Pharmacogenomics: time to rethink its role in precision medicine

The complex genetic landscape of human cancer is evident not only across cancers from different primary sites, but also amongst cancers of the same histopathologic subtype. Understanding the contribution of this genetic landscape to relevant clinical end points such as overall survival (OS), treatment response, and toxicity has helped facilitate the evolution and application of precision clinical oncology [1, 2]. Over the past several years, specific challenges posed by genetic heterogeneity have led to the implementation of novel biomarker-based clinical trial designs for drug development, which have led to improved survival for patients with a wide variety of tumor types [3]. However, whereas many of these successful biomarker-based clinical trials have utilized somatic mutation profiling, relatively fewer studies have harnessed the area of pharmacogenomics and germline variation.

For colorectal cancer (CRC), the role of germline variation in the efficacy and toxicity of cytotoxic chemotherapy has been the subject of widespread investigation [4]. Dihydropyrimidine dehydrogenase (DPYD) gene variation is a well-established example, whereby deleterious single-nucleotide polymorphisms in DPYD have been associated with severe toxicity to 5-fluorouracil (5-FU) therapy [5, 6]. However, despite multiple lines of evidence that specific DPYD variants can reliably predict 5-FU toxicity, a number of issues currently limit pre-treatment DPYD testing from standard clinical practice, namely: regional differences in population allele frequency, technical variation in genotyping methods, and a paucity of large-scale randomized studies [7]. Germline variation in UDP-glucuronosyltransferase 1A1 (UGT1A1) presents a similar example, in which the UGT1A1*28 polymorphism is associated with an increased risk of irinotecan toxicity due to decreased drug metabolism [8–10]. As in the case for DPYD, widespread testing for UGT1A1 polymorphisms in CRC patients remains controversial. It is noteworthy that neither the National Comprehensive Cancer Network (NCCN) [11] nor European Society of Medical Oncology (ESMO) [12] guidelines currently recommend routine clinical testing of DPYD and UGT1A1 polymorphisms. This not only reflects the practical challenges of incorporating germline variability into therapeutic decision-making, but also signifies an opportunity to discover novel germline biomarkers through innovative approaches.

In this issue of Annals of Oncology, Abad and Martinez-Balibrea et al. describe the results of a rigorous multi-center study that examined the feasibility and clinical utility of using germline DNA biomarkers to select front-line chemotherapy for patients with metastatic CRC (mCRC) [13]. Using a randomized, phase II, open-label design, a total of 195 Spain-based patients with mCRC were randomized to receive either standard front-line chemotherapy with XELOX plus bevacizumab (control group, n = 61) or a genotype-driven regimen (experimental group, n = 130) selected based upon germline variants in thymidylate synthetase (TYMS, 1494del6bp) and excision repair 1, endonuclease non-catalytic subunit (ERCC1, c.354T>C). The genotype-driven regimens included XELOX plus bevacizumab, XELIRI plus bevacizumab, FUOX plus bevacizumab, and FUIRI plus bevacizumab. With respect to the efficacy of this approach, no significant difference in progression-free survival (PFS) was observed between the control and experimental group (9.4 versus 10.1 months)—and thus the primary end point of the study was not met. However, modest significant improvements were observed in response rate (control 33% versus experimental 48%) and R0 metastatic resection rate (44% versus 86%). Notably, toxicities also varied between the treatment groups, with significantly lower rates of neuropathy but higher rates of grade 3 diarrhea in the experimental group.

Thymidylate synthetase carries out a critical step in the generation and maintenance of intracellular deoxythymidylate, which in turn is necessary for DNA maintenance. As the primary intracellular target for fluoropyrimidines (such as 5-FU), it has been hypothesized that TYMS gene expression and enzymatic activity are important mediators of treatment efficacy and/or toxicity with fluoropyrimidines-based chemotherapy regimens [14–17]. The TYMS 1494del6bp variant has a global allele frequency of 49% (based on 1000 Genomes project data) and is thought to confer decreased message RNA stability by disruption of 3’ untranslated region [18, 19]. Similarly, ERCC1 is an important component of the nucleotide excision repair pathway, which recognizes DNA adducts formed by platinum-containing agents (e.g. oxaliplatin) and thus is hypothesized to modulate their potency [20–23]. ERCC1 c.354T>C is a common variant (global allele frequency 33%) that is associated with decrease mRNA expression in vitro models [24].

For CRC, the efficacies of 5-FU- and/or oxaliplatin-based treatment regimens have been well-established across (neo)adjuvant and metastatic disease settings. Thus, understanding both extrinsic and intrinsic factors that regulate TYMS and ERCC1 activity is particularly relevant for patients with mCRC. The work by Abad and Martinez-Balibrea et al. contributes additional data on the relative value of TYMS 1494del6bp and ERCC1 c.354T>C as clinically useful biomarkers in the treatment of mCRC patients with standard cytotoxic chemotherapy. Yet, as the primary end point of their study was not met, it would not be reasonable to launch future pharmacogenomic biomarker-driven trials involving solely the determination of TYMS and ERCC1 to guide therapy compared with current standards of practice with the goal to improve OS.

Broadly speaking, Abad and Martinez-Balibrea et al. demonstrate that it is technically and logistically feasible to implement a germline biomarker-driven strategy in CRC. At least two features of their study design were key to its successful implementation: (i) centralized genotyping pipelines and (ii) careful attention to regional variation in genetic substructure. However, their results again highlight important challenges and limitations facing the pharmacogenomics community. The complex interplay between clinical covariates, germline variations, and somatic alterations is likely to be a major determinant of whether a germline biomarker-driven strategy translates into improved PFS or OS. To help advance this strategy further, it is time to rethink the role of pharmacogenomics within the big picture of predictive biomarkers that are available in the space of CRC oncology. In fact, it would be helpful to integrate both worlds: germline and somatic mutation profiling into future large-scale, prospective clinical investigations. This could enable exploration of a complex combinatorial space and broaden applicability—principles that are important not only for cytotoxic therapy, but also the next generation of targeted and biological therapies.

Funding

None declared.

Disclosure

The authors have declared no conflicts of interest.