Sex‐Dependent Effects of CYP2D6 on the Pharmacokinetics of Berberine in Humans

Abstract

An over‐the‐counter product berberine (a major alkaloid in goldenseal) is a substrate of the uptake transporter OCT1 and the metabolizing enzyme CYP2D6. The two genes exhibit common functional polymorphisms. Approximately 9% of Europeans and white Americans are either poor CYP2D6 metabolizers or poor OCT1 transporters. In this study, we investigated the effects of OCT1 and CYP2D6 polymorphisms on berberine pharmacokinetics in humans. We confirmed in vitro that berberine is an OCT1 substrate (K _M of 7.0 μM, CL_int of 306 ± 29 μL/min/mg). Common OCT1 alleles *3 to *6 showed uptake reduced by at least 65% and Oct1/2 knockout mice showed 3.2‐fold higher AUCs in liver perfusion experiments. However, in humans, poor OCT1 transporters did not show any differences in berberine pharmacokinetics compared with reference participants. In contrast, CYP2D6 polymorphisms significantly affected berberine metabolism, but exclusively in females. Females who were poor CYP2D6 metabolizers had an 80% lower M1‐to‐berberine ratio. General linear model analyses suggest strong synergistic, rather than additive, effects between female sex and CYP2D6 genotype. Overall, berberine displayed low oral bioavailability, yet females had a 2.8‐fold higher AUC and a 3.6‐fold higher C _max than males (P < 0.001). These effects were only partially attributable to the sex‐CYP2D6 genotype interaction. In conclusion, despite berberine being an OCT1 substrate, OCT1 deficiency did not affect berberine pharmacokinetics in humans. In contrast, CYP2D6 emerges as a critical enzyme for berberine metabolism in females, but not in males, highlighting sex‐specific differences. We suggest that factors beyond CYP2D6 metabolism are determining berberine's systemic exposure, especially in males (NCT05463003).

Study Highlights.

• WHAT IS THE CURRENT KNOWLEDGE ON THE TOPIC?

Berberine is a plant compound used over‐the‐counter for multiple indications. Berberine can cause pharmacokinetic interactions with drugs, but little is known about the pharmacokinetics of berberine itself.

• WHAT QUESTION DID THIS STUDY ADDRESS?

We analyzed the pharmacokinetics of berberine in humans. Furthermore, we leveraged naturally occurring genetic polymorphisms to assess the impact of the hepatic uptake transporter OCT1 and metabolizing enzyme CYP2D6.

• WHAT DOES THIS STUDY ADD TO OUR KNOWLEDGE?

Berberine has very low oral bioavailability. OCT1 does not affect berberine pharmacokinetics in humans. CYP2D6‐deficiency causes altered berberine plasma concentrations and changes the urinary metabolites profile in females, but not in males. This suggests the existence of a CYP2D6‐independent metabolic pathway for berberine in males, which is less active in females.

• HOW MIGHT THIS CHANGE CLINICAL PHARMA‐COLOGY OR TRANSLATIONAL SCIENCE?

Our pharmacokinetic analysis suggests that the proposed therapeutic effects of berberine are of intestinal rather than systemic origin. Additionally, we report sex‐specific influences of CYP2D6 on xenobiotic metabolism, which could extend to other substances.

Berberine is a nutritional supplement isolated from the root, rhizome, stem, and bark of plants in the Berberis family (e.g., Hydrastis canadensis, Berberis vulgaris, and Coptis chinensis). Berberine is one of the two major alkaloids of goldenseal (Hydrastis canadensis), a herb that was originally used in traditional Chinese medicine for the treatment of arterial hypertension, type II diabetes mellitus, and dyslipidemia. ¹ The expected low risk of side effects due to the herbal origin are common argument for patients to take berberine as an alternative treatment or as a supplement to their regular therapy. However, goldenseal has been reported to affect the pharmacokinetics of metformin and midazolam in humans. ² Gurley et al. described a decreased CYP2D6 activity after the intake of goldenseal. ³ , ⁴ Nguyen et al. demonstrated that CYP3A4 inhibition, most likely by ß‐hydrastine (the other major alkaloid in goldenseal), results in an increased midazolam AUC in humans after both acute and chronic goldenseal exposure. ⁵ Molecular mechanisms involving interaction of berberine with membrane transporter have been suggested to explain the interaction with metformin. ⁶ , ⁷ These goldenseal‐drug interactions suggest the relevance of CYP enzymes and drug transporters in the pharmacokinetics of berberine. Berberine together with other natural products received increasing scientific attention in the last years, ⁸ , ⁹ but less is known about the pharmacokinetic of berberine itself and the effects of genetic polymorphisms on it.

Structurally, berberine (C₂₀H₁₉NO₅) is a quaternary amine with a molecular weight of 336.4 g/mol. ¹⁰ , ¹¹ Studies in overexpressing cell lines have shown that berberine is a substrate of the organic cation transporter OCT1. ¹² , ¹³ OCT1 is an uptake transporter in hepatocytes that affects the uptake of drugs, such as metformin, morphine, sumatriptan, and fenoterol ¹⁴ , ¹⁵ , ¹⁶ , ¹⁷ and endogenous substrates like thiamine. ¹⁸ OCT1 is genetically highly polymorphic. Six different alleles of the OCT1 gene (SLC22A1) are known in the Caucasian population, ¹⁹ that may result in a severe reduction or a complete loss of function of OCT1 activity. ²⁰ , ²¹ Homozygous or compound heterozygous carriers of the nonfunctional alleles OCT1*3, *4, *5, *6, the so‐called poor OCT1 transporters, are at increased risk of experiencing adverse effects and variations in efficacy and toxicity of OCT1‐transported drugs. ¹⁵ , ¹⁷

Berberine is metabolized in vivo to nine metabolites, three phase I metabolites: berberrubine (M1), demethyleneberberine (M2), and jatrorrhizine (M3), and six phase II metabolites: jatrorrhizine‐3‐O‐glucuronide (M4), jatrorrhizine‐3‐O‐sulfate (M5), thalifendine‐10‐O‐glucuronide (M6), berberrubine‐9‐O‐glucuronide (M7), demethyleneberberine‐2‐O‐sulfate (M8), and demethyleneberberine‐2‐O‐glucuronide (M9). ²² Using human liver microsomes, CYP2D6, and to a lesser extent, CYP1A2 have been confirmed as potential metabolizing enzymes of berberine in humans. ²³ CYP2D6 is also highly genetically polymorphic ²⁴ , ²⁵ and this may additionally affect berberine pharmacokinetics in human.

In this study, we analyzed the effects of OCT1 and CYP2D6 polymorphisms on berberine pharmacokinetics in humans. To this end, we performed in vitro uptake analyses in HEK293 cells overexpressing reference and variant OCT1 alleles, berberine perfusion in the liver of Oct1/2 knockout compared with wild‐type mice, and most importantly, a pharmacokinetic study in healthy volunteers pre‐selected for their OCT1 and CYP2D6 genotypes. Additional aims of the study included analyzing the effects of sex on berberine pharmacokinetics, and exploring the use of berberine as a probe substance for phenotyping OCT1 and CYP2D6 in humans.

METHODS

OCT1 transport activity

Berberine transport was measured using HEK293 cells overexpressing human OCT1 alleles *1 (reference allele), *2 (Met420del), *3 (Arg61Cys), *4 (Gly401Ser), *5 (Gly465Arg, Met420del), *6 (Cys88Arg, Met420del) and control cells stably transfected with the empty pcDNA5 vector. Berberine was obtained from (Santa Cruz Biotechnology, Heidelberg, Germany). The generation of the OCT1 overexpressing has been described previously. ²⁶ HEK293 cells overexpressing OCT3 were a kind gift from Prof. Koepsell and Dr Gorboulev (University of Würzburg, Germany). Berberine uptake was characterized in a concentration‐dependent and time‐dependent manner as described elsewhere. ²⁷ , ²⁸ For concentration‐dependent uptake experiments, berberine concentrations ranging from 0.1 to 50 μM were used and the uptake was measured after 2 minutes of incubation. For time‐dependent experiments, concentrations of 3 nM and 0.1 μM berberine were used and the cells were incubated with berberine for 1, 2, 3, 5, 10, and 45 minutes. For single‐concentration experiments, 3 nM and 0.1 μM berberine were used and uptake was allowed for 2 minutes. Berberine intracellular accumulation was measured using LC–MS/MS as described below. Intracellular berberine was normalized to the amount of total protein in the samples as measured by using the bicinchoninic acid assay. ²⁹

Mouse liver perfusion

Mouse liver perfusion was performed in FVB mice (wild‐type) and in Oct1/2 knockout mice aged 9–16 weeks (mean age was 12 weeks). The FVB mice were obtained from Charles River. The Oct1/2 knockout mice were initially generated by Jonker et al. ³⁰ and obtained from Dr Giuliano Ciarimboli (University Hospital Münster, Münster, Germany). The experimental details are described in the Supplementary Methods.

Clinical study

The Ethics Committee of the University Medicine Greifswald approved the study (registration number BB 078/22). The study was registered on ClinicalTrials.gov (NCT05463003). All participants gave written informed consent before participation.

Three independent groups of the participants were studied, OCT1 poor transporters, CYP2D6 poor metabolizers, and a reference group. Poor OCT1 transporters were defined as homozygous or compound heterozygous carriers of OCT1 alleles *3, *4, *5, or *6. Poor CYP2D6 metabolizers were defined as homozygous or compound heterozygous carriers of CYP2D6 alleles *3, *4, *5 or *6. In the reference group, we included carriers of two fully active alleles of OCT1 (OCT1*1) and CYP2D6 (CYP2D6*1, *2, and *35) without duplication. In the poor OCT1 transporter group, subjects were allowed to carry a maximum of one reduced activity or duplicated CYP2D6 allele. Similarly, the group of CYP2D6 poor metabolizer were selected to carry at least one OCT1 alleles with high function, i.e. OCT1*1 and OCT1*2. Genotyping procedures are described in the Supplementary Methods.

The primary endpoint was genotype‐dependent differences in berberine AUC from administration to the last measurement timepoint (AUC_0–48h). The secondary pharmacokinetic endpoints included genotype‐dependent changes in C _max, T _max, and metabolite‐to‐berberine ratios in urine, as well as sex‐dependent differences in any of these pharmacokinetic parameters.

We aimed to include at least 10 participants in each of the groups of OCT1 poor transporters and CYP2D6 poor metabolizers. This sample size will provide us with 80% power to detect an increase of at least 50% in the AUC, with an alpha error of 0.05. Power calculations were performed using https://clincalc.com/stats/samplesize.aspx. We estimated the standard deviation based on our previous studies involving poor OCT1 transporters. ¹⁶

Participants in the control group were matched with those from the other groups for sex, age, BMI and smoking status. Exclusion criteria were serious pre‐existing medical conditions, regular use of more than three drugs, or use of medications that interact with OCT1 or CYP2D6. In addition, all participants had normal electrocardiogram (ECG) readings, blood pressure within the normotensive range, and clinical chemistry results and blood counts within the normal range (hemoglobin, hematocrit, erythrocytes, leukocytes, thrombocytes, aspartate aminotransferase, alanine aminotransferase, y‐glutamyl transferase, bilirubin, creatinine, glomerular filtration rate (MDRD equation)). The female participants were tested twice to exclude pregnancy: at recruitment and immediately before berberine administration. A single dose of 1,000 mg berberine as two capsules of 500 mg (ProNaturalHealth, Heilbronn, Germany) was administered with 250 mL of non‐carbonated water after an overnight fast. The manufacturer's recommendations vary, but most suggest a daily dose of 1,000–1,500 mg of berberine. Therefore, we decided on a single dose of 1,000 mg. One hundred ml of carbonated water was administered at 1, 2, 3, and 4 hours after berberine intake. Coffee or tea was allowed at 2 hours after berberine intake and a meal was served at the research unit after 4 hours. The meal consisted of three meat‐based and three vegetarian options of ready‐to‐eat meals from which the participants could choose. Blood samples were taken through an intravenous catheter at 1, 1.5, 2, 3, 4, 5, 6, 8, 10, 24, and 48 hours after intake. The EDTA‐blood samples were used and centrifuged for 10 min at 2000  g , 4 °C. The plasma, obtained from the supernatant, was used for the analyses.

Urine was collected from 0 to 4 hours and from 4 to 10 hours. The methods used to determine the concentration of berberine in plasma and berberine and its metabolites in urine are described in the supplementary methods.

Blood pressure was measured at 1, 2, 3, 4, and 10 hours after intake and two additional ECG readings were obtained at 2 and 4 hours after intake.

Pharmacokinetics and statistical analyses

Pharmacokinetic parameters were calculated using Analyst Software v1.7.1, Microsoft Excel, and SimBiology® (MathWorks, Natick, MA, USA). Maximum plasma concentration C _max and time to C _max were directly obtained from the plasma concentration values. The area under the plasma concentration curve from intake to the last measured timepoint (AUC_0–48h) was calculated using the trapezoidal method. For excretion parameters, the amount of berberine excreted (Ae) was calculated as excreted berberine multiplied by urine volume. Renal clearance (CL_R) was calculated as the ratio of the Ae_0–10h to AUC_0–10h. All metabolites‐to‐berberine ratios were calculated using urine concentrations.

Statistical analyses were conducted in SPSS v.29.0.1.0. Due to the non‐normal distribution of data, the median and the 25th and 75th quartiles were used to present them. Statistical analyses were conducted using the Welch's t‐test on log‐transformed values to compare different genotypes or sex groups. Combined sex‐genotype effects were analyzed using linear regression following the logarithmic transformation of the dependent variable (log₂ for M1‐to‐berberine ratio and log₁₀ for AUC).

RESULTS

Berberine as a substrate of OCT1

The ability of OCT1 to mediate the cellular uptake of berberine was analyzed in HEK293 cells stably transfected to overexpress human OCT1. OCT1 rapidly transported berberine. We observed a 4.5‐fold higher berberine uptake in the OCT1‐overexpressing compared with the control cells already after 1 minute of incubation (0.9 ± 0.09 and 0.2 ± 0.03 pmol/mg protein, respectively). Uptake was linear for the first 3 minutes and saturated at 10 minutes (Figure 1 a ). In concentration‐dependent experiments, OCT1 showed high‐affinity uptake of berberine (K _M 7.0 ± 1.6 μM, Figure 1 b ). With a v _max of 2041 ± 318 pmol/min/mg protein, the intrinsic clearance was 306 ± 29 μL/min/mg protein. These intrinsic clearance values are higher than those previously reported for fenoterol and sumatriptan—two OCT1 substrates whose pharmacokinetics have been shown to be dependent on OCT1 in humans. ¹⁶ , ¹⁷

Figure 1.

In vitro evidences for berberine uptake by OCT1. (a) Time‐dependence of berberine uptake in OCT1‐overexpressing (green circles) and control cells stably transfected with the empty vector (open circles). The uptake measurements were performed by applying 3 nM berberine. (b) Concentration dependence of berberine uptake by OCT1 for 2 min. On the left, the absolute uptake values are shown for both OCT1‐overexpressing (green circles) and control cells (open circles). On the right, only the OCT1‐mediated transport is shown (calculated by subtracting the uptake in the control from the uptake in the OCT1‐overexpressing cells). K _M and v _max were determined by nonlinear regression using the Michaelis–Menten equation in GraphPad Prizm 8. All graphs are shown with mean and standard errors of the mean. (c) Comparison of berberine uptake via human OCT1 (green bars) and OCT3 (gray bars) at two different concentrations. The bars represent mean and standard errors of the mean, the dots represent single values measured.

OCT3 is another uptake transporter of organic cation known to be expressed in human hepatocytes. ³¹ OCT3 showed 80% of the OCT1 uptake of berberine at both 3 and at 100 nM (Figure 1 c ). However, OCT3 is more than 15‐fold lower expressed in human hepatocytes than OCT1, ³¹ , ³² suggesting a less prominent contribution to berberine uptake in the liver. Taken together, these data confirm berberine as a high‐affinity substrate of OCT1 and suggest that OCT1 may play an important role in the hepatic uptake of berberine in humans.

Effects of OCT1 polymorphisms on berberine uptake in vitro and of Oct1 knockout on hepatic uptake in mouse

We analyzed the effects of common OCT1 polymorphisms on berberine uptake using overexpressing HEK293 cells (Figure 2 a ). While OCT1*2 retained 82% of the reference allele (OCT1*1) activity, the other common alleles showed less than 40% of the OCT1*1 activity (ranging from 0.7% of the OCT1*1 activity for OCT1*5 to 33% for OCT1*3). Therefore, for the clinical study, we considered *3–*6 as OCT1 alleles with reduced OCT1 uptake activity. Correspondingly, homozygous or compound heterozygous carriers of these alleles were considered to be poor OCT1 transporters of berberine.

Figure 2.

OCT1 genetic variants in humans and lack of Oct1 in knock‐out mice caused a substantial reduction of berberine uptake in vitro and ex vivo. (a) Effects of the six common OCT1 alleles on the cellular uptake of berberine. HEK293 cells stably transfected to overexpress OCT1*1 (reference allele), or the other five common OCT1 alleles were incubated with 3 nM berberine for 2 min. The uptake of the control cells transfected with an empty vector was subtracted from the OCT1‐overexpressing cells and represented as the percentage of OCT1*1 uptake. OCT1 alleles with 75% of OCT1*1 activity or higher are highlighted in green and alleles with 50% or less are highlighted in red. The bars represent the mean and standard errors of the mean, the dots represent single values measured. (b, c) Effects of Oct1 knockout on the berberine concentration in the perfusates collected during mouse liver perfusion. Freshly isolated mouse livers were perfused with 0.1 μM berberine dissolved in HBSS at perfusion rates were 2 mL/min. And 100 nM berberine was added to the perfusion solution between 1 and 11 minutes and washed after that. Wild‐type (n = 8) and Oct1/Oct2 knockout mice (Oct1/2 KO, n = 6) were compared. Mean and standard deviations of perfusate berberine concentrations (b) and berberine AUCs for each mouse (c) are shown. Group comparison by Mann–Whitney‐U test.

Next, we used a mouse liver perfusion model to analyze the role of OCT1 in berberine hepatic uptake in a more complex system. We perfused livers of wild‐type and Oct1/2 knockout mice with a 0.1 μM solution of berberine and measured berberine concentrations in the perfusates. The Oct1/2 knockout mice showed a time‐dependent increase in berberine concentrations that was absent in the wild‐type mice (Figure 2 b ). This resulted in a 3.2‐fold higher AUC of berberine in the knockout compared with the wild‐type mice (P = 0.008, Figure 2 c ). Taken together, these data suggest that lack of OCT1 activity (either due to genetic polymorphisms in human or in the mouse knockout model) may substantially affect berberine hepatic uptake and, as a consequence, the pharmacokinetics of berberine.

Berberine pharmacokinetics in humans

Based on their OCT1 and CYP2D6 genotypes 44 participants were recruited from the healthy participant database of the Institute of Pharmacology (University Medicine Greifswald): 10 poor OCT1 transporters, 13 poor CYP2D6 metabolizers, and 21 matched control participants who carried two fully active OCT1 and CYP2D6 alleles (see Table 1 for the main characteristics of the groups). We administered a single 1,000 mg oral dose of berberine to each participant and measured berberine plasma concentrations at 11 timepoints over 48 hours post‐administration.

Table 1.

Participant demographics and pharmacokinetics of berberine stratified by OCT1 and CYP2D6 genotype and sex

Parameter	All participants (n = 44)		Different Genotypes				Different Sexes
			Reference (n = 21)	OCT1 Poor Transporter (n = 10)	CYP2D6 Poor Metabolizer (n = 13)		Female (n = 22)	Male (n = 22)
Demographic data
Male sex (Number of participants (%))	22 (50%)		10 (48%)	6 (60%)	6 (46%)
Age (years)	23.0 (21.3/25.8)		23.0 (22.0/25.0)	23.5 (19.8/26.3)	23.0 (21.0/25.0)	22.5 (21.0/24.0)		23.5 (21.5/26.3)
BMI (kg/m2)	22.8 (20.9/24.9)		22.5 (21.4/25.0)	23.5 (20.6/25.2)	21.2 (20.5/23.8)	21.5 (20.2/23.2)		23.9 (21.7/25.1)
Tobacco smokersa	6		2	2	2	1		5
Pharmacokinetics
C max (ng/mL)	0.434 (0.218/1.13)		0.441 (0.234/1.09)	0.351 (0.181/1.30)	0.736 (0.204/1.02)	0.851 (0.523/1.30)		0.238*** (0.177/0.431)
T max (h)	3 (2/5)		3 (2/5)	4 (2.63/5.0)	3 (2.5/5)	3 (2/5)		3 (1.5/5)
AUC0–48h (min*ng/mL)	556 (302/985)		578 (391/924)	481 (273/987)	907 (265/1005)	977 (632/1089)		354*** (264/508)
AUC0–10h (min*ng/mL)	154 (85.7/315)		149 (96.6/312)	133 (75.8/360)	270 (70.2/344)	307 (165/377)		94.8*** (67.1/143)
CLR,10 (mL/min)b	45.0 (32.4/55.5)		38.8 (27.5/53.8)	51.2 (38.4/59.1)	46.9 (36.4/72.1)	42.5 (30.1/64.6)		46.9 (36.4/54.8)
M1‐to‐berberine ratio (4–10 h)	0.588 (0.327/0.972)		0.708 (0.551/1.16)	0.472 (0.195/0.826)	0.307* (0.136–0.747)	0.357 (0.171/0.800)		0.706** (0.533/1.11)
M2‐to‐berberine ratio (4–10 h)	0.517 (0.286/0.733)		0.631 (0.476/0.839)	0.456 (0.128/0.644)	0.298** (0.101/0.445)	0.380 (0.127/0.767)		0.519 (0.358/0.646)
M3‐to‐berberine ratio (4–10 h)	1.98 (1.29/3.62)		2.30 (1.60/3.50)	1.48 (0.789–4.06)	1.61 (1.21/3.23)	1.46 (1.04/2.08)		2.92* (1.96/4.51)

Shown are the median and the 25th/75th percentiles except for sex, where the number of individuals and percentage of all are shown. Significance of pharmacokinetic variables are calculated using Welch's t‐test on log‐transformed value for each genotype compared with reference group, or for the two sexes against each other. *P < 0.05, **P < 0.01, ***P < 0.001. The version of the table with log₂ transformed values is available in the supplementary (Table S2 ).

^a Tabacco smocers are shown as number of participants.

^b The renal clearance (CL_R,10) is an estimate based on the amount of drug excreted in urine from 0 to 10 hours following administration, divided by the AUC for the same time period. This may lead to lower values compared with the true CL_R. Due to an incomplete urine collection, we had to exclude one male participant from the reference group at all the 4–10 hours urine parameters.

The median berberine plasma concentration‐time curve is shown in Figure 3 a . The median maximal plasma concentration C _max occurred between 3 and 5 hours after intake (Table 1 , Figure 3 a ). We observed an inter‐individual variability of 29‐fold in AUC_0–48h (Figure 3 b ) and 32‐fold in C _max (Figure 3 c ). At the last timepoint measured (48 hours after intake), the blood concentration was still 28% of the C _max (range 21% up to 41%, Figure 3 a ; Figure S1 ).

Figure 3.

Berberine pharmacokinetics in humans. (a) Median (solid line) and the 25th/75th percentiles (dashed lines) of the berberine plasma concentrations over time of all participants are shown. The berberine concentration at 48 hours after intake is still ~28% of C _max. Insert shows berberine plasma concentration with Y‐axis in a log10 scale. (b, c) Histograms illustrating the variability of AUC and C _max among all participants are shown. (d, f) The amount excreted (Ae) of the phase I and phase II metabolites in the urine of all participants are shown. Ae represents the amount over the entire collection period of 0–10 hours. M3 is the most abundant phase I metabolite and M8 is the most abundant phase II metabolite. (in d and e the y‐axis is shown as a log₁₀ scale).

A median of 6.35 μg berberine was excreted unmodified in the urine. This corresponds to 0.0006% of the oral dose. We also analyzed the main phase I and II metabolites of berberine in urine. The most abundant phase I metabolite was M3 (Figure 3d ).

Due to the lack of reference substances, the absolute quantification of the phase II metabolites was not possible. However, regarding the peak areas in the LC–MS/MS measurements, the apparently most abundant phase II metabolite was M8, followed by M6 and M7 (Figure 3 e ; Table S1 ).

Effects of OCT1 and CYP2D6 deficiency on berberine pharmacokinetics

No significant differences were observed in berberine plasma concentrations or AUC_0–48h between poor OCT1 transporters, poor CYP2D6 metabolizers, and the reference group (Figure 4 a,b, Table 1 ).

Figure 4.

Inter‐individual variability in berberine pharmacokinetics and dependence on OCT1 and CYP2D6 genotypes. (a) Berberine plasma concentrations of reference participants, poor OCT1 transporters, and poor CYP2D6 metabolizers. Represented are individual (in light green, red, and blue lines and closed circles) and median (bold green, red, or blue colors and open circles, respectively). (b) Effects of OCT1 and CYP2D6 genotypes on berberine AUC_0–48h in each participant for the time period of 48 hours after intake. (c) Effects of OCT1 and CYP2D6 genotypes on M1‐to‐berberine ratios in urine collected 4–10 hours after administration. Group comparison by Welch's t‐test on log₂‐transformed values, compared each genotype group to the reference group. M2‐to‐berberine and M3‐to‐berberine ratios are shown in Figure S2 .

In urine, we found a significant reduction in the M1‐to‐berberine and M2‐to‐berberine ratios in 4–10 h urine in poor CYP2D6 metabolizers (42% and 47% less than in the reference group for M1‐to‐berberine and M2‐to‐berberine ratios, respectively, Table 1 , Figure 4 ; Figure S2 ). Similar effects were observed for the phase II metabolites. M6‐to‐berberine ratios were significantly reduced in poor CYP2D6 metabolizers (Table S1 ).

Effect of sex on berberine pharmacokinetics

We observed strong sex‐dependent differences in the pharmacokinetics of berberine. Females showed significantly higher plasma concentrations of berberine than males (Figure 5 a , Table 1 ). This resulted in a 2.8‐fold higher AUC_0–48h in females compared with males (P = 4.1 × 10⁻⁶, Welch's t‐test on log₂ values). C _max was 3.6‐fold higher and AUC_0–10h was 3.2‐fold higher (P < 0.001 for each of the parameters, Welch's t‐test on log₂ values, Table 1 , Figure S3 ). There were no significant sex‐dependent differences in T _max and in the renal clearance. The sex‐dependent difference in AUC_0–48h remained significant (P = 0.002) also after adjustment for age, BMI, and genotype effects.

Figure 5.

Sex‐dependent differences in berberine pharmacokinetics. (a) Means and 95% confidence intervals of the berberine plasma concentrations in females (closed circles) and males (open circles). (b) Berberine area under the time–concentration curve (AUC), in females and males. (c) M1‐berberine ratio in 4–10 h urine of all participants stratified by genotype and sex. Figures with the M2‐to‐berberine and M3‐to‐berberine ratios are shown in the Figure S5 . Significance was calculated using the Welch's t‐test on log₂‐transformed values.

Regarding urinary metabolites, females showed significantly lower M1‐to‐berberine and M3‐to‐berberine ratios (Table 1 ). Looking at phase II metabolites, females showed significantly lower metabolite‐to‐berberine ratios for M4, M5, M6, and M7 (Table S1 ).

Effects of smoking on berberine pharmacokinetics

There were 6 smokers (thereof 5 males) among the study participants. Despite the small number of participants, we observed a highly significant increase in metabolite‐to‐berberine ratios in urine for the phase I metabolite M2 and the resulting phase II metabolites M8 and M9 (Figure S4 ). Male smokers showed a 1.5‐fold higher M2‐ and M8‐to‐berberine ratio and a 2.9‐fold higher M9‐to‐berberine ratio in urine compared with male nonsmokers (P < 0.01 for each). The other phase I metabolite‐to‐berberine ratios, M1 and M3, did not show a significant increase. The metabolite‐to‐berberine ratios of the other phase II metabolites, M4, M6, and M7, were also significantly increased in smokers, but to a lesser extent.

Combined genotype‐sex effects on berberine pharmacokinetics

The effects of CYP2D6 on berberine pharmacokinetics were only observed in females and not in males (Figure 5 c ). Females who were poor CYP2D6 metabolizers showed 79% lower M1‐to‐berberine ratios compared with females with reference genotype (P = 2.3 × 10⁻⁴, Welch's t‐test on log₂ values). In contrast, no genotype‐dependent difference was observed in males. Importantly, males who were poor CYP2D6 metabolizers showed identical M1‐to‐berberine ratios to reference males and females. The same was true for M2‐to‐berberine and M3‐to‐berberine ratios (Figure S5 ). Accordingly, significant differences in metabolite‐to‐berberine ratios between both sexes were observed only in the group of poor CYP2D6 metabolizers.

Analyses using linear regression on log‐transformed variables suggest that the interaction between sex and CYP2D6 genotype is completely responsible for the effects on the metabolites‐to‐berberine ratio in urine (Table 2 ). Consistently, sex‐dependent differences in the metabolites‐to‐berberine ratios were observed only in the poor CYP2D6 metabolizer group. However, the sex‐dependent effects on AUC were observed in all genotype groups (Figure S6B ).

Table 2.

Effects of sex and genotype on systemic berberine exposure (AUC) and metabolite ratios in urine

Characteristic	AUC 0‐48h berberine (min*ng/mL)		Ratio M1/berberine (4–10 hours urine)
	Beta (95% CI)	P‐value	Beta (95% CI)	P‐value
Sex
Male	—		—
Female	0.30 (0.09 to 0.51)	0.007	−0.18 (−0.86 to 0.49)	0.60
Genotype
Reference alleles	—		—
Poor OCT1 transporter	0.04 (−0.21 to 0.29)	0.75	−0.60 (−1.4 to 0.20)	0.15
Poor CYP2D6 metabolizer	−0.17 (−0.42 to 0.08)	0.18	−0.01 (−0.80 to 0.79)	0.99
Sex * genotype
Female * poor OCT1 transporter	−0.02 (−0.40 to 0.35)	0.90	−0.56 (−1.8 to 0.64)	0.37
Female * poor CYP2D6 metabolizer	0.34 (0.00 to 0.68)	0.055	−2.1 (−3.2 to −1.0)	<0.001

CI, confidence interval.

Effects of OCT1 and CYP2D6 genotypes and sex on pharmacodynamics and adverse effects of berberine

We observed a maximal decrease in blood pressure of 1.5 mmHg (median), with a large range −7.0 and +4.5 mmHg in 25th and 75th percentiles, respectively. There was no difference in the change of mean arterial blood pressure between the three genotype groups (Table S1 , Figure S7A ). Females showed an overall lower mean arterial blood pressure than males (Table S1 , Figure S7A ). This is a well‐known sex difference. ³³ No difference was observed at any of the other timepoints measured or in the heart rate of the participants (Figure S7B ).

Regarding potential adverse reactions, five participants (three in the control group and two in the poor OCT1 transporter group) reported a headache ~4 hours after intake and one participant reported moderate abdominal pain in the evening after intake. All adverse events had resolved by the next morning (24 h after intake) and could not be clearly attributed to berberine intake, as no control group without berberine administration was available. We did not observe clinically relevant changes in the liver enzymes ASAT, ALAT, and γGT at 5, 10, 24, and 48 hours after berberine compared with baseline (Figure S8 ).

DISCUSSION

Our analyses showed that CYP2D6 but not OCT1 polymorphisms affect berberine pharmacokinetics in healthy humans. The CYP2D6 effects were strongly sex‐specific and were observed only in females. The lack of OCT1 effects was in contrast to the in vitro and mouse data.

Effects of the lack of CYP2D6 activity were only observed for the differences in metabolite‐to‐berberine ratios, but not in the AUC of berberine itself (Figure 4 , Table 1 ). In vitro data suggested that affected metabolites M1 and M2 are potential products of CYP2D6 metabolism ²² and our data confirmed this in humans.

However, CYP2D6 deficiency affected metabolite‐to‐berberine ratios exclusively in females. Quantitatively, we suggest that CYP2D6 is responsible for producing more than 75% of M1 or M2 metabolites in females, but plays no substantial role in berberine metabolism in males. Apparently, there is another enzyme responsible for berberine metabolism to M1 and M2 in males. That CYP2D6 itself is not responsible for the sex differences is in line with a previous study, which found no differences in CYP2D6 activity between sexes and no variation during different phases of the menstrual cycle. ³⁴ One possible candidate for stronger metabolizing of berberine in males is CYP1A2. On one hand, experiments using recombinant human CYP enzymes suggest that CYP1A2 is the second strongest producer of M1 and M2. ²³ On the other hand, CYP1A2 is also known to be more active in males than in females. ³⁵ , ³⁶ Furthermore, CYP1A2 expression and activity are known to increase with tobacco smoking. ³⁵ , ³⁷ , ³⁸ In our study, tobacco smokers showed significantly higher M2‐to‐berberine ratios and increased levels of phase II metabolites M8 and M9. M2 is the primary metabolite produced by CYP1A2. ²³ These findings support our hypothesis that CYP1A2 compensates for the lack of CYP2D6. Whether CYP1A2 is the only responsible enzyme for the stronger metabolism in males remains to be further analyzed.

Our statistical model suggests that the interaction between the female sex and CYP2D6 genotype, but not the female sex alone, affects metabolite‐to‐berberine ratios in urine. Current systematic analyses using UK biobank data suggest that such polymorphism–sex interactions are not very common, but are present in some phenotypes. ³⁹ Indeed, there also are some recent data about sex‐dependent effects of CYP2D6 genotypes. Lopes et al. report that opioid addiction was dependent on CYP2D6 genotypes only in females. ⁴⁰

Sex differences in the berberine pharmacokinetics have not been described by others. ⁴¹ , ⁴² On the one hand, although not our primary endpoint, due to the higher number of individuals, we were better powered to detect such differences than the previous studies. On the other hand, and maybe a more plausible explanation, we have enriched our population with poor CYP2D6 metabolizers and poor OCT1 transporters. As the sex differences were much stronger in these populations (especially in poor CYP2D6 metabolizers), we may overestimate the sex effects in an unselected population.

To the best of our knowledge, with 44 healthy participants, this is the largest study analyzing berberine pharmacokinetics in humans to date. The number of participants in this study is twice as large as in the second‐biggest study on berberine pharmacokinetics. ⁴³ , ⁴⁴ Furthermore, analyzing major metabolites of berberine in humans is unique.

Although not determined experimentally, our results indirectly suggest that berberine has very low oral bioavailability. A median of only 0.003% (mol/mol) of the orally administrated berberine dose either was recovered in the urine as unmetabolized berberine (0.0006%) or as any phase I metabolite (0.0026%). We were unable to quantify the phase II metabolites due to a lack of reference substances. However, even if we assume that the amount of phase II metabolites exceeds the amount of all other products excreted in urine by a factor of 100, oral bioavailability would still be less than 1%. This is in line with an absolute oral bioavailability of 0.68% that was experimentally determined in rats. ⁴⁵ The low oral bioavailability of berberine suggests that the proposed therapeutic effects of berberine are more likely to be due to direct intestinal actions than to systemic effects.

Our initial hypothesis that the low oral bioavailability of berberine stems from an extensive first‐pass effect through OCT1 and CYP2D6 synergy was not supported, as deficiencies in these transporter and enzymes did not significantly affect berberine plasma levels. Alternative explanations might include poor solubility, low intestinal absorption, or significant first‐pass metabolism of berberine occurring outside the liver. A solubility of 9 mM berberine in water has been reported before. ¹¹ In our hands, the maximal solubility was 2 mM. Using this solubility rate and the 250 mL water volume for oral administration, we estimate that 186 mg, or 19% of the dose, dissolves. Therefore, solubility issues alone cannot fully explain the notably low oral bioavailability of berberine.

Further, the limited reduction of berberine plasma concentrations between 10 and 48 hours after intake suggests an extensive enterohepatic circulation of berberine. The substantial enterohepatic circulation is further supported by the second peak in berberine plasma concentrations observed 5 hours post‐intake (Figure 3 ; Figure S1 ). Four hours after administration, participants were served lunch. This double peak likely indicates the enterohepatic circulation of berberine, where food intake triggers gallbladder emptying. ⁴⁶ All this may account for the huge variances in berberine plasma concentrations between the participants. Based on our results, however, it seems that if there is an enterohepatic circulation of berberine, it is most likely OCT1‐independent.

Our study (Figure 1 ) and others before ¹² , ¹³ showed that berberine is a substrate of OCT1 in vitro. Also, the model of mouse liver perfusion, which we applied, suggests significant effects of OCT1 on berberine plasma concentrations (Figure 2 ). In contrast, we did not observe any significant effects of OCT1 genotypes on berberine pharmacokinetics in humans. One potential explanation is that other major factors like low oral bioavailability are the major determinants of berberine plasma concentrations. Another explanation could be the presence of an as‐yet‐undiscovered transporter mediating berberine uptake in the human liver. Indeed our in vitro data are based on OCT1‐overexpressing (gain‐of‐function) cell models and based on this we cannot exclude the involvement of other endogenous transporters in the hepatic uptake of berberine. We exclude OCT3 as this alternative transporter because at low berberine concentrations, close to those measured in human plasma, OCT3 transport rates are comparable with those of OCT1 but OCT3 expression is 15‐fold lower compared with OCT1. ³¹ , ³² Thus, OCT3 may contribute only to a very limited extend to berberine uptake in hepatocytes. There is information in the literature on the existence of alternative transport mechanisms for organic cations in human hepatocytes like the human proton organic cation antiporter (H⁺OC). ⁴⁷ Several substrates of OCT1 are substrates of H⁺OC, too, whether berberine also belongs to the substrate spectrum of this transporter is unknown. ⁴⁸

One aim of our study was to evaluate berberine as potential probe substance for phenotyping OCT1 and CYP2D6 in humans. Given berberine's status as an over‐the‐counter natural product and its demonstrated in vitro transport by OCT1 and metabolism by CYP2D6, it emerged as a promising candidate for such a probe, for instance, in evaluating drug–drug interactions involving OCT1. However, our data clearly show that berberine cannot be used as an in vivo probe for OCT1 and is subordinate as a CYP2D6 probe compared with dextromethorphan ⁴⁹ or more recently suggested yohimbine. ⁵⁰

Our study has several limitations. First, we could not quantify the phase II metabolites of berberine due to a lack of reference substances. Therefore, we could not determine the total amount of berberine metabolites in urine and thus have a more accurate estimate of its oral bioavailability. Second, the determination of sex differences was not the primary endpoint of this study, thus we did not specifically match the different sexes for participant characteristics. The sample size of the study was not calculated to identify sex‐specific differences, but we conducted an a priori power calculation for an F‐test in a multiple linear regression with 5 predictors and 2 predictors, testing a large effect with a recommended value of f ² = 0.35. This results in a total sample size of 40 participants to achieve 90% power at alpha = 5%. This means that strong sex‐dependent effects, such as CYP2D6‐dependent effects on metabolic ratios, can be demonstrated with the given data. Less pronounced effects of OCT1 on berberine metabolism cannot be excluded (Figures 4 c and 5 c ). A higher number of poor OCT1 transporters, especially females, is needed to address this. However, all participants were young and healthy. Therefore, even without matching, the differences between both groups were limited. Another limitation of our study is that the duration of sample collection was insufficient to accurately estimate AUC_0‐inf and related parameters, such as oral clearance.

In conclusion, CYP2D6 deficiency strongly reduced berberine metabolism only in females and is apparently not the rate‐limiting metabolizing step in males. Although clearly an in vitro substrate of OCT1, berberine pharmacokinetics in humans were not strongly dependent on OCT1 polymorphisms. The low oral bioavailability observed by us questions possible systemic therapeutic effects of berberine.

CONFLICT OF INTEREST

All other authors declared no competing interests in this work.

FUNDING

No funding was received for this work.

AUTHOR CONTRIBUTIONS

J.A.B., M.T, and S.E. wrote the manuscript; J.A.B., M.M‐T., S.E., and M.T. designed the research; J.A.B., F.M., J.B., and V.R. performed the research; J.A.B., M.M.‐T., F.M., M.V., and V.R. analyzed the data; F.M., V.R., M.M‐T., and M.T. contributed new reagents/analytical tools.

Supporting information

Data S1.