Signatures of anthocyanin metabolites identified in humans inhibit biomarkers of vascular inflammation in human endothelial cells

Scope The physiological relevance of contemporary cell culture studies is often perplexing, given the use of unmetabolized phytochemicals at supraphysiological concentrations. We investigated the activity of physiologically relevant anthocyanin metabolite signatures, derived from a previous pharmacokinetics study of 500 mg 13C5‐cyanidin‐3‐glucoside in eight healthy participants, on soluble vascular adhesion molecule‐1 (VCAM‐1) and interleukin‐6 (IL‐6) in human endothelial cells. Methods and results Signatures of peak metabolites (previously identified at 1, 6, and 24 h post‐bolus) were reproduced using pure standards and effects were investigated across concentrations ten‐fold lower and higher than observed mean (<5 μM) serum levels. Tumor necrosis factor‐α (TNF‐α)‐stimulated VCAM‐1 was reduced in response to all treatments, with maximal effects observed for the 6 and 24 h profiles. Profiles tested at ten‐fold below mean serum concentrations (0.19–0.44 μM) remained active. IL‐6 was reduced in response to 1, 6, and 24 h profiles, with maximal effects observed for 6 h and 24 h profiles at concentrations above 2 μM. Protein responses were reflected by reductions in VCAM‐1 and IL‐6 mRNA, however there was no effect on phosphorylated NFκB‐p65 expression. Conclusion Signatures of anthocyanin metabolites following dietary consumption reduce VCAM‐1 and IL‐6 production, providing evidence of physiologically relevant biological activity.


Introduction
The consumption of anthocyanins has been linked to a reduced risk of cardiovascular disease [1,2], though their mechanisms of action are not fully understood. Traditionally, in vitro studies have explored the activity of parent anthocyanins mechanistically, however their low plasma concentrations and rapid clearance kinetics suggests they are not the bioactive forms responsible for in vivo activity. It is therefore probable that anthocyanin bioactivity in vivo results from the lesser studied, though more bioavailable, phenolic metabolites, and we have recently demonstrated they are more active on inflammatory biomarkers than their precursor structures [3][4][5].
The understanding of anthocyanin metabolism is relatively contemporary, though it is commonly accepted that their degradation is a result of their chemical instability and the impact of bacterial catabolism, resulting in a number  of circulating phenolic metabolites [6,7]. As anthocyanin metabolites do not circulate in isolation following ingestion, but exist as complex mixtures or profiles of metabolites at various concentrations [8][9][10], it is important that this is also reflected in the design of experiments exploring the bioactivity of anthocyanins. The metabolism of a common dietary anthocyanin, cyanidin-3-glucoside (C3G), was recently investigated and 29 metabolites were identified following the consumption of 500 mg 13 C-labelled cyanidin-3-glucoside (C3G) [8,11]. Briefly, eight healthy male participants were fed a single 500 mg oral bolus dose of isotopically labeled C3G ( 13 C 5 -C3G) following a 7-day washout period (avoiding anthocyanin-rich foods), where blood was collected at baseline, 0.5, 1, 2, 4, 6, 24, and 48 h. Three distinct peak serum metabolite profiles (or signatures) were observed post consumption at 1, 6, and 24 h (Fig. 1). Similar groupings of metabolites sharing C max and producing distinct biosignatures or peaks of metabolites during clearance have also been observed following consumption of cocoa flavan-3-ols [12] and citrus flavanones [13], suggesting this is a common response in the clearance kinetics of flavonoid metabolites. Given that these phenolics circulate at higher concentrations and for longer duration relative to their precursor structures, there is scope to investigate the collective activity of blood profiles of phenolic metabolites on inflammatory processes.
Anthocyanin metabolites have been shown to inhibit the expression of a number of inflammatory biomarkers, such as those involved in vascular adhesion and chemotaxis, including soluble vascular cellular adhesion molecule-1 (VCAM-1) and interleukin-6 (IL-6) [14,15], both of which are markers of cardiovascular disease risk and mortality [16,17] and logical targets for exploring the potential mechanisms of action of anthocyanin metabolites. The aims of the present study were therefore: investigate the effects of unique C3G metabolite signatures, observed to peak in vivo at 1, 6, and 24 h post consumption, on VCAM-1 and IL-6 protein secretion by two cell types, human umbilical vein endothelial cells (HUVEC) and human coronary artery endothelial cells (HCAECs); investigate the effects of metabolite signatures across a range of concentrations, reflecting levels tenfold lower (<0.5 M) and ten-fold higher (<50 M) than mean concentrations (<5 M) observed previously [8,11]; finally, identify mechanistic effects on VCAM-1 and IL-6  mRNA by targeting a key inflammatory transcriptional target, NFB.

Treatment metabolite profiles
Stock solutions for cell culture experiments were prepared in 100% DMSO at 200 mM and stored at -80ЊC with the exception of cyanidin-3-glucoside, which was prepared at 40 mM, and the sulfate-conjugated phenolic acids, which were prepared at 25 mM in 50% DMSO (50% PBS) to maintain stability whilst reducing final DMSO concentrations in working solutions. Working solutions of 1 mM were prepared in supplemented media before being diluted to their working concentrations (Table 1) and stored at 4ЊC until experimental commencement (with the exception of cyanidin-3-glucoside, which was added immediately prior to the final dilutions to maintain stability).
Solutions were subsequently diluted in supplemented media as required (Table 1) immediately prior to experiment commencement.

Cell culture
HUVECs and HCAECs were routinely cultured in fibronectin coated T75 flasks (0.25 g/cm 2 ), using large vessel endothelial growth medium and endothelial cell medium MV, respectively, at 37ЊC and 5% CO 2 . Cells were sub-cultured using a trypsin passage pack or Detach Kit, according to the manufacturer's instructions. HUVECs were used at passage 4 and HCAECs were used between passages 3 and 6. All cells were incubated in supplemented media for 24 h at 37ЊC, 5% CO 2 , in a humidified atmosphere, prior to experiment commencement.

VCAM-1 and IL-6 protein expression
HUVEC or HCAEC were seeded at 80 000 cells/well in fibronectin coated 24-well plates. Cells were treated for 30 min with peak metabolite profiles identified previously at 1, 6, 24 h post consumption (Table 1) or 0.01% DMSO (vehicle control) prior to the addition of 10 ng/mL (HUVEC) or 0.1 ng/mL (HCAEC) TNF-␣ (determined as providing maximal induction while maintaining physiologically relevant concentrations following time-and concentration-response experiments), and incubated for 24 h at 37ЊC, 5% CO 2 , in a humidified atmosphere. Supernatants were collected on ice, centrifuged at 2000 ×g for 10 min at 4ЊC, and stored at -80ЊC prior to ELISA. Samples were thawed at room temperature and vortexed for 3 × 5 s immediately prior to analysis. Supernatants were diluted 1:1 in Reagent Diluent (R&D Systems) and protein expression of soluble VCAM-1 (sVCAM-1; hereby referred to as VCAM-1) and IL-6 were determined by commercially available DuoSet enzyme-linked immunosorbent assay (ELISA) (R&D Systems), according to the manufacturer's instructions. Absorbance values for all ELISA plates were recorded using an OMEGA plate reader from BMG LABTECH (Bucks, UK).

VCAM-1 and IL-6 mRNA expression
HCAEC were seeded at 200 000 cells/well in fibronectin coated 6-well plates. Cells were pre-treated for 30 min with the highest working concentrations of the metabolite profiles (19,20, 44 M, reflecting serum concentrations at 1, 6, 24 h respectively) of each serum profile or 0.01% DMSO (vehicle control) prior to the addition of 0.1 ng/mL TNF-␣, and incubated for 4 h at 37ЊC, 5% CO 2 , in a humidified atmosphere. Cell culture supernatants were removed and cells washed 3× with PBS. Total RNA was extracted from HUVECs and reverse transcribed to cDNA using conditions previously described [3]. Real-time quantitative PCR (RT-qPCR) was carried out using 25 ng cDNA with the addition of VCAM-1 primers (forward primer, 5'-CAGGCTAAGTTACATATTGATGACAT-3'; reverse primer, 5'-GAGGAAGGGCTGACCAAGAC-3'), or IL-6 primers (forward primer, 5'-GCAGAAAACAACCTGAACCTT; reverse primer, 3'-ACCTCAAACTCCAAAAGACCA-5') and real-time PCR Precision master mix with SYBR green (Primer Design, UK). RT-qPCR was carried out using the ABI7500 (Applied Biosystems, UK) system, using cycle methods previously described [3]. Relative changes in gene expression from the TNF-␣ control were quantified using the comparative Ct method [19]. The difference between recorded Ct values for treatment and positive control samples were calculated in the first instance for all genes. geNORM analysis was carried out using qbasePLUS software (version 2.3; Biogazelle, Belgium) to determine stable reference genes (VIPAS39 and PRDM4), the geometric mean of which were used to normalize the data in subsequent experiments.

Data analysis
VCAM-1 and IL-6 protein (pg/mL) or mRNA (fold change) were recorded as the mean of two technical duplicates, and reported relative to the TNF-␣ positive control (containing TNF-␣ without DMSO) as the mean of three independent experiments. Phospho-NFB p65 expression (infrared density) data were normalized to GAPDH reference gene and data were presented graphically as a fold change of vehicle control (DMSO). Treatment effects were established by one-way analysis of variance (ANOVA) with post-hoc least significant difference (LSD) conducted using SPSS for Windows (version 22.0; IBM, New York, USA). Untreated controls were not included in the ANOVA for treatment effect but presented graphically, where Students t-test established difference relative to vehicle control (DMSO). All data represent the mean ± SD of three biological replicates (n = 3).   Figure 2. Effect of peak metabolite signatures on TNF-␣ stimulated VCAM-1 secretion by A) HUVEC, B) HCAEC. Cells were treated with three concentrations of three serum metabolite profiles (representing ten-fold lower and ten-fold higher concentrations than the mean concentrations observed by Czank et al. [8]; Table 1) prior to the addition of 10 or 0.1 ng/mL TNF-␣ for 24 h. Data were normalized to a TNF-␣ control (no DMSO) and columns represent the mean ± SD, n = 3 biological replicates. Labeled means without a common letter differ, p ࣘ 0.05 (ANOVA with post hoc LSD). *Different from DMSO, p ࣘ 0.05 (t-test). Abbreviations: HUVEC, human umbilical vein endothelial cells; TNF-␣, tumor necrosis factor-␣; VCAM-1, soluble vascular adhesion molecule-1.

Effect of peak cyanidin-3-glucoside metabolite signatures on VCAM-1 protein expression
The effects of signatures of cyanidin-3-glucoside metabolites (Fig. 1) on TNF-␣ stimulated VCAM-1 secretion was explored at the mean concentrations observed clinically, as well as concentrations ten-fold lower and ten-fold higher (Table 1), in HUVEC and validated in HCAEC. No treatments were cytotoxic at any tested concentration as established utilizing the WST-1 cytotoxicity assay (Roche, United Kingdom). HUVEC-secreted VCAM-1 ( Fig. 2A) was reduced relative to the vehicle control in response to all treatments at all tested concentrations, with the maximal effects observed for 6 and 24 h metabolite signatures (−65.12 ± 0.37% and −66.24 ± 2.88, respectively; p ࣘ 0.001) at cumulative concentrations of 20 and 44 M (respectively). The activity expressed in HCAEC was slightly less than that of the HUVEC, however similarly, inhibition of HCAEC-secreted VCAM-1 (Fig. 2B)

Effect of peak cyanidin-3-glucoside metabolite signatures on IL-6 protein expression
The effects of cyanidin-3-glucoside metabolite signatures (Fig. 1) on TNF-␣ stimulated IL-6 secretion was explored at the mean concentration observed clinically (Table 1), as well as ten-fold lower and ten-fold higher concentrations in HU-VEC and validated in HCAEC. HUVEC-secreted IL-6 ( Fig. 3A) was reduced relative to the vehicle control in response to all treatments, with the exception of the concentrations tested 10-fold below the mean serum levels (0. 19

Effect of peak cyanidin-3-glucoside metabolite signatures on phosphorylated NFB p65 expression
Peak metabolite signatures were further explored for their effect on TNF-␣ stimulated NFB transcription factor p65 in HCAEC (Fig. 5), where there was no apparent activity beyond that of the vehicle control (p > 0.05).

Discussion
The present study is the first to explore the activity of physiologically relevant signatures of anthocyanin metabo-lites. Here we utilized three unique treatments based on peak concentrations observed in serum post consumption ( Fig. 1) [8,11], with the aim of elucidating in vivo activity. Peak signatures of metabolites (Table 1) displayed significant inhibitory effects on VCAM-1 protein secretion (Fig. 2) at concentrations observed in vivo [11], suggesting physiologically achievable bioactivity. The greatest inhibition of VCAM-1 was observed in response to the 24 h metabolite signature, suggesting metabolites of lower intestinal microbial origin are responsible for fasting or chronic anti-inflammatory effects. Peak metabolite signatures appeared to have biological activity, despite the extremely low concentrations of their individual constitutes. Here responses were often greater then activities previously reported by our group for the same metabolites in isolation or in mixtures at equal molar concentrations [4,5]. Previous studies have explored this concept of physiologically relevant concentrations of metabolite profiles by applying extracted serum/plasma to cell culture models. For example, Koga et al. found serum containing metabolites of (+)-catechin significantly reduced U937 adhesion to human aortic endothelial cells (HAEC) relative to the pure metabolites in isolation [21]. Other studies have used both extracted animal and human plasma in a similar manner [22][23][24], however the limitation of this study design is that plasma contains many bioactive constituents other than flavonoid metabolites, making it difficult to compare treatments to controls and eliminate confounders stemming from varying endogenous analytes. There is scope to validate the present findings using this model, however, these limitations still exist. The novelty of the present study is that the experiments can be appropriately controlled for vehicle, which allows a more direct exploration of mechanism of action. It could however be argued that this in itself is a limitation, as metabolites may act differentially in serum relative in cell culture media; Indeed, certain flavonoids have been shown to interact with serum albumin in in vitro or ex vivo conditions [25]. Furthermore, in utilizing average reported metabolite concentrations, compositions do not reflect individual participants' blood profiles. Finally, the study design cannot capture the impact of other blood metabolites/analytes escaping sample preparation and detection methodologies.
The highest mean concentration of metabolites detected following the 500 mg bolus of 13 C-labeled anthocyanins (equivalent to the consumption of approximately 100 g of blackberries [26]) was observed at 24 h post consumption and totaled 4.38 M [11]. In this study, a high inter-individual variation in metabolism was observed. For example, the serum C max for hippuric acid was 1962 ± 1389 nM, indicating the mean concentration varies greatly between individuals and could be in excess of 3000 nM for a single metabolite. The present study sought to address this issue by utilizing treatment concentrations reflecting the lowest and highest concentrations reported between individuals (0.80-13.18 M [8,11]). As such, we used three metabolite concentrations, representing 10-fold higher and lower concentrations   Figure 3. Effect of peak metabolite signatures on TNF-␣ stimulated IL-6 secretion by A) HU-VEC, B) HCAEC. Cells were treated with three concentrations of three serum metabolite profiles (representing ten-fold lower and ten-fold higher concentrations than the mean concentrations observed by Czank et al. [8]; Table 1) prior to the addition of 10 or 0.1 ng/mL TNF-␣ for 24 h. Data were normalized to a TNF-␣ control (no DMSO) and columns represent the mean ± SD, n = 3 biological replicates. Labeled means without a common letter differ, p ࣘ 0.05 (ANOVA with post hoc LSD). *Different from DMSO, p ࣘ 0.05 (t-test). Abbreviations: HCAEC, human coronary artery endothelial cells; TNF-␣, tumor necrosis factor-␣; IL-6, interleukin-6.
then the observed mean. Surprisingly, there was very little difference in the inhibition of VCAM-1 protein expression between the mean and ten-fold lower and higher concentrations of the metabolites. This suggests that either there is a threshold activity or that there is something unique about these mixtures of metabolites which, when combined, have some additive or synergistic activity. This outcome is important as these concentrations reflect dietary achievable levels [27]. As phenolic metabolites are common to a number of dietary flavonoids and food sources [28], it is possible that metabolite signature concentrations utilized in the present study could be exceeded following a habitual polyphenol-rich diet, given that consumption of polyphenols in Europe has been estimated between 744-1786 mg/day [29]. As effects were observed at the lowest concentrations in the present study (between 0.19 and 0.44 M), even low levels of dietary polyphenol consumption would have beneficial effects on inflammatory status.
Although peak metabolite profiles inhibited IL-6 secretion at 1, 6, and 24 h in HUVEC, this was not reflected in HCAEC, though it is possible that effects were masked by the large variation between replicates. It is possible that the reduced stimulus (TNF-␣) concentration (0.1 ng/mL in the present study relative to 10 ng/mL) increased variation as a result of low IL-6 induction, making it difficult to quantify significant activity.
would reflect changes in protein expression. In our recent study [5], it was observed that only protocatechuic acid (PCA) inhibited VCAM-1 mRNA expression at the highest tested concentration, 100 M. It is interesting that VCAM-1 and IL-6 mRNAs were reduced by half in response to the three metabolite profiles which reflect cumulative total metabolite concentrations of only 19, 23, and 44 M, respectively. Given the low concentrations of metabolites in the present treatment mixtures, it appears that certain metabolites are acting additively or synergistically, potentially through effecting multiple pathways simultaneously. Multiple pathways are indeed thought to be affected following anthocyanin metabolite treatment, for example, aortas of ApoE-deficient mice fed an anthocyanin-rich bilberry extract, demonstrated the modulation of 1261 genes which code for proteins involved in the regulation of cellular processes, including adhesion and inflammatory biomarker expression [14].
NFB is a key transcription factor pathway in the TNF-␣ stimulated expression of adhesion molecules in endothelial cells [30]. In the present study, no effect was observed on the expression of phosphorylated p65, suggesting alternative mechanisms which influence adhesion molecule expression are at play, such as AP-1 activity via p38 and JNK MAP kinase. A recent study of ours [5] as well as that of Krga et al. [31] also demonstrated no activity of flavonoid metabolites on NFB. Data from the present study suggest that the metabolite profile with the maximum inflammatory effect was observed at 24 h post-consumption, suggesting fasting or chronic effects are possible. Conversely, improvements in flow-mediated dilation (FMD) and blood pressure in response to feeding anthocyanins are most often observed acutely, between 1 and 6 h post-consumption (maximum response at 2 h) [32]. In the study of Rodriguez-Mateos et al. [32], following the consumption of a drink containing blueberry anthocyanins, benzoic and vanillic acids positively correlated with FMD at 1-2 h, whereas hippuric, hydroxyhippuric, and homovanillic acids correlated with the FMD at 6 h. These data suggests an acute-phase modulatory vascular response of phenol metabolites. Based on these findings and those of the present study it is possible that sudden vascular responses are mediated by very low levels of parent flavonoids and their rapid degradation products, which are succeeded by a delayed anti-inflammatory response, mediated by products of lower intestinal bacterial catabolism and hepatic phase II conjugation. This ultimately suggests a dual mechanistic activity of flavonoids. This hypothesis requires conformation in randomized-control trials designed having inflammation as a primary endpoint and utilizing populations both responsive to inflammatory and vascular intervention following dietary manipulation (i.e., neither completely healthy nor chronically unhealthy), as flavonoids are likely to be effective as a preventative strategy rather than a pharmacological therapy.
In conclusion, the present study identified that signatures of anthocyanin metabolites, identified post consumption of dietary achievable levels of anthocyanins, have inhibitory effects on inflammatory protein secretion. Further work is required to elucidate the multiple mechanisms potentially at play, ultimately informing our understanding of how anthocyanins and other flavonoids impact health.