Comparative dietary sulfated metabolome analysis reveals unknown metabolic interactions of the gut microbiome and the human host

Contents lists available at ScienceDirect

Free Radical Biology and Medicine

journal homepage: www.elsevier.com/locate/freeradbiomed

Original article

Comparative dietary sulfated metabolome analysis reveals unknown metabolic interactions of the gut microbiome and the human host

Mario S.P. Correia

, Abhishek Jain

, Wafa Alotaibi

, Paul Young Tie Yang

, Ana Rodriguez-Mateos

, Daniel Globisch

a Department of Medicinal Chemistry, Science for Life Laboratory, Uppsala University, Box 574, SE-75123, Uppsala, Sweden

b Department of Nutritional Sciences, School of Life Course Sciences, Faculty of Life Sciences and Medicine, King's College London, UK

A R T I C L E I N F O

Microbiome Metabolomics Sulfatase Mass spectrometry Polyphenols Xenobiotics

A B S T R A C T

The gut microbiome converts dietary compounds that are absorbed in the gastrointestinal tract and further metabolized by the human host. Sulfated metabolites are a major compound class derived from this co-meta- bolism and have been linked to disease development. In the present multidisciplinary study, we have in- vestigated human urine samples from a dietary intervention study with 22 individuals collected before and after consumption of a polyphenol rich breakfast. These samples were analyzed utilizing our method combining enzymatic metabolite hydrolysis using an arylsulfatase and mass spectrometric metabolomics. Key to this study is the validation of 235 structurally diverse sulfated metabolites. We have identified 48 significantly upregulated metabolites upon dietary intervention including 11 previously unknown sulfated metabolites for this diet. We observed a large variation in subjects based on their potential to sulfate metabolites, which may be the foun- dation for classification of subjects as high and low sulfate metabolizers in future large cohort studies. The reported sulfatase-based method is a robust tool for the discovery of unknown microbiota-derived metabolites in human samples.

1. Introduction

Microbiota metabolism has been directly linked to human phy- siology and disease development [1]. Trillions of microbes reside in the human body carrying out biochemical conversions of metabolites or- thogonal to the human hosts’ biochemical potential [2,3]. Metabolites produced through microbiome metabolism are known to have bioactive properties that can either be beneficial or toxic to the human body [4,5]. Many dietary components are converted by these microbial communities in the gut including phytochemicals such as flavonoids, ellagitannins, and lignans [6,7]. Single dietary compounds undergo several metabolic processes by diverse microbial species that result in a wide range of metabolites. Upon absorption of these microbe-derived metabolites, the human metabolic clearance mechanism further con- verts these compounds prior to excretion from the body mainly through urine [8,9]. While several diet-specific metabolite classes have been

identified, a plethora of metabolites produced after food consumption remains unknown as common analytical techniques are limited in their detection using mass spectrometric methods.

The human clearance process facilitates removal of xenobiotics in- cluding dietary compounds through metabolic steps that include phase I and phase II modification of the metabolite scaffold [10]. Phase I biotransformation of xenobiotics as well as microbiota-derived com- pounds include oxidation through P450 oxidases, reduction and hy- drolysis reactions followed by phase II modification to insert a hydro- philic moiety for clearance through the kidney. In particular, sulfation of metabolites has been under recent focus as a major compound class of the co-metabolism of microbes and their host [11]. Metabolites produced or converted by the gut microbiome are oxidized and/or sulfated by the human host prior to excretion via urine samples. Me- tabolites in this class have also been suggested as key regulators of bacterial interaction and virulence with their host [12,13]. This co-

https://doi.org/10.1016/j.freeradbiomed.2020.09.006

Received 3 September 2020; Accepted 4 September 2020

Abbreviations: HMDB, Human Metabolome Database; UPLC-MS/MS, ultra-performance liquid chromatography coupled with tandem mass spectrometry; Hp-AS, Helix pomatia arylsulfatase; ASPC, arylsulfatase obtained from KuraBiotech; D3

-NATOS, Deuterated N-acetylserotonin-O-sulfate; MS/MS, tandem mass spectrometry;

m/z, mass to charge ratio; ppm, parts per million; BMI, body mass index; AAA, aromatic amino acid

∗

Corresponding author.

∗∗

Corresponding author.

E-mail addresses: Ana.rodriguez-mateos@kcl.ac.uk (A. Rodriguez-Mateos), Daniel.globisch@scilifelab.uu.se (D. Globisch).

1

M.S.P.C. and A.J. contributed equally to this work.

Available online 11 September 2020

0891-5849/ © 2020 The Author(s). Published by Elsevier Inc. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).

T

gies for the selective enzyme-based analysis of sulfated and glucur- onidated metabolites in human urine and fecal samples [9,22]. We have now extended the scope of this method to facilitate comparative ana- lysis of the urinary sulfate metabolome in a human dietary intervention study. Volunteers (n = 22) collected urine samples before and after consumption of a polyphenol rich breakfast containing raspberries, soy milk and flaxseeds for 3 days (Fig. 1) that were analyzed using our enzymatic and mass spectrometric method in a two-tiered approach:

Step 1 – identification of the chemical structure of sulfated metabolites in these samples; step 2 – comparative sulfate metabolome (sulfatome) analysis based on identified sulfated metabolites in the first step.

2. Materials and methods 2.1. Study design

An open-label, single-arm study investigating the variability in polyphenol gut microbial metabolism and the impact on vascular re- sponse in healthy males and females is currently ongoing. The final cohort has been powered accordingly and expects to recruit a total population of 250 volunteers. In the present work, we analyze a sub- sample of a total of 22 volunteers, recruited from King's College London and surrounding areas. Inclusion criteria comprised good general health, age between 20 and 70 years old, and BMI range between 18.5 and 35 kg/m

. The population recruited was middle age (average 34 years old), overall healthy and with an average BMI of 24.7 kg/m

, at the limit of the overweight range (Supplementary Table 1). Exclusion criteria included history of cardiovascular disease, hypertension, his- tory of diabetes, metabolic syndrome, terminal renal failure or malig- nancies, abnormal heart rhythm (below 60 or above 100 bpm), allergy to berries, flaxseed or soy, smoke an irregular number of cigarettes, taking medications that can affect the cardiovascular system, recent loss of more than 10% of weight, pregnancy or planning to become preg- nant in the next 6 months, or participation in another study in the past month. Participants were asked to refrain from vegetables, fruits, wine, cocoa, chocolate, tea and coffee 24 h prior to the first visit to reduce the influence of background diet. All subjects gave written informed con- sent before their participation in the study and agreed to maintain their eating/drinking and exercise habits for the duration of the study.

Subjects consumed a (poly)phenol breakfast containing 30 g of milled flaxseeds (containing 300 mg of lignans), 40 g of freeze-dried raspberry powder (containing 153 mg of ellagitannins) and 250 ml of soy milk (containing 22 mg of isoflavones) for 3 days. Spot urine samples were collected in a fasted state on day 1 and a 24 h urine sample was collected after consumption of the last breakfast on day 3.

Volume was recorded and samples were stored at −80 °C. The study was conducted in accordance to the guidelines stated in the current revision of the Declaration of Helsinki, and informed consent was ob- tained for all subjects. All procedures involving human subjects were

min and used as negative control in the enzymatic assay. Both assays were shaken at 300 rpm for 17 h at 21 °C and subjected to protein precipitation by addition of cold methanol (4 × the sample volume) for 15 min at 0 °C. After centrifugation (13,780 g for 5 min), the super- natant was collected and dried in vacuo. Afterwards, the remaining pellet was dissolved in 150 μL of water/acetonitrile (95/5, v/v), vig- orously shaken for 30 s and then centrifuged (13,780 g for 5 min). Each supernatant was collected and transferred to a HPLC vial for UPLC-MS/

MS analysis, alternating injection of control and assay samples to avoid biased results.

Step 2 – Sulfatome metabolome analysis: Proteins in urine samples (20 μL) were precipitated with 80 μL of ice-cold methanol for 30 min.

Upon protein precipitation and centrifugation at 13,780 g for 5 min of the urine sample, the supernatant containing the extracted urine me- tabolite mixture was dried in vacuo at ambient temperature.

Afterwards, the remaining pellet was dissolved in 20 μL of water/

acetonitrile (95/5, v/v), vigorously shaken for 30 s and then cen- trifuged (13,780 g for 5 min). Each supernatant was collected and transferred to a HPLC vial for UPLC-MS/MS analysis.

2.3. UHPLC-MS/MS analysis

Mass spectrometric analysis was performed on an Acquity UPLC system connected to a Synapt G2 Q-TOF mass spectrometer, both from Waters Corporation (Milford, MA, USA). The system was controlled using the MassLynx software package v 4.1, also from Waters. The se- paration was performed on an Acquity UPLC® HSS T3 column (1.8 μm, 100 × 2.1 mm) from Waters Corporation. The mobile phase consisted of A) 0.1% formic acid in MilliQ water and B) 0.1% formic acid in LCMS-grade methanol. The column temperature was 40 °C with the gradient : 0–2 min, 0% B; 2–15 min, 0–100% B; 15–16 min, 100% B;

16–17 min, 100-0% B; 17–21 min, 0% B, with a flow rate of 0.2 mL/

min.

The samples were introduced into the q-TOF using negative elec- trospray ionization. The capillary voltage was set to −2.50 kV and the cone voltage was 40 V. The source temperature was 100 °C, the cone gas flow 50 L/min and the desolvation gas flow 600 L/h. The instru- ment was operated in MSE mode, the scan range was m/z = 50–1200, and the scan time was 0.3 s. In low energy mode, the collision energy was 10 V and in high energy mode the collision energy was increased from 25 to 45 V. A solution of sodium format (0.5 mM in 2-propa- nol:water, 90:10, v/v) was used to calibrate the instrument and a so- lution of leucine-encephalin (2 ng/μL in acetonitrile: 0.1% formic acid in water, 50:50, v/v) was used for the lock mass correction at an in- jection rate of 30 s.

2.4. Data analysis

Step 1: Data analysis was performed using the XCMS metabolomics

software package under R (version 1.4.414), using a script designed to identify features with a m/z difference of 79.9568 Da. The results were processed in Excel 2016 and potential sulfate esters were identified using the criteria: 1.2-fold change for the control group, an intensity level higher than a 20,000 ion count and 10 ppm mass accuracy. The sulfate esters were confirmed by MS/MS fragmentation experiments. In low energy mode, the collision energy was 10 V and in high energy mode the collision energy was ramped from 30 to 40 V.

Step 2: Data analysis was performed using the XCMS metabolomics software package under R (version 1.1.414). A list of mass spectro- metric features was obtained with m/z, retention time and peak area information. We specifically selected 235 sulfate esters using the sulfate esters library generated from step 1. Next, we normalized the peak area as per the procedure described below.

2.5. Data normalization

Data was normalized using the internal standard D

-N-acetylser- otonin-O-sulfate (D

-NATOS) for MS technical correction and for the urine's concentration of creatinine levels. To each sample a total of

160 nM of D

-NATOS was spiked in before metabolite extraction. The concentration of creatinine was determined using clinical standard methods. The levels of creatinine for each sample are described in Supplementary Table 2.

2.6. Quantification of enterolactone sulfate

Quantitative mass spectrometric analysis of enterolactone sulfate was performed on a UHPLC coupled to a Triple Quadrupole Mass Spectrometer LCMS-8060 (Shimadzu, Kyoto, Japan). Urine samples were collected from participants and centrifuged immediately after collection (1700 g, 15 min at 4 °C). Supernatants were combined with 2% formic acid to stabilize metabolites and stored in −80 freezer until analysis time. Urine samples were defrosted, filtered using 0.22 μM PTFE filters and diluted (DF = 12) prior to injection. 5.0 μL were in- jected through a Zorbax Eclipse Plus C18 2.1 × 5 mm, 1.8-μm column (Agilent Technologies, California, United States) with mobile phase solvents A: water with 0.1% formic acid (FA) and solvent B: acetonitrile with 0.1% FA in a 6.5 min gradient program. The flow rate 0.5 ml/min started at 1% solvent B and increased to 20% after 1 min, to 99%

Fig. 1. Overview of the workflow for comprehensive UPLC-MS/MS metabolomics analysis. Urine samples were collected before (V1) and after (V2) dietary inter-

vention. Step 1: Sulfatase-based assay for selective identification of sulfated metabolites. Step 2: Global analysis of the sulfated metabolome (sulfatome).

and OPLS-DA plots) was performed using MetaboAnalyst 4.0 [23].

3. Results

We sought to perform a detailed analysis of sulfated metabolites for large-scale identification of microbiota-host co-metabolism and dis- covery of unknown bioactive metabolites in a two-tiered analysis. We employed our recently developed and efficient enzymatic assay for identification of sulfated metabolites in urine samples from different individuals, which requires enzymatic sample pretreatment followed by bioinformatic analysis for selective and sensitive identification of con- verted sulfated metabolites using the metabolomics software package XCMS in R [24]. As the microbiome is known to modulate and produce phenolic compounds from human diet, we performed a comparative sulfated metabolome analysis as the second step to identify changes in sulfated metabolites after 3-day consumption of a dietary intervention in 22 individuals with raspberries, soy milk and flaxseeds, rich in el- lagitanins, isoflavones and lignans, respectively.

3.1. Large scale targeted analysis of sulfated metabolites

Our previously reported assay was designed to pretreat urine sam- ples with purified arylsulfatase from the snail H. pomatia (Hp-AS) [9,25]. The substrate promiscuity of this enzyme warrants conversion of different metabolite scaffolds and the detection of these converted metabolites using bioinformatic analysis. This method also includes a control sample that is incubated in parallel with denatured enzyme to resemble the exact same mass spectrometric background. While this method is perfectly applicable for small sample numbers, analysis of large sample cohorts would require unnecessary amounts of biological samples and reagents. To minimize sample volume and enzyme quan- tities for the control sample, we have now modified our previous pro- cedure without limiting the selectivity. We designed the experiment by systematically grouping samples from four different individuals. 20 μL of each sample were pooled to constitute the control sample and treated with denatured enzyme (Supplementary Table 3). Urine samples from each of these four individuals were extracted separately (80 μL each) and treated with Hp-AS. These samples were analyzed in comparison to the corresponding pooled and matched control sample (Fig. 2). This procedure ensures the same background derived from the enzyme so- lution and an average metabolite content of each sample. Metabolites were extracted from each urine sample using methanol precipitation of proteins and enzymes followed by standard metabolite extraction pro- cedures. To test the suitability of the pooled control sample strategy, we initially compared our developed sulfate identification method based on individual control samples to one pooled control of these four urine samples (Supplementary Figs. 1 and 2). To ensure comparison of all samples, the sample lists were randomized for UPLC-MS analysis in- cluding four injections of each control sample. This adapted analysis led

3.2. Identification and structure elucidation of 235 sulfated metabolites As a first step of our investigation, we determined the structure of sulfated metabolites present in volunteer samples. UPLC-MS analysis was performed for 32 selected urine samples from individuals before (V1) and after (V2) dietary intervention (16 each group), followed by analysis using XCMS in the R software package [24]. Samples were analyzed in negative mode and mass spectrometric features were se- lected based on these criteria: i) fold change > 1.2; ii) ion count >

20,000; and iii) 10 ppm mass accuracy. We specifically searched for features with a difference in m/z of 79.9568 (the loss of a sulfate moiety) to identify sulfated metabolites. We successfully confirmed a sulfate ester moiety present in the metabolite scaffold of 235 com- pounds through selective MS/MS fragmentation for each detected mass spectrometric feature. We confirmed a sulfate moiety in these MS/MS fragmentation experiments but no putative structure could be assigned for 105 metabolites. These metabolites carrying a sulfate ester are classified as level 3 (Table 1). The structure for 130 sulfated metabolites was validated at higher confidence levels via MS/MS fragmentation experiment evaluation. In total, 27 metabolites were validated through co-injection experiments of chemically synthesized or commercially available authentic standards (level 1). Hereby, the same chromato- graphic properties and identical MS/MS fragmentation pattern of the natural and synthetic molecule unambiguously validated the metabolite structure (Fig. 3A/B) and represents the highest level of confidence for structure elucidation as even regioisomers can be distinguished. Fur- thermore, the chemical structure was determined for an additional 105 metabolites by correlation of mass spectrometric fragmentation pattern of the sulfated or the corresponding hydrolyzed metabolite using da- tabases or computational tools including HMDB, SIRIUS, Metlin, and MZmine (Fig. 3C) [27–29]. This method represents the second and third highest level of certainty for structure validation of metabolites [9,30].

We separated comparison with experimental data available in data- bases and literature (level 2a) and state-of-the-art computational frag- mentation tools (level 2b). The structure of the metabolite can be de- termined at both levels but validation of the correct regioisomer and substitution pattern of aromatic compounds requires an authentic standard for co-injection experiments. Validation of the molecular structure using MS/MS fragmentation pattern at levels 2a/2b was possible through comparison with the corresponding unsulfated meta- bolites present in databases as the sulfate ester gets cleaves easily under applied mass spectrometric fragmentation conditions. The spectra of both molecules are almost identical and the only differences derive from the loss of the MS-labile sulfate moiety (Supplementary Fig. 3).

The validation of 130 metabolite structures at confidence levels 1–2

exceeds the number identified in previous dietary studies as these

merely report targeted analyses of specific metabolite classes. The

large-scale coverage of metabolite structures includes a wide range of

different compound classes of sulfated metabolites such as flavonoids,

phenolic acids, cinnamic acids, acetophenones, phenylacetic acids, coumarins, phenylpropanoids, and phenylpropenes. We identified the sulfated metabolites of vanillic acid, ferulic acid, 5'-(3′,4′-dihydrox- yphenyl)-gamma-valerolactone, sinapic acid, dihydrocaffeic acid, p- coumaric acid, 3-hydroxyphenylpropionic acid, hydroxybenzoic acid, caffeic acid, enterolactone, enterodiol as well as several daidzein and genistein metabolites (Supplementary Table 4). These compounds have previously been detected in human samples after raspberry, flaxseed and soy consumption. Furthermore, we identified 25 metabolites that are of bacterial origin as the corresponding unsulfated compound are a product of microbial metabolism (Supplementary Table 5). The detec- tion of these sulfated metabolites validates the suitability of our ana- lytical methodology. All identified metabolites were used for global comparative sulfatome analysis of all 44 samples collected.

3.3. Global human sulfatome analysis

We next performed a global sulfate analysis on the variation of all confirmed 235 sulfated metabolites in 22 volunteers. We decided to include all metabolites at any validation level identified in step 1 to cover the entire detectable sulfate content in this second step analysis (Fig. 1). Over the past decade it has been revealed that dietary inter- ventions can be monitored through common metabolic signatures at a population level. However, it has not yet been possible to observe heterogeneous and individual shifts in metabolism. Personalized me- tabolic signatures of dietary interventions stem from long-term dietary habits, the individual genome, a baseline gut microbiome composition as well as other factors such as physiology, age, sex, and BMI [7,31]. As all these factors may affect sulfated metabolites excreted through urine from the human body, we sought a targeted and comprehensive ana- lysis of this metabolite class has a high potential to identify diet-specific metabolic pattern. A 1.5-fold change in mass spectrometric areas

between V2 and V1 was applied. The number of metabolites was de- termined that were either upregulated, downregulated, or within this fold change and revealed that the same polyphenol-rich diet affected each individual in a personalized manner. The subjects were arranged (from left to right) based on the percentage of upregulated sulfated metabolites after consumption of the polyphenol-rich breakfast for 3 days (Fig. 4). We observed a large individual variety between subjects for their potential to sulfate metabolites. Nine subjects have more than 118 upregulated sulfated metabolites (> 50%), while for the other 13 volunteers the number of upregulated sulfate esters is below 50%. In- terestingly, we also identified three subjects with more than 100 sul- fated metabolites down-regulated after the dietary intervention, which is in stark contrast to all other volunteers. It is tempting to speculate that these individual responses of subjects as a result of a dietary in- tervention can be used to classify the subjects as the high and low metabolizers in large scale future studies that support the idea of per- sonalized nutrition [31]. This observation can be the result of different reasons such as a similar diet at the baseline sample collection (V1) or poor metabolic potential for these dietary compounds.

3.4. Investigation of specific diet-induced sulfated metabolites derived from gut microbiome-host co-metabolism

The selective identification of 235 sulfated metabolites allows for a comprehensive comparative analysis before and after dietary inter- vention. Multivariate analysis revealed a distinction between the two groups before and after diet based on their sulfated metabolite com- position (Supplementary Fig. 4). Due to the distinct link of sulfation to the co-metabolism of the human host and its microbiota, this compar- ison allows for gaining unparalleled insights in this interspecies co- metabolism.

Our analysis revealed 48 sulfated metabolites with significantly Fig. 2. Overview of the analytical strategy for selective detection and validation of urinary sulfated metabolites. Comparative analysis of both groups, before (V1) and after (V2) dietary intervention, were combined for the pooled control sample to reduce sample volume and enzyme. Representative structure validation for each group are exemplified for p-hydroxyhippuric acid sulfate (top) and 3-hydroxycinnamic acid sulfate (bottom). Details for the analytical strategy are described in

Table 1

Number of identified sulfated metabolites with the level of confidence for their structure validation (More details are provided in Supplementary Table S4).

Level of confidence Validation via UPLC-MS/MS sample analysis Metabolite no.

1 Validation with authentic synthetic or commercial standards 27

2a Metabolite structure validation based on unambiguous matching of MS² spectra with experimental spectra from literature or library

sources 59

2b Identification of the molecular formula and MS² fragmentation pattern comparison using computational tools 44

3 MS²-validation of sulfate ester moiety in the metabolite 105

Total identified sulfate esters 235

increased levels upon dietary intervention (p < 0.05, Supplementary Table 6). Several significantly upregulated metabolites in this study have previously been reported as metabolic signatures upon con- sumption of raspberries, flaxseeds and soy products, which further va- lidates the suitability and strengths of our method. Among these me- tabolites, we discovered the four new metabolites dihydrodaidzein sulfate, tetrahydrodaidzein sulfate, 4-hydroxyphenylpyruvic acid sul- fate, and dihydroxyconiferyl alcohol sulfate. These have not yet been reported in human samples, although their corresponding unsulfated metabolites are known to be intermediates of polyphenol gut microbial metabolism [32]. Most other metabolites are known metabolites de- rived from raspberry, flaxseeds or soy consumption and linked to major dietary molecule classes but have never been analyzed in parallel in a single study.

Flaxseeds (and raspberries, to a much lower extent) are a rich source of lignans, a class of phytoestrogens, which lead to the formation of the gut microbial metabolites enterolactone and enterodiol [33]. As ex- pected, our analysis identified significantly increased levels of their corresponding sulfated metabolites (Supplementary Table 6). More- over, two bis-conjugated metabolites, enterolactone glucuronide sulfate and enterodiol glucuronide sulfate were among the upregulated meta- bolites. These compounds have previously been only detected after seven days of a flaxseed diet [34]. To confirm the validity of our study Fig. 3. Identification and structure elucidation of 235 sulfated metabolites. (A) Co-injection experiments of natural urinary and synthesized sinapic acid sulfate. (B) Fragmentation pattern comparison of the synthetic and the metabolite in the urine sample (10 V). (C) Example for structure validation of sulfated metabolites by comparison of the fragmentation pattern of the corresponding unsulfated metabolite in databases (10 V). A complete list with validation level and structure assignments for each identified sulfated metabolite is presented in Supplementary Table 4.

Fig. 4. Global human sulfatome analysis. Classification of subjects based on

sulfated metabolite level changes of all identified 235 sulfate esters before and

after dietary intervention. Individual data is listed according to the number of

upregulated sulfated metabolites.

design and normalization procedure, we precisely quantified en- terolactone sulfate in each V2 urine sample using an authentic standard on a triple quadruple mass spectrometer. These quantitative values perfectly correlated with the semi-quantitative method described in this study (R

= 0.9173, Supplementary Fig. 5). This experiment validates our semi-quantitative correlation results to draw reliable biological conclusions.

We also observed the two sulfated metabolites indoxyl sulfate, and hippuric acid sulfate at reduced levels after dietary intervention in eight individuals (> 1.5-fold). These microbial metabolites are derived from aromatic amino acid (AAA) metabolism. Consumption of polyphenol rich diets have previously been associated with a diet-induced sup- pression of AAA metabolism [35]. However, these metabolites were not significantly decreased but should be included in future studies of larger sample cohorts to investigate a potential correlation of these metabo- lites and this dietary intervention.

Recent studies have suggested that impact of dietary changes on human host metabolism is person-specific and no one-size-fits-all diet can be beneficial for everyone [31]. Thus, we have also investigated how the 28 most significantly upregulated metabolites (p < 0.01) are distributed among the 22 subjects at an individual level (Fig. 5A and Supplementary Fig. 6). As expected, none of these metabolites showed a consistent increase in all 22 individuals. For example, 4-ethylphenyl sulfate was more than 1.5-fold upregulated in 20 individuals, while it was more than 1.5-fold downregulated in one individual and remained unchanged in one individual. We only observed a few exceptions for all 28 metabolites. The chemical structure was determined for 23 out of these 28 sulfated compounds. The individual response to the (poly) phenol rich diet of these individuals can possibly be driven through specific (meta)genomic host and microbiome differences and by their background diet. Our observation suggests that personalized evaluation through comprehensive analysis of a variety of nutritional metabolites could be used to classify individual host and microbiome composition to design individual beneficial diets.

3.5. Discovery of unknown sulfated metabolites

Several previously unidentified compounds in human samples were among the 235 sulfated metabolites described in this study. We de- tected 33 sulfated metabolites for the first time that have only been registered in HMDB as their corresponding desulfated compound (Fig. 5B) [29]. The metabolite structures were determined at level 2a/

2b and the structure for one tentative regioisomer is illustrated. These metabolites can be divided into different metabolite classes based on their scaffolds: 17 polyphenolic compounds (red/1–17), four N-het- erocyclic compounds (purple/18–21), three indole molecules (blue/

22–24), three new hippuric acids sulfates (green/25–27), and six with diverse scaffolds different from the first four groups that also include aliphatic alcohols (black/28–33).

The majority of the newly discovered sulfates are polyphenolic compounds, of which many are derived from main dietary components.

1-Methylpyrogallol, harmalol, methyl-3-hydroxyindole-3-acetate or N- methyltyramine are common metabolites in daily diets such as fruits, cereals, beverages or flavoring agents and we have discovered their corresponding phase II conjugates (13, 18, 24, and 31). All three hip- puric acid metabolites are well described microbiome metabolites and either derived from fatty acid or phenylalanine metabolism. The two metabolites 4-hydroxyhippuric acid sulfate (26) and 3-hydroxyhippuric acid sulfate (27) were confirmed with synthetic standards. Other me- tabolites have been described as common products from either human metabolism or dietary intake. Metabolites such as cytosine or 5-me- thylcytosine are part of the pyrimidine metabolism and were detected as sulfated metabolites for the first time (19, 20).

Hydroxyacetophenone and 4-hydroxyphenylpyruvic acid are molecules derived from bacterial degradation processes or from food sources.

Other metabolite classes are involved in specific metabolic pathways,

such as indoles 5,6-dihydroxyindole sulfate (22) and 5-hydro- xytryptophol sulfate (23). Both indole-based metabolites are bi-pro- ducts of serotonin and microbiome metabolism. Some discovered me- tabolites have also been linked to diseases such as the corresponding phenolic compound of hydroxy-acetophenone sulfate (6) was described to be dysregulated in patients with type II diabetes and 5,6-dihydrox- yindole sulfate (22) is a metabolite altered in patients with malignant melanoma.

4. Discussion

The interdisciplinary analysis of a dietary intervention study re- vealed that targeted investigation of urinary sulfated metabolites is a versatile tool for identification of diet induced metabolic changes. This comprehensive and comparative urinary sulfate analysis led to identi- fication of previously unreported sulfated metabolites and unknown metabolic signatures derived from raspberry, flaxseed and soy milk consumption. Key to our findings was the parallel analysis of 235 sul- fated metabolites using state-of-the-art metabolomics software tool for comparative analysis of urine samples before and after dietary con- sumption. We have elucidated and validated the chemical structure of 130 sulfated metabolites through chemically synthesized reference metabolites (level 1) or database comparison of UPLC-MS/MS frag- mentation spectra (level 2a/2b).

Most of these investigated metabolites are linked to microbiota metabolism that we found to be significantly upregulated after dietary intervention. Combined these are demonstrating the coverage of dif- ferent metabolic pathways of human and microbiome co-metabolism.

This compound class contains microbiota and food-derived metabolites that were further transformed by the human host. We have identified 11 previously unknown sulfated molecules produced after consumption of raspberries, flaxseeds and soybeans. These metabolites should be validated in future studies to evaluate their potential as biomarkers for consumption of one of the three diets. This is the first targeted analysis of this co-metabolism compound class and exceeds previous dietary studies that were mainly focused on single compound classes.

Furthermore, this comprehensive study also led to the discovery of a plethora of unknown sulfate esters with unknown bioactivity and can be applied for comparative analysis of important sulfated metabolites in any type of human sample to uncover unknown links of human and microbiome co-metabolism.

The main components of soy are the isoflavonoids daidzin and genistin, which are metabolized by microbes in the gut into the bioactive phytoestrogens daidzein and genistein [36]. Their detected sulfated compounds can be considered as markers for gut microbiota- human host co-metabolism. In here, we identified five metabolites de- rived from daidzein metabolism (daidzein sulfate, daidzein-7-glucur- onide-4-sulfate, dihydrodaidzein sulfate, tetrahydrodaidzein sulfate, 8,2′-dihydroxyflavone sulfate) and two metabolites derived from gen- istein metabolism (genistein-7-glucuronide-4-sulfate and 4-ethylphenyl sulfate) (Fig. 6A). Especially, the bisconjugated metabolites containing one glucuronide and one sulfate moiety are difficult to detect and not commonly reported in dietary intervention studies. Our results are in close agreement with Hosoda et al., who reported the presence of genistein-7-glucuronide-4-sulfate and daidzein-7-glucuronide-4-sulfate in urine after ingestion of baked soybean flour [36]. Importantly, di- hydrodaidzein sulfate and tetrahydrodaidzein sulfate have not been described in dietary intervention studies yet and this is the first dis- covery of these molecules in human samples. They have only been re- ported in rat urine after high dosage of the pure parent compound daidzein that exceed concentrations in common diets [37]. Another soy diet product is 4-ethylphenyl sulfate, which was also found to be up- regulated in the dietary intervention group (V2) [38].

Anthocyanins are one of the main components of raspberries. This

compound class produces many commonly known gut microbiota-de-

rived metabolites such as p-coumaric acid, caffeic acid, ferulic acid, and

4-hydroxyphenylpyruvate. Three sulfated analogues of these com- pounds were significantly upregulated (Fig. 6B). The only exception is p-coumaric acid sulfate for which no significant alteration was ob- served. Upregulated levels were also detected for the two sulfated downstream metabolic products of ferulic acid, dihydroconiferyl al- cohol and sinapic acid. Caffeic acid 3-sulfate and ferulic acid 4-O-sulfate have been associated with raspberry rich diets [39]. Sinapic acid sulfate was detected in human urine after tea intake as well as in rat urine after whole rye consumption [40]. The discovered upregulation in these samples is to the best of our knowledge the first study to describe this metabolite as markers for raspberry, soybeans or flaxseed polyphenol consumption.

Additionally, we observed significantly upregulated levels of 3,4- dihydroxyphenylacetic acid sulfate, hydroxytyrosol sulfate, tyramine sulfate, homovanillic acid sulfate, and homovanillyl alcohol sulfate, which are involved in dopamine oxidative metabolism. Hydroxytyrosol and tyrosol are commonly present in polyphenol rich diets and are also synthesized in the human body from the neurotransmitter

-dopamine.

The major metabolite of dopamine in biological matrices is 3,4-dihy-

droxyphenylacetic acid. In a minor metabolic pathway, dopamine is

also metabolized to hydroxytyrosol. Homovanillic acid is the metabolite

generated from 3,4-dihydroxyphenylacetic acid, while homovanillyl

alcohol is the methylated metabolite of hydroxytyrosol. Tyramine is

synthesized through decarboxylation of tyrosine through microbial

metabolism and a precursor of tyrosol, which in turn act as a precursor

of hydroxytyrosol [41]. To the best of our knowledge, the phase II

metabolites homovanillyl alcohol sulfate, tyramine sulfate, hydro-

xytyrosol sulfate, dihydroxyphenylacetic acid sulfate, and homovanillic

acid sulfate have not yet been considered as potential metabolites of

soy, flaxseeds or raspberry consumption. However, some of these me-

tabolites have been detected in other (poly)phenol rich diets. Homo-

vanillic acid sulfate has been associated with consumption of cranberry

juice and grape extracts [42,43]. Dihydroxyphenylacetic acid sulfate

was detected after intake of almonds [44]. An elevated level of hy-

droxytyrosol sulfate was linked to red wine consumption. An increased

level of three gut microbiota-derived sulfate conjugates of two

Fig. 5. Individual significantly upregulated sulfated metabolites and newly discovered metabolites. (A) Individual distribution of 28 metabolites that were sig-

nificantly upregulated after dietary intervention (V2/V1; p < 0.01). (green: upregulated metabolites by at least 1.5-fold; blue: downregulated metabolites by at least

1.5-fold; grey: metabolites that were in-between these factors). Please see Supplementary Fig. 6 for individual fold changes and Supplementary Table 6 for p-values

and fold changes for each metabolite. (B) Previously unreported sulfated metabolites according to the HMDB. Molecules are color coded and grouped based on their

molecular scaffolds: red – (poly)phenols; blue – indoles; green – hippuric acids; purple – Heterocycles; black – diverse scaffolds. One potential structure is included for

metabolites with more than one phenolic or aliphatic alcohol (details of validation level are provided in Supplementary Table 4).

dihydroxyphenyl-γ-valerolactone analogues as well as dihydronar- ingenin after dietary intervention complements previous studies on polyphenol rich diets including raspberries and tea [40,45].

One of our objectives was to evaluate the potential of this general compound class for identification of different metabolic potentials of individuals as a result of the dietary intervention. Based on these me- tabolite sulfate signatures we indeed found that individuals have a substantially different metabolic potential to sulfate microbiome-de- rived metabolites. We suggest that this new targeted mass spectrometric metabolomics strategy to uncover and compare the co-metabolism of host and gut microbiota to be termed as sulfatome (sulfate metabo- lome) analysis.

5. Conclusion

This is the first report of distinct individual dietary differences based on large-scale analysis of sulfated metabolites. The strength of this approach compared to other dietary intervention studies is the clus- tering of individuals based on 235 metabolites rather than only single metabolite classes. Among these sulfated metabolites, we have identi- fied previously unknown metabolites in human samples as well as metabolites that have not been linked to this diet previously. We vali- dated the structure for 130 of these metabolites at a high level of confidence, which are now available for correlation analysis in other dietary and urinary metabolomics analyses. We also discovered pro- found individual metabolic differences for this dietary intervention that could lead in future studies with larger cohorts to classification of subjects based on their potential to sulfate metabolites.

Declaration of competing interest

We declare no financial or other relationships that may lead to a conflict of interest in this study.

Acknowledgements

We are grateful for funding by the Swedish Research Council [VR 2016-04423], the Swedish Cancer Foundation [19 0347 Pj], Carl Tryggers Foundation [CTS 2016:155/CTS 2018:820], a project grant from SciLifeLab [SLL 2018/12], and a generous start-up grant from SciLifeLab (to D.G.). The StratiPol study is funded by an unrestricted grant from the National Processed Raspberry Council and Washington Red Raspberries (PI A.R.M.) and a scholarship by King Faisal University to W.A. We highly appreciate discussion with Tobias Sjöblom (Uppsala University) and Kura Biotech (Jose Luis Callejas, Camila Berner and Manuel Rozas) for providing the sulfatase ASPC. Thanks are also ex- tended to Anthony Sullivan, Chris Titman and Neil Loftus from Shimadzu UK Ltd. for their support on the LC-MS 8060 instrument used.

Appendix A. Supplementary data

Supplementary data to this article can be found online at https://

doi.org/10.1016/j.freeradbiomed.2020.09.006.

Author contributions

Conceptualization, D.G.; Methodology, M.S.P.C and A.J.;

Investigation, M.S.P.C., A.J., W.A., and P.Y.T.Y.; Visualization, M.S.P.C., A.J. and D.G.; Resources, all authors; Writing – Original Draft, M.S.P.C., A.J. and D.G.; Writing – Review & Editing, all authors;

Supervision, A.R.M. and D.G.; Funding Acquisition, A.R.M. and D.G.

References

[1] M. Arumugam, J. Raes, E. Pelletier, D. Le Paslier, T. Yamada, D.R. Mende, G.R. Fernandes, J. Tap, T. Bruls, J.-M. Batto, M. Bertalan, N. Borruel, F. Casellas, L. Fernandez, L. Gautier, T. Hansen, M. Hattori, T. Hayashi, M. Kleerebezem, K. Kurokawa, M. Leclerc, F. Levenez, C. Manichanh, H.B. Nielsen, T. Nielsen, N. Pons, J. Poulain, J. Qin, T. Sicheritz-Ponten, S. Tims, D. Torrents, E. Ugarte, E.G. Zoetendal, J. Wang, F. Guarner, O. Pedersen, W.M. de Vos, S. Brunak, J. Dore, J. Weissenbach, S.D. Ehrlich, P. Bork, H.I.T.C. Meta, Enterotypes of the human gut

Fig. 6. Investigation of specific diet-induced sulfated metabolites derived from gut microbiome-host co-metabolism. An overview of significantly altered sulfated metabolites after dietary intervention of 22 individuals. Highlighted are the main dietary sources for each metabolite class. (A) Depiction of metabolic changes of sulfated analogues of metabolites of phytoestrogens. Scatter plots for sulfated isoflavones present in soy and lignans mainly present in flaxseeds demonstrate semi- quantitative changes of each metabolite. (B) Depiction of metabolic changes of anthocyanin-derived sulfated metabolites that are mainly present in raspberries.

Values are normalized mass spectrometric peak areas of the EICs for each metabolite (Mean values ± SEM). p-values: * < 0.05, ** < 0.01, *** < 0.001 (detailed p

values are reported in Supplementary Table 6).

trometric methodology for the selective investigation of gut microbiota-derived metabolites, Chem. Sci. 9 (29) (2018) 6233–6239.

[10] C.H. Johnson, A.D. Patterson, J.R. Idle, F.J. Gonzalez, Xenobiotic metabolomics:

major impact on the metabolome, Annu. Rev. Pharmacol. Toxicol. 52 (2012) 37–56.

[11] N. Koppel, V. Maini Rekdal, E.P. Balskus, Chemical transformation of xenobiotics by the human gut microbiota, Science 356 (6344) (2017).

[12] S.K. Hatzios, C.R. Bertozzi, The regulation of sulfur metabolism in Mycobacterium tuberculosis, PLoS Pathog. 7 (7) (2011) e1002036.

[13] J.D. Mougous, R.H. Senaratne, C.J. Petzold, M. Jain, D.H. Lee, M.W. Schelle, M.D. Leavell, J.S. Cox, J.A. Leary, L.W. Riley, C.R. Bertozzi, A sulfated metabolite produced by stf3 negatively regulates the virulence of Mycobacterium tuberculosis, Proc. Natl. Acad. Sci. U. S. A 103 (11) (2006) 4258–4263.

[14] A. Visconti, C.I. Le Roy, F. Rosa, N. Rossi, T.C. Martin, R.P. Mohney, W. Li, E. de Rinaldis, J.T. Bell, J.C. Venter, K.E. Nelson, T.D. Spector, M. Falchi, Interplay be- tween the human gut microbiome and host metabolism, Nat. Commun. 10 (1) (2019) 4505.

[15] C.D. Kay, G. Pereira-Caro, I.A. Ludwig, M.N. Clifford, A. Crozier, Anthocyanins and flavanones are more bioavailable than previously perceived: a Review of recent evidence, Annu. Rev. Food Sci. Technol. 8 (2017) 155–180.

[16] G. Garcia, S. Nanni, I. Figueira, I. Ivanov, G.J. McDougall, D. Stewart, R.B. Ferreira, P. Pinto, R.F.M. Silva, D. Brites, C.N. Santos, Bioaccessible (poly)phenol metabolites from raspberry protect neural cells from oxidative stress and attenuate microglia activation, Food Chem. 215 (2017) 274–283.

[17] E. Van Rymenant, J. Van Camp, B. Pauwels, C. Boydens, L.V. Daele, K. Beerens, P. Brouckaert, G. Smagghe, A. Kerimi, G. Williamson, C. Grootaert, J.V. de Voorde, Ferulic acid-4-O-sulfate rather than ferulic acid relaxes arteries and lowers blood pressure in mice, JNB (J. Nutr. Biochem.) 44 (2017) 44–51.

[18] B.M. Burton-Freeman, A.K. Sandhu, I. Edirisinghe, Red raspberries and their bioactive polyphenols: cardiometabolic and neuronal health links, Adv. Nutr. 7 (1) (2016) 44–65.

[19] M.-T. García-Conesa, K. Chambers, E. Combet, P. Pinto, M. Garcia-Aloy, C. Andrés- Lacueva, S. De Pascual-Teresa, P. Mena, A. Konic Ristic, J.W. Hollands, A.P. Kroon, A. Rodríguez-Mateos, G. Istas, A.C. Kontogiorgis, K.D. Rai, R.E. Gibney, C. Morand, C.J. Espín, A. González-Sarrías, Meta-analysis of the effects of foods and derived products containing ellagitannins and anthocyanins on cardiometabolic bio- markers: analysis of factors influencing variability of the individual responses, Int.

J. Mol. Sci. 19 (3) (2018).

[20] N. Garg, L.P. Conway, C. Ballet, M.S.P. Correia, F.K.S. Olsson, M. Vujasinovic, J.M. Lohr, D. Globisch, Chemoselective probe containing a unique bioorthogonal cleavage site for investigation of gut microbiota metabolism, Angew. Chem. Int. Ed.

57 (42) (2018) 13805–13809.

[21] W. Wang, Q. Yang, Y. Du, X. Zhou, X. Du, Q. Wu, L. Lin, Y. Song, F. Li, C. Yang, W. Tan, Metabolic labeling of peptidoglycan with NIR-II dye enables in vivo ima- ging of gut microbiota, Angew Chem. Int. Ed. Engl. 59 (7) (2020) 2628–2633.

[22] M.S.P. Correia, M. Rao, C. Ballet, D. Globisch, Coupled enzymatic treatment and mass spectrometric analysis for identification of glucuronidated metabolites in human samples, Chembiochem 20 (13) (2019) 1678–1683.

[23] J. Chong, D.S. Wishart, J. Xia, Using MetaboAnalyst 4.0 for comprehensive and integrative metabolomics data analysis, Curr. Protoc. Bioinformatics 68 (1) (2019) e86.

[24] X. Domingo-Almenara, J.R. Montenegro-Burke, J. Ivanisevic, A. Thomas, J. Sidibe, T. Teav, C. Guijas, A.E. Aisporna, D. Rinehart, L. Hoang, A. Nordstrom, M. Gomez- Romero, L. Whiley, M.R. Lewis, J.K. Nicholson, H.P. Benton, G. Siuzdak, Xcms- Mrm, Metlin-Mrm, A cloud library and public resource for targeted analysis of small

[31] A.A. Kolodziejczyk, D.P. Zheng, E. Elinav, Diet-microbiota interactions and perso- nalized nutrition, Nat. Rev. Microbiol. 17 (12) (2019) 742–753.

[32] F. Rafii, The role of colonic bacteria in the metabolism of the natural isoflavone daidzin to equol, Metabolites 5 (1) (2015) 56–73.

[33] N.K. Horner, A.R. Kristal, J. Prunty, H.E. Skor, J.D. Potter, J.W. Lampe, Dietary determinants of plasma enterolactone, Cancer Epidemiol. Biomark. Prev. 11 (1) (2002) 121–126.

[34] A. Quartieri, R. García-Villalba, A. Amaretti, S. Raimondi, A. Leonardi, M. Rossi, F. Tomàs-Barberàn, Detection of novel metabolites of flaxseed lignans in vitro and in vivo, Mol. Nutr. Food Res. 60 (7) (2016) 1590–1601.

[35] Y. Zhou, N. Zhang, Y.A. Arikawa, C. Chen, Inhibitory effects of green tea poly- phenols on microbial metabolism of aromatic amino acids in humans revealed by metabolomic analysis, Metabolites 9 (5) (2019).

[36] K. Hosoda, T. Furuta, K. Ishii, Metabolism and disposition of isoflavone conjugated metabolites in humans after ingestion of Kinako, Drug Metabol. Dispos. 39 (9) (2011) 1762.

[37] W. Zhao, T. Yuan, C. Piva, E.J. Spinard, C.W. Schuttert, D.C. Rowley, D.R. Nelson, The probiotic bacterium phaeobacter inhibens downregulates virulence factor transcription in the shellfish pathogen Vibrio coralliilyticus by N-acyl homoserine lactone production, Appl. Environ. Microbiol. 85 (2) (2019) e01545-18.

[38] V. Neveu, A. Moussy, H. Rouaix, R. Wedekind, A. Pon, C. Knox, D.S. Wishart, A. Scalbert, Exposome-Explorer: a manually-curated database on biomarkers of exposure to dietary and environmental factors, Nucleic Acids Res. 45 (D1) (2017) D979–D984.

[39] I.A. Ludwig, P. Mena, L. Calani, G. Borges, G. Pereira-Caro, L. Bresciani, D. Del Rio, M.E.J. Lean, A. Crozier, New insights into the bioavailability of red raspberry an- thocyanins and ellagitannins, Free Radic. Biol. Med. 89 (2015) 758–769.

[40] J.J.J. van der Hooft, R.C.H. de Vos, V. Mihaleva, R.J. Bino, L. Ridder, N. de Roo, D.M. Jacobs, J.P.M. van Duynhoven, J. Vervoort, Structural elucidation and quantification of phenolic conjugates present in human urine after tea intake, Anal.

Chem. 84 (16) (2012) 7263–7271.

[41] D. Globisch, A.Y. Moreno, M.S. Hixon, A.A.K. Nunes, J.R. Denery, S. Specht, A. Hoerauf, K.D. Janda, Onchocerca volvulus-neurotransmitter tyramine is a bio- marker for river blindness, Proc. Natl. Acad. Sci. U. S. A 110 (11) (2013) 4218–4223.

[42] R.P. Feliciano, E. Mecha, M.R. Bronze, A. Rodriguez-Mateos, Development and validation of a high-throughput micro solid-phase extraction method coupled with ultra-high-performance liquid chromatography-quadrupole time-of-flight mass spectrometry for rapid identification and quantification of phenolic metabolites in human plasma and urine, J. Chromatogr. A 1464 (2016) 21–31.

[43] J. van Duynhoven, J.J.J. van der Hooft, F.A. van Dorsten, S. Peters, M. Foltz, V. Gomez-Roldan, J. Vervoort, R.C.H. de Vos, D.M. Jacobs, Rapid and sustained systemic circulation of conjugated gut microbial catabolites after single-dose black tea extract consumption, J. Proteome Res. 13 (5) (2014) 2668–2678.

[44] R. Llorach, I. Garrido, M. Monagas, M. Urpi-Sarda, S. Tulipani, B. Bartolome, C. Andres-Lacueva, Metabolomics study of human urinary metabolome modifica- tions after intake of almond (prunus dulcis (mill.) DA webb) skin polyphenols, J.

Proteome Res. 9 (11) (2010) 5859–5867.

[45] F. Sánchez-Patán, C. Cueva, M. Monagas, G.E. Walton, G.R. Gibson M, J.E. Quintanilla-López, R. Lebrón-Aguilar, P.J. Martín-Álvarez, M.V. Moreno- Arribas, B. Bartolomé, Vitro fermentation of a red wine extract by human gut mi- crobiota: changes in microbial groups and formation of phenolic metabolites, J.

Agric. Food Chem. 60 (9) (2012) 2136–2147.