Polypeptide N-acetylgalactosaminyltransferase 6 expression in pancreatic cancer is an independent prognostic factor indicating better overall survival

Background: The family of polypeptide N-acetylgalactosaminyltransferases (GalNAc-Ts) is responsible for the altered glycosylation in cancer. The purpose of our study was to investigate the clinical significance of two isoforms, GalNAc-T6 and -T3, and their correlation with the prognosis of pancreatic cancer. Methods: Immunohistochemistry was used to analyse GalNAc-T6 and -T3 expressions in 70 clinicopathologically characterised pancreatic cancer cases. Results: Positive expressions of GalNAc-T6 and -T3 were immunohistochemically identified in 51% (36 of 70) and in 77% (54 of 70) of patients, respectively. A close relationship was noted between GalNAc-T6 positive expression and pathological well/moderate differentiated type (P=0.001), small tumour size (P=0.044), absence of vascular invasion (P=0.009), and low stage of the American Joint Committee on Cancer systems (P=0.043). The expression of GalNAc-T3 significantly correlated with good differentiation (P=0.001), but not with other clinicopathologic features. Furthermore, univariate and multivariate analyses revealed that GalNAc-T6 expression was an independent prognosis indicator for the disease, whereas GalNAc-T3 expression had no impact on clinical outcome, even though 33 of 36 GalNAc-T6-positive cases also had a positive expression of GalNAc-T3 (P=0.001, r=0.356). Conclusion: Both GalNAc-T6 and -T3 expressions correlated significantly with tumour differentiation, whereas only GalNAc-T6 expression predicted prognosis in pancreatic cancer.

Pancreatic cancer is responsible for B227 000 deaths per year in the world (Parkin et al, 2005). The 5-year overall survival rate for all patients diagnosed with pancreatic cancer is o4%. At present, surgical resection offers the only chance of cure, whereas longterm survival is rare, with the overall 5-year survival rates ranging from 10% to 25% (Cleary et al, 2004;Katz et al, 2008). Various clinicopathologic features, such as tumour size, and vascular involvement in resectable tumours, or initial patient performance status and the absence of distant metastasis in unresectable tumours, have been proposed as prognostic indicators, although the results remain inconsistent and inconclusive to date. The clinical importance of biological markers is also under evaluation. For example, most famously, initial serum levels of carbohydrate antigen (CA) 19-9, and more recently, Mucin-4, have been shown to be independent prognosis indicators of pancreatic cancer (Saitou et al, 2005;Berger et al, 2008). All these factors are useful guides in standardised clinical management and in establishing an individualised treatment plan for a patient.
Carbohydrate antigen 19-9 and mucins belong to cell-surface CAs, which comprise various categories of tumour markers and are essential in monitoring cell growth and the invasive and metastatic status of solid tumours. In mammals, the most common forms of CAs are the mucin-type O-glycans, which may constitute up to 80% of them (Bresalier, 1994;Brockhausen, 1999;Hollingsworth and Swanson, 2004;Nagata et al, 2007). The aberrant glycosylation of mucin-type O-glycans is thought to be associated with the functional alteration of cancer cells, including their antigenic and adhesive properties, as well as their potential for invasion and metastasis (Brockhausen, 2006). The enzymes responsible for glycosylation of CA -the glycosyltransferases and glycosidases -thus gain more and more attention (Ohtsubo and Marth, 2006). Here, we focus on the family of polypeptide N-acetylgalactosaminyltransferases (GalNAc-Ts) which catalyses the transfer of N-acetyl-a-D-galactosamine:polypeptide (GalNAc) from the sugar donor uridine diphosphate (UDP)-GalNAc to the serine and threonine residues of glycoproteins to synthesise the sugar chains which catalyse the initial step in the synthesis of mucin-type O-glycans (Marth, 1996).
Despite this seemingly simple function, the GalNAc-Ts form a big family which includes at least 15 distinct members and as many as 24 human isozymes. They are localised differentially within cellular compartments, display tissue-specific expression, and have different, although partly overlapping, kinetic properties and acceptor substrate specificities (Ten Hagen et al, 2003). Aberrant expression of the GalNAc-T isoforms results in aberrant O-glycosylation and participates in the progression of cancer (Hollingsworth and Swanson, 2004). According to the previous literature, GalNAc-T1 was considered to decrease ovarian cancer risk (Phelan et al, 2010), and the higher expression levels of GalNAc-T14 correlated with lower histological grade in breast invasive ductal carcinoma and tumour progression . The most extensively studied GalNAc-T was GalNAc-T3, which was proved by our group and others to be expressed in adenocarcinoma cell lines, but not in other carcinoma cell lines (Sutherlin et al, 1997;Nomoto et al, 1999). Recent reports showed that the expression of GalNAc-T3 was a useful indicator of prognosis in patients with colorectal, gastric, and lung cancer (Shibao et al, 2002;Onitsuka et al, 2003;Gu et al, 2004). GalNAc-T6 exhibited a high sequence similarity to GalNAc-T3 in the coding region and shared similar kinetic properties and acceptor substrate specificities which are distinct from the other GalNAc-Ts (Bennett et al, 1999). The expression of GalNAc-T6 has been reported in most ductal breast carcinomas showing a significant correlation with T1 stage , whereas in gastric carcinomas it is significantly associated with the presence of vascular invasion (VI) (Gomes et al, 2009). GalNAc-T6 thus could be considered as an intriguing marker of glycosylation modifications in cancer. To date, however, no study has assessed GalNAc-T6 expression and its correlation with GalNAc-T3 in pancreatic cancer.
In this study, we evaluated the expressions of GalNAc-T6 and -T3 in pancreatic cancer specimens and their correlation with the clinicopathologic features of the patients. The results indicated that GalNAc-T6 and -T3 were highly expressed in half of human pancreatic cancer cases, all of which were well to moderately differentiated type, while no or very weak staining was observed in normal pancreas ductal epithelium or poorly differentiated cancer. Furthermore, positive expression of GalNAc-T6, but not GalNAc-T3, is associated with good clinical prognosis of patients.

Patient information, tumour specimens, and blood samples
The use of specimens from human subjects was approved by the Ethics Committee of the University of Occupational and Environmental Health (UOEH) Medical Center. Pathology reports were reviewed to identify patients who had undergone pancreaticoduodenectomy or distal pancreatectomy for pancreatic cancer between 1994 and June 2009 at the hospital of UOEH. Two patients who had suffered perioperative deaths, defined as death during the patient's initial hospitalisation or within 30 days of surgery, were excluded. A total of 70 patients with available follow-up data comprised the cohort of this retrospective study. Three pathologists examined all the resected specimens to confirm the histopathologic features. The tumour node metastasis system of the American Joint Committee on Cancer (AJCC) 6th edition was used for staging. Clinical information was gathered from the patients' records, and the survival duration was set as the period from the date of surgery to death or the most recent clinic visit. Formalin-fixed, paraffinembedded tissue blocks were obtained from our department of pathology. Serum CA19-9 and carcinoembryonic antigen (CEA) levels were measured at the time of final diagnosis. Normal human tissue was obtained from non-tumour portions of the surgically resected specimens.

Preparation of antibody against GalNAc-Ts
Polyclonal antibody against GalNAc-T6 was raised by multiple immunisations of New Zealand white rabbits with synthetic peptides. The sequence of the synthetic peptide was GFYT-PAELKPFWERPPQDP. The specificity of antibody was confirmed by western blotting and immunohistochemistry with peptide competition. Anti-GalNAc-T3 polyclonal antibody was generated in the same way as described previously (Nomoto et al, 1999).

Immunohistochemistry of tissue samples
Immunohistochemical staining was performed by the antibodylinked dextran polymer method (EnVision; Dako Cytomation, Inc., Carpinteria, CA, USA). Deparaffinised and rehydrated 5 mm sections were incubated in 3% H 2 O 2 for 10 min to block endogenous peroxidase activity. The sections were then rinsed and incubated with rabbit polyclonal GalNAc-T6 or -T3 antibody for 30 min. The second antibody-peroxidase-linked polymers were then applied, and the sections were incubated with a solution consisting of 20 mg of 3.3 0 -diaminobenzidine tetrahydrochloride, 65 mg of sodium azide, and 20 ml of 30% H 2 O 2 in 100 ml of Tris-HCl (50 mM, pH 7.6). After having been counterstained with Meyer's haematoxylin, the sections were observed under a light microscope. For immunohistochemistry of the neoplastic tissue, positive areas comprising o10% of the neoplasms were considered as negative staining. Positive areas that were equal to or 410% were defined as positive staining and were graded into three categories: weak, positive area of 11 -30%; moderate, 31 -80%; and strong, 480% positive area.

Cell cultures
Human pancreatic invasive ductal carcinoma cell lines PANC1 (American Type Culture Collection) and PK1 (a kind gift from the Institute of Development, Aging and Cancer, Tohoku University, Japan) were maintained in a DMEM medium containing 10% fetal calf serum (GIBCO BRL, Rockville, MD, USA) at 371C in an atmosphere of 95% air and 5% CO 2 .

Immunofluorescence of pancreatic cancer cell lines PANC1 and PK1
The pancreatic cancer cell lines PANC1 and PK1 were cultured on coverslips, fixed with 95% acetone for 5 min, and allowed to air dry. The cells were then incubated with anti-GalNAc-T6 antibodies for 1 h at room temperature (RT), washed with PBS, and reacted with fluorescein isothiocyanate-conjugated goat anti-rabbit IgG for 1 h at RT. After having been washed with PBS, the specimens were observed under a Nikon ECLIPSE E600 inverted fluorescence microscope (Nikon, Tokyo, Japan).

Cell fractionation and western blotting
The cells were washed and lysed with RIPA buffer. For nuclear and cytoplasmic fractions, the cells were scraped in 100 ml ice-cold PBS and centrifuged at 500 g for 5 min. The cell pellet was resuspended in 100 ml Buffer A (50 mM NaCl, 10 mM HEPES pH 8.0, 500 mM sucrose, 1 mM EDTA, 0.2% Triton X-100, freshly added protease inhibitors and 7 mM 2-mercaptoethanol), vortexed at high speed for 45 s and centrifuged at 2000 g for 2 min at 41C. The resulting supernatant was used as a cytoplasmic extract. The pellet was resuspended in 100 ml Buffer B (Buffer A with 25% glycerol and GalNAc-T6 in pancreatic cancer Z Li et al 0.1 mM EDTA) and nuclei were pelleted at 2000 g for 2 min at 41C. The pellet was dissolved in 50 ml Buffer B, incubated on ice for 30 min with intermittent high speed vortexing and spun down at 11 000 g for 15 min at 41C. The supernatant was diluted to 100 ml with PBS and used as a nuclear extract. Equal amounts of samples were analysed on an SDS -PAGE gel and then were transferred onto a nitrocellulose membrane. The membranes were incubated with the primary antibody and visualised with a secondary antibody coupled to horseradish peroxidase (Cell Signaling Technology, Beverly, MA, USA) and SuperSignal West Pico Chemiluminescent Substrate (Pierce, Rockford, IL, USA). The bands on the western blots were analysed densitometrically using Scion Image software (version 4.0.2; Scion Corp., Frederick, MD, USA).

Statistical analysis
Spearman's correlation test was used to assess the relationships between variables. Survival curves were plotted by the Kaplan-Meier method and compared with the log-rank test. Hazard ratios and 95% confidence intervals (95% CIs) were estimated using univariate or multivariate Cox proportional-hazard models. All statistical tests were two-tailed, with Po0.05 considered significant. SPSS statistical software (SPSS for Windows, version 16.0.0 (SPSS, Chicago, IL, USA)) was used for the above statistical analyses.

Association of GalNAc-Ts expression with clinicopathologic variables
To identify the association of GalNAc-Ts expression (GalNAc-Ts negative vs positive) with the clinical and pathological variables of the cohort, the variables were dichotomised as shown in Table 1. There were no significant differences between the patients with GalNAc-T3 negative and positive tumour expressions regarding age, gender, tumour location, size, lymph node involvement, disease stage, margin status, lymphatic invasion, VI, perineural invasion or serum CA19-9 or CEA level (P40.05), whereas GalNAc-T3 staining status significantly affected tumour differentiation (P ¼ 0.001; Figures 2C, E, and G). The rate of well to moderately differentiated tumours in the GalNAc-T3-negative group was 10 out of 16 (62.5%), but 51 out of 54 (94.4%) in the GalNAc-T3-positive group. These data were consistent with the results of a previous study using the same antibody as ours (Yamamoto et al, 2004). Positive GalNAc-T6 staining also correlated significantly with well to moderately differentiated type (P ¼ 0.001; Figures 2D, F, and H). Moreover, it was associated with early disease stage and absence of VI (P ¼ 0.043 and 0.009, respectively) in all the tumours and with decreasing tumour size in the resectable tumours (P ¼ 0.044). Thirty out of thirty-six cases (83.3%) in the GalNAc-T6-positive group were classified into stage I/II of the AJCC TNM classification, compared with 21 out of 34 (61.8%) in the GalNAc-T6-negative group.
The 5-year overall survival rate of the cohort was 6%. In a Kaplan -Meier analysis (Figure 3), GalNAc-T3 expression showed no prognostic significance in overall survival. In contrast, patients with GalNAc-T6-positive expression had significantly longer overall postoperative survival (median, 21.8 months) compared with those who had GalNAc-T6-negative expression (median, 9.3 months; P ¼ 0.014). For the patients with GalNAc-T6-positive tumours, an increasing degree of GalNAc-T6 expression was not associated with better clinicopathologic features or longer survival (data not shown).

GalNAc-T6 represents an independent prognostic indicator in pancreatic cancer
To assess whether GalNAc-T6 expression was an independent predictor of overall postoperative survival, a Cox proportionalhazards model was created in a forward manner including only the covariates that had a statistically significant correlation (inclusion threshold, Po0.05) with overall survival (Table 2). Univariate analysis demonstrated that poor tumour differentiation, advanced stage and GalNAc-T6-negative status were significant predictors of poorer survival (P ¼ 0.003, o0.001, and 0.014, respectively). Margin status had no significant impact on patient survival in our relatively small cohort (P ¼ 0.124).
Furthermore, multivariate analysis showed that, after correction for confounding variables, GalNAc-T6 expression remained an independent prognostic indicator of overall survival (P ¼ 0.024).

Analysis of GalNAc-T6 expression in human pancreatic cancer cell lines
The immunofluorescence staining of the pancreatic adenocarcinoma cell lines (PANC1 and PK1) showed a marked expression of GalNAc-T6 in the cytoplasm in a perinuclear manner (Figure 4). Western blotting analysis showed GalNAc-T6 expression in the cytoplasm, whereas very weak nuclear expression was detectable in both cell lines (Figure 4). Additionally, the expression levels of GlaNAc-T6 in PANC1 were significantly higher than those in the PK1 cell line by both methods.

DISCUSSION
Prior studies revealed that GalNAc-T6 expression was detected in a few kinds of malignancies, and breast cancer was the most studied type in this field. Interestingly, one group showed that GalNAc-T6 was expressed in most ductal carcinoma in situ and was significantly associated with the T1 stage rather than higher stage tumours . Other studies showed that GalNac-T6 mRNA positively correlated with bone marrow metastasis (Freire et al, 2006), and knockdown of GalNAc-T6 suppressed the growth of breast cancer cells (Park et al, 2010). These conflicting results   GalNAc-T6 in pancreatic cancer Z Li et al are probably due to the heterogeneity of breast cancer, which can be supported by the fact that distinctive chemotherapy drugs showed good efficacy on it. By contrast, pancreatic cancer has been supposed to be a relatively homogeneous cancer with a low frequency of either molecular or histological variants (Niederhuber et al, 1995) and would be a better model to demonstrate the function of GalNAc-T6 in malignancies. Thus, we examined the clinical relevance of GalNAc-T6 expression in pancreatic cancer. Expression of GalNAc-T6 was observed in half of pancreatic cancer specimens, associated closely with well to moderately differentiated type (Po0.001), VI (P ¼ 0.009), early AJCC stage (P ¼ 0.043), as well as small tumour size in resectable cases (P ¼ 0.044). In both univariate and multivariate analyses, a positive GalNAc-T6 status predicted improved overall survival. To our knowledge, this is the first study to reveal that GalNAc-T6 was an independent prognostic indicator in patients with pancreatic cancer. Further analysis of the correlation between GalNAc-T6-positive expression and better 1-year or 2-year overall survival (P ¼ 0.0003 and 0.006, respectively; Supplementary Table 2) indicated that positive GalNAc-T6 expression significantly predicted a survival advantage, especially in the early phase after resection, very similar to postoperative serum CA19-9 (Berger et al, 2008). Thus, negative GalNAc-T6 status recommended more aggressive treatment. Because GalNAc-Ts are in external secretions such as colostrum, based on the previous literature (Hagen et al, 1993;Bennett et al, 1996;Kitazume et al, 2001), it is very likely that GalNAc-T6 could be secreted to body fluids too. Therefore, GalNAc-T6 is considered as a useful candidate for surveillance because it is well known that 480% postoperative relapse (local or distant) occurs within the first 2 years (Stathis and Moore, 2010).
Additionally, the hypothesis that GalNAc-T6 could be a good indicator of local aggression and metastasis in pancreatic cancer is supported by our result that GalNAc-T6 was negatively associated with VI (P ¼ 0.009, r ¼ À0.181). Not only that, but the primary pancreatic cancer cell line (PANC1) demonstrated, by western blotting and immunocytochemistry methods, a significantly higher expression of GalNAc-T6 (Figure 4) than that of PK1 derived from liver metastasis (Kobari et al, 1984;Motomura et al, 2000), which was considered as a more aggressive subtype. These results were consistent with the data of Bennett et al (1999) in which northern blotting analysis showed that GalNAc-T6 mRNA could only be detected in one primary cancer cell line (miaPaca2), but not in other metastatic cancer cell lines such as ASPC1 and COLO357. Very similarly, our preliminary data showed a relatively higher expression of GalNAc-T6 in primary colon cancer specimens than that in asynchronous liver metastatic specimens (data not shown).
The molecular mechanism to explain this tight relationship between GalNAc-T6 status and local invasion or metastasis remains to be elucidated. However, recent studies focusing on the modulation of GalNAc-Ts to cell adhesion function, as well as the degradation of connective tissue, provide some useful clues. Several GalNAc-Ts were shown to affect leukocyte adhesion through modulating E-and P-selectin counter receptors (Lowe, 2003;Tenno et al, 2007); polypeptide GalNAc-T3, in Drosophila, was found to regulate integrins and to promote cell adhesion (Zhang et al, 2008; and GalNAc-T3 was found to modulate the activities of metalloproteinases (Chefetz et al, 2009). GalNAc-T6 was also found to regulate the cell adhesion molecular E-cadherin and b-catenin in breast cancer (Park et al, 2010). Therefore, it is reasonable to hypothesise that GalNAc-T6 can suppress invasion and metastasis by impacting cell -cell adhesion and cell -stromal interaction in pancreatic cancer, and we attribute this to the early survival advantage of GalNAc-T6 expression mentioned above.
Interestingly, we found that positive GalNAc-T6 expression was closely associated with GalNAc-T3 positivity (P ¼ 0.002, r ¼ 0.356) in our cohort. We used the anti-human polyclonal GalNAc-T6 antibody raised against a completely distinctive synthetic peptide (GFYTPAELKPFWERPPQDP) from GYYTAAELKPVLDRPPQDS of GalNAc-T3 for both immunohistochemistry and western blotting (Nomoto et al, 1999). The results revealed GalNAc-T6or GalNAc-T3-specific staining in clinical specimens (Figurse 2A and B) and a distinct specific band of GalNAc-T6 (Figure 4) or GalNAc-T3 (data not shown) detected in the cytoplasmic fraction of pancreatic cancer cell lines, respectively. Additionally, the preliminary immunohistochemical examinations of different kinds of adenocarcinomas demonstrated ubiquitous expression of GalNAc-T3 in most specimens of lung, breast, gastric, and colon cancers, whereas GalNAc-T6 was expressed in only a few cases of them. All these confirmed the specificity of our GalNAc-T3 and -T6 antibodies.
Indeed, among all the isoforms, GalNAc-T6 and -T3 formed a subfamily with high sequence similarity in the coding region and shared similar kinetic properties and acceptor substrate specificities which are distinct from the other GalNAc-Ts (Bennett et al, 1999). It has been reported that there are at least two kinds of interaction -competition or complement -among the isoforms of GalNAc-Ts, as proven by experiments in vitro (Brockhausen et al, 1990;Bennett et al, 1998). Thus, the interaction of GalNAc-T6 and -T3 is a prospective topic awaiting illumination. At first, we hypothesised that the survival duration might be different between the double negative group and either positive group if the relationship between GalNAc-T3 and -T6 were complementary. On the other hand, if GalNAc-T3 and -T6 competed with each other, there might be some differences in the survival duration between the GalNAc-T3 positive-only group and the GalNAc-T6 positive-only group. Hence, the cohort was divided into two such manners (double negative vs either positive, and GalNAc-T3 positive-only vs GalNAc-T6 positive-only) and the Kaplan -Meier method was used to verify them. Unexpectedly, the overall survival of patients showed no significant difference (P ¼ 0.618 and 0.451, respectively) in either classification approach, suggesting that there was no apparent competitive or complementary correlation between GalNAc-T3 and -T6. Based on our analysis with a small cohort, GalNAc-T3 and -T6 expressed together, but might function separately, which was very interesting and needs further confirmation in larger cohorts later.
In summary, the present findings demonstrate that the positive expression of GalNAc-T6 in pancreatic cancer has a significantly close relationship with the well to moderately histopathologic phenotype, absence of VI, and high incidence of early stage of the AJCC system, namely well-differentiated and non-invasive characteristics. Moreover, the outcome of the patients who had tumours with positive GalNAc-T6 expression is significantly better than that with negative GalNAc-T6 expression, especially in the early period after surgery. On the other hand, most GalNAc-T6-positive cases expressed GalNAc-T3 as well, whereas they influenced overall survival separately. Based on these features, we can, for the first time, implicate that GalNAc-T6 is an independent novel and useful marker of prognosis in patients with pancreatic cancer, but GalNAc-T3 is not. Compared with more conventional markers like CA19-9, CEA, or mucins, GalNAc-T6 is suggested to have a potential but special advantage to predict not only the prognosis of patients, but local aggression and metastasis too. It is of clinical value to identify the danger of local invasion and metastasis, which are the primary causes of the poor outcome for this cancer. By contrast, we admit the disadvantage of this molecular marker. Its negative expression in undifferentiated cancer makes it difficult to be used as a diagnostic marker in pancreatic cancer. However, it might be a candidate for combination with the conventional markers mentioned above.
Actually, since this cohort was not big, margin status did not predict the ultimate outcomes in the univariate analysis. It is expected that the status and the prognostic value of GalNAc-T6 will be evaluated in a larger cohort. On the other hand, it is GalNAc-T6 in pancreatic cancer Z Li et al postulated that GalNAc-T6 had a key role in various types of cancer, thus evaluation of its prognostic value in other cancer types would be another important and intriguing topic. Furthermore, it is worth doing more molecular experiments on the potential invasive and metastatic suppressor function of GalNAc-T6 in pancreatic cancer. Finally, we plan to perform a high-throughput assay to evaluate the functions of other GalNAc-Ts in pancreatic cancers. It would be very interesting to further study the relationships between GalNAc-Ts and mucins.