CCDC12 promotes tumor development and invasion through the Snail pathway in colon adenocarcinoma

Integrative expression Quantitative Trait Loci (eQTL) analysis found that rs8180040 was significantly associated with Coiled-coil domain containing 12 (CCDC12) in colon adenocarcinoma (COAD) patients. Immunohistochemical staining and western blotting confirmed CCDC12 was highly expressed in COAD tissues, which was consistent with RNA-Seq data from the TCGA database. Knockdown of CCDC12 could significantly reduce proliferation, migration, invasion, and tumorigenicity of colon cancer cells, while exogenous overexpression of CCDC12 had the opposite effect. Four plex Isobaric Tags for Relative and Absolute Quantitation assays were performed to determine its function and potential regulatory mechanism and demonstrated that overexpression of CCDC12 would change proteins on the adherens junction pathway. Overexpressed Snail and knocked down CCDC12 subsequently in SW480 cells, and we found that overexpression of Snail did not significantly change CCDC12 levels in SW480 cells, while knockdown of CCDC12 reduced that of Snail. CCDC12 plays a significant role in tumorigenesis, development, and invasion of COAD and may affect the epithelial to mesenchymal transformation process of colon cancer cells by regulating the Snail pathway.


Background
The incidence of colon cancer is the fourth highest for malignant tumors, with over one million new colon cancer patients worldwide each year (1). In the Asian population, approximately 90% of colon cancers are histologically classi ed adenocarcinomas, and the prognosis of these patients is poor (2,3). Refractory and metastatic colon adenocarcinoma (COAD) has been a major problem worldwide (4)(5)(6). Somatic mutations and activation of key oncogenic pathways have often been observed in COAD. Hence, it is essential to understanding the mechanism of refractory and metastatic COAD. In addition, identifying effective prognostic biomarkers for high risk patients with COAD is vital.
Coiled-coil domain containing 12 (CCDC12) is an evolutionarily conserved protein that encodes a coiledcoil domain. It is located in the 3p21.31 region of human chromosome 3 (7). CCDC12 has been reported to be associated with erythroid differentiation (8), and split-ubiquitin system (9). In addition, coiled-coil domain containing family members have been associated with tumor cell proliferation. CCDC106 is associated with the progression and poor prognosis of non-small cell lung cancer (10), while CCDC67 has been demonstrated to inhibit the proliferation of papillary thyroid carcinomas (11). A genome-wide association study (GWAS) in China identi ed colorectal cancer risk single nucleotide polymorphism (SNP) rs1076394 as an expression Quantitative Trait Loci (eQTL) for CCDC12 (12). However, the speci c carcinogenesis of CCDC12 has not been deciphered.
In this study, we demonstrated that high-expression levels of CCDC12 in COAD was closely associated with tumor development and aggressive. CCDC12 promoted COAD tumor cell proliferation, invasion, migration, and inhibited apoptosis in in vitro and in vivo experiments. Furthermore, CCDC12 could regulate epithelial to mesenchymal transformation (EMT) of COAD cells through the Snail pathway. Our study demonstrated a biological link between CCDC12 and COAD, which could be used as a potential therapeutic target.

Colorectal cancer (CRC)-associated SNPs and germline genotype data
CRC-associated SNPs were extracted from the National Human Genome Research Institute (NHGRI) GWAS database including 50 CRC risk loci (Supplement to table S1). The datasets for germline genotypes, ancestry, expression pro les, methylation, somatic copy number aberrations, germline copy number aberrations regarding CRC were downloaded from The Cancer Genome Atlas (TCGA) portal. SNP loci with minor allele frequency (MAF) > 0.05 from TCGA (subjects) and HapMap cell lines (controls) were downloaded on EIGENSTRAT and the top two principal components were retrieved ( Figure S1). We calculated the average of segmented copy-number scores of genetic interval between the transcription start and end sites as gene-based somatic copy-number measure ( Figure S2A). CpG methylation status was determined by discretization CpG methylation value with cut-off values of 0.2, 0.4, 0.6, 0.8, and 1.0 ( Figure S2B). Afterward, we calculated the expression levels for each gene as TPM values.

Association analysis and eQTL analysis
eQTL analysis was performed according to the owchart described in Li et al. (13). Brie y, the expression data of gene was adjusted for somatic copy-number effects and CpG methylation status using a multivariate linear model. The P value corresponds to the regression coe cient based on residual expression levels and germline genotype. We performed cis-eQTL analysis between 656 857 SNP loci and corresponding mRNA transcripts. Then excluded 243 359 SNP loci with MAF < 0.05 and their genes with absent calls >90% and false discovery rate (FDR) > 0.1. With the 50 SNPs from NHGRI GWAS database, 18 SNPs were present in TCGA germline genotype. For variants not directly genotyped, we used proxy SNPs (nearest SNP with linkage disequilibrium > 0.5). SNAPinfo software was selected to obtain pairwise linkage disequilibrium between SNPs (13). In cis-eQTL analysis, we evaluated the association between genotype of given SNP locus and the transcripts located within ± 1Mb regions. For risk SNP locus, the relationship between it and transcripts at a genome-wide level was evaluated. For each target gene,  Kb regions on either side of transcription start site were considered putative enhancer regions. They were overlapped using ENCODE DNaseI hypersensitivity data from HCT116 cell line and then analyzed for transcription factor (TF) DNA binding motif enrichment. Hypergeometric distribution test was used for overlap analyses with a signi cance level of P value < 0.05. TFs that satis ed the above criteria were considered as candidates for trans-acting risk SNPs.

Immunohistochemistry staining
Immunohistochemical staining was performed using Power-Vision two-step tissue staining kit (ZSGB-BIO, Cat. PV-6001, Beijing, China). After depara nization and rehydration, tissue slides were incubated with 3% H 2 O 2 for 10 minutes at room temperature. 10 mmol/L EDTA solution used for antigen retrieval.
Primary antibodies were incubated overnight at 4℃ and then secondary antibodies incubated at 37℃ for 30 minutes. After washed, stained with DAB-H 2 O 2 and counterstained with hematoxylin. The results of IHC were evaluated using H-scores by 3 researchers independently (Table S2).

Over expression of Snail in SW480 cell lines
Lentivirus expressing SNAI1 was used to infect SW480 cell line and expressed red uorescent protein. Seventy-two hours later, uorescence of red uorescent protein was observed under a uorescent microscope (Olympus, IX71, Japan) then they were used for subsequent experiments.

Western blotting
The extracted proteins from cells and tissues were electrophoresed on a 10% SDS-PAGE, and then transferred onto a 0.45μm Immobilon-P Transfer Membrane (Millipore, Cat.IPVH00010, USA) using the wet transfer method. Incubated with primary antibody overnight at 4℃ and then incubated with corresponding secondary antibody for 1 hour at room temperature. Bands were visualized using ECL kit (Millipore, Cat.WBKLS0500, USA) and the Amersham Imager 680 system. Primary antibodies used for western blotting are listed in Table S2. 2.9 Real-time Quantitative Polymerase Chain Reaction (RT-PCR) Total RNA was extracted with RNA isolation kit (Invitrogen, USA) and reverse transcribed to cDNA using the Reverse Transcription System (Promega, USA). Quantitative RT-PCR was performed with SYBR Green qPCR SuperMix (Invitrogen, USA) and ABI PRISM® 7500 Sequence Detection System based on the manufacturer's instructions. 18srRNA was used as the internal reference control. Primer sequences were as follows: 18srRNA forward 5'-CCTGGATACCGCAGCTAGGA-3', reverse 5'-GCGGCGCAATACGAATGCCCC-3', CCDC12 forward 5'-CTGACTGGGACCTCAAGAGA-3', reverse 5'-CCTTTCAGCCTTTCACGGAT-3', Snail forward 5'-GAGGCGGTGGCAGACTAGAGT-3', reverse 5'-CGGGCCCCCAGAATAGTTC-3'.
2.10 Colony-forming and MTS assays 100 cells in logarithmic growth phase were resuspended in 300μl medium and then seeded into a 6-well plate. After colony formation, stained with 1% crystal violet solution for 20 minutes. Number of colonies were counted under a microscope.
In MTS assay, 1×10 4 cells were seeded into a 96-well plate. CellTiter 96 ® A Q ueous One Solution Cell Proliferation Assay (Promega Cat.G3582, USA) was used to measure cell proliferation and was performed based on the manufacturer's instructions. OD was measured using an Multiscan MK3 microplate reader at 490nm.

Wound-healing assays
Cells were seeded into a 6-well plate and cultured until 95% con uent. Monolayer cell was scraped off using a pipette tip in the middle of plate. Cell migration was measured every 6 hours using the Image Pro-Plus 6.0 and migration rate was calculated with (Distance 0h -Distance different time points ) / Distance 0h .

Cell invasion assays
Transwell chambers (BD, Cat.353097) with Matrigel (BD, Cat.356234, USA) were used for invasion assays. 100μl of cells suspension (1×10 5 cells) with serum-free medium were placed in the upper chamber. The bottom chamber contained medium with 20% serum. Cells were incubated for 24 hours at 37℃, 4% paraformaldehyde was used to x the cells for 15 minutes, and then stained with crystal violet solution. The cells passing through the chamber were observed with a microscope.

Apoptosis assays
The Annexin V-FITC apoptosis detection kit (Keygen, Cat.KGA106, Jiangsu, China) was used per manufacturer's instructions. 1.25μl Annexin V-FITC reagent was added to 500μl of cell suspension (1×10 6 /ml), and then incubated for 15 minutes at room temperature in the dark. After centrifuged at 1 000×g for 5 minutes, supernatant was removed, and cells were resuspended in 0.5 ml pre-cooled bindingbuffer. Then, 10 μl Propidium Iodide was added and incubated in the dark before being read on the BD FACSCalibur CellSorting System.

Cell cycle analysis
Cell Cycle Detection Kit (Keygen, Cat.KGA511, Jiangsu, China) was used for cell cycle analysis. 5 μl (10mg/ml) of RNase A was added to cells and incubated at 37℃ for 1 hour. Afterward, 50 μg/ml PI and 0.2% Triton X-100 were added and incubated at 4℃ in the dark for 30 minutes. BD FACSCalibur CellSorting System was used to measure cell cycle phases. 2-3×10 4 cells were counted and analyzed using ModFit software.

Xenograft mouse models
4-week-old BALB/c nude mice were purchased from Charles River Laboratories (Beijing, China) fed on ordinary diet. Xenograft tumors were established by subcutaneous injection of 200μl cell suspension (5×10 5 cells) into the underarms or backs of nude mice. The tumor volume was calculated using the following formula, Tumor Volume(mm 3 ) = (Long diameter × Short diameter 2 )/2. Mice were euthanized 30 days after inoculation, and tumors removed for subsequent analysis.

Statistical Analysis
All statistical analysis was performed using R 3.5.1. Data were expressed as mean ± SD. One-way analysis of variance and Student's T test were used to analyzed differences among groups. χ 2 test or linear correlation was used to determine the correlation between CCDC12 expression and clinicopathological features. Kaplan-Meier method was used to generate survival curves with log-rank test. MAF > 0.05, FDR < 0.1, α = 0.05 and Pvalue < 0.05 with two sides were considered statistically signi cant.

SNP rs8180040 is signi cantly associated with CCDC12 expression based on integrative eQTL analysis
Ancestry veri cation con rmed that the 130 samples in our study was HapMap CEU, and table S3 summarizes their clinical and pathologic data. Of all the risk SNP loci, only rs10505477 and rs6983267 were in high pairwise linkage disequilibrium (R 2 = 0.875). After ltering out FDR < 0.1, we obtained 5 029 762 SNP-genes and identi ed 2 030 signi cant associations with P value < 4.03×10 -5 . They mapped to a total of 1 964 SNP loci and 478 unique target genes (Table S4). Of the target genes, 332 (69.5%) were found to be regulated by a single cis-acting SNP locus. After adjusting for correlated loci by stepwise feature selection, 254 genes (53.1%) could be explained by multiple SNP loci (median = 4), which suggested the existence of multiple independent eQTLs. Of the 1,964 cis-acting eQTL loci, 1 899 were associated with one target gene, and suggested that the associations were highly locus speci c. Transcript levels of 6 638 genes (55.6%) were signi cantly affected by the somatic copy-number changes in the corresponding coding regions (FDR < 0.1). In addition, among the 478 target genes of the cis-acting SNP loci, 299 (62.6%) were signi cantly associated with somatic copy number. We identi ed 6 662 transcripts (55.8%) that were affected by CpG island methylation in the promoter region (FDR < 0.1). Among these genes, 302 (63.2%) were also target genes of eQTLs ( Figure 1A). Subsequently, we identi ed 25 SNP-gene expression associations in 22 loci with nominal P value < 0.05 for genes within 2Mb of the risk SNPs (Table S5). After correcting for multiple-testing (FDR < 0.05), only one SNP-gene association was found to be signi cant (SNP rs8180040 with genes CCDC12, Table S5 and Figure 1A).

High expression of CCDC12 in COAD is associated with poor prognosis
Our previous study found that CCDC12 may be associated with colon cancer and was a potential protooncogene (14). To determine the expression levels of CCDC12 in COAD, we performed immunohistochemistry (IHC) with TMAs that contained 75 pairs of COAD and adjacent normal tissues. Positive staining rate of CCDC12 in COAD tissues were signi cantly higher compared to normal colon tissues (51/75 vs 8/75, P value < 0.001, Figure 1B). In addition, we performed western blotting with fresh tumor and adjacent normal tissues from 12 patients. The results showed that the expression levels of CCDC12 in 10 colon cancer tissue samples were signi cantly higher compared to normal adjacent tissues ( Figure 1C). We then analyzed RNA-Seq data from TCGA database, which included 456 COAD patients and 41 corresponding normal tissues. It found that CCDC12 was statistically signi cantly overexpressed in COAD (P value < 0.001, Figure 1D). Subsequently, we determined the correlation between expression level of CCDC12 and prognosis of patients with COAD. Our analysis demonstrated that higher expression of CCDC12 was correlated with poor prognosis of COAD patients grouped by gender (P value = 0.0042, Figure 1E).

Knockdown of CCDC12 inhibits proliferation, migration, invasion and promoted apoptosis.
To investigate the biological role of CCDC12 in colon cancer, we rst measured the expression levels of CCDC12 in ve different colon cancer cell lines by qRT-PCR, and the CCD-18Co as the control ( Figure S3). The expression levels of CCDC12 in LOVO and SW480 cell lines were relatively high, hence we reduced their expression levels using si-CCDC12 RNA. qRT-PCR results demonstrated that siRNA1 had the best knockdown e ciency for CCDC12 in both SW480 and LOVO cell lines ( Figure S4). So, we used siRNA1 for all subsequent functional assays. In vitro experiments were performed with CCDC12 knockdown group (CCDC12-KD, cells transfected with siRNA1), normal group (NC, un-transfected cells) and blank control group (Vec, cells transfected with empty vector).
Colony-forming assays demonstrated that both SW480 and LOVO cell lines in CCDC12-KD group had reduced ability to form colonies compared to cells in the NC and Vec groups (Figure 2A). Cell cycle assay demonstrated that compared to cells in others group, cells in the CCDC12-KD group were predominately in the G0/G1 phase, with a reduced proportion of cells in the S phase. In addition, it was observed that in SW480 cell lines, the ratio of cells in the G2/M phase were decreased ( Figure 2B). Using MTS assays, we observed that the rates of cell proliferation in the CCDC12-KD group were signi cantly lower on the third day compared to that in the NC and Vec groups ( Figure S5A). In wound-healing assays, we observed slight differences in migration at 6hrs between cells in the CCDC12-KD group and NC group. At 24hrs, cell migration in the CCDC12-KD group were signi cantly reduced ( Figure 2C). Matrigel-coated transwell chambers were then used to evaluate differences in cell invasion between the three treatment groups. We found that the cell invasion ability in the CCDC12-KD group were much lower, especially in the SW480 cell line ( Figure 2D). Using Annexin V-FITC to measure apoptosis rates, we observed that the total apoptosis rate (UR+LR) of cells in the CCDC12-KD group were increased. There were no differences in early apoptosis between three groups, however there were differences in late apoptosis ( Figure 2E).We then injected CCDC12-KD SW480 or NC SW480 cells subcutaneously into the back of BALB/c nude mice to generate xenograft models. After 30 days, we observed that the volume and weight of tumors in mice in CCDC12-KD group were much smaller ( Figure 2F). All these results indicated that knockdown of CCDC12 could effectively inhibit colon cancer cell proliferation, invasion, migration, and promoted apoptosis in vivo and in vitro.

Overexpression of CCDC12 promotes proliferation, migration, invasion, and inhibits apoptosis
Based on expression levels of CCDC12 in colon cancer cell lines ( Figure S3), we selected HCT116 cell line and CCDC12-knocked SW480 cell line to overexpress CCDC12. In addition, we overexpressed CCDC12 stably (OE-CCDC12 group) in these two cell lines ( Figure S6) and set normal group (NC) and blank control group (Vec) as described in the last section.
In colony formation assays, cells in the OE-CCDC12, NC, and Vec groups formed colonies on the seventh day of culture. However, cells in the OE-CCDC12 group generated a higher number of colonies compared to the other two groups ( Figure 3A). In addition, we observed a higher proportion of OE-CCDC12 HCT116 cells in the G0/G1 phase, with a lower proportion of cells in the S and G2/M phase. For SW480-KD cells in the OE-CCDC12 group, cells were more arrested in the G0/G1 phase, while reduced in the G2/M phase ( Figure 3B). We found a higher statistically signi cant rate of cell proliferation in OE-CCDC12 group through MTS assays ( Figure S5B). Next, we evaluated cell migration ability using wound-healing assays. There were slight differences in migration ability between cells in OE-CCDC12 group and NC group at 24 hours. However, after 48 hours, the cell migration ability in OE-CCDC12 group was much higher ( Figure   3C). We then measured the invasive ability of cells using matrigel-coated transwell chambers. As shown in Figure 3D, the invasive ability of cells in OE-CCDC12 group was higher compared to cells in NC and Vec groups, and especially for SW480-KD cell lines. Furthermore, we observed that the total apoptotic rate of cells was reduced by approximately 50% in OE-CCDC12 group ( Figure 3E). This was more evident in HCT116 cell lines, speci cally, the ratio of cells in early apoptosis was reduced. We next performed in vivo studies by injecting OE-CCDC12 HCT116 or NC HCT116 cells subcutaneously into the axilla of nude mice. Mice were then euthanized 4 weeks later to determine tumor growth. Tumor volume and weight in nude mice injected with OE-CCDC12 HCT116 cells were larger compared to NC group ( Figure 3G). These ndings demonstrate that over-expression of CCDC12 increases colon cancer cell proliferation, migration, and invasiveness both in vivo and in vitro while reducing apoptosis levels.
3.5 CCDC12 induces epithelial-mesenchymal transition (EMT) to aggravate COAD iTRAQ was used to quantitatively determine differential protein expression in cells after CCDC12 overexpression, and we compared their expression in OE-CCDC12 HCT116 cells and NC HCT116 cells. We set the threshold for fold change (FC) to ≥ 1.3 or ≤ 0.77 and q value < 0.05. 127 proteins had higher expression in OE-CCDC12 HCT116 cells, while 42 proteins had lower expression ( Figure 4A). We analyzed the differentially expressed proteins using the COG database to predict possible functions and functionally classify these proteins ( Figure 4B). Enriched protein classi cations included translation, ribosomal structure, and biogenesis; transcription; replication, recombination, and repair; posttranslational modi cation, protein turnover, chaperones; signal transduction mechanisms, etc., all of which are important functional classi cations associated with cancer. We then performed cluster analysis on differentially expressed proteins. The top 50 differentially expressed proteins clustered into several groups ( Figure 4C) to classify CCDC12 interacting proteins, and the full heatmap shown in Figure S7. GO analysis was used to further classify the function of these differentially expressed proteins. We have separately annotated the increased and decreased proteins, and we only describe the increased proteins. As for Biological Process (BP) classi cations, proteins were mostly annotated into transcription (DNAtemplated), negative regulation of transcription from RNA polymerase promoter, cell division and cell proliferation ( Figure 4D); for Molecular Function (MF), the majority were protein binding, enzyme binding, protein kinase binding and Zinc ion binding ( Figure 4E); for Cellular Component (CC), cytoplasmic ribonucleoprotein granule and nucleus ( Figure 4F). In terms of signaling pathways, adherens junction, viral carcinogenesis, and epithelial cell signaling in Helicobacter pylori infection were annotated ( Figure  4G). These results suggested that CCDC12 may regulate EMT in COAD, more speci cally via the regulation of intercellular adherens junction. To support these ndings, we performed western blotting using HCT116 cells overexpressing CCDC12 and compared them with naive HCT116 cells. As shown in Figure 4H, CCDC12 expression was associated with the biomarkers in EMT process. Overexpression of CCDC12 induced signi cant changes in E-cadherin, Vimentin, Fibronectin, Matrix Metallopeptidase 9 (MMP 9), Snail, and Slug protein levels. In particularly, the expression levels of Snail and Slug dramatically changed in HCT116 cells after overexpressing CCDC12. In summary, CCDC12 regulated COAD by altering the expression levels of several biologically functional proteins that were associated with EMT, especially zinc nger transcription factors.
After determining the rate of transfection by a uorescence microscope, we measured expression levels of Snail by western blotting ( Figure S8B) and qPCR ( Figure S8C). We then used siRNA1 to knockdown CCDC12 in SW480 cells, where overexpressed Snail. Western blotting demonstrated that compared to cells in NC group, expression levels of CCDC12 in OE-Snail cells did not change. Knockdown of CCDC12 resulted in the expression level of Snail decreasing to a certain extent ( Figure 5A). To determine whether CCDC12 regulates EMT dependent on Snail, we used transwell chamber to evaluate the invasive ability of cells knocked down CCDC12 after overexpressing Snail. Overexpression of Snail signi cantly increased cell invasion. However, knocking down CCDC12 in Snail overexpressing cells reduced cell invasion but was still higher compared to cells in the NC group ( Figure 5B). These results suggested that CCDC12 regulates EMT in COAD by affecting Snail expression.

Discussion
Adenocarcinoma is the main pathological type of colon cancer and accounts for more than 95%, especially in Asia. Based on its biological characteristics, COAD is highly metastatic, which results in poor ve-year patient survival (1-3). Patients who have undergone systemic chemotherapy after surgery have not shown improvements in survival rates (15). With the development and innovation of surgical techniques, the prognosis of patients with colon cancer seems to have improved. However, these new surgical technologies have not been fully adopted (16,17). Concurrently, large clinical trials using targeted therapies have been performed worldwide (4,18,19), to nd treatments that could effectively improve patient prognosis and inhibit colon cancer metastasis.
The CCDC12 gene, located on 3p21.31 of chromosome 3, has been reported to play a role in colorectal tumorigenesis (12). CCDC12 has been demonstrated to accelerate the growth of K562 cells by upregulating CD235, ε-globin, and γ-globin in human chronic myeloid leukemia (8). Anne et. al. reported that CCDC12 may be associated with ubiquitination (9), which is particularly critical in tumor cells and could participate in the modi cation and degradation of some cancer factors to affect the biological behavior of tumors. In addition, Ke et al. using GWAS found that CCDC12 may be a potential risk gene for colorectal cancer and associated with a potential regulatory variant, rs1076394 (12). In this study, we demonstrated that CCDC12 was highly expressed in colon adenocarcinomas and may affect cancer metastasis by regulating EMT.
We analyzed GWAS data from NHGRI database, and then performed deep mining of colorectal cancerrelated data from TCGA. Our team found numerous colorectal cancer-related risk SNPs, which were veri ed through 130 additional samples. With this, we identi ed rs10505477 and rs6983267 that had high pairwise linkage disequilibrium. After integrative eQTL-based analysis, we identi ed 25 SNP-gene expression associations in 22 risk SNP loci, and multiple corrections were performed to improve the reliability. We found SNP rs8180040 and CCDC12 had a signi cant correlation, both of which were located on chromosome 3 and had a strong geometric correlation. SNP rs8180040 was located at chr3: 47347457 (GRCh38.p12) with T > A allele (T = 0.5992 and A = 0.4008). The gene frequency of rs8180040 was slightly different in different populations. The Asian population had the highest allele A frequency of 0.5, especially in East Asian populations; the African population had the lowest frequency of 0.23; the European population was close to the average (A = 0.4088), (12,20,21) and our results were similar to previous studies. Hence, we inferred that rs8180040 may affect the occurrence of colorectal cancer by regulating CCDC12 expression.
We then measured the expression levels of CCDC12 in cancer and normal tissues by IHC staining. Using 75 pairs of COAD and matched adjacent normal mucosa samples demonstrated that CCDC12 was frequently overexpressed in COAD (68.0%). We sequentially performed con rmatory western blotting through 12 fresh COAD samples and demonstrated higher expression levels of CCDC12 compared to matched adjacent normal mucosa samples. Based on a large-scale RNA-Seq dataset in TCGA, we con rmed that CCDC12 was frequently overexpressed in COAD tissues. Cox regression analysis performed by gender showed that CCDC12 was a prognostic factor. Interestingly, in female patients, CCDC12 overexpression leads to a poorer prognosis, but not in male patients. This was the focus of our study in subsequent experiments. These ndings demonstrated that CCDC12 plays a potential role in COAD tumorigenesis and development.
We next performed a series of in vivo and in vitro experiments to determine how CCDC12 regulates the proliferation, migration, and invasion of COAD cells to understand its biological role. After knockdown of CCDC12, cell proliferation, invasion, and migration were reduced, while rate of apoptosis increased. In addition, knockdown of CCDC12 resulted in a higher proportion of cells in the G0/G1 phase. In contrast, overexpression of CCDC12 increased cell proliferation, invasion, and migration, while reducing apoptosis levels. Interestingly, overexpression of CCDC12 still resulted cells in a block in G0/G1 phase. If cells do not enter S phase and are in a quiescent state, which meaning not participate in cell division. Additional studies are required to understand these ndings in the future. Using xenograft tumor mouse models, knockdown of CCDC12 signi cantly reduced tumor size and weight, while CCDC12 overexpression promoted xenograft tumor growth. These ndings strongly demonstrate that increased expression of CCDC12 was essential for COAD tumorigenesis and development.
We next investigated the role of CCDC12 in COAD metastasis. Through iTRAQ assays, we observed that overexpression of CCDC12 could induce expression levels changes in 169 proteins. In COG analysis, about 1,000 proteins annotated with de nite functions and enriched in the following annotations: translation, ribosomal structure and biogenesis (22); transcription(23); replication, recombination and repair (24); posttranslational modi cation, protein turnover, chaperones (25); and signal transduction mechanisms (26), all of which have been strongly associated with cancer. Cluster analysis divided the differentially expressed proteins into two groups and was in good agreement with the original groups. For proteins that changed in expression levels after CCDC12 overexpression, we performed GO analysis to annotate their functions. For BP classi cation, the majority of proteins were annotated to transcription (DNA-templated), negative regulation of transcription from RNA polymerase promoter, cell division, and cell proliferation. These constituted the core functions involved in cancer biology, which were based on changes in protein-coding genes and non-coding regulatory elements (27), especially for DNA template transcription. We also classi ed differentially expressed proteins into MF terms, which included zinc ion binding, protein binding, enzyme binding, and protein kinase binding. Metal ions form the basis for higher-order structures in protein assembly and provide short peptides chemical reactivity (28). In addition, zinc ions play a key role in homeostasis, immune function, and apoptosis (29), and may also induce p53 misfolding (30). With regards to CC classi cation, several proteins were annotated to the nucleus, cytoplasmic ribonucleoprotein granule, and nuclear chromatin. Chromatin state signi cantly affects the occurrence and treatment of cancers through speci c regulatory mechanisms (31), additionally, the state of plasma membrane is critical for intercellular communication (32). What's more, we observed that CCDC12 was localized in the nucleoplasm, which suggested that CCDC12 may affect proteins in the nucleus, such as Snail (localized in the nucleus and cytosol). KEGG pathway enrichment analysis demonstrated that cancer-related pathways such as adherens junction and viral carcinogenesis were annotated. Early EMT is associated with the overall deterioration of cell-cell adhesion, which triggers front-rear polarization of cells required for migration(33). It is gratifying that in the adherens junction pathway, Snail and Slug are its typical representatives, both of which could be detected in the nucleus.
And bioinformatics analysis based on the results of iTRAQ assay further con rmed that overexpression of CCDC12 induced changes in expression levels of proteins, especially in the nucleus. All these ndings demonstrated that CCDC12 may regulate EMT of colon cancer cells through Snail located in the nucleus.
To con rm that CCDC12 overexpression was involved in the invasion and malignancy of COAD, we measured typical EMT biomarkers by western blotting. Overexpression of CCDC12 reduced E-cadherin level, but increased expression levels of Fibronectin, Vimentin, MMP 9, Snail, and Slug. During EMT, epithelial cells lose their cell polarity and connection with the basement membrane, resulting in the ability to migrate and invade, resist apoptosis, and degrade extracellular matrix(34), which were consistent with our in vitro experiments. iTRAQ assays and western blotting data suggested that Snail may be a downstream target of CCDC12, because not only Snail expression levels positively correlated with that of CCDC12 but also showed the highest fold change in expression levels by western blotting ( Figure 4H). Through lentivirus, we made Snail overexpressed in SW480 cells, and then knocked down CCDC12.
Overexpression of Snail had no signi cant impact on CCDC12 expression levels, while knockdown of CCDC12 decreased Snail expression levels ( Figure 5A). Additionally, we performed transwell assays and demonstrated that Snail overexpression effectively increased the invasion ability of colon cancer cells, and after subsequent knockdown of CCDC12, the invasive ability promoted by overexpression of Snail was corrected to some extent. These demonstrated that Snail was a downstream target of CCDC12. As to how CCDC12 regulates Snail to affect colon cancer EMT requires additional follow-up studies.

Conclusion
We demonstrated that CCDC12 plays an important role in the tumorigenesis and development of COAD.
Higher expression levels of CCDC12 in COAD was signi cantly associated with malignancy and metastasis. CCDC12 may be a potential oncogene that regulates EMT in COAD by affecting the Snail pathway, which and then affects the prognosis of patient. Additionally, CCDC12 is expected to be a potential therapeutic target for COAD.