A CREB1/miR-433 reciprocal feedback loop modulates proliferation and metastasis in colorectal cancer

Downregulation of miR-433 participates in the progression of CRC

Firstly, we conducted qRT-PCR in 35 paired fresh colorectal cancer and adjacent normal mucosa specimens. As displayed in Figure 1a-b, miR-433-3p exhibited lower expression in cancer relative to normal mucosa. We then performed real-time PCR in CRC cell lines, as shown in Figure 1c, and found that miR-433-3p tended to present relatively low expression across the 9 cell lines. The miRWalk database was adopted to carry out the KEGG (Kyoto Encyclopedia of Genes and Genomes) pathway analysis, and the results indicated that miR-433 is significantly correlative to the Wnt, Ras, MAPK, cAMP signaling pathways and cell adhesion molecules, which are commonly mis-regulated in cancer progression (Figure 1d). Last, we utilized the LinkedOmics database to determine the prognostic value of miR-433 in CRC and found that higher expression of miR-433 tended to be a protective variable in prognosis, but not significantly (Figure 1e).

miR-433 suppresses CRC cell proliferation in vitro and in vivo

To assess the biological impact of miR-433 on CRC cells, gain-of-function and knockdown approaches were carried out. As shown in Figure 2a, miR-433-3p mimics notably repressed SW480, SW620, LoVo and RKO cells propagation compared with NC, and its inhibitor conferred more robust cell viability on these cells relative to the counterpart groups. Similarly, miR-433-3p mimics and inhibitor remarkably attenuated or reinforced the colony formation ability of CRC cells compared with their respective control groups (Figure 2b). Then, we cloned a miR-433-3p overexpression vector into lentivirus and infected LoVo cells to carry out subcutaneous oncogenesis. From the fourth week, we repeatedly measured the diameter of the subcutaneous tumors in the nude mice. With time elapsing, the tumors in the LoVo/LV3-NC group grown more and more robustly compared with the counterpart. Eight weeks after inoculation, the mice were sacrificed. As presented in Figure 2c, nude mice in the LoVo/LV3-miR-433-3p group apparently formed smaller tumors than those in the counterpart group. The tumor growth curve and weights between the LV3-miR-433-3p and LV3-NC groups were also significantly different (Figure 2d-e). The mean tumor volume of LV3-NC group was more than quintuple than that in the counterpart group (2575.76 ± 909.63 vs 476.81 ± 245.58 mm³). Furthermore, the mean tumor weight of LV3-NC group was more than sextuple than that in the LoVo/LV3-miR-433-3p group (1.74 ± 0.30 vs 0.27 ± 0.14 g). Hematoxylin and eosin stained slides were imaged and are displayed in Figure 2g. IHC revealed less Ki-67 staining in LoVo/LV3-miR-433-3p tumors relative to counterparts.

miR-433 abrogates CRC invasion and liver metastasis in vitro and in vivo

As shown in Figure 3a and S3, miR-433-3p mimics obviously inhibited the migration and invasion activity of LoVo and RKO cells compared with NC; and miR-433-3p inhibitor upregulated the cell migration and invasion properties relative to inhibitor NC groups. An intra-spleen injection CRC liver metastasis model was performed in mice, as exhibited in Figure 3b-c, and the LoVo/LV3-miR-433-3p group presented much fewer visible metastatic nodules than the LoVo/LV3-NC group on the mouse livers. H&E staining demonstrated that the tissues that infiltrated into the liver was derived from LoVo cells (Figure 3d).

CREB1, CCAR1 and JNK1 are direct targets of miR-433 in CRC

The Oncomine database was applied to explore the expression of CREB1, CCAR1 and JNK1 in CRC tissues. As illustrated in Figure 4a, CREB1, CCAR1 and JNK1 were all overexpressed in CRC relative to normal tissues across a series of subsets. Next, we conducted real-time PCR in our collected 35 paired fresh CRC and adjacent normal mucosa specimens, and a consistent result was observed (Figure 4b). Meantime, Spearman correlation analyses revealed that miR-433-3p is negatively correlated with CREB1, CCAR1 and JNK1 mRNA levels (Figure 4c).

The TargetScan database was adopted to predict the binding sites between miR-433 and its targets. As shown in Figure 4d, two putative binding sites existed in the 3’UTR region of CREB1; however, only a single conserved site presented in the 3’UTR of CCAR1 and JNK1. Likewise, a negatively correlated model between CREB1, CCAR1, JNK1 and miR-433 was observed when we up- or down-regulated miR-433 through transient or stable transfection in CRC cells at the mRNA and protein level (Supplementary Figure S1a-b and Figure 4e). Furthermore, the same phenomenon was again found in the subcutaneous tumors (Figure 2f).

The dual luciferase assays were used to determine the targeted binding activity between miR-433 and its putative targets. As shown in Figure 4f, there was no difference between the wt-1 CREB1 3’UTR and mut-1 CREB1 3’UTR. Inversely, a remarkable decrease in luciferase activity was found when wt-2 CREB1 3’UTR was co-transfected with the miR-433 mimics, which indicated that miR-433 targeted CREB1 by binding to the second presumed site. Positive binding results were also observed among wt CCAR1 3’UTR, wt JNK1 3’UTR and miR-433 mimics.

Next, we conducted IHC in 200 FFPE CRC samples. An immune-staining intensity score system was adopted as mentioned above. Representative images of the hierarchic staining of CREB1, CCAR1 and JNK1 are displayed in Supplementary Figure S2. As shown in Table 1, chi-square analyses revealed no evident association between CREB1 and clinicopathological factors, although CREB1 appeared to be connected with distant metastasis. CCAR1 was relative to tumor depth, distant metastasis and pTNM stage, but not to age, gender, tumor location, tumor size, histological grade, lymph node invasion, CEA, CA125 or CA19-9. JNK1 was correlated with lymph node invasion, pTNM stage and CA19-9, but not with age, gender, tumor location, tumor size, histological grade, tumor depth, distant metastasis, CEA or CA125.

Then, a Cox proportional hazard regression model was established. As shown in Table 2, univariate analysis revealed that tumor depth, lymph node invasion, distant metastasis, pTNM stage, CEA, CA19-9, CREB1, CCAR1 and JNK1 may act as potential prognosticators. In light of that the pTNM stage system integrates the clinical significance of the tumor, lymph node and distant metastasis, only pTNM stage, CEA, CA19-9, CREB1, CCAR1 and JNK1 were incorporated into the multivariate analyses. The results indicated that pTNM stage, CEA, CCAR1 and JNK1 could serve as independent variables in prognosis. In addition, CREB1 was another potent borderline candidate.

Last, Kaplan-Meier analyses combined with the log-rank method demonstrated that the patients with higher expression of CREB1, CCAR1 or JNK1 were predisposed to a much shorter OS than those with lower expression (Figure 4g).

CREB1 transactivates miR-433 by binding to its promoter

To elucidate the underlying mechanism involved in the downregulation of miR-433 in CRC, we considered the possibility of transcriptional modulation. The DBTSS database was utilized to explore the presumed promotor region of miR-433. The corresponding sequence was acquired from the UCSC gene browser. Finally, a combination approach using JASPAR, CONSITE and LAGASNA databases was adopted to predict the putative transcription factor binding site. As demonstrated in Figure 5a, all three of the datasets predicted a probable CREB1 binding site. Therefore, we purchased wt-pmirGlo-mir433 and mut-pmirGlo-mir433 luciferase vectors from GenePharma (Figure 5b).

Next, we found that the expression of miR-433-3p was downregulated when we successfully knocked down CREB1 at the mRNA and protein level (Figure 5c-d). A similar result was also discovered when we detected the expression of miR-433 after LV5-CREB1 infection of LoVo and RKO cells (Figure 5e-f). As shown in Figure 5g, a ChIP-qPCR assay confirmed that anti-CREB1 could enrich much more DNA fragment containing putative CREB1 binding site on the miR-433 promotor region relative to IgG. Luciferase assays indicated that overexpression of CREB1 could significantly increase the luciferase activity in LoVo and RKO cells relative to counterpart cells (Figure 5h-i). A concordant phenomenon was observed when RNA interference was conducted in LoVo and RKO cells (Figure 5j-k).

miR-433 modulates cell cycle progression and epithelial-mesenchymal transition

To determine the specific pathway underlying miR-433 involvement in CRC proliferation and metastasis, we detected some specific markers. As shown in Figure 6a, SW480, LoVo and RKO cells stably transduced with LV3-miR-433-3p presented much higher expression of miR-433-3p compared with their counterparts. Overexpression of miR-433-3p inhibited CREB1, CCAR1 and JNK1 expression in CRC cells, with an accompanying downregulation of phospho-Smad2, phospho-c-Jun, CDK2, vimentin, snail and slug and upregulation of p21, p27, E-cadherin and β-catenin (Figure 6b). Last, as demonstrated in the schematic in Figure 6c, CREB1 and miR-433 composed a feedback loop to regulate CRC proliferation and metastasis by targeting CCAR1 and JNK1.

Bioinformatics

The LinkedOmics (www.linkedomics.org) database was used to plot a survival curve according to the median of miR-433 expression. The miRWalk (http://mirwalk.umm.uni-heidelberg.de/) dataset was adopted to conduct KEGG pathway analysis. The Oncomine (www.oncomine.org) database was applied to explore the expression of CREB1, CCAR1 and JNK1 in CRC specimens. The TargetScan (www.targetscan.org) dataset was employed to predict the targets of miR-433 and its binding sequences. UCSC gene browser allied DBTSS (https://dbtss.hgc.jp), JASPAR, CONSITE (http://consite.genereg.net/) and LASAGNA-Search 2.0 (http://biogrid-lasagna.engr.uconn.edu/lasagna_search/) were utilized to discover the putative promotor region of miR-433 and potential transcription factor binding sites (TFBS).

Patients and study materials

Thirty-five paired fresh colorectal cancer and adjacent normal mucosa tissues from patients who were diagnosed with adenocarcinoma of the colon or rectum in histology were collected in the Department of General Surgery, Shanghai Jiao Tong University Affiliated Sixth People’s Hospital with the informed consent of the patients. Another two hundred FFPE (formalin-fixed and paraffin-embedded) CRC patient sections between January 2010 and January 2012 were acquired from the Department of Pathology, Shanghai Jiao Tong University Affiliated Sixth People’s Hospital. Patients with incomplete follow-up were excluded. The histological type was determined as well, moderately or poorly differentiated. Tumor stage and pathological classification were defined according to the American Joint Committee on Cancer/Union for International Cancer Control 8th Edition. 164 patients (stage I-III) received radical colectomy or proctectomy and lymphadenectomy. Other 36 metastatic CRC patients received palliative chemotherapy or best supportive care. Postoperative adjuvant chemotherapy was given to 97 patients with pathological T4 tumors, T3 tumors with high-risk factors (poorly differentiated histology; lymphatic/vascular invasion; bowel obstruction; <12 lymph nodes examined; perineural invasion; localized perforation; indeterminate or positive margins) for relapse or lymph node metastasis. Among these 97 patients, 74 (stage III) were treated with FOLFOX (5-fluoro-2,4(1H, 3H)-pyrimidinedione, leucovorin, oxaliplatin), XELOX (xeloda, oxaliplatin) or FOLFIRI (5-fluoro-2,4(1H, 3H)-pyrimidinedione, leucovorin, irinotecan) regimens, and another 23 stage II patients with high-risk factors for recurrence were treated with xeloda, XELOX or FOLFOX regimens. The median follow-up time was 34 months (range: 1–88). A total of 130 patients died of CRC-related causes during the study period. Written informed consent was obtained from all patients in accordance with the local ethical guidelines. The whole procedure was approved by the Ethics Committee of Shanghai Jiao Tong University Affiliated Sixth People’s Hospital.

Cell culture

The human CRC cell lines SW480, SW620, LoVo, HCT116, LS174-T, Caco-2, DLD-1, RKO and SW1116 were purchased from Cell Bank of Chinese Academy of Sciences (Shanghai, China) and cultured in Dulbecco’s modified Eagle’s medium (DMEM) (Gibco, Carlsbad, USA) supplemented with 10% fetal bovine serum (FBS) (Gibco, Carlsbad, USA) and 1% penicillin/streptomycin (KeyGen BioTECH, Nanking, China) in a humidified atmosphere of 5% CO2 at 37°C.

Oligonucleotides transfection and lentivirus infection

miR-433 mimics, inhibitor, their respective negative controls, and the siRNA of CREB1 were purchased form GenePharma (Soochow, China) and successfully transfected into the corresponding cells according to the manufacturer’s instructions in the presence of Lipofectamine 2000 (Life Technologies, Carlsbad, USA). Similarly, the lentivirus packages of miR-433 and CREB1 overexpression vectors were also purchased from GenePharma, and successfully infected CRC cells by following the instructions.

RNA isolation and quantitative real-time PCR

One milliliter of TRI Reagent^® (Sigma, St. Louis, MO, USA) was applied to dissociate the specimens and cells. Isopropanol (Sangon Biotech, Shanghai, China) was used to precipitate the total RNA. After measuring the concentration of RNA with a spectrophotometer (Tiangen, Peking, China), 1000 ng of total RNA was reverse transcribed to cDNA, and then, 1 µl cDNA was employed to conduct quantitative PCR (polymerase chain reaction) following the instructions of the manufacture (GeneCopoeia, Rockville, MD, USA). The specific primers of CREB1, CCAR1, JNK1, GAPDH, miR-433 and snRNA U6 were also purchased from GeneCopoeia. The 2^-△△CT method was adopted to analyze the relative expression.

Western blotting

After 48-72 hours of relevant treatment, the cultured cells were washed twice with PBS (phosphate-buffered solution), and then, 50-100 µl RIPA lysis buffer (Sangon Biotech, Shanghai, China) supplemented with PMSF (phenylmethane sulfonyl fluoride), phosphatase and protease inhibitor (KeyGEN BIOTECH, Nanking, China) was used to elute the total protein. After measurement of protein concentration with a BCA (bicinchoninic acid) kit (Merck Millipore, Darmstadt, Germany), 50-100 µg of total protein was loaded and separated on a 10% SDS polyacrylamide gel, transferred to a PVDF (polyvinylidene difluoride) membranes, and incubated with anti-p21, anti-p27, anti-CDK2, anti-phospho-c-Jun or anti-phospho-Smad2 monoclonal rabbit antibody (1:1000, CST, Danvers, MA, USA); anti-E-cadherin, anti-β-catenin, anti-vimentin, anti-snail, anti-slug, anti-c-Jun, anti-Smad2, anti-CREB1, anti-JNK1 or anti-GAPDH rabbit polyclonal antibody (1:1000-1:5000, Proteintech, Chicago, IL, USA); or anti-CCAR1 rabbit polyclonal antibody (1:1000, SAB, Baltimore, MD, USA) overnight at 4°C. Protein bands were visualized using a chemiluminescence kit (Merck Millipore, Darmstadt, Germany).

Immunochemistry

FFPE slides were heated at 60°C for two hours and then submerged in xylene for 10 mins (minutes) three times, ethyl alcohol twice for 5 mins, 95% ethanol, 85% ethanol, 75% ethanol for 5 mins, respectively. Then, antigen retrieval was performed with 10 nM citrate antigen retrieval solution (Sangon Biotech, Shanghai, China) in a microwave oven. After endogenous peroxidase inhibition and blocking procedures completed, the sections were incubated with anti-CREB1 rabbit polyclonal antibody (1:400, Proteintech, Chicago, IL, USA), anti-CCAR1 rabbit polyclonal antibody (1:400, SAB, Baltimore, MD, USA) and anti-JNK1 rabbit polyclonal antibody (1:300, Proteintech, Chicago, IL, USA). The other procedures, H&E (hematoxylin & eosin) staining and immune scoring of the slides were conducted as described in our previous work [41].

Dual luciferase reporter assay

To determine the binding between miR-433 and its targets, a certain number of HEK293T cells were seeded in a 24-well plate in advance, and then the cells were co-transfected with 20 pmol of miR-433 mimics or negative control with wild-type (wt) or mutant (mut) 3’UTR vectors using Lipofectamine 2000. After 48 hours, the cells were washed with PBS twice and lysed with passive lysis buffer (Promega, Madison, WI, USA). For the binding between CREB1 and the mir433 promoter, LV5-NC- and LV5-CREB1-infected cells were similarly counted and seeded in a 24-well plate prior to the experiment, and concurrently, the relevant cells were pre-transfected with CREB1 siRNA and the negative control 24 hours prior to the experiment, and then, the CRC cells were transfected with wt and mut mir433 promoter dual luciferase plasmids accordingly in the presence of Lipofectamine 2000. The subsequent procedures were performed following the instructions of the manufacturer (Promega, Madison, WI, USA). At last, the luciferase activity was collected with a microplate reader (BioTeke, Peking, China).

Chromatin immunoprecipitation (ChIP) assay

For the ChIP assay, CRC cells were infected with the CREB1 overexpression lentivirus LV5-CREB1. Anti-CREB1 polyclonal antibody (Abcam, Cambridge, UK) and a Chromatin Immunoprecipitation (ChIP) Assay Kit (Merck Millipore, Darmstadt, Germany) were adopted in this assay. Other procedures were performed as described previously [42].

MTT assay

After relevant treatments were conducted, the CRC cells were counted and adjusted to a concentration of 1 x 10⁶/ml, and 1 x 10³ cells per well were seeded into a 96-well plate. Then, 50 µl of 1 x MTT (methyl thiazolyl tetrazolium) reagent (KeyGen BioTECH, Nanking, China) was added into the medium accordingly. The OD (optical density) value was obtained with a microplate reader.

Transwell assay

Similarly, after treatment, 1-2 x 10⁵ CRC cells were seeded in the upper chamber of 8-µm polycarbonate membrane filters (Corning, NY, USA) with or without Matrigel (Corning, NY, USA), and 600 µl of medium containing 10% FBS as a chemokine was placed in the lower chamber. After 24-48 hours, the cells in the lower chamber were fixed in 4% paraformaldehyde and stained with 0.5% crystal violet. Then, five random fields were photographed.

Colony formation assay

CRC cells from different treatment groups were seeded in 6-well plates (1 x 10³ cells/well) and then cultured for two weeks. After, the cells were fixed in pre-cooled formaldehyde and stained with 0.5% crystal violet. The colony numbers were counted accordingly.

Subcutaneous tumorigenicity and liver metastasis mouse model

Approximately 5 x 10⁶ CRC cells were inoculated into the right flank of female BALB/c nu-nu mice. The mice were observed every week. After optical tumor appeared, a vernier caliper was applied to measure the length and width each week. Eight weeks later, the mice were sacrificed, and the tumors were collected and weighed. For the CRC liver metastasis model, a 1-cm single incision was made in the upper left lateral abdomen of the mice to expose the spleen in the intraperitoneal administration of 10% chloral hydrate. The other procedures were followed according to a classic method reported previously [43]. All animal experimental procedures were approved by the Ethics Committee of Shanghai Jiao Tong University Affiliated Sixth People’s Hospital and the mice were cared in accordance with the institution guidelines.

Statistical analyses

All data were analyzed through SPSS13.0 or GraphPad Prism 7.0 software using Student’s t test, χ² test or one-way analysis of variance (ANOVA) as appropriate and are presented as the mean ± standard deviation (SD). Spearman correlation was adopted to evaluate the relationship between miR-433 and its targets. The Cox proportional hazard regression model and Kaplan-Meier analyses were applied to identify the independent prognostic variables and to plot survival curves using a log-rank test. p<0.05 was considered statistically significant.