GSTM1, GSTT1 and GSTP1 Ile105Val polymorphisms in outcomes of head and neck squamous cell carcinoma patients treated with cisplatin chemoradiation

Cisplatin (CDDP) combined with radiotherapy (RT) is employed in head and neck squamous cell carcinoma (HNSCC) with variable toxicities and clinical response. Glutathione S-transferases (GSTs) participate in CDDP excretion from cells, and genes encoding GSTs, GSTM1, GSTT1and GSTP1, are polymorphic in humans. This prospective study aimed to evaluate the roles of GSTM1, GSTT1, and GSTP1 Ile105Val polymorphisms in outcomes of HNSCC patients treated with CDDP chemoradiation. Ninety patients were genotyped by multiplex PCR. Urinary CDDP measurements were performed by HPLC. Treatment side effects and response were analysed by conventional criteria. Patients with GSTT1 genes showed 7.23- and 5.37-fold higher likelihood of presenting vomiting and ototoxicity, lower glomerular filtration rate (GFR), and lower elimination of CDDP in urine relative to patients with deleted genes. Patients harbouring the GSTP1 IleVal or ValVal genotypes showed 4.28-fold higher likelihood of presenting grade 2 or 3 vomiting and lower GFR with treatment than those harbouring the IleIle genotype. In multivariate Cox analysis, patients with the GSTP1 105ValVal genotype had 3.87 more chance of presenting disease progression than those with the IleIle or IleVal genotype (p < 0.01). Our findings provide preliminary evidence that inherited abnormalities in CDDP metabolism, related to GSTT1 and GSTP1 Ile105Val polymorphisms, alter outcomes of HNSCC patients treated with CDDP and RT.


Results
Patients and laboratory characteristics. A total of 90 HNSCC patients with a median age of 56 years were enrolled in this study. The majority of the enrolled subjects were males, drinkers and smokers, and presented moderately differentiated and advanced tumours. Nearly two-thirds of the cases had tumours in the pharynx. All analysed cases tested negative for human papillomavirus (HPV) type 16. Most patients showed medium or high adherence to anti-emetics (Table 1).
All patients were homogenously treated with RT at a total dose of 75 Gy and 80-100 mg/m 2 of CDDP as initial dose as previously described (100 mg/m 2 of CDDP was administered to patients with Karnofsky Performance Scale (KPS) 80-100% and without comorbidities, and patients with KPS 60-70% without comorbidities or KPS higher than 70% with comorbidities received CDDP at dose of 80 mg/m 2 ). As recommended by the institutional protocol, 13 patients who presented adverse side effects of grades 3 or 4 received CDDP at lower dose (50-75 mg/m 2 ) in further administrations or had suspension of CDDP. A total of 68 patients (75.5%) received three infusions of CDDP, while 22 patients (24.5%) received only two CDDP infusions because of hematologic or renal toxicities. CDDP was measured in the urine samples of 43 patients after three CDDP infusions, and cumulative CCDP was found to be higher in urine samples collected in the first 12 h than in samples collected between 12 to 24 h (p = 0.001) and between 24 to 48 h (p = 0.001) ( Table 1) (Fig. 1A). Similar urinary CDDP levels were observed in urine samples collected between 0 to 12 h (76.91 ± 50.26, 81.97 ± 68.74, and 56.39 ± 40.10 µg CDDP/mg creatinine; p = 0.06), 12 to 24 h (8.34 ± 7.02, 9.90 ± 7.82, and 7.59 ± 3.38 µg CDDP/mg creatinine, p = 0.10), and 24 to 48 h (5.35 ± 4.11, 7.83 ± 8.85, and 6.96 ± 4.10 µg CDDP/mg creatinine; p = 0.10) after the first, second, and third CDDP infusions, respectively.
Nearly four fifths and two thirds of available cases presented nausea and vomiting, respectively. The majority of patients showed hematologic toxicities; a quarter of the cases presented nephrotoxicity, and half of the patients showed ototoxicity (Table 2). Among the patients with available data on response rate (RR), 15 (20.5%) patients achieved complete response, 53 (72.6%) showed partial response (PR), and 5 (6.9%) showed stable disease (SD) status. Similar gastrointestinal and haematological toxicity (Table S1 Supplement), nephrotoxicity, ototoxicity, and response to therapy (Table S2 Supplement) were observed in patients stratified based on clinicopathological aspects and cumulative dose of intravenous CDDP.
All cases were followed-up for a median period of 21 months (range: 3.0-74). The cumulative probabilities of event-free survival (EFS) and overall survival (OS) at 24 months follow-up were 35.0% and 40.0%, respectively. On the date of analysis (December 2017), 23 patients were alive, of which 6 had HNSCC and 17 did not have HNSCC. A total of 67 patients died, of which 59 mortalities were associated with tumour effects and eight were attributed to unrelated causes.
Patients harbouring the GSTT1 genes were 7.23-and 5.37-fold more likely to present vomiting and ototoxicity of any grade (G1 to G3), respectively, relative to patients with the GSTT1 null genotype. Patients with the GSTP1 IleVal or ValVal genotype were 4.28-fold more likely to present vomiting (G2 or G3) than those carrying the IleIle genotype. Patients stratified based on nausea and nephrotoxicity symptoms showed similar frequencies of the GSTM1, GSTT1, and GSTP1 Ile105Val polymorphisms (Table 3). In addition, patients harbouring different genotypes of the analysed polymorphisms showed no significant differences in haematological toxicity and response to therapy (Table S3 Supplement).
Reduction in glomerular filtration rate (GFR) after chemoradiation was found to be more pronounced in patients harbouring the GSTT1 genes and in patients with GSTP1 IleVal or ValVal genotype relative to patients with the GSTT1 null genotype and IleIle genotype, respectively ( Table 4). The mean cumulative urinary CDDP level (in the three cycles) was higher in patients with the GSTT1 null genotype than in those carrying the GSTT1 gene (429.58 ± 116.24 vs. 253.42 ± 95.20 g CDDP/mg creatinine, p = 0.04; power test = 83.4%) (Fig. 1B). A unique patient with combined GSTM1 null, GSTT1 null, and GSTP1 IleIle genotypes showed the highest reported total CDDP elimination rate (545.42 µg CDDP/mg creatinine).
(26.3% vs. 83.3%, p = 0.02) and patients with the GSTP1 ValVal genotype (0.0% vs. 31.7%, p = 0.047) had shorter event-free survival (EFS) than those with localised tumours and those harbouring the GSTP1 IleVal or ValVal genotype (Fig. 1C). Only patients with advanced tumours showed shorter OS (83.3% vs. 21.7%, p = 0.009) (Kaplan-Meier estimates). In univariate Cox analysis, patients with advanced tumours showed 7.50-and 8.88-fold higher likelihood of disease progression and mortality than those with localised tumours, respectively. Results of multivariate Cox analysis showed that patients with advanced tumours and patients harbouring the GSTP1 105ValVal genotype had 8.90-and 3.87-fold higher likelihood of presenting disease progression than those with localised tumours and harbouring the GSTP1 IleIle or Ile Val genotype, respectively. The significance of survival differences were confirmed by the bootstrapping method (Table 5).

Discussion
Our findings on the clinicopathological aspects of HNSSC patients, toxicity and RR to chemoradiation, and survival were similar to those described in previous studies 18,19 . A history of tobacco and alcohol consumption but the low prevalence of HPV infection indicate that history of tobacco and alcohol consumption were in fact the most important factors influencing tumour development in the analysed patients.
The presence of GSTT1 and the GSTP1 IleVal or ValVal genotype were found to be associated with a higher likelihood of presenting vomiting with chemoradiation, as previously suggested 20 . In a previous publication, we www.nature.com/scientificreports www.nature.com/scientificreports/ analysed almost the same patients enrolled in the current study (n = 88) and focused on the roles of GSTM1, GSTT1, GSTP1, and other genetic polymorphisms in genes involved in the repair of DNA damage induced by CDDP and apoptosis of cells with CDDP-induced lesions, only in severity of vomiting presented by HNSCC patients treated with CDDP chemoradiation 14 . Our previous findings indicated that the GSTP1 IleVal or ValVal genotype, but not the presence of GSTT1, was associated with vomiting; in the current study, vomiting was strongly associated with the presence of both GSTT1 and the GSTP1 IleVal or ValVal genotype. The contrasting www.nature.com/scientificreports www.nature.com/scientificreports/ results between the two studies can be attributed to the differences in patient stratification based on the severity of vomiting grade. In the previous publication, patients were stratified based on grade 0 or 1 vs. grade 2 or 3, whereas patients in the present study (n = 90) were stratified based on grade 0 vs. grades 1 to 3.
GSTT1 null patients were expected to show more severe vomiting than patients harbouring GSTT1; however, opposite results were obtained in the present study. Nutrient deficiency in patients with malignancies was postulated to lead to a lack of GSH, thereby causing impaired detoxification of free radicals of CDDP by GSTT1 and more severe toxicity 13 . In fact, reduction in body mass was found to be more predominantly observed in HNSCC patients receiving CDDP and RT treatment 21 , which could have modified the effects of the GSTT1 gene.  www.nature.com/scientificreports www.nature.com/scientificreports/ CDDP is well known to cause vomiting by inducing DNA damage in epithelial enterocromaffin cells of the intestine, thereby leading to serotonin release and stimulation of the chemoreceptor trigger zone and vomiting centre 22 . On the other hand, in addition to its role in CDDP detoxification, GSTP1 regulates JNK signalling pathways by forming a complex with c-Jun-JNK, which inactivates JNK signalling and inhibits apoptosis 23 . Thus, increased apoptosis of CDDP-damaged enterochromaffin cells of the intestines, which is associated with the less active "Val" allele of the GSTP1 Ile105Val polymorphism, may have influenced the severity of vomiting in the HNSCC patients.
In addition, the presence of GSTT1 was associated with ototoxicity in the patients. Oldenburg et al. 24 demonstrated that the GSTM1 null genotype conferred protection against hearing impairment in testicular cancer patients treated with CDDP. Talach et al. 13 reported that early treatment with CDDP induced ototoxicity in adults with various tumours types, including HNSCC, and two copies of the GSTT1 gene. The GSTP1 "Val" allele was associated with ototoxicity in patients with medulloblastoma who were treated with CDDP-based therapy 25 .
CDDP induces ototoxicity through the generation of free radicals and subsequently activates cell death pathways, such as the c-Jun N terminal kinase (JNK) and p38 mitogen activated protein kinase (MAPK) pathways, which in turn induce hair cell apoptosis and hearing loss 26 . GSTs represent the major enzyme family involved in the excretion of CDDP, which protects cells from the deleterious effects of free radicals 9 . Again, we expected more severe ototoxicity in GSTT1 null patients than in patients harbouring GSTT1; however, contrasting results were obtained in the present study. Talach et al. 13 observed that patients with testicular tumours tended to have increased ototoxicity and presence of GSTT1. They postulated that nutrient deficiency with lack of GSH production and impaired detoxification of free radicals of CDDP in the sensory inner ear cells consequently increased ototoxicity. The same mechanisms are likely to occur in HNSCC patients in the present study.
Our current findings indicated that the GSTM1, GSTT1, and GSTP1 Ile105Val polymorphisms did not alter myelotoxicity and RR to CDDP and RT. Cabelguenne et al. 9 found no associations of the GSTM1, GSTT1, and GSTP1 Ile105Val polymorphisms with RR in HNSCC patients treated with low doses of CDDP (25 mg/m²) and 5-FU. We observed pronounced myelotoxicity and a durable complete response in one HNSCC patient treated with both CDDP (100 mg/m²) and RT, which was attributed to inherited deficiency in CDDP detoxification (GSTT1 null genotype), deficiency in repair of CDDP cellular damage (MSH3 1045ThrThr), and increased apoptosis induction during CDDP cellular damage (GSTP1 105IleIle) 12 .
Patients harbouring the GSTT1 and GSTP1 IleVal or ValVal genotype showed more pronounced reduction in 51 Cr-EDTA GFR after chemoradiation than those with the remaining genotypes. Khrunin et al. 27 observed that patients with epithelial ovarian cancer and harbouring the GSTT1 null genotype who were treated with CDDP (100 mg/m 2 ) plus cyclophosphamide were 3.31-fold more likely to develop nephrotoxicity.
Given its low molecular weight, CDDP is conjugated with reduced glutathione (GSH) in the liver. CDDP-GSH is freely filtered in the glomerulus and is completely recovered in the urine. CDDP-GSH reaches high concentrations in the proximal tubular cells of the internal renal cortex and the outer medullary layer. The mechanism of CDDP-induced tubular damage is complex and involves several mechanisms, such as CDDP accumulation-mediated membrane transport, conversion to nephrotoxin, DNA damage, mitochondrial dysfunction, oxidative stress, inflammatory response, activation of transducers and intracellular messengers, and activation of apoptotic pathways 28 .  www.nature.com/scientificreports www.nature.com/scientificreports/ To our knowledge, the roles of GSTT1 and GSTP1 in renal damage by CDDP remain unknown. Thus, the high CDDP-GSH concentrations in proximal tubule cells, associated with the GSTT1 genes, are likely to influence the severity of renal dysfunction in our patients. Furthermore, patients harbouring the GSTP1 IleVal or ValVal genotype showed more pronounced renal dysfunction. These results are unexpected because the protein encoded by the "Val" allele is known to be less efficient in CDDP-GSH production than that encoded by the "Ile" allele 11 . Again, this association could be attributed to effects of the protein encoded by the "Val" allele, which induces apoptosis in proximal tubule cells 23 .
Nearly 90% of CDDP excretion in our patients was measured within the first 24 h after infusion. Patients with the different GSTT1 genotypes showed variable CDDP levels, suggesting that gene polymorphisms influence the kinetics of CDDP elimination. Notably, pharmacokinetic studies usually analyse plasma samples obtained from patients receiving therapy. However, in the present study, sequential blood plasma collection cannot be obtained from outpatients. Therefore, we determined CDDP levels in urine for this pharmacokinetic analysis. The HPLC method used was well established 29 and considered essential to pharmacokinetics and pharmacodynamics studies 30 . Lanjwani et al. 29 postulated a correlation between urine and plasma CCDP levels and that both parameters could be used to pharmacokinetics studies of the drug. To the best of our knowledge, our study is the first to analyse the effects of GST polymorphisms on CDDP pharmacokinetics in HNSCC patients. Results of a unique study  Table 5. Association of clinicopathological characteristics and GSTM1, GSTT1 and GSTP1 Ile105Val genotypes with survival of head and neck squamous cell carcinoma patients treated with chemoradiotherapy in univariate and multivariate Cox analyses. (n) number of patients; (HR) hazard ratio; (CI) confidence interval; (NA) Not aplicable. Multivariate analysis was adjusted by tumour stage and GSTP1 Ile105Val genotypes. * pbootstrap = 0.03; ** pbootstrap = 0.01, *** pbootstrap = 0.02, **** pbootstrap = 0.001. Significant differences between groups are presented in bold letters.
Surprisingly, patients with the GSTT1 null genotype showed higher cumulative urinary CDDP levels compared to patients with the GSTT1 genotype. In this case, the deficiency in an individual GST isoenzyme could be compensated for by other isoforms 32 . Consistent with the above hypothesis, the only patient with GSTM1 null, GSTT1 null, and GSTP1 IleIle genotypes showed the highest total CDDP elimination rate.
Finally, our findings indicated that patients carrying the GSTP1 Ile105Val polymorphism had altered EFS. There were no associations between the GSTM1 and GSTT1 genotypes and survival in early-stage radiation-treated HNSCC patients 17 . In addition, disease-specific survival was observed in patients with SCC of oral cavity and harbouring GSTM1 who were treated with radiotherapy 16 . Shorter survival rates were observed in GSTM1 null patients with SCC of oral cavity who were treated with CDDP-based chemotherapy 15 . To the best of our knowledge, no published studies have focused on the role of the GSTP1 Ile105Val polymorphism in survival of HNSCC patients treated with CDDP and RT. The GSTP1 105IleIle genotype was previously associated with higher PFS in epithelial ovarian cancer patients treated with CDDP (100 mg/m²) and cyclophosphamide 27 and oesophageal cancer patients treated with CDDP and 5-FU/paclitaxel 33 . The observed association between GSTP1 ValVal genotype and shorter EFS in the current study was based on a small sample size and cannot be explained by direct drug detoxification by GSTP1, which will lead to higher survival. Free radicals produced by CDDP that are not detoxified due to the lack of GSTP1 are likely to induce new mutations in residual tumour cells, thereby facilitating survival and proliferation of the tumour cells 34 .
In summary, our study is the first to provide preliminary evidence that inherited GSTT1 and GSTP1 Ile105Val polymorphisms can alter gastrointestinal status, nephrotoxicity, ototoxicity, pharmacokinetics, and survival in HNSCC patients treated with CDDP chemoradiation. In this context, GST genotypes can be used as instruments for assessing toxicity of CDDP-based therapies by selecting HNSCC patients for the use of specific anti-emetics and protective renal and auditory system agents, thereby saving patients from the side effects of chemotherapy 14,35,36 and reducing treatment costs 37 . However, we are aware that the number of patients enrolled in study was not large, and that further larger studies and functional analyses of relevant polymorphisms are required to confirm the roles of these GSTs in disease.

Material and Methods
Subjects and clinical variables. In this prospective study, HNSCC patients diagnosed at the Clinical Oncology Service of the General Hospital of University of Campinas between June 2011 and February 2014 were selected to undergo CCDP chemoradiation as definitive treatment due to locoregional unresectable tumour, refusal of surgery related to expected functional, or anatomic sequels or an organ preservation protocol. The following exclusion criteria were considered: (1) refusal to participate in the study; (2) low KPS score; (3) renal dysfunction; (4) previous hearing dysfunction; and/or (5) other therapeutic protocol. The present study was conducted according to the Declaration of Helsinki and was approved by the Ethics Committees of the University of Campinas (n° 274/2011; CAAE: 0218.0.146.000-11). All patients provided written informed consent after they were informed about the study and any associated risks.
Data relating to age, gender, tobacco and alcohol consumption, hematologic and biochemistry exams, tumour location, histological grade, and stage were obtained using specific questionnaires or patient charts. Subjects were classified according to smoking 38 and drinking habits 39 as previously reported. Tumours were diagnosed based on standard criteria 40 and staged based on criteria specified by the American Joint Committee on Cancer 41 . HPV type 16 was investigated in tumour fragments by immunohistochemistry and in situ hybridization as previously reported 42 .
Toxicities (nausea, vomiting, hematologic, nephrotoxicity, and ototoxicity) were assessed using data on adverse effects, hematologic exams, GFR measured with EDTA labelled with chrome 37 ( 51 Cr-EDTA GFR), and audiometric tests performed before and after chemoradiotherapy. Toxicities were evaluated according to the National Cancer Institute based on common terminology criteria for adverse events version 4.0 (CTCAE) 43 . For each patient, the worst grade for each toxicity was included in analysis.
Patients were homogeneously treated with CDDP chemoradiation according to the institutional protocol 42 . RR to treatment was assessed using the Response Evaluation Criteria in Solid Tumors (RECIST) guidelines version 1.1 44 . Surgical tumour resection was indicated to patients with good clinical condition and partial response or tumour relapse.
On each day of CDDP infusion, patients received hydration (3 L of 0.9% saline solution), 125 mL of 20% mannitol as diuretic, 20 mL of 19.1% potassium chloride and 10 mL of magnesium sulphate as electrolytes, and 20 mg of dexamethasone plus 24 mg of ondansetron as prophylaxis to acute emesis. Patients were orally administered with 8 mg of dexamethasone (every 12 h) and 10 mg of metoclopramide (every 6 h) for three additional days 42 . Anti-emetic adherence was classified as previously described 42 .
Patient follow-up was performed in three-month intervals. The end of the follow-up period was September 2018.
DNA extraction and genotyping. Genomic DNA of all subjects was isolated from 5 mL of peripheral blood sample following the proteinase K and lithium chloride method 45 . GSTM1 and GSTT1 genotypes were amplified from genomic DNA by the multiplex-polymerase chain reaction (multiplex-PCR) 46 . GSTP1 Ile105Val genotypes were identified by PCR and enzymatic digestion 47 . Positive and negative controls were used in all genotyping reactions. The amount of 15% of genotype determinations was carried out twice in independent experiments and showed 100% of concordance.
Urinary CCDP excretion kinetics. Urine  www.nature.com/scientificreports www.nature.com/scientificreports/ CDDP derivatization and extraction from patient urine and a sample of standard urine (without CDDP) were conducted as previously described 49 . Nickel chloride and chloroform were obtained from Merck, Darmstadt, Germany, and diethyldithiocarbamate was obtained from Sigma-Aldrich, India. Samples were analysed using a high-performance liquid chromatography (HPLC) Separation Module system with dual absorbance detector (UV-visible detector, wavelength 254 nm) (Waters 2487, Milford, MA, USA) and a Hypersil ODS C18 column (150 mm × 4 mm with a particle size of 4 μm (Thermo, Waltham, MA, USA). The HPLC conditions and the mobile phase were set as described by Lopez-Flores et al. 50 . Sample measurement and standard curves were performed in duplicate. Urinary CDDP was normalized with urinary creatinine and measured using akinetic kit (Creatinine Laborclin kit, Parana, Brazil). The final concentration of urinary CDDP was calculated as the sum of all measurements obtained after each CDDP administration. Statistical analysis. The HWE was tested with chi-square (χ 2 ) statistic for the goodness-to-fit. Differences between groups were analysed by χ 2 or Fisher's exact test. A logistic regression model was generated to obtain odds ratios (OR) values to verify associations between genotypes, nausea, vomiting, nephrotoxicity, ototoxicity, haematological toxicity, and RR. Comparisons between GFR before and after CCDP treatment were evaluated by two-way ANOVA. Student's t-test was performed to determine associations between average CDDP urinary elimination rate in each collection period (0-12, 12-24, and 24-48 h) and to determine significant associations of the total CDDP elimination rates with genotypes. The statistical power of a test was calculated using the Researcher's Toolkit 51 .
EFS and OS were calculated for each participant from the date of diagnosis until the date of tumour progression, tumour relapse, death attributed to tumour effects, or last follow-up and from the date of diagnosis until the date of death by any cause or last follow-up contact, respectively. Kaplan-Meier method was used to plot EFS and OS curves, and log-rank test was conducted to determine significant differences between curves. Multivariate Cox regression was used to estimate hazard ratios (HRs) adjusted for potential possible discrepancies in clinical aspects (p ≤ 0.10 in univariate Cox regression). Bootstrapping (n = 1,000) based on repeatedly random sampling was applied to ensure the stability of the model by applying the bias-corrected and accelerated method.
P-values were two-sided and considered statistically significant at p < 0.05. Statistical analyses were conducted using SPSS 21.0 software (IBM Corporation, Armonk, NY, USA).