CYP3A4*22 Genotype and Systemic Exposure Affect Paclitaxel-Induced Neurotoxicity

Paclitaxel is a highly active antimicrotubular agent used for the treatment of various solid tumors and has a large interpatient variability in pharmacokinetics and toxicity (1). Neurotoxicity is frequently observed during paclitaxel treatment and is often dose limiting. The degree of neurotoxicity is highly variable between individual patients (2, 3). Axonal degeneration and demyelization are the primary underlying causes of this neurotoxicity (4).

Genetic variants in enzymes involved in paclitaxel metabolism could contribute to interindividual differences in toxicity and efficacy of paclitaxel treatment. Paclitaxel is metabolized by cytochrome 450 (CYP) enzymes CYP2C8 and CYP3A4 (5, 6). Recently, a new intron 6 single-nucleotide polymorphism (SNP), encoding the CYP3A4*22 variant allele, was discovered. This variant allele is associated with decreased CYP3A4 hepatic mRNA levels and consequently lower enzymatic activity (7). In vivo, the CYP3A4*22 variant allele was shown to be associated with altered therapeutic parameters in several CYP3A4-metabolized drugs (e.g., tacrolimus, simvastatin, and cyclosporine; refs. 8–10).

The majority of patients treated with paclitaxel will develop peripheral neurotoxicity in the course of their treatment (11). The incidence and severity of neurotoxicity has been associated with pharmacokinetic exposure parameters such as area under the curve (AUC), and time above total paclitaxel concentrations of 0.05 μmol/L (T_>0.05; ref. 12). Mielke and colleagues studied the association between paclitaxel pharmacokinetics and neurotoxicity in 24 patients and found that drug exposure (AUC × weeks of paclitaxel therapy) was higher in the group that developed neurotoxicity (12). Furthermore, Green and colleagues showed in 23 patients that paclitaxel pharmacokinetics and severity of neurotoxicity were correlated (13). Studies in larger cohorts on the relationship between paclitaxel exposure and neurotoxicity have not been published so far.

The aim of the current study was to evaluate the influence of several SNPs in genes encoding drug-metabolizing enzymes and transporters on the pharmacokinetics of paclitaxel and development and severity of sensory neuropathy. In addition, we aimed to further clarify potential associations between paclitaxel pharmacokinetic parameters and the development and severity of peripheral neuropathy in a large cohort of patients.

Patients with cancer treated with paclitaxel for different tumor types within a prospective trial in which pharmacokinetics, pharmacodynamics, and pharmacogenetics was studied (registered at www.trialregister.nl as NTR2311, ethics board study number MEC 03.264) were included in the exploratory cohort (n = 261). The influence of genetic variants on the pharmacokinetics and frequency and severity of paclitaxel-induced neurotoxicity were studied. The findings were subsequently validated in an independent cohort of paclitaxel-treated patients (n = 239) from whom whole blood for DNA analysis and neurotoxicity data were available (ethics board study number MEC 02.1002; this study involves a large dataset of patients with cancer who provided blood for DNA analysis for pharmacogenetic purposes). In this validation cohort, the association between CYP3A4*22 carrier status and development and severity of neuropathy were studied.

The inclusion criteria for the exploratory cohort were (i) histologic or cytologic confirmed diagnosis of cancer treated with paclitaxel, (ii) age 18 years or older, (iii) WHO performance score 0–1, and (iv) adequate hematopoietic, hepatic, and renal functions. The use of CYP3A4 and CYP2C8 inducers or inhibitors was not allowed. In the validation cohort, patients were included if whole blood and neurotoxicity data were available. The trials were approved by the Ethics Board of the Erasmus University Medical Center and supported by the Dutch Cancer Society. All patients provided written informed consent before study participation.

During the entire treatment course with paclitaxel, neurotoxicity was graded by the treating physician according to National Cancer Institute—Common Terminology Criteria for Adverse Events criteria version 2–4. During each hospital visit, the highest neurotoxicity score of the previous cycle was assessed. In both cohorts, the highest neurotoxicity score during paclitaxel treatment was used in the analyses.

Paclitaxel pharmacokinetics, using a validated limited sampling strategy, were assessed in up to 3 treatment cycles for each patient in the exploratory cohort. Pharmacokinetic sampling was allowed during any treatment cycle. Lithium heparin was used as anticoagulant for all samples. Paclitaxel was quantitated by a validated UV detection high-performance liquid chromatography method (14) or by a validated or liquid chromatography/tandem mass spectrometry (LC/MS-MS) method (15).

Individual pharmacokinetic parameters were calculated on the basis of measured plasma samples and a previously developed population pharmacokinetic model for paclitaxel (16–18). Individual pharmacokinetic parameters were estimated as Empirical Bayes estimates within the nonlinear mixed-effect modeling software NONMEM version 7 (Icon Development Solutions). AUCs were obtained by integrating the predicted concentration–time profile up to 96 hours after start of the infusion. The time above 0.05 μmol/L (T_>0.05) was predicted for each patient.

Genomic DNA was isolated from 200 μL EDTA whole blood using MagnaPure LC (Roche Diagnostics GmbH). Genotyping was conducted using TaqMan (Applied Biosystems) assays for CYP2C8*3 (rs10509681, C_25625782_20, 1196A>G), CYP2C8*4 (rs1058932, C_361406_1, 792C>G), ABCB1 3435 C>T (rs1045642, C_7586657_20), and CYP3A4*22 (rs35599367, C_59013445_10, intron 6 C>T), using 20 ng genomic DNA on the ABI PRISM 7500 fast real-time PCR Systems (Applied Biosystems) according to the manufacturer's instructions. Assays were validated by sequencing.

Human dorsal root ganglia isolated from the lumbar position 4 (L4) were obtained from the National Disease Research Interchange (NDRI) and RNA was extracted using the RNEasy Mini Kit (Qiagen). Expression of CYP3A4 was measured by quantitative real-time PCR (qRT-PCR) using SYBR green and the gene-specific primers (forward: 5′-CACAGATCCCCCTGAAATTAAGCTTA-3′; reverse: 5′-AAAATTCAGGCTCCACTTACGGTG-3′). Gene expression was determined by C_t relative to the housekeeping gene, GAPDH, which was measured using a gene specific TaqMan probe (HS02758991_g1; Applied Biosystems).

Data are presented as medians with ranges, unless stated otherwise. To test whether patients with different grades of neurotoxicity had different pharmacokinetic parameters, the Kruskal–Wallis test was used. To study the relationship between genetic variants and severity of neurotoxicity, the Fisher exact test was used. To test the association between severity of neurotoxicity and CYP3A4*22 carrier status, logistic regression was conducted. The analysis was conducted separately for males and females because of the reported gender difference in paclitaxel pharmacokinetic parameters (19). Also, to correct for different dosing regimens, the analysis was stratified on weekly and 3 weekly schedules of paclitaxel. To test whether all studied genetic variants were in Hardy–Weinberg equilibrium, the χ² test was used. A P value below 0.05 was considered statistically significant. All statistical analyses were conducted with SPSS version 20.0 and Stata (StataCorp), release 12.

In the exploratory cohort, 261 patients (135 male, 126 female) were included. Median age was 61 years (range: 18–82 years) and 96% of patients were Caucasians (Table 1). Esophageal cancer was the main diagnosis (46%) in this cohort. Patients were treated with a median dose of 180 mg paclitaxel during each cycle (range: 75–560 mg). The median cumulative dose in this cohort was 975 mg (range: 280–3,910 mg). In 7 patients, genotyping could not be conducted because of poor DNA quality.

In the validation cohort, 239 patients (129 male, 110 female) were included. Median age was 63 years (range: 24–83 years) and 95% of patients were Caucasians. Most patients in this cohort were diagnosed with esophageal cancer (64%; Table 1). Patients were treated with a median dose of 165 mg paclitaxel during each cycle (range: 70–480 mg). The median cumulative dose in this cohort was 1,140 mg (range: 200–2,975 mg). In 2 patients, genotyping could not be conducted because of poor DNA quality.

Patients in both cohorts received paclitaxel weekly or every 3 weeks in different combination regimens. Also, patients receiving chemotherapy in combination with radiotherapy as a preoperative regimen for resectable esophageal cancer were included (20). These patients received a weekly dose of 50 mg/m². The cumulative dose of paclitaxel did not differ between CYP3A4*22 carriers and noncarrier in both cohorts together (P = 0.30). In the training set, the cumulative dose did not differ between CYP3A4*22 carriers and noncarriers in males (P = 0.93) and females (P = 0.66). In the validation cohort, the cumulative dose was also not significantly different between CYP3A4*22 carriers and noncarriers, both in males (P = 0.66) and females (P = 0.12).

Systemic exposure (AUC) of paclitaxel was significantly associated with severity of neurotoxicity in both females and males (P ≤ 0.001; Table 2). Also, T_>0.05 and the maximum observed concentration after administration (C_max) were significantly associated with neurotoxicity (P ≤ 0.001; Table 2). Paclitaxel exposure (_logAUC) and development and severity of neurotoxicity showed a relationship (R = 0.52; P < 0.000001).

All tested genetic variants were in Hardy–Weinberg equilibrium (Supplementary Table S1).

In the exploratory cohort, neurotoxicity was observed in 106 of 261 patients (41%). There were significantly more females than males who developed neurotoxicity (67% vs. 33%; P < 0.0001). In this cohort, severity of neurotoxicity was differently distributed between female CYP3A4*22 carriers and noncarriers (P = 0.043), whereas male CYP3A4*22 carriers and noncarriers had an even distribution of neurotoxicity (P = 0.90; Table 3). The other SNPs tested showed no association with severity of neurotoxicity (Table 4). CYP3A4*22 carrier status in both males and females was not associated with pharmacokinetic parameters (unbound CL, AUC, T_>0.05, and C_max) of paclitaxel (data not shown). There was no influence of CYP2C8*3 or CYP2C8*4 carrier status on pharmacokinetics of paclitaxel (data not shown). Furthermore, the ABCB1 3435C>T SNP was also not associated with paclitaxel pharmacokinetics (data not shown). Cumulative dosages of patients with grade 3 neurotoxicity are summarized in Table 5.

To confirm the relationship observed in the exploratory cohort between CYP3A4*22 carrier status and severity of neurotoxicity, we studied this association in an independent validation cohort. In this cohort, 113 of 239 patients (47%) developed neurotoxicity. Significantly, more females than males developed neurotoxicity (65% vs. 29%; P < 0.0001). In this cohort, in both females and males, the grade of neurotoxicity was differently distributed in CYP3A4*22 carriers than in CYP3A4*22 noncarriers (P = 0.036 and P = 0.025, respectively; Table 3). The risk of developing grade 3 neurotoxicity was higher in CYP3A4*22 carriers than in noncarriers (OR = 19.1; P = 0.001; 95% confidence interval = 3.3–110), confirming the observation in females in the exploratory cohort and showing this time a comparable effect in males. Cumulative dosages of patients with grade 3 neurotoxicity are summarized in Table 5.

Grade 3 neurotoxicity may be a result of the cumulative dose of paclitaxel, and is a reason to discontinue paclitaxel treatment. Therefore, we also conducted an exploratory Cox regression analysis in patients of both cohorts together because of the small number of neurotoxicity grade 3, taking cumulative dose into account. In this analysis, the occurrence of grade 3 neurotoxicity was included as the event, whereas the cumulative dose of paclitaxel was included as the time-to-event variable. The prognostic impact of CYP3A4*22 was then evaluated and adjusted for cohort and gender. Again, neurotoxicity grade 3 was more often seen in CYP3A4*22 carriers (HR = 22.1, 95% confidence interval = 4.7–105, P < 0.001).

We found that CYP3A4 was expressed in human dorsal root ganglia in 2 separate patient samples as shown by amplified products that were detected by qRT-PCR (Supplementary Fig. S1). CYP3A4 transcripts were expressed with a C_t value of 28.71 ± 0.074 in dorsal root ganglia of patient 1 and 28.27 ± 0.009 in the dorsal root ganglia of patient 2, relative to the control gene, GAPDH, which was expressed with a C_t value of 21.96 ± 0.008 in dorsal root ganglia of patient 1 and 25.40 ± 0.090 in the dorsal root ganglia of patient 2.

In this study, we showed that systemic exposure to paclitaxel was highly correlated with the development of (severe) neurotoxicity. Importantly, systemic exposure to paclitaxel measured during one course is already predictive for both development and severity of neuropathy in males and females. This result is in line with the study of Mielke and colleagues who observed that the time above the threshold of 0.05 μmol/L paclitaxel was associated with development of neuropathy (12) and the study of Greenand colleagues, reporting a relationship between paclitaxel exposure and neurotoxicity (13).

In addition, we showed that females carrying the reduced function CYP3A4*22 variant allele had an increased risk of developing severe neurotoxicity. This was shown in our exploratory cohort, and subsequently confirmed in the independent validation cohort. Interestingly, in the exploratory cohort, only female carriers of CYP3A4*22 were found to have an increased risk of neurotoxicity, whereas in the validation cohort, there was an increased risk of grade 3 neuropathy in both males and females carrying the CYP3A4*22 allele. It should be noted that the low incidence of grade 3 neurotoxicity in our cohort makes the absolute risk of developing neurotoxicity during paclitaxel treatment difficult to interpret. The lack of statistical significance in the male CYP3A4*22 carriers in the exploratory cohort could possibly be explained by the fact that there were no male patients with grade 3 neurotoxicity in this cohort. Because of the observed discrepancy between exploratory and validation cohorts, it is not yet possible to present a conclusion on the risk of neurotoxicity for male CYP3A4*22 carriers.

Recently, it was shown that taxane-induced neuropathy is not a pharmacodynamic marker of treatment outcome (21). Therefore, a predictive marker for neuropathy during paclitaxel therapy could be of particular clinical usefulness. CYP3A4*22 carrier status has the potential to aid medical oncologists in selecting female patients sensitized to development of neurotoxicity during paclitaxel therapy. It would be clinically relevant to predict grade 3 (or higher) neurotoxicity because this toxicity often leads to dose reductions or preliminary discontinuation of paclitaxel therapy. For a patient in whom severe neurotoxicity should absolutely be avoided (e.g., those with disabling peripheral neurologic disorders, or those with preexisting neuropathy from previous chemotherapy), pretreatment knowledge of the CYP3A4*22 carrier status might help choosing the appropriate (chemo-) therapy for an individual patient. If alternative drugs are available, these patients should preferably not be exposed to paclitaxel.

In this study, systemic pharmacokinetic parameters did not differ between CYP3A4*22 carriers and noncarriers. This is in contrast with the altered tacrolimus pharmacokinetics observed in CYP3A4*22 carriers (9) and increased cholesterol reduction in simvastatin-treated patients who are CYP3A4*22 carriers (8). It is also in contrast to the increased risk of delayed graft function and worse creatinine clearance in cyclosporine-treated kidney patients who carry the CYP3A4*22 allele (10). This discrepancy could possibly be due to the fact that CYP3A4 in the liver is only a minor elimination pathway of paclitaxel when compared with CYP2C8, which indeed has a 2.3-fold greater metabolite production than CYP3A4 (22). However, none of the CYP2C8 SNPs nor ABCB1 3435 C>T showed an association with paclitaxel pharmacokinetics or the development of paclitaxel-induced neuropathy in our study. These findings are in line with several other pharmacogenetic studies in paclitaxel-treated patients (1, 23). Bergmann and colleagues also did not find an association between CYP2C8*3, and ABCB1 3435 C>T and sensory neuropathy and overall survival in patients with ovarian cancer (24). More recently, these authors reported that paclitaxel clearance was 11% lower in CYP2C8*3 carriers than in noncarriers (25). In our study, we did not observe pharmacokinetic differences between patients, also not in a subgroup analysis of patients with ovarian cancer (data not shown). We also could not confirm the findings by Leskala and colleagues and Green and colleagues, who reported an association between CYP2C8*3 and neurotoxicity in patients treated with paclitaxel (13, 26). Because of the discrepancy in results in these studies, the potential of genetic variants to predict individual paclitaxel pharmacokinetics is still under debate. We are currently conducting a large study associating 1,936 relevant SNPs in drug-metabolizing enzymes and transporters (DMET) with paclitaxel pharmacokinetics to elucidate this issue further.

A higher systemic exposure to paclitaxel could not explain the higher incidence of neurotoxicity seen in CYP3A4*22 carriers. Therefore, a possible explanation might be that the effect of CYP3A4*22 is not systemic but localized in the peripheral neurons. Ghosh and colleagues suggested a potential cytoprotective role for CYP3A4*22 in central nerves (27, 28). It was already known that CYP3A4 is expressed by endothelial cells in the blood brain barrier (27), but these authors observed that CYP3A4 was expressed in approximately 75% of neurons of epileptic brain tissue (28). In CYP3A4-transfected cells, incubated with toxic concentrations of carbamazepine, a remarkably reduced cell death was observed, suggesting a cytoprotective effect of CYP3A4 (28). In the current study, we found that CYP3A4 is also expressed in peripheral nerves in particular dorsal root ganglia and this could explain the possible cytoprotective mechanism against toxic CYP3A4 substrates such as paclitaxel. This localization of CYP3A4 in peripheral neurons provides a potential mechanistic explanation for the observation that female carriers of the CYP3A4*22 variant allele, which is associated with reduced CYP3A4 function, have a higher risk to develop severe neuropathy in our study. The observation that CYP3A4*22 is also expressed in peripheral neurons is only an indication that CYP3A4 might protect against neurotoxicity during paclitaxel therapy. It is, however, too early to provide a mechanistic explanation for our observations. Further research into the underlying biologic principles of this potential protective role of CYP3A4 is needed.

Unfortunately, multivariate analyses were not warranted in both cohorts because of the relatively low incidence of grade 3 neurotoxicity. Therefore, we cannot exclude the possibility that the effect of CYP3A4*22 on neurotoxicity is influenced by confounders. Therefore, our preliminary findings have to be validated in future research to explore the clinical potential of CYP3A4*22 as a marker for development of neurotoxicity.

Recently, several other SNPs identified in large genome-wide association studies (GWAS) were associated with paclitaxel-induced neuropathy. Schneider and colleagues presented results of an interim analysis of their E5103 phase III trial, comparing chemotherapy plus concurrent bevacizumab or chemotherapy with concurrent and sequential bevacizumab as adjuvant treatment for early-stage breast cancer (29). They found that SNPs in RWDD3 (rs2296308) and TECTA (rs1829) were associated with the time of first reporting ≥grade 2 neuropathy. Not much is known about RWDD3 or TECTA, but involvement of TECTA in sensoric hearing loss and cellular stress has been suggested (30), making an association with the development of neuropathy biologically plausible. Bergmann and colleagues aimed to validate these findings in an independent cohort but could not confirm any association between these SNPs and time to neurotoxicity (31). In another GWAS, Baldwin and colleagues found a SNP in FGD4 (rs10771973) to be associated with the onset of peripheral neuropathy and validated this finding in an independent European and African American cohort (32). In this study, there was also evidence that 2 other markers in EPHA5 (rs7349683) and FZD3 (rs7001034) were associated with onset or severity of paclitaxel-induced neuropathy.

Neurotoxicity is not only a side effect attributable to taxanes, but is also frequently seen in treatments with several other drugs metabolized by CYP3A4. For example, bortezomib and thalidomide, used for the treatment of multiple myeloma, have incidences of grade ≥3 neurotoxicity of 8% and 5%, respectively (33). Also, vincristine (34) and ixabepilone (35) have been reported to frequently cause severe peripheral neuropathy. Therefore, further clinical research should elucidate the possible effects of CYP3A4*22 carrier status on the development of neurotoxicity during treatment with these agents.

In conclusion, we identified a relationship between CYP3A4*22 carrier status in women and occurrence of neurotoxicity during paclitaxel therapy. In our study, female carriers of CYP3A4*22 had an increased risk of neurotoxicity, although paclitaxel pharmacokinetic profiles were similar to those of noncarriers. This novel SNP could potentially be used as a predictive factor for paclitaxel-induced neurotoxicity in females, but further research is necessary to confirm our preliminary findings. Also, the predictive value of CYP3A4*22 carrier status in other CYP3A4-metabolized drugs remains to be established.

No potential conflicts of interest were disclosed.

Conception and design: A.-J.M. de Graan, A. Sparreboom, J. Verweij, R.H.J. Mathijssen, R.H.N. van Schaik

Development of methodology: A.-J.M. de Graan, J.A. Sprowl, R.H.J. Mathijssen, R.H.N. van Schaik

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): A.-J.M. de Graan, J.A. Sprowl, A. Sparreboom, F.K. Engels, F.A.L.M. Eskens, E. AC Wiemer, J. Verweij, R.H.J. Mathijssen, R.H.N. van Schaik

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): A.-J.M. de Graan, L. Elens, J.A. Sprowl, A. Sparreboom, L.E. Friberg, B. van der Holt, P.J. de Raaf, J. Verweij, R.H.J. Mathijssen, R.H.N. van Schaik

Writing, review, and/or revision of the manuscript: A.-J.M. de Graan, L. Elens, J.A. Sprowl, A. Sparreboom, L.E. Friberg, B. van der Holt, P.J. de Raaf, P. de Bruijn, F.K. Engels, F.A.L.M. Eskens, E. AC Wiemer, J. Verweij, R.H.J. Mathijssen, R.H.N. van Schaik

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): A.-J.M. de Graan, P.J. de Raaf, E. AC Wiemer

Study supervision: R.H.J. Mathijssen, R.H.N. van Schaik

The authors thank Samira Elbouazzaoui and Marina Hooghart for genotyping procedures, Inge Ghobadi Moghaddam-Helmantel, Mei-Ho Lam, and Walter Loos for pharmacokinetic analyses, and their physicians for including patients in the clinical trials. The authors also acknowledge the use of tissues procured by the NDRI with support from NIH grant 5 U42 RR006042.

This study was sponsored by the Dutch Cancer Society (EMCR 2010-4664), NIH (grant 5 U42 RR006042), USPHS Cancer Center Support Grant 3P30CA021765, and NCI Grant 5R01CA151633-01.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.