Evidence for Chemopreventive and Resilience Activity of Licorice: Glycyrrhiza Glabra and G. Inflata Extracts Modulate Estrogen Metabolism in ACI Rats

In 2018, breast cancer will account for nearly one-third of new cancer cases in women (1). A decline in hormone therapy (HT) usage was observed after the Women's Health Initiative study in 2002 due to the increased breast cancer risk caused by the estrogen + progestin HT regimen (2). Besides the well-known hormonal estrogen carcinogenesis pathway, breast cancer risk is also influenced by changes in estrogen oxidative metabolism (3). In the breast, P450 1A1 and P450 1B1 catalyze the metabolism of estrogens to 2- and 4-hydroxylated catechols, 2-OHE₁/E₂ and 4-OHE₁/E₂, respectively (Fig. 1; ref. 4). The 2-hydroxylated metabolites are strongly associated with reduced breast cancer risk (5) because they inhibit E₂-induced proliferation (6) and are converted to nontoxic quinones (Fig. 1). On the other hand, P450 1B1 is linked to carcinogenesis because it is overexpressed in malignant tissues (7). P450 1B1 produces 4-OHE₁/E₂ which are oxidized by peroxidases/P450s to genotoxic quinones (4-OHE₁/E₂-Q) that alkylate DNA and generate depurinating adducts (estrogen chemical carcinogenesis, Fig. 1; refs. 4, 8). Hence, upregulation of the 2-hydroxylation (P450 1A1) and downregulation of the 4-hydroxylation pathway (P450 1B1) may significantly inhibit estrogen chemical carcinogenesis in the breast. In addition, NAD(P)H quinone oxidoreductase 1 (NQO1) decreases depurinating estrogen-DNA adducts (4) due to reduction of the reactive 4-OHE₁/E₂-Q to its catechol. Catechol-O-methyltransferase (COMT) also prevents quinone formation through methylation of estrogen catechols to produce stable metabolites, 2-MeOE₁/E₂ and 4-MeOE₁/E₂ (Fig. 1; ref. 9).

Due to fear of increased breast cancer risk with HT, many women have turned to botanical dietary supplements (BDS) as a complementary approach therapy for menopausal symptom relief. Female consumers generally perceive BDS as safer modalities because BDS often contain constituents that are reported to be anti-inflammatory and antioxidant, and induce detoxification enzymes (10). However, the effects of most BDS on estrogen oxidative metabolism are unknown. Although licorice belongs to one of the most popular botanicals contained in BDS used for women's health issues, several licorice species (Glycyrrhiza sp., Fabaceae) are used to source these BDS without discrimination (11). Also, clinical evidence for efficacy is generally lacking (10).

Botanically, over 30 species of licorice exist (12, 13). From a pharmacopoeial perspective, three species are currently used in BDS interchangeably: Glycyrrhiza glabra (GG), G. inflata (GI), and G. uralensis. Notably, these three species have distinctive and very different chemical profiles, as already demonstrated in previous studies (Table 1; refs. 12, 14). Cultivated in China, GI is naturally more popular in Asia. The roots from this species contain an abundant species-specific chemopreventive Michael acceptor, licochalcone A (LicA; ref. Fig. 2). GG is the most popular licorice species in the United States and Europe, and its species-specific compound is glabridin (Table 1; Fig. 2; refs. 15, 16). Both licorice species contain isoliquiritigenin (LigC, C for chalcone), also a chemopreventive Michael acceptor, and liquiritigenin (LigF, F for flavanone), a phytoestrogen with estrogen receptor (ER) β preferential properties (refs. 13, 17; Fig. 2). Both LigC and LigF are spontaneously interconvertible isomers (Fig. 2; ref. 18), found mainly as glycosides (e.g., liquiritin and isoliquiritin) in the roots (Table 1). Due to the species-specific compound profile, each licorice species has a unique bioactivity that could lead to differential clinical activity.

Previously, we showed that GG and LigC increased estrogen oxidative metabolism (2- and 4-hydroxylation), whereas GI and LicA decreased metabolism in MCF-10A “normal” breast epithelial cells (11). The purpose of the current study was to determine the chemopreventive effects of GI and GG on estrogen oxidative metabolism in the ACI rat model, which is frequently used for in vivo estrogen carcinogenesis studies (19). LicA was administered in parallel to determine its role in GI's bioactivity and its metabolic profile. LicA's distribution in liver and mammary tissues was determined and compared with LigC and LigF. Levels of 2-MeOE₁ in serum were quantified by LC-MS/MS, as a biomarker for overall estrogen oxidative metabolism, because rat P450 1B1 predominantly performs estrogen 2-hydroxylation (20). CYP1A1 and CYP1B1 expressions were determined in mammary tissues, and NQO1 activity was measured in liver and mammary tissues. Studies in MCF-10A cells with GI revealed different results between in vitro and in vivo studies and identified potential bioactive compounds. The current experiments carried out in ACI rats provide crucial information to continue studying these licorice species and their effects on estrogen carcinogenesis in long-term studies with ACI rats and eventually estrogen metabolism in women. The data herein explore additional chemopreventive modes to help develop more beneficial and safer standardized licorice dietary supplements for women's health.

All chemicals and reagents were purchased from Fisher Scientific or Sigma except for the following: 2-methoxyestrone (MeOE₁)-1,4,16,16-d4 was obtained from CDN isotope, E₂ and 2-/4-MeOE₁/E₂ reference compounds were acquired from Steraloids Inc., LigF and LigC were obtained from ChromaDex, and rat serum for initial standardization studies was purchased from BioreclamationIVT.

GI was provided as a gift to SNC by Dr. Liang Zhao (Lanzhou Institute of Chemical Physics CAS), and GG was purchased from Mountain Rose Herbs. Raw materials were identified by macroscopic/microscopic analyses and DNA barcoding, as previously described (14). Ground roots of GG and GI were extracted by maceration and percolation at room temperature with a solvent mixture [ethanol (200 USP proof), isopropanol, and water (90:5:5, v/v)] at the ratio of botanical to solvent as 1:15 (g/mL). The extracts were concentrated and freeze dried to yield 12% w/w of the initial ground roots. Extracts were analyzed by UHPLC-UV to quantify major chalcone and flavanone constituents, as previously reported (21, 22). Briefly, a standard curve containing the following 11 reference standards was used for their quantitation in both licorice extracts. The AUC was taken at 360 nm for all chalcones (isoliquiritin, isoliquiritin apioside, licuraside, LigC, and LicA), and at 275 nm for all flavanones (liquiritin, liquiritin apioside, liquiritigenin 7-O-apiosylglucoside, and LigF) and for glabridin. Quantitative results obtained for each LigF glycoside or LigC glycoside were converted by their molecular weight, thereby leading to their concentration as LigF or LigC equivalents, respectively (Table 1).

The crude LicA sample, enriched from GI (eLicA, purity of LicA, ∼ 50%), was a gift to SNC from Qinghai Lake Medicinal CO., Ltd. A loss-free countercurrent separation was implemented for the purification of LicA from the eLicA, as follows: TLC-based solvent system strategy (23) was performed for screening a proper solvent system. Among screened solvent systems (Supplementary Table S1), LicA was eluted by the organic phase of n-hexane-ethyl acetate-methanol-water (HEMWat, 4:6:5:5, v/v) to R_f = 0.43 on a precoating normal-phase Si TLC plate (MACHEREY-NAGEL). Considering the main principles of the GUESS method (Generally Useful Estimate of Solvent Systems; ref. 24), these data suggested that the corresponding solvent system may elute the target analyte into a high-resolution separation range (the sweet spot, partition coefficient, K value from 0.25 to 4 of a countercurrent separation method; ref. 25). Regarding verification of the selected solvent system, an analytical scale of high-speed countercurrent chromatography (HSCCC; 16 mL, Tauto Biotech) was applied, which achieved a higher purity of LicA than that from the literature reported solvent system on the same instrument (Supplementary Fig. S1; ref. 26). The scaled-up separation is described in the Supplementary Information. The purity of LicA used in this animal study was determined as 95% (w/w) by qHNMR.

Female August-Copenhagen Irish (ACI) rats were purchased from Harlan Laboratories at 5 weeks of age, acclimated for 1 week, fed a phytoestrogen-free diet (AIN-76A), and randomly divided into five groups with 6 rats each: vehicle, estradiol benzoate (EB, 1 mg/kg/day), LicA (80 mg/kg/day) + EB, GG (2 g/kg/day, gavage) + EB, and GI (2 g/kg/day, containing 141 mg LicA) + EB. At 6 weeks of age, one vehicle was applied subcutaneously (s.c., sesame oil) and one by gavage (50% corn oil with 50% PEG/H₂O), EB and LicA were given s.c., and the licorice extracts were administered by gavage for 4 days. The rats were sacrificed by CO₂ asphyxiation on day 5. Blood was collected, and serum prepared immediately after collection; mammary tissues, uterus, and liver were collected and snap frozen in liquid nitrogen and stored at −80°C until analysis. The animal protocol complied with the Guide for the Care and Use of Laboratory Animals, and all procedures were approved by UIC's Institutional Animal Care and Use Committee (Protocol No. 16-033).

LicA was quantified in serum, liver, and mammary gland samples from the LicA + EB and GI + EB treatment groups (Fig. 3C; Supplementary Table S2). After thawing at room temperature, serum (50 μL) was transferred to a 1.5 mL Eppendorf tube and mixed with 10 μL of ACN containing naringenin (500 nmol/L) as the internal standard (IS). Liver and mammary tissues (500–800 mg for liver and 50–100 mg for mammary gland) were weighed and homogenized in 70% aqueous methanol (containing 0.1% formic acid) at 5 mL for liver and 1 mL for mammary tissues. The homogenate (200 μL) was taken and spiked with 20 μL naringenin (500 nmol/L). Ice-cold ACN (600 μL) was added, and the mixture was centrifuged for 15 minutes at 13,000 x g at 4°C for protein precipitation. The supernatant (400 μL) was transferred to a new Eppendorf tube and evaporated to dryness under a stream of nitrogen. The residue was reconstituted in 100 μL of 20% ACN, and 5 μL was injected into LC-MS/MS for analysis. The same rat samples were analyzed for LicA metabolites.

UHPLC-MS/MS analysis was performed as described previously (27, 28). Briefly, a Shimadzu LCMS-8060 triple quadrupole mass spectrometer equipped with a Shimadzu Nexera UHPLC system was used for analysis. For quantitative analysis, analytes were separated on a Waters Acquity UPLC BEH C18 2.1 × 50 mm column (1.7 μm particle size). Mass-spectrometer parameters were as follows: nebulizing gas flow: 2.5 L/min; heating gas flow: 10 L/min; interface temperature: 300°C; DL temperature: 250°C; heating block temperature: 400°C; drying gas flow: 10 L/min. The data were acquired using selected reaction monitoring (SRM) with positive ion electrospray as follows: naringenin ([M+H]⁺ m/z 273 to 153 and m/z 273 to 147, IS); and LicA ([M+H]⁺ m/z 339 to 121 and m/z 339 to 297).

For determination of LicA metabolites, the positive ion electrospray SRM transitions for each analyte were established as m/z 515 to 339 for LicA monoglucuronides, m/z 646 to 339 for the glutathione conjugate of LicA, m/z 419 to 339 for LicA sulfate, m/z 531 to 339 for monooxygenated LicA glucuronides, and m/z 545 to 369 for catechol-O-methylated LicA glucuronides.

For qualitative determination of LigC, LigF, LicA, and glabridin in the GI and GG extracts, serum, and tissues, the licorice extracts (GG and GI) were dissolved in 50% aqueous methanol at 10 μg/mL. Rat serum and tissue samples were prepared as described above. Samples were analyzed using UHPLC-MS/MS with negative ion electrospray and SRM (28) as follows: liquiritin and isoliquiritin, m/z 417 to 255; liquiritin apioside, isoliquiritin apioside, and licuraside, m/z 549 to 255; LigF and LigC, m/z 255 to 119; glabridin, m/z 323 to 201; and glycyrrhetinic acid, m/z 469 to 425.

LC-MS/MS analysis was performed as previously described with minor modifications (29). Serum samples (150 μL) were incubated at 37°C for 4 hours after adding glucuronidase and sulfatase hydrolysis buffer (300 μL) and the IS, 2-methoxyestrone-d₄. After the incubation, sample preparation and analysis were conducted as previously described (29). The SRM transitions were as follows: m/z 534.3 to 171.2, m/z 504.3 to 171.0, and m/z 506.3 to 171.0, for 2-MeOE₁, E₁, and E₂, respectively. Results are expressed as fold change from average amount of analytes of rats treated with EB alone.

The NQO1 activity in frozen liver and mammary tissue was determined as described previously (30). The NQO1 activity was determined in a clear supernatant solution of the tissue homogenate (5 μg of liver protein and 30 μg of mammary gland protein) as previously described (31). The absorbance was measured at 610 nm, and the results were expressed as fold induction of NQO1 activity relative to the control group.

Mammary tissues (100 mg) were homogenized in TRIzol reagent. The total RNA was extracted using the RNeasy lipid tissue Kit (Qiagen), and RNA (5 μg) was reverse transcribed with Invitrogen's SuperScript III First-Strand Synthesis System. RT-qPCR was performed using TaqMan rat CYP1A1, CYP1B1, and ACTB primers with FAM/MGB probe (Applied Biosystems) as described previously (29). MCF-10A cells were obtained from the ATCC and authenticated via short tandem repeat profiling (Promega). MCF-10A cells were cultured as described previously (29). For the in vitro CYP1A1/CYP1B1 induction experiments, cells having approximately 15 through 20 passages were plated in 96-well plates and treated with vehicle (DMSO), GG, GI, and licorice compounds for 24 hours. RT-qPCR was performed as previously described (29) using TaqMan 1-Step RT-PCR Master Mix, and CYP1A1 and CYP1B1 primer with FAM-MGB probe, and GAPDH primer with VIC-MGB probe. Data were analyzed with the comparative C_T (ΔΔC_T) method and expressed as fold induction relative to the vehicle control group.

The data were expressed as mean ± SEM from 6 animals per group or ± SEM for three independent experiments in MCF-10A cells. Significance was determined using the Student t test to compare two samples or one-way ANOVA with Dunnett posttest to compare multiple samples with the control (*, P < 0.05).

Extracts were analyzed by UHPLC-UV, and all compounds were expressed as % w/w of each extract (Table 1). Glabridin (Fig. 2) was present at 1.34% of total GG extract, whereas LicA represents one of the major compounds of GI crude extract (7.07%). Glabridin and LicA exist as aglycones, but LigC and LigF occur primarily as glycosylated compounds, which exceeded the corresponding aglycones by 15-fold. LigC glycosides (isoliquiritin, isoliquiritin apioside, and licuraside) and LigF glycosides (liquiritin, liquiritin apioside, and liquiritigenin-7-O-apiosylglucoside; Fig. 2) together with the corresponding aglycones are represented as total LigC and LigF equivalents, as the glycosides are deglycosylated in vivo (Fig. 3A and B). GG extract contained 1.5-fold more LigC equivalents and over 2-fold more LigF equivalents than GI. The most abundant LigF glycoside (>70%) in GG extract was liquiritin apioside (Table 1).

LigC/LigF, LigC/LigF glycosides, glabridin, and LicA were qualitatively determined in crude extracts, serum, liver, and mammary tissues by UHPLC-MS/MS analysis. The concentration of glabridin was low in the serum, and it was not detected in liver and mammary tissues (Fig. 3A). The three other licorice compounds, LicA, LigF, and LigC, were all available in the liver and mammary gland. The most striking observation was that most of the LigC/LigF glycosides were hydrolyzed, which increased LigC/LigF aglycone concentrations in the serum and both tissue samples. LigC and LigF concentrations were similar in mammary tissues, although LigC was notably higher in serum but significantly lower than LigF in liver tissues (Fig. 3A and B). In the GI crude extract, LicA levels were much higher than free LigC; however, because LigC equivalents were hydrolyzed in vivo, the concentration of free LigC exceeded LicA concentrations in serum and mammary tissues (Fig. 3B). After LicA s.c. and GI gavage administration for 4 days, LicA was also quantified by UHPLC-MS/MS in serum, liver tissue, and mammary gland. Interestingly, after both applications, free LicA was available in serum, liver, and mammary gland 24 hours after the last dose (day 5; Fig. 3C). Although LicA is highly glucuronidated in vivo (Fig. 1B; Supplementary Fig. S2), considerable amounts of free LicA were observed in the target tissues after oral administration of the extract (Supplementary Table S2; Fig. 3C).

LicA metabolites were analyzed in serum, liver, and mammary gland samples by UHPLC-MS/MS after s.c. administration of LicA. In serum, liver, and mammary tissues, glucuronidation reactions dominated LicA metabolism and resulted in two major (MG1 and MG2) and one minor (MG3) LicA glucuronide metabolites (Fig. 1B; Supplementary Fig. S2A–S2D). LicA also formed a GSH conjugate, which was a major metabolite in the liver (Supplementary Fig. S2; Fig. 1B) and a minor metabolite in the serum and mammary tissues (Supplementary Fig. S2A and S2C). Sulfation was minor in both serum and liver (Supplementary Fig. S2A and S2B; Fig. 1B) and not detectable in mammary tissue (Supplementary Fig. S2C). LicA also formed phase I metabolites (M1, M2, and M3); however, these metabolites were only detectable in vivo after their glucuronides were hydrolyzed with β-glucuronidase/sulfatase (Supplementary Fig. S2D). These LicA phase I and II metabolites have been previously determined from incubations of LicA in liver microsomes (27).

NQO1 activity was measured in liver and mammary tissues. EB treatment alone did not affect the NQO1 activity significantly in liver tissue (Fig. 4). EB treatment slightly reduced NQO1 activity in the mammary gland which is consistent with previous reports (Supplementary Fig. S3; ref. 32).

Upon cotreatment of EB with botanicals, GG and GI significantly induced NQO1 activity to 2-fold of the EB control groups in liver tissue. LicA significantly increased NQO1 activity to 1.5-fold (Fig. 4). No induction in NQO1 activity by the licorice extracts or LicA was observed in mammary tissue (Supplementary Fig. S3).

As rat P450 1B1 preferentially catalyzes estrogen 2-hydroxylation (20), the levels of 2-MeOE₁ were measured in serum samples as a biomarker for overall estrogen oxidative metabolism. The results showed that all three groups, GG, GI, and LicA, significantly reduced 2-MeOE₁ by almost 60%, 70%, and 70% (Fig. 5A; Supplementary Fig. S4) of the EB treatment group, respectively. To analyze the influence of the botanicals on the overall amount of E₂/E₁ levels, the E₂/E₁ concentrations were also determined in serum. GG and GI slightly reduced E₂/E₁ levels; however, the difference did not reach significance (Supplementary Fig. S5).

To assess the effect on P450 1A1 and 1B1 expression in the mammary gland, CYP1A1 and CYP1B1 mRNA expression levels were measured. Treatment with EB significantly suppressed the expression of both, CYP1A1 and CYP1B1, to 3- and 5-fold lower than the vehicle control group, respectively (Fig. 5B). GG, GI, and LicA were administered with EB; therefore, their effects were compared with the EB treatment group to determine statistical significance. Interestingly, both GG, but also GI, upregulated CYP1A1 to 7- and 2-fold, respectively. Only GI caused significant downregulation of CYP1B1 expression, reducing it 3-fold (Fig. 5B); however, LicA-dosed rats showed no statistical difference from the EB-dosed rats.

To identify potential compounds that might be responsible for the observed upregulation of CYP1A1 in mammary tissue by GG and especially GI, CYP1A1 and CYP1B1 expressions were determined in MCF-10A cells after treatment with GI, GG, LicA, LigC, and LigF. As expected from previous results (11), GG caused an increase in both genes, CYP1A1 (3-fold) and CYP1B1 (1.5-fold); however, after GI treatment, CYP1A1 and CYP1B1 expressions were reduced to nearly 13-fold and 6.5-fold less than basal levels, respectively (Fig. 6A). LicA (20 μmol/L) was responsible for decreases in both CYP1A1 and CYP1B1 expressions seen by GI (Fig. 6B), which were 34-fold and 10-fold lower than basal levels, respectively. In contrast, LigC (20 μmol/L) caused an increase in CYP1A1 and CYP1B1 expression. Interestingly, LigC preferentially increased CYP1A1 to 4-fold compared with a 2.5-fold induction of CYP1B1, and LigF significantly increased CYP1B1 to 2-fold of control (Fig. 6B).

Licorice is contained in very popular BDS utilized for women's health and regarded as a potential chemopreventive agent (10, 33). It is also one of the prevalent plants in Traditional Chinese Medicine (TCM) as about 1/3 of TCM formulae contain licorice (Gan Cao; ref. 34). In the United States, GG is the most frequently utilized species. Besides GG, GI and also G. uralensis can be found in European dietary supplements. In China, all three species are cultivated and utilized without discrimination as Gan Cao. GG, GI, and their compounds demonstrated differential effects on estrogen oxidative metabolism in MCF-10A cells due to their distinct chemical profiles (11). In our previous studies, LigF had no effect, and GG and LigC treatments led to an increase in estrogen oxidative metabolism, whereas GI and LicA inhibited estrogen metabolism (11). In the ACI rat model, we explored GI's effect on detoxification (CYP1A1) and genotoxic (CYP1B1) pathways involved in estrogen chemical carcinogenesis and compared it with GG's effect. In addition, the contribution of LicA to GI's bioactivity and LicA's metabolism and distribution were analyzed. Although LicA was significantly conjugated with glucuronic acid (Fig. 1B; Supplementary Fig. S2A–S2D), free LicA was still detected in serum, liver, and mammary tissues 24 hours after the last LicA injection and oral administration of the GI extract (Supplementary Table S2; Supplementary Fig. S2; Fig. 3). This demonstrated that LicA was bioavailable in its free, bioactive form in target tissues. The present study corroborated previous reports (30, 35) that LigC and LigF were bioavailable (Fig. 3A and B).

GG, GI, and LicA induced NQO1 activity in the liver, but not in mammary tissue (Fig. 4; Supplementary Fig. S3). The lack of NQO1 activity in the mammary gland is consistent with lower LicA levels in mammary tissue compared with the liver (Supplementary Table S2; Fig. 3C). In addition, the inducibility of NQO1 in mammary epithelial cells has been demonstrated to be much lower than in liver cells (36). In support of this, other known NQO1 inducers, 4′bromoflavone, hops, and xanthohumol, caused only low (4′bromoflavone) or no NQO1 induction in the mammary gland compared with significant NQO1 induction in liver tissues of Sprague–Dawley rats (37). In comparison with vehicle, EB moderately reduced NQO1 activity in the mammary gland in this study (Supplementary Fig. S3). Estradiol has been shown before to reduce NQO1 in ER-positive cells and estrogen-sensitive tissue (32). However, EB did not change NQO1 activity in the liver in the present and previous investigations (Fig. 4; ref. 32).

The increase in NQO1 activity observed in liver tissues from GG, GI, and LicA groups confirmed the observed in vitro NQO1-inducing properties in liver cells (21). LicA (Fig. 2) contains a Michael acceptor group that reacts with sulfhydryl groups, such as in Keap1. LicA GSH conjugates were detected in liver tissues which show reaction with sulfur nucleophiles (Fig. 1B; Supplementary Fig. S2B and S2D). Activation of the Keap1/Nrf2 and ARE pathway by LicA and also LigC has been demonstrated previously in in vitro studies (21, 38) leading to reduction of oxidative stress (39, 40). However, although LigC formed GSH conjugates in vivo (Sprague–Dawley rats) similar to LicA, LigC did not induce NQO1 activity in in vivo models (21, 30). In previous in vitro studies, LigF had no effect on NQO1 activity (21). In the case of GG, these data suggest that other phytoconstituents other than LigC or LigF must contribute to the observed NQO1-inducing activity (Fig. 4).

GG, GI, and LicA caused a significant decrease in 2-MeOE₁ levels (Fig. 5A) in serum. P450 1B1 in rats primarily catalyzes estrogen 2-hydroxylation; therefore, 2-MeOE₁ was used as a general biomarker for estrogen oxidative metabolism (20). Estrogen hydroxylation in the liver is higher than in any other tissue and is predominantly performed by P450 3A4 and P450 1A2 (41). P450 1A1/2 and P450 1B1 are regulated by the aryl hydrocarbon receptor (AhR) and the xenobiotic response element (XRE; Fig. 1). LicA is an AhR antagonist (Fig. 1; ref. 11), and because P450 1A2 is regulated by AhR (42), it can be downregulated by LicA similarly to P450 1A1 (Fig. 6B) leading to reduced formation of 2-MeOE₁ levels in serum. It is interesting that GG, which led to an induction of 2-MeOE₁ in MCF-10A cells (Fig. 6), downregulated 2-MeOE₁ levels in serum in this study (Fig. 5A). Literature data demonstrate that GG as well as GI, and licorice compounds inhibit P450 3A4 and P450 1A2 activity leading to reduction of estrogen oxidative metabolism as demonstrated by reduced 2-MeOE₁ levels (Fig. 5A; refs. 43, 44). Specifically, P450 1A2 and P450 3A4 activity was moderately inhibited by GI and LicA and only weakly by GG, glabridin, and LigC (43, 44). LicA inhibited P450 3A4 irreversibly as a mechanism-based inhibitor (44) and decreased P450 3A4 gene expression in HepG2 cells (45). These results suggest that GG, GI, and LicA might interfere with general P450 metabolism (44); however, it is unclear at this point if these two licorice species lead to clinically relevant drug–botanical interactions. Clinical studies to analyze the influence of GG and GI on P450 enzymes and its drug interaction potential are warranted. In contrast to the 2-MeOE₁ levels, E₁/E₂ levels were only slightly influenced by GG, GI, and LicA, suggesting that the decrease in estrogen oxidative metabolism did not lead to an increase in E₁/E₂ levels (Supplementary Fig. S5).

To gauge the effects of GG, GI, and LicA on estrogen metabolism in mammary tissues, CYP1A1 and CYP1B1 expressions were analyzed. These data closely correlated with 2-MeOE₁ and 4-MeOE₁ levels in our previous estrogen metabolism studies in MCF-10A and MCF-7 cells (11, 29, 46). A significant reduction in CYP1A1 and CYP1B1 by EB was observed in mammary tissues (Fig. 5B). This estrogen effect on CYP1A1 has previously been described in MCF-7 cells (46). As GG and GI caused upregulation of CYP1A1 and GI additionally caused CYP1B1 downregulation, GG and GI promote estrogen detoxification and GI also reduces the estrogen genotoxic pathway in mammary tissue (Fig. 5B). Our current study and literature reports indicated that LigC preferentially increased biomarkers of the 2-hydroxylation pathway, 2-MeOE₁ and CYP1A1, in MCF-10A cells (Fig. 6B; ref. 11) and increased CYP1A1 levels (6.84-fold) in mammary tissues from female Sprague–Dawley rats (30). Thus, LigC is suggested to significantly contribute to the reversal of E₂-mediated CYP1A1 downregulation as seen after GG and GI administration (Fig. 5B). Conversely, LicA, an AhR antagonist that inhibited CYP1A1/CYP1B1 expression in MCF-10A cells (Fig. 6B) and XRE-luciferase reporter activity in HepG2 cells (11), primarily downregulated CYP1B1 expression in vivo after GI administration (Fig. 5B). Surprisingly, GI did not decrease CYP1A1 expression in ACI rats (Fig. 5B), although GI and LicA treatments caused a significant reduction in CYP1A1 well below basal levels in MCF-10A cells (Fig. 6A). The GI crude extract used in MCF-10A cells and ACI rats (Table 1; Fig. 3B) contained LigC glycosides that were hydrolyzed extensively in the gut/intestine, increasing the concentration of LigC aglycone and consequently the LigC:LicA ratio and CYP1A1 induction in mammary tissues (Figs. 3B and 5B). LigF did not affect CYP1A1 expression in MCF-10A cells in these and previous studies (Fig. 6B; ref. 11). Considering that GG exhibits a very complex phytochemical profile, other constituents of GG may add to the observed CYP1A1 induction (16).

In summary, this study investigated the effects of two popular licorice species on biomarkers of detoxification and genotoxic estrogen metabolism pathways in ACI rats and breast epithelial cells. GG and GI increased NQO1 activity in the liver (Fig. 4) and the detoxification estrogen 2-hydroxyation pathway in the mammary tissues of ACI rats (Fig. 5B). In addition, GI also decreased CYP1B1 expression (genotoxic pathway), because it contains the chemopreventive AhR antagonist, LicA (Fig. 5B). Other studies suggest that GI has superior chemopreventive properties to GG (11, 47). GI had the greatest anti-inflammatory activity among medicinal licorice species in macrophage cells because it contains two major anti-inflammatory compounds, LicA and LigC (11). Also, GI contains the ERβ preferential agonist 8-prenylapigenin (47), which may have a better safety profile. Long-term studies are planned to compare the effect of GG and GI on E₂-induced mammary carcinogenesis in ACI rats and ultimately estrogen metabolism in women. This study highlights the fact that the chemopreventive bioactivity of licorice species cannot be reduced to the activity of single bioactive compounds, but is rather a result of multiple constituents leading to polypharmacologic actions. Furthermore, knowing that licorice species have very different chemical profiles (14) that profoundly affect their bioactivity in vitro as well as in vivo should help design safer and more efficacious botanicals with the greatest chemopreventive potential in women. These data suggest that GI containing the chemopreventive compounds, LicA and LigC, might be the optimal licorice species used for women's health.

No potential conflicts of interest were disclosed.

Conception and design: S. Wang, L. Huang, G.F. Pauli, B.M. Dietz, J.L. Bolton

Development of methodology: S. Wang, L. Huang, Y. Liu, J.L. Bolton

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): S. Wang, T.L. Dunlap, L. Huang, C. Simmler, D.D. Lantvit, J. Crosby, H. Dong, R.B. van Breemen

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): S. Wang, T.L. Dunlap, L. Huang, C.E. Howell, H. Dong, B.M. Dietz

Writing, review, and/or revision of the manuscript: S. Wang, T.L. Dunlap, L. Huang, C. Simmler, C.E. Howell, S-N. Chen, G.F. Pauli, R.B. van Breemen, B.M. Dietz, J.L. Bolton, B.M. Dietz

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): S. Wang, T.L. Dunlap, L. Huang, R.B. van Breemen, B.M. Dietz, J.L. Bolton

Study supervision: S. Wang, G.F. Pauli, B.M. Dietz, J.L. Bolton

Other (prepared LicoA for in vivo assays): Y. Liu

Other (botanical integrity to ensure consistency and reproducibility): S.-N. Chen, C. Simmler, G.F. Pauli

This work was supported by NIH grant P50 AT000155 from the NIH Office of Dietary Supplements (ODS) and the National Center for Complementary and Integrative Health (NCCIH) to the UIC/NIH Center for Botanical Dietary Supplements Research and by NIH grant U41 AT008706 from NCCIH/ODS to the Center for Natural Product Technologies. The construction of the UIC CSB NMR facility and instrumentation was funded by NIGMS grant P41 GM068944 to Dr. Peter Gettins. We thank Shimadzu for providing the LCMS-8060 mass spectrometer used in this investigation. We also thank Dr. Liang Zhao at Lanzhou Institute of Chemical Physics, CAS, and Qinghai Lake Medicinal CO., Ltd. for their generous gifts.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.