Evaluating the impact of pharmacogenomic variants on aromatase inhibitor-induced bone toxicity using a quantitative systems pharmacology model of bone remodelling

Objectives: Aromatase inhibitors are a therapy of choice for women with early-stage hormone receptor-positive breast cancer. Despite their efficacy, aromatase inhibitors often produce toxicity in the bone, such as increased bone loss and osteoporosis, leading to therapy discontinuation. Genetic markers have been suggested to affect the risk of toxicity. However, non-linear effects of genetic variants and linkage disequilibrium in allele distributions make predictions of toxicity challenging. The objective of this study was to integrate genetic variants in bone remodeling pathways into our quantitative systems pharmacology (QSP) model of aromatase inhibitor-induced bone toxicity. Methods: We used previously published mechanistic models of bone remodeling [ 1– 3] that describe cellular and molecular networks affecting bone metabolism and extended them to incorporate the effect of estrogen. The base model was parameterized using published in-vitro biochemistry data, mouse cellular dynamics data, and human molecular concentration and clinical data, and was validated using data on bone mineral density (BMD) in healthy postmenopausal women [ 4], as well as in women treated with romosozumab [ 5]. Common polymorphisms in ESR1, CYP19A1, OPG, RANK, RANKL, SOST, LRP5 and TGFB1 were simulated by modifying production rates or binding affinities of the respective species in the model. The functional form of the genetic effect was determined from the genomic location of the variant; the effect size was estimated from previous association studies. A previously published PKPD model of exemestane [ 6] was adapted to simulate the effect of aromatase inhibition. To evaluate the impact of genetic variation, we created a set of virtual patients with population-specific minor allele frequencies and linkage disequilibrium patterns, and simulated BMD dynamics with and without aromatase inhibitors. Results: The final QSP model effectively captured expected BMD decline in postmenopausal women and was able to replicate bone gain observed after treatment with romosozumab. The model predicted that aromatase inhibition would increase the rate of bone loss 2.75-fold. The primary mechanism by which estrogen depletion affected toxicity was through modulation of osteoclast apoptosis. The simulations in virtual populations demonstrated that variability in bone loss due to genetic variation was higher after 1 year of exemestane treatment [IQR: 1.80–2.50%] than under normal conditions [IQR: 0.70–0.86%]. Finally, the model predicted that individuals carrying ESR1 and OPG variants were more likely to develop severe toxicity, which was mitigated in SOST rs18810925 carriers. Conclusions: We have developed a QSP model of aromatase inhibitor-induced bone toxicity and evaluated the impact of genetic variation in bone remodeling pathways on bone loss. The model facilitates our understanding of genetic contribution to aromatase inhibitor-induced bone loss and enables personalized predictions of toxicity.