Identification of genetic variants associated with capecitabine-induced hand-foot syndrome through integration of patient and cell line genomic analyses

Abstract

Objective: A primary challenge in identifying replicable pharmacogenomic markers from clinical genomewide association study (GWAS) trials in oncology is the difficulty in performing a second large clinical trial with the same drugs and dosage regimen. We sought to overcome this challenge by incorporating GWAS results from cell-based studies using the same chemotherapy as a clinical cohort.

Methods: In this study, we test whether the overlap between genetic variants identified in a preclinical study and a clinical study on capecitabine is more than expected by chance. A GWAS of capecitabine-induced cytotoxicity was performed in 164 lymphoblastoid cell lines derived from the CEU HapMap population and compared with a GWAS of hand-foot syndrome (HFS), the most frequent capecitabine-induced adverse drug reaction, in Spanish breast and colorectal cancer patients (n=160) treated with capecitabine.

Results: We observed an overlap of 16 single nucleotide polymorphisms associated with capecitabine-induced cytotoxicity (P<0.001) in lymphoblastoid cell lines and HFS (P<0.05) in patients, which is a greater overlap than expected by chance (genotype-phenotype permutation empirical P=0.015). Ten tag single nucleotide polymorphisms, which cover the overlap loci, were genotyped in a second patient cohort (n=85) and one of them, rs9936750, was associated with capecitabine-induced HFS (P=0.0076).

Conclusion: The enrichment results imply that cellular models of capecitabine-induced cytotoxicity may capture components of the underlying polygenic architecture of related toxicities in patients.

Figure 1. Comparison of individual genome-wide association study (GWAS) results

Left: capecitabine-induced cytotoxicity in LCLs. Right: capecitabine-induced HFS in patients. Black triangles correspond to SNPs with the same directions of effect in both GWAS and P-values < 0.001 in the LCL data and < 0.05 in the patient data. Dashed lines indicate the thresholds used in the enrichment analysis, while solid lines indicate genome-wide significance (P < 5 × 10⁻⁸).

Figure 2. Patient capecitabine-induced HFS SNPs are enriched for SNPs associated with capecitabine-induced cytotoxicity in LCLs, but not for SNPs associated with cytotoxicity induced by other chemotherapeutics

Distribution of chemotherapy-induced cytotoxicity SNP count (P < 0.001) in 1000 permutations of the patient HFS genotype-phenotype connections (P < 0.05). The dot is the observed SNP overlap at these thresholds for (A) capecitabine, empirical P = 0.015, (B) carboplatin, empirical P = 0.608, (C) cisplatin, empirical P = 0.812, and (D) paclitaxel, empirical P = 0.749.

Figure 3. The association of one capecitabine patient/LCL overlap SNP, rs9936750, also associated with HFS in a second patient cohort

rs9936750 genotype is associated with (A) capecitabine-induced cytotoxicity in LCLs (P = 8.7 × 10⁻⁴), as measured by area under the concentration curve (AUC). (B) Kaplan–Meier analysis of cumulative dose of capecitabine up to the development of grade 3 HFS, by rs9936750 genotype in patient cohort 1 (P = 0.036), and (C) in patient cohort 2 (P = 0.0076). The C allele is associated with increased cytotoxiciy (decreased survival) and increased risk of HFS.