Multidrug Transporter ABCG2/Breast Cancer Resistance Protein Secretes Riboflavin (Vitamin B₂) into Milk

Abstract

The multidrug transporter breast cancer resistance protein (BCRP/ABCG2) is strongly induced in the mammary gland during pregnancy and lactation. We here demonstrate that BCRP is responsible for pumping riboflavin (vitamin B₂) into milk, thus supplying the young with this important nutrient. In Bcrp1^−/− mice, milk secretion of riboflavin was reduced >60-fold compared to that in wild-type mice. Yet, under laboratory conditions, Bcrp1^−/− pups showed no riboflavin deficiency due to concomitant milk secretion of its cofactor flavin adenine dinucleotide, which was not affected. Thus, two independent secretion mechanisms supply vitamin B₂ equivalents to milk. BCRP is the first active riboflavin efflux transporter identified in mammals and the first transporter shown to concentrate a vitamin into milk. BCRP activity elsewhere in the body protects against xenotoxins by reducing their absorption and mediating their excretion. Indeed, Bcrp1 activity increased excretion of riboflavin into the intestine and decreased its systemic availability in adult mice. Surprisingly, the paradoxical dual utilization of BCRP as a xenotoxin and a riboflavin pump is evolutionarily conserved among mammals as diverse as mice and humans. This study establishes the principle that an ABC transporter can transport a vitamin into milk and raises the possibility that other vitamins and nutrients are likewise secreted into milk by ABC transporters.

BCRP belongs to the ATP binding cassette (ABC) family of transmembrane drug transporters. It has a broad substrate specificity and actively extrudes a wide variety of drugs, carcinogens, and dietary toxins from cells (19). By its apical localization in epithelia of intestine, kidney, and placenta and in the hepatocyte bile canalicular membrane, Bcrp1/BCRP can reduce the systemic and tissue uptake of its substrates and mediate their extrusion from the body. It thus protects the body from harmful xenotoxins (12, 18). We recently demonstrated that BCRP is expressed in mammary gland alveolar epithelial cells during pregnancy and lactation, where it actively secretes a variety of drugs, toxins, and carcinogens into milk (13). In apparent contradiction with BCRP's detoxifying role elsewhere in the body, this BCRP-mediated contamination of milk exposes suckling infants and young to xenotoxins. Yet, the evolutionarily conserved expression and induction of BCRP in the lactating mammary glands of mice, cows, and humans point to an important physiological function. We considered that BCRP in the mammary gland might function to provide the milk with an essential nutrient, offsetting the coincident risk of contaminating milk with xenotoxins. However, obvious candidates, such as the known BCRP substrates folic acid and dehydroepiandrosterone-sulfate or the porphyrin vitamin B₁₂, were found not to be actively secreted into milk by Bcrp1 (12, 13, 19).

One of the important vitamins occurring in milk is riboflavin (vitamin B₂). Milk is a primary dietary source of riboflavin, and vitamin B₂ deficiency is often endemic in human populations that subsist on diets poor in dairy products and meat (17). Riboflavin is converted to the essential coenzymes flavin mononucleotide (FMN) and flavin adenine dinucleotide (FAD), which participate as electron carriers in many key enzymatic redox reactions in the body (Fig. 1) (17). Riboflavin cannot be synthesized by mammals, and there is a only limited, short-term storage capacity for this vitamin in the liver (8, 17). Humans are vulnerable to develop a vitamin B₂ deficiency during periods of dietary deprivation or stress, and this may lead to a variety of clinical abnormalities, including growth retardation, anemia, skin lesions, and degenerative changes in the nervous system (7). Therefore, this water-soluble vitamin should be present in the diet on a daily basis.

FIG. 1.

Molecular structures of riboflavin (vitamin B₂) (R = R1) and its two derived cofactors, FMN (R = R2) and FAD (R = R3).

Huang and Swaan (9) demonstrated an apically directed transport of riboflavin in Caco-2 cells and rat ileum, suggesting the existence of an apically located transport mechanism. We considered that the presence of Bcrp1 in small and large intestines could be responsible for this vectorial transport of riboflavin. Because Bcrp1 is expressed in the lactating mammary gland, where it actively secretes its substrates into milk, we investigated the possible role of Bcrp1 in riboflavin transport into milk.

MATERIALS AND METHODS

Animals.

Mice were housed and handled according to institutional guidelines complying with Dutch legislation. Animals used were Bcrp1^−/− and wild-type mice, all of >99% FVB genetic background. They received a standard diet containing 12 to 14 mg/kg riboflavin (AM-II; Hope Farms, Woerden, The Netherlands) or diets containing 6 mg/kg or 1 mg/kg riboflavin or without riboflavin supplementation (Dyets Inc., Bethlehem, PA) and acidified water ad libitum.

Chemicals.

[³H]riboflavin (20 Ci mmol⁻¹), riboflavin, FMN, and FAD were from Sigma Chemical Co. (St. Louis, MO). d-[carbonyl-¹⁴C]biotin (58.0 mCi mmol⁻¹) was from Amersham Biosciences (Little Chalfont, Buckinghamshire, United Kingdom). Methoxyflurane (Metofane) was from Medical Developments Australia Pty. Ltd. (Springvale, Victoria, Australia).

In vitro transport assays.

Transwell transport assays were carried out as previously described (18). Vectorial transport of [³H]riboflavin (10 nM and 500 nM, in addition to an estimated 200 nM in Optimem culture medium) across the monolayer was determined, either with or without the Bcrp1 inhibitor Ko143 (1). Vesicular transport assays were performed as described at pH 7.4 (4), using vesicles from Sf9 cells infected with human BCRP-expressing or wild-type baculovirus in the presence or absence of 4 mM ATP.

Western analysis.

Milk retrieved from wild-type or Bcrp1^−/− mice was loaded on a polyacrylamide gel in dilution series. Blots were probed with horseradish peroxidase-labeled antibody against murine immunoglobulin G (IgG) or IgA (sc-2005 and sc-3791; Santa Cruz Laboratories, CA). Equal loading across the lanes was confirmed with total protein staining (Ponceau S and India ink).

Hematological analysis of plasma.

Standard hematological analyses on plasma were performed on a Beckman Coulter A^c T diff analyzer (Beckman Coulter, Mijdrecht, The Netherlands) to determine white and red blood cell counts, hemoglobin, hematocrit, and platelets.

Pharmacokinetic experiments.

For the determination of radiolabeled riboflavin or biotin in plasma, milk, and organs, riboflavin or biotin was administered by intravenous (i.v.) injection into the tail veins of mice lightly anesthetized with methoxyflurane. Riboflavin and biotin were dissolved in phosphate-buffered saline. Plasma of adult males, virgin females, lactating females, and pups for measurement of endogenous levels of riboflavin, FMN and FAD, or radiolabeled riboflavin or biotin was obtained by cardiac puncture under methoxyflurane anesthesia. Mice were sacrificed by cervical dislocation or, in the case of pups, decapitation. Organs were removed and solubilized in Solvable (Perkin-Elmer, MA). Intestinal contents were separated from intestinal tissue. Milk was collected from lactating wild-type or Bcrp1^−/− females with pups of ∼10 days old. Approximately 10 min prior to milk sampling, mice received oxytocin (200 μl of a 1-IU/ml solution) subcutaneously to stimulate milk secretion. Milk samples were collected bilaterally from all mammary glands by gentle vacuum suction, and blood samples were taken by cardiac puncture under methoxyflurane anesthesia. Levels of radioactivity in plasma, milk, feces, and tissue homogenates were determined by liquid scintillation counting. Endogenous levels of riboflavin, FMN, and FAD were quantified by high-pressure liquid chromatography (HPLC).

HPLC analysis of riboflavin, FMN, and FAD.

Riboflavin, FMN, and FAD were determined as described previously (22), with modifications. Samples were protected from light. Methanol (four times the sample volume) was added to plasma and milk samples to precipitate proteins. After mixing, samples were centrifuged for 10 min at 10,500 × g. Supernatants were evaporated to dryness at 40°C under a stream of nitrogen. Samples were reconstituted in 100 μl of methanol-50 mM ammonium acetate (pH 5) (30:70, vol/vol), and 25-μl aliquots were injected into the HPLC system. The chromatographic system consisted of a Perkin-Elmer 200 series pump and ISS 200 autosampler (Perkin-Elmer, Norwalk, CT). Chromatographic separation was performed on a Zorbax SB C₁₈ column (150 by 4.6 mm [inside diameter]; 3.5-μm particle size) (Rockland Technologies Inc., Newport, DE). The mobile phase consisted of a mixture of 50 mM ammonium acetate (pH 5) and methanol, and the following gradient was used: 0 to 5 min with 30% to 90% methanol, 5 to 7 min with 90% methanol, 7 to 8 min with 90% to 30% methanol, and 8 to 12 min with 30% methanol. The flow rate was 1.0 ml/min, and the detection was performed fluorimetrically using a FP920 Intelligent fluorescence detector (Jasco International Co. Ltd., Tokyo, Japan) with excitation and emission wavelengths set at 372 and 520 nm, respectively, and a 40-nm bandwidth. The capacity of the flow cell of the fluorescence detector was 16 μl. Riboflavin, FMN, and FAD eluted after approximately 5.3, 3.9, and 2.5 min, respectively. Processed samples were stable in the autosampler for at least 24 h. Calibration concentrations were between 0.001 and 10 μg/ml. Calibration curves were calculated by least-squares linear regression using a weighting factor of the reciprocal of the concentration. The lower limit of quantification was 0.01 μg/ml.

Measurements of vitamins in milk.

Milk samples retrieved from wild-type or Bcrp1^−/− lactating dams were pooled (n = 3), and the vitamins listed in Table 2 were measured (by TNO, Zeist, The Netherlands) by HPLC (vitamins A, B₁, B₆, C, E, and K₁), competitive protein binding (vitamin B₁₂), or microbiological turbidity assay (vitamin B₉ and biotin). Biotin was measured in separate milk samples of wild-type and Bcrp1^−/− mice (n = 3).

TABLE 2.

Levels of vitamins in wild-type versus Bcrp1^−/− milk

Compound	Concna in:
	Wild-type mice	Bcrp1−/− mice
Vitamin A (retinol)	1,300 IE/100 ml	1,200 IE/100 ml
Vitamin B1 (thiamine)	0.25 mg/100 ml	0.26 mg/100 ml
Vitamin B2 (riboflavin)	0.23 mg/100 ml	0.0036 mg/100 ml
Vitamin B6 (pyridoxine)	0.10 mg/100 ml	0.16 mg/100 ml
Vitamin B9 (folate)	31 μg/100 ml	22 μg/100 ml
Vitamin B12 (cyanocobalamin)	7.6 μg/100 ml	8.7 μg/100 ml
Vitamin C (ascorbic acid)	0.64 mg/100 ml	0.68 mg/100 ml
Vitamin E (tocopherol)	0.83 mg/100 ml	0.96 mg/100 ml
Vitamin H (biotin)	163 μg/100 ml	47 μg/100 ml
Vitamin K1 (phylloquinone)	0.59 μg/100 ml	0.59 μg/100 ml

^aMilk samples from dams of each genotype (n = 3) were pooled, and vitamin concentrations were measured. Values for vitamin B2 (riboflavin) and vitamin H (biotin) are means of values for separately measured samples (n = 3).

Pharmacokinetic calculations and statistical analysis.

All values are given as means ± standard deviations (SD), unless indicated otherwise. Mean concentrations (nanomolar) for each time point were used to calculate the area under the plasma concentration-versus-time curve from time zero to the last sampling point by the linear trapezoidal rule; standard errors were calculated by the law of propagation of errors (2). A two-tailed unpaired Student t test was used to assess the significance of differences between two sets of data. Differences were considered to be statistically significant when the P value was <0.05.

RESULTS

Consequences of absence of Bcrp1 expression in the mammary gland.

We hypothesized that Bcrp1 could secrete nutrients or other essential compounds into milk and monitored the health status of heterozygous Bcrp1^+/− pups nurtured by Bcrp1^−/− or wild-type mothers. This approach circumvents independent effects of the pup genotype itself, including prenatal effects of fetal Bcrp1 expression in the placental barrier. Despite the highly conserved expression of Bcrp1 in the mammary gland, we could not detect differences in hematological parameters (white and red blood cell counts, hemoglobin, hematocrit, and platelets) between pups nurtured by wild-type or Bcrp1^−/− dams. There was also no significant difference in liver or total body weight between 5-day-old pups nurtured by wild-type or Bcrp1^−/− mothers fed on standard chow (n = 10 two neonates/litter; total body weight, 3.05 ± 0.4 versus 2.82 ± 0.5 g [P = 0.27]; liver weight, 0.11 ± 0.0 versus 0.10 ± 0.0 g [P = 0.48]). In addition to nutrient transfer, an immunological protective role has been attributed to milk, and we therefore assessed IgG or IgA contents in wild-type and Bcrp1^−/− milk by Western blot analysis. No differences were found, nor were there differences in the general milk protein content or composition as detected by India ink and Ponceau S staining of Western blots (not shown).

Bcrp1 secretes [³H]riboflavin into milk.

To investigate the role of Bcrp1 in the secretion of riboflavin into milk, we administered a trace amount of [³H]riboflavin (1.67 nmol/kg) i.v. to lactating wild-type and Bcrp1^−/− mice and collected blood and milk at 30 min after administration. We observed a 13-fold lower concentration of [³H]riboflavin in milk from Bcrp1^−/− versus wild-type mice (Fig. 2A). After correction for plasma [³H]riboflavin concentrations, which were 1.7-fold lower in wild-type than in Bcrp1^−/− mice, we observed a 22-fold difference in the milk-to-plasma ratio of [³H]riboflavin (Fig. 2A). These data indicate that Bcrp1 plays a substantial role in the secretion of riboflavin into milk.

FIG. 2.

(A) Milk and maternal plasma riboflavin concentrations (nM equivalent) and milk-to-plasma ratio at 30 min after i.v. administration of [³H]riboflavin (1.67 nmol/kg) to male wild-type and Bcrp1^−/− mice. (B) Plasma riboflavin concentrations (nM equivalent) at 7.5, 15, 30, and 60 min after i.v. administration of [³H]riboflavin (1.67 nmol/kg) to male wild-type and Bcrp1^−/− mice. (C) Tissue distribution and accumulation in intestinal contents (percentage of administered dose) of [³H]riboflavin in male wild-type and Bcrp1^−/− mice 30 min after i.v. administration of [³H]riboflavin (1.67 nmol/kg). L, liver; K, kidney; SI, small intestine; C, colon. Data represent the means; error bars indicate SD (n = 3) (*, P < 0.05; **, P < 0.01; ***, P < 0.001).

Role of Bcrp1 in systemic and tissue availability and excretion of [³H]riboflavin.

To assess whether Bcrp1 expression influences systemic riboflavin availability, we administered a trace amount of [³H]riboflavin (1.67 nmol/kg) i.v. to male Bcrp1^−/− and wild-type mice and measured plasma radioactivity levels (Fig. 2B). The area under the plasma concentration-versus-time curve from 7.5 to 60 min was 1.7-fold higher in Bcrp1^−/− than in wild-type mice (0.52 ± 0.01 versus 0.31 ± 0.01 h·nM; P < 0.01) (Fig. 2B), supporting the idea that Bcrp1 contributes to riboflavin elimination.

The [³H]riboflavin tissue distribution was determined at 30 min after i.v. administration (1.67 nmol/kg). There were 1.8-fold-higher plasma and liver [³H]riboflavin levels in Bcrp1^−/− than in wild-type mice (Table 1). Levels of [³H]riboflavin in most other organs were similarly increased. The percentage of the dose of [³H]riboflavin retrieved from the contents of small intestine was 6.2-fold lower in Bcrp1^−/− than in wild-type mice (Fig. 2C), indicating substantial Bcrp1-mediated hepatobiliary or direct intestinal excretion of [³H]riboflavin into the lumen of this organ. The same appeared to occur in colon (Fig. 2C). Bcrp1 thus has a marked influence on riboflavin availability in plasma and tissues and on its excretion into the gut lumen.

TABLE 1.

Levels of radioactivity in male wild-type and Bcrp1^−/− mice at 30 min after i.v. administration of 1.67 nmol/kg [³H]riboflavin

Tissue	Avg 3H concn (nmol equivalent kg−1 or liter−1) ± SDa in:		Concn ratio (Bcrp1−/−/wild type)
	Wild-type mice	Bcrp1−/− mice
Plasma	0.26 ± 0.01	0.47 ± 0.04**	1.8
Brain	0.27 ± 0.01	0.36 ± 0.02**	1.3
Spleen	1.22 ± 0.05	1.45 ± 0.13*	1.2
Kidney	3.67 ± 0.32	5.23 ± 0.75*	1.4
Liver	2.32 ± 0.21	4.13 ± 0.01***	1.8
Small intestine	3.76 ± 0.76	2.95 ± 0.33	0.8
Colon	2.04 ± 0.18	2.31 ± 0.11	1.1

^an = 3. *, P < 0.05; **, P < 0.01; ***, P < 0.001.

The dose of 1.67 nmol/kg [³H]riboflavin (∼0.05 nmol per mouse) is comparatively low, as substantially more endogenous riboflavin is present in the blood compartment alone (in male wild-type mice, ∼0.2 nmol). At 30 min after a considerably higher i.v. dose (664 nmol/kg, or ∼20 nmol per mouse [i.e., a 100-fold excess over the blood riboflavin pool]), there was again a twofold-higher plasma [³H]riboflavin level in Bcrp1^−/− than in wild-type mice (72 ± 8.8 versus 37 ± 8.5 nM; P < 0.01), and liver accumulation of [³H]riboflavin was 1.5-fold higher in Bcrp1^−/− than in wild-type mice (9.5% ± 0.5 versus 6.2% ± 1.9; P < 0.05). Bcrp1 thus has a marked effect on riboflavin pharmacokinetics at both low and high exposure levels.

Riboflavin transport by murine Bcrp1 and human BCRP in vitro.

Riboflavin transport by Bcrp1 in vitro was tested using the polarized canine kidney cell line MDCK-II and a subclone transduced with murine Bcrp1 cDNA. Cell lines were grown to confluent polarized monolayers on porous membrane filters, and vectorial transport of [³H]riboflavin across the monolayer was determined. Parental MDCK-II cells displayed a high basolaterally directed translocation of [³H]riboflavin, whereas apically directed translocation was very low, indicating the presence of an active transepithelial absorptive riboflavin transport process (Fig. 3A). This is in concordance with the accumulation of riboflavin in basolateral spaces and fluid-filled “domes” of MDCK-II cells (15) and may reflect the mechanism(s) responsible for reabsorption of riboflavin in the kidney and possibly in the intestine (9, 20). Bcrp1 is apically located and should transport its substrates to the apical side of the monolayer, possibly counteracting the endogenous absorptive process (19). Indeed, in Bcrp1-transduced MDCK-II cells the basolaterally directed translocation was markedly decreased (5.2-fold) and the apically directed translocation was increased (3.3-fold), resulting in net apical transport of riboflavin (Fig. 3B). Similar results were obtained with 10 or 500 nM riboflavin. The apical riboflavin transport by Bcrp1 was completely inhibited by Ko143, a selective Bcrp1 inhibitor (Fig. 3C and D) (1). Qualitatively similar results were obtained with MDCK-II cells expressing human BCRP (not shown). MDCK-II or polarized pig kidney LLC-PK1 cells expressing human multidrug resistance protein 2, human MDR1, or murine Mdr1a did not display apically directed transport of riboflavin (not shown).

FIG. 3.

(A to D) Transepithelial transport of [³H]riboflavin (10 nM) in MDCK-II cells, either nontransduced (A and C) or transduced with murine Bcrp1 cDNA (B and D). Ko143 (5 μM) was added where indicated (C and D). At t = 0, radiolabeled riboflavin was applied in one compartment (basolateral or apical), and the percentage of radioactivity appearing in the opposite compartment at t = 1, 2, 3, and 4 h was measured (n = 3). ○, translocation from the basolateral to the apical compartment; •, translocation from the apical to the basolateral compartment. Error bars (often smaller than the symbols) indicate SD. (E) BCRP-mediated uptake of [³H]riboflavin (0.25 μM) in membrane vesicles from Sf9 cells infected with BCRP-expressing or wild-type baculovirus. Data represent the mean uptake ± standard error at various time points (n = 6 to 9; **, P < 0.01). Uptake of [³H]riboflavin in Sf9-BCRP vesicles in the presence of ATP was significantly higher than uptake in the absence of ATP in these vesicles or uptake in Sf9-wild-type vesicles in the presence or absence of ATP, at all time points (P < 0.01).

Riboflavin transport by human BCRP was further investigated using inside-out vesicles obtained from BCRP- or wild-type baculovirus-infected Sf9 cells. Vesicular uptake over time confirmed ATP-dependent transport of [³H]riboflavin by human BCRP (Fig. 3E). BCRP-dependent riboflavin secretion into milk and other pharmacokinetic changes may thus be explained by riboflavin transport by mouse Bcrp1 and human BCRP.

Endogenous plasma riboflavin levels display Bcrp1-dependent gender-specific differences.

Endogenous riboflavin and its cofactors FMN and FAD were measured by HPLC in plasma from adult wild-type and Bcrp1^−/− mice fed a standard chow, which contains 12 to 14 mg/kg riboflavin (Fig. 4). Both sexes were tested, as males have higher hepatic levels of Bcrp1 than females, resulting in gender-specific pharmacokinetic differences (16). Whereas FMN levels did not differ between wild-type and Bcrp1^−/− mice or between sexes, riboflavin levels in male Bcrp1^−/− mice were 1.5-fold higher than those in male wild-type mice. In contrast, females did not display differences in endogenous plasma riboflavin levels, and these also were not different from the levels in Bcrp1^−/− males (Fig. 4). Endogenous plasma riboflavin levels thus display Bcrp1-dependent gender-specific differences. Differences in FAD levels between wild-type and Bcrp1^−/− mice were minor in both sexes (Fig. 4).

FIG. 4.

Endogenous levels of riboflavin (Rf) and its coenzyme forms FMN and FAD (detected by HPLC) in plasma of adult wild-type and Bcrp1^−/− males and females fed normal chow containing 12 to 14 mg/kg riboflavin. Data represent the means; error bars indicate SD (n = 5 to 14) (*, P < 0.05; **, P < 0.01; ***, P < 0.001).

Bcrp1 secretes riboflavin into milk, and this secretion is dose dependent.

When we measured endogenous riboflavin levels in milk samples from lactating females kept on standard chow, we observed a 63-fold-lower milk secretion of riboflavin in Bcrp1^−/− than in wild-type mice (96 ± 22 versus 6073 ± 631 nM; P < 0.001) (Fig. 5A). Whereas in wild-type mice riboflavin was actively concentrated into milk, as indicated by a milk-to-plasma ratio of 25 ± 6.7, this active secretion was abolished in the Bcrp1^−/− mice (milk-to-plasma ratio, 0.37 ± 0.11; P < 0.001). Bcrp1 can thus actively pump endogenous riboflavin into milk. Levels of FMN, the phosphorylated form of riboflavin, were reduced ca. sixfold in Bcrp1^−/− milk compared to wild-type milk (125 ± 20 versus 719 ± 174 nM; P < 0.001), suggesting that it may also be transported into milk by Bcrp1 (milk-to-plasma ratios, 26 ± 20 in wild-type mice and 8.3 ± 3.6 in Bcrp1^−/− mice).

FIG. 5.

(A) Endogenous levels of riboflavin, FMN, and FAD (detected by HPLC) in milk from lactating wild-type and Bcrp1^−/− dams fed a diet containing 12 to 14 mg/kg riboflavin. (B) Secretion of [³H]riboflavin into milk 1 h after oral supplementation (1 mg/kg) of wild-type and Bcrp1^−/− lactating dams fed a riboflavin-deficient diet for the previous 5 days. (C) Endogenous levels of riboflavin, FMN, and FAD in plasma of 7- to 12-day-old milk-fed pups (Bcrp1^+/−) nurtured by wild-type or Bcrp1^−/− dams fed a diet containing 12 to 14 mg/kg riboflavin. LLQ, lower limit of quantification. Data represent the means; error bars indicate SD (n = 3 to 6) (**, P < 0.01; ***, P < 0.001).

Our standard chow contains a high level of riboflavin (12 to 14 mg/kg). Whether maternal riboflavin status influences riboflavin secretion in milk was investigated by changing the food conditions of lactating mice to chows with less riboflavin, i.e., 6 mg/kg or (deficiency inducing) 1 mg/kg, throughout lactation. Whereas wild-type females on standard chow produced milk with ∼6,000 nM riboflavin (Bcrp1^−/−, 96 nM), on 6-mg/kg and 1-mg/kg riboflavin diets this was decreased to 416 ± 60 and 20 ± 27 nM, respectively. Maternal intake of riboflavin thus has a profound effect on Bcrp1-dependent riboflavin secretion into milk.

We investigated whether oral supplementation of riboflavin to the lactating mother could increase the riboflavin content of the milk. Indeed, a single oral dose of riboflavin (1 mg/kg, equivalent to the daily intake on standard chow) given to dams which had been on chow without added riboflavin for 5 days partly restored milk riboflavin secretion 1 hour after administration, and this was Bcrp1 dependent (Fig. 5B). The milk riboflavin concentration after supplementation in these dams was still far lower than that in dams on standard chow (containing 12 to 14 mg/kg riboflavin), probably due to the depleted riboflavin state of the dams on deficient chow (compare Fig. 5A and B).

Riboflavin equivalents in pups from Bcrp1^−/− mothers are due to cofactor FAD in milk.

To assess the consequences of riboflavin deficiency in milk for pups, we analyzed the riboflavin status of Bcrp1^+/− pups nurtured by wild-type and Bcrp1^−/− mothers. This avoids confounding effects from different handling of riboflavin by Bcrp1^+/+ and Bcrp1^−/− pups. Due to the almost complete absence of riboflavin in the milk of Bcrp1^−/− lactating dams, plasma riboflavin levels in 7- to 12-day-old exclusively milk-fed pups with Bcrp1^−/− dams were 1.6-fold decreased compared to those in pups with wild-type dams (84 ± 19 versus 134 ± 27 nM; P < 0.01 [dams were on standard chow]). This is a very modest reduction considering the 63-fold difference in milk riboflavin secretion (Fig. 5A and C). Moreover, despite negligible amounts of riboflavin in Bcrp1^−/− milk, pups nurtured by Bcrp1^−/− dams thrive (12, 19), contradicting a severe riboflavin shortage.

Riboflavin can also be derived from dietary FMN and FAD, which are broken down in the intestinal lumen to riboflavin before absorption (8). When we tested milk from mothers on standard chow, FMN levels were quite low and ca. sixfold reduced in Bcrp1^−/− milk, but FAD levels were substantial and unchanged in Bcrp1^−/− milk (Fig. 5A). Assuming complete conversion of FMN and FAD to riboflavin, Bcrp1^−/− milk still contained 26% of riboflavin equivalents compared to wild-type milk, which is apparently more than enough to salvage riboflavin delivery to the pups. Indeed, in 7- to 12-day-old milk-fed pups (Bcrp1^+/−) nurtured by wild-type or Bcrp1^−/− dams, the riboflavin derivative FAD was present in comparable amounts in plasma (Fig. 5C). This suggests that there is sufficient riboflavin equivalent availability in pups drinking milk from Bcrp1^−/− dams on standard chow.

When the transfer of total vitamin B₂ equivalents (riboflavin, FAD, and FMN) was limited in milk of wild-type and Bcrp1^−/− lactating mothers by replacing their standard chow with a diet without riboflavin supplementation, all pups suffered from vitamin B₂ deficiency (lack of growth, fatty degeneration of liver, and severe anemia) (not shown). Intraperitoneal supplementation of riboflavin to the mothers salvaged these deficiencies in the suckling pups, however, without any obvious difference between pups with wild-type or Bcrp1^−/− mothers. This is presumably due to the conversion to and subsequent secretion of FAD into milk of the wild-type and Bcrp1^−/− mothers (not shown). Clearly, there are at least two independent pathways involved in secretion of vitamin B₂ equivalents (riboflavin, FMN, and FAD) into milk. We note that the secretion of FAD into milk is likely also active, as the milk-to-plasma ratio is 16 ± 5.2, indicating a concentrative process.

Bcrp1 deficiency reduces biotin (vitamin H) levels in milk.

As Bcrp1 is a broad-specificity transporter, riboflavin might not be the only vitamin transported into milk by Bcrp1. Indeed, when we tested a range of other vitamins in pooled wild-type versus Bcrp1^−/− milk samples, we observed a threefold-lower level of biotin (vitamin H) in Bcrp1^−/− milk. None of the other vitamins tested was substantially lower in Bcrp1^−/− milk (Table 2). Analysis of separate milk samples (n = 3) revealed a 3.5-fold-lower concentration of biotin in milk of Bcrp1^−/− compared to wild-type mice (47 ± 24 versus 163 ± 32 μg/100 ml; P < 0.01). Additionally, when we administered [¹⁴C]biotin (690 nmol/kg) i.v. to lactating wild-type and Bcrp1^−/− mice and collected blood and milk at 30 min after administration, we observed a 2.3-fold-lower concentration of [¹⁴C]biotin in milk from Bcrp1^−/− versus wild-type mice (322 ± 73 versus 779 ± 246 nM; P < 0.05). When corrected for plasma [¹⁴C]biotin concentrations, which were 1.2-fold higher in Bcrp1^−/− than in wild-type mice, we observed a 2.6-fold difference in the milk-to-plasma ratio of [¹⁴C]biotin (2.2 ± 1.0 versus 5.8 ± 1.8; P < 0.05). This indicates that in addition to reducing riboflavin levels, Bcrp1 deficiency also reduces biotin levels in milk, albeit not as profoundly.

DISCUSSION

This study identifies a molecular mechanism responsible for concentrating a vitamin into milk. We found that mouse Bcrp1 and human BCRP transport riboflavin and that Bcrp1 reduces the tissue distribution and plasma levels of this vitamin in mice. It also mediates excretion of riboflavin into the intestinal lumen. We then demonstrated that Bcrp1 mediates active riboflavin secretion into milk. Whereas riboflavin transport into milk is solely dependent upon the expression of Bcrp1 in the lactating mammary gland, its cofactor FAD can enter the milk independently of Bcrp1. Our findings illustrate that there are two independent pathways involved in secretion of vitamin B₂ equivalents (riboflavin, FMN, and FAD) into milk. This partial redundancy explains why no obvious vitamin B₂ deficiency was observed in Bcrp1^−/− pups.

BCRP has broad substrate specificity and actively extrudes a wide variety of drugs, carcinogens, and dietary toxins from cells. Due to its expression in intestine, liver, placenta, and blood-brain barrier, it has an important role in protection of the organism from these noxious compounds (5, 19). Surprisingly, rather than evolving a dedicated transporter, the mammary gland has recruited this broad-specificity transporter to pump riboflavin into milk. The tissue distribution of BCRP/Bcrp1 combined with the dual utilization of BCRP/Bcrp1 as a vitamin and a xenotoxin transporter potentially compromises essential biological functions of the respective organs in which it is expressed, i.e., in the breast, production of safe, xenotoxin-free milk and in the intestine, liver, placenta, brain, etc., efficient systemic and tissue uptake of dietary riboflavin. In spite of this paradox, the dual utilization of BCRP/Bcrp1 is evolutionarily conserved, at least between humans and mice, but presumably also in cows and other mammals (13). Riboflavin must be provided in the diet of mammals or synthesized by bacteria present in the intestinal lumen. In either case it must be absorbed from the intestine. We demonstrated that Bcrp1 can reduce the systemic and tissue levels of riboflavin in mice, but the overall effects were modest. Apparently, the efficient riboflavin uptake systems present in the intestine and in other blood-tissue barriers and cells, combined with subsequent metabolic trapping of riboflavin by conversion to FAD, are sufficiently effective to offset the riboflavin extrusion capacity of BCRP/Bcrp1 and thus avoid overall vitamin B₂ deficiency.

Vitamin B₂ is essential for many key enzymatic reactions in the body. However, despite negligible amounts of riboflavin in Bcrp1^−/− milk, pups nurtured by Bcrp1^−/− dams did not show any signs of malnourishment or growth retardation (13, 19; this study), contradicting a severe vitamin B₂ shortage. Indeed, although pup plasma riboflavin levels were 1.6-fold decreased in Bcrp1^−/− versus wild-type milk (Fig. 5), plasma levels of the cofactor FAD did not differ. This indicates that under the favorable conditions of our animal facility, with an ad libitum supply of food containing ample riboflavin for nursing dams, pups of Bcrp1^−/− dams still received sufficient vitamin B₂ equivalents. Clearly, in addition to the transport of riboflavin by BCRP, there is at least one other mechanism to deliver vitamin B₂ equivalents to the young, in the form of FAD. We do not know the nature of the mechanism(s) responsible for FAD secretion into milk. However, we hypothesize that this redundancy may be important to ensure vitamin B₂ supply to pups under different environmental circumstances, as there may be conditions where either FAD or riboflavin secretion into milk is compromised. We showed earlier that a BCRP inhibitor could nearly abrogate Bcrp1-mediated secretion into milk (13), and other conditions might affect the FAD secretion process. Under more natural conditions than in a mouse facility, an erratic food supply of variable content for nursing dams will be common. Moreover, (intestinal) infections and other stresses for the pups may result in reduced dietary intake or intestinal uptake capacity for riboflavin or in increased physiological riboflavin demand. Under such circumstances, the ability to secrete large amounts of vitamin B₂ equivalents into milk and the possibility of utilizing several independent secretion mechanisms may be essential for survival of the pups.

Suboptimal BCRP activity may also be caused by genetic polymorphisms. In humans, BCRP activity can vary extensively due to known genetic polymorphisms (3, 11, 21) or due to intentional or coincidental use of drugs that inhibit BCRP (5). This might compromise riboflavin secretion into milk and increase dependence on the alternative FAD secretion pathway. It will be of interest to test whether reduced BCRP activity in humans also diminishes riboflavin secretion into human milk. In Holstein cattle, a correlation was recently found between a polymorphism in BCRP (Y581S) and modestly altered fat and protein contents and quantity of milk production (6). Whether the activity or proper membrane localization of BCRP is affected by this amino acid substitution remains to be determined, but it will be interesting to test whether this polymorphism affects riboflavin transfer into milk.

We considered the possibility that riboflavin secretion into milk might not be a primary function of BCRP in the mammary gland, as there was no immediate deficiency seen in the pups. However, if secretion of other vitamins or nutrients (in addition to the ones we tested) was the primary function of BCRP in the mammary gland, the question still arises of why there is no obvious deficiency phenotype in pups. An entirely different hypothesis proposes that xenotoxin secretion by BCRP into milk might be biologically useful. By supplying milk with modest amounts of xenotoxins to which the mother is exposed, BCRP could help start the induction of appropriate xenotoxin detoxification proteins in the pups during lactation. As a consequence, by the time that the pups have to switch from milk to solid food (likely identical to the food ingested by the mother), they are already prepared for exposure to the xenotoxins present in this food. However, while we consider this a hypothesis worth testing in the future, it still would not explain why riboflavin transport is evolutionarily conserved in BCRP. If xenotoxin transport was the only relevant function of BCRP, mice (and humans) would be better off with a BCRP that does not transport riboflavin, thus avoiding the complications of reduced systemic riboflavin availability. Of course, it is possible that both functions of BCRP in the mammary gland are biologically important, which would resolve most of the paradox in our findings.

The transport of riboflavin by BCRP/Bcrp1 can explain earlier findings. For instance, the apically directed transport of riboflavin observed in Caco-2 cells and rat ileum (9) could be explained by the expression of Bcrp1 at the apical membrane in small and large intestinal epithelial cells; we demonstrated a more-than-sixfold reduction in excretion of riboflavin into the intestinal lumen due to the absence of Bcrp1 in mice. Yanagawa et al. (20) observed that riboflavin can be secreted by proximal tubules in the kidney, and this might be explained by the apical expression of Bcrp1 in this organ (12). Similarly, the observed competition for transport in mouse leukemia cells between the anticancer drug methotrexate and riboflavin, both Bcrp1 substrates, could be due to Bcrp1 expression in these cells (14).

We note that efficient riboflavin extrusion by BCRP and Bcrp1 might cause problems during cell culture of highly BCRP-expressing cells, since cells need riboflavin for survival. Analogous to the situation observed earlier with the BCRP substrate folate (10), high BCRP expression might result in the need for higher vitamin (folate or riboflavin) concentrations in the medium for optimal cell growth.

While this study establishes the principle that an ABC transporter can actively concentrate a vitamin into milk, it gives rise to many questions for future studies. For instance, it will be of interest to investigate what mechanism is responsible for the active secretion of FAD into milk and to establish what the respective roles of the free riboflavin and FAD secretion processes are in supplying the infant with vitamin B₂ equivalents. It will also be interesting to establish whether other vitamins and nutrients are also transported into milk by ABC transporters.