Moving away from candidate genes: a “genome-wise” discovery of gemcitabine myelotoxicity

The paper from Kiyotani et al.¹ provides novel markers of myelotoxicity of gemcitabine, an antimetabolite highly used in the treatment of pancreatic cancer patients, as well as in other diseases². Myelotoxicity is the main side effect of gemcitabine and is clinically relevant, causing dose reductions and interruptions during the course of a patient therapy. Despite almost two decades of pharmacology explorations on gemcitabine^3,4, there are no available markers that can be used to guide therapy with gemcitabine in the clinic today, as stated by the authors of this paper.

The importance of the findings of Kiyotani et al.¹ is to have mapped genomic regions associated with increased risk of neutropenia/leukopenia in Japanese patients treated with gemcitabine, moving away from the traditional, candidate gene pharmacogenetic studies and using a genome-wide approach. Leveraging the Japanese BioBank infrustructure of thousands of patients already interrogated for their genomic content, a case-control design (presence/absence of severe myelotoxicity) has identified four novel genomic regions. The relatively small sample size for a genomic study might be a limitation, although we have witnessed important discoveries of toxicity markers with similar sample sizes of the discovery set⁵. The use of replication, like it was done in this study, should limit the chance for false discoveries.

Out of the new four regions, two are intergenic and without an apparently plausible link to the observed associations. Although long-distance regulatory mechanisms have been proposed for SNPs in gene deserts⁶, distant intergenic SNPs are still regarded to as regulatory “dark matter”. The study mapped DAPK1 and PDE4B, both putatively involved in mediating the cytotoxicity of gemcitabine. The authors do not provide the mechanistic basis for why these SNPs might change the gene phenotype at the cellular level. Such studies are quite critical, in particular when, after stringent Bonferroni correction of the replication p value, SNPs in these two genes have a p value that is higher than the adjusted p value for replication.

Genome-wide association studies (GWAS) provide very rich datasets. Another crucial analysis of this paper is the evaluation of the association with SNPs in genes of gemcitabine disposition and mechanism of action. In essence, none of the interrogated SNPs these genes showed an association with the myelotoxicity phenotype. Now, although it is not clear how many markers in each of the candidate genes were tested, it is well known that the SNP coverage of genes of pharmacogenetics is incomplete for most of the GWAS platforms⁷. Kiyotani et al.¹ provide a first indication that SNP predictors of gemcitabine myelotoxicity might lay outside of the candidate genes, and additional testing in other GWAS⁸ might be required to rule out gemcitabine disposition genes.

Kiyotani et al.¹ provides the first GWAS of gemcitabine toxicity. Such studies generate unprecedented opportunities for major discoveries, as the knowledge base of the genomic content of the study patients can be exploited for multiple phenotypes and to address numerous scientific questions, directing translational research to novel genomic territories and, hopefully, to improved patient management.