A pharmacogenetic study of docetaxel and thalidomide in patients with castration-resistant prostate cancer using the DMET genotyping platform

Abstract

The anticancer agent docetaxel shows significant inter-individual variation in its pharmacokinetic and toxicity profile. Thalidomide is an active anticancer agent and also shows wide pharmacological variation. Past pharmacogenetic research has not explained this variation. Patients with prostate cancer enrolled in a randomized phase II trial using docetaxel and thalidomide versus docetaxel alone were genotyped using the Affymetrix DMET 1.0 platform, which tests for 1256 genetic variations in 170 drug disposition genes. Genetic polymorphisms were analyzed for associations with clinical response and toxicity. In all, 10 single-nucleotide polymorphisms (SNPs) in three genes were potentially associated with response to therapy: peroxisome proliferator-activated receptor-δ (PPAR-δ), sulfotransferase family, cytosolic, 1C, member 2 (SULT1C2) and carbohydrate (chondroitin 6) sulfotransferase 3 (CHST3). In addition, 11 SNPs in eight genes were associated with toxicities to treatment: spastic paraplegia 7 (pure and complicated autosomal recessive) (SPG 7), CHST3, cytochrome P450, family 2, subfamily D, polypeptide 6 (CYP2D6), N-acetyltransferase 2 (arylamine N-acetyltransferase) (NAT2), ATP-binding cassette, sub-family C (CFTR/MRP), member 6 (ABCC6), ATPase, Cu + + transporting, alpha polypeptide (ATP7A), cytochrome P450, family 4, subfamily B, polypeptide 1 (CYP4B1) and solute carrier family 10 (sodium/bile acid cotransporter family), member 2 (SLC10A2). Genotyping results between drug metabolizing enzymes and transporters (DMET) and direct sequencing showed >96% of concordance. These findings highlight the role that non-CYP450 metabolizing enzymes and transporters may have in the pharmacology of docetaxel and thalidomide.

Introduction

Individualizing therapy for patients who are treated with pharmaceutical agents is an overarching goal of basic and clinical research in this first part of the twenty-first century. In no area of medicine is this goal more critical than in medical oncology. Genetic variation in genes encoding proteins involved in the metabolism and transport of drugs may account for some of the wide variation observed in the response to, and toxicity from, anticancer agents. Pharmacogenetic research holds the promise of reducing the unpredictable inter-individual variability in the efficacy and toxicity from using chemotherapy drugs.

Enzymes and transporters mediate the absorption, distribution, metabolism and excretion of endogenous as well as exogenous substrates, including drugs. However, for only a relatively small number of absorption, distribution, metabolism and excretion genes have single nucleotide polymorphisms (SNPs) and other genetic variations been identified that mediate the efficacy and toxicity of chemotherapy drugs. These include irinotecan and UGT1A1, 6-mercaptopurine and thiopurine methyltransferase, 5-fluoruracil and dihydropyrimidine dehydrogenase and tamoxifen and cytochrome P450, family 2, subfamily D, polypeptide 6 (CYP2D6).¹

Docetaxel (Taxotere) is a semisynthetic taxane that is currently used as first-line therapy in castration-resistant prostate cancer (CRPC). It is thought that the major route of metabolism is through the CYP3A family of enzymes.^2–4 The pharmacokinetic profile of docetaxel shows wide inter-individual variability, with measures such as clearance and area under the concentration curve varying as much as sixfold between patients.^5,6 Even in an ethnically homogenous patient population, variability is as high as 3.5-fold.⁷ Research using probes of CYP3A4 activity have been unsuccessful in explaining this large variation.^7–10 In fact, the activity of CYP3A4/3A5 accounts for as little as two-thirds of the inter-individual variable pharmacology of the drug.^6,11 This led researchers to look at other metabolizing enzymes, including glutathione-S-transferases, as well as drug transporters such as ABCB1 (P-glycoprotein), to help explain this large variability in drug pharmacology.^11–14 The pharmacogenetic factors mediating docetaxel pharmacokinetics have not been fully characterized.¹⁴ We have previously proposed that a more comprehensive analysis into the expression and genetic polymorphisms of multiple drug enzymes and transporters may improve our understanding of the pharmacokinetics of docetaxel.¹⁵

Thalidomide, a potent teratogen that causes dysmelia in humans, has been approved for the treatment of multiple myeloma and is under active clinical analysis for its use in treating other malignancies. The anticancer mechanism of action of thalidomide is complex and has not yet been fully charecterized. In vitro data suggest that it inhibits angiogenesis. Non-enzymatic spontaneous hydrolysis is a major route of metabolism, with at least 12 different products identified by this process.¹⁶ However, enzymatic activation of these hydrolysis products is necessary for thalidomide to exert its anticancer activity.^17–19 Two metabolites, 5-hydroxythalidomide and 5’-hydroxythalidomide, are formed by CYP2C19 and to a lesser extent CYP2B6.²⁰ CYP1A1 may have a role in parent drug activation.²¹ Other metabolism products, including dihydroxylated and glucuronide conjugates, have been found in animal models.²² Past pharmacogenetic studies analyzing whether SNPs in CYP2C19 accounted for the inter-individual variation in pharmacokinetics and toxicities have been inconclusive.^19,23

Testing for the presence of genetic variations and SNPs is typically performed by DNA amplification using PCR and sequencing of specific gene fragments. This work is specialized and labor intensive, which in turn limits the institutions and investigators who can pursue genotyping and pharmacogenetic research. One of the main obstacles facing pharmacogenetic researchers is the lack of a proven, scalable genotyping technology that screens for SNPs in multiple genes with sufficient sensitivity and high-throughput capabilities to be useful in translational and clinical research. The new Affymetrix drug-metabolizing enzyme and transporter (DMET) genotyping platform (Affymetrix, Inc., Santa Clara, CA, USA) screens for 1256 genetic variations in 170 absorption, distribution, metabolism and excretion genes, including 50 CYP450 genes, 72 non-CYP genes, 39 transporters and 9 other proteins involved in drug disposition (Table 1).²⁴ We tested this new genotyping platform, termed DMET, using patient samples and clinical data from our previously completed phase II trial in patients with CRPC to analyze the pharmacogenetics of thalidomide and docetaxel.

Table 1.

Genes included in the DMET 1.0 panel

P450 Enzymes		Non-P450 enzymes			Transporters		Other
CYP1A1	CYP4F2	ADH1A	CROT	NQO1	ABCB1	SLC15A2	AHR
CYP1A2	CYP4F3	ADH1B	DYPD	SULT1A1	ABCB4	SLC16A1	NR1I2
CYP1B1	CYP4F8	ADH4	EPHX1	SULT1A2	ABCB7	SLC19A1	NR3C1
CYP2A13	CYP4Z1	ADH5	FMO1	SULT1A3	ABCB11	SLC22A1	PPARD
CYP2A6	CYP46A1	ADH6	FMO2	SULT1B1	ABCC1	SLC22A2	PPARG
CYP2A7	CYPA1	ADH7	FMO3	SULT1C1	ABCC2	SLC22A3	RALBP1
CYP2B6	CYP51A1	ALDH1A1	FMO4	SULT1C2	ABCC3	SLC22A4	RPL13
CYP2B7P1	CYP7A1	ALDH2	FMO5	SULT1E1	ABCC4	SLC22A5	SPG7
CYP2C9	CYP7B1	ALDH3A1	FMO6	SULT2A1	ABCC5	SLC22A6	XDH
CYP2C8	CYP8A1	AOX1	GSTA1	SULT2B1	ABCC6	SLC22A8
CYP2C19	CYP8B1	COA	GSTA2	SULT4A1	ABCG2	SLC28A1
CYP2C18	CYP11A1	CHST1	GSTA3	TPMT	ATP7A	SLC28A2
CYP2D6	CYP11B1	CHST2	GSTA4	UGT1A1	ATP7B	SLC28A3
CYP2E1	CYP11B2	CHST3	GSTA5	UGT1A3	SLC5A6	SLC29A1
CYP2F1	CYP17A1	CHST4	GSTM2	UGT1A4	SLC7A5	SLC29A2
CYP2J2	CYP19A1	CHST5	GSTM3	UGT1A6	SLC7A7	SLCO1A2
CYP281	CYP20A1	CHST6	GSTM4	UGT1A7	SLC10A1	SLCO1B1
CYP3A4	CYP21A2	CHST7	GSTO1	UGT1A8	SLC10A2	SLC1B3
CYP3A43	CYP24A1	CHST8	GSTP1	UGT2A1	SLC13A1	SLCO2B1
CYP3A5	CYP26A1	CHST9	HNMT	UGT2B11	SLC15A1
CYP3A7	CYP26C1	CHST10	MAOB	UGT2B15
CYP4A11	CYP27A1	CHST11	NAT1	UGT2B28
CYP4B1	CYP27B1	CHST13	NAT2	UGT2B4
CYP4F11	CYP39A1	COMT	NNMT	UGT8
CYP4F12	POR

Abbreviations: DMET, drug-metabolizing enzymes and transporters; SNP, single-nucleotide polymorphism.

The DMET assay and panel genotype samples for variations in 170 drug disposition genes, including 50 CYP450 genes, 72 non-CYP genes, 39 transporters and 9 other proteins involved in drug disposition. Assay genotyping identifies homozygote wild-type, homozygote variant or heterozygote alleles in 1256 SNPs and other genetic variation sites in these genes.

Patients and methods

Patients

Patients diagnosed with CRPC were enrolled in a randomized phase II clinical trial comparing docetaxel to docetaxel with thalidomide. The results of this trial have been previously reported.²⁵ In brief, 74 patients with CRPC were enrolled with a 2:1 randomization to receive either docetaxel and thalidomide, or docetaxel alone. Docetaxel was dosed at 30m gm⁻² weekly for 3 weeks followed by a 1-week rest. For patients assigned to the combination arm, they received docetaxel on the same schedule, as well as thalidomide at 200 mg orally each day. In all, 49 patients were in the combination arm and 25 patients were in the control arm.²⁵ The study was approved by the institutional review board of the National Cancer Institute.

Response to treatment was determined using the prostate-specific antigen Working Group consensus criteria of a drop in serum prostate-specific antigen of ≥50% with no other evidence of disease progression. Toxicities were graded using the Common Toxicity Criteria (version 2.0) of the Cancer Therapy Evaluation Program/National Cancer Institute. For this pharmacogenetic study, only adverse reactions graded as 3 (no grade 4 adverse events were experienced on this trial) and identified as being possibly, probably or definitely related to study medications were included. Events of bone pain, which is common in this patient population, as well as events of hyperglycemia (which were likely caused by the use of dexamethasone as a docetaxel pre-medication) were not included. Finally, events of venous thrombosis were not included because the trial was modified to add low-molecular-weight heparin to patients receiving thalidomide after an interim analysis found a high rate of such events early in the trial. Toxicity was used as a pharmacodynamic measure as well as a surrogate marker for drug pharmacokinetics as, at least in the case of docetaxel, clearance and pharmacokinetics have been previously found to correlate well with drug toxicity.^26–28

Patient samples

For 47 of the 74 originally enrolled patients, stored buffy coat samples were archived and available for DNA extraction. This included 33 patients on the combination arm of the trial and 14 patients on the docetaxel-alone arm. Blood was collected from patients, and the buffy coat layers were resuspended and frozen in Trizol (Invitrogen, Carlsbad, CA, USA). Genomic DNA was extracted from serum or white blood cell buffy coat layers of whole blood from patients using either the QiAamp Ultrasens Viral DNA kit (serum) or the QIAamp DNA blood kit (buffy coat) as described by the manufacturer (Qiagen, Valencia, CA, USA).

Preparation and processing of the drug-metabolizing enzyme and transporter platform (DMET)

DNA processing and genotype identification for each patient sample were performed using the Affymetrix DMET platform as recently described.²⁴ Genotypes were determined for each SNP site, reported as homozygous wild-type, heterozygous, homozygous variant or ‘no call’. Four samples were run in duplicate and the repeatability from both experiments was compared.

Direct genotyping

Genotyping was conducted through direct sequencing for variants in eight genes that were previously thought to correlate with prostate cancer risk or the pharmacology of docetaxel and thalidomide (Table 5). PCR was performed on gene fragments for each variant site using primers and PCR conditions as previously performed for CYP17,²⁹ CYP2C19,³⁰ CYP1B1,³¹ CYP3A5,³² ABCB1³³ and ABCG2.³⁴ A 50-μl reaction was prepared for PCR amplification. The reaction consisted of 1 × PCR buffer, 2 mmoll⁻¹ of each of the four deoxynucleotide triphosphates, 1.5 mmol l magnesium chloride, 20 mmoll⁻¹ of the forward and reverse primers and 1 unit of Platinum Taq DNA polymerase. Direct nucleotide sequencing was performed using the dRhodamine Terminator Cycle Sequencing Ready Reaction kit or the Big Dye Terminator Cycle Sequencing Ready Reaction kit version 1.1 on an ABI 310 genetic analyzer (Applied Biosystems, Foster City, CA, USA). The CYP17 polymorphism was analyzed using the PCR-restriction fragment length polymorphism assay. For peroxisome proliferator-activated receptor-δ (PPAR-δ), genomic DNA containing the 798T >C SNP site was amplified using PCR. Primer sequences were as follows: F1, 5’-AGCAAC ACTCACCGCCGTGTG-30 and R1, 5’-ACCTCTGACATCCCCA TCCCTT-3’. After an initial denaturation at 94 °C for 5m in, 40 cycles of amplification with denaturation at 94 °C for 30 s, annealing at 64 °C for 30 s and extension at 72 °C for 30 s was performed, followed by a final extension step of 7 min at 72 °C. Direct nucleotide sequencing PCR was conducted using the Big Dye Terminator Cycle Sequencing Ready Reaction kit V1.1 on an ABI Prism 3130 × l genetic analyzer (Applied BioSystems). The primer sequences for these reactions were as follows: F2 5’-TGGCTTTGCCGGTGAGGA TGC-3’ and R2 5’-GCAGTCAGCAAGGAGCCCAG-30. Genotyping results by direct sequencing were previously reported and not found to correlate with clinical parameters.²⁵

Table 5.

Concordance between DMET and direct sequencing genotyping

Gene	SNP	Concordance
CYP17	A1A1	100
CYP2C19	*2	97
CYP2C19	*3	100
CYP1B1	*3	96
CYP3A5	*3C	100
ABCB1	2677G>T	96
ABCG2	Q141K	100
PPAR-δ	Rs2076167	100

Abbreviations: DMET, drug-metabolizing enzymes and transporters; SNP, single-nucleotide polymorphism.

Individual genotyping results using DMET and direct sequencing were compared, and concordance between these two data sets was compared. A result of 100% indicates that the results were in complete agreement.

Statistical analysis

In an exploratory manner, the association between the SNP parameters and clinical response or toxicity were tested using Mehta’s modification to Fisher’s exact test.³⁵ Results of potential interest were limited to those in which the P-value was < 0.01. P-values < 0.001 would suggest even stronger association. Results are to be interpreted as hypothesis generating and in the context of a large number of exploratory evaluations performed.

Results

Patients

Of the 75 patients in the original study, sufficient DNA was available from 47 individuals. Demographic and clinical data on these patients are presented in Table 2. The proportion of patients in each arm differed from the original 2:1 randomization, with slightly more patients on the combination arm (70 vs 67% in the parent study). Of the 14 patients who received docetaxel alone and were included in this pharmacogenetic study, 50% had a clinical response (partial or complete response), compared with a response rate of 37% in the overall study. Of the patients who received thalidomide and docetaxel, 73% of the patients included in this study had a clinical response compared with 53% in the parent trial. Whether a sample was archived and was of sufficient volume to enable DNA extraction was in no way related to the patient’s response on the trial, and the higher response rates in this subset of patients cannot be readily explained. Toxicities experienced during the trial of grade 3 are listed in Table 3, along with the percentage of study patients experiencing each toxicity.

Table 2.

Patient characteristics

No. of patients	47
Age, yearsa	66.1 (41–79)
Race
Caucasian	43
African American	3
Asian	1
Gleason at diagnosisa	7.9 (6–10)
Baseline PSAa	252 (5–1068)
Trial armb
Docetaxel	14
Docetaxel and thalidomide	33
Response
Docetaxel	7 (50%)
Docetaxel and thalidomide	24 (73%)

^aData are expressed as the median value and range.

^bThe original trial consisted of 25 patients receiving monotherapy, and 49 patients receiving the combination. There were 9/24 (37.5%) responses for the single agent, and 25/47 (53%) responses for the combination.²⁵ We included the maximum number of patients from the original trail who had a sufficiently large DNA sample for analysis with DMET.

Characteristics of the 47 patients in this pharmacogenetic study. The parent study enrolled 75 patients in a 2:1 randomization between the two arms. Response rates using the prostate-specific antigen (PSA) Working Group consensus criteria were slightly higher in the subset of patients included in this study when compared with the patients in the parent study.

Table 3.

Incidence of adverse reactions

Adverse reaction	No. of patients	Incidence (%)
Constitutional
Fatigue	3	6
Myalgias/weakness	3	6
Allergic reaction	2	4
Cardiovascular
Arrhythmia	2	4
Chest pain/ischemia	2	4
Neurological
Dizziness	1	2
Syncope	4	9
Hallucinations	1	2
Seizure	1	2
Hematological
Neutropenia	2	4
Infection	7	15
Anemia	2	4
Thrombocytopenia	1	2
Bleeding	3	6
Metabolism
Electrolyte disturbances	9	19
Hepatic dysfunction	3	6
Gastrointestinal
Constipation	1	2
Diarrhea	3	6
Ileus	1	2
Pulmonary
Dyspnea	5	11

Type and incidence rates of adverse reactions included in the toxicity analysis of the study using the National Cancer Institute (NCI) Common Toxicity Criteria (version 2.0). Only adverse events graded as 3 (no grade 4 adverse reactions were experienced on this trial), and those that were judged to be possibly, probably or definitely related to study medication were included.

DNA extraction and processing of the drug-metabolizing enzyme and transporter platform (DMET)

Extracted DNA content from archived samples ranged between 3.0 and 930.9 ng. Although the recommended minimal DNA sample content for optimal DMET analysis was 2000 ng, we obtained call rates of between 88 and 98% for all SNP loci for those samples containing at least 92 ng of DNA. No-call rates in those samples with < 92 ng (n = 15) were lower, ranging between 21 and 91%. Four samples were run in duplicate and the results were compared between the two genotyping results across the 1256 genetic variation sites. Repeatability of results ranged between 98.1 and 99.9%.

Single-nucleotide polymorphisms in the 170 DMET genes were identified. An SNP is defined as a genetic variant that occurs in at least 1% of a population. Frequency of the variant alleles was in general < 10% (or > 9 0 % in those cases in which the reference ‘wild-type’ allele is the less common one). For 373 of the 1256 polymorphisms, or 30% of all loci, variant allelic frequencies were between 10 and 90% (Figure 1).

Figure 1.

Frequency of variant alleles. The frequencies of the single-nucleotide polymorphism (SNP) variant alleles in the 170 genes are graphed. The frequency of variant alleles was in general < 10% (or >90% in those cases in which the reference ‘wild-type’ allele is the less common one). For 373 of the 1256 SNPs, or 30% of all loci, variant allelic frequencies were between 10 and 90%.

Single-nucleotide polymorphisms (SNPs) associated with clinical response

In an initial screening procedure that used the generalized Fisher’s exact test, 28 SNPs were identified in which the P-value was < 0.05. Given the exploratory nature of this study, and to limit false-positive results, we further limited reported results to those with a P-value of < 0.01. Using this criterion, 10 SNPs in three genes were found to be associated with response to therapy: PPAR-δ, sulfotransferase family, cytosolic, 1C, member 2 (SULT1C2) and carbohydrate (chondroitin 6) sulfotransferase 3 (CHST3) (Table 4).

Table 4.

Genes and SNPs correlating with clinical response and toxicity

Gene	SNP	Variant frequency	Clinical response	Toxicity
PPAR-δ	rs6922548	0.15	0.0011	NS
	rs2016520	0.32	0.0056a	NS
	rs1883322	0.32	0.0061a	NS
	rs3734254	0.73	0.0089a	NS
	rs7769719	0.90	0.0055a	NS
CHST3	rs4148943	0.58	0.0001	NS
	rs4148947	0.43	0.0023	NS
	rs12418	0.42	0.0005	NS
	rs730720	0.40	0.0034	NS
	rs4148950	0.42	0.024b	0.006
	rs1871450	0.44	0.048b	0.006
	rs4148945	0.61	0.011b	0.010
SULT1C2	rs1402467	0.22	0.0083	NS
SPG7	rs2292954	0.73	NS	0.0004
	rs12960	0.82	NS	0.004
CYP2D6	CYP2D6*19 (2539–2542del)	0.48	NS	0.002
NAT2	rs1799931	0.92	NS	0.003
ABCC6	rs2238472	0.67	NS	0.006
ATP7A	rs2227291	0.78	NS	0.006
CYP4B1	rs4646487	0.08	NS	0.008
SLC10A2	rs2301159	0.34	NS	0.010

Abbreviations: NS, not significant(P>0.05); SNP, single-nucleotide polymorphism.

^aResults are from analyses restricted to docetaxel and thalidomide trial arm.

^bResult did not meet statistical significance as defined in the methods of P<0.01. SNPs were tested for an association with clinical response and toxicities using Mehta’s modification to Fisher’s exact test. Reported genes and SNPs were limited to those in which P <0.01. SNP identification is by reference sequence number. Frequency of the variant allele in this patient sample set is also listed.

Single-nucleotide polymorphisms (SN Ps) and toxicity

For toxicities from therapy, 11 SNPs in eight genes were found to be associated at a significance level of P < 0.01 (Table 4). These included two SNPs in spastic paraplegia 7 (pure and complicated autosomal recessive) (SPG7), three in CHST3 and single SNPs in CYP2D6, N-acetyltransferase 2 (arylamine N-acetyltransferase) (NAT2), ATP-binding cassette, sub-family C (CFTR/MRP), member 6 (ABCC6), ATPase, Cu + + transporting, alpha polypeptide (ATP7A), cytochrome P450, family 4, subfamily B, polypeptide 1 (CYP4B1) and solute carrier family 10 (sodium/bile acid cotransporter family), member 2 (SLC10A2). There was no association between patients with the CYP3A5*3C genotype, a poor metabolizing phenotype previously found to be associated with docetaxel pharmacology, and docetaxel toxicities (P = 0.51). No other variant in CYP3A5 or CYP3A4 was associated with toxicity (data not shown).

Genotyping with the drug-metabolizing enzyme and transporter platform (DMET) versus direct sequencing genotyping

Genetic variations in eight genes were identified using direct sequencing, and the results were compared with those results using the DMET platform. Concordance rates between these results ranged between 96 and 100% (Table 5). All of the polymorphisms in Tables 4 and 5 were equally distributed across the two trial arms (P>0.05;generalized Fisher’s exact test). It is to be noted that in cases in which there were disagreements between the two result sets, and sufficient DNA was available, patient samples were resequenced. In each case tested, the initial results found by direct sequencing were found to be in error, and the results from the DMET platform processing were confirmed (data not shown).

Discussion

We present the findings from an exploratory pharmacogenomic study that related SNPs with the efficacy and toxicity of docetaxel and thalidomide used in a clinical trial of patients with CRPC. These data revealed that polymorphisms in three genes (that is, PPARδ, SULT1C2 and CHST3) were associated with clinical outcome measures whereas polymorphisms in eight genes (that is, SPG7, CHST3, CYP2D6, NAT2, ABCC6, ATP7A, CYP4B1 and SLC10A2) were associated with toxicity. To our knowledge, these associations represent novel genes that may be related to either thalidomide- or docetaxel-based therapy, although this remains unclear given that no reports have indicated that any of the above genes are involved in the metabolism or disposition toward either agent. Moreover, the allelic variations in genes that are known to be involved in metabolism, transport and activity of docetaxel and thalidomide were not associated with any of the study end points. Thus, further studies are warranted to validate and explore these relationships, and we can only speculate as to why such strong relationships exist. Although all of these genes may be related to drug metabolism directly, and thus could be related to pharmacokinetics, they also participate in pathways that may affect drug action and could therefore be involved in pharmacodynamic interactions as well. Finally, some of these alleles may be related to disease state and not necessarily alter drug metabolism or disposition.

Peroxisome proliferator-activated receptor-δ, one of three members of the PPAR family, is a ligand-activated transcription factor that forms heterodimers with the retinoid X receptor, which then binds to peroxisome proliferator response elements. The gene product has been identified for its role in lipid and cholesterol metabolism,³⁶ whereas the polymorphisms in the gene have been related to inter-individual variation in cholesterol metabolism,³⁷ insulin sensitivity,³⁸ psychological disorders³⁹ and reduced height.⁴⁰ Although no study has yet evaluated PPARδ SNPs in relation to antiangiogenesis therapy, PPARδ gene expression is a major hub node in the angiogesis network,⁴¹ thus opposing thalidomide action. Given that we observed strong relationships with several SNPs in PPARδ in only those patients who received both docetaxel and thalidomide, but not docetaxel alone, it seems that allelic variation in PPARδ may influence therapeutic efficacy of the antiangiogenesis agent, thalidomide. However, this relationship is unclear as PPARδ is a nuclear receptor that is responsible for the induction of genes involved in metabolic pathways as well.

Carbohydrate (chondroitin 6) sulfotransferase 3 encodes a member of the carbohydrate sulfotransferase family, termed chondroitin 6-O-sulfotransferase, as it is known to transfer sulfate groups from 3’-phosphoadenosine 5’-phosphosulfate to the six position of N-acetylgalactosamine of chondroitin. CHST3 is involved in extracellular matrix remodeling,⁴² and may thus be involved in invasion/metastasis, or in endothelial cell adhesion.⁴³ It is also involved in drug metabolism within the liver.⁴⁴ Numerous SNPs that may affect function have been identified.⁴⁵ An intriguing finding in this study was that genetic variations in CHST3 were associated with both efficacy and toxicity, increasing the likelihood that this finding of association may indicate a role in docetaxel or thalidomide pharmacokinetics and/or pharmacodynamics. Further analyses into this gene and its possible role in drug pharmacology are warranted.

Sulfotransferase family, cytosolic, 1C, member 2 encodes a sulfotransferase that is thought to be involved in the metabolism of drugs and endogenous substrates, and is expressed both intra- and extra-hepatically as well as in the adult kidney.^46,47 Genetic variation in this gene and other cytosolic sulfotransferases were recently reported.⁴⁸ SULT1C2 seems to be downregulated during breast cancer progression in which it may be involved with steroid hormone metabolism and detoxification reactions.^49,50 Sulfation reactions are not known to be involved in the metabolism of docetaxel or thalidomide. Given the strong extrahepatic expression of SULT1C2, it is equally likely that SULT1C2 is involved in the bodily disposition toward these drugs through direct metabolism, or indirect metabolism of other endogenous/exogenous agents that influence the activity of docetaxel and/or thalidomide. It is also possible that these SNPs are merely related to disease state, and not drug activity.

Single-nucleotide polymorphisms in eight genes were found to be associated with docetaxel and thalidomide toxicities. SPG7 encodes a mitochondrial metalloprotease protein known as paraplegin. Genetic variations in SPG7 have been found in patients with spastic paraplegia, a degenerative neurological condition.⁵¹ A closely related gene, SPG6, which is also associated with spastic paraplegia, was recently found to function as a transporter.⁵² CYP2D6 and CYP4B1 are members of the P450 family of phase I metabolic enzymes. CYP2D6 is known to be involved in the metabolism of a wide range of drugs.¹ Inhibition of this enzyme was not found to affect the formation of hydroxyldocetaxel, which is thought to be mediated by CYP3A4.⁵³ W hether this enzyme is involved in the formation of other docetaxel metabolites is unknown. Human microsome studies did not find a significant role for CYP2D6 and other CYP enzymes in thalidomide metabolism.⁵⁴ It is also possible that certain CYP2D6 alleles may confer a multiple-chemical sensitivity phenotype that increases the likelihood of development of toxicities after drug treatment in general.⁵⁵

N-acetyltransferase 2 (arylamine N-acetyltransferase) encodes a polymorphic drug and xenobiotic-metabolizing enzyme that acetylates substrates. Genetic variants in NAT2 can cause a ‘slow’, ‘intermediate’ or ‘rapid’ acetylator phenotype. These variants can predict efficacy and toxicity of a wide range of drugs, including sulfasalazine⁵⁶ and isoniazid.⁵⁷ ABCC6 encodes a xenobiotic transporter from the C subfamily of the ATP-binding cassette (ABC) family of transporters. ABCC6, also termed MRP6, is known to be involved in the transport of glutathione conjugates and may mediate drug resistance to agents, such as etoposide, teniposide, doxorubicin and daunorubicin.⁵⁸ ATP7A encodes a transporter that is characterized best for its role in copper disposition, and is also thought to mediate cisplatin and carboplatin transport and resistance.⁵⁹ Finally, SLC10A2 encodes a sodium-dependent transporter highly expressed in the ileum and liver involved in the enterohepatic circulation of bile acids.⁶⁰ Polymorphisms in these enzymes and transporters might alter the disposition toward xenobiotics in general, and could potentially be involved in pharmacokinetic and pharmacodynamic relationships with docetaxel and thalidomide. However, as the mechanism behind these associations is unclear, it is difficult to ascertain the importance of these findings.

This is the first study to report that these enzymes and transporters may be involved in the disposition of docetaxel and thalidomide, and/or their metabolites. This raises the possibility that these results are false positives and reflect only random associations that are found to be statistically significant. The risk of false-positive findings can be common in pharmacogenetic research as more polymorphisms, combinations of polymorphisms or phenotypes are included in studies with limited number of study subjects.⁶¹ Given the breath of the DMET platform, with 170 genes and 1256 polymorphisms, this was a potential shortcoming in our study. To reduce the risk of false-positive results, we restricted reported results to those in which the P-value was < 0.01, using Mehta’s modification to Fisher’s exact test. As an exploratory study analyzing associations between SNPs and clinical parameters, this was meant to be hypothesis generating and provide targets for future—and larger—pharmacogenetic clinical studies.⁶¹ Research to confirm these exploratory findings are ongoing.

The promise of the emerging field of pharmacogenetics is that someday we may enter an era of ‘individualized medicine’ in which physicians can choose the right drug at the right dose for each patient. In no area of pharmacotherapy is this more important than in medical oncology. Given that we still only poorly understand the relative importance of enzymes and transporters involved in the pharmacokinetics of chemotherapy drugs,¹ the challenge facing pharmacogenetic and oncology researchers is the ability to screen for polymorphisms in many genes in order that we may better narrow our search for the key players in each drug’s metabolic pathway.

We believe that a research strategy for understanding the pharmacogenetics of anticancer agents can be pursued using exploratory screening for polymorphisms that may correlate with pharmacokinetics and/or pharmacodynamics. These exploratory studies can then be followed by high-throughput analyses using larger clinical sample sets.^11,61 New tools, such as the DMET platform, will make pharmacogenetic research available to more researchers, which may lead to answering clinical questions in medical oncology and other fields. This, in turn, may hasten the day when the promise of individualized pharmacogenetic therapy becomes a reality for more of our cancer patients.

Abbreviations

CRPC

castrate-resistant prostate cancer

DMET

drug-metabolizing enzymes and transporters

SNP

single-nucleotide polymorphism