Identification and Clinical Evaluation of Potential Biomarkers for Breast Cancer Resistance Protein (BCRP/ ABCG2 )

Abstract

Clinical inhibition and genetic variation of the Breast Cancer Resistance Protein (BCRP/ABCG2) efflux transporter can significantly influence drug exposure, highlighting the need for reliable BCRP functional biomarkers. This study aimed to identify and evaluate biomarkers predictive of BCRP function in humans. A comprehensive analysis of metabolomic genome‐wide association studies (mGWAS) was conducted to discover potential BCRP biomarkers, followed by evaluation in in vitro transporter assays and a clinical drug–drug interaction (DDI) study. Across multiple mGWAS datasets, plasma concentrations of three herbicide derivatives—4‐hydroxychlorothalonil (4HC), 3‐bromo‐5‐chloro‐2,6‐dihydroxybenzoic acid (BCDBA), and 3,5‐dichloro‐2,6‐dihydroxybenzoic acid (DCDBA)—were significantly elevated (P < 5E‐8) in individuals carrying reduced function ABCG2 polymorphisms. These compounds were confirmed as novel BCRP substrates via transporter uptake assays and selected for clinical evaluation alongside riboflavin, a known BCRP substrate and potential BCRP biomarker. In a DDI study with 11 healthy subjects, eltrombopag, a BCRP inhibitor, increased rosuvastatin concentrations by approximately twofold (P = 0.002). No significant changes in the plasma concentrations of organic anion transporting polypeptide 1B (OATP1B) biomarkers (CP‐I and CP‐III) or potential BCRP biomarkers (4HC, BCDBA, DCDBA, or riboflavin) were observed. Notably, two subjects were heterozygous carriers for the ABCG2 p.Q141K variant and exhibited significantly higher baseline concentrations of 4HC (P = 0.004) and BCDBA (P = 0.0003), consistent with reduced BCRP function. These findings suggest that 4HC and BCDBA are promising biomarkers for baseline BCRP function in specific populations, such as those harboring reduced function genetic polymorphisms, but do not appear suitable for detecting acute BCRP inhibition.

Study Highlights.

• WHAT IS THE CURRENT KNOWLEDGE ON THE TOPIC?

BCRP is a major determinant of exposure and response to various drugs. Although biomarkers have been evaluated for several clinically important transporters, reliable BCRP biomarkers are limited. Riboflavin is a proposed BCRP biomarker; however, its complex pharmacokinetics require further investigation and highlight the need for additional and more robust biomarkers to assess clinical BCRP function.

• WHAT QUESTION DID THIS STUDY ADDRESS?

Which compounds commonly found in human plasma are substrates of BCRP and could serve as reliable indicators of BCRP activity or expression levels in healthy subjects?

• WHAT DOES THIS STUDY ADD TO OUR KNOWLEDGE?

Three herbicide derivatives—4HC, BCDBA, and DCDBA—were identified as novel BCRP substrates detectable in human plasma. A clinical DDI study confirmed that eltrombopag increases rosuvastatin concentrations, likely by inhibiting intestinal BCRP. Neither the herbicide derivatives nor riboflavin increased following BCRP inhibition, suggesting these compounds may not be sensitive to acute modulation of the transporter. Elevated pre‐treatment concentrations of 4HC and BCDBA in carriers of the ABCG2 variant p.Q141K demonstrate their potential as markers of baseline BCRP function.

• HOW MIGHT THIS CHANGE CLINICAL PHARMACOLOGY OR TRANSLATIONAL SCIENCE?

Transporter biomarkers have advanced the assessment of membrane transporter function in drug development and precision medicine. While these findings do not support using the identified BCRP substrates for predicting acute inhibition, 4HC and BCDBA show promise as biomarkers of baseline BCRP function. With further validation, these compounds could add to the growing list of transporter biomarkers available for clinical pharmacology studies in specific populations such as children, older adults, patients with chronic diseases, and individuals harboring genetic polymorphisms.

Breast Cancer Resistance Protein (BCRP/ABCG2) is a critical efflux transporter that limits the body's exposure to a wide range of drugs through reduced intestinal absorption. Variations in BCRP function, including clinical inhibition and genetic factors, can alter drug exposure. ¹

In light of this, regulatory agencies recommend screening new molecular entities for potential BCRP inhibition as part of drug–drug interaction (DDI) risk assessments, and many marketed drugs are BCRP inhibitors. ² , ³ Beyond its role in DDIs, BCRP is also a key factor in therapeutic response to certain drugs, with ABCG2 recognized as an important pharmacogene. ⁴ , ⁵ For example, individuals with the p.Q141K missense mutation exhibit reduced allopurinol efficacy and are potentially at increased risk of statin‐induced myopathy. ⁶ , ⁷ Therefore, accurate assessment of clinical BCRP function is essential for drug development and precision medicine.

In recent years, endogenous biomarkers have emerged as valuable tools for measuring transporter function in vivo. ⁸ Traditionally, DDI predictions rely on in vitro inhibitory data compared to expected drug concentrations at interaction sites. However, this approach is prone to false‐positive predictions, leading to unnecessary clinical DDI studies. In contrast, transporter biomarkers offer a more streamlined approach and enable the detection of in vivo inhibition across dose ranges while reducing the need for dedicated DDI studies. ⁹ Transporter biomarkers are also increasingly applied in specific populations, such as patients with genetic polymorphisms or chronic diseases. For example, hepatic OATP1B biomarkers have been used to identify SLCO1B1 p.V174A allele status and to characterize altered transporter activity in patients with rheumatoid arthritis and chronic kidney disease. ¹⁰ , ¹¹ , ¹² , ¹³

Despite advances in identifying hepatic and renal transporter biomarkers, the development of biomarkers for BCRP and other intestinal efflux transporters remains a challenge. Riboflavin (vitamin B2), a known BCRP substrate, has shown potential as a BCRP biomarker in preclinical studies and early human trials. For example, Bcrp knockout rats and mice exhibited significantly elevated plasma riboflavin concentrations compared to their wild‐type counterparts, while clinical BCRP inhibition in healthy subjects resulted in modest increases in plasma riboflavin concentrations. ¹⁴ , ¹⁵ , ¹⁶ , ¹⁷ However, riboflavin pharmacokinetics are complex, with multiple factors influencing its intestinal absorption—including specific riboflavin uptake transporters (RFVTs)—while the role of BCRP in riboflavin efflux remains unclear. ¹⁸ , ¹⁹ , ²⁰ Furthermore, riboflavin is an MRP4 substrate, potentially complicating its utility as a biomarker for drugs that inhibit both BCRP and MRP4. ¹⁷  Thus, the development of additional BCRP biomarkers and further evaluation of riboflavin are necessary for improving the assessment of BCRP function in the setting of DDIs and in specific populations.

This study aimed to identify novel plasma BCRP substrates and evaluate their potential as functional biomarkers using a multifaceted approach that integrated genomic and metabolomic database analysis, in vitro transporter assays, and clinical investigation (Figure 1 ). A comprehensive analysis of metabolomic genome‐wide association studies (mGWAS) identified three pesticide derivatives—4HC, BCDBA, and DCDBA—elevated in reduced function ABCG2 variants. In vitro validation confirmed these as BCRP substrates and assessed their selectivity. These compounds, along with riboflavin, were evaluated as potential BCRP biomarkers in a clinical DDI study examining the effects of transporter inhibition and reduced function ABCG2 variants on plasma concentrations in 11 healthy subjects. While eltrombopag increased rosuvastatin concentrations and had no effect on OATP1B biomarkers (indicating BCRP inhibition), an increase in the plasma concentrations of potential BCRP biomarkers was not observed. However, 4HC and BCDBA concentrations were significantly higher in heterozygous ABCG2 p.Q141K carriers than in homozygous reference carriers. These findings suggest 4HC and BCDBA may serve as baseline BCRP function biomarkers in specific populations but are not sensitive to acute BCRP inhibition.

Figure 1.

Approach for identifying and evaluating potential BCRP biomarkers. BCRP biomarker candidates were identified by mining public genomic and metabolomic databases for metabolite associations with reduced function ABCG2 genetic variants. These compounds were subsequently screened as BCRP substrates and evaluated for interactions with OATP1Bs and P‐gp. A clinical study in 11 healthy subjects assessed the proposed biomarkers' utility as indicators of BCRP modulation in both acute (clinical inhibition) and chronic (genetic variation) contexts. Created in BioRender.com.

MATERIALS AND METHODS

Genomic and metabolomic database analysis

Summary statistics from mGWAS mapping ABCG2 as a gene of interest were extracted from the NHGRI‐EBI GWAS Catalog (www.ebi.ac.uk/gwas). ²¹ Focusing on metabolic phenotypes, we identified several studies reporting plasma metabolite associations with ABCG2 genetic variants. Briefly, these studies employed non‐targeted metabolomic profiling using gas chromatography–mass spectrometry (GC–MS) or liquid chromatography–mass spectrometry (LC–MS) quantification methods to detect known and unknown biochemicals from fasted or non‐fasted plasma samples. Metabolite associations reaching genome‐wide significance (P < 5E‐8) were selected for further evaluation. Additional information on these compounds was obtained from published literature and the Human Metabolome Database (www.hmdb.ca). ²²

Cell culture and in vitro assays

HEK293 cell lines overexpressing BCRP, OATP1B1, and OATP1B3 were established for transporter substrate assays and cultured in Dulbecco's Modified Eagle Medium (DMEM) supplemented with 10% fetal bovine serum (FBS), 1% penicillin/streptomycin, and appropriate antibiotics specific to each cell line. Cultures were maintained at 37°C in a humidified incubator with 5% CO₂. For transporter inhibition assays, HEK293 cell lines overexpressing P‐gp, RFVT1, RFVT2, and RFVT3 were cultured under the same conditions. BCRP membrane vesicles were prepared according to the manufacturer's instructions.

Transporter uptake assays were conducted to evaluate the proposed BCRP biomarkers as both substrates and inhibitors and to assess eltrombopag as an inhibitor of relevant transporters. The data were applied to FDA criteria for assessing transporter‐mediated DDI potential (see Supplementary Information ). A comprehensive description of the materials, cell lines, test compounds, uptake conditions, and procedures is provided in the Supplementary Information .

Clinical DDI study

Eleven healthy subjects were enrolled in an open‐label, fixed‐sequence crossover study. Participant demographics are available in the Supplementary Information . Eligible participants were males and non‐pregnant females aged 18–64 who provided written informed consent and were considered healthy through medical history, physical examination, and clinical laboratory evaluation. All participants were non‐smokers, with no concurrent use of prescription or non‐prescription drugs and no known allergies to the study drugs. The study was approved by the Institutional Review Board at the University of California, San Francisco (UCSF) and conducted at the UCSF Clinical Research Center. The study is registered at the ClinicalTrials.gov database (NCT04542382).

The study design was adapted from a published study initially reporting this DDI. ²³ A detailed schematic is provided (Figure S1 ). Briefly, participants received a single oral dose of rosuvastatin 10 mg tablet (Crestor; AstraZeneca). Following a washout period, participants received an oral dose of eltrombopag 75 mg tablet (Promacta; Novartis) daily for 5 days. The final eltrombopag dose was co‐administered with an oral 10 mg rosuvastatin tablet. All doses were administered with water following an overnight fast, with fasting continuing for 4 hours to minimize food effects on rosuvastatin absorption. Venous blood samples were collected in sodium heparin tubes. Samples were immediately placed on ice, centrifuged, and stored at −70°C until analysis.

Genotyping of ABCG2 polymorphism

All sample genotyping was conducted in a blinded manner using coded ID samples. Genomic DNA was extracted using the QIAamp DNA Blood Mini Kit (QIAGEN; Catalog No. 51104). Target regions containing the ABCG2 c.421C>A polymorphism were amplified by PCR with the Q5 Hot Start High‐Fidelity 2X Master Mix (New England Biolabs; Catalog No. M0494S) using primers previously described in the literature ²⁴ :

• ABCG2 rs2231142_F: 5′‐TCATTGTTATGGAAAGCAACCA‐3′

• ABCG2 rs2231142_R: 5′‐GGCAAATCCTTGTATGAAGCAG‐3′

The PCR cycling conditions were set as follows: initial denaturation at 98°C for 2 minutes, followed by 35 cycles of 98°C for 10 seconds, 62°C for 30 seconds, and 72°C for 30 seconds, with a final extension at 72°C for 5 minutes. The PCR products were cleaned using the ExoSAP‐IT™ Express PCR Product Cleanup Reagent (Applied Biosystems; Catalog No. 75001.200.UL) and subjected to Sanger sequencing. The results were analyzed using FinchTV software, and genotype information was recorded.

Drug and biomarker quantification

Plasma concentrations of rosuvastatin, eltrombopag, 4HC, BCDBA, DCDBA, riboflavin, CP‐I, and CP‐III were measured using liquid chromatography‐mass spectrometry (LC–MS). Detailed materials and methods for LC–MS quantification are available ( Supplementary Information ).

Pharmacokinetic endpoints

Drug and biomarker parameters were estimated from individual plasma concentration‐time data using noncompartmental methods (Phoenix WinNonlin version 8.3; Certara). Baseline concentration (C₀), maximum plasma concentration (C_max), and time to maximum plasma concentration (T_max) were obtained directly from the observed data. The linear‐up and logarithmic‐down trapezoidal method was used to calculate the area under the concentration‐time curve (AUC) parameters.

Statistical analysis

Data are reported as geometric means (90% CI) and ratios as geometric mean ratios (90% CI) unless otherwise noted. All parameters and ratios were logarithmically transformed prior to analysis. Paired t‐tests were used for statistical comparison across arms for all parameters except T_max, for which the Wilcoxon signed‐rank test was used. Unpaired t‐tests were used for comparisons in genetic and sex‐based subgroup analyses. Correlations were analyzed using Pearson's correlation coefficient, and the coefficient of determination (R ²) was reported. Data analysis and visualization were conducted using GraphPad Prism version 10.2.2.

RESULTS

Three potential BCRP biomarkers (4HC, BCDBA, DCDBA) discovered through analysis of genomic and metabolomic databases

Leveraging data from the NHGRI‐EBI GWAS Catalog (www.ebi.ac.uk/gwas), a database compiling summary statistics from over 45,000 human GWAS, ²¹ we identified metabolomic alterations linked to genetic variants of ABCG2 (BCRP). This analysis revealed three compounds as potential substrates of BCRP (Table 1 ). Plasma concentrations of these compounds—4HC, BCDBA, and DCDBA—were significantly elevated in individuals with the reduced function variant of ABCG2, p.Q141K (rs2231142), or other SNPs reported to be in strong linkage disequilibrium (D′ and R ² > 0.9) with this mutation, including rs141471965, rs74904971, rs45499402, rs1481012, and rs4148155. Associations between ABCG2 reduced function variants and the three compounds reached genome‐wide significance (P < 5E‐8) across multiple GWAS, with the strongest association reaching a P value of 9.21E‐206.

Table 1.

Overview of the four proposed BCRP biomarkers

Proposed BCRP biomarkers	Chemical structure and information	Metabolomic GWAS data			Animal data summary	Human data summary	Dietary source	References
		Size and population	Effect (β)	P value
4HC (4‐hydroxy chlorothalonil)	Type: Fungicide derivative MW: 247.46 g/mol LogP: 3.19 (Predicted)	19,994 European ancestry 11,840 Multi‐ethnic 8,299 European ancestry 6,136 Finnish men 1,960 European ancestry 1,768 European ancestry	0.58 0.58 0.52 0.39 0.20 0.57	9.21E‐206 1.20E‐125 1.80E‐106 2.62E‐26 5.26E‐19 1.56E‐131	Toxicology findings reported by the EPA indicate 4HC (SDS‐3701) is moderately toxic to small mammals on an acute oral basis. 4HC exhibited potent toxicity to zebrafish embryo compared to its parent compound chlorothalonil	Quantified in serum samples of pregnant women. Detected in human breast milk. Concentrations elevated in urban versus rural populations in China	Fruit, berries, vegetables, tobacco, coffee, tea, soybeans, peanuts, potatoes, maize, mushrooms	Chemical Information 22 GWAS 25 , 26 , 27 , 28 , 29 , 30 Animal Data 31 , 32 Human Data 33 , 34 , 35 Dietary Source 36 , 37
BCDBA (3‐bromo‐5‐chloro‐2,6‐dihydroxy benzoic acid)	Type: Herbicide derivative MW: 267.46 g/mol LogP: 3.7 (Predicted)	11,840 Multi‐ethnic 8,299 European ancestry 6,136 Finnish men	0.49 0.36 0.54	1.87E‐26 4.84E‐51 1.32E‐32	‐‐	Associated with lower overall breast cancer in a Mendelian randomization study	Red meat, processed meat, poultry, milk, almond milk, rice milk	Chemical Information 22 GWAS 26 , 27 , 28 Human Data 38 Dietary Source 36 , 37
DCDBA (3,5‐dichloro‐2,6‐dihydroxy benzoic acid)	Type: Herbicide derivative MW: 223.01 g/mol LogP: 3.53 (Predicted)	11,840 Multi‐ethnic 8,299 European ancestry 6,136 Finnish men	0.39 0.36 0.48	1.41E‐17 4.45E‐52 3.17E‐42	‐‐	Concentrations reduced in hypertrophic cardiomyopathy patients following surgical myectomy	Garlic, red meat, processed meat, poultry, milk, almond milk, rice milk	Chemical Information 22 GWAS 26 , 27 , 28 Human Data 39 Dietary Source 36 , 37
Riboflavin	Type: Micronutrient (Vitamin B2) MW: 376.36 g/mol LogP: −1.46 (Reported)	–	–	–	1.5‐fold increased plasma concentrations in Bcrp KO vs. WT mice (P = 0.001). 2.64‐fold increased plasma concentrations in Bcrp/Pgp dKO vs. WT rats (P = 0.0002). 1.8‐fold increased plasma concentrations in Bcrp KO vs. WT mice (P < 0.01) after IV administration of [3H]riboflavin. A 6.2‐fold lower dose (P < 0.001) was retrieved from the small intestines of Bcrp KO vs. WT mice	1.25‐fold increase in riboflavin AUC0−4 (P = 0.003), 1.14‐fold increase in AUC0−24 (P = 0.009), and 1.11‐fold increase in C max (P = 0.025) in 14 healthy males with administration of BCRP inhibitor BMS‐986371. 1.20‐fold increase in riboflavin C max/C0 ratio (P = 0.006) in nine healthy volunteers with administration of BCRP inhibitor ticagrelor. Comparable mean riboflavin AUC0–24 values observed with rosuvastatin alone (160.7 ± 77.4) and in combination with the BCRP inhibitor cedirogant (149.2 ± 82.0) in 11 healthy volunteers	Eggs, organ meats, lean meats, milk, vegetables, grains and cereals	Chemical Information 22 Animal Data 15 , 40 Human data 16 , 17 , 41 , 42 Dietary Source 43

AUC_0−X, area under the concentration‐time curve for 0 to X time interval; C _max, maximum concentration; dKO, double knockout; GWAS, genome‐wide association study; IV, intravenous; KO, knockout; LogP, octanol‐water partition coefficient; MW, molecular weight.

Metabolomic GWAS data: The associations listed are with ABCG2 missense mutation p.Q141K (rs2231142) or other SNPs that are reported to be in strong linkage disequilibrium (D′ and R ² > 0.9) with these polymorphisms, including rs141471965, rs74904971, rs45499402, rs1481012, rs4148155. Positive effect size (β) indicates that the analyte is elevated in the reported association. Dietary source: Dietary data presented for 4HC is based on foods reportedly used with chlorothalonil, the parent compound of 4HC.

Further investigation through the Human Metabolome Database (www.hmdb.ca) and published literature revealed that these compounds are herbicide derivatives associated with various foods. ²² , ³⁶ Specifically, 4HC is a derivative of chlorothalonil, while BCDBA and DCDBA are presumably metabolites of Dicamba (3,6‐dichloro‐o‐anisic acid). ³¹ , ⁴⁴ Table 1 summarizes genetic associations and additional information for these compounds, including reported toxicity and detection in human studies. For example, 4HC has exhibited toxicity in animals and is detectable in human breast milk. ³¹ , ⁴⁵

Riboflavin selected for further evaluation as a BCRP biomarker in clinical DDI studies

Published studies in Bcrp knockout (KO) models have reported elevated plasma riboflavin concentrations in Bcrp KO compared to wild‐type species, suggesting that Bcrp plays a role in riboflavin disposition in mice and rats. ¹⁴ , ¹⁵ , ⁴⁰ Additionally, two recent DDI studies have demonstrated modest yet significant increases in riboflavin concentrations in the presence of BCRP inhibitors. ¹⁶ , ¹⁷ Accordingly, riboflavin was selected to further evaluate its suitability for predicting BCRP‐mediated DDIs (Table 1 ).

Transporter uptake studies confirm 4HC, BCDBA, and DCDBA are BCRP substrates and not substrates of OATP1B1 and OATP1B3

In vitro assays were performed to determine whether 4HC, BCDBA, and DCDBA are BCRP substrates and investigate their interactions with other relevant drug transporters. The uptake of all three compounds was significantly reduced in HEK293 cells stably expressing BCRP compared to the same cell line treated with the BCRP inhibitor Ko143 (Figure 2 ). In contrast, no increased uptake was observed in OATP1B1‐ and OATP1B3‐expressing cells (Figure 2 ). Additionally, inhibitor screenings in P‐gp‐expressing cells showed that 4HC, BCDBA, and DCDBA did not affect the intracellular accumulation of the P‐gp substrate [³H]digoxin (Figure S2 ). These findings confirm that 4HC, BCDBA, and DCDBA are BCRP substrates and suggest that they are not substrates for OATP1B1 or OATP1B3 and do not inhibit P‐gp.

Figure 2.

Screening of proposed biomarkers as substrates of BCRP and OATP1B1/1B3 in cellular assays. Top Panel: Uptake of 4HC, BCDBA, and DCDBA was significantly reduced in HEK293 cells stably expressing the BCRP efflux transporter compared to the same cell line with inhibitor Ko143 added and in HEK293 empty vector cells. Bottom Panel: Uptake of these compounds was similar across cells expressing OATP1B1/1B3, OATP1B1 cells treated with the inhibitor CsA, and empty vector cells. All screenings were performed using 10 μM and 30 μM concentrations of each compound, with uptake quantified using normalized peak areas. 4HC, 4‐hydroxychlorothalonil; BCDBA, 3‐bromo‐5‐chloro‐2,6‐dihydroxybenzoic acid; BCRP, breast cancer resistance protein; CsA, cyclosporine A; DCDBA, 3,5‐dichloro‐2,6‐dihydroxybenzoic acid; OATP1B1/1B3, organic anion transporting polypeptide 1B1/1B3. Bars represent mean ± SD. Significance Levels: *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001.

Eltrombopag inhibits BCRP, OATP1B1, and riboflavin transporters at clinically relevant concentrations

Inhibitor screenings of transporters critical to rosuvastatin disposition demonstrated that eltrombopag is a potent in vitro inhibitor of BCRP, OATP1B1, and OATP1B3, with respective IC₅₀ values of 4.9 ± 0.6 μM, 1.4 ± 1.2 μM, and 4.0 ± 3.1 μM (Figure S3 ). Further investigation revealed that eltrombopag also inhibits key intestinal riboflavin uptake transporters RFVT1, RFVT2, and RFVT3, with IC₅₀ values of 44.7 ± 18.3 μM, 28.2 ± 15.7 μM, and 26.3 ± 34.0 μM, respectively (Figure S3 ). Applying these data to the FDA framework for evaluating transporter‐mediated DDIs, a therapeutic dose of 75 mg of eltrombopag was predicted to inhibit intestinal BCRP, all three RFVTs, and hepatic uptake mediated by OATP1B1 ( Supplementary Information ). Notably, the FDA framework does not currently include specific criteria for RFVT inhibition. Therefore, the recommended ratio and cutoff value used for intestinal P‐gp and BCRP were also applied to the RFVTs, given the intestinal role of these transporters.

Eltrombopag co‐administration increased rosuvastatin concentrations in 11 healthy subjects

An open‐label, fixed‐sequence crossover study was conducted in 11 healthy subjects to investigate the reported DDI between eltrombopag and rosuvastatin, as well as the potential impact of eltrombopag on the proposed BCRP and established OATP1B biomarkers. ²³ The study aimed to elucidate the mechanism of this interaction, given that rosuvastatin is a substrate of both BCRP and OATP1B1, which are predicted to be inhibited by eltrombopag. Additionally, the study evaluated 4HC, BCDBA, DCDBA, and riboflavin as potential biomarkers of acute BCRP modulation.

Co‐administration of eltrombopag significantly increased rosuvastatin concentrations, resulting in a twofold increase in C_max (P = 0.002), a 2.1‐fold increase in AUC₀₋₄ (P = 0.002), and a 1.8‐fold increase in AUC_0−INF (P = 0.002) compared to rosuvastatin alone (Figure 3 , Table 2 ). Eltrombopag markedly reduced rosuvastatin apparent clearance from 1.96 to 1.18 L/h/kg (P = 0.02) and apparent volume of distribution from 29.3 to 14.9 L/kg (P = 0.004). The half‐life of rosuvastatin remained consistent across study arms (10.4 vs. 8.8 h, P = 0.13). Eltrombopag pharmacokinetics are presented in Figure S4 .

Figure 3.

The Effect of Eltrombopag on Rosuvastatin Pharmacokinetics and OATP1B Biomarker Coproporphyrin‐I in 11 Healthy Subjects. Plasma concentrations of rosuvastatin (Top Panel) and coproporphyrin‐I (Bottom Panel) in a single‐sequence crossover study with 11 healthy subjects. In the first arm, subjects received a single 10 mg oral dose of rosuvastatin (Rosuvastatin Alone). Following a washout period of ≥7 days, subjects received a 75 mg oral dose of eltrombopag every 24 hours for 5 days, with the last dose co‐administered with a 10 mg oral dose of rosuvastatin (Rosuvastatin + Eltrombopag). Concentration‐time profiles are presented as geometric means with a 90% geometric coefficient of variation. Parameters were compared using paired t‐tests, with P values reported on the respective graphs. AUC_0‐X (Area under the concentration‐time curve for 0 to X time interval); C_max (Maximum concentration).

Table 2.

The Effect of eltrombopag on rosuvastatin pharmacokinetics, established OATP1B biomarkers, and proposed BCRP biomarkers

PK parameter	Rosuvastatin alone arm	Rosuvastatin + eltrombopag arm	Geometric mean ratio (90% CI) of rosuvastatin + eltrombopag arm to rosuvastatin alone arm; P value
Rosuvastatin
Cmax (ng/mL)	5.4 (3.8–7.7)	10.9 (8.8–15.2)	2.01 (1.59–2.53); P = 0.002
Tmax (h)	5 (1.5–5)	4 (3.5–5)	N/A; P = 0.66
AUC0−4 (ng*h/mL)	11.8 (8.2–16.9)	25.0 (17.8–35.0)	2.12 (1.71–2.62); P = 0.002
AUC0−INF (ng*h/mL)	69.1 (48.5–98.5)	115.6 (94.5–163.2)	1.80 (1.49–2.17); P = 0.002
CL/F/kg (L/h/kg)	1.96 (1.41–2.71)	1.18 (0.91–1.52)	0.56 (0.46–0.67); P = 0.02
V/F/kg (L/kg)	29.3 (22.8–37.7)	14.9 (11.5–17.8)	0.49 (0.38–0.63); P = 0.004
T1/2 (h)	10.4 (9.2–11.7)	8.8 (7.5–10.2)	0.88 (0.78–0.99); P = 0.13
CP‐I
C0 (nM)	0.81 (0.74–0.90)	0.74 (0.67–0.82)	0.90 (0.83–0.98); P = 0.05
Cmax (nM)	0.93 (0.84–1.03)	0.89 (0.79–1.00)	0.95 (0.88–1.03); P = 0.33
AUC0−24 (nM*h)	16.9 (15.3–18.6)	16.2 (14.2–18.4)	0.96 (0.88–1.04); P = 0.50
CP‐III
C0 (nM)	0.12 (0.10–0.14)	0.09 (0.07–0.11)	0.75 (0.64–0.88); P = 0.01
Cmax (nM)	0.13 (0.11–0.15)	0.11 (0.09–0.13)	0.83 (0.70–0.98); P = 0.08
AUC0−24 (nM*h)	1.83 (1.60–2.09)	1.70 (1.46–1.99)	0.93 (0.80–1.08); P = 0.44
4HC
C0 (ng/mL)	2.79 (1.79–4.35)	2.64 (1.65–4.22)	0.95 (0.80–1.12); P = 0.65
Cmax (ng/mL)	3.65 (2.38–5.59)	3.88 (2.58–5.83)	1.06 (0.90–1.26); P = 0.62
AUC0−4 (ng*h/mL)	10.8 (6.6–17.4)	10.1 (6.3–16.2)	0.94 (0.89–0.99); P = 0.06
AUC0−24 (ng*h/mL)	65.6 (41.4–103.9)	62.6 (39.0–100.5)	0.96 (0.90–1.02); P = 0.20
BCDBA
C0 (ng/mL)	0.64 (0.47–0.88)	0.57 (0.40–0.82)	0.99 (0.89–1.10); P = 0.95
Cmax (ng/mL)	0.70 (0.50–0.98)	0.70 (0.50–0.98)	1.01 (0.93–1.09); P = 0.96
AUC0−4 (ng*h/mL)	2.13 (1.47–3.10)	2.10 (1.45–3.04)	0.98 (0.91–1.07); P = 0.78
AUC0−24 (ng*h/mL)	16.9 (15.3–18.6)	16.2 (14.2–18.4)	0.95 (0.89–1.02); P = 0.38
DCDBA
C0 (ng/mL)	0.73 (0.57–0.95)	0.80 (0.62–1.04)	1.10 (1.00–1.21); P = 0.07
Cmax (ng/mL)	1.03 (0.84–1.28)	0.98 (0.77–1.24)	0.95 (0.85–1.06); P = 0.49
AUC0−4 (ng*h/mL)	3.02 (2.42–3.76)	2.97 (2.37–3.72)	0.98 (0.93–1.04); P = 0.72
AUC0−24 (ng*h/mL)	18.6 (14.6–23.5)	19.2 (14.9–24.7)	1.03 (0.97–1.10); P = 0.39
Riboflavin
C0 (ng/mL)	2.72 (2.22–3.35)	2.79 (2.03–3.86)	1.03 (0.82–1.28); P = 0.42
Cmax (ng/mL)	4.16 (3.29–5.26)	3.61 (2.69–4.86)	0.87 (0.71–1.07); P = 0.31
AUC0−4 (ng*h/mL)	10.6 (8.6–12.9)	10.2 (7.6–13.6)	0.96 (0.77–1.20); P = 0.88
AUC0−24 (ng*h/mL)	65.9 (55.6–78.2)	65.6 (49.6–86.7)	0.99 (0.82–1.21); P = 0.63

No elevation in coproporphyrin concentrations was observed with eltrombopag administration

OATP1B biomarkers coproporphyrin (CP)‐I and CP‐III were measured to investigate the role of OATP1B1/1B3 in the observed DDI. The administration of eltrombopag showed no effect on plasma CP‐I and CP‐III concentrations, as the AUC_0–24 and C_max of both biomarkers remained consistent across arms (Figure 3 , Table 2 ). Notably, CP‐III baseline concentrations (C₀) were lower in the rosuvastatin + eltrombopag arm compared to the rosuvastatin alone arm.

ABCG2 reduced function variant linked to elevated baseline concentrations of 4HC and BCDBA

To assess the impact of the ABCG2 reduced function variant p.Q141K on the proposed BCRP biomarkers, baseline concentrations of these compounds were measured in the clinical study subjects, stratified by ABCG2 genotype. The two individuals heterozygous for the p.Q141K variant exhibited significantly higher baseline concentrations of 4HC and BCDBA than those homozygous for the reference allele (Figure 4 ). While DCDBA concentrations showed an upward trend in individuals with the p.Q141K variant, this increase was not statistically significant. Riboflavin concentrations showed no association with the p.Q141K variant. As expected, increased concentrations of the positive control rosuvastatin were observed in participants with the p.Q141K variant compared to the reference allele. These individuals demonstrated a 2.4‐fold higher C_max (P = 0.02), a 2.9‐fold higher AUC_0–4 (P = 0.02), and a 2.6‐fold higher AUC_0–INF (P = 0.01).

Figure 4.

The Effect of ABCG2 p.Q141K Variant on Baseline Concentrations of Proposed BCRP Biomarkers. Baseline (pre‐treatment; T = 0 h) concentrations of 4HC, BCDBA, DCDBA, and riboflavin in nine individuals homozygous for the ABCG2 reference allele (“REF (CC)”) compared to two individuals heterozygous for the reduced function variant (“p.Q141K (CA)”). Concentrations were compared across groups using unpaired t‐tests, with P values reported on the respective graphs.

No proposed BCRP biomarkers increased with eltrombopag administration

Acute administration of eltrombopag over 5 days, which caused a twofold change in rosuvastatin concentrations through presumed BCRP inhibition, did not lead to corresponding increases in 4HC, BCDBA, DCDBA, or riboflavin (Figure 5 ). Specifically, C_max, AUC₀₋₄, or AUC₀₋₂₄ did not significantly change across study arms for any proposed BCRP biomarker (Table 2 ).

Figure 5.

The Effect of Eltrombopag on Proposed BCRP Biomarkers. Plasma concentrations of 4HC, BCDBA, DCDBA, and riboflavin in 11 healthy subjects. Concentration‐time data are presented as geometric means with a 90% geometric coefficient of variation.

Further investigation of riboflavin concentrations reveals potential inhibition of riboflavin uptake, sex differences, and food effects

Given that in vitro data suggested eltrombopag might inhibit riboflavin uptake, we investigated the relationship between eltrombopag and riboflavin concentrations in vivo. The C_max of eltrombopag and riboflavin were negatively correlated (R ² = 0.59, P = 0.006), though this was not significant for AUC₀₋₂₄ (R ² = 0.18, P = 0.19) (Figure S5 ). No relationships were observed between eltrombopag and 4HC, BCDBA, or DCDBA concentrations.

To explore the potential effect of sex differences on riboflavin, we conducted a subgroup analysis of our participants to assess fold change in riboflavin AUC and C_max stratified by sex. Riboflavin AUC₀₋₂₄ fold change was significantly higher in males compared to females (1.21 vs. 0.79, P = 0.03). However, this difference was not significant for fold change in AUC₀₋₄ (1.17 vs. 0.76, P = 0.07) or C_max (1.02 vs. 0.71, P = 0.21) (Figure S6 ).

Considering the substantial amount of riboflavin consumed daily, we investigated whether any increases occurred following food intake during the study (Figure S7 ). In the presence of eltrombopag, there was a modest yet significant increase in mean riboflavin AUC in the 4 hours following food intake (AUC₄₋₈) compared to the fasted state (AUC₀₋₄) (11.3 vs. 9.7, P = 0.001). In the rosuvastatin alone arm, mean riboflavin AUC₄₋₈ and AUC₀₋₄ were similar but did not reach statistical significance (11.7 vs. 10.6, P = 0.12).

DISCUSSION

Transporter biomarkers have proven effective for understanding how clinical inhibition, genetic variation, or disease states affect transporter function and, in turn, drug exposure and response. Although BCRP is a major determinant of intestinal drug absorption, reliable biomarkers for assessing BCRP function are currently not available. To address this gap, our study aimed to identify BCRP substrates in the plasma and assess their utility in detecting BCRP modulation in acute (e.g., clinical inhibition) and chronic (e.g., genetic variation) settings.

Through a comprehensive analysis of public genomic and metabolomic databases, we identified three novel BCRP substrates: 4HC, BCDBA, and DCDBA. Across multiple populations, plasma concentrations of these compounds were significantly elevated in individuals carrying reduced function ABCG2 variants, consistent with impaired BCRP‐mediated efflux (Table 1 ). All three compounds are derivatives of commonly used agricultural herbicides and are ingested through produce and various food products. Potential toxicities of these herbicides have been explored in other studies and are not the subject of this manuscript. ³¹ , ³²

Through in vitro assays, we confirmed that BCRP transports 4HC, BCDBA, and DCDBA, marking their first validation as BCRP substrates (Figure 2 ). A persistent challenge in predicting BCRP‐mediated DDIs is the overlap of substrates and inhibitors with other transporters, such as OATP1B1/1B3 and P‐gp. ⁴⁶ , ⁴⁷ , ⁴⁸ Further uptake studies showed that these compounds are not OATP1B substrates and do not inhibit P‐gp, indicating they are unlikely to undergo P‐gp efflux. Thus, these compounds were proposed as selective BCRP biomarkers for further evaluation.

In a study of 11 healthy subjects, two heterozygous carriers of the ABCG2 p.Q141K variant exhibited significantly elevated baseline 4HC and BCDBA plasma concentrations compared to the nine individuals homozygous for the reference allele (Figure 4 ). This targeted analysis, supported by in vitro and untargeted mGWAS data, suggests that 4HC and BCDBA are promising markers of reduced BCRP function in vivo. Although DCDBA showed a non‐significant increasing trend, in vitro and mGWAS data also support further investigation of DCDBA as a potential BCRP biomarker.

Beyond the p.Q141K variant, other factors may influence BCRP expression and activity, contributing to variability in drug concentrations across populations. For example, BCRP expression varies with age, with adults having higher intestinal BCRP abundance than children. ⁴⁹ Additionally, chronic administration of BCRP inhibitors can mimic the effects of the p.Q141K mutation. ⁵⁰ Less characterized ABCG2 variants, along with epigenetic and post‐transcriptional modifications, may further affect BCRP function. ⁴ Given these potential influences on BCRP function, investigating 4HC and BCDBA in these contexts is warranted, particularly in specific populations where variability in drug response is a concern.

In addition to assessing chronic BCRP function, we aimed to evaluate the biomarkers' ability to detect acute BCRP inhibition for DDI prediction. Administration of eltrombopag at steady‐state resulted in a twofold increase in rosuvastatin concentrations, consistent with previous reports. ²³ Intestinal BCRP inhibition was identified as the likely mechanism, supported by unchanged OATP1B biomarkers (CP‐I and CP‐III), no significant effect on rosuvastatin elimination half‐life, and a greater fold‐change in rosuvastatin AUC₀₋₄ compared to AUC_0−INF. A prior study with intravenous ceftriaxone suggested that eltrombopag does not inhibit hepatic BCRP, reinforcing the intestine as the inhibition site in our study. ⁵¹ While additional transporters (e.g., OATP2B1, MRP2, OAT1/3) and enzymes (e.g., CYP2C9, CYP2C19) contribute to rosuvastatin disposition, there are insufficient data to suggest that eltrombopag modulates these pathways. ⁵² , ⁵³

While rosuvastatin showed a significant increase with acute BCRP inhibition, plasma concentrations of 4HC, BCDBA, and DCDBA remained unchanged. Several factors may explain this. First, as organochlorinated herbicides, these compounds likely exhibit lipophilic, stable, and bioaccumulative properties, with potentially long biological half‐lives. ⁴⁵ , ⁵⁴ These properties could make them less responsive to short‐term BCRP inhibition. Additionally, the roles of intestinal versus systemic (i.e., hepatic or renal) BCRP in their elimination remain unclear. Since eltrombopag inhibits intestinal but not hepatic BCRP, if these compounds are primarily cleared via hepatic BCRP, an acute increase from eltrombopag administration would not be expected, although the genetic effect could still be observed.

This presents a broader challenge for developing biomarkers for BCRP and other efflux transporters, such as P‐gp, that predominantly impact intestinal drug absorption but are widely expressed throughout the body. Ideally, circulating plasma concentrations of these markers would reflect intestinal transporter activity. However, these compounds may need to be excreted into the bile and then reabsorbed through the intestine to detect intestinal BCRP inhibition. Therefore, future studies of these and other BCRP biomarkers should investigate potential enterohepatic circulation and elimination pathways in greater detail to determine whether these markers are appropriate to assess intestinal BCRP activity or rather reflect transporter function more broadly.

Despite two clinical studies showing modest increases in plasma riboflavin with acute BCRP inhibition, no such effect was observed in our study. ¹⁶ , ¹⁷ This led us to investigate the effect of eltrombopag on riboflavin transporters (RFVTs). In vitro findings showed that therapeutic concentrations of eltrombopag inhibit the three known RFVTs, which facilitate riboflavin uptake in the intestine. ¹⁸ , ¹⁹ , ²⁰ , ⁴¹ The negative correlation between eltrombopag and riboflavin plasma concentrations also suggests reduced riboflavin absorption and attenuation of BCRP inhibition effects. This finding resembles ticagrelor inhibition of MRP4, proposed to explain lower‐than‐expected riboflavin increases despite BCRP inhibition. ¹⁷ Similarly, a recent study by Savaryn et al. observed no riboflavin increase with the BCRP inhibitor cedirogant, though the involvement of other transporters was not examined. ⁴² As eltrombopag was not screened for MRP4 inhibition in this study, this mechanism should also be explored as a potential contributor. Future studies employing riboflavin as a BCRP biomarker should investigate RFVT and MRP4 inhibition.

Since 4HC, BCDBA, DCDBA, and riboflavin are all ingested through diet, the timing of food intake is an important consideration. To minimize the effect of food on rosuvastatin absorption, participants followed an overnight fast, continuing until 4 hours after drug administration. We investigated whether concentrations of these compounds differed between the fasted (AUC₀₋₄) and fed (AUC₄₋₈) states. Riboflavin AUC₄₋₈ was modest but significantly higher than AUC₀₋₄, but only in the presence of eltrombopag co‐administration (Figure S7 ). This suggests that when transporters are inhibited by eltrombopag, riboflavin concentrations may be proportional to dietary intake, whereas without inhibition, transporter capacity may govern total intake rather than dietary dose. No food effects were observed for 4HC, BCDBA, or DCDBA, likely due to their trace amounts in food compared to riboflavin. ⁴³

Given the reported differences in micronutrient intake between sexes, we conducted a subgroup analysis of riboflavin based on biological sex. ⁵⁵ Males exhibited significantly higher fold changes in AUC₀₋₂₄ than females, with the average fold change in males aligning with the findings from a prior pilot DDI study involving 14 healthy males, which demonstrated riboflavin as a BCRP biomarker (1.21‐fold vs. 1.14‐fold). However, no significant differences were observed between sexes in AUC₀₋₄ or C_max fold changes. Due to the small sample size, it remains unclear whether these effects reflect physiological differences between sexes. Future studies should prioritize female enrollment to allow for a more comprehensive evaluation.

This study had several notable limitations. First, the small sample size and single‐site design limit the generalizability of the findings. Although 4HC, BCDBA, and DCDBA were identified across multiple populations in mGWAS, potential variations due to demographic factors, cultural differences, and dietary habits cannot be ruled out. Additionally, the single‐sequence design, without control over dietary intake, may have influenced biomarker concentrations due to variations in diet. Future research should involve larger, more diverse populations, utilize randomized controlled designs, and incorporate standardized dietary controls. Further studies on the ADME (absorption, distribution, metabolism, and excretion) profiles of these herbicide derivatives, including in animal models, are needed to better understand their pharmacokinetics and potential utility. For example, estimating the biological half‐life of these compounds and determining whether they are substrates of key transporters, such as P‐gp and OATP2B1, would provide valuable insights into their behavior and applicability as biomarkers.

Overall, our study identified three novel BCRP substrates, 4HC, BCDBA and DCDBA, and demonstrated 4HC and BCDBA as potential biomarkers for baseline BCRP function. These compounds may serve as valuable indicators of BCRP activity in specific populations, including those with reduced‐function ABCG2 genetic polymorphisms, and potentially in other groups with altered BCRP function. However, none of the biomarkers were able to detect acute BCRP inhibition, highlighting the need for further exploration of BCRP biomarkers specifically suited for predicting acute DDIs.

FUNDING

This publication was made possible by Grant Number U01 FD004979/U01 FD005978 from the FDA, which supports the UCSF‐Stanford Center of Excellence in Regulatory Sciences and Innovation. Its contents are solely the responsibility of the authors and do not necessarily represent the official views of the HHS or FDA. This project was fully funded by federal sources, with a total award amount of $355,763. A.M.R. is supported by the National Institute of General Medical Sciences of the National Institutes of Health (NIH) under award number 5T32GM007546. K.T. was supported by an appointment to the Research Participation Program at the Center for Drug Evaluation and Research, administered by the Oak Ridge Institute for Science and Education through an interagency agreement between the US Department of Energy and the FDA.

CONFLICT OF INTEREST

R.L., X.L., and Y.L. are employees of Gilead Sciences, Inc., and own stock and/or stock options in Gilead Sciences, Inc. All other authors declared no competing interests for this work.

AUTHOR CONTRIBUTION

A.M.R. and K.M.G. wrote the manuscript. A.M.R., S.W.Y., and K.M.G. designed the research. A.M.R., S.W.Y., J.Y., C.M.B., K.T., X.Z., R.L., X.L., R.H., and Y.X. performed the research. A.M.R., Y.L., H.S., L.Z., X.Y., Q.L., S‐M.H., and K.M.G. analyzed the data. Y.L., H.S., and K.M.G. contributed new reagents/analytical tools.

DISCLAIMER

The contents of this article reflect the views of the authors and should not be construed to represent the FDA's views or policies. No official support or endorsement by the FDA is intended or should be inferred. As Deputy Editor‐in‐Chief of Clinical Pharmacology & Therapeutics, Kathleen Giacomini was not involved in the review or decision process for this paper.

Supporting information

Data S1.

Figure S1.