• We are sorry, but NCBI web applications do not support your browser and may not function properly. More information
Logo of ajhgLink to Publisher's site
Am J Hum Genet. Sep 2007; 81(3): 427–437.
Published online Aug 1, 2007. doi:  10.1086/519850
PMCID: PMC1950832

Identification of Genetic Variants Contributing to Cisplatin-Induced Cytotoxicity by Use of a Genomewide Approach

Abstract

Cisplatin, a platinating agent commonly used to treat several cancers, is associated with nephrotoxicity, neurotoxicity, and ototoxicity, which has hindered its utility. To gain a better understanding of the genetic variants associated with cisplatin-induced toxicity, we present a stepwise approach integrating genotypes, gene expression, and sensitivity of HapMap cell lines to cisplatin. Cell lines derived from 30 trios of European descent (CEU) and 30 trios of African descent (YRI) were used to develop a preclinical model to identify genetic variants and gene expression that contribute to cisplatin-induced cytotoxicity in two different populations. Cytotoxicity was determined as cell-growth inhibition at increasing concentrations of cisplatin for 48 h. Gene expression in 176 HapMap cell lines (87 CEU and 89 YRI) was determined using the Affymetrix GeneChip Human Exon 1.0 ST Array. We identified six, two, and nine representative SNPs that contribute to cisplatin-induced cytotoxicity through their effects on 8, 2, and 16 gene expressions in the combined, Centre d’Etude du Polymorphisme Humain (CEPH), and Yoruban populations, respectively. These genetic variants contribute to 27%, 29%, and 45% of the overall variation in cell sensitivity to cisplatin in the combined, CEPH, and Yoruban populations, respectively. Our whole-genome approach can be used to elucidate the expression of quantitative trait loci contributing to a wide range of cellular phenotypes.

Cisplatin, a platinating agent, is commonly used to treat head and neck, testicular, lung, and gynecological cancers.13 It has been shown that cisplatin exerts its antitumor activity by binding preferentially to the nucleophillic positions on guanine and adenine of DNA, resulting in the formation of intra- and interstrand crosslinks. Eventually, crosslinks lead to DNA-strand breaks and, ultimately, to cell death.4,5 Despite its wide usage, dose-limiting toxicities—in particular, nephrotoxicity6 and neurotoxicity7—have hindered the utility of this agent. In addition, treatment-induced ototoxicity can result in dose reduction or discontinuation of cisplatin treatment.8 The incidences of cisplatin treatment–induced toxicities are highly variable and are associated with cumulative treatments or dose intensities.7

High levels of drug efflux transporters, detoxifiers, and DNA-repair proteins and a low Bax:Bcl-2 ratio have all been suggested to play a role in cisplatin resistance.9 Genetic variants in candidate genes have demonstrated an association with clinical response to or toxicity from cisplatin. For example, two common SNPs of ERCC1 are correlated with an increased risk of toxicity and with the survival of cisplatin-treated patients with non–small-cell lung cancer.10,11 Polymorphisms in cytokine-promoter genes (e.g., TNF, IL1, IL6) have been suggested to be associated with toxicities induced by treatment with 5-fluorouracil and cisplatin.12 Glutathione S-transferase genetic polymorphisms have also been associated with treatment outcomes of paclitaxel- and cisplatin-based chemotherapy.13 An illustration of candidate genes involved in the mechanism of cisplatin activity can be found at the PharmGKB Web site. Although the study of candidate genes and pathways has increased our understanding of the mechanism of action of platinating agents, our understanding of genetic variants important in determining a patient’s likelihood of response or toxicity is extremely limited. Thus, the development of comprehensive, unbiased models is critical to the identification of genetic variants and genes contributing to interindividual variation in drug effect. Genomewide approaches open up the possibility of identifying genetic and/or expression signatures that can be evaluated in clinical trials, for validation.

Previously, we used Epstein-Barr virus (EBV)–transformed B-lymphoblastoid cell lines (LCLs) derived from healthy individuals within 10 large CEPH pedigrees and demonstrated that 38%–47% of human variation in susceptibility to cisplatin-induced cytotoxicity is due to genetic components.14 To better elucidate the genetic variants important in cisplatin-induced cytotoxicity, we employed a genomewide association study, using the International HapMap cell lines derived from trios of northern and western European and Yoruban populations. These well-genotyped samples provide an extremely rich data set for genotype–drug effect correlations.15 We performed gene-expression analysis on these HapMap cell lines, using the Affymetrix GeneChip Human Exon 1.0 ST Array, and phenotyped the samples for susceptibility to cisplatin-induced cytotoxicity. The focus of this article is the description of genetic variants in two populations that contribute, through variation in gene expression, to cisplatin-induced cytotoxicity. To this end, we designed a three-way model, correlating genotype, gene expression, and cytotoxicity data, to identify potentially functional SNPs and/or haplotypes associated with cisplatin-induced cytotoxicity. Cell lines derived from individuals of European and African descent allow us to define a set of genetic variants unique to and common among the populations. The long-term goal is to identify, through a genetic signature, patients at risk for adverse events associated with these agents.

Material and Methods

Material

EBV-transformed LCLs derived from 30 CEPH trios (i.e., mother, father, and child) collected from Utah residents with northern and western European ancestry (CEU [HAPMAPPT01]) and from 30 trios collected from Yoruba in Ibadan, Nigeria (YRI [HAPMAPPT03]), were purchased from the Coriell Institute for Medical Research. Cell lines were maintained and were diluted as described elsewhere.16 Cisplatin and dimethyl sulfoxide (DMSO) were purchased from Sigma-Aldrich.

Cytotoxicity Assay

Cell-growth inhibition was evaluated at concentrations of 0, 0.5, 1, 2.5, 5, 10, 20, 40, and 80 μM of cisplatin. Cisplatin was dissolved in DMSO immediately before use. DMSO concentrations did not exceed 0.1% in the cells. The cytotoxic effect of cisplatin was determined using the nontoxic colorimetric-based assay alamarBlue, as described elsewhere.16 The concentration required to inhibit 50% of cell growth (IC50) was determined by curve fitting the percentage of cell survival against the concentration of cisplatin.

Genotype and Cytotoxicity Association Analysis

SNP genotypes were downloaded from the International HapMap database (release 21). To perform a high-quality genomewide association study, we employed several data filters. To reduce possible genotyping errors, we excluded 100,536 and 138,533 SNPs with Mendelian allele-transmission errors in 22 autosomes in the 30 CEU and 30 YRI HapMap trios, respectively. To exclude the extreme outliers and to increase the power of the association studies within our limited number of samples, we included only the SNPs that met the criteria of having three genotypes and containing a minimum of two counts for each genotype in the unrelated individuals of each population. To obtain functionally relevant SNPs, we further filtered the SNPs by their location. Only SNPs located in genes or within 10 kb up- or downstream of a gene were included. Thus, our final data set consisted of 387,417 very informative SNPs covering 22,667 well-annotated genes.

All 175 IC50 values (from 86 CEU and 89 YRI) were log2 transformed to obtain normally distributed data. The quantitative transmission/disequilibrium test (QTDT) was performed to identify any genotype-cytotoxicity associations, with the use of QTDT software.17 Because of the possible heterogeneity between and within populations, we performed association studies in these two ethnic groups separately, using sex as a covariate, and together, using sex and race as covariates. P[less-than-or-eq, slant].0001 was considered statistically significant.

Gene-Expression Assessment

RNA from 87 CEU and 89 YRI cell lines was extracted after four dilutions, by use of RNeasy Plus Mini Kits (QIAGEN). RNA quality was assessed using the RNA 6000 Nano Assay (Agilent Technologies). For each cell line, ribosomal RNA was depleted from 1 μg of total RNA by use of the RiboMinus Human/Mouse Transcriptome Isolation Kit (Invitrogen). cDNA was generated using the GeneChip WT cDNA Synthesis and Amplification Kit (Affymetrix), per the manufacturer’s instructions. cDNA was fragmented and end labeled using the GeneChip WT Terminal Labeling Kit (Affymetrix). Approximately 5.5 μg of labeled DNA target was hybridized to the Affymetrix GeneChip Human Exon 1.0 ST Array at 45°C for 16 h, per the manufacturer’s recommendation (see Affymetrix Web site for additional information). Hybridized arrays were washed and stained on a GeneChip Fluidics Station 450 and were scanned on a GCS3000 Scanner (Affymetrix). Resulting probe-signal intensities were sketch-quantile normalized using a subset of the 1.4 million probe sets. Gene-expression levels were summarized using the robust multiarray average (RMA). A constant of 16 was added for variance stabilization, and summarized signals were log2 transformed.18 This was done with signals generated on a core set of well-annotated exons (~200,000) within the Affymetrix Exon Array Computational Tool (ExACT) software package. To prevent confounding interpretations of gene-expression variation, we removed data from exons for which probe sets contained two or more probes harboring SNPs, before summarizing expression. All raw exon-array data have been deposited into Gene Expression Omnibus (GEO) (accession number GSE7761).

Genotype and Gene-Expression Association Analysis

A second QTDT test that integrated mRNA gene expression and significant SNPs found in the genotype and cytotoxicity association analysis was performed to identify possible association with gene expression. Significant SNPs generated from the genotype-cytotoxicity association in CEU, YRI, or combined populations were tested for their association with gene expression in the same population. Genes with average intensity >5 from Affymetrix GeneChip Human Exon 1.0 ST Array analysis were considered expressed and were included in this association analysis. The QTDT test was performed using gene-expression analysis in CEU and YRI populations separately and combined, with sex and race as covariates in the combined samples. We examined not only the cis-acting gene, defined as gene expression associated with SNP(s) within 5 Mb on the same chromosome, but also the trans-acting gene, defined as gene expression associated with SNP(s) on different chromosome(s) or >5 Mb away on the same chromosome. A Bonferroni correction (P<.05) that used a number of transcript clusters in the analysis was used to adjust raw P values after QTDT analysis.

Gene Expression and Cisplatin IC50 Linear-Regression Analysis

To examine the relationship between gene expression and sensitivity to cisplatin, we constructed a general linear model with log2-transformed cisplatin IC50 as the dependent variable and RMA-summarized log2-transformed gene-expression level and an indicator for sex as the independent variables. The dependent variable was transformed to satisfy the assumption of normality. Trios were treated as units of analysis, and members of different families were considered independent. The covariance structure within a trio was modeled using a Toeplitz structure with two diagonal bands, such that the trios were ordered father, then offspring, and then mother. With this covariance structure, mother and father IC50 values were independent, but the offspring’s value was allowed to covary with both the father’s and mother’s values. If a SNP was significantly associated with cisplatin IC50 and the same SNP was significantly associated with gene expression, then the above approach was used to test whether gene expression significantly predicted IC50. In the CEU population, 4 transcript clusters were tested for their expression correlation with cisplatin IC50, whereas 19 transcript clusters were tested in the YRI population, and 19 transcript clusters were tested in the combined CEU and YRI populations. With the combined approach, predictors of population and sex were included in the model. Sex was also tested in the separate CEU and YRI populations as a predictor of cisplatin IC50. P<.05 was considered statistically significant. The model was programmed using the PROC MIXED procedure in SAS/STAT software version 9.1.19 The REPEATED statement was used to model the Toeplitz covariance structure. The linkage disequilibrium (LD) of significant SNPs within each population was evaluated using Haploview version 3.32.

Multivariate Model to Predict Association of Cisplatin IC50 with Genotypes

To examine the overall genetic contributions to sensitivity of cisplatin, additional general linear models were constructed with transformed cisplatin IC50 as the dependent variable. The independent variables included all the significant SNP genotypes (with assumption of an additive genetic effect) that were selected using the three-way model in the combined populations and in the two populations independently. These SNP genotypes were significantly associated with cisplatin IC50 through their effect on gene expression. For the model of combined populations, indicators of race and sex were also included as predictors. Trios were analyzed as independent units. The covariance was modeled as described above. Models were reduced using a backward-elimination approach. SNPs included in each of the final models were statistically significant at the α=.05 level. By use of the final model, predicted transformed IC50 values were computed. For the unrelated individuals (parents from the trios and, separately, offspring from the trios), R2 was estimated between observed IC50 and the predicted IC50 from the final model. Lastly, a weighted average of the two R2 estimates was computed to quantify the amount of variation in cisplatin IC50 explained by the selected SNP genotypes.

Alternative Methods

Alternative methods were considered to evaluate the endpoints of the analytical experiments with use of a different initial step but with the same statistical cutoffs. The first alternative approach involved analyzing the SNP genotype and gene-expression association, testing the association of the significant SNPs with cisplatin IC50, and then performing linear regression between gene expression and cisplatin IC50. The second alternative approach involved evaluating the correlation between gene expression and cisplatin IC50, followed by analyzing the SNPs associated with gene expression and testing the association between those SNPs and cisplatin IC50.

Results

QTDT Genotype-Cytotoxicity Association

Elsewhere, we reported the median IC50 as 5.1 μM and 6.3 μM for cell lines derived from CEU (n=86) and YRI (n=89) trios, respectively, after exposure to increasing concentrations of cisplatin (0.5–80 μM) for 48 h.16 Interindividual variation in the IC50 was 17-fold for CEU and 49-fold for YRI.16 Using 387,417 SNPs representing 22,667 genes (~85% of genes in the entire genome), we evaluated whether genetic variation was associated with sensitivity to cisplatin by use of the IC50 value. An arbitrary P value threshold (P[less-than-or-eq, slant].0001) resulted in the identification of 96, 57, and 138 SNPs significantly associated with cisplatin IC50 in the combined, CEU, and YRI populations, respectively (table 1). These SNPs were located in or within 10 kb up- or downstream of 67, 36, and 88 genes, respectively.

Table 1.
Significant Results from the Three-Way Model with Combined, CEU, and YRI Populations[Note]

QTDT Genotype and Gene-Expression Association

We generated expression data on 176 LCLs (87 CEU and 89 YRI), using the Affymetrix GeneChip Human Exon 1.0 ST Array (an exon array). A total of 14,722 transcript clusters with a mean log2-transformed gene-expression intensity of >5, indicating expression in both CEU and YRI samples, were included in the analysis. The QTDT association analysis was conducted between gene expression and the SNPs that were significantly associated with cisplatin IC50. We found 2 cis- and 32 trans-acting relationships in the combined populations, 1 cis- and 3 trans-acting relationships in CEU, and 2 cis- and 36 trans-acting relationships in YRI (Bonferroni-corrected P<.05). Among all observed cis- and trans-acting relationships, some SNPs were significantly associated with more than one gene expression, and some gene expressions were associated with more than one SNP. Therefore, the final cis- and trans-acting relationships were represented by 8 SNPs that were significantly associated with 22 gene expressions in the combined population, by 3 SNPs that were significantly associated with 4 gene expressions in CEU, and by 11 SNPs that were significantly associated with 25 gene expressions in YRI (table (table11 and the tab-delimited ASCII file, which can be imported into a spreadsheet, of data set 1).

Linear Regression of Gene Expression and Cisplatin IC50

We examined the correlation between gene expression and cisplatin IC50, using a general linear model that was constructed to reflect the trio relationships in our data. Since some genes shared the same transcript cluster identification numbers (IDs) on the exon array, the expression of 19 transcript clusters (representing 22 genes identified above) were evaluated in the combined population. Eight genes had significant correlation with cisplatin IC50 (P<.05) (table 1). In the same manner, we found 2 and 17 genes whose expression significantly correlated with cisplatin IC50 in the CEU and YRI populations, respectively (P<.05) (table 1, current approach). A summary of SNPs that were found to be significantly associated with cisplatin IC50 through gene-expression analysis of the CEU, YRI, and combined populations is shown in table 2.

Table 2.
SNPs Associated with Cisplatin IC50 through Gene-Expression Analysis of the CEU, YRI, and Combined Populations

Alternative methods were considered to evaluate the endpoints of the analytical experiments with use of a different initial step but with the same statistical cutoff. If the initial step is the analysis of the SNP genotype and gene-expression association followed by tests of association of the significant SNPs with cisplatin IC50 and linear-regression analysis between gene expression and cisplatin IC50 (table 1, alternative approach 1), we find identical results. If, however, the initial step is an analysis of gene expression and cisplatin IC50 followed by analysis of the SNP associated with gene expression and then association analysis of those SNPs with cisplatin IC50 (table 1, alternative approach 2), the same genetic variants with additional variants are identified. Additional SNPs found through alternative approach 2 are indicated in table 1.

When the results generated from association tests between genotype, cisplatin IC50, and gene expression—as well as the linear-regression results between gene expression and cisplatin IC50—were combined, we identified 6, 2, and 10 SNPs that were significantly associated with cisplatin IC50 through regulation of 8, 2, and 17 gene expressions in the combined, CEU, and YRI populations, respectively (table 1). One example was the significant association between the genotype of rs456998 (located within an intron of the FCHSD1 gene on chromosome 5) and cisplatin IC50 (P=4×10-5) (fig. 1A). This SNP was associated with the expression of DNA-damage–inducible transcript 4 (DDIT4; on chromosome 10; P=2×10-6); never in mitosis gene A–related kinase 2 (NEK2 [MIM 604043]; on chromosome 1; P=2×10-6); serine hydroxymethyltransferase, mitochondrial (SHMT2 [MIM 138450]; on chromosome 12; P=1×10-6); WDR58 (on chromosome 16; P=2×10-7); and FRAG1 (on chromosome 11; P=3×10-6) genes (fig. 1B), whose expression also significantly correlated with cisplatin IC50 in the combined populations (P=1×10-4, 8×10-6, 2×10-4, 7×10-4, and 4×10-2, respectively) (fig. 1C).

Figure  1.
Relationship between SNP genotype, gene expression, and cisplatin IC50 in combined CEU and YRI populations. A, Genomewide association between SNP genotypes and log2-transformed cisplatin IC50 in the combined populations. The X-axis represents the chromosomal ...

In addition, we identified a significant association between cisplatin IC50 and the genotype of SNP rs8094647 located in the intron of the myosin Vb gene (MYO5B [MIM 606540]; P=2×10-5) on chromosome 18 (fig. 2A). This same SNP genotype was significantly associated with the expression of the v-myc avian myelocytomatosis viral oncogene homolog gene (MYC [MIM 190080]; P=1×10-6) located on chromosome 8 (fig. 2B). The GG genotype of rs8094647 was associated with higher MYC gene expression and lower cisplatin IC50. This was further indicated by the negative correlation found between MYC gene expression and cisplatin IC50 (P=5×10-11) (fig. 2C).

Figure  2.
Relationship between rs8094647, MYC gene expression, and cisplatin IC50 in combined CEU and YRI populations. A, Association between rs8094647 genotype and log2-transformed cisplatin IC50. B, Association between rs8094647 genotype and log2-transformed ...

In the individual CEU population, we identified rs1649942, located in the intron of the neuregulin 3 gene (NRG3 [MIM 605533]; P=7×10-5) on chromosome 10, associated with cisplatin IC50 (fig. 3A) and expression of basic helix-loop-helix domain–containing protein, class B, 3 gene (BHLHB3 [MIM 606200]; P=3×10-6) (fig. 3B). We also found a strong correlation between the BHLHB3 expression and cisplatin IC50 (P=4×10-3) (fig. 3C).

Figure  3.
Relationship between SNP genotype, gene expression, and cisplatin IC50 in CEU population. A, Genomewide association between SNP genotypes and log2-transformed cisplatin IC50 in CEU population. The X-axis represents the chromosomal location of SNPs. The ...

In the YRI population, we identified a strong association between the genotype of SNPs rs12278731 and rs9988868, both located in the intron of the GALNTL4 gene on chromosome 11, and cisplatin IC50 (P=4×10-5) (fig. 4A). These two SNPs are in complete LD (D=1; r2=1). These SNP genotypes were also significantly associated with the expression of eight genes: FAM57A (on chromosome 17; P=5×10-7), WDR54 (on chromosome 2; P=1×10-6); 6-phosphofructo-2-kinase/fructose-2,6-biphosphatase 4 (PFKFB4 [MIM 605320]; on chromosome 3; P=3×10-6), l-arginine:glycine amidinotransferase (GATM [MIM 602360]; on chromosome 15; P=3×10-6), histone deacetylase 10 (HDAC10 [MIM 608544]; on chromosome 22, P=3×10-6), mitogen-activated protein kinase 12 (MAPK12 [MIM 602399]; on chromosome 22; P=3×10-6), KIAA1799 (on chromosome 1; P=3×10-6), and phosphoglucomutase 1 (PGM1 [MIM 171900]; on chromosome 1; P=3×10-6) (fig. 4B). The CC genotype of rs12278731 was associated with higher expression of all eight genes and lower cisplatin IC50. This was further indicated by the inverse correlation found between these gene expressions and cisplatin IC50 (all P<.05) (fig. 4C).

Figure  4.
Relationship between SNP genotype, gene expression, and cisplatin IC50 in YRI population. A, Genomewide association between SNP genotypes and log2-transformed cisplatin IC50 in YRI population. The X-axis represents the chromosomal location of SNPs. The ...

Multivariate Models to Predict Cisplatin IC50 with Genotypes

To examine the overall contributions of our selected genetic variants to sensitivity of cisplatin, we constructed additional general linear models. All SNP genotypes that were significantly associated with cisplatin IC50 through their effects on gene expression were included as the independent variables to predict cisplatin IC50, the dependent variable in each population. The backward-elimination approach was applied for model reduction. In the combined population, four of the six tested SNPs were included in the final model (P<.03 for all SNPs). Specifically, rs1566347, rs2136241, rs456998, and rs7244679 were all significant predictors of cisplatin IC50. The indicator of race is not a significant predictor of cisplatin IC50 (P=.13). Computing a weighted sum of R2 from each group of unrelated individuals gives an overall estimate of R2=0.27, which indicates that 27% of the variation in cisplatin IC50 can be explained by these four SNPs in the combined populations. In the CEU population, both rs1649942 and rs7550918 were included in the final model (P=.002 and P=.001, respectively). The overall estimate of R2=0.29 indicates that 29% of the cisplatin IC50 variation can be explained by these two SNPs in the CEU population. In the YRI population, 4 of the 10 tested SNPs were included in the final model (P<.02 for all SNPs). Specifically, rs11236836, rs12278731, rs4751143, and rs935196 were all significant predictors of cisplatin IC50. The overall estimate of R2=0.45 indicates that 45% of the cisplatin IC50 variation can be explained by these four SNPs in the YRI population. The indicator of sex is not a significant predictor of cisplatin IC50 within CEU (P=.91); it is, however, a significant predictor (P=.03) in the YRI population, which agrees with our previous finding of a significant difference in cisplatin IC50 between female and male subjects within the YRI population.16

Discussion

Given the wide usage of cisplatin, the high variability of cisplatin-induced toxicity, and, most importantly, our current inability to use genetic variation to identify patients at risk for toxicity associated with cisplatin, the construction of a genetic model that provides leads for clinical testing should significantly improve the utility of this agent. The leads come from a genomewide, unbiased model that integrates genotype, gene expression, and drug sensitivity. This is the first attempt to integrate, on a genomewide scale, SNP-pattern analysis, gene-expression profiling, and a pharmacologic phenotype (cisplatin-induced cytotoxocity) for the discovery of important, novel genetic information with significant clinical potential. We identified 17 genetic variants significantly associated with cisplatin-induced cytotoxicity through the expression of 26 genes in cell lines derived from the CEU and/or YRI populations. These genetic variants can explain ~27%, ~29%, and ~45% of the overall observed cisplatin IC50 variations in the combined, CEU, and YRI populations, respectively.

Using cell lines derived from large pedigrees, our laboratory was the first, to our knowledge, to demonstrate that a significant genetic component contributed to cisplatin susceptibility.14 More recently, our laboratory has demonstrated that cisplatin-induced cytotoxicity is a highly heritable trait, with 34% of the variation in IC50 for 324 LCLs (derived from 27 large CEPH pedigrees) due to genetic factors (P=1×10-7) (S. Shukla, S. Duan, J. Badner, X. Wu, and M. E. Dolan, unpublished data). Linkage analysis revealed suggestive and significant LOD scores >2 at two chromosomal regions. Since no significant population differences in cisplatin-induced cytotoxicity were observed between the CEU and YRI populations, we compared the linkage-scan results obtained from the large CEPH pedigrees with the present association findings.16 Among all significant SNPs that were found to be associated with cisplatin IC50 through gene-expression analysis, two of them are also under suggestive linkage peaks (11p15.4-15.1; LOD >2). These SNPs, rs12278731 and rs9988868, are in complete LD (D=1; r2=1), are located within the intron regions of GALNTL4 on chromosome 11, and are significantly associated with the expression of eight genes (FAM57A, WDR54, PFKFB4, GATM, HDAC10, MAPK12, KIAA1799, and PGM1) in the YRI population. The different allele frequencies of the SNPs (FST=0.14) between the CEU and YRI populations may account for the significant association observed in YRI but not in CEU when our current approach was used.

In the current study, we evaluated the HapMap trios, of which extensive genotypic information provided reasonable power to detect genetic variants that were significantly associated with drug sensitivity. LCLs derived from two HapMap populations, CEU and YRI, were evaluated together and separately to detect significant genetic variants that contribute to cell sensitivity to cisplatin in each population and in both populations combined. Given the stringent statistical cutoffs, we did not observe any overlap of SNPs between the CEU and YRI populations. This was not surprising, given the heterogeneity between the CEU and YRI samples. Nor did we observe SNPs that overlapped between the combined populations and the individual populations. The larger number of samples in the combined populations provides more power to detect smaller genetic effects on cytotoxicity and gene expression. However, by combining the CEU and YRI data, the unique genetic findings in one population may be masked by the noise produced by the lack of a genetic effect in the other population. Thus, we interpreted the final results in each separate population as population-specific genetic variants that contribute to cisplatin toxicity, whereas those in the combined population are genetic variants important to drug-induced toxicity, regardless of population tested.

Furthermore, our three-step model links genotype, gene expression, and sensitivity to cisplatin, to decrease the potential false-discovery rate. A genomewide association test between genotype and cisplatin IC50 yielded 96, 57, and 138 significant SNPs (P[less-than-or-eq, slant].0001) associated with cisplatin IC50 from the combined and CEU- and YRI-derived cell lines, respectively. All these could be functionally important; however, the consideration of a subgroup of those that act through gene expression narrows the list to 6, 2, and 10 SNPs significantly associated with the expression of 8, 2, and 17 genes and significantly correlated with cisplatin IC50 (P<.05) in combined, CEU, and YRI populations, respectively. Our long-term goal is to identify, before chemotherapy, patients who are at risk for toxicity. Since genotyping is clinically more practical than gene-expression measurements, each approach considered (current and alternatives 1 and 2) results in a genetic signature that considers SNPs that are important in cytotoxicity and expression. The first alternative approach is also a SNP-oriented approach and produced exactly the same results as our current approach. The second alternative approach focused on correlating gene expression with cytotoxicity, then finding SNPs associated with those genes, and evaluating the final SNPs with cytotoxicity. For SNP and gene-expression association, all approaches used Bonferroni correction with the number of transcript clusters in that particular analysis. Since the second alternative approach had fewer transcript clusters in the denominator (only gene expression that was significantly correlated with cisplatin IC50), the P value cutoff was less stringent and resulted in a higher number of SNPs.

Although all 26 genes should be considered important, there are a number of genes with considerable evidence in the literature—for example, DDIT4, NEK2, and MYC. DDIT4 is a known mediator of reactive-oxygen species generation.20 The role of DDIT4 in cell sensitivity to cisplatin has been suggested in the Affymetrix Human HG-U133A chip analysis in NT2/D1 (human embryonal carcinoma) cell lines after cisplatin treatment, with up-regulation of DDIT4 gene expression after cisplatin treatment coincident with a 60% decrease in cell viability.21 Furthermore, the higher DDIT4 gene expression was also observed in the SQ20b (head and neck cancer) cell line after treatment with O6-benzylguanine plus cisplatin, compared with cisplatin alone. This higher gene expression was concomitant with higher cisplatin-induced cytotoxicity, further suggesting that higher DDIT4 expression level may lead to enhanced sensitivity to cisplatin.22 Our study showed that the TT genotype of SNP rs456998 was associated with higher DDIT4 gene expression and higher sensitivity to cisplatin, in agreement with literature evidence.22 Given the evidence of inducibility of this gene by cisplatin, we are currently evaluating DDIT4 expression changes over time after treatment.

NEK2, another gene whose expression is strongly associated with the rs456998 genotype, is required for the correct execution of mitosis.23 Zhang et al. have shown that irradiation-induced DNA damage leads to inhibition of Plk1, which leads to subsequent inhibition of Nek2 activity and thus prevents centrosome separation in HeLa and U2OS cell lines.24 The depletion of Nek2 leads to an apparent arrest in Hela cell proliferation and an increase in apoptosis, possibly as a result of mitotic errors.25,26 Our current study showed that the TT genotype of SNP rs456998 was associated with higher NEK2 gene expression and higher sensitivity to cisplatin. Using the Affymetrix U133A microarray, we found that Nek2 expression patterns were significantly different between cisplatin-sensitive and cisplatin-resistant LCLs after 8 h of cisplatin treatment (S. Duan, W. Bleibel, C. Cheng, and M. E. Dolan, unpublished data).

Another interesting finding is the significant association of the genotype of SNP rs8094647 with cisplatin IC50 (P=2×10-5), as well as with the expression of the MYC gene (P=1×10-6). The protein encoded by MYC is a multifunctional, nuclear phosphoprotein that plays a role in cell-cycle progression, apoptosis, and cellular transformation. It functions as a transcription factor that regulates transcription of specific target genes. Pretreatment with 10058-F4, a small-molecule c-Myc inhibitor, increased the chemosensitivity of HepG2 (hepatocellular carcinoma) cells to low-dose doxorubicin, 5-fluorouracil, and cisplatin.27 We found that the GG genotype of rs8094647 was associated with higher MYC gene expression and lower cisplatin IC50. Because it has been shown that cisplatin treatment down-regulates MYC gene expression at both the mRNA level (in NT2/D1 cells)21 and protein level (in Hela cervical carcinoma cells),28 we are currently investigating the relationship between sensitivity to cisplatin, SNP genotype, and MYC gene expression after cisplatin treatment.

Because of the relatively small sizes of HapMap samples, our approach focused on SNPs within or close to gene regions with fairly high allele frequencies, to increase the power to detect significant genetic variants associated with the cytotoxicity. Furthermore, the multivariate model used to test association between multiple SNPs and IC50 demonstrated the power of our approach. With a limited number of SNP genotypes (four, two, and four SNPs in the combined, CEU, and YRI populations, respectively), we were able to predict 27%, 29%, and 45% of the cisplatin IC50 variation in these LCLs. Although the full implications and biological significance of other genes and networks identified through our approach are not yet completely understood, these studies will likely help to direct clinical studies by providing a strong list of candidate genes on which to focus. They may serve as a platform for the further exploration of relevant mechanisms and may improve our understanding of the molecular basis of cisplatin-induced cytotoxicity. We are currently validating these findings in a set of unrelated CEPH cell lines.

One potential limitation of these cell lines is that EBV transformation could have an effect on drug sensitivity and/or expression profiles. Therefore, the model could miss some genes of importance because either they are not expressed in these cell lines or, upon transformation, the genes are down-regulated. Another important caveat of the model is the use of gene-expression analysis to narrow the number of associated genetic variants. The potential to miss genetic variants that affect cytotoxicity through effects on protein structure, activity, and/or degradation exists. However, the association test between cisplatin IC50 and genotype (current approach, first step) is inclusive of all genetic variants that act through these means (see the tab-delimited ASCII file of data set 1). Furthermore, the dense genotyping within HapMap cell lines allows systematic, genomewide association analysis that would not be possible in other systems. We recognize that there likely are expression and posttranslational-modification differences in various tissues. Of the 26 genes we identified in LCLs, 18 were also expressed in human kidney tissue, which is one of the major sites of cisplatin-induced toxicity (GeneCards Web site). The genetic variants that are associated with cisplatin cytotoxicity within genes in human kidney are decent candidates to evaluate with patients that experience nephrotoxicity. Furthermore, the model can be applied to uncover the genetic signatures contributing to a wide range of cellular phenotypes.

Supplementary Material

Data Set 1:

Acknowledgments

We are grateful to Wasim K. Bleilbel for excellent technical support and to Dr. Jeong-Ah Kang for maintenance of the cell lines. This Pharmacogenetics of Anticancer Agents Research Group study was supported by National Institutes of Health/National Institute of General Medical Sciences grants GM61393 and GM61374.

Web Resources

The accession number and URLs for data presented herein are as follows:

The Coriell Institute for Medical Research, http://www.coriell.org/
GEO, http://www.ncbi.nlm.nih.gov/geo/ (for all raw exon array data [accession number GSE7761])
International HapMap Database, http://www.hapmap.org/
Online Mendelian Inheritance in Man (OMIM), http://www.ncbi.nlm.nih.gov/Omim/ (for NEK2, SHMT2, MYO5B, MYC, NRG3, BHLHB3, PFKFB4, GATM, HDAC10, MAPK12, and PGM1)

References

1. Chaney S, Campbell S, Bassett E, Wu Y (2005) Recognition and processing of cisplatin- and oxaliplatin-DNA adducts. Crit Rev Oncol Hematol 53:3–11 [PubMed] [Cross Ref]10.1016/j.critrevonc.2004.08.008
2. Siddik Z (2003) Cisplatin: mode of cytotoxic action and molecular basis of resistance. Oncogene 22:7265–7279 [PubMed] [Cross Ref]10.1038/sj.onc.1206933
3. Wang D, Lippard S (2005) Cellular processing of platinum anticancer drugs. Nat Rev Drug Discov 4:307–320 [PubMed] [Cross Ref]10.1038/nrd1691
4. Zorbas H, Keppler B (2005) Cisplatin damage: are DNA repair proteins saviors or traitors to the cell? Chembiochem 6:1157–1166 [PubMed] [Cross Ref]10.1002/cbic.200400427
5. Decatris MP, Sundar S, O’Byrne KJ (2005) Platinum-based chemotherapy in metastatic breast cancer: the Leicester (U.K.) experience. Clin Oncol R Coll Radiol 17:249–257 [PubMed]
6. Daugaard G (1990) Cisplatin nephrotoxicity: experimental and clinical studies. Dan Med Bull 37:1–12 [PubMed]
7. Verstappen C, Heimans J, Hoekman K, Postma T (2003) Neurotoxic complications of chemotherapy in patients with cancer: clinical signs and optimal management. Drugs 63:1549–1563 [PubMed] [Cross Ref]10.2165/00003495-200363150-00003
8. Rybak L, Kelly T (2003) Ototoxicity: bioprotective mechanisms. Curr Opin Otolaryngol Head Neck Surg 11:328–333 [PubMed] [Cross Ref]10.1097/00020840-200310000-00004
9. Masters J, Koberle B (2003) Curing metastatic cancer: lessons from testicular germ-cell tumours. Nat Rev Cancer 3:517–525 [PubMed] [Cross Ref]10.1038/nrc1120
10. Zhou W, Gurubhagavatula S, Liu G, Park S, Neuberg D, Wain J, Lynch T, Su L, Christiani D (2004) Excision repair cross-complementation group 1 polymorphism predicts overall survival in advanced non-small cell lung cancer patients treated with platinum-based chemotherapy. Clin Cancer Res 10:4939–4943 [PubMed] [Cross Ref]10.1158/1078-0432.CCR-04-0247
11. Suk R, Gurubhagavatula S, Park S, Zhou W, Su L, Lynch T, Wain J, Neuberg D, Liu G, Christiani D (2005) Polymorphisms in ERCC1 and grade 3 or 4 toxicity in non-small cell lung cancer patients. Clin Cancer Res 11:1534–1538 [PubMed] [Cross Ref]10.1158/1078-0432.CCR-04-1953
12. Sakamoto K, Oka M, Yoshino S, Hazama S, Abe T, Okayama N, Hinoda Y (2006) Relation between cytokine promoter gene polymorphisms and toxicity of 5-fluorouracil plus cisplatin chemotherapy. Oncol Rep 16:381–387 [PubMed]
13. Medeiros R, Pereira D, Afonso N, Palmeira C, Faleiro C, Afonso-Lopes C, Freitas-Silva M, Vasconcelos A, Costa S, Osorio T, et al (2003) Platinum/paclitaxel-based chemotherapy in advanced ovarian carcinoma: glutathione S-transferase genetic polymorphisms as predictive biomarkers of disease outcome. Int J Clin Oncol 8:156–161 [PubMed] [Cross Ref]10.1007/s10147-003-0318-8
14. Dolan ME, Newbold KG, Nagasubramanian R, Wu X, Ratain MJ, Cook EH Jr, Badner JA (2004) Heritability and linkage analysis of sensitivity to cisplatin-induced cytotoxicity. Cancer Res 64:4353–4356 [PubMed] [Cross Ref]10.1158/0008-5472.CAN-04-0340
15. The International HapMap Consortium (2005) A haplotype map of the human genome. Nature 437:1299–1320 [PMC free article] [PubMed] [Cross Ref]10.1038/nature04226
16. Huang RS, Kistner EO, Bleibel WK, Shukla SJ, Dolan ME (2007) Effect of population and gender on chemotherapeutic agent-induced cytotoxicity. Mol Cancer Ther 6:31–36 [PMC free article] [PubMed] [Cross Ref]10.1158/1535-7163.MCT-06-0591
17. Abecasis G, Cardon L, Cookson W (2000) A general test of association for quantitative traits in nuclear families. Am J Hum Genet 66:279–292 [PMC free article] [PubMed]
18. Irizarry RA, Hobbs B, Collin F, Beazer-Barclay YD, Antonellis KJ, Scherf U, Speed TP (2003) Exploration, normalization, and summaries of high density oligonucleotide array probe level data. Biostatistics 4:249–264 [PubMed] [Cross Ref]10.1093/biostatistics/4.2.249
19. SAS Institute (1997) SAS/STAT software release 9.1. Cary, NC
20. Martindale J, Holbrook N (2002) Cellular response to oxidative stress: signaling for suicide and survival. J Cell Physiol 192:1–15 [PubMed] [Cross Ref]10.1002/jcp.10119
21. Kerley-Hamilton J, Pike A, Li N, DiRenzo J, Spinella M (2005) A p53-dominant transcriptional response to cisplatin in testicular germ cell tumor-derived human embryonal carcinoma. Oncogene 24:6090–6100 [PubMed] [Cross Ref]10.1038/sj.onc.1208755
22. Fishel M, Rabik C, Bleibel W, Li X, Moschel R, Dolan M (2006) Role of GADD34 in modulation of cisplatin cytotoxicity. Biochem Pharmacol 71:239–247 [PubMed] [Cross Ref]10.1016/j.bcp.2005.10.039
23. Faragher A, Fry A (2003) Nek2A kinase stimulates centrosome disjunction and is required for formation of bipolar mitotic spindles. Mol Biol Cell 14:2876–2889 [PMC free article] [PubMed] [Cross Ref]10.1091/mbc.E03-02-0108
24. Zhang W, Fletcher L, Muschel R (2005) The role of Polo-like kinase 1 in the inhibition of centrosome separation after ionizing radiation. J Biol Chem 280:42994–42999 [PubMed] [Cross Ref]10.1074/jbc.M505450200
25. Fletcher L, Cerniglia G, Nigg E, Yen T, Muschel R (2004) Inhibition of centrosome separation after DNA damage: a role for Nek2. Radiat Res 162:128–135 [PubMed] [Cross Ref]10.1667/RR3211
26. Fletcher L, Cerniglia G, Yen T, Muschel R (2005) Live cell imaging reveals distinct roles in cell cycle regulation for Nek2A and Nek2B. Biochem Biophys Acta 1744:89–92 [PubMed] [Cross Ref]10.1016/j.bbamcr.2005.01.007
27. Lin C, Liu J, Chow J, Liu C, Liu H (2007) Small-molecule c-Myc inhibitor, 10058-F4, inhibits proliferation, downregulates human telomerase reverse transcriptase and enhances chemosensitivity in human hepatocellular carcinoma cells. Anticancer Drugs 18:161–170 [PubMed] [Cross Ref]10.1097/CAD.0b013e3280109424
28. Yim E, Lee K, Kim C, Park J (2006) Analysis of differential protein expression by cisplatin treatment in cervical carcinoma cells. Int J Gynecol Cancer 16:690–697 [PubMed] [Cross Ref]10.1111/j.1525-1438.2006.00586.x

Articles from American Journal of Human Genetics are provided here courtesy of American Society of Human Genetics
PubReader format: click here to try

Formats:

Related citations in PubMed

See reviews...See all...

Cited by other articles in PMC

See all...

Links

Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...