Absorption, metabolism and excretion of flavanones from single portions of orange fruit and juice and effects of anthropometric variables and contraceptive pill use on flavanone excretion

Chemicals and reagents

HPLC-grade methanol was purchased from Fisher Scientific (Loughborough, UK). Acetonitrile (HPLC grade), trifluoroacetic acid, β-glucuronidase (Helix pomatia type H5), sulfatase (H. promatia type H1), N,O-bis-(trimethylsilyl) trifluoracetamide, rhamnetin, naringenin, hesperetin, perillic acid, ethylbenzoic and propylbenzoic acids, cobalt(II) bromide and 4,7-diphenyl-1,10-phenanthroline (4,7-dpphen) were purchased from Sigma-Aldrich (Poole, Dorset, UK). Pelargonidin-3-glucoside and galangin were obtained from Extrasynthèse (Genay, France).

Sampling of orange fruits and juices

Orange fruits and juices were purchased locally from a variety of outlets including major supermarkets and smaller retailers. The fruits were of the Maroc, Shamouti, Navel, Navelina, Salustiana, Moro and Lane Late varieties. Juices were a mixture of branded, economy and supermarket own-branded products. For the interventions, fresh oranges (Delta seedless variety, South Africa) were obtained in bulk from a local supermarket and stored at 4°C until consumption. All volunteers (with the exception of three) consumed oranges from the same batch. Fresh oranges were prepared by mixing together the segments (eight per fruit) of three whole oranges and portioning a 150 g representative sample for consumption and another 100 g sample for analysis. The orange juice was a commercially available supermarket own-branded product made from concentrate.

Subjects and study design

Twenty apparently healthy volunteers (ten men and ten women) aged between 20 and 65 years were recruited to participate in this study. All study participants were assessed for eligibility on the basis of a health questionnaire and the results of clinical laboratory tests. The following exclusion criteria applied: smokers; long-term medical conditions such as asthma (unless untreated within the past 2 years), heart disease, gastrointestinal disease, diabetes, cancer; regular prescribed medication (except hormone replacement therapy and oral contraceptives); supplement (unless judged not to affect study outcome) or antibiotic use within 4 weeks before the start of the study; pregnancy; blood donation within 4 months before the start of the study; BMI <18·5 or >35 kg/m²; clinical results at screening judged by the medical advisor to affect study outcome or be indicative of a health problem. Subject characteristics were: weight 73·9 (sd 14·1) kg (range 50·3–101 kg), BMI 24·8 (sd 3·0) kg/m² (range 20·7–32·2 kg/m²); age 49 (sd 11) years (range 26–64 years). The study was explained to participants and written informed consent was obtained before participation. The study protocol was approved by the Human Research Governance Committee of the Institute of Food Research and the Norwich Research Ethics Committee.

The study was a randomised two-phase cross-over design investigating the bioavailability of flavanones from fresh and processed oranges. Each test phase comprised a 5 d period of intervention separated by a washout period of at least 1 week. During each period of intervention, subjects followed a low-polyphenol diet and to aid compliance a list of authorised and prohibited foods were given. On day 3 of the intervention, fasted subjects had an intravenous catheter inserted and a baseline blood sample (10 ml) was obtained. Subjects were given a standard breakfast consisting of two slices of white toast (72 g) with spread (10 g) followed by either 150 g fresh orange segments or 300 g orange juice. To limit variation in food and drink intakes, subjects refrained from drinking and eating for 1·5 and 4 h respectively. Blood samples (10 ml) were collected into lithium heparin tubes at 15, 30 and 45 min and 1, 1·5, 2, 3, 4, 6, 8, 24 and 48 h and immediately centrifuged at 1500 g for 10 min. Plasma samples were subsequently mixed with ascorbate (final concentration 1 mm) to prevent flavanone degradation and frozen at −20°C until analysis. Urine was collected in 24 h fractions the day before consumption of orange flavanones and for 48 h after consumption into collection bottles containing 1 g ascorbic acid. The amount of urine in each fraction was measured, and samples were stored at −20°C until analysis.

Quantification of flavanone glycosides in orange fruits and juices

Orange fruits and juices were analysed using HPLC. All solvents used were HPLC grade, all water was ultra-pure and the purity of commercial standards was confirmed by HPLC. All analyses were performed at least in triplicate. Fresh oranges were weighed and samples of fruit purée (30 g) were mixed with 70 ml of 70 % aqueous methanol and blended to homogeneity. A sample (1 ml) was removed and mixed with 50 μl rhamnetin in methanol (0·1 mg/ml; internal standard). Juices were prepared by adding 700 μl methanol and 50 μl rhamnetin (0·1 mg/ml) to 300 μl of juice. Both fruit and juice samples were heated to 70°C for 20 min with occasional vortex mixing, centrifuged (13 000 rpm × 10 min; 4°C) and filtered before HPLC analysis.

Samples were analysed by HPLC with online UV-diode array and LC-MS detectors (see later). Absorbance at 270 nm was used for quantification, and tandem MS (positive-ion mode) to confirm the identity of the analytes. Recoveries of internal standards for the food and beverage analyses were 93·1 (sd 4·8) and 95·4 (sd 4·4) % for fruits and juices, respectively. Intra-day variance values for narirutin, hesperidin and didymin in fruits and juices were <10 %.

Extraction and quantification of flavanones and metabolites in plasma and urine

Acidified plasma samples (2·0 ml) were hydrolysed enzymically by incubating with 2·0 ml phosphate buffer (pH 5·0), 0·2 ml β-glucuronidase (2000 units) and 0·2 ml sulfatase (200 units) at 37°C for 2 h. Acidified urine samples (10·0 ml) were incubated with 5·0 ml phosphate buffer (pH 5·0), 1·0 ml β-glucuronidase (10 000 units) and 1·0 ml sulfatase (1000 units) at 37°C for 2 h. Then 50 μl galangin (0·1 mg/ml) was added as an internal standard before incubation. Flavanones in hydrolysed plasma and urine samples were extracted using a solid-phase extraction (SPE) cartridge (Varian Bond Elute C₁₈) conditioned with methanol (5 ml) followed by water (10 ml). Following application of urine or plasma, the cartridge was washed with water (10 ml) and flavanones eluted directly into vials with 1 % HCl in methanol (1·0 ml for urine) or 1 % HCl in acetonitrile (0·5 ml for plasma). Pelagonidin-3-glucoside (50 μl of 0·1 mg/ml) was added to the SPE eluate as a volume marker immediately before HPLC analysis.

Samples (1 μl) of hydrolysed plasma and urine extracts were analysed by HPLC (Agilent HP1100; Agilent Technologies, Waldbronn, Germany) using a Gemini C₁₈ column (150 × 2·00 mm, 5 μm particle size; Phenomonex, Macclesfield, Cheshire, UK) eluted at 0·3 ml/min with a gradient of increasing solvent B (0·1 % trifluoroacetic acid in acetonitrile) from solvent A (0·1 % aqueous trifluoroacetic acid) at 30°C over 36 min (plasma) or 65 min (urine). The eluent was scanned over 200–600 nm by a diode array detector and subsequently an ESI-MS (Agilent Technologies, Waldbronn, Germany). The mass spectrometer was operated in negative ionisation mode (cone voltage 22 V, source block temperature 120°C, desolvation temperature 300°C) with multiple reaction monitoring. Quantification was based on peak areas at 270 nm. Quantification of orange flavanones in plasma and urine samples was based on standard curves (range 0·1–100 μg/ml) for naringenin and hesperetin. Standard curves were linear with regression coefficients >0·99 and recoveries of the internal standard were 91·4 (sd 9·38) %. The precision of the urine and plasma assays was assessed using replicate (n 4) analyses of spiked blank samples: precision was excellent with mean values of 97·8 (sd 2·0) and 99·8 (sd 2·5) % of expected analyte recovered for hesperetin and naringenin, respectively. Assay repeatability was assessed by repeated analyses (four different days) of replicate (n 5) spiked samples; the inter-day variance was <5% (sem as percentage of the mean). Recoveries of internal standards for the plasma analyses were 91 (sd 10·5) %.

Identification of flavanone metabolites in urine after orange consumption

A sample of urine from a typical volunteer was analysed as above, except directly without any enzyme hydrolysis. An authentic sample of synthetic naringenin 7-glucuronide⁽²⁹⁾ was run in comparison to allow its identification in the urine sample. Samples treated separately with glucuronidase or sulfatase were compared with untreated equivalents to assist identification of glucuronidated and sulfated metabolites. A second analysis of the urine samples was performed independently and blindly at another facility (see below). Liquid chromatography-tandem MS (LC-MSⁿ) was the only identification technique employed for this parallel analysis. Negative-ion-mode LC-MSⁿ provided the mass of each metabolite, the number of sulfate and/or glucuronic acid moieties attached, and the identity of the flavonoid aglycone. The locations of the glucuronic acid moieties were determined by metal complexation with the Co²⁺ ion and an auxiliary ligand, 4,7-dpphen. The fragmentation patterns of the [Co(II) (flavonoid glucuronide–H) (4,7-dpphen)₂]⁺ complexes are highly indicative of the position of the glucuronic acid moieties, as demonstrated in a recent study⁽²⁹⁾. Similar types of metal complexes have been used recently to identify flavonoid glycosides in various plant extracts^(30,31). Methanol (7·5 ml) was added to 2·5 ml samples of urine for protein precipitation. The mixture was vortexed and centrifuged for 15 min at 1380 g. The supernatant fraction was collected and evaporated with N₂ down to approximately 500 μl. Injection volumes were 20–40 μl. Chromatography was performed on an Alliance 2695 HPLC system (Waters, Milford, MA, USA). The column was a Waters Symmetry C18 (2·1 × 50 mm, 3·5 μm particle size). The mobile phases were water containing 0·05 % formic acid (A) and methanol containing 0·05 % formic acid (B). The gradient began at 25 % B, increased to 30 % B over 10 min, then increased to 50 % B over 10 min, then to 100 % B over 5 min, and was held constant at 100 % B for an additional 2 min. The flow rate was 0·3 ml/min throughout. The column effluent was directed without splitting into an LCQ Duo quadrupole ion trap mass spectrometer (ThermoFinnigan, San Jose, CA, USA) equipped with an ESI source. For negative-ion-mode experiments, the spray voltage was −4·5 kV. For metal complexation experiments, a methanolic solution of 5 μm-CoBr₂ and 4,7-dpphen was added post-column via a T joint and syringe pump flowing at 20 μl/min. The positive-ion mode was employed for these experiments, using a spray voltage of +4·5 kV. Collision-induced dissociation (CID) with the He bath gas was used to fragment ions of interest for structural characterisation.

Data analysis

Statistical analyses were performed using the R data analysis software (R Foundation for Statistical Computing, Vienna, Austria; http://www.r-project.org). In the comparison of data from the fruit and juice cross-over study, repeated-measures models were used to analyse the data and ‘volunteer’ was included as a ‘random effect’. Regression diagnostics were checked to determine if data transformations, outlier omissions, or alternative non-parametric models were required. All results from the models were considered significant if P<0·05. When the metabolite data were split into their two separate components (naringenin and hesperetin) and analysed for differences in fruit v. juice, a Wilcoxon rank sum test was used. Again, results were considered significant if P<0·05. For the single portion juice intervention, the urinary excretion data were analysed using standard ANOVA models. For all models, regression diagnostics were checked to determine if data transformations, outlier omissions, or alternative non-parametric models were required. All results from the models were considered significant if P<0·05.

Comparison of flavanone absorption and excretion from orange fruit and juice

In a randomised cross-over study, twenty volunteers ingested orange juice and orange fruit segments and the appearance of flavanones in plasma and urine was measured. The oranges used in the interventions contained 11·8 (sd 5·5) mg naringenin (as narirutin) and 79·7 (sd 17·7) mg hesperetin (as hesperidin)/150 g portion. The orange juice used in the interventions contained 9·4 (sd 0·7) mg naringenin (as narirutin) and 71·8 (sd 8·1) mg hesperetin (as hesperidin)/300 g portion. The fruits and juices also contained small amounts of didymin (data not shown). The stereochemical nature of the flavanones was not determined directly, but it is usual for citrus flavanones to be present as the 2S conformer.

Plasma and urine samples were analysed for total flavanone aglycones (naringenin and hesperetin) following hydrolysis with aryl sulfatase and β-glucuronidase. Flavanones were absent from all baseline samples, indicating good compliance with the flavonoid avoidance diet. Low concentrations of flavanones were detected in plasma within 15 min of ingestion of the fruit or juice. The pharmacokinetic profile for total flavanones in plasma was similar for both naringenin and hesperetin; a small peak at 1–1·5 h followed by the major peak at about 6 h with a subsequent decline to near baseline over the next 48 h (Fig. 2). The absolute mean peak plasma concentration (C_max) and area under the plasma concentration time curve (AUC) for hesperetin were significantly higher than that for naringenin for both the fruit and juice matrices (Table 1). However, since the hesperetin oral dose was substantially lower (about 7-fold) than the naringenin dose, naringenin was more efficiently absorbed to the plasma than hesperetin on a dose-adjusted basis. The mean time to reach C_max was greater for hesperetin compared with naringenin but the effect was not significant. Urinary excretion (% of oral dose) of naringenin was greater than for hesperetin for both the juice and fruit matrices, which reflects the more efficient absorption of naringenin to plasma. There were no significant differences between fruit and juice in the urinary yields (% of dose excreted) for hesperetin or naringenin. For all the parameters described (C_max, time to peak concentration, AUC, urinary yield), the individual variation was substantial and the variance was high (see Table 1). The stereochemical nature of the flavanones in plasma and urine was not determined.

The observed individual variation in absorption and excretion parameters detailed in Table 1 was investigated further. First, the association between parameters was compared as individual paired dose-adjusted data (n 20). The correlation (Spearman’s) between AUC and C_max was strong for naringenin from juice (R ² 0·88) and fruit (R ² 0·74), and hesperetin from juice (R ² 0·73) and fruit (R ² 0·92) (all P values <0·01). The correlations between percentage urinary excretion and AUC or C_max were weaker for both hesperetin and naringenin for both matrices (R ² values 0·00–0·45 (P>0·05) in all cases except for naringenin from fruit).

Urinary excretion of orange flavanones in an extended study investigating the relationship with various subject characteristics

The urinary excretion of flavanones from orange juice was extended to include a further 109 volunteers aged between 18 and 80 years to produce a large dataset suitable for evaluating the effects of volunteer anthropometric and other variables. For the extended study population, the excretion of naringenin and hesperetin was 14·5 (sd 11·9) and 3·9 (sd 3·6) % (n 129), respectively. The range for naringenin was 1·6–59·0 % (37-fold), with a median of 9·9 %. The range for hesperetin was between not detected (three volunteers) to 25·4 %, with a median of 3·1 %. The naringenin:hesperetin ratio was 4·4 (sd 2·8) (range 0·7–13, median 3·7).

The urinary excretion data were fitted to a standard linear model (ANOVA) with log (percentage excretion) as the response and several subject variables (age, BMI, sex, contraceptive pill use amongst the female volunteers) as the explanatory variables. After systematically removing non-significant terms and re-fitting the data, a possible effect of age on excretion of hesperetin (P=0·033) and naringenin (P=0·058) was detected. For naringenin, the apparent effect of increasing age was to decrease excretion in males but increase excretion in females. However, separation of the data into males and females did not indicate a significant effect of age. For hesperetin, the effect of increasing age was to decrease excretion, although the magnitude of the effect was very small – equivalent to a 4 % decrease in urinary excretion between the ages of 20 and 80 years (Fig. 3). There were no significant effects of sex, BMI or contraceptive pill use on the excretion of either flavanone (Table 1).

Structural identification of flavanone metabolites in plasma and urine

The urine of a typical volunteer, analysed by reverse-phase chromatography, and monitored at 270 nm, is shown in Fig. 4. The sample was simultaneously analysed by full scan positive- and negative-ion ESI-MS. Data for labelled peaks, which were absent from a baseline chromatogram, are summarised in Table 2. Each displayed [M + H]⁺ and [M − H]⁻ ions consistent with a glucuronidated or sulfated flavanone, and each was removed by a combined glucuronidase and sulfatase treatment. Hence the labelled components were tentatively identified as follows: A and B, naringenin monoglucuronides; C and D, hesperetin monoglucuronides; E, a hesperetin monosulfate. Component A was confirmed as naringenin 7-glucuronide by comparison with an authentic synthetic standard⁽²⁹⁾. The UV spectra of components B to E were then examined; all were consistent with a flavanone structure. Naringenin-7-glucuronide (A) and C exhibited very similar spectra (Table 1 and Fig. 2), strongly suggesting that C was also a 7-substituted flavanone; from this, C was provisionally identified as hesperetin 7-glucuronide. D and E exhibited very similar spectra, suggesting that they were both substituted in the same position. (We have observed that the UV spectra of, for example, quercetin glucuronides and sulfates with the same glucuronidation pattern display almost identical UV profiles (PW Needs and PA Kroon, unpublished results).) B displayed its own distinctive spectrum. 5-Substituted flavonoid metabolites have not been reported; and if 5-substution is discounted, B must be naringenin 4′-glucuronide; and hence D and E are hesperetin 3′-glucuronide and hesperetin 3′-sulfate respectively.

These assignments were independently confirmed using LC-MSⁿ techniques as well as recently discovered metal complexation techniques⁽²⁹⁾. Negative-ion-mode analysis of components A (retention time (RT) = 11·7 min) and B (RT = 12·9 min) revealed the masses of both of these compounds to be 448 Da. In each case, CID led to the loss of 176 Da, corresponding to a glucuronic acid moiety and giving the product of m/z 271. The second-generation fragment ions stemming from m/z 271 are shown in Figs. 5 (a) and (b), and match the fragmentation pattern of deprotonated naringenin (Fig. 5 (c)). Post-column metal complexation was then performed as described in the Materials and methods section in order to provide metal complexes of the target compounds that are useful for pinpointing the specific sites of glucuronidation⁽²⁹⁾. The fragmentation of the metal complexes of A and B are shown in Figs. 5 (d) and (e). The former shows a fragmentation pattern highly indicative of glucuronidation at the 7 position based on characteristic losses such as those of the aglycone group and of the glucuronic acid moiety. However, the latter pattern is consistent with B-ring glucuronidation, as the only fragment ions observed stem from the loss of an auxiliary ligand, with or without the glucuronic acid moiety. The only possible site for B-ring glucuronidation on naringenin is the 4′ hydroxyl group. Hence A was identified as naringenin 7-glucuronide and B as naringenin 4′-glucuronide. Similar evidence was used to identify unknowns C (RT = 14·9 min) and D (RT = 17·8 min). Negative-mode ESI-LC-MS revealed two compounds of mass 478 Da (based on the formation of deprotonated molecular ions at m/z 477) that fragmented to yield a single neutral loss of 176 Da, corresponding to the glucuronic acid moiety. The second-generation fragment ions from subsequent MSⁿ experiments are indicative of hesperetin (Figs. 6 (a)–(b)), as revealed based on comparison with the fragmentation pattern of deprotonated hesperetin (Fig. 6 (c)). Fragmentation of the metal complex of C indicates a glucuronic acid moiety at the 7 position (Fig. 6 (d)), whereas the metal complex of D indicates B-ring glucuronidation (Fig. 6 (e)), i.e. at the 3′ position of hesperetin. Thus C was positively identified as hesperetin 7-glucuronide and D as hesperetin 3′-glucuronide. E was partially identified as a hesperetin sulfate based on its mass (382 Da), the loss of 80 Da upon fragmentation (corresponding to SO₃), and second-generation fragment ions that mirror those of hesperetin (data not shown). To our knowledge, there is no purely MS-based method for determining the location of sulfate groups on flavonoids.

In addition to the above compounds, which were also discovered in the previously described LC-UV-MS analysis, the LC-MSⁿ analysis additionally revealed several di-substituted metabolites in the urine samples. A hesperetin glucuronide sulfate (RT = 16·6 min) was identified based its mass (558 Da), neutral losses of 80 Da and 176 Da upon CID, and an aglycone fragmentation signature matching that of hesperetin. Surprisingly, metal complexation was able to reveal the location of the glucuronic acid moiety of this compound. The metal complexation technique was developed originally for the characterisation of flavonoid monoglucuronides, but the hesperetin glucuronide sulfate also formed a metal complex of the correct stoichiometry, observed at m/z 1280 in the positive-mode ESI mass spectrum. The first stage of CID led to the loss of the sulfate group (−80 Da), effectively resulting in a metal complex of a flavonoid monoglucuronide, which can be characterised based on its fragmentation pattern as described for the other compounds above. The second stage of fragmentation, using m/z 1200 as the precursor ion, yielded a CID mass spectrum similar to that shown in Fig. 6 (d). This strongly suggests that the glucuronic acid moiety of this hesperetin glucuronide sulfate is located at the 7 position. A similar strategy has been used previously to partially identify diglycosyl flavonoids while employing metal complexes designed to characterise monoglycosyl flavonoids⁽³⁰⁾. Finally, a few diglucuronidated flavonoids were found in the urine samples. Two early-eluting components (RT = 4·8 and 9·9 min) were identified as hesperetin diglucuronides. These analytes both had a mass of 654 Da, exhibited two sequential losses of glucuronic acid (−176 Da), and yielded fragmentation patterns of the aglycone group that corresponded to hesperetin. In addition, there was at least one other low-intensity compound resembling a naringenin diglucuronide that appeared in only a few urine samples. This analyte had a mass of 624 Da, had an RT of 1·9 min, and lost two glucuronic acid moieties upon sequential stages of CID. However, the low abundance of this analyte combined with signal losses due to scattering during each stage of fragmentation led to an MS⁴ spectrum with a poor signal:noise ratio. While there is evidence of naringenin aglycone, the spectrum is too noisy to make a conclusive identification. Metal complexes could not be formed for any of the purported naringenin or hesperetin diglucuronides, possibly due to low abundance of these compounds.