POFUT1 as a Promising Novel Biomarker of Colorectal Cancer

Background: While protein O-fucosyltransferase 1 (POFUT1) overexpression has been recently proposed as a potential biomarker for different cancer types, no study was carried out on POFUT1 implication in colorectal cancer (CRC). Methods: Data from 626 tumors and 51 non-tumor adjacent tissues available in FireBrowse had been used in this study. Statistical analyses on POFUT1 expression and gene copy number, NOTCH receptors (main targets of POFUT1 enzymatic activity) expression and association of POFUT1 and NOTCH1 expressions with clinical parameters were investigated. Data were completed by POFUT1 histological labeling on six tumor tissues from patients with CRC. Results: We found that POFUT1 is overexpressed from the stage I (p < 0.001) and 76.02% of tumors have a 20q11.21 amplification, associated in 90.13% of cases with a POFUT1 overexpression, compared to non-tumor adjacent tissues. The POFUT1 copy number in tumors is mainly between 2 and 3. POFUT1 is positively correlated with NOTCH1 (rs = 0.34, p < 0.001), NOTCH3 (rs = 0.087, p = 0.0297), and NOTCH4 (rs = 0.097, p = 0.0148) expressions, while negatively correlated with NOTCH2 expression (rs = −0.098, p = 0.0142). POFUT1 overexpression is markedly associated with rectal location, non-mucinous adenocarcinoma and cancer stages IV and M1. NOTCH1 overexpression is only associated with rectal location and non-mucinous adenocarcinoma. Conclusion: We conclude that POFUT1 is overexpressed in CRC from stage I, and its high expression is associated with metastatic process, probably through NOTCH pathway activation. Then, POFUT1 could represent a potential novel biomarker for CRC diagnosis.


Introduction
Colorectal cancer (CRC) is the third most commonly diagnosed cancer in males and the second in females with 1.65 million new cases and almost 835,000 deaths in 2015 [1]. The majority of CRC (75%) has a sporadic origin but in some cases the origin is related to familial heredity or due to inflammatory bowel diseases [2]. Although the mortality associated with CRC declined over the past decades, identification of new biomarkers for an early diagnosis and the improved treatment of CRC are crucial. Previous studies have demonstrated the association between glycosylation changes and tumorigenesis [3,4]. Glycosylation is the main post-translational modification of proteins. Nand/or O-glycans play major roles as in protein conformation then modulating their functional activity [5], in ligand-receptor complex formation for cell-cell interactions [6], and in cellular metabolism [7]. Fucose LGG: brain lower grade glioma, LIHC: liver hepatocellular carcinoma, LUAD: lung adenocarcinoma, LUSC: lung squamous cell carcinoma, MESO: mesothelioma, OV: ovarian serous cystadenocarcinoma, PAAD: pancreatic adenocarcinoma, PCPG: pheochromocytoma and paraganglioma, PRAD: prostate adenocarcinoma, READ: rectum adenocarcinoma, SARC: sarcoma, SKCM: skin Cutaneous Melanoma, STAD: stomach adenocarcinoma, STES: stomach and esophageal carcinoma, TGCT: testicular germ cell tumors, THCA: thyroid carcinoma, THYM: Thymoma, UCEC: uterine corpus endometrial carcinoma, UCS: uterine carcinosarcoma, UVM: uveal melanoma. COADREAD RNAseq data extracted from FireBrowse database containing 626 CRC and 51 normal adjacent tissues show that POFUT1 is significantly overexpressed in tumor tissues (B) and from the first stage of tumor classification (C). For B. and C., bar graph represented mean of log2 RSEM ± SEM. Statistical significance was assessed using a two-tailed Student test; * p < 0.05, *** p < 0.001. To investigate in detail POFUT1 expression in colorectal cancer (CRC), an in silico analysis using the RNAseq data of COADREAD samples extracted from FireBrowse was performed using 626 tumor and 51 adjacent non tumor tissues. POFUT1 expression is significantly higher in 459 (72.8%) tumor compared to normal tissues (p < 0.001) ( Figure 1B). The distinction between cancer stages showed a significant (p < 0.001) increase in POFUT1 expression whatever the stage is, therefore at the first signs of the tumor growth ( Figure 1C). Stage II presented a lesser amount of POFUT1 transcripts compared to other stages. POFUT1 immunolabeling performed on tumors representing each CRC pathological stage confirms that POFUT1 is overexpressed in tumor (Figure 2A). To demonstrate anti-POFUT1 antibody (ab74302) specificity, we performed an immunofluorescence detection on two human colorectal cancer cell lines (HCT 116 and SW620) stably transfected or not, with shRNAs targeting POFUT1. As shown in Figure 2B, HCT 116 shPOFUT1 cell line, whose POFUT1 expression is 30% lesser (quantification by Taqman probe qRT-PCR method), has a lower staining compared to HCT 116. This result is more accentuated with SW620 shPOFUT1 cell line, which has 60% POFUT1 expression decrease compared to SW620. As the POFUT1 antibody (ab74302) was ineffective in immunoblotting, we used another antibody raised against Pofut1 and produced in our laboratory [39]. This antibody has been proven in different studies especially in mice [24,40]. Despite a low quality of protein migration due to the Optimal Cutting Temperature (OCT) embedded colorectal tissues, we observed an increase (1.134 and 1.565 fold) of POFUT1 labeling in tumor samples compared to normal tissues ( Figure 2C). Such a result was confirmed on human colorectal cancer cell lines HCT 116, HT-29 and SW620 where the expression levels were respectively 2.680, 2.418 and 2.608 fold higher compared to the human embryonic colon cell line CCD841CoN ( Figure 2D). melanoma. COADREAD RNAseq data extracted from FireBrowse database containing 626 CRC and 51 normal adjacent tissues show that POFUT1 is significantly overexpressed in tumor tissues (B) and from the first stage of tumor classification (C). For B. and C., bar graph represented mean of log2 RSEM ± SEM. Statistical significance was assessed using a two-tailed Student test; * p < 0.05, *** p < 0.001.
To investigate in detail POFUT1 expression in colorectal cancer (CRC), an in silico analysis using the RNAseq data of COADREAD samples extracted from FireBrowse was performed using 626 tumor and 51 adjacent non tumor tissues. POFUT1 expression is significantly higher in 459 (72.8%) tumor compared to normal tissues (p < 0.001) ( Figure 1B). The distinction between cancer stages showed a significant (p < 0.001) increase in POFUT1 expression whatever the stage is, therefore at the first signs of the tumor growth ( Figure 1C). Stage II presented a lesser amount of POFUT1 transcripts compared to other stages. POFUT1 immunolabeling performed on tumors representing each CRC pathological stage confirms that POFUT1 is overexpressed in tumor ( Figure  2A). To demonstrate anti-POFUT1 antibody (ab74302) specificity, we performed an immunofluorescence detection on two human colorectal cancer cell lines (HCT 116 and SW620) stably transfected or not, with shRNAs targeting POFUT1. As shown in Figure 2B, HCT 116 shPOFUT1 cell line, whose POFUT1 expression is 30% lesser (quantification by Taqman probe qRT-PCR method), has a lower staining compared to HCT 116. This result is more accentuated with SW620 shPOFUT1 cell line, which has 60% POFUT1 expression decrease compared to SW620. As the POFUT1 antibody (ab74302) was ineffective in immunoblotting, we used another antibody raised against Pofut1 and produced in our laboratory [39]. This antibody has been proven in different studies especially in mice [24,40]. Despite a low quality of protein migration due to the Optimal Cutting Temperature (OCT) embedded colorectal tissues, we observed an increase (1.134 and 1.565 fold) of POFUT1 labeling in tumor samples compared to normal tissues ( Figure 2C). Such a result was confirmed on human colorectal cancer cell lines HCT 116, HT-29 and SW620 where the expression levels were respectively 2.680, 2.418 and 2.608 fold higher compared to the human embryonic colon cell line CCD841CoN ( Figure 2D).

In CRC, 20q11.21 Chromosomic Region is Often Amplified, Which Induced POFUT1 Copy Number Alteration
To determine if a link exists between POFUT1 chromosomic region state (20q11.21) and its overexpression, an in silico analysis was performed. The study showed that among 613 patients with CRC, 76.02% had an amplification of the 20q11.21 region, which correlated, in 90.13% of cases, with the increase in POFUT1 expression compared to healthy patients ( Figure 3A). Interestingly, around 80% of CRC patients who had no 20q11.21 amplification presented a lower POFUT1 expression compared to healthy patients. A significant positive correlation exists between copy number and POFUT1 expression (rS = 0.774, p < 0.001) ( Figure 3B), suggesting that POFUT1 transcript quantity is predominantly due to the gene copy number. Furthermore, copy number analysis of POFUT1 gene showed that 20q11.21 chromosomic region amplification mostly generates between two and three POFUT1 copies per genome (49%) and no more than six copies ( Figure 3C). Copy number analysis performed on six selected CRC tumors including those immunolabeled by anti-POFUT1 revealed an increase of POFUT1 copy number in five patients with in majority of cases between 2 and 3 copies, like in bioinformatics analysis ( Figure 3D).

In CRC, 20q11.21 Chromosomic Region is Often Amplified, Which Induced POFUT1 Copy Number Alteration
To determine if a link exists between POFUT1 chromosomic region state (20q11.21) and its overexpression, an in silico analysis was performed. The study showed that among 613 patients with CRC, 76.02% had an amplification of the 20q11.21 region, which correlated, in 90.13% of cases, with the increase in POFUT1 expression compared to healthy patients ( Figure 3A). Interestingly, around 80% of CRC patients who had no 20q11.21 amplification presented a lower POFUT1 expression compared to healthy patients. A significant positive correlation exists between copy number and POFUT1 expression (r S = 0.774, p < 0.001) ( Figure 3B), suggesting that POFUT1 transcript quantity is predominantly due to the gene copy number. Furthermore, copy number analysis of POFUT1 gene showed that 20q11.21 chromosomic region amplification mostly generates between two and three POFUT1 copies per genome (49%) and no more than six copies ( Figure 3C). Copy number analysis performed on six selected CRC tumors including those immunolabeled by anti-POFUT1 revealed an increase of POFUT1 copy number in five patients with in majority of cases between 2 and 3 copies, like in bioinformatics analysis ( Figure 3D). expression is significantly correlated with its copy number. To only view additional copies of POFUT1 gene, a subtraction of two copies corresponding to a physiological state is applied for each sample. (C) CRC patients have in the majority of cases between two and three POFUT1 gene copies.
(D) POFUT1 copy number analysis performed on gDNA extracted from six CRC and one normal tissues shows an increase of POFUT1 copy number in five CRC cases compared to the normal sample.

Correlation between POFUT1 and NOTCH Receptor Expressions
Since the cross talk between POFUT1 and NOTCH receptors has been demonstrated in hepatocellular carcinoma and gastric cancer [32,33], a Spearman's correlation coefficient was used to determine their relationships in CRC. A significant positive correlation was observed between POFUT1 and NOTCH1 (rs = 0.34, p < 0.001), NOTCH3 (rs = 0.087, p = 0.0297) and NOTCH4 (rs = 0.097, p = 0.0148) receptors ( Figure 4). Furthermore, a significant negative correlation was detected between have a POFUT1 overexpression compared to the POFUT1 mean expression in non-tumor adjacent tissues. (B) Spearman Rho correlation analysis in 613 CRC patients shows that POFUT1 expression is significantly correlated with its copy number. To only view additional copies of POFUT1 gene, a subtraction of two copies corresponding to a physiological state is applied for each sample. (C) CRC patients have in the majority of cases between two and three POFUT1 gene copies. (D) POFUT1 copy number analysis performed on gDNA extracted from six CRC and one normal tissues shows an increase of POFUT1 copy number in five CRC cases compared to the normal sample.

Correlation between POFUT1 and NOTCH Receptor Expressions
Since the cross talk between POFUT1 and NOTCH receptors has been demonstrated in hepatocellular carcinoma and gastric cancer [32,33], a Spearman's correlation coefficient was used to determine their relationships in CRC. A significant positive correlation was observed between POFUT1 and NOTCH1 (r s = 0.34, p < 0.001), NOTCH3 (r s = 0.087, p = 0.0297) and NOTCH4 (r s = 0.097, p = 0.0148) receptors ( Figure 4). Furthermore, a significant negative correlation was detected between POFUT1 and NOTCH2 (r s = −0.098, p = 0.0142). Among all correlations, POFUT1/NOTCH1 one was the strongest.

NOTCH Signaling Pathway is Deregulated in CRC
Although the expressions of POFUT1 and NOTCH1 receptor are significantly and positively correlated, it is necessary to characterize the expression of the NOTCH target genes in order to highlight a potential deregulation of the signaling pathway. We were interested in HES/HEY transcription factor gene family especially HES1 and HEY1 widely studied in NOTCH pathway analysis, p21 (CDKN1A) and Cyclin D1 (CCND1) that encode cell cycle regulators, c-Myc (MYC) which is an oncogene, Snail 1 (SNAI1) implicated in EMT and Survivin (BIRC5) related to apoptosis regulation ( Figure 5). All genes, except HES1, are significantly modified in their expression levels in tumor compared to normal tissues (p < 0.001). Cyclin D1 and c-Myc that induce proliferation, Snail 1 that promotes EMT and Survivin that inhibits apoptosis are overexpressed. p21, a negative regulator of cell cycle, and HEY1 transcription factor mediator of Notch signaling, are downregulated. Taken together these results demonstrate that NOTCH signaling is altered in CRC.

NOTCH Signaling Pathway is Deregulated in CRC
Although the expressions of POFUT1 and NOTCH1 receptor are significantly and positively correlated, it is necessary to characterize the expression of the NOTCH target genes in order to highlight a potential deregulation of the signaling pathway. We were interested in HES/HEY transcription factor gene family especially HES1 and HEY1 widely studied in NOTCH pathway analysis, p21 (CDKN1A) and Cyclin D1 (CCND1) that encode cell cycle regulators, c-Myc (MYC) which is an oncogene, Snail 1 (SNAI1) implicated in EMT and Survivin (BIRC5) related to apoptosis regulation ( Figure 5). All genes, except HES1, are significantly modified in their expression levels in tumor compared to normal tissues (p < 0.001). Cyclin D1 and c-Myc that induce proliferation, Snail 1 that promotes EMT and Survivin that inhibits apoptosis are overexpressed. p21, a negative regulator of cell cycle, and HEY1 transcription factor mediator of Notch signaling, are downregulated. Taken together these results demonstrate that NOTCH signaling is altered in CRC.

Relationship between POFUT1, NOTCH1 Expression,s and Clinical Features
To further explore the association between POFUT1 and NOTCH1 in CRC progression, analysis of their expressions compared to the mean value of healthy patients was studied in the light of various clinical parameters in CRC patients (Tables 1 and 2). POFUT1 expression is significantly associated with tumor issue site (p = 0.0001), overexpressed in 68.9% of colon and 84.6% of rectum tissues (Table 1). It is linked to pathologic stage (p = 0.00019) and markedly overexpressed in 79% of stage I, 63.3% of stage II, 74.9% of stage III and 85.2% of stage IV. POFUT1 expression is associated with M classification (p = 0.01087) and overexpressed in 70.7% of M0 stage and 83.9% of M1 stage. In addition, POFUT1 is differently expressed according to histological type (p = 0.00001) with an overexpression in 75% of colon adenocarcinoma, 30.6% of colon mucinous adenocarcinoma, 88.4% of rectal adenocarcinoma and 46.2% of rectal mucinous adenocarcinoma. However, no correlation was observed between POFUT1 expression and gender, age, T and N classifications.

Relationship between POFUT1, NOTCH1 Expressions, and Clinical Features
To further explore the association between POFUT1 and NOTCH1 in CRC progression, analysis of their expressions compared to the mean value of healthy patients was studied in the light of various clinical parameters in CRC patients (Tables 1 and 2). POFUT1 expression is significantly associated with tumor issue site (p = 0.0001), overexpressed in 68.9% of colon and 84.6% of rectum tissues (Table 1). It is linked to pathologic stage (p = 0.00019) and markedly overexpressed in 79% of stage I, 63.3% of stage II, 74.9% of stage III and 85.2% of stage IV. POFUT1 expression is associated with M classification (p = 0.01087) and overexpressed in 70.7% of M0 stage and 83.9% of M1 stage. In addition, POFUT1 is differently expressed according to histological type (p = 0.00001) with an overexpression in 75% of colon adenocarcinoma, 30.6% of colon mucinous adenocarcinoma, 88.4% of rectal adenocarcinoma and 46.2% of rectal mucinous adenocarcinoma. However, no correlation was observed between POFUT1 expression and gender, age, T and N classifications. NOTCH1 expression was significantly associated with tumor issue site (p = 0.00099) and overexpressed in 70.9% of colon and 84% of rectum tissues ( Table 2). It is associated to histological type (p = 0.00173) and NOTCH1 is overexpressed in 73.2% of colon adenocarcinoma, 59.7% of colon mucinous adenocarcinoma, 84.4% of rectal adenocarcinoma and 76.9% of rectal mucinous adenocarcinoma. No correlation was observed between NOTCH1 expression and gender, age, pathological stage, T, N and M classifications. In addition, we classified CRC patients into four groups according to their combined expression status of POFUT1 and NOTCH1 as follow: high POFUT1/high NOTCH1, low POFUT1/low NOTCH1, high POFUT1/low NOTCH1 and low POFUT1/high NOTCH1. The threshold value which allows to classify individuals in the high and low groups is the mean expression value of POFUT1 and NOTCH1 in healthy patients. The associations between these groups and clinical features were analyzed in Table 3. Combined POFUT1/NOTCH1 expressions were significantly associated with tumor issue site (p = 0.00004), pathologic stage (p = 0.00498) and histological type (p < 0.001). It should be noted that in all cases, the majority of CRC are located in high/high group.

Discussion
Comprehension of mechanisms which initiate tumor development is crucial since an early diagnosis of cancer can trigger treatment and increase the patient chances of recovery. Therefore, research of new potential diagnostic markers of cancer occupies a substantial part in the scientific field. Recently, many studies focused on glycosylation, especially in malignant tumor development [41]. In that context, O-fucosylation, linked to expression of Protein O-fucosyltransferase 1 (POFUT1) and its activity on EGF-like domains, appears promising. POFUT1 adds O-fucose on S or T residues within the consensus sequence C 2 X 4 (S/T)C 3 of EGF-like domains [42] present in some cell surface and secreted proteins [43]. In humans, 87 putative POFUT1 targets had been referenced [44]. Among those, NOTCH receptors are the most described in literature and their O-fucosylation was shown to be essential for their interaction with ligands and therefore for NOTCH signaling [24]. POFUT1 and NOTCH cross talk had been described in two cancer types. In breast cancer, an overexpression of POFUT1 and NOTCH1 was associated with lymph node metastasis and advanced tumor stage [45]. In hepatocellular carcinoma, POFUT1 overexpression induced an aberrant activation of NOTCH pathway switching on HES1, which in turn promoted migration and cell proliferation [32]. Currently, no study focused on the implication of POFUT1 in colorectal cancer, although it is a major public health issue. Colorectal cancer is one of the cancers where POFUT1 is the most overexpressed. Here, bioinformatics combined with immunohistochemistry, western blot and gene copy number analysis had been used as an approach to determine if POFUT1 could be a potential novel CRC biomarker. Among the data of 626 CRC patients available in FireBrowse database, 459 (72.8%) had a POFUT1 overexpression compared to healthy patients. The overexpression was detected from the first stage of CRC. POFUT1 labeling on CRC biopsies confirmed the overexpression in tumor compared to the adjacent non-tumor tissues. The chromosomic region 20q11.21, where POFUT1 gene is located, appears to be unstable leading to gene copy number variation, which could explain the expression increase [46]. In the CRC panel, we observed a 20q11.21 amplification in 466 cases over 613 (76.02%), which induces an increase of POFUT1 gene copy number to 5.7 copies. Our POFUT1 copy number analysis performed on six CRC tissues follows the same trend as the bioinformatics analysis. As expected, the small sample size does not allow observing the whole range of copy number alterations. A strong positive correlation (r s = 0.774) between POFUT1 copy number and its expression argues for a direct link as already noticed [38]. It should be noted that the significant decrease of POFUT1 expression in stage II ( Figure 2B) is mainly due to a greater proportion of cases without 20q11.21 chromosomic region amplification (33.9%) compared to other stages (Stage I, 23.1%; Stage III, 18.0%; Stage IV, 13.8%). Significant correlations between POFUT1 and NOTCH receptor expressions were measured with the strongest for POFUT1/NOTCH1 association (r s = 0.34). POFUT1, NOTCH1 and POFUT1/NOTCH1 high expressions are significantly associated with the tumor issue site, preferentially overexpressed in rectum tissue (84.6%, 84%, 76.1%, respectively). Several other studies highlighted different gene expressions and genetic features associated with carcinogenesis between colon and rectum [47,48]. Interestingly, POFUT1, and not NOTCH1, expression appears to be significantly associated with M classification. POFUT1 is predominantly overexpressed in colorectal metastasis (83.9%) and could O-fucosylate other protein targets than NOTCH receptors, such as AGRIN which was shown to enhance tumor progression by activating cell migration and invasion in oral cancer [49]. Overexpression of POFUT1 and NOTCH1 is preferentially observed in non-mucinous adenocarcinoma histological type. This observation can be explained by the fact that mucinous adenocarcinoma are characterized by a markedly reduced rate of copy-number aberrations compared to adenocarcinoma [50]. Indeed, in mucinous adenocarcinoma, only 42.6% of cases had an amplified 20q11.21 region. Regarding NOTCH signaling activation, it is known that in tumor tissues a greater activation of NOTCH pathway is involved in cell proliferation and metastasis process [51,52]. In this study on CRC, an increase of NOTCH activation is supported by the higher expression levels of its target genes such as p21, Cyclin D1, c-Myc, Survivin and Snail 1. However, the expression of HES1, a proved Notch signaling downstream target, is not modified in tumor compared to healthy tissues ( Figure 5). Nevertheless, studies in relation with HES1 expression in CRC are controversial [53][54][55] suggesting that it is not a good marker of NOTCH signaling activation in colorectal cancer. We also showed that HEY1 expression is downregulated although this gene is also known to be activated by Notch signaling [56,57]. As well, conflicting studies showed that this NOTCH target transcriptional factor was overexpressed [58] or non-expressed [59] in colorectal cancer. Surprisingly, significant positive correlations between the expressions of POFUT1 and HES1 or HEY1 are found for healthy tissues and not for tumor ones ( Figure S1). It could be explained by the cell heterogeneity of tumors. Therefore, the consequences of POFUT1 overexpression on NOTCH signaling activation could be opposite depending on the NOTCH target genes. The effect of POFUT1 overexpression on O-fucosylation levels and NOTCH signaling would be cell-type dependent. Overexpression of POFUT1 most likely does not result in increased O-fucosylation of NOTCH receptors. Indeed, in HEK293T cells, most EGF-like repeats containing O-fucose consensus sequences are O-fucosylated at high stoichiometry degree [60]. If it is the case in colorectal cancer, the overexpression of POFUT1 may affect the O-fucosylation state of other proteins. Nevertheless, it is important to note that in the majority of CRC cases analyzed in the present study, both POFUT1 and NOTCH1 are overexpressed (60.5%), suggesting that overexpression of POFUT1 is necessary to ensure O-fucosylation of additional NOTCH receptors in the tumor. Finally, in addition to its O-fucosyltransferase activity, it had been demonstrated an independent chaperone function for the POFUT1 orthologue in Drosophila melanogaster [61]. However, this additional function is still controversial in mammals [24]. Ajima et al. (2017) [62] showed that in mouse it is not possible to dissociate the possible chaperone contribution from its O-fucosyltransferase activity, which could also be the case for human POFUT1. Our study focuses on POFUT1 expression level related to its copy number determined by 20q11.21 chromosomic region state. Nevertheless we cannot exclude that gene expression level can also be modified by other mechanisms such as mutations within promoter or by miR-34 family regulation [63]. The lack of these informations in the database did not allow us to explore these expression regulatory mechanisms.

The Cancer Genome Atlas Data Analysis
Data for colorectal carcinoma were extracted from FireBrowse database (http://www. firebrowse.org). A total of 626 tumor samples and 51 normal samples were studied. Gene expression levels were merged from COADREAD.uncv2.mRNAseq_RSEM_normalized_log2.txt found in COADREAD.mRNAseq_Preprocess.Level file.

Statistical Analysis
Statistical analyses were performed using Past3 3.20 version [64] and GraphPad Prism 7 (GraphPad Software Inc, San Diego, CA, USA). mRNA expression data were referenced as mean ± SEM and a t-Student test was applied to compare values between normal and tumor tissues. Bivariate correlation analysis between POFUT1 and NOTCH receptor mRNA expressions was performed using Spearman's Rho. Associations between POFUT1, NOTCH1 expressions and clinicopathological parameters were estimated by a Chi-square test. Results were considered statistically significant if the p-value was less than 0.05.

Genomic DNA Extraction and POFUT1 Copy Number Analysis
Genomic DNA was extracted from normal and tumor tissues with Maxwell ® 16 FFPE Plus LEV DNA Purification Kit and Maxwell ® 16 IVD device (Promega, Madison, WI, USA) according to the manufacturer's protocol. Genomic DNA concentration was determined using Quantifluor ® ONE dsDNA system (Promega) and measured with Quantus™ Fluorometer (Promega) following manufacturer's recommendations. Taqman™ copy number assay for POFUT1 (Hs02487189_cn) and RNAse P reference assays (4403326) were used with Gene Expression Master Mix (Applied Biosystems™, Thermo Fisher Scientific, Waltham, MA, USA), according to product literature. Twenty nanograms of gDNA were run in triplicate on QuantStudio 3 real-time PCR system (Applied Biosystem™). POFUT1 copy number was estimated using ∆∆Ct method [65].

POFUT1 Labelling by Immunohistochemistry
Paraffin-embedded blocks of six colorectal adenocarcinomas corresponding to each CRC pathological stage were obtained from the Tumor Bank (CRBiolim) of Limoges University Hospital. Immunohistochemical analysis was performed on five-µm-thick paraffin sections with anti-POFUT1 antibody (1/25, ab74302, Abcam, Cambridge, UK). Slides were automatically processed (Ventana Benchmark ULTRA, Roche, Meylan, France) according to the protocol supplied by the manufacturer. Images were acquired with NanoZoomer RS 2.0 Hamamatsu (Hamamatsu Photonics, Massy, France). All samples were used in accordance with French bioethics laws regarding patient information and consent. Ethics approval (CRB-CESSION-2018-016) was obtained from the "Comité médico-scientifique de la tumorothèque de l'Hôpital Dupuytren", the bioethics committee of our hospital.

POFUT1 Labeling by Immunofluorescence
Colorectal cancer cell lines, HCT 116 and SW620 obtained from ATCC and the stably transfected cell lines, HCT 116 shPOFUT1 and SW620 shPOFUT1, created by our team were fixed with 4% paraformaldehyde in PBS for 30 min at room temperature and permeabilized with HEPES Triton buffer (20 mM HEPES, 300 mM sucrose, 50 mM NaCl, 3 mM MgCl2, 0.5% Triton X-100, pH 7.4) for 30 min at 4 • C. After three washes with PBS, non-specific binding sites were saturated for 1 h at room temperature with a blocking solution containing 10% goat serum, 1% BSA, 0.1% Triton X-100 in PBS. After one wash with PBS/0.2% BSA, immunolabeling was performed with anti-POFUT1 (ab74302) antibody diluted at 1:100 in PBS/1% BSA overnight at 4 • C. After washes with PBS and PBS/0.2% BSA/0.1% Tween-20, cells were incubated with the F(ab')2 fragment of goat anti-rabbit IgG (H + L) secondary antibody Alexa fluo ® 546 conjugated (Molecular Probes, Life Technology, Eugene, OR, USA) used at 1:1000 in PBS/1% BSA for 15 min in dark at room temperature. After new washes, nuclei were stained with DAPI (Thermo Fisher Scientific) at 1 µg.mL −1 in PBS, 5 min in dark at room temperature. Finally, after three PBS washes, cells were mounted on slides with Fluoromount-G ® (SouthernBiotech, Birmingham, AL, USA) and sealed with glass coverslips. We used the MetaMorph ® software (Molecular Devices, Sunnyvale, CA, USA) to acquire images with a LEICA microsystem DMI6000B inverted epifluorescence microscope.

Conclusions
In conclusion, our findings indicated that POFUT1 is overexpressed in colorectal cancer driven in majority of cases by a 20q11.21 chromosomic region amplification. This aberrant expression may promote carcinogenesis by NOTCH pathway activation. Finally, targeting POFUT1 seems to be a promising strategy for CRC diagnosis.