Research ArticleAging

Sulfate Metabolites Provide an Intracellular Pool for Resveratrol Generation and Induce Autophagy with Senescence

See allHide authors and affiliations

Science Translational Medicine  02 Oct 2013:
Vol. 5, Issue 205, pp. 205ra133
DOI: 10.1126/scitranslmed.3005870


The phytochemical resveratrol has been shown to exert numerous health benefits in preclinical studies, but its rapid metabolism and resulting poor bioavailability may limit translation of these effects to humans. Resveratrol metabolites might contribute to in vivo activity through regeneration of the parent compound. We present quantitation of sulfate and glucuronide conjugates of resveratrol in human plasma and tissue after repeated ingestion of resveratrol by volunteers and cancer patients, respectively. Subsequent pharmacokinetic characterization of a mixture of resveratrol-3-O-sulfate and resveratrol-4′-O-sulfate in mice showed that these metabolites are absorbed orally but have low bioavailabilities of ~14 and 3%, respectively. Sulfate hydrolysis in vivo liberated free resveratrol, which accounted for ~2% of the total resveratrol species present in mouse plasma. Monosulfate metabolites were also converted to the parent in human colorectal cells. The extent of cellular uptake was dependent on specific membrane transporters and dictated antiproliferative activity. Sulfate metabolites induced autophagy and senescence in human cancer cells; these effects were abrogated by inclusion of a sulfatase inhibitor, which reduced intracellular resveratrol. Together, our findings suggest that resveratrol is delivered to target tissues in a stable sulfate-conjugated form and that the parent compound is gradually regenerated in selected cells and may give rise to the beneficial effects in vivo. At doses considered to be safe in humans, resveratrol generated via this route may be of greater importance than the unmetabolized form.


Preclinical evidence in model systems suggests that the phytochemical resveratrol (trans-3,5,4′-trihydroxystilbene) has cancer-preventive properties (1), beneficial effects on cardiovascular (2) and neurodegenerative diseases (3), promotes longevity in lower organisms (4), and delays or attenuates many age-related changes and early mortality that result from obesity in mice (5, 6). A wealth of mechanistic data supports the role of resveratrol in the management of these conditions by virtue of its antioxidant, anti-inflammatory, antitumorigenic, and calorie restriction–mimetic properties (7). However, a limitation in translating these observations to efficacy in humans stems from resveratrol’s poor bioavailability, which results from rapid and extensive phase 2 metabolism, and toxicity concerns, which prohibit simply increasing the dose to overcome the metabolism issue (8, 9). Repeated ingestion of resveratrol at doses exceeding 1.0 g/day is associated with gastrointestinal side effects, which, although mild, would certainly prohibit the long-term use of such doses in high-risk or healthy populations (10, 11). Consumption of 1.0 g of resveratrol affords maximal plasma concentrations of ~0.6 μM in humans (10), but most of the reported in vitro studies, particularly those relating to cancer, necessitate concentrations above this for detectable activity. This raises the question of whether sufficient levels of resveratrol can be safely attained in humans or whether resveratrol metabolites might contribute to the beneficial effects associated with the parent compound and negate this concern. It has long been speculated, but not proven, that resveratrol conjugates may undergo hydrolysis in vivo to regenerate the parent compound or may themselves elicit biological changes (12).

To date, only a limited number of in vitro studies and no in vivo ones have attempted to address the potential effects of resveratrol metabolites. Conjugated derivatives of dietary polyphenols often have reduced antiproliferative activity compared to the parent compound, and this seems to be the case for resveratrol (13, 14). However, individual monosulfate metabolites have been found to have comparable or greater potency than resveratrol against specific molecular targets, namely, cyclooxygenase (COX), quinone reductase 1 (NQO1), and nuclear factor κB (NFκB), as well as a similar ability to scavenge free radicals (15, 16).

Here, we present the quantitation of resveratrol conjugates in human plasma and colorectal tissue after repeated ingestion of resveratrol, which defined the concentration range suitable for further preclinical mechanistic investigations in mice and in human cancer cells. The data demonstrate that resveratrol sulfates may contribute to in vivo efficacy by delivering resveratrol to target tissues in a stable conjugated form, thus enabling gradual regeneration of the active parent compound in selected cells.


Concentrations of resveratrol conjugates greatly exceed previous estimations

Acquisition of detailed human pharmacokinetic information is essential for the rational development of all pharmaceuticals and dietary agents, and may be especially valuable for resveratrol, with its numerous potential indications. Optimal concentrations of the active species required for efficacy may vary considerably, depending on the disease being treated or prevented. Therefore, all resveratrol-derived species with biological activity must be identified and accurately measured in plasma and other tissues.

Because of the lack of available authentic standards, clinical pharmacokinetic investigations, including our own, report estimated concentrations of the major resveratrol conjugates generated in vivo (10, 17). Such approximations are based on a resveratrol standard curve, which is affected by the extraction and spectroscopic characteristics of resveratrol itself. We have now synthesized sufficient quantities of resveratrol sulfate and glucuronide standards to enable reanalysis, with our validated high-performance liquid chromatography with ultraviolet detection (HPLC-UV) assay, of a representative set of plasma and colorectal samples from our recent clinical trials that involved repeated administration of resveratrol capsules to volunteers (10) and cancer patients (17) (Fig. 1). Standard curves used for the analysis were reproducible and displayed R2 values of ≥0.99, which indicated linearity over the concentration range measured. The limits of quantitation were 10, 12, and 8 ng/ml for resveratrol-4′-O-glucuronide, resveratrol-4′-O-sulfate, and resveratrol-3-O-sulfate, respectively.

Fig. 1. Resveratrol and its metabolites can be accurately measured in human plasma and colorectal tissue.

(A) Representative HPLC-UV chromatograms of (i) plasma taken 1 hour after the last dose of resveratrol from a healthy volunteer who received resveratrol (1.0 g) daily for 29 days, (ii) colorectal cancer tissue from a patient who ingested 1.0 g of resveratrol daily for 8 days before surgery, and (iii) authentic resveratrol and metabolite standards. Peaks were assigned by comparison of retention times with synthetic standards where available, and the identity was confirmed by LC-MS/MS in negative ionization mode. (B) Representative analysis of plasma extracts by LC-MS/MS with multiple reaction monitoring (MRM) for the mass/charge ratio (m/z) transitions designated. Plasma was taken from a healthy volunteer 1.5 hours after dosing with 5 g of resveratrol on a day during the last week of a 29-day intervention with 5 g of resveratrol administered daily. Where multiple peaks are present for a transition, asterisks indicate the peak of interest. Metabolites not previously identified in human plasma are bolded and underlined. (C) Concentrations of resveratrol and its metabolites in human plasma estimated using resveratrol to generate the standard curve (resveratrol equivalents, left) or accurately measured using synthetic metabolite standards (right). Healthy volunteers received resveratrol (0.5, 1.0, 2.5, or 5.0 g) daily for 29 days (10). Concentrations were determined by HPLC-UV analysis of plasma taken on a day during the last week of intervention. Values are means + SD of maximum plasma concentrations (Cmax) for three randomly selected individuals per dose group.

The use of metabolite standard curves revealed that the average maximum plasma concentrations (Cmax) in humans for monoglucuronide, 4′-O-sulfate, and 3-O-sulfate derivatives of resveratrol were actually ~2.6-, 3.8-, and 2.9-fold higher, respectively, than previously described (Fig. 1C and table S1A). This means that repeated oral dosing with 1 g of resveratrol daily can yield plasma concentrations of the major 3-O-sulfate of ~22 μM (range, 8 to 32 μM), whereas the monoglucuronides typically reach ~7 to 8 μM (range, 2 to 18 μM) (table S1B). Reanalysis of 3-O-sulfate concentrations in colorectal tissue also indicated significant previous underestimation, although not as pronounced as that in plasma samples (table S2); concentrations were 1.7-fold higher in the current study, with 3-O-sulfate concentrations in tissue originating from the right and left side of the colon averaging ~54 and 1 nmol/g, respectively (overall range, 0 to 638 nmol/g), after ingestion of 1 g of resveratrol for 8 days before surgical resection. This discrepancy between matrices suggests that metabolites are extracted by methanol with an efficiency closer to that of resveratrol when isolated from tissues compared to plasma. Selected samples were also subjected to liquid chromatography–tandem mass spectrometry (LC-MS/MS) analysis, which not only confirmed metabolite identity for the major products but also revealed the presence of metabolites not formerly detected in human plasma or tissues, namely, resveratrol trisulfate, a disulfate glucuronide, a dihydroresveratrol monosulfate, and a glucuronide (Fig. 1B). Although observed by LC-MS/MS, the reduced metabolite dihydroresveratrol was not detected by HPLC-UV analysis because the limit of detection was ~1000 ng/ml, making the method at least 200-fold less sensitive for measuring this particular derivative, and presumably its conjugates, than for resveratrol (18).

Resveratrol monosulfates regenerate resveratrol in mice

The synthetic scheme adopted for the production of resveratrol monosulfates afforded a 3:2 mixture of 3-O-sulfate and 4′-O-sulfate in ~30% yield (99.49% purity; fig. S1). Because both metabolites are clinically relevant, the mixture was used for the pharmacokinetic studies and initial biological evaluation described here. It was anticipated that any evidence of activity would provide justification for undertaking the time-consuming separation procedures required to isolate the individual isomers in sufficient quantities for future investigations (19). When administered as a mixture to mice by gavage or intravenous injection (for comparison), resveratrol monosulfates were shown, by pharmacokinetic profiling, to be systemically absorbed by the oral route but exhibited poor bioavailability (~14% for the 3-O-sulfate and 3% for resveratrol-4′-O-sulfate). This can be attributed, at least in part, to rapid metabolism; regardless of the route of administration, the monosulfates were subject to secondary transformation, generating glucuronides, a disulfate, and, most importantly, the parent resveratrol in plasma, intestinal mucosa, liver, lung, and pancreas (Fig. 2 and fig. S2). Detection of abundant resveratrol-3-O-glucuronide in plasma and tissues is also consistent with deconjugation of the sulfate isomers in vivo. After oral administration of the resveratrol sulfate mixture, the peak plasma concentrations of resveratrol attained were ~20% of the combined monosulfate Cmax value. In the various organs examined, the resveratrol Cmax values represent ~1 to 7% of the combined maximum sulfate concentrations, which, although lower than those of plasma, still constitute a considerable degree of conversion from sulfate metabolite to the parent compound in tissues (table S3 and fig. S3). Furthermore, resveratrol was present in plasma and liver 6 hours after dosing and persisted for as long as 24 hours within intestinal mucosa, suggesting the potential for prolonged exposure of tissues to resveratrol when formed via this route. Although the average concentration of resveratrol in mouse mucosa increased at 6 hours relative to that at 2 hours, the difference was not statistically significant; therefore, it is not possible to conclude that the 6-hour measurement corresponded to a second peak of resveratrol generated within intestinal tissue, which may have been indicative of increased sulfatase activity.

Fig. 2. Resveratrol sulfates generate resveratrol in vivo.

(B to D) Representative HPLC-UV chromatograms of extracts of plasma (B and C) or colorectal mucosa (D) taken, 1 hour after dosing, from mice that received resveratrol (120 mg/kg) (B) or resveratrol sulfates (120 mg/kg) (3-O-sulfate and 4′-O-sulfate in a 3:2 ratio) (C and D) intragastrically. (A) Chromatogram shows the resveratrol sulfate mixture used for dosing. Metabolites were identified by comparison with synthetic standards where available and confirmed by LC-MS/MS. The internal standard (IS) is naringenin. Graphs (a) and (b) show the 24-hour kinetic profiles of resveratrol generated in plasma and mucosa, respectively (mean + SD, three mice per time point), after administration of resveratrol sulfates.

To quantify the potential contribution of resveratrol regeneration via monosulfate intermediates to the measured plasma levels, we conducted an identical pharmacokinetic study in mice that were administered resveratrol itself (120 mg/kg). After gavage dosing, resveratrol monosulfates accounted for 10.5% of the total resveratrol species present in plasma based on area under the curve (AUC) values (16,644 nmol/g per hour for the 3-O-sulfate versus 158,681 nmol/g per hour for all resveratrol species combined; table S4A). According to the previous sulfate pharmacokinetic study, parent resveratrol equates to 1.9% of the total plasma AUCall in mice (table S4A). Therefore, when resveratrol is taken orally, it can be estimated that 10.5% will be converted to monosulfates, and 1.9% of this conjugated form will be hydrolyzed back to resveratrol in the plasma. Overall, this represents ~0.2% of the initial resveratrol dose. Comparison of the plasma half-life of resveratrol when administered as the parent compound (11.0 hours) or in sulfate form (2.1 hours) suggests that conversion to sulfate conjugates, rather than elimination of resveratrol, may be the rate-limiting step (table S4B).

Sulfate metabolites provide an intracellular reservoir of resveratrol

The propensity for human cells to liberate resveratrol from the sulfate conjugates was then assessed in a panel of colorectal cell lines. Monitoring of the monosulfate concentrations in culture medium (37°C, 5% CO2) over the course of 7 days by HPLC-UV analysis offered no evidence of hydrolysis to the parent resveratrol in the absence of cells. Moreover, both sulfates were stable under these conditions, with no significant change in resveratrol-3-O-sulfate (56.6 ± 2.1 μM versus 57.8 ± 2.5 μM; n = 3; P = 0.56, Student’s t test) and just a small reduction in resveratrol-4′-O-sulfate concentrations (from 20.4 ± 1.1 μM to 16.4 ± 1.1 μM; n = 3; P = 0.01). The human adenocarcinoma cell lines HCA-7 and HT-29, which were developed from malignant colorectal cancers, and human colonic epithelial cells (HCECs), which were derived from normal colonic epithelia, were incubated with clinically achievable concentrations of the resveratrol monosulfate mixture (75 μM) or resveratrol (10 μM), and the uptake kinetics and metabolite profile were determined by analysis of the medium and intracellular contents. These concentrations were chosen to mimic the differential between plasma concentrations of resveratrol and total monosulfates, but were also governed by practical considerations and the balance between using a high enough concentration to maximize the chances of detecting intracellular resveratrol species while avoiding significant toxicity, which would reduce the number of intact cells available for analysis. HT-29 cells appeared the most metabolically active, generating a sulfate glucuronide as the prominent metabolite, and smaller amounts of resveratrol-4′-O-glucuronide in medium after 24 hours of incubation with the monosulfates (Fig. 3). The concentration of these metabolites together with the disulfate and resveratrol-3-O-glucuronide, which appeared subsequently, increased over 7 days to the extent that the 3-O-sulfate accounted for only 3% of the total resveratrol species (fig. S4). However, resveratrol itself was not detected in the medium at any time point. HCA-7 cells produced a qualitatively similar pattern of metabolites in the medium, but the extent of conversion of resveratrol monosulfates to other conjugates was less than that in HT-29 cells; after 7 days, resveratrol-3-O-sulfate still represented 36.0 ± 5.0% (mean ± SD) of the total species in the medium taken from HCA-7 cells, whereas the resveratrol-4′-O-sulfate accounted for 21.8 ± 0.6%, compared to 3.0 ± 1.6% and 11.2 ± 2.2%, respectively, in medium from HT-29 cells (n = 3) (fig. S4). In contrast, HCEC cells displayed minimal metabolic capacity, generating only trace amounts of extracellular resveratrol disulfate and resveratrol-4′-O-glucuronide. This differential activity was also reflected in incubations containing resveratrol; in HT-29 and HCA-7 cells, resveratrol was completely converted (<0.2% remaining) to monoglucuronides and the 3-O-sulfate within 24 hours, whereas resveratrol remained the predominant species in medium from the normal epithelial cells (equating to 76.9 ± 16.8% of the total, mean ± SD, n = 3), with monosulfates the only metabolites detectable (Fig. 3 and fig. S4).

Fig. 3. Resveratrol sulfates are deconjugated and glucuronidated in colorectal cells.

(A to F) Representative HPLC-UV chromatograms of cell culture medium extracted after 24-hour incubations of HCA-7 (A and D), HT-29 (B and E), or HCEC (C and F) cells with either a 3:2 mixture of resveratrol-3-O-sulfate and resveratrol-4′-O-sulfate (75 μM) (A to C) or resveratrol (10 μM) (D to F).

Formation of metabolites implies that resveratrol sulfates can cross cell membranes, and this was confirmed by their presence within HT-29 and HCA-7 cells throughout the entire 7 days of incubation with the sulfate mixture (Fig. 4). Despite its absence from the medium, free resveratrol was also apparent at every time point in the cancer cell extracts, albeit at relatively low amounts, but was never reliably detected in extracts from HCECs. Maximal intracellular resveratrol concentrations were attained at 24 hours and were about threefold higher in the HT-29 (0.16 ± 0.05 ng/mg) compared to the HCA-7 cells (0.05 ± 0.02 ng/mg), consistent with the greater absorption of sulfates by the former. Sulfate entry into HCA-7 cells was rapid, reaching peak concentrations of 3.2 ng/mg (~10 μM) by 15 min of incubation time (for the 3-O-sulfate), but considerably higher amounts were achieved in the HT-29 cells (14.6 ng/mg, or ~45 μM at 24 hours). It seems that the lack of metabolites detected in HCEC medium may be attributed to particularly poor uptake of the sulfates by these cells because intracellular concentrations were consistently below the limit of quantitation (0.001 ng/mg).

Fig. 4. Human colorectal cells generate resveratrol intracellularly from resveratrol sulfates.

(A to F) Representative HPLC-UV chromatograms of extracts of HCA-7 (A and D), HT-29 (B and E), or HCEC (C and F) cells obtained after 24-hour incubations with either a 3:2 mixture of resveratrol-3-O-sulfate and resveratrol-4′-O-sulfate (75 μM) or resveratrol (10 μM). Graphs (1) (HCA-7) and (2) (HT-29) show the changes in intracellular resveratrol-related species after incubation with the sulfate mixture over 7 days. Intracellular concentrations in HCEC cells were below the limit of quantitation and are not shown. Graphs (3) (HCA-7), (4) (HT-29), and (5) (HCEC) show the changes in intracellular resveratrol-related species after incubation with 10 μM resveratrol. (G) Comparison of intracellular resveratrol concentrations achieved in each of the three cell lines over time after incubation with resveratrol sulfates or resveratrol. Error bars indicate SD. n = 3 separate experiments for each cell type and each compound added.

After exposure to resveratrol, maximal concentrations of the parent compound achieved in all three cell lines were similar, ranging from 0.01 to 0.02 ng/mg. In HT-29 and HCA-7 cells, however, resveratrol concentrations were vastly superseded by the amounts of 3-O-sulfate produced, with maximum concentrations reaching ~0.33 and 0.36 ng/mg, respectively (Fig. 4). In contrast, peak 3-O-sulfate concentrations were only fourfold higher than resveratrol concentrations in the normal HCEC cells, and beyond ~30 min of incubation time, resveratrol itself was the major species observed in the HCEC extracts. Therefore, it appears that absorption of free resveratrol, as well as the sulfate conjugates, is restricted in HCEC cells compared to the malignant cells, although another contributing factor influencing the profile could be low expression of the sulfotransferases responsible for resveratrol conjugation or high expression of sulfatases. Whether this lack of resveratrol uptake and metabolism is representative of all noncancer colon cells requires further investigation.

Sulfate uptake correlates with expression of specific membrane transporters

Although the passage of resveratrol across cell membranes can be achieved through both passive diffusion and active processes (20), sulfate conjugates are likely to necessitate a transport mechanism. Comparison of the basal gene expression profiles of candidate transporters in the three cell lines using an array format revealed significant variations that may explain the differences in sulfate uptake (Fig. 5A and fig. S5). Of the 84 genes analyzed, 11 followed a pattern consistent with the sulfate kinetic data, with expression numerically decreasing in the rank order HT-29 > HCA-7 > HCEC, and the ΔCt value [threshold cycle (Ct) minus average Ct value of the housekeeping genes for each plate array] for at least one of the possible pairings being significantly different from each other (Fig. 5A). To date, the products of four of these genes have been ascribed a role in small-molecule drug transport and may conceivably influence uptake of resveratrol sulfate conjugates; the remaining genes encode proteins known to be required for the passage of other substrates such as glucose or nucleosides. Likely contenders based on known cargo specificity are the organic anion transporter SLC22A9 and the organic anion–transporting polypeptides (OATPs) SLCO1B1 and SLCO1B3. Analysis of OATP protein concentrations revealed that OATP1B3 mirrored the gene expression data, with higher expression in both cancer cell lines compared to HCEC cells (Fig. 5B). In contrast, OATP1B1 protein was undetectable in all cell lines. When cells were coincubated with resveratrol sulfate conjugates and ursolic acid, a compound that inhibits both OATP1B1 and OATP1B3 but has greater selectivity for the latter (21), intracellular concentrations of resveratrol-3-O-sulfate were significantly reduced in HT-29 cells by 28 and 14% at 15 and 60 min, respectively, compared to control cells without ursolic acid (P ≤ 0.05, Student’s t test; Fig. 5C). In the same incubations, uptake of the 4′-O-sulfate was also impaired, but the effect failed to reach significance (Fig. 5C). Together, these results suggest that OATP1B3-mediated transport influences intracellular concentrations of resveratrol sulfates, although further experimental confirmation is needed.

Fig. 5. Resveratrol sulfates have selective antiproliferative activity, which correlates with membrane transporter expression.

(A) Comparison of the basal expression of drug transporter genes in cells, determined with an RT2 Profiler PCR array. The array contains 84 test genes, but the chart shows only those genes for which the relative expression profiles correlate with resveratrol sulfate uptake (that is, HT-29 > HCA-7 > HCEC) and for which significant differences [P < 0.01, two-way analysis of variance (ANOVA)] were detected in ΔCt values between at least one of the possible pairings (HT-29 versus HCA-7, HCA-7 versus HCEC, and HT-29 versus HCEC); see Supplementary Materials and Methods for further details. Experiments were conducted in triplicate, and data are expressed relative to gene expression levels in HCA-7 cells (set as 1.0) to facilitate comparisons to cells that attained higher (HT-29) and lower (HCEC) intracellular concentrations of resveratrol monosulfates. *P < 0.01, cases in which transporter gene expression in HCA-7 cells was significantly different from that in HT-29 or HCEC cells; #P < 0.01, genes expressed at significantly higher levels in HT-29 compared to HCEC cells. (B) Basal protein expression of OATP1B1 and OATP1B3 in cell lines (100 μg of total protein loaded in each lane of the gel); positive control is a HepG2 cell lysate (25 μg of protein loaded). (C) Effect of ursolic acid (+UA) coincubation on resveratrol sulfate uptake in HT-29 cells (mean + SD, n = 3). *P ≤ 0.05, Student’s t test, significant decrease compared to control incubations without ursolic acid (−UA). (D) Proportion of cells remaining, relative to control, after incubation with resveratrol sulfates (black) or resveratrol (green) for 7 days (mean + SD of n = 3 separate experiments, each performed in triplicate). *P ≤ 0.0005, one-way ANOVA, significant reduction in cell number compared to dimethyl sulfoxide (DMSO)–treated control cells. (E to G) Expression of LC3-I/II and p21 proteins in HT-29 cells incubated with resveratrol (green) or resveratrol sulfate mixture (black) for up to 72 hours, measured by Western blotting. Signficant increases in protein expression compared to levels in solvent-treated (DMSO) control cells are indicated by *P ≤ 0.05 and **P = 0.01 (Student’s t test). (H) Treatment of HT-29 cells with resveratrol sulfates (black) significantly increased senescence-associated β-galactosidase (SA-β-gal) activity in the cells relative to control cells incubated with DMSO alone, as measured with the Senescence β-Gal Staining Kit after 72 hours of incubation (*P = 0.02, **P = 0.002, Student’s t test); incubation of HT-29 cells with resveratrol (green) had no significant effect on SA-β-gal activity. Data in (E), (F), and (H) are means + SEM from three separate cell culture experiments. (I and J) Representative electron microscopy images of HT-29 cells incubated with 250 μM resveratrol monosulfates (I) or vehicle (DMSO) (J) for 72 hours. Arrows illustrate autophagosomes, late-stage autophagic compartments, and autolysosomes. Morphologically normal-looking nuclei (N) and mitochondria (M) are also indicated. Magnifications for treated and control cells are ×25,000 and ×30,000 respectively.

Vesicular transport experiments have previously shown that resveratrol sulfates are a substrate for breast cancer resistance protein (BCRP; ABCG2), a member of the adenosine triphosphate–binding cassette (ABC) superfamily of membrane transporters (22). ABCG2 was highly expressed in HT-29 cells, with mRNA levels ~350-fold greater than those in HCA-7 cells; however, ABCG2 was also present in the normal colon cell line at amounts ~18 times higher than those in HCA-7 cells (fig. S5). The presence of ABCG2 would be expected to have a negative impact on intracellular concentrations of resveratrol sulfates given that ABC transporters are principally efflux proteins. However, expression of ABCG2 in HCEC cells may not actually contribute to the balance of influx/efflux because the kinetic studies in Fig. 4 suggest that resveratrol sulfates fail to enter these cells to any measurable extent. The relatively higher expression of ABCG2 in HT-29 cells compared to HCA-7 directly contrasts with what might be predicted from the intracellular levels of resveratrol monosulfates; this finding may indicate differences in sulfate uptake or that the contribution of other efflux proteins overrides the effect of ABCG2.

Clinically relevant concentrations of resveratrol monosulfates attenuate cell growth

The ability of clinically relevant levels of resveratrol monosulfates and glucuronide conjugates to inhibit the growth of colorectal cancer and normal epithelial cells was assessed over 7 days and compared to the activity of resveratrol at a concentration achievable in human colon (10 μM). Although both glucuronide conjugates had little effect on cell numbers, even at 250 μM, the mixture of monosulfates produced significant dose-dependent reductions at concentrations equal to or exceeding 25 μM for HCA-7 and 50 μM for HT-29 cancer cells (Fig. 5D). Consistent with the relative intracellular concentrations, the most pronounced growth inhibition was evident in the HT-29 cells. Furthermore, the normal epithelial cell line HCEC was completely unaffected by the presence of resveratrol sulfates over the dose range investigated, although resveratrol retained a degree of activity similar to that measured in HT-29 cells, causing an ~32% maximum reduction in cell number (Fig. 5D).

The decreased cell numbers cannot be explained entirely by simple growth arrest or apoptosis because a concentration of resveratrol monosulfates that caused ~50% growth inhibition (IC50) in HT-29 cells (75 μM) failed to alter the distribution of cells within each phase of the cell cycle over the course of 72 hours or to significantly increase the extent of apoptosis or necrosis above background levels in control cells (fig. S6). Similarly, resveratrol at clinically attainable tissue concentrations (10 μM) was unable to induce significant cell death or growth arrest of HT-29 cells, as measured by propidium iodide staining, which does not distinguish between the G0 and G1 phases of the cell cycle. As expected, considering the lack of an effect of resveratrol sulfates on HCEC growth, we observed no indication of sulfate-stimulated apoptosis, necrosis, or cell cycle arrest in these cells, even when incubated with a concentration of 250 μM (fig. S6).

Alternative processes that may contribute to cell growth inhibition observed are senescence, a stable form of cell cycle arrest, and autophagy, a lysosomal-dependent cellular catabolic pathway required for the quality control of proteins and organelles and for maintenance of energy homeostasis. Both programs can be triggered by cellular stresses and serve as tumor suppressor mechanisms. Moreover, it has recently become apparent that the two pathways are functionally intertwined (23, 24). Treatment of HT-29 cells for 24 hours with the resveratrol sulfate metabolites, but not with resveratrol, significantly enhanced the conversion of soluble microtubule-associated protein 1 light chain 3 (LC3-I) to lipid-bound LC3-II, a constituent of autophagosomal membranes and a marker of autophagy initiation (Fig. 5, E and G). This activity was confirmed by transmission electron microscopy, which revealed characteristic hallmarks of autophagy, including the presence of numerous vesicles with distinct double membranes (Fig. 5, I and J) (25). The link between an antiproliferative effect, or lack thereof, and autophagy was further reinforced by the discovery that resveratrol sulfates failed to stimulate LC3-II production in HCEC cells (fig. S7). In addition, sulfate concentrations of 75 and 250 μM caused a persistent and significant up-regulation of p21 protein expression and amplified SA-β-gal staining at pH 6.0 (two established markers of senescence) in HT-29 cells, but not in the normal colon cell line (fig. S7). Conversely, there was no discernible increase in p21 expression or SA-β-gal staining in cancer cells incubated with resveratrol itself at concentrations of 5 and 10 μM (Fig. 5, F to H), which may be explained by the lower levels of intracellular resveratrol species achieved (Fig. 4 and table S5).

Autophagy and senescence induction are mediated by intracellular resveratrol

To ascertain the direct contribution of resveratrol sulfates to the activity observed in vitro, we coincubated HT-29 cells with the metabolites (75 μM) and a nontoxic concentration of estrone 3-O-sulfamate (EMATE, STX64), a potent active site–directed inhibitor of steroid sulfatase (26). As predicted, EMATE inhibited the intracellular hydrolysis of resveratrol sulfates, significantly reducing conversion to resveratrol by 42% over 24 hours while elevating resveratrol sulfate concentrations by ~31% (Fig. 6A). Higher concentrations of EMATE were not associated with greater inhibition of hydrolysis, implying that resveratrol sulfates serve as substrates for other sulfatase enzymes. The shift in metabolite pattern toward lower resveratrol concentrations was accompanied by a ~71% reduction in LC3-II accumulation relative to cells treated with the sulfates alone, and a ~34% decrease in the expression of the cyclin-dependent kinase inhibitor p21, although this latter effect was not significant. These observations suggest that the parent compound, rather than the sulfate conjugates, is responsible for inducing autophagy and possibly senescence (Fig. 6).

Fig. 6. Autophagy is mediated by intracellular resveratrol.

(A to D) Effects of sulfatase inhibitor EMATE on intracellular generation of resveratrol (A) and LC3-I/II and p21 levels (B to D). HT-29 cells were incubated with EMATE (50 μM) for 1 hour, after which resveratrol sulfates (75 μM of a 3-O-sulfate and 4′-O-sulfate 3:2 mixture) were added, and incubations were terminated 24 hours later. (A) Generation of resveratrol in cell pellets was determined by HPLC-UV analysis. Bars indicate means ± SD of four separate cell culture experiments; significant differences compared to control incubations (−EMATE) are indicated by *P < 0.0001 (Student’s t test). (B) Representative Western blot of lysates from cells incubated with resveratrol sulfates in the presence or absence of EMATE. Densitometric quantitation of LC3-II (C) and p21 (D) protein levels in cells incubated with resveratrol sulfates, EMATE alone, or the combination. Incubations and Western blot analysis were performed on three separate occasions, and values represent means + SEM of protein levels normalized to β-actin and relative to the solvent-treated (DMSO) control cells. *P < 0.05, Student’s t test, significant differences between protein concentrations measured in cells incubated with resveratrol sulfates with or without EMATE.


In recent years, resveratrol has received considerable scientific and public attention for its numerous potential health benefits. However, because of resveratrol’s rapid metabolism and resulting poor bioavailability, doubts persist over whether the promising effects seen in preclinical studies can be translated to humans (8). Whether resveratrol’s major metabolic products—sulfate and glucuronide conjugates—can contribute to activity in vivo has important implications for the future clinical translation of resveratrol, particularly whether the development of alternative prodrugs or drug delivery systems that resist metabolism is indicated.

The mouse pharmacokinetic study presented here provides the first direct demonstration that resveratrol can be generated from its sulfate conjugates and that formation via this route results in sustained exposure to the parent compound. Therefore, the potential exists for prolonged intracellular regeneration of the parent within internal target tissues for as long as these conjugates persist, which is at least 24 hours for the 3-O-sulfate in human plasma (10). Irrespective of the actual species responsible, it is encouraging that pharmacologically achievable concentrations of resveratrol sulfates, as defined in our clinical trials, induce favorable biological effects, namely, inhibition of cancer cell division through contributions from autophagy and senescence. Plasma sulfate concentrations of 20 to 30 μM can be attained in humans with repeated ingestion of 1 g of resveratrol daily, which probably equates to the upper dose limit for prevention purposes, determined on the basis of previous resveratrol tolerability and safety studies in people (11). Resveratrol sulfate concentrations within this range (25 μM) inhibited the proliferation of HT-29 and HCA-7 cancer cells in vitro by ~20% while sparing the normal epithelium–derived HCEC cells. Moreover, the levels reached in human colorectal tissue originating from the right side of the intestine averaged 50 μM, but could reach as high as ~640 μM, which surpasses the concentrations required for 95% inhibition of HT-29 cell growth in culture (250 μM).

Nothing is known about the metabolite profile of resveratrol in human tissues other than the colon, and in the absence of data to the contrary, it seems reasonable to assume that the blood supply will dictate the pattern, with phase 2 conjugates dominating in tissues distant to the gastrointestinal tract. Although resveratrol metabolism has long been considered a major clinical limitation, the demonstration that sulfate metabolites are taken up into human cells and can provide a reservoir for regenerating the parent in situ suggests that the metabolites contribute appreciably to the intracellular concentration and activity of resveratrol. Furthermore, it is feasible that intracellular resveratrol generated by this route may actually play a greater role than the unchanged parent in vivo because the maximum concentration produced by incubation with 75 μM resveratrol sulfates was more than 10-fold higher than that detected in cells treated with 10 μM resveratrol—a concentration that exceeds, by a factor of 16, that which is attainable in human plasma from a 1.0-g daily dose (10). These observations may also help to explain why resveratrol has been shown to exert efficacy in numerous in vivo mouse models that rely on systemic delivery to the target organ, such as suppression of pancreatic and prostate cancer (27, 28), protection against diet-induced metabolic heart disease (29), and delaying of age-related deterioration (30), as well as emerging evidence from clinical trials showing that resveratrol confers metabolic benefits in humans (31).

Other investigators have concluded that resveratrol sulfates are unable to affect the viability of SK-N-AS and NGP neuroblastoma cells (13). This inactivity was attributed to lack of uptake, as detected by fluorescence multiphoton microscopy, which measures the intrinsic fluorescence of resveratrol species. In contrast, individual resveratrol sulfates have some, albeit poor, cytotoxicity in breast cancer cell lines but are considerably less potent than resveratrol (32). These accounts may be rationalized by our findings that uptake of resveratrol sulfates is cell-specific and dependent on the expression of certain transporters. SLCO1B3 seems to contribute to this process, although other members of the solute carrier (SLC) families, such as SLC22A9, may also play a role (33). In colon cells, antiproliferative activity correlated with the amount of resveratrol generated, which was governed by the efficiency of uptake of resveratrol sulfates. Although this process may be the initial determining factor of efficacy, variations in intracellular sulfatase activity will also have an impact on cellular response (34). Although only basal levels of membrane transporters were assessed in this study, exposure to resveratrol derivatives may modulate expression or activity of these proteins over time (35). Many cancer tissues and cell lines have altered expression of OATPs; SLCO1B3, for example, is normally liver-exclusive but is expressed at the mRNA and protein levels in a variety of cancers, including colorectal adenocarcinomas (36, 37). This higher expression may render cancer cells more susceptible to the antiproliferative effects of resveratrol via sulfate intermediates.

In contrast to resveratrol sulfates, the few published studies plus our current findings have consistently shown glucuronide conjugates to be ineffective in biological assays (13). This extends beyond the context of cancer. Resveratrol itself causes cytotoxicity in cultured human peripheral blood mononuclear cells (PBMCs) and, when used in combination with nucleoside analogs, synergistically inhibits virus replication in phytohemagglutinin-activated PBMCs infected with HIV-1; in contrast, resveratrol glucuronides did not cause cytotoxicity and had no impact on HIV-1 infection, even at concentrations of 300 μM (14). Resveratrol glucuronides have relatively high affinity for multidrug resistance protein 3 (MRP3; ABCC3) and are also substrates for ABCG2, albeit at a much lower affinity (22). Although both are expressed at the mRNA level in all three cell lines used in this study, ABC transporters are generally considered to be responsible for drug efflux; therefore, it is unclear whether their presence might aid or hinder activity in our model systems. The complete lack of antiproliferative effects suggests that, if resveratrol glucuronides are taken up by cells, they either are rapidly pumped out or fail to generate resveratrol at sufficient concentrations for activity. Kinetic studies similar to those performed for resveratrol sulfates would help ascertain whether glucuronides play any role in resveratrol efficacy in vivo.

The finding that clinically achievable concentrations of resveratrol sulfates can induce autophagy and potentially senescence may have clinical implications, given the diverse pathologies affected by these processes. The ability of resveratrol to stimulate autophagy is well recognized in experimental systems and is believed to contribute to neuroprotection in animal models of Alzheimer’s disease and Parkinson’s disease, to the attenuation of human prion-mediated neurotoxicity in cultured cells, to the life span–prolonging effects of caloric restriction (3841), and to the preservation of cardiac function during aging (42). The role of autophagy in carcinogenesis is paradoxical; it can act as an oncogenic or tumor-suppressing mechanism (43). Consistent with the present study, resveratrol triggers autophagic cell death in chronic myelogenous leukemia cells (44). Conversely, resveratrol enhances the therapeutic effect of temozolomide, an alkylating agent, through inhibition of autophagy in malignant glioma cells, thereby promoting apoptosis (45).

Autophagy has recently been identified as a new effector mechanism of senescence, important for the rapid protein remodeling needed to make the efficient transition from a proliferative to a senescent state (23). Senescence, which can be triggered by redox stress, DNA damage, or oncogene activation, independently serves as a protective mechanism against cancer, arresting the growth of cells at risk for tumorigenesis and causing immune-mediated clearance (46, 47). Although chronic exposure to resveratrol has been shown to induce senescence-like growth arrest of cancer cells in vitro (48), we found that, using clinically relevant concentrations, senescence was only apparent in resveratrol sulfate–treated cells, which also displayed signs of autophagy; the fact that resveratrol itself failed to cause a significant effect strengthens the hypothesis that resveratrol generated in situ may be of greater importance than the unchanged parent for eliciting efficacy in humans.

Although sulfatase inhibition experiments described above indicate that resveratrol, rather than the sulfate metabolites, caused autophagy and senescence, it remains possible that the conjugates have some intrinsic properties. In the most comprehensive mechanistic assessment of resveratrol metabolites to date, Hoshino et al. reported that at least one of the five possible sulfates investigated could inhibit tumor necrosis factor–α (TNF-α)–induced NFκB activity, decrease the production of nitric oxide by nitric oxide synthase, and induce NQO1 activity in cell-based assays (15). Additional mechanistic investigations, particularly at earlier time points than those studied here, are required to elucidate the underlying molecular changes responsible for the autophagy and senescence caused by exposure to resveratrol sulfate.

Resveratrol has received considerable attention because of its pharmacological properties in preclinical systems. The findings described here suggest a coherent, albeit complicated, mechanistic scenario that explains how the major resveratrol metabolites in humans may contribute to activity; this is particularly important for justifying the use of resveratrol in the prevention or treatment of systemic diseases. Although there is considerable commercial interest in developing resveratrol prodrugs and delivery systems aimed at resisting metabolism, the results of this study suggest that such formulations may not be necessary to deliver efficacious concentrations to target tissues and support further clinical evaluation of resveratrol.


Resveratrol metabolites were synthesized according to adaptations of published methods (19, 49). Details of the clinical trials involving healthy volunteers and patients (registered at as NCT 00098969 and 00433576) have been described previously (10, 17). Human plasma and colorectal mucosa samples were extracted and analyzed with our validated HPLC-UV assay (9, 10, 17) or LC-MS/MS (18). Mouse studies were approved by Leicester University Ethical Review Panel and licensed by the U.K. Home Office. C57BL/6J adult mice (three per group) received a single dose of resveratrol-3-O-sulfate and resveratrol-4′-O-sulfate (3:2 ratio) by either intravenous injection (6 mg/kg in saline) or intragastrically (120 mg/kg in saline) and were culled at various times after dosing (0, 5, 15, 30, and 60 min and 2, 6, and 24 hours). Bioavailability in mice was calculated by comparing AUC values after intravenous and intragastric administration of resveratrol sulfates. An identical pharmacokinetic study was also performed in mice administered intragastric resveratrol (120 mg/kg). Cell cycle and apoptosis analysis was performed by flow cytometry (fig. S6). LC3-I/II and p21 expression were determined by Western blotting and analysis of senescence in cultured cells with the Senescence β-Gal Staining Kit (Cell Signaling Technology). Gene expression was measured with the RT2 Profiler PCR Array (Qiagen). More information and details of statistical analysis are provided in Supplementary Materials and Methods.


Materials and Methods

Fig. S1. 1H nuclear magnetic resonance spectrum of resveratrol sulfate.

Fig. S2. Metabolite profile in mouse tissues after oral resveratrol sulfate.

Fig. S3. Metabolite pharmacokinetics in mice.

Fig. S4. Kinetics of resveratrol/metabolite formation in cell medium.

Fig. S5. Expression of transporter genes.

Fig. S6. Cell cycle and apoptosis analysis.

Fig. S7. Autophagy and senescence markers in HCEC cells.

Table S1. Levels of resveratrol and metabolites in human plasma.

Table S2. Comparison of accurately quantified and estimated colorectal tissue levels of resveratrol metabolites.

Table S3. Mouse pharmacokinetic parameters after resveratrol sulfate administration.

Table S4. Comparison of pharmacokinetic parameters for resveratrol and its metabolites.

Table S5. Intracellular concentrations in HT-29 cells.


  1. Acknowledgments: We thank S. Hyman and N. Allcock (University of Leicester) for electron microscopy assistance and A. Purohit (Imperial College London) for advice on the sulfatase inhibition experiments. Funding: Supported by a Cancer Research UK programme (C325/A6691) with assistance from the Leicester Experimental Cancer Medicine Centre (C325/A15575, funded by Cancer Research UK/UK Department of Health) and U.S. National Cancer Institute (NCI-N01-CN-25025). Author contributions: K.R.P., C.A., R.G.B., E.H.-G., R.S., and K.B. designed and/or performed all laboratory experiments; K.R.P., A.K., and S.S. performed in vivo studies; K.R.P., K.B., C.A., E.H.-G., R.G.B., A.K., and A.J.G. analyzed the data; V.A.B., D.E.B., W.P.S., and A.J.G. designed and/or conducted clinical trials; K.B., A.J.G., W.P.S., and D.E.B. provided funding; K.B. and K.R.P. wrote the paper. Competing interests: The authors declare that they have no competing interests.
View Abstract

Navigate This Article