Clinicopathological significance of core 3 O-glycan synthetic enzyme, β1,3-N-acetylglucosaminyltransferase 6 in pancreatic ductal adenocarcinoma

Mucin-type O-glycans are involved in cancer initiation and progression, although details of their biological and clinicopathological roles remain unclear. The aim of this study was to investigate the clinicopathological significance of β1,3-N-acetylglucosaminyltransferase 6 (β3Gn-T6), an essential enzyme for the synthesis of core 3 O-glycan and several other O-glycans in pancreatic ductal adenocarcinoma (PDAC). We performed immunohistochemical and lectin-histochemical analyses to detect the expression of β3Gn-T6 and several O-glycans in 156 cases of PDAC with pancreatic intraepithelial neoplasias (PanINs) and corresponding normal tissue samples. The T antigen, Tn antigen, sialyl Lewis X (sLeX) antigen, and sLeX on core 2 O-glycan were more highly expressed in PDAC cells than in normal pancreatic duct epithelial cells (NPDEs). Conversely, the expression of 6-sulfo N-acetyllactosamine on extended core 1 O-glycan was found in NPDEs and was low in PDAC cells. These glycan expression levels were not associated with patient outcomes. β3Gn-T6 was expressed in ~20% of PDAC cases and 30–40% of PanINs but not in NPDEs. Higher expression of β3Gn-T6 was found in PDAC cells in more differentiated adenocarcinoma cases showing significantly longer disease-free survival in both univariate and multivariate analyses. In addition, the expression of β3Gn-T6 in PDAC cells and PanINs significantly correlated with the expression of MUC5AC in these cells, suggesting that β3Gn-T6 expression is related to cellular differentiation status of the gastric foveolar phenotype. Thus, it is likely that β3Gn-T6 expression in PDAC cells is a favorable prognostic factor in PDAC patients, and that the expression of β3Gn-T6 correlates with the gastric foveolar phenotype in pancreatic carcinogenesis.

Introduction Mucin-type O-glycans play roles in various biological functions, including lymphocyte homing and gastric mucosal defense against Helicobacter pylori [1][2][3]. Cancer cells express unique and characteristic glycan structures [4], some of which are involved in cancer initiation, progression, and metastasis, mainly through cellular recognition and/or cell adhesion [5]. Although these unique characteristics have the potential to be used for diagnostic and therapeutic research and development, limited information is currently available regarding the biological roles and clinicopathological significance of O-glycans in cancer.
Glandular epithelial cells produce mucin consisting of core proteins and abundant O-linked glycans [6]. The synthesis of O-glycans is initiated by the addition of N-acetylgalactosamine to Ser/Thr to form the Tn antigen (Fig 1). Based on the Tn antigen, core 1 (T antigen) or core 3 structures are formed, which are then branched to give rise to core 2 or core 4 structures, sequentially. These core structures can be further extended thus resulting in complex glycans, such as several blood type antigens (Fig 1). It has been reported that both core and peripheral modified glycans are expressed specifically in some types of cancer and are related to biological characteristics of the cancer cells, thereby representing tumor markers and prognostic markers [7][8][9][10].
Pancreatic ductal adenocarcinoma (PDAC) is a highly malignant disease [11]. Despite advances in diagnosis and treatments, the 5-year survival rate is less than 10% in PDAC [12]. To improve the patient outcomes, we have to understand PDAC more deeply. PDAC is known to express the Tn antigen and its sialylated derivative, the STn antigen [13]. These truncated O-glycans in PDAC cells are associated with aggressive characteristics [14,15]. Sialyl Lewis A (sLeA), alternatively called CA19-9, and sialyl Lewis X (sLeX) are reported to be unfavorable prognostic factors in PDAC [15,16]. Although this has been previously demonstrated by studies involving cell lines or animal models, only a few reports have addressed the clinicopathological and biological roles of glycans using human clinical samples. In addition, these two unique glycans, 6-sulfo N-acetyllactosamine on extended core 1 O-glycan detected by antibody MECA-79 [3,17], and sLeX on core 2 O-glycan detected by antibody NCC-ST-439 (ST-439) [18,19] have not yet been evaluated in PDAC. The presence of MECA-79 antigen is used as a marker of high endothelial venules of lymph nodes and the antigen is a part of the glycan structure of an L-selectin ligand [3], and is expressed in pancreatic ductal epithelial cells [20].
The presence of core 3 O-glycan that is induced in cancer cells by β1,3-N-acetylglucosaminyltransferase 6 (β3Gn-T6), which is an essential enzyme for the synthesis of core 3 O-glycan [8], has been reported to reduce malignant characteristics (proliferation, invasion, and metastasis) in colon, prostate, and pancreatic cancers according to in vitro assays with animal model experiments [8,21,22]. However, clinicopathological significance of core 3 O-glycan and β3Gn-T6 has not been evaluated yet.
The aim of this study was to investigate the clinicopathological impact of core 3 O-glycan on PDAC through immunohistochemical detection of β3Gn-T6, rather than measuring the structure of core 3 O-glycan, for which currently there is no specific antibody or lectin. We examined the expression of β3Gn-T6 in 156 consecutive cases of PDAC along with normal pancreatic tissue and the most popular premalignant lesion of pancreatic intraepithelial neoplasia (PanIN) [23], and compared the clinicopathological features. We also examined the clinicopathological impact of several O-glycans in PDAC.

Ethics approval and consent to participate
This study was approved by the Institutional Review Board of the National Cancer Center, Japan (#2005-077). The written informed consent was obtained from all participants involved in the study, and all clinical investigations were conducted in line with the principles of the Declaration of Helsinki.

Study population
Clinical and pathological data and specimens used for this study were obtained through a detailed retrospective review of the medical records of 156 consecutive patients with PDAC who had undergone surgical resection between 2009 and 2011 at the National Cancer Center Hospital, Tokyo. None of the patients had received any therapy before surgery. All patients included in this study underwent macroscopic curative resection, and all cases involved conventional ductal carcinomas. The clinicopathological characteristics of the study participants are summarized in Table 1. The median follow-up period after surgical treatment was 29.1 (1.7-126.5) months. Recurrence was suspected when a new local or distant metastatic lesion was found on serial images, and an increase in tumor marker levels was observed. At the census date (September 2018), we checked whether the patients were dead or alive; 62 patients (39.7%) were alive, 82 (52.6%) had died of pancreatic cancer, and 12 (7.6%) had died of other causes. All M1 (TNM classification [24]) patients showed only nodal metastasis around the abdominal aorta.

Pathological evaluation
All carcinomas were examined pathologically and classified according to the World Health Organization (WHO) classification [11,23], Union for International Cancer Control (UICC) TNM classification [24], and the Classification of Pancreatic Carcinoma of the Japan Pancreas Society [25]. Surgically resected specimens were fixed in 10% formalin and cut into serial 5-mm-thick slices and all sections were stained with hematoxylin and eosin (HE) for pathological examination. Representative tissue blocks were selected for subsequent analyses. We used PanINs and normal pancreatic tissue in this study as follows: the areas containing PanIN were apart from cancer cells during microscopic observation, and normal pancreatic tissues were more than 2 cm away from the tumor cells.

Immunohistochemistry and lectin-histochemistry
Immunohistochemistry was performed on 4-μm-thick formalin-fixed paraffin-embedded tissue sections using the avidin-biotin complex method as described previously [26]. Lectin-histochemical analysis was performed in the same way as the immunohistochemical analysis, except lectin was used instead of the primary antibody. The primary antibodies and lectins used in this study are listed in Table 2. Immunohistochemical analysis without the primary antibody was carried out as a negative control. Positive findings are shown in S1 Fig.

Evaluation of immunohistochemistry and lectin-histochemistry
After the immunohistochemical and lectin-histochemical analyses, the antigen expression levels were assessed via a semiquantitative scoring system that incorporated percentages of stained cells with the categorized staining intensity. The staining intensity was recorded in comparison to internal positive controls as 0, negative; 1+, positive but weaker than an internal positive control; 2+, equal to the internal positive control; and 3+, stronger than the internal positive control. The percentage of stained cells was determined by the comparison of the number of cancer cells with each staining intensity to the total number of cancer cells. The sum of products obtained by multiplying the staining intensity and the percentage of corresponding intensity was defined as an expression score (Exp-score). Two observers, i.e., Japanese certified pathologists (ND and NH), who had no access to the patient data, independently evaluated the Exp-score. For statistical analyses, patients were subdivided into two groups by means of the medians as a cutoff.

Immunofluorescence
Cells were seeded on a chamber slide. The staining procedure was previously described [27]. Immunofluorescence images were obtained using a BZ-X710 all-in-one fluorescence microscope (Keyence, Japan).

Extraction of RNA and quantitative RT-PCR (qRT-PCR)
Total RNA was extracted from pancreatic cancer cells, as described previously [28]. All samples were treated with rDNase during isolation, in accordance with the manufacturer's instructions. qRT-PCR for target genes and non-target housekeeping control genes was performed with a Quantstudio 3 (Thermo scientific) using FastStart Universal Probe Master (ROX) and probes from the Universal Probe Library (Roche Diagnostics Corp., Indianapolis, IN), as described previously [26]. The sequences of the primers and the respective Universal Probe Library probes are given in S1 Table. The CT values were normalized to that of GAPDH, and the ΔΔCT method was utilized to compare the expression levels of the genes.
To analyze the structures of glycans attached to MUC5AC, immune complexes were subjected to western blot analysis. The immune complexes were separated by SDS-PAGE in a 4-12% gradient gel (Invitrogen) and were transferred to a Polyvinylidene fluoride (PVDF) membrane, which was then blocked by incubation with PBS-Tween containing 5% of bovine serum albumin as described elsewhere [30]. After that, the membrane was incubated with a primary antibody and biotin-conjugated secondary antibody or with biotin-conjugated lectins ( Table 2) followed by ABC reagents (Vector laboratories).

Statistical analysis
Comparison analyses were performed using the nonparametric test. Post-operative overall survival (OS) and disease-free survival (DFS) rates were calculated using the Kaplan-Meier method and analyzed by the log-rank test. The factors found to be significant by univariate analysis were subjected to multivariate analysis using the Cox proportional hazards model (backward elimination method). Differences at P<0.05 were considered statistically significant. Statistical analyses were performed using SPSS software version 26 (IBM Corp., Armonk, USA).

Expression of glycans and β3Gn-T6 in PDAC cells, premalignant cells, and normal tissues
We evaluated the expression of the T antigen (staining with Peanut agglutinin, PNA) [ [3,17], sLeX (staining with antibodies CSLEX1 and HECA-452) [33], sLeX on core 2 O-glycan (staining with antibody ST-439) [18,19], and β3Gn-T6 (Fig 1) in PDAC cells, PanINs, and noncancerous tissues, normal pancreatic duct epithelial cells (NPDEs), and other normal tissues. Representative immunohistochemical and lectin-histochemical features are shown in Fig 2 and S1 Fig. PDAC is usually composed of variously differentiated cancer cells, with varied frequency and intensity of glycan expression in PDAC cells in the same case. We first analyzed glycan expression in each component of PDAC (Table 3). Next, to determine a representative value for overall expression of antigens in PDAC cells in each PDAC case, we calculated the Exp-score. All glycan antigens, except MECA-79 antigen, were expressed significantly more highly in PDAC cells than in NPDEs (Fig 3). In contrast, MECA-79 antigen expression in PDAC cells was significantly lower than that in NPDEs (Fig 3). The T antigen: This antigen was found to be expressed in some of the well or moderately differentiated PDAC cells but not in NPDEs. Over 30% of low-grade PanINs expressed the T  (Table 3). T antigen expression was mildly higher in PDAC cells compared to NPDEs (Fig 3).
The Tn antigen: The majority of PDAC cells and PanINs expressed the Tn antigen, but most NPDEs did not (Table 3). Tn antigen expression in PDAC cells was markedly higher, and most of the cases were strongly positive, i.e., the Exp-score was >100 (Fig 3).
MECA-79 antigen: In contrast to NPDEs, which usually express MECA-79 antigen, PDAC cells were found to rarely express it, whereas PanINs did not express it at all (Table 3 and Fig  3).
SLeX: PDAC cells expressed sLeX strongly and at a high frequency, regardless of the antibodies applied. However, PanINs and NPDEs showed different profiles depending on the antibodies used (Table 3 and Fig 3). Antibody ST-439 identified the limited sLeX antigen, sLeX on core 2 O-glycan, so that ST-439 + PDAC cells also stained with antibodies CSLEX1 or HECA-452. Antibodies CSLEX1 and HECA-452 recognize both O-linked and N-linked sLeX. Antibodies HECA-452 and ST-439, but not CSLEX1, can recognize sulfated sLeX [2,18], These features can be summarized: (1) both staining frequency and area were ranked as follows, in ascending order: HECA-452, CSLEX1, and ST-439; (2) potentially sulfated sLeX was found in MECA-79 + NPDEs, where HECA-452 staining was sometimes present while CSLEX1 staining was not; (3) low-grade PanINs were positive for HECA-452 staining and CSLEX1 staining, and normal epithelial cells were positive for HECA-452 staining but almost negative for ST-439 staining. These results suggest that PanINs and epithelial cells did not express core 2 O-glycan.

Relation between glycan expression and clinicopathological variables
The correlation between glycan expression in PDAC cells and various clinicopathological factors was examined next. Significant correlations were found only between higher β3Gn-T6 expression in PDAC cells and a lower histological grade, between higher MECA-79 antigen expression and a higher histological grade, and between higher sLeX expression (staining with antibody HECA-452) and a lower histological grade (Table 1 and S2 Fig).
Next, we evaluated correlations among expression levels of different glycans. We compared the Exp-scores of all glycans by Spearman's test (Table 4). Scores on sLeX detected by different antibodies correlated positively. In addition, a few significant correlations were found between β3Gn-T6 and MUC5AC (ρ = 0.49) and between β3Gn-T6 and T antigen (ρ = 0.16), and a negative correlation was found between β3Gn-T6 and MECA-79 antigen (ρ = −0.17).

MUC5AC carries core 3 O-glycan generated by B3GNT6 gene expression in pancreatic cancer cells
To investigate whether MUC5AC contains core 3 O-glycan, we examined glycosylation status of MUC5AC. To select suitable pancreatic cancer cells for the assay, we analyzed the

PLOS ONE
expression of genes B3GNT6 (encoding β3Gn-T6) and MUC5AC, together with genes encoding core 2 synthases (GCNT1, GCNT3, and GCNT4) by qRT-PCR. We chose Capan-1 cells because they show almost no expression of B3GNT6 and higher expression of MUC5AC compared to the other cell lines (Fig 5A). The immunofluorescence assay revealed that GFP-positive stably B3GNT6-transduced cells, Capan1-B3GnT6 expressed the β3Gn-T6 protein ( Fig  5B). The N-acetylglucosaminyl terminus of core 3 O-glycan can be detected by a lectin called GS-II [21]. MUC5AC that was immunoprecipitated from the lysates of Capan1-B3GnT6 and Capan1-mock cells was subjected to SDS-PAGE followed by western blotting with GS-II ( Fig  5C). An intense band was produced by MUC5AC isolated from Capan1-B3GnT6 cells. In contrast, no band was yielded by the MUC5AC isolated from Capan1-mock cells, even though anti-MUC5AC stained bands were comparable between them. These results indicated that MUC5AC from Capan1-B3GnT6 had core 3 O-glycan. In support of this finding, PNA binding to MUC5AC was lower in Capan1-B3GnT6 cells compared with Capan1-mock cells ( Fig  5C). It was confirmed that core 3 O-glycan was present on MUC5AC isolated from Capan1-B3GnT6 cells, and that this glycan was synthesized by β3Gn-T6.

Prognostic significance of glycan-related antigens in PDAC cells
Kaplan-Meier survival analyses revealed a statistically significant association between higher expression of β3Gn-T6 in PDAC cells and longer DFS (Fig 6). Patients with higher sLeX (staining with CSLEX1) expression tended to have shorter DFS. No significant association was found between any other glycan expression and patient outcomes (DFS or OS). No significant association was found between OS and β3Gn-T6 expression (S3 Fig).
Cox proportional analysis of the groups categorized by each glycan expression and β3Gn-T6 expression in PDAC cells as well as conventional clinicopathological variables are shown in Table 5. Data on variables found to be significant by univariate analysis were subjected to multivariate analysis. In the latter, several variables (age, lymph node metastasis, nerve plexus invasion, tumor histological grade, and β3Gn-T6 expression) were found to be significantly associated with DFS.

Discussion
Mucin-type O-glycan is known to be involved in tumor development and malignant characteristics. However, its clinicopathological significance has not yet been sufficiently elucidated. Here, we investigated the clinicopathological significance of both O-glycan cores and Table 4

β3Gn-T6
T peripheral modified glycans in PDAC. Higher β3Gn-T6 expression was noted in more differentiated adenocarcinoma in PDAC patients. These PDAC cases showed significantly longer DFS. Together with previous reports indicating that forced expression of β3Gn-T6 reduces the aggressiveness of cancers in vitro and in vivo [8,21,22], our findings suggest that β3Gn-T6 expression in PDAC cells is a favorable prognostic indicator. In addition, the expression of β3Gn-T6 in PDAC cells and PanINs significantly correlated with the expression of MUC5AC in these cells, implying that β3Gn-T6 expression is related to cellular differentiation status of the gastric foveolar phenotype. The expression of the T antigen, Tn antigen, sLeX antigen, and sLeX on core 2 O-glycan was higher in PDAC cells. Unexpectedly, we did not find any significant association with patient outcome in our cohort. However, 6-sulfo N-acetyllactosamine on extended core 1 O-glycan (MECA-79 antigen) was underexpressed in PDAC cells compared to NPDEs and was not associated with patient outcome. Core 3 O-glycan is widely distributed throughout the gastrointestinal tract [35] and is synthesized only by β3Gn-T6 expressed normally in gastric foveolar epithelial cells and colonic goblet cells [8]. β3Gn-T6 is not expressed in normal pancreatic tissue, and the induction of β3Gn-T6 has been previously found in some low-grade PanINs and differentiated PDAC cells (Table 3). Takano et al. have reported that MUC5AC + PDAC tends to be a more differentiated adenocarcinoma [36]. In our study, these tumors were found to start expressing β3Gn-T6, which significantly correlated with MUC5AC expression (Fig 4 and Table 4). These results suggest that this induction of expression may be associated with gastric metaplasia. Carcinoma cells, such as colonic, prostate, and pancreatic cancer cells [8,21,22], reduce their aggressiveness in vitro or in vivo when forced to express β3Gn-T6. Forced expression of the core 3 structure destabilizes oncoprotein MUC1 [22], affecting downstream signals and upregulating cell cycle inhibitor p21 [22]. The expression of β3Gn-T6 also leads to a reduction in the formation of the α2β1 integrin complex, subsequently reducing the level of phosphorylated FAK relative to total FAK, thereby leading to decreased tumor progression [21,22]. β3Gn-T6 alters cancer cell invasion through impairment of actin stress fiber organization [21,37]. Here, we demonstrated clinical significance of β3Gn-T6 expression, which turned out to be a favorable prognostic factor in PDAC, consistent with the known effects of β3Gn-T6 in cancer biology.
In the biosynthesis of O-glycans (Fig 1), the processes of formation of core 1 and core 3 structures compete with each other for the same substrate, although the expression levels of T antigen and β3Gn-T6 only weakly correlated in PDAC. This finding suggests that the core 1 structure (T antigen) is mostly modified, i.e., by extension, branching, or sialylation, which are not recognized by PNA. The extended core 1 structure detected by antibody MECA-79, which was only a limited 6-sulfated structure of extended core 1, weakly but statistically significantly negatively correlated with β3Gn-T6 levels ( Table 4).
The Tn antigen is one of the representative truncated structures, whose expression is abundant in many types of carcinoma cells owing to the loss of Cosmc, a chaperone for core 1 synthase [9,38]. Our study revealed that the positivity and Exp-score of the Tn antigen in PDAC cells were high (Table 3 and Fig 3), despite the high expression of sLeX in PDAC cells, suggesting that Cosmc inactivation does not entirely explain the presence of the Tn antigen in our cohort.
This study has several limitations. First, data collection and analyses were performed retrospectively. Second, it was difficult to investigate detailed glycan structural alterations and their changed biosynthesis in cancer cells in clinical samples, because of the lack of antibodies specific for various glycan structures. Therefore, our conclusions are drawn from speculation based on the limited findings. Further studies would be warranted to clarify the molecular mechanism of glycan alterations in PDAC.
In summary, this study is the first to report on the clinicopathological significance of β3Gn-T6 expression in PDAC. β3Gn-T6 was found to be expressed most highly in PDAC cells in differentiated adenocarcinoma and was significantly associated with longer DFS in PDAC patients. Our findings on the molecular mechanisms underlying the induction of core 3 O-glycan in PDAC provide a basis for its use as a therapeutic tool.