Immunohistochemical analysis of NAD(P)H:quinone oxidoreductase and NADPH cytochrome P450 reductase in human superficial bladder tumours: Relationship between tumour enzymology and clinical outcome following intravesical mitomycin C therapy

Abstract

A central theme within the concept of enzyme-directed bioreductive drug development is the potential to predict tumour response based on the profiling of enzymes involved in the bioreductive activation process. Mitomycin C (MMC) is the prototypical bioreductive drug that is reduced to active intermediates by several reductases including NAD(P)H:quinone oxidoreductase (NQO1) and NADPH cytochrome P450 reductase (P450R). The purpose of our study was to determine whether NQO1 and P450R protein expression in a panel of low-grade, human superficial bladder tumours correlates with clinical response to MMC. A retrospective clinical study was conducted in which the response to MMC of 92 bladder cancer patients was compared to the immunohistochemical expression of NQO1 and P450R protein in archived paraffin-embedded bladder tumour specimens. A broad spectrum of NQO1 protein levels exists in bladder tumours between individual patients, ranging from intense to no immunohistochemical staining. In contrast, levels of P450R were similar with most tumours having moderate to high levels. All patients were chemotherapy naïve prior to receiving MMC and clinical response was defined as the time to first recurrence. A poor correlation exists between clinical response and NQO1, P450R or the expression patterns of various combinations of the 2 proteins. The results of our study demonstrate that the clinical response of superficial bladder cancers to MMC cannot be predicted on the basis of NQO1 and/or P450R protein expression and suggest that other factors (other reductases or post DNA damage events) have a significant bearing on tumour response. © 2004 Wiley-Liss, Inc.

The ability to predict tumour response to chemotherapy has been and continues to be a major objective in cancer research with the emphasis currently being placed on the identification of molecular markers of tumour response.1, 2 Within the field of bioreductive drug development, the ability to individualise chemotherapy based on the activity of specific reductases and hypoxia in tumours forms one of the cornerstones of a concept known as enzyme-directed bioreductive drug development.3 The key requirements for the successful clinical application of this concept include the development of drugs that are bioreductively activated by specific reductases and good correlations between tumour response and enzymology in experimental tumour models. Mitomycin C (MMC) is a clinically active agent used to treat a number of tumours and is regarded as the prototypical bioreductive drug.4 Its mechanism of action has been extensively reviewed elsewhere5, 6, 7, 8 and involves a complex interplay between tumour physiology (oxygen tension and extracellular pH) and several reductase enzymes (both 1 and 2 electron reductases). The ability to predict cellular and tumour response to MMC based on tumour enzymology has been addressed by several research groups but the results are controversial, and there are conflicting reports on the relationship between tumour enzymology and chemosensitivity in the literature. This is particularly true in the case of the 2 electron reductase NAD(P)H:quinone oxidoreductase-1 (NQO1), which has been implicated in the bioreductive activation of MMC, especially under aerobic conditions.5, 6, 9 Reports of good correlations between NQO1 activity (and mRNA expression) and response to MMC contrast sharply with reports of poor correlations in both in vitro and in vivo models.10, 11, 12, 13, 14, 15, 16

In view of these findings and in conjunction with the complex nature of MMC activation, Cummings et al.6 have suggested that the ability to predict tumour response to MMC on the basis of the activity of a single enzyme is too simplistic and unlikely to be of clinical benefit. In this context, a recent study has demonstrated that analysis of NQO1 and cytochrome P450 reductase (P450R, which plays a role in MMC activation predominantly under hypoxic conditions17) mRNA by semiquantitative RT-PCR correlates with MMC response of histocultures derived from 21 human bladder cancer patients.18 Whilst the response of histocultures in vitro has been shown to correlate with tumour response to MMC in vivo,19, 20, 21 further clinical studies to validate these findings in a larger number of patients are required particularly in view of the conflicting reports in the literature regarding the role of NQO1 in MMC activation. The primary objective of our study is to determine whether NQO1 and P450R protein expression in low and intermediate grade (G1/G2) superficial (Ta/T1) bladder cancer correlates with clinical response to MMC. The use of bladder cancer as a clinical model to address this question has several advantages over other cancer types. MMC is routinely used as a single agent to treat superficial disease22 and direct intravesical administration ensures that variations in drug exposure parameters between individual patients are reduced compared to systemic drug administration. Furthermore, the exclusion of G3 tumours (which have poorer prognosis than low grade tumours) from our study removes a possible source of variation in clinical response caused by a more aggressive tumour. These facts greatly reduce the number of variable factors that could influence clinical response, thereby facilitating a critical evaluation of the hypothesis in question.

MATERIAL AND METHODS

Tissue and clinical data collection

Following local research ethical committee approval, formalin-fixed, paraffin-embedded human bladder tumour specimens were collected from Bradford Royal Infirmary Pathology archives. Tissue specimens and medical records were obtained for patients treated between 1995 and 2000, and all patient details were anonymised to ensure confidentiality. All patients were chemo- and immuno-therapy naïve at the time of surgical resection. All patients received a single dose of MMC (40 mg/40 ml) within at least 24 hr of surgery. Patients with multiple lesions, tumours of > 5cm diameter or recurrent G2pT1 tumours subsequently received a course of MMC (40 mg/40 ml given weekly for 6 weeks), commencing 2 weeks after surgery, so as to allow time for obtaining histological confirmation. Tissue specimens were obtained from 92 patients and sample collection was restricted to histological grades 1 and 2 and stages Ta and T1. Patients were routinely followed up by check cystoscopy 3, 6 and 12 months (and annually thereafter) after chemotherapy. The clinical end point used to measure response to MMC was time to first recurrence following treatment with MMC (identified at check cystoscopy or if patients developed clinical symptoms).

Immunohistochemistry

Formalin-fixed paraffin-embedded tissue sections were deparaffinised in xylene and rehydrated through graded ethanol washes to phosphate-buffered saline (PBS). Endogenous peroxidase activity was quenched via incubation in 3% hydrogen peroxide for 20 min at room temperature. Following a blocking step for nonspecific staining, and primary and secondary antibody incubations, signals were amplified using a horseradish peroxidase labelled avidin-biotin complex (ABC Vectastain, Vector Labs, Burlingame, CA). Immunocomplex visualisation was performed using diaminobenzidine (DAB) for 5 min at room temperature. Slides were counterstained with Harris haematoxylin. Negative controls were performed by either omitting the primary antibody or substituting it with mouse (in the case of NQO1) or goat (in the case of P450R) IgG1 negative controls (Dako Ltd, UK).

Detection of NQO1 and P450R

Nonspecific staining was blocked using normal horse serum (in the case of NQO1) or normal rabbit serum (in the case of P450R). Detection of NQO1 was performed according to previously published protocols23 using a mouse anti-human NQO1 monoclonal antibody (IgG1) kindly supplied by Dr. Siegel and Dr. Ross (University of Colorado). Sections were incubated for 1 hr at room temperature with the anti-NQO1 antibody diluted 1:1 in TBST-M [10 mM Tris-HCl, 150 mM NaCl and 0.2% Tween 20, 5% nonfat milk powder]. Detection of P450R was performed using a goat antihuman P450R polyclonal antibody (Santa Cruz Biotechnology, Santa Cruz, CA). Sections were incubated for 90 min at room temperature with the anti-P450R antibody diluted at 1:100 in PBS. Immunodetection was performed using a biotinylated antimouse IgG (NQO1) or antigoat IgG (P450R) (Vector Labs.) followed by signal amplification and detection.

Semiquantitative analysis of immunohistochemistry

Positive immunostaining in tumour cells (staining in stromal cells was not scored) was scored semiquantitatively by 3 independent observers. Each observer scored 6 randomly chosen fields of views (×20 magnification) and 3 sections per block were scored. A scoring system was devised consisting of 3 grades of staining intensity and distribution: 1 (no staining), 2 (weak to moderate staining) and 3 (strong staining). An average score was calculated for each of the tissue sections from the results of the 3 independent observers.

Data analysis and statistics

Kaplan-Meier plots of percentile disease-free patients against time following MMC treatment was constructed based on NQO1 and P450R protein scores as determined above. A further Kaplan-Meier plot was constructed comparing patients with high levels of both enzymes against patients with any other combination of NQO1 and P450R scores. A log-rank test was performed comparing time to first recurrence of the groups demonstrated on each plot. Finally, a Cox proportional hazards regression model was fitted using the factors: single dose or course, NQO1 level (low, moderate or high), P450R level (moderate or high) and a dichotomous factor indicating whether or not the levels of both enzymes were high. Hazard ratios generated by this analysis are analogues to the risk ratio. For instance, when comparing the risk of recurrence between 2 conditions (0 and 2 NQO1 phenotype scores, for example), a hazard ratio of 2.0 means there is twice the risk of recurrence in someone with NQO1 score 2 compared to the reference group (NQO1 score 0 in this example). All analyses were performed using Stata statistical software (Software release 7.0, Stata Corporation, College Station, TX).

RESULTS

Immunohistochemical analysis

The results of the immunohistochemical analysis are presented in Table I and Figure 1. Staining for NQO1 was cytosolic and largely confined to the tumour cell population with little or no staining in stromal cells (Fig. 1a,c). Considerable interpatient heterogeneity in NQO1 staining intensity exists within patients segregated on the basis of grade, stage or stage plus grade (Table I). Within this group of patients 39.1% and 43.5% of tumours had moderate or intense NQO1 staining patterns, respectively (Table I). No obvious relationship between stage, grade (or stage plus grade) and NQO1 staining intensity was shown (Table I). Intra-tumoural heterogeneity with regards to NQO1 protein expression exists within many specimens with regions of intense NQO1 staining in close proximity to tumour cells with moderate or no NQO1 staining (Fig. 1i,j). Analysis of NQO1 distribution in tumours with an overall NQO1 score of 2 demonstrated no significant difference in the NQO1 distribution between good and poor responders to MMC (Fig. 2). P450R staining was cytoplasmic, but in contrast to NQO1, staining intensity in both normal stromal and tumour tissues was comparable (Fig. 1d,f). In addition, interpatient heterogeneity in P450R immunostaining was reduced compared to NQO1 with all tumours examined having moderate to high expression of P450R (Table I).

Table I. Summary of Immunohistochemical Results Segregated on the Basis of Tumour Stage or Grade and Stage Plus Grade Combined1

Stage and Grade (n = number of patients)	NQO1 (IHC score 1)	NQO1 (IHC score 2)	NQO1 (IHC score 3)	P450R (IHC score 1)	P450R (IHC score 2)	P450R (IHC score 3)
1 Values presented represent percentage of patients in each category.
G1 (n = 31)	22.5	35.5	42.0	0	67.7	32.3
G2 (n = 61)	14.7	41.0	44.3	0	49.2	50.8
Ta (n = 64)	18.8	40.6	40.6	0	60.9	39.1
T1 (n = 28)	14.3	35.7	50.0	0	42.9	57.1
G1/pTa (n = 24)	20.8	37.5	41.7	0	66.7	33.3
G1/pT1 (n = 7)	28.6	28.6	42.8	0	71.4	28.6
G2/pTa (n = 40)	17.5	42.5	40.0	0	57.5	42.5
G2/pT1 (n = 21)	9.5	38.1	52.4	0	33.3	66.7
All patients (n = 92)	17.4	39.1	43.5	0	55.4	44.6

Figure 1. Immunohistochemical analysis of NQO1 and P450R in 4 patients with superficial bladder cancer (G2 pTa). (a,c,e,g) NQO1 immunohistochemistry (IHC) staining. (b,d,f,g) represent P450R IHC staining. (a–d) Two cases that have high NQO1 and moderate P450R protein levels. Clinical response to MMC for case 1 (a,b) and case 2 (c,d) was no recurrence 60 months after treatment and tumour recurrence at 3 months, respectively. Similarly, (e–h) represent a further 2 cases which have low NQO1 and moderate P450R. Clinical response to MMC for case 3 (e,f) and case 4 (g,j) was no recurrence 49 months after treatment and tumour recurrence at 4 months, respectively. (i,j) NQO1 staining in tumours from 2 patients with a broad spectrum of clinical response (recurrence within 6 months and no recurrence 60 months after MMC therapy, respectively). Both tumours show marked intratumour heterogeneity in the pattern of NQO1 distribution. The bar represents 100 μm in all cases.

Figure 2. Analysis of protein distribution within NQO1 rich (score 2) tumours. Each bar represents the mean ± standard deviation for 10 tumours in each group of good (no recurrences 49 or more months after MMC treatment) and poor (recurrences between 3 and 11 months after MMC therapy) responders to MMC.

Clinical response data

The response of bladder cancer tumours following intravesical administration (single dose and course of MMC combined) of MMC is presented in Figure 3. A broad spectrum of response (as measured by time to first recurrence) was observed with some patients recurring within 6 months, whereas others remain disease free at greater than 48 months after treatment. This heterogenous pattern of response is consistent throughout all patients and is independent of both tumour stage and grade (Fig. 3).

Figure 3. Response of superficial bladder cancer to MMC: Relationship with tumour stage and grade.

Relationship between NQO1, P450R and clinical response to MMC

Kaplan-Meier plots depicting the relationship between clinical response (time to first recurrence) following intravesical MMC and NQO1 and/or P450R are presented in Figure 4 and Cox proportional hazards statistical analysis model is described in Table II. No statistical difference exists between the Kaplan-Meier plots for both NQO1 (Fig. 4a, Table II) and P450R (Fig. 4b, Table II) when immunohistochemical scores are compared to clinical response to MMC. Similarly, the clinical response of patients whose tumours have both high NQO1 and P450R protein levels is not significantly different from patients with all other possible combinations of NQO1 and P450R protein levels (Fig. 4c, Table II). A hazard ratio of 1.85 (Table II) was obtained, suggesting that there is a risk of earlier recurrence in patients with high NQO1 and P450R although this difference did not reach statistical significance (p = 0.395). A hazard ratio of 1.41 was also obtained for patients treated with a course of MMC as opposed to a single dose of MMC (Table II), suggesting that patients treated with a course of MMC have an increased risk of earlier recurrence. As this difference did not however reach statistical significance (p = 0.323, Table II), further analysis and stratification of protein expression data on the basis of whether patients received a course or single dose of MMC was deemed unnecessary. On an individual patient basis, examples of tumours that represent the extremes of NQO1 expression (with moderate P450R protein levels) and clinical response are presented in Figure 1. In the case of high NQO1 and moderate P450R, 1 patient (Fig. 1a,b) had a good response to MMC (no recurrences 60 months after treatment), whereas a different patient (Fig. 1c,d) with similar NQO1/P450R protein expression profiles had a poor clinical outcome (recurrence within 3 months). The same pattern exists for the 2 other patients presented in Figure 1e–h. In these cases, both patients had low NQO1 and moderate P450R protein levels but 1 patient (Fig. 1e,f) had a good response to MMC (no recurrences with in 49 months of treatment), whereas the other patient (Fig. 1g,h) had recurrences 4 months after intravesical MMC therapy.

Figure 4. Kaplan-Meier plots of time to first recurrence following MMC treatment stratified according to NQO1 IHC scoring (a), P450R IHC scoring (b) and patients whose tumours have high levels of both NQO1 and P450R compared to all other possible combinations of NQO1 and P450R levels (c). No statistically significant differences exist in the Kaplan Meier plots with p values of 0.727, 0.446 and 0.395 for data in (a), (b) and (c), respectively.

Table II. Statistical Analysis of Data Using the Cox (Proportional Hazards) Model1

	Hazard ratio	95% CI (lower, upper)	p value
1 Hazard ratios are all with respect to baseline levels of the factors in the model: single dose, low NQO1, moderate P450R and any combination of NQO1 and P450R other than both high.
Course of MMC (vs. single MMC)	1.41	(0.71, 2.77)	0.323
Moderate NQO1 (score 2) vs. NQO1 (score 1)	1.35	(0.53, 3.45)	0.524
High NQO1 (score 3) vs. NQO1 (score 1)	0.89	(0.25, 3.22)	0.868
High P450R vs. P450R score 2.	1.11	(0.44, 2.76)	0.824
High NQO1 and P450R vs. all other combinations of NQO1 and P450R scores.	1.86	(0.45, 7.76)	0.395

DISCUSSION

The ability to tailor chemotherapy to individual patients has been an integral part of bioreductive drug development for many years with considerable attention being paid to the relationship between tumour enzymology and response.3, 5, 6 In our study, the protein expression of 2 key enzymes implicated in the activation of MMC have been characterised and retrospectively compared to clinical response to MMC. The results of the immunohistochemical studies clearly demonstrate that a broad spectrum of NQO1 expression exists with a significant number of patients' tumours containing high NQO1 levels (Table I). These results are entirely consistent with previous studies24, 25 and suggest that NQO1 is a good target for drug development in bladder cancer. A narrow range of P450R protein levels were observed with most tumours having moderate to high protein expression, a result which is again consistent with previous findings in bladder, lung and breast cancers.25, 26 The response of bladder cancer patients to MMC is highly heterogenous with some patients suffering early recurrences, whereas others with histologically identical disease experiencing no recurrences many years after therapy (Fig. 3). In direct contrast to previous reports of a correlation between NQO1, P450R mRNA levels and the response of human bladder cancer histocultures to MMC in vitro,18 the major finding of our study is that the clinical response of superficial bladder cancers to MMC cannot be forecast on the basis of NQO1, P450R protein levels or various combinations of the 2 protein expression patterns (Figs. 1 and 4).

The reasons for the conflicting findings are not clear although it is important to acknowledge that both studies have employed different methodologies to analyse enzymology, protein expression and sensitivity to MMC. Whilst a detailed discussion of the relative merits of immunohistochemistry and semiquantitative RT-PCR is beyond the scope of this article, issues such as the relationship between gene expression (or immunoreactive protein) and enzyme activity, spatial localisation of mRNA or protein in tumour as opposed to stromal cells and the semiquantitative nature of both assays are frequently raised points that affect both methodologies used to a greater or lesser extent. The antibodies used in our study have previously been shown to selectively stain for both NQO123, 24 and P450R (data not shown) with good correlations between immunoblotting, protein staining and enzyme activity in cell cultures and tumours. The presence of intratumour heterogeneity with regard to NQO1 protein expression (Figs. 1 and 2) is however important in this context as it raises several issues that could influence the analysis of NQO1 and the outcome of therapy. These include the fact that analysis of whole tumour specimens (by RT-PCR or enzyme assays) may not be representative of NQO1 mRNA, protein or enzyme activity in all tumour cells within the tumour mass. Furthermore, the presence of “pockets” of low NQO1 expressing cells within an otherwise NQO1 rich tumour raises the possibility that poor tumour response maybe due to these cells escaping the activity of MMC and subsequently repopulate the tumour. This is however unlikely since no significant differences in the distribution pattern of NQO1 protein expression exist between MMC sensitive and resistant tumours in our study (Fig. 2). With respect to the measurement of MMC sensitivity, the end point used in our study was time to first recurrence following MMC therapy. In our opinion, this is the most relevant endpoint to use as patients receive no further chemotherapy or immunotherapy following MMC treatment and so time to first recurrence should provide a good measure of MMC efficacy. Whilst the study by Gan et al.18 used an in vitro-based assay to assess tumour response, this approach is also valid as the clinical relevance of human tumour histocultures has been substantiated in several studies.19, 20, 21 Finally, the patients selected for inclusion in our study all had superficial bladder cancers (stage Ta and T1) and no G3 tumours were included. This eliminates a potential problem in the assessment of tumour response as G3 tumours are typically more aggressive and have a poorer prognosis than low-grade tumours. It is of interest to note that G3 (n=6), T2 (n = 2) and T3 (n = 3) tumours were included in the 21 patients analysed by Gan et al18 and therefore the relevance of IC₅₀ values obtained in histocultures to clinical outcome is difficult to interpret since these tumours have a much poorer prognosis than low grade Ta/T1 tumours.

Whilst differences in experimental design could potentially explain the discrepancies between our study and Gan et al.,18 several other issues exist that fundamentally question whether or not response to MMC can realistically be predicted on the basis of NQO1 and/or P450R levels. In the late 1980s for example, there was considerable doubt as to whether or not MMC was actually a substrate for human NQO1.27, 28 It is now widely acknowledged that MMC is a substrate for NQO1 but only under mild acidic conditions.29 Under physiological pH conditions, MMC serves as a suicide inhibitor of NQO1.29 NMR studies have however demonstrated that intracellular pH remains neutral or slightly alkaline when extracellular pH is acidic.30 In addition, several other enzymes have been implicated in the activation of MMC. These include other reductases (e.g., cytochrome b5 reductase, xanthine oxidase and dehydrogenase), nitric oxide synthase, the DNA repair enzyme HAP1 and other enzymes that have not yet been fully characterised.31, 32, 33, 34, 35, 36 Dissecting the relative roles that these enzymes play in the bioreductive activation of MMC is complex and the relationship between enzyme activity and chemosensitivity requires further study. In addition, physiological conditions within tumours could also influence the final outcome of chemotherapy as both hypoxia and low extracellular pH have been shown to modulate the activity of MMC in vitro.37, 38

Efforts to predict response to bioreductive drugs in general have almost exclusively focused on the process of bioreductive activation itself with relatively little attention being focused on other biologically and pharmacologically important processes that determine the outcome of therapy. Other factors such as how the cell responds to potentially lethal DNA damage in terms of repair or ability to undergo apoptosis will also determine how a cell responds to chemotherapy. Several studies have in fact demonstrated that other biochemical properties of cells could have a significant bearing on how the cell responds to the damage induced by MMC.39, 40, 41, 42, 43, 44 In addition, recent studies have identified a novel role for NQO1 in the stabilisation of p53,45, 46 which could also have an impact on response to MMC. Further studies are required to determine whether or not these factors are important in terms of developing novel assays that can predict tumour response to MMC.

In conclusion, the results of our study clearly demonstrate that the clinical response of a series of human superficial bladder tumours to MMC cannot be forecast on the basis of NQO1 and/or P450R levels as determined by immunohistochemical methods. These results suggest that factors other than NQO1 and P450R influence the final outcome of chemotherapy and significant improvements in the design of predictive assays are required before clinical decisions as to which individual is suited for MMC therapy can be made. Nevertheless, the response of bladder cancer patients to MMC is clearly heterogenous and the need to develop other approaches to individualise patient chemotherapy remains.47 Whilst the results of our study demonstrate that MMC response cannot be predicted on the basis of NQO1 and P450R protein levels, many superficial bladder tumours possess elevated levels of NQO1, suggesting that NQO1 is a potential target for drug development in this disease. Compounds where NQO1 plays a pivotal role in bioreductive activation may be suitable candidates for clinical evaluation. In these cases, prediction of response based upon NQO1 activity may be more feasible.