Truncated O‐glycans promote epithelial‐to‐mesenchymal transition and stemness properties of pancreatic cancer cells

Abstract Aberrant expression of Sialyl‐Tn (STn) antigen correlates with poor prognosis and reduced patient survival. We demonstrated that expression of Tn and STn in pancreatic ductal adenocarcinoma (PDAC) is due to hypermethylation of Core 1 synthase specific molecular chaperone (COSMC) and enhanced the malignant properties of PDAC cells with an unknown mechanism. To explore the mechanism, we have genetically deleted COSMC in PDAC cells to express truncated O‐glycans (SimpleCells, SC) which enhanced cell migration and invasion. Since epithelial‐to‐mesenchymal transition (EMT) play a vital role in metastasis, we have analysed the induction of EMT in SC cells. Expressions of the mesenchymal markers were significantly high in SC cells as compared to WT cells. Equally, we found reduced expressions of the epithelial markers in SC cells. Re‐expression of COSMC in SC cells reversed the induction of EMT. In addition to this, we also observed an increased cancer stem cell population in SC cells. Furthermore, orthotopic implantation of T3M4 SC cells into athymic nude mice resulted in significantly larger tumours and reduced animal survival. Altogether, these results suggest that aberrant expression of truncated O‐glycans in PDAC cells enhances the tumour aggressiveness through the induction of EMT and stemness properties.

of the immature truncated O-glycans (Tn and STn antigens) in ~40% of the PDAC tumours. 7 Further, COSMC is the molecular chaperone for Core 1 synthase (C1GalT1) that stabilizes the enzyme. 8 Loss of COSMC and aberrant expression of truncated O-glycans induce oncogenic features, including compromised adhesion, increased cell migration, reduced apoptosis, loss of tissue architecture, invasive growth and metastasis. 7,9 However, the molecular mechanism whereby these truncated O-glycans enhance the tumorigenic properties of PDAC cells remains elusive.
Studies have demonstrated that epithelial-to-mesenchymal transition (EMT) is a critical phenomenon for tumour metastasis where switching of epithelial phenotype to mesenchymal phenotype facilitates the tumour cell migration through the matrix to the distant organ sites. This orchestrated biological event is complemented by the down-regulation of the epithelial markers, such as E-cadherin, claudin and occludin, and up-regulation of mesenchymal markers, such as N-cadherin, vimentin, transcription factors Snail and Slug. 10,11 In human PDAC, EMT is closely associated with the invasive tumour, development of metastasis and negatively affect overall patient survival. 12 A recent study has demonstrated that overexpression of N-acetylgalactosaminyl transferase 6 (GALNT6) a glycosyltransferase involved in O-glycosylation potentially disrupts cell morphogenesis and induced cellular changes similar to EMT in mammary epithelial cells. 13 Further, hyperglycemia, especially Oglycosylation induced EMT in lung adenocarcinoma cells to promote tumour metastasis has also been reported. 14 In the light of these reports, we hypothesized the possible role of EMT as the underlying mechanism for the enhanced tumorigenic

| Human Rapid Autopsy PDAC samples
De-identified matched sets of human primary pancreatic tumour (n = 6) and metastatic lesions (n = 6), and unaffected normal pancreas tissues (n = 6) were procured from the Rapid Autopsy Pancreas Program tissue bank at the University of Nebraska Medical Center (UNMC).

| PCR analysis of COSMC/C1GALT1C1 and Core 1 synthase/C1GALT1
Total RNA was isolated from T3M4 and Capan-2 (WT, SC and SC-R) cells, and cDNA was synthesized using Verso cDNA synthesis kit (Thermo Fisher Scientific) according to the manufacturer's instruction. C1GALT1C1 (COSMC), C1GALT1 (Core-1 synthase) and GAPDH genes were amplified using cDNA-specific primers (Table   S1). PCR was carried out at 95°C for 3 minutes followed by 30 cycles of 95°C for 30 seconds, 55°C for 30 seconds and 72°C for 1 minutes followed by a single incubation at 72°C for 5 minutes. PCR products were resolved by electrophoresis on a 1.2% agarose gel.

| In vitro invasion and migration assays
Invasion and migration analyses were performed as described previously. 7 Briefly, T3M4 WT and SC cells (5 × 10 4 ) were seeded on top portion of the polyethylene terephthalate inserts and matrigelcoated Boyden chambers (BD Biosciences). The cells that migrated or invaded through the membrane barrier were fixed, stained and counted in five different fields using a light microscope (20×). The number of invaded or migrated cells was expressed as the mean number of cells that transited to the lower part of the membrane.
These experiments were performed in triplicates.

| Antibody labelling and Fluorescence-activated cell sorting analysis
For side cell population analysis, T3M4 WT and SC cells (1 × 10 6 cells/mL) were re-suspended in DMEM supplemented with 2% foetal bovine serum (FBS) at 37°C. Then, the cells were labelled with Hoechst 33342 (Thermo Fisher Scientific) at a concentration of 5 µg/mL for 2h at 37°C in dark with gentle agitation with or without verapamil (50 µg/mL). The cells were centrifuged at 500 g for 5 minutes at 4°C; re-suspended in ice-cold PBS containing 2% FBS. Cells were counterstained with 5 µg/mL propidium iodide (Sigma-Aldrich), and cell sorting was performed using a

| Orthotopic pancreas tumour model and animal survival
T3M4 WT and SC cells were orthotopically implanted into the mice pancreas as described earlier. 16 Briefly, cells (0.25 × 10 6 /30 µL PBS) were orthotopically implanted into the pancreas of athymic nu/nu mice (Crl:NU-Foxn1 nu ) (n = 13/group). After 28 days of implantation, the animals were killed and the tumour weight, volume and incidence of metastases were determined. For animal survival analyses, the same experiment was performed in athymic nude mice (n = 15/ group). Animal survival was monitored on a daily basis, or animals were killed at a pre-determined end-point, if the tumour has been grown more than 2 cm in diameter. All the animals were housed under standard housing conditions at the University of Nebraska Medical Center animal core facilities. Animal procedures included in this study were reviewed and approved by the UNMC institutional animal care and use committees (IACUC).

| Western blot analysis
Cell lysates were prepared from T3M4 (WT, SC and SC-R) and Capan 2 (WT and SC) cells. For mouse tissue sample, the tissue homogenate was prepared in homogenizing buffer. 30 μg of proteins were resolved in a gradient (4%-20%) denaturing polyacrylamide gel (Bio-Rad) and transferred to polyvinylidene difluoride (PVDF) membranes (Millipore). After blocking with 5% BSA, the membranes were incubated with the respective primary antibodies (Table S2).
After incubation with HRP-conjugated secondary antibodies, the antigen-antibody complex was developed using Bio-Rad enhanced chemiluminescence (ECL) Prime Western Blotting detection reagent (General Electric Healthcare Life Sciences).

| Immunohistochemistry
For the analysis of STn antigen expression in RAP samples, the paraffin-embedded tissue sections were deparaffinized with xylene, hydrated with series of ethanol and quenched with hydrogen peroxide. Antigen retrieval was performed with citrate buffer (pH 6.0); blocked with universal blocker (Thermo fisher Scientific) and incubated with TKH2 monoclonal antibody (a kind gift from Dr Ulla Mandel, University of Copenhagen, Denmark) for 2 hours at room temperature. For the analysis of E-cadherin, N-cadherin and CD 133 in mouse tissue sections, the paraffin-embedded slides were processed as describes above and incubated with rabbit anti-N-cadherin (ab18203), rabbit anti-E-cadherin (ab15148) and for 30 minutes at room temperature and developed using DAB according to the kit instructions. All the slides were dehydrated with ethanol series, and after xylene washes, the slides were mounted with coverslip. The protein expression was analysed by a pathologist. The histological scoring was performed based on stain proportion (0%-100%) and intensity (0-negligible, 1-low, 2moderate and 3-high). The histoscore was generated by multiplying the stain proportion score (1 = <5%, 2 = 5%-25%, 3 = 26%-50%, 4 = 51%-75%, 5 = >75%) with the intensity score (0-3) to obtain values between 0 and 15.
The tumour sphere numbers were counted under phase-contrast microscope at 10× magnification.

| Statistical analysis
In vitro cell migration and invasion was analysed between the groups by unpaired t test. Protein expressions by Western blotting were quantified using ImageJ software, and the statistical significance between groups was analysed by two-way ANOVA. The significance of tumour weight, volume and metastasis between the groups was calculated by two-tailed Student's t test and two-tailed Fisher's exact test. Animal survival rates were graphed using Kaplan-Meier method, and the statistical significance between the groups was calculated by log-rank statistical analysis. P value < .05 was considered as statistically significant.

| Expression of Sialyl-Tn antigen in primary and metastatic PDAC tumour tissues
We have assessed the clinical relevance of Sialyl-Tn (STn) antigen in PDAC by analysing its expression on tissues of normal pancreas (n = 6), primary pancreatic tumour (n = 6), liver (n = 6) and lung (n = 6) metastases (obtained from PDAC patients who underwent rapid autopsies) by immunohistochemistry (IHC). Consistent with available reports, we found no expression of STn antigen in normal pancreas.
However, aberrant expression of STn antigen was detected in primary tumour. Interestingly, STn antigen is more highly expressed in liver as well as lung metastases ( Figure 1A). Figure 1B shows the graphic representation of IHC intensity score for the relative expression of STn antigen in each autopsy patient's primary tumour, liver mets and lung mets samples.

| Genetic deletion of COSMC enhances tumorigenic features in PDAC cells
We previously demonstrated that reduced expression of C1GalT1 and/or hypermethylation of its molecular chaperone COSMC are the major causative factors for the truncation of mucin-type O-glycosylation in clinical specimens of PDAC. 7 In continuation with the previous work, to explore the biological role of truncated O-glycans, we have utilized COSMC knockout PDAC cells (T3M4 and Capan-2).

Disruption of COSMC and loss of its expression in T3M4 and
Capan-2 were observed by PCR ( Figure 1C), while the mRNA expression of C1GALT1 (core-1 synthase) remained same in both T3M4 and Capan-2 WT and SC cells ( Figure 1C) which is in accordance with the previous reports. 7,9 We further examined the COSMC deletion induced degradation of C1GalT1 protein in PDAC SC cells ( Figure 1D).
To confirm these results, we re-expressed the COSMC in T3M4 SC cells and confirmed at the mRNA levels by PCR ( Figure 1E

| COSMC deletion promotes PDAC tumorigenicity through the induction of epithelial-tomesenchymal transition
The  Figure 2C and Figure S1A). In addition to this, we also observed a decreased expression of occludin in T3M4 SC cells (P = 0.0067) as compared to T3M4 WT cells ( Figure 2C and Figure S1A). Interestingly, either low or no expression of occludin was seen in Capan-2 SC cells as compared to Capan-2 WT cells ( Figure 2C and Figure S1A). Previous studies have reported that matrix degrading enzymes cleave the stromal basement epithelium of extracellular matrix (ECM) and therefore facilitates the invasion of transmitted cells. 11 We, therefore,  Figure 2D and Figure   S1B). Next, we analysed the expression of alpha smooth muscle actin (α-SMA), a commonly used marker for the cells in the transition phase. 19 T3M4 SC and Capan-2 SC cells showed an increased expression of α-SMA (P < 0.0004 and P = 0.0007, respectively) as compared to their WT cells ( Figure 2D and Figure S1B). In order to further explore whether mesenchymal transition is ensued, we have analysed the expression of major mesenchymal markers such as N-cadherin, vimentin and the transcription factors Snail and Slug in PDAC cells. We detected a significant fold increase in the expression of mesenchymal markers N-cadherin and vimentin in T3M4 SC (P = .0005 and P = 0.0161, respectively) and Capan-2 SC (P = 0.0005 and P = 0.0346, respectively) cells as compared to WT cells ( Figure 2E and Figure S1C). Also, in the protein expres-

| Truncation of O-glycans enhance the stemness feature of PDAC cells
It is well known that PDAC is composed of both undifferentiated or poorly differentiated cancer stem cells and more differentiated malignant cells that are derived from cancer stem cells. 20 Since we

| Truncated O-glycans modulate PDAC cells in vivo tumour growth and animal survival
We  Figure 5A). In contrast to this, we detected an increased expression of N-cadherin in SC cells implanted mice tumour tissues, whereas its expression was little in WT cells implanted mice tumour tissues (P < 0.0001) ( Figure 5B). Also, we validated these results by Western blotting; expectedly, a more prominent cadherin switching with decreased expression of E-cadherin and increased expression of Ncadherin was observed in SC cells implanted tumours ( Figure 5C).
In addition to this, we have analysed the expression of CD133 in mice tumour tissues. More interestingly, the protein expression of CD133 was significantly high in SC cells implanted mice tumours as compared to WT cells implanted tumours (P = 0.0169) ( Figure 5D).
Taken together, these results suggest that aberrant expression of It is well understood that mucins play a protective role on epithelial surfaces and exert significant control on cellular signalling.  50 Also, very recently, co-expression of CD133 with STn antigen in a subset of human ovarian cancer cells has been reported. 51 Interestingly, a recent study has demonstrated that CD44 is one of the major carriers of truncated O-glycans which is accompanied by increased hyaluronan binding potential that affects matrix shedding. 52 Altogether, these studies strongly support our hypothesis that aberrant glycosylation plays a critical role in maintaining the stemness features of cancer cells that may promote the induction of EMT ( Figure 6). However, further studies are warranted to prudently elucidate the underlying mechanism of aberrant glycosylation in inducing stemness and EMT in cancer cells.

CO N FLI C T O F I NTE R E S T
The authors of this manuscript declare no conflicts of interest.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available from the corresponding author upon reasonable request.