Glyoxalase 1 expression is associated with an unfavorable prognosis of oropharyngeal squamous cell carcinoma

Tumor specimens for this retrospective study were obtained from patients with primary OPSCC, who have given informed consent to participate and were treated at the University Hospital Heidelberg between 1990 and 2008. Formaldehyde fixed and paraffin embedded tissue specimens were provided by the tissue bank of the National Center for Tumor Disease (Institute of Pathology, University Hospital Heidelberg) after approval by the local institutional review board (ethic vote: 176/2002 and 206/2005). The study was performed according to the ethical standards of the Declaration of Helsinki. For all tumor samples, clinical and follow-up data were available from the Department of Otolaryngology, Head and Neck Surgery at the University Hospital Heidelberg and are listed in Additional file 1: Table S1. HPV16 DNA and viral transcript status was determined previously [20] and generation of tissue microarrays has been described elsewhere [21].

Tissue microarrays were assembled with independent cores from distinct areas of a formalin-fixed and paraffin embedded pre-treatment tumor tissue and for most tumors at least two independent probes were available. Tissue microarrays were incubated with anti-AP or anti-GLO1 antibodies that are listed in Additional file 2: Table S2. Immunostaining was visualized with the TSA Amplification Kit (Perkin Elmer, Rodgau, Germany) and DAB peroxidase substrate (Vector Laboratories, Burlingame, USA) according to the manufacturers instructions. Counterstaining was done by hematoxylin to visualize tissue integrity. Stained tissue microarrays were scanned using the Nanozoomer HT Scan System (Hamamatsu Photonics, Japan). Protein expression was evaluated by three independent observers using the NDP Viewer software (version 1.1.27). For AP the staining intensity (score 1 = no, score 2 = low, score 3 = moderate, score 4 = high) was evaluated. For GLO1 the relative amount of positive tumor cells (score 1 = no positive cell, score 2 ≤ 33%, 33% > score 3 ≤ 66%, score 4 > 66%) and the staining intensity (score 1 = no, score 2 = low, score 3 = moderate, score 4 = high) was evaluated (Additional file 3: Table S3). Both scores were significantly correlated (Pearson correlation coefficient 0.696, p < 0.005 and Spearman correlation coefficient 0.586, p < 0.005) and multiplied to calculate the final immunoreactivity score (IRS, range 1–16). The cut-off value for further analysis was: GLO1^high > 8 and GLO1^low ≤ 8. The intracellular staining of GLO1 was evaluated as predominant nuclear (GLO1^nuc) versus cytoplasmic (GLO1^cyt) (Additional file 4: Table S4).

Human cell lines (FaDu and Cal27) were purchased from ATCC (http://​www.​lgcstandards-atcc.​org; catalog numbers: HTB-43™ and CRL-2095™) and were maintained in http://​www.​sigmaaldrich.​com (Sigma, Germany) supplemented with 10% fetal bovine serum (Invitrogen, Germany), 2 mM L-Glutamine (Invitrogen, Germany) and 50 μg/ml Penicillin-Streptomycin (Invitrogen, Germany) in a humidified atmosphere of 6% CO2 at 37 °C. Authentication of both cell lines was confirmed by the Multiplex Human Cell Line Authentication Test (Multiplexion, Germany). For Western blot analysis 1 × 10⁶ FaDu or Cal27 cells were seeded on 96 mm TC plates and cultured over night. Cells were treated with indicated concentrations of MG and 24 h following administration cells were harvested for protein lysates. MG was synthesized by acid hydrolysis as described [22], and the concentration of the stock solution was determined by derivatization with aminoguanidine [23]. The cell permeable GLO1 inhibitor, S-p-bromobenzylglutathione cyclopentyl diester (BrBzGSHCp₂) was prepared and characterized as described [24, 25]. For colony formation assays 100 and 300 FaDu cells or 300 and 1000 Cal27 cells were seeded per 6-well plate, respectively. Cells were treated with the indicated concentration of MG or the GLO1 inhibitor every second day. 10 days upon seeding cells were PFA-fixed and the amount of colonies was determined after crystal violet staining as described in [26]. The survival fraction was computed according to [27].

Whole cell protein lysate was extracted using RIPA (Radioimmunoprecipitation assay) buffer [28] and protease and phosphatase inhibitor cocktails (Sigma-Aldrich). 20 μg of denatured protein were separated by Sodiumdodecylsulfate-polyacrylamide gel electrophoresis (SDS-PAGE) and transferred to polyvinyl difluoride (PVDF) membranes (Millipore, Germany). After blocking with 5% milk (Roth, Germany), membranes were incubated with primary and horseradish peroxidase coupled-secondary antibodies, which are listed in Supplementary Table S2. Membranes were incubated in enhanced chemiluminiscence solution (Thermo Scientific, Germany) and developed with an E.O.S. developer (Agfa, Germany).

The concentrations of MG in supernatant of cultured cells were determined by derivatization with 1,2-diamino-4,5-dimethoxybenzene and HPLC of the quinoxaline adduct by fluorescence detection as described elsewhere [22, 29].

Tissue microarrays were analyzed by immunohistochemistry with an anti-AP antibody as a surrogate marker for MG accumulation in tumor cells of primary OPSCC. In parallel, GLO1 protein levels were analyzed by immunohistochemistry on serial sections of tissue microarrays. Data on AP and GLO1 immunoreactivity were available for tumor specimens of 134 patients, and revealed no AP staining in 3.7% (n = 5), low staining pattern in 20.9% (n = 28), moderate staining pattern in 31.4% (n = 42), and a high staining pattern in 44% (n = 59) (Fig. 1A, C, E and G). A heterogeneous staining pattern was also evident for GLO1 considering the staining intensity as well as the relative amount of positive tumor cells (Fig. 1B, D, F and H). It is worth noting that in addition to the expected positive staining in the cytoplasm of tumor cells (Fig. 1F), we also detected a prominent nuclear GLO1 staining in a substantial amount of OPSCC samples (Fig. 1H). Comparison of the relative GLO1 immunoreactivity score in subgroups of OPSCC patients with no, low, moderate or high AP staining demonstrated a significant positive correlation (Fig. 1I).

The data presented so far suggested that induced GLO1 expression is a compensatory mechanism of OPSCC tumor cells to counteract and survive the accumulation of MG. To further support this assumption FaDu and Cal27, two well-established head and neck cancer cell lines, were cultured in the presence of increasing amounts of MG. In both cell lines, administration of exogenous MG increased GLO1 protein levels as determined by Western blot analysis (Fig. 2A). It is worth noting that GLO1 levels were lower in Cal27 as compared to FaDu cells under control conditions as well as the presence of exogenous MG.

Next, the cytotoxic profile of MG treatment on tumor cell viability was analyzed by a colony-forming assay. No major difference between both cell lines was found at lower MG concentrations (≤ 40 μM, Additional file 5: Fig. S1). However, at higher MG concentrations FaDu cells displayed an improved survival as compared to Cal27 cells, which was statistical significant at a MG concentration of 160 μM (p ≤ 0.049, Fig. 2B). These data indicated a better tolerance of tumor cells with higher basal or inducible GLO1 expression under conditions of MG accumulation.

To assess the consequence of GLO1 inhibition on tumor cell survival we cultured both cell lines in the presence of the cell permeable inhibitor S-p-bromobenzylglutathione cyclopentyl diester [3]. Quantification of MG levels in cell culture supernatant after treatment with the GLO1 inhibitor revealed a concentration dependent increase in FaDu but not Cal27 cells (Additional 6: Fig. S2). However, both cell lines showed a comparable decrease in the survival fraction with increasing amount of the GLO1 inhibitor (Fig. 2C-D). These data strongly suggested that the cytotoxic effect of GLO1 inhibition does not exclusively rely on an accumulation of MG levels.

To evaluate the clinical relevance of GLO1 staining patterns in the cohort of OPSCC patients, subgroups with a low (GLO1^low, n = 67) or a high immunoreactivity score (GLO1^high, n = 89) were identified. Furthermore, we considered a cytoplasmic (GLO1^cyt, n = 94) or a predominant nuclear staining (GLO1^nuc, n = 43) in the subgroup of OPSCC patients with a positive GLO1 staining. A GLO1^high staining was significantly correlated with shorter progression-free (PFS) and disease-specific survival (DSS), while nuclear GLO1 staining revealed a trend towards shorter PFS and was significantly correlated with shorter DSS (Fig. 3A-D). These findings strongly indicated that a nuclear GLO1 accumulation in tumor cells with a high expression serves as a risk factor for unfavorable prognosis of OPSCC patients. This assumption was further assessed by a combinatorial subgroup analysis taking into account both features (GLO1^high/nuc) in comparison to all other staining patterns (GLO1^others), and confirmed a highly significant shorter PFS and DSS (Fig. 3E-F). Moreover, a high and nuclear GLO1 staining was significantly associated with a larger tumor size (Table 1), but not with any other patient characteristic tested (e.g. gender, age, TNM status, pathological grading, clinical staging, alcohol and tobacco consumption or HPV status). Next, multivariate Cox proportional hazard model analysis confirmed that a GLO1^high/nuc staining pattern served as independent risk factors for unfavorable PFS and DSS (Table 2), taking into consideration relevant prognostic risk factors based on univariate analysis (Additional 7: Table S5).

Finally we addressed the question, whether the prognostic value of GLO1 staining patterns correlates with the mode of treatment and conducted subgroup analysis by univariate Cox regression analysis. While the prognostic value of nuclear GLO1 staining was largely independent of treatment with (n = 109) or without surgery (n = 46), a high GLO1 immunoreactivity score was significantly correlated with an unfavorable PFS after surgery (with or without adjuvant radio- or radiochemotherapy, Additional file 8: Fig. S3A). A similar trend was also found for DSS (Additional file 9: Fig. S3B). In contrast, a high GLO1 immunoreactivity score was significantly correlated with a poor clinical outcome in the absence of chemotherapy (n = 108), which was not observed for patients with chemotherapy (n = 47, Additional file 10: Fig. S4). All GLO1 staining patterns (high or nuclear GLO1 staining and their combination) were correlated with unfavorable PFS and DSS after radiotherapy (n = 131, adjuvant and definitive), but not for surgery only (n = 24, Additional Fig. S5). However, it is worth noting that the amount of cases without radiotherapy was limited in this retrospective cohort.