BACKGROUND: The enzymes encoded by the GALNT [UDP-N-acetyl-α-d-galactosamine:polypeptide N-acetylgalactosaminyltransferase (GALNAC-T)] gene family catalyze the first step of O-glycosylation. Little is known about the link between expression of the genes encoding GALNAC-T enzymes and tumor progression in neuroblastoma, a pediatric cancer that can be classified as either low or high risk. We assessed the expression of genes in the GALNT family in a large cohort of neuroblastoma patients and characterized members of this family that might be used as new prognostic markers.
METHODS: Reverse-transcription PCR analysis of 14 GALNT genes with a panel of neuroblastoma cell lines identified the GALNT9 gene as playing a potential role in disease progression. We used the log-rank test and the multivariable Cox proportional hazards model with a cohort of 122 neuroblastoma patients to analyze the relationship between GALNT9 expression and overall survival or disease-free survival.
RESULTS: In the high-risk neuroblastoma experimental model IGR-N-91, GALNT9 expression was present in neuroblasts derived from primary tumors but not in neuroblasts from metastatic bone marrow. Moreover, GALNT9 in neuroblastoma cell lines was expressed in substrate adherent (S)-type cell lines but not in neuronal (N)-type lines. In the tumor cohort, GALNT9 expression was associated with high overall survival, independent of the standard risk-stratification covariates. GALNT9 expression was significantly associated with disease-free survival for patients currently classified as at low risk (P < 0.0007).
CONCLUSIONS: GALNT9 expression correlates with both improved overall survival in low- and high-risk groups and an improved clinical outcome (overall and disease-free survival) in low-risk patients. Thus, the GALNT9 expression may be a prognostic marker for personalized therapy.
Neuroblastoma, the most common malignant solid tumor diagnosed during infancy (median age at diagnosis, 17 months), accounts for 10% of all childhood cancers (1). Its biology may be extremely variable: Certain tumors regress spontaneously, whereas others are highly aggressive. The combination of age at diagnosis, tumor burden, histopathology, DNA index, and MYCN 7 [v-myc myelocytomatosis viral related oncogene, neuroblastoma derived (avian)] gene status is used to stratify risk categories (2). The low-risk group consists of non-MYCN–amplified tumors that exist either as localized forms (stages 1, 2, and 3) or as metastatic forms in children younger than 18 months (stage 4S). The high-risk group comprises all MYCN-amplified neuroblastomas, regardless of stage and age of the child, plus non-MYCN–amplified stage 4 neuroblastomas for children older than 18 months. The low-risk group has a survival rate of up to 90%, but the high-risk group consists of aggressive tumors that more frequently lead to death (5-year survival rate, approximately 35%).
Alterations in glycan profiles are a hallmark of cancer development linked to the expression of tumor-associated carbohydrate antigens (3). Abnormal O-glycans produced by cancer cells contribute to the malignant phenotype and play an important functional role in cell adhesion, invasion, and metastasis (4). The most abundant form of O-linked glycosylation in higher eukaryotes, termed “mucin-type,” is initiated in the Golgi apparatus by the covalent linkage of an α-N-acetylgalactosamine (GalNAc)8 residue to the hydroxyl group of Ser and Thr residues (5), a reaction catalyzed by the UDP-GalNAc:polypeptide N-acetylgalactosaminyltransferases (GALNAC-Ts) (EC 18.104.22.168) (6). These enzymes are a complex family of isoenzymes, with 20 members having been characterized to date (7). The genes encoding certain isoforms are broadly expressed, whereas other isoforms are restricted in their distributions and activities to certain cells or tissues (6). Individual GALNAC-Ts have distinct activities (8), and the O-glycosylation process in a given cell is regulated by the repertoire of the isoenzyme genes expressed in that cell (4).
The various genes encoding GALNAC-Ts are differentially expressed in malignant tissue, compared with normal tissue (9–11). Overexpression of the GALNT3 [UDP-N-acetyl-α-d-galactosamine:polypeptide N-acetylgalactosaminyltransferase 3 (GALNAC-T3)] gene promotes the growth of pancreatic cancer cells (12), whereas somatic and germline mutations in GALNT12 that encode a nonfunctional enzyme are associated with colon cancer development (13). Increasing evidence suggests that GALNAC-Ts might be useful tumor markers. GALNAC-T3 enzyme production, for example, has been shown to correlate with prognosis in patients with colorectal and gall bladder cancer (14, 15). The gene encoding GALNAC-T6 is expressed in breast cancer but not in normal breast epithelium, and this difference is apparent at both the mRNA (16) and protein (17) levels. In previous work, we used a reverse-transcription PCR (RT-PCR) assay to identify an association between GALNAC-T6 production in bone marrow samples and poor clinical outcome in lymph node–negative breast cancer patients (16). We have also shown that GALNT13, the gene encoding the GALNAC-T13 isoenzyme (18), might be a novel molecular marker of bone marrow involvement in neuroblastoma patients (19). When we applied microarray gene expression analysis to the xenograft-derived cell model of human neuroblastoma [the so-called IGR-N-91 cell line (20)], we also found that GALNT13 was the gene most strongly upregulated in metastatic malignant neuroblasts, compared with a primary-tumor xenograft (19). In the same study, we measured tyrosine hydroxylase, ganglioside D2 synthase, DOPA decarboxylase, and GALNT13 transcripts in bone marrow aspirates from the same neuroblastoma patients to evaluate whether GALNT13 might be useful for detecting disseminated disease. We found that GALNT13 expression in bone marrow at diagnosis was the strongest predictor of a poor clinical outcome. This evidence prompted the present study, in which we analyzed the expression of 14 GALNT genes at the mRNA level in a panel of neuroblastoma cell lines, including the IGR-N-91 model. Because we found the brain-specific isoform, encoded by GALNT9 (21), to be produced primarily in neuroblasts displaying a low aggressive behavior, we assessed the production of this enzyme in a cohort of 122 primary neuroblastoma tumors. Our findings indicate that GALNT9 gene expression is associated with a better clinical outcome in neuroblastoma patients.
Materials and Methods
PATIENTS AND CLINICAL SAMPLES
Tumor samples from a retrospective cohort of patients (n = 122) who had been staged according to the International Neuroblastoma Staging System (22) were collected between 1987 and 2009 at the Institut Gustave Roussy with the approval of the appropriate ethics committees and according to the national law applicable to people taking part in biomedical research. Primary-tumor tissues obtained from patients either via Tru-Cut (CareFusion Corporation) biopsy or after surgery were immediately snap-frozen and stored in liquid nitrogen until nucleic acid extraction. The main clinical features of the patients and the tumor characteristics are described in Table 1 in the Data Supplement that accompanies the online version of this article at http://www.clinchem.org/content/vol59/issue1.
TUMOR CRYOSECTIONS AND EXTRACTION OF TOTAL RNA
The first and last cryosections were used to select tumor tissues with a malignant tumor cell content of >60% (23, 24). Total RNA was isolated with TRIzol Reagent (Invitrogen) according to the manufacturer's protocols. Nucleic acid concentration and purity were measured with a Nanodrop ND-1000 spectrophotometer (NanoDrop/Thermo Scientific), and nucleic acid quality was checked with a 2100 Bioanalyzer (Agilent Technologies).
The neuroblastoma metastatic-cell model, IGR-N-91, was derived from a high-risk neuroblastoma through in vitro culture of malignant neuroblasts that had been collected from the bone marrow of an 8-year-old boy and then xenografted into nude mice, as has been described (20). Derived sublines were established from the primary-tumor xenograft (PTX) and bone marrow metastasis (BM) in nude mice. Other neuroblastoma cell lines [IMR-32, SH-SY5Y, SK-N-AS, SK-N-BE (2), LAN-1, and LAN-5] showing the typical phenotypes of established malignant neuroblasts [i.e., neuronal (N) and substrate-adherent (S)] were obtained from various sources (25). The cells were maintained in DMEM at 37 °C in a 5% CO2 humidified atmosphere. Total RNA was isolated with TRIzol Reagent according to the manufacturer's protocols.
Total RNA (1 μg) prepared from cell lines and primary tumors was reverse transcribed by Moloney murine leukemia virus reverse transcriptase (Amersham/GE Healthcare Life Sciences). The reaction mixture consisted of 200 U enzyme, 2 μL of 10 mmol/L of each deoxynucleoside triphosphate (dNTP), and 1 μL of 250 ng random hexamers in a 20-μL total reaction volume. Different RT-PCR reactions were run with the appropriate negative controls to amplify GALNT family members. The primer sequences are shown in Table 2 in the online Data Supplement. The B2M (β2-microglobulin) gene was amplified to verify cDNA quality. We added 1 μL cDNA to a final 25-μL PCR reaction volume containing 1× provided enzyme buffer, 2 mmol/L MgCl2, 200 μmol/L dNTPs, 400 nmol/L of each primer, and 1 U Taq DNA polymerase (Fermentas). The amplification conditions consisted of 35 cycles of 30 s at 95 °C, 30 s at the annealing temperature, and 1 min at 72 °C. The annealing temperature was 60 °C for GALNT1, 2, 4, 5, 6, 7, 9, and 15; 56 °C for GALNT3, 10, 11, 12, and 14; and 62 °C for GALNT13. The PCR products (15 μL) were analyzed by electrophoresis on 2% agarose gels and by direct visualization after ethidium bromide staining.
We performed several analyses to determine whether the expression levels of the GALNT genes of interest were significantly related to overall and disease-free survival of the patients (end points of cancer-specific death and relapse, respectively). Overall survival was defined as the time from diagnosis to the date of death or last follow-up; disease-free survival was defined as the time from diagnosis to the date of death or the first appearance of relapse. SAS software (SAS Institute) was used to construct Kaplan–Meier survival curves and to perform univariate and multivariable Cox proportional hazards analyses. To generate survival curves, we stratified patients into groups of GALNT gene expression (i.e., expression, +; lack of expression, −). Survival curves were compared by the log-rank test and judged significantly different at P values <0.05. We undertook univariate Cox proportional hazards regression analyses with SAS software and defined statistical significance as P values <0.05. To ascertain whether expression of the GALNT genes of interest were independent prognostic factors, we used Cox proportional hazards regression analyses to examine the joint effects of covariates. Patients were dichotomized by age into patients younger than 18 months and patients 18 months of age and older. Stage was dichotomized for the entire neuroblastoma cohort with respect to metastatic progression (i.e., stage 1, 2, and 3 vs stage 4. Stage 4S was included with stages 1, 2, and 3, because the prognosis for these patients is usually considered good. For the GALNT gene of interest, patients were categorized into groups as described above. All variables with P values <0.05 in univariate analyses were selected for inclusion in the multivariable model.
EXPRESSION OF THE GALNT GENE FAMILY IN HUMAN NEUROBLASTOMA CELL LINES
We investigated the production of GALNAC-Ts in the IGR-N-91 neuroblastoma experimental model, which offered the opportunity to compare neuroblastoma cell lines established from a primary tumor (PTX) and matched metastatic neuroblasts (BM). An RT-PCR assay was designed for 14 members of the GALNT gene family (GALNT1 to GALNT15 except GALNT8). In comparing the 2 cell sublines, we found that several GALNT genes were expressed in both, (PTX and BM neuroblasts expressed GALNT1, 2, 4, 6, 7, 12, and 14), but the expression of 3 genes (GALNT3, 5, and 10) were undetectable in these 2 cell lines. We found a marked difference between PTX and BM malignant neuroblasts in the expression GALNT9, 11, 13, and 15. The GALNT11 and GALNT13 genes were expressed in the BM cell line but not in the PTX cell line, and conversely, GALNT9 and GALNT15 genes were expressed in the PTX line but not in the BM cell line (Fig. 1). Consequently, we measured the expression of these genes in a panel of 6 neuroblastoma cell lines. GALNT11, 13, and 15 were found in almost all the cells lines and hence were broadly expressed. In sharp contrast, GALNT9 displayed a very restricted expression pattern, being found only in SK-N-AS and SK-N-BE (2) cell lines (Fig. 2). These results led us to choose GALNT9 gene for further evaluation of clinical samples.
GALNT9 EXPRESSION IN PRIMARY HUMAN NEUROBLASTOMA TUMORS
On the basis of the in vitro results, we used RT-PCR to assess GALNT9 expression in samples of primary neuroblastoma tumors from a tumor cohort from 122 patients (Table 1; see Table 1 in the online Data Supplement). We found GALNT9 expression in 77 (63%) of the 122 tumors. We then investigated whether the lack of GALNT9 expression was associated with various clinical features (Table 2). The relationship between disease stage and GALNT9 was not significant (P = 0.0596). GALNT9 was expressed not only in most patients with localized and 4S stages (47 of 66 patients, 71%) but also in 30 (54%) of 56 patients with stage 4 disease. GALNT9 was expressed in most patients with no MYCN amplification (66 of 96 patients, 69%) and in patients younger than 18 months at diagnosis (42 of 56 patients, 75%); however, GALNT9 expression occurred in only 42% of patients (11 of 26) with MYCN amplification and 53% of patients (35 of 66) ≥18 months. Thus, the lack of GALNT9 expression was associated with MYCN amplification (P = 0.0209) and age at diagnosis (P = 0.0147), indicating a link between lack of GALNT9 expression and the high-risk phenotype.
EXPRESSION OF GALNT9 AS A PROGNOSTIC MARKER OF PATIENT SURVIVAL
Statistical analyses were performed to determine the significance of GALNT9 expression as a prognostic factor for the entire cohort (n = 122). Disease outcome (survival and relapse events) was documented for all patients. The final follow-up date was February 2011, with a median follow-up period of 51 months. An event (death or relapse) was registered for 54 patients (44%), with death recorded for 42 patients (34%). A univariate Cox proportional hazards regression analysis showed that patients with a lack of GALNT9 expression in their tumors had an overall and disease-free survival rate significantly lower than patients whose tumors expressed that gene (Table 3). Consistent with the results of the univariate Cox regression analysis, the Kaplan–Meier survival analysis highlighted that patients with GALNT9 expression displayed a better prognosis than patients with no expression, in terms of both overall survival and disease-free survival (P < 0.0001, and P = 0.0004, respectively; Fig. 3A). Analyses with the multivariable Cox proportional hazards model assessed whether GALNT9 expression was an independent prognostic factor (Table 3) and found it to be an independent prognostic factor for overall survival (hazard ratio, 2.066; 95% CI, 1.078–3.963; P = 0.0290) but not for disease-free survival (hazard ratio, 1.446; 95% CI, 0.824–2.538; P = 0.1986).
GALNT9 EXPRESSION PREDICTS SURVIVAL FOR BOTH HIGH- AND LOW-RISK NEUROBLASTOMA PATIENTS
We first investigated the association of GALNT9 expression with survival in the group of high-risk patients (n = 58). Kaplan–Meier curves showed a significant association of GALNT9 expression with overall survival (P = 0.0391) but not with disease-free survival (P = 0.4178) (Fig. 3B).
Finally, we investigated the prognostic value of GALNT9 expression in low-risk patients (n = 64). Kaplan–Meier survival curves revealed that the clinical outcome (overall and disease-free survival) of patients with GALNT9 expression was strongly correlated with better survival than patients with no GALNT9 expression (P = 0.0001, and P = 0.0007, respectively) (Fig. 3C).
Altered glycosylation is a universal feature of cancer cells, and some glycan structures are well-known markers of tumor progression. Several studies have linked specific ganglioside changes in human neuroblastoma tumors to differences in biological behavior and clinical outcome (26, 27). We previously reported that GALNT13 is highly expressed (12 times higher) in metastatic neuroblasts than in primary tumors (19). The present study has shown that GALNT9 is expressed in PTX cells but not in BM cells in the IGR-N-91 model. This gene is also expressed in less aggressive S-type neuroblasts (SK-N-AS cell line), but not in N-type neuroblasts (SH-SY5Y, IMR-32, LAN-1, and LAN-5 cell lines). These findings could be of particular importance, because S-type cells have been reported to exhibit weaker invasiveness and metastatic properties, a lower growth capability in vivo (28), and a higher spontaneous apoptosis rate, compared with N-type cells (29).
GALNT9 expression was initially thought to be exclusive to the brain, with mRNA concentrations being more abundant in the cerebellum than in the cerebral cortex, frontal lobe, temporal lobe, and putamen (21). GALNT9, GALNT8, GALNT18, and GALNT19 constitute a subfamily that differs markedly in sequence from the other members of the GALNT family (30). Interestingly, the GALNAC-T8, -T9, and -T18 enzymes do not have catalytic activity toward classic peptide substrates, which are derived from mucin 1 (MUC1) variant MUC1/A, MUC5AC, and mono-GalNAc/Thr7th-EA2 (a glycopeptide containing 1 GalNAc in the threonine at the 7th position, named EA2) (31), suggesting that the GALNAC-T members of this subfamily may glycosylate very specific substrates. Functional analysis has shown that GALNAC-T9 is able to glycosylate a synthetic peptide (SDC284) derived from syndecan-3 (expressed in neuronal cells), but it possesses no activity for the transfer of GalNAc to 3 other peptides (SDC106, SDC155, and SDC165) derived from syndecan-3 (18). These results indicate that GALNT9 displays a very restricted specificity for this acceptor substrate, in contrast to that observed for GALNAC-T13. These functional differences between GALNAC-T9 and GALNAC-T13, and the fact that the genes encoding these 2 enzymes display an opposing expression pattern, lead to the hypothesis that they could be of biological relevance in neuroblastoma behavior.
Our analysis of 122 neuroblastoma primary tumors revealed an association between GALNT9 gene expression and survival. For the high-risk patients, GALNT9 expression was correlated with overall survival but not with disease-free survival. In contrast, the survival analysis of the low-risk neuroblastoma subgroups clearly highlighted the prognostic value of GALNT9 expression for both overall and disease-free survival of patients. In other words, patients with tumors having GALNT9 expression displayed higher survival rates than patients with no expression. More specifically, within the low-risk neuroblastoma subgroup, a relapse can be predicted for tumors that lack GALNT9 expression, whereas a positive outcome can be predicted for tumors with GALNT9 expression. Therefore, assessment of GALNT9 gene expression in these patients could provide an additional method of biological classification of tumors. Expression of the GALNT9 gene therefore appears to have predictive value beyond that for risk stratification based on age, disease stage, and MYCN amplification, and its expression in particular may constitute a marker of relapse for neuroblastomas classified as low risk. Very recently, nonrandom chromosomal abnormalities (numerical and segmental changes) have been shown to improve the standard risk stratification (32, 33). In this regard, studies of GALNT9 gene expression with independent tumor cohorts are necessary to refine the forthcoming risk-stratification algorithm (age, stage, MYCN gene status, and genomic changes).
The low expression of the genes encoding some GALNAC-T isoenzymes has been associated with more aggressive tumors. GALNAC-T6 is reportedly produced in fibroblasts, placenta, and normal pancreas cells. The absence of this enzyme in pancreatic tumors was correlated with a poor clinical outcome (34). Similarly, a loss of GALNAC-T3 was correlated with a higher metastatic potential in a mouse colon cancer model (35), and strong GALNAC-T3 protein production, as demonstrated in human colorectal carcinomas with immunochemistry staining, significantly enhanced the likelihood of patient survival (36). One possible explanation for neuroblastoma is that GALNT9 expression is a marker for more-mature stages of neuroblastic tumor cells, which are associated with less aggressive tumor cells. In neuroblastoma, S-type cells are more susceptible to apolipoprotein 2 ligand/tumor necrosis factor–related apoptosis-inducing ligand (Apo2L/TRAIL)-mediated apoptosis than N-type cells, owing to their expression of caspase-8 (29). O-glycosylation pathways have been found to play a role in the regulation of cell growth by modifying the rate of apoptosis of cancer cells (37, 38).
In conclusion, our data support the use of GALNT9 as a prognostic marker that is independent of the conventional risk stratification, and it may be a valuable marker that may be helpful in guiding specific therapy in low-risk patients.
We thank French surgeons and oncologists for providing material and clinical reports, and Sabrina Cantais for technical assistance.
↵† These authors contributed equally to this work.
↵7 Human genes:
- v-myc myelocytomatosis viral related oncogene, neuroblastoma derived (avian);
- UDP-N-acetyl-α-d-galactosamine:polypeptide N-acetylgalactosaminyltransferase 3 (GALNAC-T3);
↵8 Nonstandard abbreviations:
- UDP-GalNAc:polypeptide N-acetylgalactosaminyltransferase;
- reverse-transcription PCR;
- primary-tumor xenograft;
- bone marrow metastasis;
- deoxynucleoside triphosphate;
- apolipoprotein 2 ligand/tumor necrosis factor–related apoptosis-inducing ligand.
Author Contributions: All authors confirmed they have contributed to the intellectual content of this paper and have met the following 3 requirements: (a) significant contributions to the conception and design, acquisition of data, or analysis and interpretation of data; (b) drafting or revising the article for intellectual content; and (c) final approval of the published article.
Authors' Disclosures or Potential Conflicts of Interest: Upon manuscript submission, all authors completed the author disclosure form. Disclosures and/or potential conflicts of interest:
Employment or Leadership: None declared.
Consultant or Advisory Role: None declared.
Stock Ownership: None declared.
Honoraria: None declared.
Research Funding: Comisión Honoraria de Lucha Contra el Cáncer, Montevideo, Uruguay, Programa Grupos de Investigación CSIC, Montevideo, Uruguay, and Société Française des Cancers de l'Enfant, La Ligue contre le Cancer Comité Oise.
Expert Testimony: None declared.
Patents: None declared.
Role of Sponsor: The funding organizations played no role in the design of study, choice of enrolled patients, review and interpretation of data, or preparation or approval of manuscript.
- Received for publication July 6, 2012.
- Accepted for publication October 15, 2012.
- © 2012 The American Association for Clinical Chemistry