This study was approved by the Institutional Review Board of the National Cancer Center, Tokyo, Japan (2019–0186). Informed consent was obtained from all participants before their participation in the study. All clinical investigations were conducted in accordance with the principles of the Declaration of Helsinki.

This study included 185 patients, 118 with dCCA and 67 with pCCA underwent surgical resection between 2006 and 2016 at the National Cancer Center Hospital, Tokyo. All patients included in this study underwent macroscopic curative resection. None of the patients received radiotherapy or chemotherapy before surgery. Recurrence was diagnosed when a new local or distant metastatic lesion was detected in imaging studies or when an increase in tumor marker levels with deterioration of the patient's condition was observed. All of the CCA were diagnosed according to the World Health Organization (WHO) classification (2), the 8th UICC/AJCC TNM Classification (3,4). Macroscopic types of eCCA and the following histopathological variables were evaluated following the Japanese Society of Biliary Surgery (JSBS) classification (4): lymphatic, venous, and perineural invasions. The clinicopathological characteristics of patients are summarized in Table SI. All patients with stage IV disease were diagnosed on the basis of para-aortic lymph node involvement.

The monoclonal antibody that reacts with core 3 synthase, called G8-144, was previously established by our laboratory (10). First, G8-144 was purified using protein-G beads. G8-144 secreted from hybridoma cells was diluted with PBS to a total volume of 25 ml and applied to 200 µl protein-G Sepharose fast flow (GE Healthcare) for antibody capture at 4°C overnight. The protein-G beads were washed and eluted with 0.1 M glycine (pH 2.5), followed by neutralization with 3 M Tris-HCl (pH 8.5). Subsequently, the purified antibody was dialyzed with pH 7.5 phosphate buffer saline (PBS; pH; Fujifilm Wako Pure Chemical Corporation), and concentrated using an Amicon ultra 0.5 ml filter cut off 100 K (Merck). Purified G8-144 was subjected to immunohistochemical analysis.

The MECA-79 antibody recognizes a sulfated N-acetyllactosamine on extended core 1 mucin-type O-glycan, Galβ1→4(sulfo→6)GlcNAcβ1→3Galβ1→3GalNAc (6-sulfo LacNAc on core 1) (20). The MECA-79 rat monoclonal IgM antibody used to stain the serial pathological sections was purchased from BD Biosciences.

To confirm the specificity of G8-144 before performing the immunohistochemical study, the purified G8-144 antibody was subjected to western blot analysis. As most of the core 3 synthase-expressing cells, which were transfected with core 3 synthase cDNA, were reported as overexpressing cell lines, there are only a few cell lines that stably express core 3 synthase (21). To prepare core 3 synthase (B3GNT6), B3GNT6 cDNA was subcloned into the pCDN3.1 vector (Thermo Fisher Scientific, Inc.), resulting in pcDNA3.1-B3GNT6. Thereafter, the purified plasmid DNA was transfected into HEK293T cells using Lipofectamine^® LTX DNA transfection reagent (Thermo Fisher Scientific, Inc.), according to the manufacturer's protocol. Forty-eight hours after transfection, the cells transiently expressing the core 3 synthase were collected and lysed with lysis buffer containing 1% NP-40, 0.1% SDS, 50 mM Tris-Cl pH 7.4, 150 mM NaCl, 5 mM EDTA, 1 mM PMSF, and protease inhibitors (Merck). Frozen tissues were obtained from resected surgical specimens and were stored at −80°C until use. For comparison between pathological specimens and cancer cell lines, frozen CCA tissue samples from the G8-144 positive cases and HAL8 cells, a lung adenocarcinoma cell line (22), were lysed using the lysis buffer. After centrifugation, the supernatant was collected, and the protein concentration was measured using a micro-BCA kit (Thermo Fisher Scientific, Inc.). Core 3 synthase might be expressed in minute amounts, so collecting them from the positive sites in the frozen tissue may provide limited yield. Thus, the samples were concentrated from frozen tissue and HAL8 cells using an Amicon 3 K (Merck) tube. The proteins were then loaded and separated using 5–20% gradient SDS-PAGE, while the HEK235T+B3GNT6 and HEK293T cell lysates were loaded at 5 µg/µl and blotted onto a polyvinylidene fluoride membrane (Bio-Rad). Proteins that reacted with G8-144 or GAPDH (Fujifilm Wako) were detected using anti-mouse IgG conjugated with HRP (Thermo Fisher Scientific, Inc.) and Immobilon Forte western HRP substrate (Merck).

The representative tumor tissue was used for all the examinations of the tissue specimen in this study. To evaluate the expression levels of core 3 synthase and 6-sulfo LacNAc on core 1 in CCA, serial CCA sections were immunohistochemically stained using G8-144 and MECA-79. Briefly, formalin-fixed and paraffin-embedded sections were deparaffinized in xylene and rehydrated using a graded ethanol series. After the sections were washed with deionized water, endogenous peroxidase was blocked with 0.3% H₂O₂ in methanol for 20 min. Antigens were retrieved via autoclaving in 1X Envision™ flex target retrieval solution at high pH (Dako). The sections were washed with PBS, and endogenous biotin and biotin-binding factors were blocked with a biotin-blocking system (Dako). After washing with PBS, the sections were incubated with each antibody overnight at 4°C (23). G8-144 (anti-core 3 synthase antibody) was used at a concentration of 0.5 µg/ml while MECA-79 (anti-6-sulfo LacNAc on core 1 antibody) was used at a concentration of 1.0 µg/ml. The sections were incubated with the secondary antibody, biotinylated goat anti-mouse IgG (Vector Laboratories) for G8-144 or biotinylated goat anti-rat IgM (Vector Laboratories) for MECA-79 for 30 min. Thereafter, they were reacted for 30 min to form the avidin-biotin-peroxidase complex (Vectastain ABC kitl Vector Laboratories). Staining was visualized with diaminobenzidine. The sections were also counterstained with hematoxylin.

Cancerous tissues were divided into two regions: non-invasive and invasive regions. Cancer cells that remained on the epithelial layer of the bile duct were determined to be non-invasive cancer cells, while cells infiltrating beyond the epithelial layer were determined to be invasive cancer cells. To evaluate the level of staining positivity, sections were quantitatively scored based on the percentage of positive cells in the total cancer cells (1–100%). When the immunolabelled positive cells were more than 1%, the cancer was judged as a positively stained. The staining scores were evaluated by two independent pathologists that were blinded to the clinical status.

The associations between characteristic variables were analyzed by chi-square or the Fisher exact tests. Postoperative overall survival (OS), disease-free survival (DFS) rates, and the expression of core 3 synthase and MECA-79 positivity were calculated using the Kaplan-Meier method and analyzed by the log-rank test (24). Factors found to be significant in univariate analysis were incorporated into the multivariate analysis using the Cox proportional hazards model (backward elimination method). Differences were considered statistically significant at P<0.05, and all statistical analyses were performed using StatView-J 5.0 software (Abacus Concepts).

To determine whether G8-144 could be applied to immunohistochemical studies for diagnostic purposes, we assessed the specificity of G8-144 by western blotting using CCA samples. HAL8, a lung adenocarcinoma cell line, was used as a positive control cell line expressing core 3 synthase [(22), data not shown]. HEK293T cells that did not express core 3 synthase were compared with HEK293T cells transfected with B3GNT6 cDNA. Lysates purified from CCA tissues that were positive for G8-144 staining were blotted. The results indicated that this enzyme was approximately 45 to 50 kDa, as detected using the G8-144 antibody (Fig. S1, upper right). The size differences may be due to differences in glycosylation between transient expression in cultured cells and human CCA tissue. Western blotting results showed several weak positive bands with higher molecular weights, which may be polymers or complexes with other proteins. Such findings indicate that a detectable amount of core 3 synthase with G8-144 antibody was expressed in the CCA tissue, unlike that in colorectal cancer (11).

To explore the expression of core 3 synthase in CCA and its clinicopathological correlation with patient prognosis, an immunohistochemical study using the G8-144 antibody was performed using 185 CCA patient specimens. Briefly, we classified the 185 CCA pathological specimens into two groups, 118 samples of dCCA and 67 of pCCA, based on their anatomical location. The dCCA specimens were retrieved from 98 males and 20 females of different ages; median age was 72 years. As shown in (Table SI), the study included 64 cases of patients ≥72 and 54 cases of patients <72 years of age. Meanwhile, 51 male and 16 female samples were retrieved for pCCA; median age was 67 years. There were 34 cases of patients ≥67 years and 33 cases of patients <67 years).

The results of immunohistochemical staining with G8-144 showed that dCCA and pCCA were positive for G8-144 in 55.1 and 59.7% of the total cases, respectively (Table I). No core 3 synthase was expressed in any of the normal epithelial cells in the bile duct (Fig. S1, lower left). It was expressed frequently in the intraepithelial non-invasive cancer cells (Fig. 2, upper left panel) and the invasive cancer cells (Fig. 2, upper right panel) at the shallow invasive lesion, that is, near mucosal layer in the bile duct wall in the eCCA. It could be found in a low frequency that cancer cells invaded deeply into the bile duct expressed core 3 synthase. These findings were observed in both of dCCA and pCCA. The frequencies of core 3 synthase expression in non-invasive cancer cells and invasive cancer cells were 52.7 and 33.6% in dCCA, respectively (Table I), and 51.9 and 38.2% in pCCA, respectively (data not shown).

To obtain further knowledge on O-glycan expression in eCCA, MECA-79, which is specific to 6-sulfated N-acetyllactosamine (6-sulfo LacNAc) on the extended core-1 O-glycans (Fig. 1), was compared to G8-144. Serial sections of the pathological specimen used for G8-144 staining were subjected to MECA-79 staining (Fig. 2). In contrast to G8-144 staining, MECA-79 antigen was expressed in the non-cancerous epithelial cells of the bile duct (Fig. S1, lower right). MECA-79 antigen was expressed in 69.5 and 55.2% in dCCA and pCCA, respectively. It tended that MECA-79 antigen was expressed more frequently in deeply infiltrating cancer cells (Fig. 2, lower left panel) and often found in cancer cells showing lymphovascular or perineural invasion (Fig. 2, lower right panel). The frequencies of MECA-79 antigen expression in non-invasive cancer cells and invasive cancer cells were 49.1 and 62.7% in dCCA, respectively (Table I), and 42.3 and 50.9% in pCCA, respectively (data not shown). Thus, cancer cells expressing core 3 synthase and those expressing MECA-79 were different cancer areas. Furthermore, we often observed that in the serial specimens, cancer cells expressing core 3 synthase did not display MECA-79-positivity in the non-invasive (Fig. 3, upper left and right) or invasive (Fig. 3, lower left and right) areas.

As summarized in Table I, the 118 dCCA patients were divided into four groups: i) positive staining in both non-invasive and invasive regions (31 patients), ii) positive in non-invasive but negative in invasive region (28 patients), iii) negative in non-invasive but positive in invasive region (6 patients), and iv) negative in both non-invasive and invasive regions (53 patients). pCCA patients were not included in this table as no correlation was found between G8-144 staining and patient prognosis as shown in later.

To analyze the correlation between core 3 synthase expression and the survival rate of dCCA patients, the classified staining results were subjected to Kaplan-Meier survival analysis (Fig. 4). As shown in the left panels of Fig. 4, the dCCA patients showing positive G8-144-staining in non-invasive or invasive cancer regions (65 of the 118 patients, 55.1%) had a significantly better prognosis than the G8-144-negative patients (53 patients). The difference between core 3 synthase -positive and -negative survival rates of dCCA patients was significant in postoperative OS (upper panels) and DFS (lower panels) (P=0.007 and P=0.024, respectively). The dCCA patients with positive staining in the non-invasive region displayed a better prognosis than those with negative dCCA (P=0.031 and P=0.035 in OS and DFS, respectively; middle panels of Fig. 4). The dCCA patients with positive staining in the invasive region also showed a slightly good correlation with OS (P=0.029) relative to those with negative dCCA; however, not significantly correlated in DFS (P=0.153) (right panels of Fig. 4). Moreover, there was no significant correlation between core 3 synthase expression and outcome in patients with pCCA (Fig. S2).

Various clinicopathological factors shown in Table II were investigated to determine whether they were prognostic. When the significant factors identified at univariate analysis were subjected to the multivariate analysis; core 3 synthase positive in cancer cells, no nodal metastasis, negative tumor margin status, and lower lymphatic invasion were closely associated with longer DFS. In addition, core 3 synthase positive in cancer cells, lower tumor status, no nodal metastasis and lower lymphatic invasion were significantly associated with longer OS. Thus, dCCA patients expressing core 3 synthase, regardless of the non-invasive or invasive region, displayed a significantly better prognosis than negative dCCA patients and it is indicated that core 3 synthase expression in any cancer cells is an independent prognosticator. As presented in Table SI, core 3 synthase displayed high positivity accounting for 100% of positive cases in tumors in situ or at the zero stage.

MECA-79 positivity in the invasive cancer region of the dCCA was significantly associated with patient survival rate in an unfavorable manner (P=0.006 for OS, P=0.004 for DFS) (right panels of Fig. 5). While MECA-79 positivity in the non-invasive cancer did not have any significant association with patient outcome (Fig. 5). When the significant factors identified at univariate analysis were subjected to the multivariate analysis, expression of MECA-79 antigen in invasive cancer cells was not a significant factor in the multivariate analyses of both DFS and OS (Table II). There was no significant survival effect of MECA-79 antigen expression in cancer cells found in pCCA (Fig. S3).

When the correlation between MECA-79 antigen expression and the various clinicopathological factors was analyzed, MECA-79 positivity was significantly associated with the node status, and cancer progression, such as lymphatic and venous invasion (Table SI).

From the total of 185 patients with eCCA, two groups, G8-144(+)-MECA-79(−) and G8-144(−)-MECA-79(+), were selected and analyzed to compare which O-glycan pathway (core 3 or core 1) was associated with a longer survival rate. From the results, we suggest that G8-144 is a favorable type of core structure in dCCA. The OS and DSF of G8-144 and MECA-79 in the eCCA cases are shown in Fig. 6. Both OS and DFS were longer in the G8-144(+)-MECA-79(−) group than in the G8-144(−)-MECA-79(+) group, except for the OS of the non-invasive eCCA cases. These results suggest that when eCCA with G8-144(+)-MECA-79(−) progresses to G8-144(+)-MECA-79(+) and G8-144(−)-MECA-79(+), the survival rate may decrease as malignant transformation progresses.

We expected to find an inverse relationship between G8-144 and MECA-79 expression due to their glycan structures. However, there was no significant correlation between the expression of core 3 synthase and MECA-79 in the CCA cells (Table SII). This may be due to the fact that both antigens are not diffusely expressed on tumor cells in the same case, but are often expressed on some tumor cells separately.

The frequency of G8-144 expression is high in non-invasive intramucosal cancers and tends to decrease as the cancer cells invade deeper into the mucosa, but the low frequency of G8-positive tumor cells makes data generation difficult. However, based on the characteristics of G8-144, which is frequently expressed in non-invasive intramucosal cancers, the relationship between G8-144 and depth of invasion should be statistically significant. This would also support that G8-144 is a good prognostic marker, but it is unclear since we have not searched for it.