Variation in human gut microbiota impacts tamoxifen pharmacokinetics

Abstract

Tamoxifen is the most prescribed drug used to prevent breast cancer recurrence, but patients show variable responses to tamoxifen. Such differential inter-individual response has a significant socioeconomic impact as one in eight women will develop breast cancer and nearly half a million people in the United States are treated with tamoxifen annually. Tamoxifen is orally delivered and must be activated by metabolizing enzymes in the liver; however, clinical studies show that neither genotype nor hepatic metabolic enzymes are sufficient to predict why some patients have sub-therapeutic levels of the drug. Here, using gnotobiotic- and antibiotics-treated mice, we show that tamoxifen pharmacokinetics are heavily influenced by gut bacteria and prolonged exposure to tamoxifen. Interestingly, 16S rRNA gene sequencing shows tamoxifen does not affect overall microbiota composition and abundance. Metabolomics, however, reveals differential metabolic profiles across the microbiomes of different donors cultured with tamoxifen, suggesting an enzymatic diversity within the gut microbiome that influences response to tamoxifen. Consistent with this notion, we found that β-glucuronidase (GUS) enzymes vary in their hydrolysis activity of glucuronidated tamoxifen metabolites across the gut microbiomes of people. Together, these findings highlight the importance of the gut microbiome in tamoxifen's pharmacokinetics.IMPORTANCEOne in eight women will develop breast cancer in their lifetime, and tamoxifen is used to suppress breast cancer recurrence, but nearly 50% of patients are not effectively treated with this drug. Given that tamoxifen is orally administered and, thus, reaches the intestine, this variable patient response to the drug is likely related to the gut microbiota composed of trillions of bacteria, which are remarkably different among individuals. This study aims to understand the impact of the gut microbiome on tamoxifen absorption, metabolism, and recycling. The significance of our research is in defining the role that gut microbes play in tamoxifen pharmacokinetics, thus paving the way for more tailored and effective therapeutic interventions in the prevention of breast cancer recurrence.

Keywords: breast cancer; human gut microbiome; tamoxifen.

Fig 1

Tamoxifen metabolism and enterohepatic circulation. (A) (1) Oral administration of TAM results in the exposure of the gut microbiota to the drug. (2) After absorption into the portal circulation and (3) transport to the liver, (4) tamoxifen undergoes bioactivation by P450 enzymes to produce endoxifen (END) and 4-hydroxytamoxifen (4HT) through two pathways: 4-hydroxylation and N-demethylation. (5) These metabolites undergo glucuronidation, leading to the formation of endoxifen glucuronide (END-G) and 4-hydroxytamoxifen glucuronide (4HT-G), respectively. (6) Approximately 75% of END-G and 4HT-G are excreted in the bile to the intestine. (7) END-G and 4HT-G necessitate hydrolysis by β-glucuronidase (GUS) (8) before END and 4HT can be reabsorbed into the systemic circulation and exert anti-cancer effects. (B) Chemical structures of tamoxifen and key metabolites and the liver enzymes responsible for their production.

Fig 2

Tamoxifen pharmacokinetics as a function of the gut microbiota. (A) Experimental groups (n = 6/group). (B) Experimental timeline. Germ-free mice were colonized with human fecal bacteria and tamoxifen was administered daily for 10 days. On day 10, all mice were dosed with 1:1 uniformly labeled ¹³C tamoxifen:natural-isotope-abundance tamoxifen to profile pharmacokinetics over a 12-hour time course. (C) Circulating levels of ¹³C-tamoxifen across groups. (D) ¹³C-tamoxifen abundance across groups in intestinal tissue and luminal contents. Data are mean ± standard error (s.e.). *P < 0.05, **P < 0.01, ****P < 0.0001 by a two-way ANOVA and Tukey’s multiple comparisons test.

Fig 3

Tamoxifen pharmacokinetics in antibiotics-treated mice. (A) Experimental groups n = 4, 3, 4 per group (Vehicle, TAM, Abx + TAM). (B) Experimental timeline. Antibiotics and tamoxifen or vehicle were administered daily for 10 days. All mice were dosed with tamoxifen to profile pharmacokinetics over a 12-hour time course. (C and D) Circulating levels of tamoxifen at the 2-hour peak time point (C) and 4HT across time points (D). Data are mean ± s.e. *P < 0.05, ***P < 0.001, ****P < 0.0001 by unpaired nonparametric Kolmogorov–Smirnov t-tests and two-way ANOVA with Tukey’s multiple comparisons test.

Fig 4

Diversity, abundance, and metabolome of bacteria in the gut microbiome of tamoxifen-treated mice. (A) Taxonomic analysis of microbial composition at the species level from cecal samples of the humanized + TAM and the humanized + vehicle mouse groups. The size of the bars represents the relative abundance of the bacterial taxa. Different species are shaded by different colors; see raw data (Table S1) for legend. Tamoxifen-treated mice showed no difference in (B) alpha diversity (richness) or (C) abundances as measured by copies of the 16S rRNA gene compared to vehicle. (D) Firmicutes/Bacteroidetes ratio of all samples. (E) Heatmap of the top 25 significant metabolites identified in the cecal content of humanized mice. Rows indicate metabolites (Table S4) and columns represent each mouse sample. (F) Volcano plot broadly indicates the statistically significant metabolites between groups at a significance threshold of 0.05. FC: fold change. N = 6 mice/group. Data are mean ± s.e. Unpaired nonparametric Kolmogorov–Smirnov t-tests were used to calculate and compare cumulative distributions between groups.

Fig 5

Metabolomic analysis of tamoxifen-treated human gut bacterial cultures. (A) Heatmap showing relative abundances of known and unknown metabolites in ex vivo cultures of fecal samples incubated with tamoxifen (right) or vehicle control (left). Rows indicate metabolites (Table S7) and columns represent each donor sample. N = 5 fecal samples (one per donor) and three biological replicates per sample. SC, sterile control (without bacteria). Boxed region indicates metabolites that are enriched in donors 1–3 only. Only anaerobic condition is shown here; data for aerobic condition are shown in Fig. S3; Table S8.

Fig 6

Distinct GUS activities across gut microbiomes from different healthy human donors. Time-course profile following incubation of fecal lysate from donor 6, the sample that is the highest producer of END, with (A) END-G or (B) 4HT-G for 120 minutes. Levels of hydrolyzed (C) END or (D) 4HT following incubation of lysates from 9 different fecal samples with glucuronidated 4-HT and END for 180 minutes. Average across donors ± s.e. boxed above graphs for each hydrolyzed product. (E) Comparison of hydrolysis of 4HT-G and END-G per donor. N = 3 technical replicates per donor. Data are mean ± s.e. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001 by ordinary one-way ANOVA and Tukey’s multiple comparisons test. Comparisons not shown were not statistically significant. ns, not significant.

Fig 7

Characterization of whole-genome shotgun data reveals a correlation between END production and B. fragilis GUS genes. (A) Stacked bar plot showing the normalized abundance, in reads per million (RPM), of bacterial GUS genes (HMGC279) and their assigned taxonomy across multiple fecal donors (1, 2, 3, 4, 6, 7, and 9; samples 5 and 8 were excluded because they could not be matched to their sample IDs). Each color represents a different taxonomic group. (B) Non-metric multidimensional scaling (NMDS) plot of HMGI3013 bacterial GUS gene compositions (Bray–Curtis). Each point represents a fecal donor (numbered), and the distances between points reflect the similarities in bacterial GUS gene composition. Points are colored according to END production, and features that separate the points are shown as pink arrows. (C and D) Scatter plots showing the Spearman’s rank correlation (ρ) between END production and normalized HMGI3013 GUS gene abundance, in RPM, per fecal donor (numbered). The black line represents the trend, and the shaded area indicates the 95% confidence interval.