Bioinformatics and Functional Analyses Implicate Potential Roles for EOGT and L-fringe in Pancreatic Cancers

Notch signaling receptors, ligands, and their downstream target genes are dysregulated in pancreatic ductal adenocarcinoma (PDAC), suggesting a role of Notch signaling in pancreatic tumor development and progression. However, dysregulation of Notch signaling by post-translational modification of Notch receptors remains poorly understood. Here, we analyzed the Notch-modifying glycosyltransferase involved in the regulation of the ligand-dependent Notch signaling pathway. Bioinformatic analysis revealed that the expression of epidermal growth factor (EGF) domain-specific O-linked N-acetylglucosamine (EOGT) and Lunatic fringe (LFNG) positively correlates with a subset of Notch signaling genes in PDAC. The lack of EOGT or LFNG expression inhibited the proliferation and migration of Panc-1 cells, as observed by the inhibition of Notch activation. EOGT expression is significantly increased in the basal subtype, and low expression of both EOGT and LFNG predicts better overall survival in PDAC patients. These results imply potential roles for EOGT- and LFNG-dependent Notch signaling in PDAC.


Introduction
Pancreatic cancer is one of the most fatal malignancies, which causes cancer mortality in most developed countries. It was the fourth leading cause of cancer death in 2017 in the United States, and approximately 57,000 adults were diagnosed with pancreatic cancer in 2020, which represented approximately 3% of the total cancer cases [1,2]. Generally, the 5-year survival rate of pancreatic cancer is approximately 4% [3]. More than 90% of pancreatic cancer is considered pancreatic ductal adenocarcinoma (PDAC) [4]. Typically, PDAC is diagnosed at a later stage of the disease, with highly aggressive malignancies. For this reason, patients do not receive adequate surgical treatment or conventional chemotherapy [5]. Therefore, it is necessary to determine the underlying molecular mechanism of tumor growth in PDAC to develop new therapeutic techniques that can prevent tumor progression.
The Notch signaling pathway is an evolutionarily conserved, intercellular signaling pathway involved in the cell fate decision in many biological contexts, and plays a vital role in oncogenesis [6,7]. Accordingly, Notch signaling is considered a molecular target for different malignancies [8,9]. In mouse models, both activation and inhibition of Notch signaling promote PDAC development and progression [10][11][12][13]. There are four types of Notch receptors (Notch 1-4) and five types of canonical Notch ligands: Delta-like 1 (DLL1), DLL3, DLL4, Jagged 1 (JAG1), and JAG2. The interaction of epidermal growth factor (EGF)-like repeats of Notch receptors with their ligands initiates the Notch signaling pathway [14]. Most of the EGF repeats are post-translationally modified by O-glycan, such as O-glucose (O-Glc), O-fucose (O-Fuc), and O-linked N-acetylglucosamine (O-GlcNAc) [7]. Notch signaling is a crucial factor in the progression of PDAC. It is also responsible for the EMT phenotype and cancer stem cell formation in PDAC. It has also been noted that Notch signaling components and their target genes are unregulated in invasive PDAC [15][16][17][18][19][20]. Nonetheless, dysregulated Notch signaling by the post-translational modification of Notch receptors remains poorly understood, with little research focusing on the glycosyltransferases modulating Notch activity.
EGF domain-specific O-GlcNAc-transferase (EOGT) is an endoplasmic reticulum (ER)-specific enzyme, which adds an O-GlcNAc moiety to the extracellular domain of secreted or membrane proteins, including Notch receptors and their ligands [21,22]. Evidence suggests that the mutation of EOGT causes a rare congenital disease, Adams-Oliver syndrome [23][24][25]. EOGT is highly expressed in endothelial cells and regulates optimal vascular integrity and development by enhancing DLL ligand-mediated Notch signaling [24,26]. However, there is no evidence showing that EOGT-mediated O-GlcNAcylation participates in the development and progression of cancer, including PDAC.
In addition to EOGT, Fringe genes are involved in the regulation of ligand-dependent Notch signaling pathway. Fringe encodes β-1,3-GlcNAc-transferase modifying O-Fuc on Notch receptors. In mammals, Fringe activity is mediated by three different genes (LFNG, RFNG, and MFNG), of which LFNG has significantly higher catalytic activity than others [27]. Unlike EOGT, LFNG enhances Delta-ligand-mediated Notch signaling but inhibits JAG1-mediated Notch signaling [28]. Though some evidence has shown that loss of Lfng in mice accelerated Kras-induced pancreatic cancer development, the roles of LFNG in human cancer cells are cell type-dependent and not fully understood [29].
To examine the contribution of EOGT and LFNG in PDAC, we conducted database analysis and functional studies in a PDAC cell line. Our study indicated critical roles for EOGTand LFNG-dependent Notch signaling in PDAC. Furthermore, we showed-for the first time-that EOGT impacts cell proliferation and migration in cancer cells.

Contribution of EOGT and LFNG to Notch Signaling in PDAC
Notch target gene expression is dysregulated in PDAC [15,16,19,20]. To date, the expression levels of EOGT and FNG genes have been experimentally shown to affect both O-glycan structures and functions in mammals [26,28,[30][31][32][33]. To investigate the expression of EOGT and FNG genes in PDAC, in silico analysis was performed using the GEPIA2 integrated database, which includes 179 tumor and 171 normal tissue samples. Among Notch target genes, HES1 and HEY1 expressions were higher in both basal and classical subtypes of PDAC. In contrast, EOGT expression was significantly increased in the basal subtype, which represents a more aggressive phenotype [34]. LFNG expression was not significantly altered ( Figure 1A). However, these data did not exclude the possibility of multiple, rather than a single, glycosyltransferase(s) contributing to dysregulated Notch signaling in PDAC.
Previous studies demonstrated that the loss of EOGT decreased DLL ligand-induced Notch signaling and downstream Notch target genes [26]. Thus, we analyzed the correlation between EOGT and the upregulated Notch target genes ( Figure 1B). Spearman correlation coefficient was used to assess monotonic correlations. Correlation analysis revealed a significant positive correlation between EOGT and HEY1 (r = 0.47, p = 4.2 × 10 −11 ). In contrast, the correlation analysis between LFNG and NOTCH target genes revealed a positive correlation between LFNG and HES1 expression (r = 0.48, p = 8.4 × 10 −12 ). A negligible correlation was observed for MFNG and RFNG [28]. These data suggested that the contribution of EOGT and LFNG to Notch signaling is qualitatively different, possibly through the differential impact on multiple Notch-ligand pairs, which leads to the trans-activation of distinct sets of Notch signaling target genes, including HES1 and HEY1 [35]. Further in-depth bioinformatics analyses included Biclustering methods, will help elucidate the pathological relevance of the observed correlations in tumor progression [36][37][38][39].

Expression of EOGT in PDAC Cell Lines
We compared endogenous expression levels of EOGT in human PDAC cell lines to a normal pancreatic ductal cell line, H6C7. From immunoblotting data, we found that four PDAC cell lines (Panc-1, BxPC3, Panc03.27, and CAPAN-2) out of ten showed higher EOGT expression compared to H6C7 cells ( Figure 2A). We also verified the expression of EOGT by immunostaining in the four PDAC cell lines ( Figure 2B). Based on the immunoblotting and immunostaining assays, we selected Panc-1, which showed prominent EOGT expression, for functional analysis.

CRISPR/CAS9-Mediated Lentiviral Knockout of EOGT in a PDAC Cell Line
To assess the role of EOGT in the growth of the PDAC cell line, we performed gene editing with the CRISPR/Cas9-mediated lentivirus technique in Panc-1 cells. After lentiviral transduction and blasticidin S selection, the lack of EOGT was confirmed by the anti-EOGT (AER61) antibody (Supplementary Figure S1).

EOGT Knockout in Panc-1 Cells Impairs Cell Proliferation and Migration
To investigate whether the decreased Notch activity inhibits Panc-1 cell proliferation, Panc-1 cells were treated with DAPT, a γ-secretase inhibitor that blocks Notch signaling. As shown in Supplementary Figure S2, the inhibitory effect of DAPT on the proliferation of Panc-1 cells was observed in a time-dependent manner. This result suggested that decreased Notch activity can substantially reduce the cell proliferation of Panc-1. These data agree with previous reports showing a similar effect on cell proliferation by DAPT treatment [40,41].
To examine the effect of knockout of EOGT in the Panc-1 cells, we observed the proliferation between wild-type and EOGT-KO cells at an interval of 6 h over 66 h. EOGT-KO cells exhibited slow growth relative to wild-type parental control cells. This result showed that EOGT promotes cell proliferation in Panc-1 cells ( Figure 2C).
It has been reported that inhibition of Notch1 affects cancer cell migration and invasion in vitro in pancreatic cancer [42,43]. We analyzed the cell motility of the parental wildtype and EOGT-KO Panc-1 by wound-healing migration assay. Analysis of the relative wound-healing density revealed that EOGT-KO Panc-1 showed significant decreases in re-capturing the wound portion compared to wild-type control cells ( Figure 2D). Compared to the control cells, EOGT-KO-1, KO-2, KO-4, KO-10 clones showed a slower migration rate, presenting a depletion of wound density 13, 4, 18, and 20%, respectively, within 18 h. The differences in the migration rate among KO clones appear to be due to cancer cells' clonal variation. According to previous studies, the doubling time of Panc-1 is 52-56 h. Therefore, the effects on the wound-healing density were predominantly due to decreased cell migration of EOGT-KO Panc-1 cells, but not to decreased proliferation [44,45].

CRISPR/Cas9-Mediated Lentiviral Knockout of LFNG in a PDAC Cell Line
To compare the effect of EOGT and LFNG in the growth and motility of Panc-1 cells, we generated LFNG-KO cells by CRISPR/Cas9-mediated lentivirus transduction. The cutting efficacy was analyzed by T7 endonuclease assay. Then, by limiting dilution, single-cell clones were isolated. GFP-positive clones were selected for sequencing to confirm LFNG knockout in Panc-1 cells (Supplementary Figure S3).

LFNG Knockout in Panc-1 Cells Impairs Cell Proliferation and Migration
To evaluate the role of LFNG in cell proliferation, Panc-1 cells lacking LFNG were analyzed by the IncuCyte ZOOM system (Sartorius, Japan). The result showed that the cell proliferation of LFNG-KO cells was lower than that of parental control cells ( Figure 3A). These data suggested that LFNG promotes cell proliferation of Panc-1 cells. According to previously reported data, a similar experiment was performed by shRNA-mediated knockdown of LFNG [29]. Although this group reported the tumor-suppressive behavior of LFNG in mice and pancreatic cancer cells, our findings suggested that LFNG promotes the proliferation activity of Panc-1 cells, which was also noted in the same study [29]. Then, we used the wound-healing assay to compare the motility of wild-type and LFNG-KO Panc-1 cells. The relative wound density revealed that the cell migration rate in controls was 10-15% faster than LFNG-KO Panc-1 cells ( Figure 3B). These results suggest that LFNG promotes cell migration in a similar way to EOGT.

Low Expression of Both EOGT and LFNG Predicts Better Overall Survival in PDAC Patients
Knowing that EOGT and LFNG modulate cell proliferation and migration, the Cancer Genome Atlas (TCGA) dataset was analyzed to evaluate the clinical outcome. For overall survival, PDAC patients were classified into high and low expression groups using the median expression of genes as the cut-off value. Kaplan-Meier survival curves revealed no association between EOGT expression and overall survival in PDAC ( Figure 4A). Additionally, no association was observed between LFNG expression and overall survival ( Figure 4B). In contrast, lower expression of both EOGT and LFNG is associated with a better prognosis of PDAC (p = 0.0195). These data suggest that the dysregulation of Notch signaling mediated by high expression of either EOGT or LFNG leads to poor prognosis of PDAC. In summary, we showed the roles of EOGT in cell proliferation and migration in cancer cells. In our study, loss of EOGT or LFNG resulted in a similar cellular phenotype, suggesting that both genes cooperate in the tumor properties of PDAC cells. Furthermore, the combination of lower LFNG and EOGT expression in PDAC serves as an excellent prognostic marker, implying that both Notch modifiers modulate disease progression. Given that EOGT and LFNG modulate ligand-induced Notch signaling, these results suggest that changes in the expression of these glycosyltransferases result in altered Notch activity, as indicated by the correlation between LFNG and HES1, or EOGT and HEY1 expression in the PDAC database. Therefore, dysregulated Notch signaling mediated by changes in the expression of EOGT and LFNG would affect the prognosis of PDAC. Although this study implicates potential roles for EOGTand LFNG-dependent Notch signaling in PDAC, further studies will be necessary to clarify the roles of Notch-modifying enzymes in dysregulated Notch activity and the development and progression of PDAC at the molecular level.

Cell growth inhibition by DAPT
Panc-1 cells (3 × 10 4 ) were seeded in a 24-well plate with 500 µl of DMEM/10% FBS. After 24 h incubation, cells were treated with 5 µM, 50 µM, and 75 µM DAPT over 69 h. Control Panc-1 cells were treated with 0.1% DMSO in the culture medium. Cell growth inhibition was measured using IncuCyte ZOOM Live-Cell Analysis System (Sartorius) [46]. Data were analyzed using phase object confluence, which quantified the cell surface area as confluence values. The images of the cells were taken every 3 h with a 4× objective.

Lentivirus Vector Construction
For the EOGT sgRNA lentivirus vector construction, EOGT sgRNA was designed using online software CRISPRdirect [47]. The sgRNA was synthesized as a g-block segment, which includes XhoI restriction site, U6 promoter region, sgRNA target site, chimeric sgRNA scaffold, and NheI restriction site, as shown in Supplementary Table S1. The gblock was cloned into a pLX-sgRNA lentivirus vector containing blasticidin S-resistance gene [48]. For selecting LFNG sgRNA, we used previously validated sgRNA for LFNG (Supplementary Table S1) [49]. The sgRNA sequence was inserted into the pLRG2.1 vector, which contains GFP as a selection marker [50]. A separate doxycycline-inducible FLAG-Cas9 expression vector (pCW-Cas9) was also used, which contains a puromycin resistance gene [48].

CRISPR/Cas9-Mediated Lentiviral Knockout of EOGT and LFNG Genes
The knockout experiments in Panc-1 cells were performed as described before with a slight modification [50,51]. Briefly, lentiviral plasmids are co-transfected with LP1, LP2, and VSVG virus-packaging plasmid in LentiX-293T cells using PEI MAX (Polysciences, USA). After 12h later, culture media were replaced with a high FBS containing medium (DMEM/40% FBS), which did not contain penicillin and streptomycin. After 24 and 48 h later, lentivirus supernatant was collected, centrifuged at 12,000 rpm for 5 min, and filtered using a 0.45 µm filter. After filtration, the virus was concentrated by LentiX concentrator according to the manufacturer protocol. Cas9-transfected Panc-1 cells were generated after selection with 2 µg/mL puromycin. Anti-FLAG antibody was used to verify Cas9 protein expression by immunoblotting. For generating EOGT-KO Panc-1 cells, the Cas9 stable Panc-1 cells were infected with lentiviral supernatant, which contains EOGT sgRNA. After 24 h, cells were selected with 20 µg/mL blasticidin S for 14 days, and treated with 1 µg/mL doxycycline for 1 day to induce Cas9 expression. For preparing LFNG-KO Panc-1 cells, a lentiviral solution was used to infect the Cas9 stable Panc-1 cells for 12 h. After 12 h, cells were treated with 1 µg/mL doxycycline for 1 day to induce Cas9 expression. The transduction efficiency was confirmed by monitoring GFP expression. Both EOGTand LFNG-KO cells were cultured in DMEM/10% FBS medium containing penicillin and streptomycin. A single cell was isolated by limiting dilution and seeding into 96-well plates for clonal selection and further analysis.

Evaluation of CRISPR/Cas9-mediated Genome Editing by T7 Endonuclease Assay and Sequencing Analysis
Genomic DNA was extracted from wild-type Panc-1, clonally selected EOGT-KO Panc-1, and LFNG-KO Panc-1 cells. Briefly, all cells were cultured in 12-well plates. After harvesting, cells were centrifuged and washed with phosphate-buffered saline (PBS). Cells were heated at 95 • C for 20 min after adding 18 µL of 50 µM NaOH for each well of 12-well plates. Then, samples were vortex briefly, added with 3 µL of 1 M Tris-HCl, pH 7.6, and centrifuged at 15,000 rpm for 6 min after vortex to collect the supernatant. The target region was amplified with sgRNA-specific primers according to the manufacturer protocol. To detect double-stranded DNA mismatches and mutations, we used T7 endonuclease according to manufacture protocol. After confirming the knockout cells, single cells were isolated by limiting dilution. For EOGT-KO, of the 22 single-cell clones picked out, 10 clones showed negative expression of EOGT in Western blotting screening (Supplementary Figure S1B). We selected four clones for further analysis. To select LFNG-KO clones, the target regions of both wild-type and knockout cells were sequenced by Sanger sequencing. Four clones were selected for LFNG-KO Panc-1 cells after comparing with the wild-type sequence (Supplementary Figure S3).

Immunostaining
Panc-1 cells were fixed with 4% paraformaldehyde in PBS for 15 min and ice-cold methanol for 5 min. Then cells were permeabilized with 0.5% Triton-X 100 in PBS for 15 min and blocked with 5% BSA in PBS. Cells were incubated with Anti-AER61 (1:200) antibody diluted with 5% BSA/PBS overnight at 4 • C. After washing with PBS, cells were incubated with DyLight 549-conjugated goat anti-rabbit IgG antibody (1:1000) diluted with 5% BSA/PBS for 1 h at room temperature. Then, cells were counter-stained with DAPI (Southern Biotech, Birmingham, AL, USA). All the slides were examined by TiEA1R Confocal Microscopy with NIS Elements (Nikon, Japan) [52].

Western Blotting
Cells were lysed in a lysis buffer containing Tris-buffered saline (TBS), pH 7.6, 1 mM CaCl 2 , and 1% Nonidet P-40. Each sample was separated by 7.5% SDS-PAGE for Western blotting and transferred onto a PVDF membrane (Merck, Japan). The membrane was blocked in 3% BSA/PBST at room temperature for 1 h, followed by incubation with anti-AER61 antibody (1:10,000) diluted in 3% BSA/PBST for 2h. After washing with PBST three times, the membrane was incubated with Anti Rabbit HRP conjugated secondary antibody (1:8000) in 3% BSA/PBST at room temperature for 1 h. After washing with PBST, bands on the membrane were visualized with Immobilon Western chemiluminescent HRP substrate (Merck) and iBright FL1500 Imaging System (Thermo Fisher).

Cell Proliferation Assay
Cells were seeded 96-well plates at 7 × 10 3 cells per well in DMEM medium containing 10% FBS. The increasing confluence of the cells was measured as an indicator of cell proliferation using IncuCyte ZOOM Live-Cell Analysis System. Images were taken at 6 h intervals with a 4× objective from 3 separate wells, and mean ± SD of confluence percentage was measured.

Cell Migration Assay
Cell migration (wound healing) assay was performed as reported previously [53]. Briefly, cells (2 × 10 4 ) were seeded into an uncoated 96-well plate image lock plate (Sartorius). After 6-8 h later, the wound was made using 96-pin WoundMaker (Sartorius). The images of the wound were taken every 2 h for 18 h. Then, data were analyzed by calculating the relative wound density with IncuCyte ZOOM Live-Cell Analysis System.

Statistical Analysis and Database Analysis
The significance of the data in functional assays was analyzed with one-way repeated measures ANOVA followed by Dunnett's Post-hoc test to show a significant difference between groups. Significance was evaluated based on the p-value. Gene expression and correlation were analyzed using the GEPIA2 [54] database for pancreatic adenocarcinoma (PAAD), in which the pre-processed RNA-seq datasets and a normalization method that calculates transcripts per million (TPM) were utilized. Survival analyses were analyzed with a TCGA dataset and EZR software (version 1.52), which is a graphical user interface for R software (version 4.0.2).