experimental Animal Committee of Xiamen University approved all animal procedures (Approval ID: SCXK2013-0006).

2.3
Cell culture and construction of cell line silencing or overexpressing RBM24 Human HCT116, SW480 and LoVo cell lines were purchased from the Cell Bank of the Chinese Academy of Sciences (Shanghai, China). CRC cells were cultivated with DMEM medium supplemented 10% FBS (Gibco) and 1% penicillin-streptomycin (Gibco) at 37 • C under 5% CO 2 in saturated humidity conditions. To establish RBM24-silenced cell lines, a specific short hairpin RNA (shRNA) of RBM24 (Table 3) was designed and packaged into lentiviral particles. After 48 h of infection, the cells were subjected to 1 μg/ml of puromycin for more than 72 h. RBM24 expression was examined using qPCR and Western blotting analysis.
To construct RBM24-overexpressing cell lines, lentiviruses overexpressing RBM24 were established. Next, the cells were infected with the lentiviruses and stable cell populations were selected for puromycin as described above. The efficiency of RBM24 overexpression was measured using qPCR analysis and western blotting assays.

Preparation of tissue sections of mouse intestine
Mice were sacrificed by cervical dislocation. Mouse colon and parts of the small intestine were excised and washed using pre-cooled PBS to ensure that the contents of the intestine were rinsed clean. Afterwards, the intestine was flattened and photographed. Thereafter, mouse intestinal tissues were fixed in 4% buffered paraformaldehyde for 48 h; rinsed 3 times with PBS for 5 min each time; dehydrated with 10% sucrose solution, followed by gradient dehydration using 20% and 30% sucrose solution in sequence. Next, the OCT-embedded tissues were snapfrozen to -80°C. Five-micrometre-thick sections were prepared for immunofluorescence.

Immunochemistry
Paraffin-embedded tissue sections were treated with H 2 O 2 solution at RT for 10 min. Following washing with PBS, antigen retrieval was conducted as described above. Next, the sections were blocked with 5% goat serum in PBS at RT for 1 h and then incubated with primary antibodies (anti-RBM24, Abcam, Cat no: ab94567, 1:300; anti-PTEN, Santa Cruz, Cat no: SC-7974, 1:300; or anti-Ki-67, Abcam, Cat no: ab15580, 1:300) overnight at 4 • C. Subsequently, the sections were subjected to HRP-conjugated secondary anti-body for 1 h at RT following washing with PBS. Color was developed using DAB solution at RT for 5 min. Tissue sections were counterstained with haematoxylin staining for 30 s, followed by dehydration in differential alcohol gradients. Sections were coverslipped with neutral balsam and then observed under microscope (Olympus, Japan).

Western blotting
Tissues or cells were treated with lysis buffer containing 1% PMSF. Protein quantitation was determined using the bicinchoninic acid assay.

Quantitative polymerase chain reaction (qPCR)
Total RNA was prepared using TRIzol reagent (Life Technology) following the guidance. EasyScript R TA B L E 2 Primary antibodies and corresponding secondary antibodies used in this study and their optimal working dilutions First-Strand cDNA Synthesis SuperMix (Beijing TransGen Biotech Co., Ltd, China) and TransScript R One-Step RT-PCR SuperMix (Beijing TransGen Biotech Co., Ltd, China) were used to synthesise cDNA and for qPCR analysis following the manufacturer's instructions, respectively. Furthermore, the target gene amplification was confirmed using agarose gel electrophoresis. All primers are listed in Table 3. Gene expression was normalised to the expression level of housekeeping gene GAPDH.

Haematoxylin and eosin (HE) staining
Tissue sections, as described above, were stained with haematoxylin at 60 • C for 60 s, subjected to 1% hydrochloric acid ethanol treatment for 3 s, rinsed in running tap water, stained with eosin for 60 s, rinsed with distilled water for 2 s and further dehydrated with ethanol and xylene in order. The slides were sealed with neutral balsam. Images were photographed under an optical microscope (Olympus, Japan).

Cell Counting Kit 8 (CCK-8) assay
CRC cells overexpressing or silencing RBM24 (1 × 10 3 cells/well) were seeded into 96-well dishes and cultured overnight. CCK-8 regent was added into cells at each time point, followed by incubation for 2 h. Cell viability was assessed by detecting the absorbance at 450 nm using a Microplate Reader (Bio-Rad). Cell-free medium with CCK-8 reagent was included as the negative control.

Colony-formation assay
To assess the role of RBM24 in CRC cells, 6-well plates were seeded with 1 × 10 3 cells and cultured for 10 days. The culture medium was changed every 2 days. Subsequently, the supernatant was discarded to remove the floating cells. After fixing with 75% ethanol at RT for 10 min, the cells were stained with 1% crystal violet (CV) for 30 min. Ten fields were reviewed in each well and were randomly selected under a microscope (Olympus, Japan). The number of colonies (≥50 cells) and the average number of cells per clone were calculated.

TA B L E 3
The sequences of all primers used in this study

Transwell migration/invasion (with Matrigel) assays
Briefly, 5 × 10 5 cells per well were plated into the upper chamber. For transwell invasion assay, the upper chamber was pre-coated with 5 mg/ml of Matrigel. The lower transwell chamber contained 500 μl of culture medium. Twelve hours later, the cells were stained using 1% crystal violet as described above. The migrated or invaded cells were counted (five fields of view for each group).

Cell apoptosis and cell cycle
Cell apoptosis and cell cycle were measured by flow cytometric analysis. Cell apoptosis was tested by Annexin V-FITC and PI double-staining (Beijing Solarbio Science and Technology Co., LTD., China). Cell cycle was evaluated by PI staining as suggested by manufacturer's protocol. Data were analysed using Flowjo 7.6 software. All experiments were independently performed a minimum of three times.

Wound-healing assay
When the cells in each group reached 95% confluence, the confluent cell monolayer was scratched with a sterile pipette tip (0 h). Afterwards, the wound closure at different time points was monitored under the microscope (Olympus, Japan).

RNA immunoprecipitation assay (RIP)
Cells transfected with RBM24-Flag plasmids were harvested 48 h post-transfection and then fully lysed using RIP lysis buffer. Anti-Flag beads (50 μl) were pre-washed five times with NT2 buffer, then 50 μl of magnetic beads were added into 800 μl of cell lysate, following incubation overnight at 4 • C with shaking. Magnetic beads were precipitated by magnetic separation and quickly washed twice with NT2 buffer. The immunoprecipitated pellet was collected and prepared for western blotting and qPCR analysis. Magnetic beads coated with anti-IgG antibodies were used as negative controls.

Statistical analysis
Data, from a minimum of at least three independent experiments, were analysed using SPSS (version 21.0, IBM Corp.) and expressed as the mean ± standard deviation (SD). Two-tailed Student's t-test, one-way analysis of variance (ANOVA), post hoc least significant difference (LSD) test, Cox test, or Pearson's correlation test was applied for data analysis. GraphPad Prism 5.0 software (GraphPad Software) was used for graphing. Asterisk represented a p value less than .05.

Rbm24-knockout mice develop spontaneous colorectal adenomas
Previously, Yang et al. confirmed that systemic Rbm24 knockout resulted in embryonic lethality. 27 Hence, Rbm24-inducible knockout mouse model was constructed (Figure S1A-D). 28 Tamoxifen (100 mg/kg daily for 5 days) was injected intraperitoneally into 8-week-old CreERT2-Rbm24 loxp/loxp mice (N = 32) to generate Rbm24 conditional knockout mice. Wild type (WT) mice were administrated with the same dose of tamoxifen ( Figure 1A). Knockout efficiency was validated using immunofluorescence, western blotting and qPCR analysis ( Figure 1B-D). We found that BrdU-positive cells (Figure 1E) in the colon and the expression levels of Ki-67 and pHH3 protein in the intestine ( Figure S1E-H) of Rbm24knockout mice increased significantly compared with the WT mice 15 days after tamoxifen administration. Besides, colonic wall thickening in the Rbm24-knockout mice was observed by HE staining (Figure 1F). At 6 months after Rbm24 knockout, an enlargement of the colon in mice was observed ( Figure 1G); the colorectal segment was strikingly longer than that of WT mice ( Figure 1H) and the number of adenomas increased significantly ( Figure 1I, J). Immunofluorescence also confirmed a significant increase in Ki-67-positive ( Figure 1K) and pHH3-positive cells ( Figure 1L) in adenoma tissues. No adenoma was found in WT mice at 6 months. These data suggest that Rbm24 knockout promotes the development of colorectal adenomas in mice.

RBM24 overexpression suppresses malignant behaviours in CRC cells
To understand the role of RBM24 in CRC, we performed cell proliferation assay, CCK-8 assay, cell cycle analysis and colony formation assay. We found that RBM24 overexpression repressed cell proliferation, contributed to cell cycle arrest at the G0/G1 phase and reduced colony-forming capability of human CRC cells (HCT116, SW480 and LoVo) (Figure 2A-I). To further investigate the two typical characteristics (invasion and metastasis) of cancer cells, 29,30 we performed wound-healing and transwell assays. We found enforced expression of RBM24 suppressed migration and invasion of CRC cells as compared to the matched controls ( Figure 2J-O). In addition, RBM24 overexpression restrained tumour growth and lung metastasis in vivo (Figure 2P). Conversely, RBM24 knockdown yielded opposing results ( Figure 3A-P). These results indicate that RBM24 suppresses CRC cell growth, both in vitro and in vivo.

RBM24 regulates PTEN expression and downstream signalling of PTEN
In our preliminary experiments, we observed an association between the expression of RBM24 and PTEN. PTEN, an important negative regulator of the PI3K pathway, is a classical tumour suppressor in various type of cancers. We assumed that RBM24 may affect PTEN expression and its downstream PI3K/Akt signalling pathway. Western blotting and qPCR analysis revealed that PTEN expression was significantly decreased and positively correlated with RBM24 expression in tumour tissues of RBM24knockout mice. Akt is a key molecule of PI3K/Akt signalling. In present study, Akt phosphorylation was significantly upregulated in tumour tissues of RBM24-knockout mice, while there was no difference in the expression of total Akt in tumour tissues of RBM24-knockout mice compared with WT mice. Similarly, the expression of Matrix metalloproteinases 2 and 9 (MMP2 and MMP9), downstream molecules of the PI3K/Akt pathway, was markedly increased and correlated inversely with RBM24 expression, whereas P21 expression showed a significant positive correlation with RBM24 expression in tumour tissues of RBM24-knockout mice in comparison to WT mice (Figure 4A-C).
Consistent with the above in vivo findings, RBM24 overexpression resulted in the upregulation of PTEN as well as a reduction of Akt phosphorylation in HCT116 cells, but had no effect on the expression of Akt ( Figure 4D); meanwhile, MMP-2 and MMP-9 expression was significantly downregulated in CRC cells compared to the control. In contrast, RBM24 knockdown produced the opposite effect ( Figure 4E).
Further, the decreased expression of MMP-2, MMP-9 and phosphorylated Akt in CRC cells with RBM24 overexpressing could be reversed by exposure to the PTEN specific inhibitor SF1670 ( Figure 4F). Additionally, RBM24 knockdown contributed to the significant downregulation of PTEN expression, increase in MMP-2 and MMP-9 expression and Akt phosphorylation compared to the control. However, Akt-specific inhibitor MK-2206 or PI3K inhibitor PI3K-IN-6 treatment could reverse the results induced by RBM24 knockdown in HCT116 cells (Figure 4G, H). Together, these results indicate that RBM24 Vimentin and PTEN were detected using qPCR (left) and Western blotting analysis (right). Data were expressed as the mean ± SD from at least three independent experiments and analysed using one-way ANOVA with post hoc LSD test. *p<.05, **p<.01, ***p<.001 attenuates the PI3K/Akt signalling pathway by regulating PTEN expression, causing the suppression of CRC cell proliferation, migration and invasion.

PTEN inhibitor SF1670 abolishes the inhibitory effect of RBM24 overexpression on CRC cells
Given our results showing that RBM24 expression was positively associated with PTEN expression and negatively influenced the PI3K/Akt signalling pathway, CRC cells overexpressing RBM24 were subjected to treatment with PTEN inhibitor SF1670. As expected, overexpression of RBM24 in HCT116 cells followed by treatment with 1 μM of SF1670 reversed the weakened cell proliferative, migratory and invasive capabilities induced by RBM24 overexpression ( Figure 4I-K). Western blotting and qPCR analysis demonstrated that RBM24 overexpression promoted P21 and E-cadherin expression, but attenuated the expression of CDK4, CyclinD1, MMP2/9 and Vimentin. However, the expression levels of the above-mentioned genes were reversed by PTEN inhibitor SF1670 ( Figure 4L, M).
In addition, RBM24 knockdown in combination with SF1670 treatment had a synergistic effect in promoting CRC cell migration and invasion, in upregulating MMP2/9 and Vimentin expression and in reducing Ecadherin expression compared to the untreated cells (Figure S2A-F). Besides, proliferative, migratory and invasive capabilities of RBM24-overexpressing HCT116 cells were restored by downregulation of PTEN expression using siRNA, whereas upregulation of PTEN expression by transfecting pcDNA3.1-PTEN in RBM24-knockdown HCT116 cells suppressed the malignant phenotypes of RBM24knockdown HCT116 cells ( Figure S2G-J).
To determine the effect of RBM24 on PI3K/Akt signalling pathway, PI3K inhibitor PI3K-IN-6, or Akt inhibitor MK-2206 was used. We found that PI3K-IN-6 or MK-2206 treatment further suppressed the proliferation, migration and invasion of RBM24-overexpressing CRC cells; meanwhile, a synergistic effect of RBM24 overexpression in combination with either PI3K-IN-6 or MK-2206 treatment on gene expression was observed in comparison with untreated controls ( Figure S3).
Conversely, RBM24 knockdown facilitated the malignant behaviour of HCT116 cells compared to the control; however, PI3K-IN-6 or MK-2206 treatment could partially reverse the phenotypes induced by RBM24 knockdown. Meanwhile, PI3K-IN-6 or MK-2206 treatment restored the expression of the above-mentioned genes in RBM24knockdown HCT116 cells ( Figure S4). Naturally, treatment with SF1670, PI3K-IN-6 or MK-2206 alone influenced the proliferative, migratory and invasive capacities of CRC cells, respectively ( Figure S5A-F). These findings suggest that RBM24 exerts its biological function by controlling PTEN expression and the PI3K/Akt signalling pathway. ATM (ataxia-telangiectasia, mutated) and ATR (ATM and Rad3-related), belonging to protein kinases, are the sensors of DNA damage. 31,32 Gamma-H2AX (γH2AX), a phosphorylated protein, is a marker of DNA double-strand breaks. 33,34 In this study, western blotting analysis revealed that 5-FU or cisplatin treatment significantly upregulated the expression of cleaved-caspase3 and cleaved-PARP (two markers of apoptotic induction) and the phosphorylation of ATM, ATR and H2AX in RBM24 overexpressing HCT116 cells as compared to the cells treated with 5-FU or cisplatin alone; however, the above-mentioned results could reverse by SF1670 treatment, or enhanced by MK-2206. Conversely, as compared to HCT116 cells treated with 5-FU or cisplatin alone, RBM24 knockdown significantly reduced the elevated phosphorylation levels of ATM, ATR and H2AX caused by 5-FU or cisplatin, while MK-2206 treatment restored the results induced by RBM24 knockdown ( Figures 5J, S5J and S6M). Overall, these data suggest that RBM24 expression is associated with resistance of CRC cells to 5-FU or cisplatin by regulating the PI3K/Akt signalling pathway. Considering the association among PI3K/Akt signalling pathway, tumour drug resistance and the regulation of PI3K/Akt signalling pathway by RBM24, we believe that RBM24 is closely related to CRC chemoresistance.

RBM24 directly binds to PTEN mRNA and enhances its mRNA stability
To understand the specific mechanism that RBM24 regulates PTEN expression in CRC, RBM24 was overexpressed or silenced in various types of CRC cells (HCT116, SW480, LoVo). Our findings revealed that expression levels of PTEN were correlated with that of RBM24 ( Figure 6A-C).
To determine whether RBM24 directly binds to PTEN mRNA, RNA immunoprecipitation (RIP) assay was performed. Data showed that PTEN mRNA co-precipitated with RBM24, indicating a direct interaction between RBM24 protein and PTEN mRNA ( Figure 6D, F, H). To further evaluate the effect of RBM24 on the stability of PTEN mRNA, CRC cells were treated with the transcriptional inhibitor actinomycin D and mRNA decay was assessed by qPCR analysis. RBM24 knockdown resulted in a decrease of the half-life of PTEN mRNA, whereas overexpression of RBM24 extended its half-life time (Figure 6E, G, I), suggesting that RBM24 stabilised PTEN mRNA. These data demonstrate that RBM24 promotes PTEN mRNA stability by directly binding to its 3′-UTR regions.
Considering that previous studies have proved that the RNA binding site of RBM24 is a GT-rich region, we further cloned the GT-rich region in the 3′-UTR region of PTEN mRNA into plasmid pcDNA3.1 expressing GFP (Figure 6J). Thereafter, pcDNA3.1-PTEN 3′-UTR and pcDNA3.1-RBM24 were co-transfected into 293T cells ( Figure 6K). Results from RIP and qPCR assays showed that among the 5 GT-rich regions in the 3′-UTR of PTEN mRNA, only the GT-rich regions at positions 8101-8251 were able to pull down GFP mRNA from total RNA ( Figure 6L). Therefore, these findings suggest that RBM24 enhances the stability of PTEN mRNA by directly binding to the GTrich region at positions 8101-8251 in the 3′-UTR of PTEN mRNA.

3.7
Rbm24 is downregulated in colorectal adenoma tissues of Apc min/+ mice APC suppresses CRC development by repressing the Wnt signalling pathway. 35 APC gene mutation and inactivation are often observed in CRC. Apc min/+ mice spontaneously develop intestinal adenomas, which is the classic animal model for studying CRC ( Figure 7A). In this study, we found that Rbm24, PTEN and P21 mRNA levels were significantly downregulated in tumour tissues of Apc min/+ mice compared to adjacent normal samples (N = 18); meanwhile, the expression levels of MMP-2 and MMP-9 were markedly elevated ( Figure 7B-F). Notably, when the protein expression of PTEN was downregulated, Akt phosphorylation was significantly increased and total Akt was not changed in tumour tissues of Apc min/+ mice, compared with para-cancer tissues ( Figure 7G). Besides, the expression levels of MMP-2, MMP-9 and P21, the downstream genes of PI3K/Akt pathway, was also significantly altered, of which MMP-2 and MMP-9 expression were increased, but P21 expression was decreased. Further correlation analysis showed a statistically significant positive correlation between the transcriptional levels of PTEN or P21 and Rbm24, an inverse correlation between MMP-2/MMP-9 expression and Rbm24 expression ( Figure 7H).

RBM24 is downregulated in human CRC tissue samples and low RBM24 expression is associated with poor prognosis in CRC patients
To further explore the role of RBM24 in human CRC, we examined RBM24 expression in human CRC tissues and paired para-cancerous samples. qPCR and western blotting analysis revealed that, compared with adjacent normal tissues, RBM24 expression in human CRC tissues was significantly reduced ( Figure 8A-C), an effect that was also associated with TNM staging ( Figure 8D) and overall survival (N = 36; Figure 8E). There was no correlation between RBM24 and other clinical characteristics, including age and gender of CRC patients. Immunohistochemistry (IHC) indicated that RBM24 and PTEN expression in human CRC samples was significantly lower than in paracancerous tissues (Figures 8F and S7). Besides, Ki-67, a proliferation marker, was significantly upregulated in CRC tissues compared to adjacent normal samples ( Figures 8F  and S8A). Correlation analysis showed a negative correlation between the IHC scores of RBM24 and Ki-67 (Figure 8G) and a positive correlation between the IHC scores of RBM24 and PTEN ( Figure 8H).
Further, qPCR results showed that mRNA expression of PTEN in human CRC tumour tissues was markedly downregulated ( Figure 8I) and positively correlated with mRNA expression of RBM24 ( Figure 8J). However, MMP-2 and MMP-9 expression in tumour tissues were significantly upregulated and both were negatively correlated with RBM24 expression; P21 expression was inversely correlated with MMP-2/MMP-9 expression ( Figure S8B-G). Western blotting analysis confirmed that PTEN and P21 expression were downregulated in tumour tissues; Akt phosphorylation was significantly upregulated and no change was observed in the expression level of total Akt ( Figure 8K). Similarly, by searching the GEPIA (Gene Expression Profiling Interactive Analysis) (http://gepia.cancerpku.cn/), GEO and TCGA datasets, we found that RBM24 was significantly downregulated in CRC tumour tissues ( Figure 8L). PTEN expression was positively correlated with RBM24 expression in CRC ( Figure S8H). Notably, ROC curves showed that RBM24 level could distinguish tumours from normal tissues ( Figure S8I). RBM24 expression level in colon adenocarcinoma (COAD) and rectum adenocarcinoma (READ) was inversely correlated with its promoter methylation level ( Figure S8J). Besides, there was no relationship between RBM24 expression and APC or KRAS mutation in CRC ( Figure S8K), indicating that the expression of RBM24 may be epigenetically affected in tumours by aberrant methylation.
Overall, schematic diagram of how RBM24 exerts it function in CRC tumourigenesis is represented (Figure 8 M). Briefly, RBM24 directly binds to the 3′-UTR of PTEN mRNA and maintains its mRNA stability, leading to the increased PTEN protein levels, thereby repressing PI3K/Akt signalling pathway. Nevertheless, abnormally low expression of RBM24 promotes degradation of PTEN mRNA, resulting in decreased PTEN protein expression, thereby relieving its inhibitory effects on PI3K/Akt pathway. Taken together, our findings suggest that RBM24 acts as a tumour suppressor gene in colorectal tumourigenesis.

DISCUSSION
Although a previous study has reported that RBM24 presents a certain correlation with Hirschsprung's disease, 36  Previous studies have shown that RBM24 also participates in cytoskeletal organisation, 20 providing new insight into the biological function of RBM24 in CRC. Cytoskeletal regulation is significantly associated with cell migration and motor ability. 37,38 Therefore, we assumed that RBM24 knockout could increase the migratory capacity of cells and favour cell migration, which is also one of the hallmark events of CRC development. 12 After overexpressing or silencing RBM24 expression, we found that the cell migratory and invasive abilities and the expression level of EMT-related genes (E-cadherin and vimentin) were significantly changed, that is, RBM24 expression negatively regulated CRC cell migration and invasion.
Apc min/+ mouse model is a classic animal model used in CRC research. The mutation of the APC gene con-tributes to the abnormal activation of the Wnt/β-catenin signalling pathway, which results in interference with the expression of multiple genes, initiating colorectal adenoma in Apc min/+ mice. [39][40][41] In this study, our data indicated that Rbm24 was significantly downregulated in intestinal tumour of Apc min/+ mice. Nevertheless, we did not find an association between RBM24 expression and APC mutation in human colorectal cancer tissues via bioinformatics analysis. This may be due to differences in the regulation of gene expression between human and murine. In addition, RBM24 expression in tumour tissues of CRC patients was dramatically reduced and the expression of RBM24 correlated significantly with the prognosis of CRC patients, which was consistent with the data from TCGA and GEPIA databases. CRC patients with high expression of RBM24 seemed to have better long-term survival. Therefore, we considered that RBM24 may be related to drug resistance of CRC, which was confirmed to a certain extent in this study. High expression of RBM24 elevated 5-FU or cisplatin-induced apoptosis rate. However, since these results were obtained only in vitro, the mechanisms by which the downregulation of RBM24 lessens sensitivity of CRC cells to chemotherapy agents need to be further explored.
Previous studies have indicated that RBM24 can bind several tumour-related genes. 21,22,24 PTEN, a phosphatase on PIP3, acts as a tumour suppressor gene and activates PI3K/Akt signalling pathway. 42,43 Moreover, the downstream target genes of Akt, including CDK4, MMP-2 and MMP-9, are involved in cancer cell proliferation, migration and invasion. RBM24 is highly homologous to RBM38 44 ; the biological functions of RBM24 and RBM38 share some similarities. In a previous study, it was reported that RBM38 functions as a tumour suppressor, which can bind PTEN mRNA and maintain its stability, increasing mRNA levels of PTEN in breast cancer tissue. 45 In the present study, our results suggested that RBM24 could bind to the mRNA of PTEN and maintain its stability, thereby regulating the PI3K/Akt signalling pathway. This notion was supported by evidence that treatment with PI3K/Akt signalling inhibitors could partially reverse phenotypes induced by RBM24 knockdown. P21 is regulated by RBM24, 24 which represses the proliferation of various tumour cells. 46 However, it has also been shown that P21 is a downstream molecule of the PI3K/Akt pathway. 47 Accordingly, our results indicate that overexpression of RBM24 in CRC cells followed by treatment with PTEN inhibitor could partially reverse phenotypes induced by RBM24 overexpression. In the RIP assay, our results further confirmed that RBM24 directly binds to PTEN mRNA, which strongly suggests that RBM24 affects PTEN mRNA abundancy and leads to the activation of its downstream signalling pathways.

CONCLUSIONS
In conclusion, our findings revealed that RBM24 suppressed colorectal tumourigenesis by increasing the stability of PTEN mRNA, which led to the suspension of the PI3K/Akt signalling axis. Since the RBM24-mediated regulatory pathway represses CRC development, targeting RBM24 holds strong promise for the diagnosis and treatment of CRC. To our knowledge, this is the first study to show that RBM24 plays an important tumour suppressive role in colorectal tumourigenesis in an animal model, which is invaluable for understanding the biological function of RBM24 in cancer.

A C K N O W L E D G E M E N T S
We thank Prof. Yong Xiong Chen (Ophthalmology Center, School of Medicine, Xiamen University) for the generous gift of the UBC-CreERT2 mice. We are also grateful to all members of the Institute of Stem Cell and Regenera-