Prognostic Value of p53, Glutathione S-Transferase π, and Thymidylate Synthase for Neoadjuvant Cisplatin-based Chemotherapy in Head and Neck Cancer¹

Cisplatin-based neoadjuvant chemotherapy is widely used in the treatment of head and neck cancer. In untreated patients with head and neck squamous cell carcinoma, overall responses of 70–90% with CRs³ of ≥30% have been reported when cisplatin is combined with 5-fluorouracil (5-FU) (1). These numbers fall dramatically in patients with relapsed tumors after surgery and/or radiotherapy who show only a 30% overall response rate to the same cisplatin-based chemotherapy (2).

Neoadjuvant chemotherapy has been used increasingly in the last decade for organ preservation and for unresectable disease in head and neck tumors (3, 4, 5, 6, 7). However, despite the high response rate to neoadjuvant chemotherapy, there is no clear evidence of improved survival for patients treated with cisplatin-based chemotherapy in this setting. Some investigators have concluded that neoadjuvant chemotherapy should not be offered to patients with locally advanced head and neck cancer if improved survival is to be the criteria for selection of treatment (8).

At the present time, there are no accepted prognostic markers that can guide the selection and treatment of patients with head and neck squamous carcinomas or predict the long-term outcome of such treatment. In theory, such markers would be very useful, especially in the neoadjuvant setting because alternative treatments including surgery could be offered to patients who are likely to fail after neoadjuvant treatment.

In head and neck cancer, a number of cellular factors may be important in clinical resistance to both cisplatin and 5-fluorouracil, which are the most widely used drugs for neoadjuvant treatment. Several cellular products are potentially important in regulating cellular resistance to cisplatin-based chemotherapy in head and neck cancer.

The tumor-suppressor gene product, p53, has been correlated with poor response to cisplatin-based chemotherapy in lung and ovarian cancers (9, 10, 11). In patients with invasive bladder cancer undergoing neoadjuvant chemotherapy, p53 mutations have been associated with poor response and prognosis (12). This may result from an impaired ability of mutant p53 to induce apoptosis following DNA damage through transactivation of bax (13).

Glutathione and other components of the glutathione metabolic pathway may also have a critical role in determining the cellular sensitivity to cisplatin-based chemotherapy. Glutathione is a ubiquitous tripeptide, which has been hypothesized to protect against the DNA damaging effects of agents such as cisplatin by conjugating toxic moieties including metal compounds in the cytoplasm and preventing DNA interaction (14, 15). More recent data suggest that the cytoprotective effect of glutathione may in fact be secondary to enhanced DNA repair and increased ability to recover from intracellular toxic events that follow cisplatin-induced DNA damage (16). Glutathione may also be important in protecting the cell from other potentially damaging molecules such as oxygen free radicals. Studies from this laboratory and elsewhere have demonstrated that cellular glutathione levels are inversely correlated with cisplatin sensitivity in head and neck cancer cells (17, 18). GSTs have been shown to catalyze conjugation of glutathione to a number of chemotherapy agents including cisplatin (19, 20). The cisplatin/glutathione complex may be ejected from the cell in an ATP-dependent fashion. The cellular protective action of glutathione has been believed to be a result from the conjugation of the peptide to the cytotoxic chemotherapy agents mediated by the catalytic enzymes known collectively as GSTs (15, 16). The predominant GST isoform in head and neck cancer is GSTπ (21). Immunohistochemical studies have shown that GSTπ expression may predict response to cisplatin-based chemotherapy in patients with non-small cell lung cancer (22).

The MRP is a 190-kDa transmembrane transport protein that is thought to function as an ATP-dependent export pump for conjugates of glutathione and cytotoxic drugs such as cisplatin (23). Overexpression of MRP might then be expected to increase cellular resistance to some chemotherapy agents.

TS catalyzes the methylation of dUMP to dTMP, an essential step in DNA biosynthesis. TS is a critical target for fluoropyrimidine drugs such as 5-FU (24). Overexpression of both TS protein and mRNA have been correlated with resistance to 5-FU chemotherapy in patients with rectal, gastric cancers as well as head and neck cancers (25, 26 ).

The c-erbB2 oncoprotein is a 185-kDa transmembrane protein transmembrane protein with tyrosine kinase activity (27). C-erbB2 is expressed in a significant subset of head and neck cancers (28). Studies in breast and ovarian cancers indicate that activation of c-erbB2 may inhibit cisplatin-induced DNA repair, thereby enhancing cytotoxicity (29, 30).

In this study, we retrospectively analyzed these tumor markers in a series of 68 patients from two institutions who received cisplatin-based neoadjuvant chemotherapy (cisplatin/5-FU or cisplatin/paclitaxel) and compared those data with treatment response and survival.

Clinical data for 68 patients with head and neck squamous carcinoma were obtained by retrospective chart review. The patients were treated between March 1989 and September 1997 at Georgetown University Hospital or Johns Hopkins University. Patients were selected for this analysis primarily on the basis of availability of adequate tissue for study. Formalin-fixed, paraffin-embedded pretreatment biopsy tissues were available for analysis in all patients, either from the treating institution or from the referring hospital. All patients received cisplatin-based neoadjuvant chemotherapy after their initial biopsy either for organ preservation or for unresectable disease. In all cases, biopsy tissue prior to chemotherapy had been obtained as part of routine medical management. Pretreatment clinicopathological characteristics of all patients are described in Table 1.

Forty-nine patients received cisplatin, 100 mg/m², on day 1, followed by a continuous infusion of 5-FU at 1000 mg/m² per day for 5 days. Patients were treated every 3 weeks for a minimum of two cycles. Nineteen patients received cisplatin, 75 mg/m², on day 1 and paclitaxel, 175 mg/m², on day 1; both medications were also given on a 3-week cycle. Patients treated with neoadjuvant chemotherapy who had a>50% (partial) response at the end of two cycles of chemotherapy had a third chemotherapy treatment and went on to receive definitive radiation therapy. Patients who presented with bulky (N2 or N3) neck disease generally had neck dissection at the completion of radiation therapy.

Clinical response scoring for this study was determined by retrospective chart review. CR was defined as complete disappearance of tumor by physical examination and for radiographic evaluation. PR was a ≥50% reduction in tumor mass measured in two dimensions either by physical examination and/or radiographic evaluation of >1-month duration. NR represented no significant change in tumor dimension, whereas PD represented a ≥25% enlargement in tumor dimension while on therapy. DFS and OS were determined from the date of initial diagnostic biopsy. DFS was measured from initial biopsy until local recurrence or distant metastasis. Death without recurrence was not counted as an event. OS was estimated until death by any cause or the date of last patient contact.

Paraffin blocks were obtained from the departments of pathology, Georgetown University and Johns Hopkins University or from the referring institution. Analysis of TS and GSH-S was carried out on only the 50 cases from Georgetown University. Paraffin sections (5 μm) were deparaffinized in xylene, rehydrated through graded alcohols, and washed in PBS. Endogenous peroxidase was blocked using 3% hydrogen peroxide solution. Following incubation in normal goat serum (Biogenex, San Ramon, CA) at room temperature for 20 min., the primary antibodies were diluted in 1% BSA and 1% sodium azide according to the following concentrations: anti-p53 mouse monoclonal antibody (BP53.12, Zymed, South San Francisco, CA), 1:50; anti-TS mouse monoclonal antibody (TS106, gift from Dr. Patrick G. Johnston, The Queen’s University of Belfast, United Kingdom; Ref. 31), 1:200; anti-GSTπ rabbit polyclonal antibody (Novocastra, Newcastle upon Tyne, United Kingdom), 1:200; anti-c-erbB2 mouse monoclonal antibody (Novocastra, Newcastle upon Tyne, United Kingdom), 1:40; and anti-MRP rabbit polyclonal antibody (gift from Dr. Gary D. Kruh, Fox Chase Cancer Center, Philadelphia, PA; Ref. 32), 1:300. The diluted antibodies were added, and the slides were incubated at room temperature for 1 h. Slides were washed in Cadenza buffer, and the reaction was visualized with the BioGenex multilink system (BioGenex, San Ramon, CA). The slides were incubated with multilink solution (bio-tinylated goat antimouse and rabbit immunoglobulin) at 37°C for 20 min followed by horseradish peroxidase-conjugated streptavidin (37°C, 20 min) according to the manufacturer’s protocol. The peroxidase reaction was developed using diaminobenzidine solution. The slides were counterstained with hematoxylin, then mounted. A positive control tissue was included with each set of sections stained. For GSTπ, normal human kidney known to be positive was used. The positive control slides for p53 and c-erbB2 were human ovarian cancers known to be positive. For TS, normal human colon tissue was used as a positive control. The positive control for MRP was normal human kidney. In negative controls, the primary antibody solution was replaced by PBS with 1% BSA and 0.1% sodium azide.

Because no antibody was available for GSH-S, in situ hybridization was performed. Briefly, paraffin sections (5 μm) were deparaffinized in xylene, rehydrated through graded alcohols, and incubated in proteinase K solution at 37°C for 30 min. The slides were incubated in hybridization buffer with ³⁵S radiolabeled antisense probe at 55°C overnight to allow for cellular localization. The antisense probe was synthesized using a cDNA for human GSH-S (gift from Dr. Philip G. Board, The Australian National University, Canberra, Australia; Ref. 33). The next day, single-stranded RNA was degraded by the addition of RNase A and was washed off in graded salt solution. The slides were rinsed in graded alcohol, air-dried, dipped in appropriate diluted liquid emulsion (NTB2, Kodak), allowed to air dry, and placed in light-tight boxes for incubation in a −70°C freezer. After a 4-week exposure, slides were developed by hand and counterstained with hematoxylin.

Immunohistochemical staining and in situ hybridization reactivity for GSH-S were assessed independently by each of two investigators who were blind to the patients’ identity and clinical outcome. For p53, specific nuclear staining of >5% of the cancer cells was considered positive (most positive cases showed nuclear staining in the majority of tumor cells). TS expression was scored positive if any specific cytoplasmic staining was detected. GSTπ expression was quantified by use of a visual grading system based on the intensity of cytoplasm staining (0–3) as follows: grade 0, no immunoreactivity; grade 1, weak immunoreactivity slightly stronger than background staining; grade 2, clear immunoreactivity in more than half of the cancer cells; grade 3, strong immunoreactivity as dark as nuclear counter stain in the majority of cancer cells. For statistical analysis, grades 2 and 3 were classified as high GSTπ immunoreactivity, and grades 0 and 1 were classified as low GSTπ immunoreactivity (34). For c-erbB2 and MRP, specific membrane staining of >5% of the cancer cells was counted as positive. Cytoplasmic staining was not scored. For GSH-S, slides were counted as positive only if there was a specific signal (accumulation of silver grains) on the tumor epithelial cells.

Univariate analyses for recurrence-free survival and OS included Kaplan-Meier survival estimation (Prism, GraphPad Software, Inc., San Diego, CA), with statistical significance assessed via the log-rank test and the Gehan-Wilcoxon test. Multivariable statistical models were generated using a proportional hazards regression approach. Stepwise model building was done manually based on the likelihood ratio test and changes to parameter estimates for variables already in the model. Interactions were tested using cross-product terms. Both forward and backward stepwise approaches yielded the same final models.

Positive staining (>5% cells) for p53 was observed in the nuclei of the cancer cells in 25 of 68 cases (37%; Fig. 1,A). GSTπ was observed mainly in the cytoplasm of cancer cells with 53% high scores (36 of 68; Fig. 1,B). TS expression was observed in the cytoplasm of the cancer cells in 29 of 50 cases (58%; Fig. 1 C). Positive membrane staining for c-erbB2 was detectable in 21 of 68 cases (31%), and positive membrane staining for MRP was found in 29/68 cases (43%). GSH-S was detected in cancer cells in 13/50 cases (26%).

Expression of tumor markers was not associated with age, gender, site of primary tumor, clinical stage, T stage, and N stage. A higher proportion of patients with high GSTπ scores or TS-positive tumors had well-differentiated tumors (GSTπ, P < 0.001; TS, P = 0.023).

The overall chemotherapy response rate for neoadjuvant treatment was 79% (54 of 68 cases). Table 2 shows the relationship between tumor markers and clinical response to neoadjuvant chemotherapy. Although there was a trend between expression of several tumor markers and clinical resistance to chemotherapy, none of the tumor markers measured were significantly correlated with response. Other parameters, such as age, gender, clinical stage, T stage, N stage, and tumor differentiation were not significantly correlated with clinical response to chemotherapy.

Table 3 shows the association of chemotherapy response and tumor markers with DFS in the 68 patients analyzed. DFS (or PFS in patients who had less than a PR to induction chemotherapy and were never free of disease) was significantly correlated with clinical response to neoadjuvant chemotherapy for all patients (P < 0.0001; Hazard ratio, 4.67; Fig. 2 A). Poor response to treatment (NR, PR) was associated with shorter PFS. At 24 months after diagnosis, the PFS rate was 60% in responders, whereas it was zero in nonresponders.

For tumor markers analyzed, DFS was approximately two times longer for patients who were p53-negative compared with patients who were p53-positive in neoadjuvant treatment cases, although the difference did not achieve statistical significance (P = 0.112; Hazard ratio, 1.69; Fig. 3 A). Expression of GSTπ was associated with decreased DFS in neoadjuvant treatment cases. DFS was 55% at 3 years after diagnosis in patients with low GSTπ scores, although it was 39% in patients with high GSTπ scores (P = 0.119; Hazard ratio, 1.7). Expression of c-erbB2, TS, MRP, or GSH-S were not significantly correlated with DFS in patients treated with neoadjuvant chemotherapy. Clinical parameters including age, gender, clinical stage, T stage, site of primary tumor, and tumor differentiation were not correlated with DFS in neoadjuvant cases. As expected, patients with no neck disease had a better prognosis than patients with clinically positive necks.

Table 4 shows the association of chemotherapy response and tumor markers with OS. OS was significantly correlated with clinical response to neoadjuvant chemotherapy (P < 0.0001; Hazard ratio, 4.51; Fig. 2,B). Poor response (PD, NR) was associated with shorter OS. For neoadjuvant treatment, expression of p53 was very significantly correlated with OS (P < 0.0001; Hazard ratio, 5.37; Fig. 3,B). OS was 81% at 3 years after diagnosis in patients with p53-negative tumors, compared to 30% in patients with p53-positive tumors. A significant association was also observed between high GSTπ scores and shorter OS for neoadjuvant treatment (P = 0.0018; Hazard ratio, 3.78; Fig. 4,A). Those patients with significant TS expression had an overall poorer survival compared with patient whose tumors did not express TS (P = 0.0071; Hazard ratio, 2.93; Fig. 4 B).

Expression of c-erbB2, MRP, and GSH-S were not correlated to OS for the patients evaluated. Among clinical parameters evaluated, female gender was associated with an improved OS (P = 0.018), as was low clinical stage, but other clinicopathological factors were not statistically associated with OS.

No variables demonstrated statistically significant associations with DFS. Four variables were significant predictors of OS: lymph node metastases, p53 overexpression, high GSTπ expression, and TS expression. There was a statistically significant interaction between p53 and GSTπ. The final multivariable model is shown in Table 5. (Because of the interaction, the effect of p53 mutation versus wild-type is shown separately for each GSTπ level.)

In this study, we observe that the expression of p53, GSTπ, and TS are associated with poor prognosis in head and neck tumor patients treated with neoadjuvant cisplatin/5-FU or cisplatin/paclitaxel.

Loss of wild-type p53 function has been associated with chemotherapy resistance in vitro (13) and in vivo (10, 11). p53 may participate in cellular pathways leading to apoptosis following treatment with DNA-damaging agents such as cisplatin (35, 36, 37). Our data indicate that overexpression of p53 protein as determined by immunohistochemistry is a strong indicator of poor prognosis. Similar data have been demonstrated in bladder cancer where p53 overexpression has also been associated with decreased survival (12). Analysis of the data from the Veterans Affairs larynx preservation trial similarly indicated that p53 was not associated with chemotherapy response but was correlated with decreased patient survival (38).

GSTπ expression was also associated with decreased survival in the patients evaluated in this study. GSTπ has been widely studied as a factor contributing to cisplatin resistance in a spectrum of neoplasms (21, 22). We previously demonstrated that expression of GSTπ is associated with poor response to treatment in neoadjuvant and relapsed patients. In that study, GSTπ was a significantly better predictor of response to treatment in relapsed patients than in patients receiving neoadjuvant chemotherapy (36). Similarly, in our present study, GSTπ was a weak predictor of response to treatment, although it did have a stronger correlation with OS. Expression of GSTπ has been shown to correlate with decreased survival in several malignancies (39, 40), but a recent study in head and neck tumors demonstrated no clear correlation between GSTπ expression and local control in patients receiving radiation treatment (41). Additionally, the association between GSTπ expression and prognosis in other tumors is controversial (42, 43).

Our analysis of TS expression supports previous studies that suggest that this enzyme may be associated with decreased treatment response and survival in patients with advanced head and neck cancer treated with 5-FU-based chemotherapy (26). In that study, TS expression was more common in moderately or well-differentiated tumors. We observed the same association between TS expression and tumor differentiation. Similar findings have been demonstrated in other malignancies, suggesting that TS expression may in part determine response to 5-FU-based chemotherapy and influence survival in a broad spectrum of tumors (25). The lack of association between TS expression and treatment response in our series may reflect the relatively small number of cases studied and the high overall response rate to neoadjuvant therapy. The observation that expression of p53, GSTπ, and TS were more predictive of poor OS than DFS may indicate that these markers reflect the aggressiveness of disease in these patients. Patients with relapsed tumors that are negative for these factors may have a more indolent course and be more likely to respond to second-line chemotherapy than patients who have relapsed tumors that are positive for p53, GSTπ, and TS. In support of this concept, we previously demonstrated that among relapsed patients, the response to cisplatin-based chemotherapy was 70% for patients whose tumors were GSTπ-negative, but only 8% for patients who had GSTπ-positive tumors (34).

Our data indicate that easily measured cellular factors may be important predictors of outcome in patients undergoing neoadjuvant treatment for head and neck malignancies. The magnitude of survival differences seen with p53 and the other markers indicate that these may ultimately be clinically useful in making treatment decisions for patients being considered for neoadjuvant treatment. For example, an attempt at organ preservation may be less attractive for a patient whose tumor profile predicts that he or she will be unlikely to respond or survive with that treatment strategy. If these data can be confirmed, it may be possible to guide the selection of chemotherapy agents based on the individual profile of gene expression in a given tumor. Patients with high TS expression but low c-erbB2 expression may be better candidates for a taxane-based regimen rather than a 5-FU-containing regimen. Additionally, poor prognosis patients may be identified as candidates for clinical trials involving new agents or dose-intensification schemes.

We thank the physicians and nurses of the head and neck tumor service at Georgetown University, including Catherine Picken, Bruce Davidson, William Harter, and Betsy Bischoff. We also thank Koleoso Olukayode for his assistance with statistics and data management.