The Journal of Nutrition xxx (xxxx) xxx
journal homepage: www.journals.elsevier.com/the-journal-of-nutrition
Nutrient Physiology, Metabolism, and Nutrient-Nutrient Interactions A water-soluble tomato extract rich in secondary plant metabolites lowers trimethylamine-n-oxide and modulates gut microbiota: a randomized, double-blind, placebo-controlled cross-over study in overweight and obese adults Ateequr Rehman 1 , Susan M. Tyree 2 , Sophie Fehlbaum 1 , Gillian DunnGalvin 2 , Charalampos G. Panagos 1 , Bertrand Guy 1 , Shriram Patel 3 , Timothy G. Dinan 4 , Asim K. Duttaroy 5 , Ruedi Duss 1 , Robert E. Steinert 1 , 6 , * 1 DSM Nutritional Products, Kaiseraugst, Switzerland; 2 Atlantia Clinical Trials, Cork, Ireland; 3 SeqBiome, Cork, Ireland; 4 Atlantia Clinical Trials, Cork, Ireland, APC Microbiome Ireland, Cork, Ireland, Department of Psychiatry and Neurobehavioral Science, University College Cork, Cork, Ireland; 5 Department of Nutrition, Institute of Basic Medical Nutrition, Faculty of Medicine, University of Oslo, Norway; 6 Department of Surgery, Division of Visceral and Transplantation Surgery, University Hospital Zurich, Zurich, Switzerland A B S T R A C T Background: Natural products rich in polyphenols have been shown to lower plasma trimethylamine-n-oxide (TMAO) known for its proatherogenic effects by modulating the intestinal microbiota. Objectives: We aimed to determine the impact of Fruit fl ow, a water-soluble tomato extract, on TMAO, fecal microbiota, and plasma and fecal metabolites. Methods: Overweight and obese adults ( n ¼ 22, BMI 28 – 35 kg/m 2 ) were included in a double-blind, placebo-controlled, cross-over study receiving 2 150 mg Fruit fl ow per day or placebo (maltodextrin) for 4 wk with a 6-week wash-out between interventions. Stool, blood, and urine samples were collected to assess changes in plasma TMAO (primary outcome) as well as fecal microbiota, fecal and plasma metab- olites, and urine TMAO (secondary outcomes). In a subgroup ( n ¼ 9), postprandial TMAO was evaluated following a choline-rich breakfast (~450 mg). Statistical methods included paired t-tests or Wilcoxon signed rank tests and permutational multivariate analysis of variance. Results: Fruit fl ow, but not placebo, reduced fasting levels of plasma ( 1.5 μ M, P 0.05) and urine ( 19.1 μ M, P 0.01) TMAO as well as plasma lipopolysaccharides ( 5.3ng/mL, P 0.05) from baseline to the end of intervention. However, these changes were signi fi cant only for urine TMAO levels when comparing between the groups ( P 0.05). Changes in microbial beta, but not alpha, diversity paralleled this with a signi fi cant difference in Jaccard distance-based Principal Component ( P 0.05) as well as decreases in Bacteroides , Ruminococccus , and Hungatella and increases in Alistipes when comparing between and within groups ( P 0.05, respectively). There were no between-group differences in SCFAs and bile acids (BAs) in both faces and plasma but several changes within groups such as an increase in fecal cholic acid or plasma pyruvate with Fruit fl ow ( P 0.05, respectively). An untargeted metabolomic analysis revealed TMAO as the most discriminant plasma metabolite between groups ( P 0.05). Conclusions: Our results support earlier fi ndings that polyphenol-rich extracts can lower plasma TMAO in overweight and obese adults related to gut microbiota modulation. This trial was registered at clinicaltrials.gov as NCT04160481 (https://clinicaltrials.gov/ct2/show/ NCT04160481?term ¼ Fruit fl ow & draw ¼ 2 & rank ¼ 2). Keywords : prebiotic, humans, TMAO, cardiovascular disease, polyphenols
Abbreviations: AEs, adverse events; BA, bile acids; BMI, body mass index; CA, cholic acid; CDCA, chenodeoxycholic acid; CLR, entroid Log Ratio; CVD, ardiovascular disease; FOS, ructooligosaccharides; FXR, farnesoid X receptor; GCA, glycocholic acid; GCDCA, glycochenodeoxycholic acid; GDCA, glycodeoxycholic acid; GOS, galactooligosaccharides; GSRS, Gastrointestinal Symptom Rating Scale; HB, Benjamini-Hochberg; hsCRP, high sensitive C-reactive protein; IP, investigational product; LPS, lipopolysaccharides; PC1, Jaccard Principal Component; PERMANOVA, Permutational Multivariate Analysis of Variance; PLS-DA, Partial Least Squares Discriminant Analysis; OTU, Operational taxonomic unit; RCT, randomized clinical trial; SAEs, serious adverse events; SCFA, short-chain fatty acids; TCDCA, taur- ochenodeoxycholic acid; TDCA, taurodeoxycholic acid; T2DM, type 2 diabetes; TMAO, trimethylamine-N-oxide; VIP, Variable Importance in Projection. * Corresponding author. E-mail address: robert.steinert@dsm.com (R.E. Steinert).
https://doi.org/10.1016/j.tjnut.2022.11.009 Received 29 August 2022; Received in revised form 26 October 2022; Accepted 23 November 2022; Available online xxxx 0022-3166/ © 2022 American Society for Nutrition. Published by Elsevier Inc. All rights reserved.
A. Rehman et al.
The Journal of Nutrition xxx (xxxx) xxx
Introduction
to investigate the effect of 4 wk of supplementation of 2 150 mg Fruit fl ow on plasma TMAO (primary outcome), as well as fecal microbiota and concentrations of urine TMAO and other fecal and plasma metabolites, including lipopolysaccharides (LPS), bile acids (BA), SCFAs, and other organic acids (sec- ondary outcome). Given that fasting plasma TMAO may un- derestimate the TMAO-producing function of gut microbiota because of host ’ s ef fi cient renal clearance of plasma TMAO, we also performed an egg challenge to determine the acute effect of the intervention on postprandial plasma TMAO concentrations.
Targeted modulation of the gut microbiota has been suggested as a preventative and/or novel therapeutic approach for several diseases, including obesity, type 2 diabetes mellitus (T2DM), and cardiovascular or intestinal in fl ammatory disease [1 – 3]. Several nutritional regimes are available today, including pre and pro- biotics; however, their full potential remains unexploited particu- larly in the case of prebiotics [4, 5]. To foster the appropriate use of theterm prebiotic , the International Scienti fi c Association of Probiotics and Prebiotics published a consensus paper in 2017, de fi ning a prebiotic as “ a substrate that is selectively utilized by host microor- ganisms conferring a health bene fi t [6]. ” Besides a few well-established prebiotics, including fructans [fructooligo- saccharides (FOS) and inulin] and galactans [gal- actooligosaccharides (GOS)] that fi t that category, several prebiotic candidates were listed to impact host microorganisms using invitro fermentation or clinical studies. However, the challenge remains to satisfy the criterion of conferring a host health bene fi t. Among the listed prebiotic candidates, plant polyphenols are also included with an estimated 90% to 95% that are not absorbed in the small intestine but reach the colon to undergo extensive biotransformation by the gut microbiota [7]. It has been hypothesized that the health bene fi ts associated with polyphenol consumption may depend on microbial utilization and metabolites produced rather than on the parent compound [8]. In addition, polyphenols may modulate the gut microbiota to confer a health bene fi t. One example is that polyphenols may attenuate trimethylamine-N-oxide (TMAO)-induced atheroscle- rosis by remodeling the gut microbiota and thus lowering TMAO synthesis [9 – 11]. TMAO is derived from trimethylamine (TMA), a microbial metabolite produced by various taxa of the gut microbiota primarily from dietary phosphatidylcholine and L-carnitine, commonly found in red meat, cheese, and eggs [12]. TMA is absorbed through the intestinal epithelium, transported to the liver, and converted into TMAO. TMAO is known for its proin fl ammatory and proatherogenic activities, and high base- line levels of TMAO have been linked to major adverse cardio- vascular events [13 – 16]. Fruit fl ow, a watery tomato extract, was the fi rst product in Europe to obtain an approved, proprietary health claim ( Water- soluble tomato concentrate helps maintain normal platelet aggrega- tion ) under Article 13 [5] of the European Health Claims Regu- lation 1924/2006 on nutrition and health claims [17,18]. It contains a range of tomato-derived secondary metabolites, including nucleosides, phenolic conjugates, and polyphenols, all showing different anti-platelet activity [17, 19 – 22]. The com- pounds present are all water-soluble and of low molecular weight; some (for example, nucleosides) are absorbed from the upper gastrointestinal tract very quickly on ingestion, whereas others (for example, fl avonoid glycosides) are absorbed much later in the digestive process, either from the lower small intes- tine or after microbial interaction in the proximal colon [17, 23]. It has been suggested that the different metabolic fates of the diverse compounds may contribute to the modulation of platelet function; however, the effect on TMAO through gut microbial changes remains unknown [24]. We hypothesized that Fruit fl ow would lower the TMAO levels by modulating gut microbiota and therefore performed a randomized, double-blind, placebo-controlled cross-over trial
Methods
Study population The study population comprised 40 overweight, and obese adults (BMI, 28 – 35 kg/m 2 ) aged 35 – 65 y. The main exclusion criteria were as follows: signi fi cant acute or chronic disease; smoking; a history of drug and/or alcohol abuse (more than 2 servings/d), pregnancy; antibiotic use within the previous 3 mo; major dietary changes in the past 3 mo; eating disorders; vege- tarians or vegans; enemas and dietary supplements including prebiotics, probiotics, or fi ber within 4 wk before the baseline visit and for the duration of the intervention; chronic medica- tions for active gastrointestinal disorders (unless the product was taken for at least 2 month before screening and the exact dosage was maintained throughout the study); and high habitual intake of tomatoes or tomato-based products as con fi rmed by a Food Frequency Questionnaire ( > 1000g per wk). Participants deemed eligible were randomly assigned to an intervention order on a 1:1 basis, where n ¼ 20 participants were assigned to the pla- cebo/Fruit fl ow arm (group 1), and n ¼ 20 participants to the Fruit fl ow/placebo arm (group 2). Study design The trial was designed as a randomized, double-blind, placebo- controlled, cross-over trial consisting of 5 visits: 1) the screening visit (visit 1); 2) after a 21-d run-in period, the start of intervention phase 1 (visit 2); 3) end of the 4-week intervention (visit 3); 4) 6- week wash-out, and the start of intervention phase 2 (visit 4); and 5) end of the 4-week intervention (visit 5) (Supplemental Figure 1). During the screening, vital signs were recorded, and a complete medical examination, including medical history and demographic or anthropometric assessment, was performed. In addition, weekly tomato consumption was queried, and a fasting venous blood sample was collected for safety pro fi ling. Partici- pants were provided with a stool collection kit including cooler bag and cooler block for transporting, instructions for collecting and storing the stool sample, and a Bristol Stool Chart [25] to be completed at the time of stool collection. At visit 2, participants arrived at the study site and fasted overnight for at least 10 hours. Participants returned the Bristol Stool Chart and stool sample collected and stored at 20 C in home freezers within 24 hours before the visit. A venous blood sample was collected for safety pro fi ling and analysis of plasma TMAO and lipopolysaccharides (LPS). In addition, a urine sample was collected for the analysis of urine TMAO, and participants completed a Gastrointestinal Symptom Rating Scale (GSRS) and Food Frequency Questionnaire before they were assigned into one of the 2 intervention groups.
2
A. Rehman et al.
The Journal of Nutrition xxx (xxxx) xxx
TMAO and LPS samples were sent to MS-Omics ApS (Bygstubben 9, Vedbaek, Denmark) for analysis using a Ultra Performance Liquid Chromatography (UPLC) system (Vanquish, Thermo Fisher Scienti fi c) coupled with a high resolution quadrupole-orbitrap mass spectrometer (Q Exactive HF Hybrid Quadrupole-Orbitrap, Thermo Fisher Scienti fi c). Plasma BA were extracted from plasma and quanti fi edusinga commercially available BA assay (Biocrates Life Sciences AG). SCFAs and other organic acids were fi rst extracted from the plasma by protein precipitation. Then plasma extracts were derivatized and reaction products extracted by liquid-liquid extraction using dichloromethane. Obtained extracts were fi nally injected into a Ultra High Performance Liquid Chroma- tography (UHPLC)-MS/MS system for analysis in combined positive and ESI MRM mode. For untargeted metabolomics, plasma samples were prepared using a previously described method with minor adaptations [26]. In short, 100 μ L of human plasma was mixed with 100 μ Lof a deuterated phosphate buffer pH ¼ 7.3, prepared by adding a Phosphate Buffered Saline (PBS) tablet (OXOID) in 100 mL of D 2 O. Maleic acid (0.5mM) was used as an internal standard, and the samples were subsequently put in 3-mm Nuclear Magnetic Resonance (NMR) tubes. All spectra were acquired at 298 K on a Bruker Avance III NMR spectrometer operating at 600 MHz proton Larmor frequency and equipped with a 5 mm triple resonance inverse cryoprobe. 1D 1 H NMR spectra were acquired using a cpmgpr1d pulse sequence, relaxation and D1 and D20 delays of 5 and 0.0003 seconds respectively; 256 scans were accumulated in 36.5 minutes per spectrum. Fecal samples Fecal samples were collected in DNA/RNA Shield and fecal collection tubes (Zymo Research) and delivered on dry ice in 80 º C compatible boxes to BaseClear BV for microbiome pro fi ling. In brief, nucleic acid was extracted from fecal samples using the ZymoBIOMICS DNA Miniprep (Zymo Research Corp.) kit per manufacturer s instruction. 16S rRNA gene variable region V3- V4 was ampli fi ed by composite primer 341F (5 ’ - CCTACGGGNGGCWGCAG-3') and 785R (5 ’ -GACTACHVGGG- TATCTAATCC – 3') and sequenced using the Illumina MiSeq sequencing platform to generate paired-end sequence reads. Subsequently, reads containing the PhiX control signal were removed by aligning the sequence reads against the PhiX genome (NC_001422.1) with Bowtie2 (2.2.6) [27] and removed from further analysis. In addition, reads containing (partial) adapters were clipped (up to a minimum read length of 50 bp). The second quality assessment was based on the remaining reads using the FASTQC quality control tool version 0.11.5. Paired-end sequence reads were collapsed into the so-called pseudoreads using sequence overlap with USEARCH version 9.2 [28]. Classi fi cation of these pseudoreads is performed based on the results of align- ment with SNAP version 1.0.23 [29] against the RDP database [30] for bacterial organisms. Alpha diversity indices (observed species, Chao1, Shannon and Simpson diversity indices) and beta diversity indices (Jaccard and Bray – Curtis) were calculated by implementing mothur version 1.35.1 [31]. Stool consistency and gastrointestinal symptoms Stool consistency was assessed using the Bristol Stool Chart: Type 1 (separate hard lumps, like nuts), Type 2 (sausage-shaped
Participants were provided with another stool collection kit as well as Bristol Stool Chart and instructed to collect a stool sample at home, within 24 hours of their next scheduled visit at week 4. At visit 3, participants again arrived at the study site, fasted overnight, and returned the stool samples and Bristol Stool Chart. Another blood and urine sample was collected, and a Gastroin- testinal Symptom Rating Scale was completed. In addition, par- ticipants returned any unused study product to assess compliance with the intervention phase 1. Participants were then sent home for the 6-week wash-out before entering the intervention phase 2, following the same experimental setup as visits 2 and 3. Subgroup egg challenge To determine the effect of the intervention on postprandial plasma TMAO concentrations, a subgroup of 9 participants completed an acute egg challenge at visits 3 and 5, respectively. For this, participants were instructed to follow a standardized menu, designed to deliver a low-choline diet for the 24-hours before the study visits. On the day of the experiments, partici- pants arrived in the morning and fasted overnight, and a urine sample was collected. Subsequently, an intravenous cannula was placed in a forearm vein, and a baseline blood sample (T0) was collected. Participants then consumed the egg breakfast within 10 minutes, and additional blood samples (6 mL) were drawn for the measurement of plasma TMAO at 120, 240, 360, and 480 minutes postprandially. In addition, all urine samples were collected over an 8-hour period and analyzed for TMAO. Par- ticipants were required to stay at the study site during this time and were provided with a standardized lunch and snacks. The egg breakfast consisted of a high-yolk scrambled egg recipe using pasteurized liquid egg yolks and egg whites to control the vol- ume of egg yolk. The total choline content of the egg breakfast was approximately 450 mg. The provided lunch and snacks were of fi xed calorie content and low in total choline. Participants were advised to fi nish each provided meal. Ethics statement The trial was conducted by Atlantia Food Clinical Trials Ltd., in Cork, Ireland in accordance with the Declaration of Helsinki and approved by the Clinical Research Ethics Committee of the Cork Teaching Hospitals. Written informed consent for partici- pation was obtained from all participants. The trial was regis- tered with clinicaltrials.gov (NCT04160481).
Measurements Blood and urine samples
Blood samples were collected using EDTA plasma tube for safety pro fi ling (hematology, chemistry, glucose, and bilirubin) and analysis of TMAO, LPSs, BAs, SCFAs, and other organic acids as well as untargeted metabolomics. In addition, a serum sample was collected for analysing the levels of high-sensitivity CRP (hsCRP) and urine samples for the analysis of urine TMAO. Safety pro fi ling, including hematology, chemistry, and glucose, was performed by standard clinical laboratory methods in Euro fi ns Biomnis. Bilirubin and hsCRP samples were sent to Euro fi ns Biomnis for analysis as per Biomins standard proced- ures. Bilirubin was assessed using a diazo reaction assessment to determine the total bilirubin and direct bilirubin values, and hsCRP was assessed via immunoturbidimetric determination using a Multigent CRP Vario assay.
3
A. Rehman et al.
The Journal of Nutrition xxx (xxxx) xxx
resulting bins were analyzed using MetaboAnalyst v4.0 [36] after being divided into 2 groups, i.e., subjects who had received Fruit fl ow (group 1, visit 5; group 2, visit 3) and subjects who had not (group 1, visit 2, 3, and 4; group 2 visit 2, 4, and 5). All data were normalized (quantile normalization) and auto-scaled (Pareto scaling) before univariate (t-test) and multivariate analysis (PLS-DA). The Q2 and R2 values for Partial Least Squares Discriminant Analysis (PLS-DA), as well as the Variable Importance in Projection (VIP) plots, were determined by MetaboAnalyst following an established work fl ow[37]. A paired Wilcoxon test was used for microbial alpha diversity to determine the statistical signi fi cance between and within groups. For beta diversity, Permutational Multivariate Analysis of Variance (PERMANOVA) was used to quantify differences within and between groups (Fruit fl ow compared with placebo) using 9999 permutations. This test evaluates whether groups or time points signi fi cantly affect the overall gut microbiota composition and structure. Within and between-group changes in the relative abundance of microbial taxa (i.e., the difference in outcome between the Fruit fl ow and placebo group) was analyzed using paired Wilcoxon test followed by multiple testing correction with Benjamini-Hochberg. In addition, for between- group changes, raw abundance counts were Centroid Log Ratio normalized to ensure that the data is scale-invariant and sub- compositionally coherent. Finally, the effect size differences for each taxon were calculated using the cohens_d function (https://easystats.github.io/effectsize/), an R-based package.
but lumpy), Type 3 (like a sausage but with cracks on its surface), Type 4 (like a sausage or snake, smooth and soft), Type 5 (soft blobs with clear-cut edges), Type 6 ( fl uffy pieces with ragged edges, a mushy stool), Type 7 (watery, no solid pieces, entirely liquid) [25]. Gastrointestinal symptoms were assessed by a 6-point scale using the GSRS utilizing a 7-point rating scale, depending on the intensity and frequency of gastrointestinal symptoms experienced during the previous wk. A high score indicated more inconvenient symptoms. Investigational product Fruit fl ow was commercially produced by DSM Nutritional Products, Basel, Switzerland, in powder format. The composition of Fruit fl ow has been described previously [32] with a standard 150 mg dose delivering up to 9 mg nucleoside derivatives, up to 10 mg simple phenolic conjugates (e.g., chlorogenic acid, other caffeic or phenolic acid derivatives), and up to 7 mg fl avonoid derivatives, of which at least 2.4 mg are quercetin derivatives. The ef fi cacious range for Fruit fl ow to affect platelet aggregation and thrombin generation capacity lies between 75 and 300 mg [33]. Maltodextrin was used as a placebo control (Essential Nutrition Ltd., Brough, UK). All supplements were encapsulated using size 00 Vegecaps (LGA, La Seyne-sur-Mer, France), and the fi nal weight of each capsule was 500 mg (weight of Fruit fl ow plus weight of tapioca starch fi ller). Participants were instructed to consume 2 capsules orally (either placebo or Fruit fl ow) in the morning at breakfast along with a glass of water resulting in a total of 300 mg Fruit fl ow per d for a 28-d intervention period. Sample size calculation and statistical analysis The sample size was determined based on the fi ndings from previous studies with nutritional interventions to reduce plasma TMAO [9, 34, 35]. For a power of 80%, the signi fi cance level of 5%, and an expected effect size [mean (SD)] of 0.5 for Fruit- fl ow compared with placebo, it was calculated that 34 partici- pants were required. To account for potential losses to follow-up, 40 participants were enrolled. The effectiveness of the wash-out was assessed using paired samples t-tests comparing the baseline of phase 1 (visit 2) with the baseline of phase 2 (visit 4). There were no statistically sig- ni fi cant differences between any parameters at phase 1 baseline and phase 2 baseline for the Fruit fl ow or placebo group. As a result, the Fruit fl ow group is a combination of group 2 – phase1 data (visit 2-visit 3) and group 1 – phase 2 data (visit 4-visit 5), whereas the placebo group is a combination of group 1 – phase1 data (visit 2-visit 3) and group 2 – phase 2 data (visit 4-visit 5). SPSS IBM V26.0 and R software were used to analyze all data. For the ef fi cacy analysis, paired t-tests (or Wilcoxon signed rank tests) were used to determine 1) whether there was a statistically signi fi cant within-group change from baseline to end of the intervention in either the Fruit fl ow group or the placebo group or 2) whether there was a statistically signi fi cant between-group difference (i.e., differences in changes between groups). All data are reported as means SE. All tests were 2-tailed, and differ- ences were considered statistically signi fi cant at P 0.05. For untargeted metabolomics, all spectra were collected and processed using ACD Labs 2012. All spectral regions containing NMR features were group processed by Intelligent Bucketing (Bucket width ¼ 0.02 ppm, width looseness ¼ 50%). The
Results
Study population The study population consisted of 40 overweight and obese adults. Three participants self-withdrew after randomization whereas the remaining 37 completed the study. An independent committee subsequently evaluated all participants case by case before unblinding the data to determine inclusion in the per- protocol analysis. Fifteen subjects were excluded from the analysis becauseofusingconcomitantmedicationssuchasantibiotics( n ¼ 9) and investigational product (IP) compliance below 80% and/or missing critical variables (because of missing blood and stool sam- ples, n ¼ 6);therefore,theper-protocoldataanalysiswasbasedon22 subjects ( see Supplemental Figure 2 for the CONSORT (Consoli- dated Standards of Reporting Trials) fl ow diagram and Supple- mentalTable1forsubjectcharacteristicsofthe fi nal22participants)
TMAO
Fasting plasma and urine TMAO. Fruit fl ow reduced fasting urine, but not fasting plasma, TMAO concentrations when compared with placebo ( P 0.05, Figure 1A, B). When comparing baseline to end of the intervention time points within each group, Fruit- fl ow, but not placebo, signi fi cantly reduced both fasting plasma ( 1.51 μ M, P 0.05) and fasting urine TMAO concentrations ( 19.09 μ M, P 0.01, Figure 1A and B). Postprandial plasma and urine TMAO during the egg challenge. The egg challenge was performed only in a subgroup of participants either at visit 3 or 5 to compare postprandial changes in plasma TMAO between groups. Overall, there was a large variation
4
A. Rehman et al.
The Journal of Nutrition xxx (xxxx) xxx
Microbial composition. When assessing between-group differences in species-level OTUs, we found that changes in relative abun- dances of 17 OTUs signi fi cantly differed between groups, of which 9 OTUs related to Ruminococcus albus _OTU#1468 ( P ¼ 0.002), Alistipes ihumii_O TU#528 ( P ¼ 0.002), Anaeromassilibacillus sene- galensis _OTU#1422 ( P ¼ 0.017), Saccharofermentans acetigene- s _OTU#1479 ( P ¼ 0.017), Oribacterium sinus _OTU#1283 ( P ¼ 0.027), Clostridium carnis _OTU#994( P ¼ 0.035), Ruthenibacterium lactatiformans _OTU#1478 ( P ¼ 0.039), Alistipes obesi _OTU#532( P ¼ 0.046), and Faecalitalea cylindroides _OTU#1564 ( P ¼ 0.046) were elevated signi fi cantly in the Fruit fl ow group when compared with placebo. In addition, 4 OTUs related to Parabacteroides gold- steinii _OTU#448( P ¼ 0.025), Bacteroides acidifaciens _OTU#364( P ¼ 0.027), Veillonella parvula _OTU#1638 ( P ¼ 0.030), and Rumi- nococcus faecis _OTU#1473 ( P ¼ 0.039) were lower in the Fruit fl ow group when compared to changes in the placebo group (Figure 3A). There were also signi fi cant between-group differ- ences at genus level with decreases in the relative abundance of Prevotella ( P ¼ 0.007)and Veillonella ( P ¼ 0.023) and increases in 5 taxa related to Saccharofermentans ( P ¼ 0.01), Alistipes ( P ¼ 0.023), Anaeromassilibacillus ( P ¼ 0.033), Ruthenibacterium ( P ¼ 0.039), and Oribacterium ( P ¼ 0.042) in the Fruit fl ow group when compared with the changes evident in the placebo group ( Figure 3B). Within-group changes from baseline to end of the intervention were assessed using ALDEx2. In the Fruit fl ow group, 4 species-level OTUs related to Ruminococcus faecis _OTU#1473 ( P ¼ 0.0005), Bacteroides uniformis _OTU#406 ( P ¼ 0.018), Bacteroides ovatu- s _OTU#392 ( P ¼ 0.025), and Hungatella hathewayi _OTU#1083 ( P ¼ 0.036) had lower relative abundance at the end of intervention than at baseline ( Figure 3C ) . In contrast, in the placebo group, 3 species-level OTUs related to Coprococcus catus _OTU#1222 ( P ¼ 0.023), Anaeromassilibacillus senegalensis _OTU#1422 ( P ¼ 0.028), and Barnesiella intestinihominis _OTU#428( P ¼ 0.043) were signif- icantly depleted at the end of intervention when compared with baseline (Figure 3D ). At the genus level, 3 taxa related to Strepto- coccus ( P ¼ 0.005), Rumniococcus ( P ¼ 0.021), and Hungatella ( P ¼ 0.039) had signi fi cantly lower relative abundance at the end of intervention than that at baseline in the Fruit fl ow group. In contrast, in the placebo group, Anaeromassilibacillus ( P ¼ 0.028), Alistipes ( P ¼ 0.038), and Eubacterium ( P ¼ 0.038) were signi fi - cantly depleted at the end of intervention when compared with baseline ( data not shown ). We did not observe any signi fi cant between or within-group changes at the phylum level ( data not shown ).
between individuals in postprandial TMAO responses (for indi- vidual pro fi les, see Supplemental Figure 3). When calculating mean area under the concentration-time pro fi le (total and in- cremental Area Under Curve) and maximum peak plasma con- centrations (C max ), we did not observe a signi fi cant difference between the groups. In addition, we failed to observe any dif- ference between groups for urine TMAO that was collected over 8 hours postprandially ( data not shown ).
Fecal microbiota
Alpha and beta diversity. No signi fi cant between-group differ- ences or within-group changes were observed for species di- versity and richness using observed species, Chao1, Shannon, and Simpson diversity indices (Supplemental Figure 4). To gain insights into the temporal dynamics of microbial communities, fecal samples were also subjected to a multivariate analysis using Bray – Curtis and Jaccard distance methods. However, no signif- icant shift in beta diversity was observed in global or pairwise PERMANOVA analysis between and within groups. In contrast, a signi fi cant shift in Jaccard Principal Component (PC1) was observed when comparing the end of intervention time points of Fruit fl ow and the placebo group ( P ¼ 0.019, Figure 2).
Plasma metabolites
Plasma untargeted metabolomics. Principal component analysis (PCA) of the NMR data shows a clear distinction between plasma samples collected after Fruit fl ow intervention and control sam- ples (Figure 4A). By means of VIP, the top 15 ranking features driving this distinction included TMAO, formic acid, valine, glucose, and lactate, with TMAO being the most discriminant metabolite (low in Fruit fl ow samples, high in control samples, Figure 4B). Plasma LPS. There was no signi fi cant between-group difference in plasma LPS concentrations. However, we observed a signi fi - cant within-group change with reduced plasma LPS
FIGURE1. Effects of 4-wk of supplementation of Fruit fl ow or placebo (maltodextrin) on trimethylamine-n-oxide (TMAO) in overweight and obese adults. Boxes represent median and interquartile range (IQR), and the diamond shape, symbol mean values of absolute changes in (A) plasma and (B) urine TMAO from baseline to end of the inter- vention in each group. There were within-group changes, i.e. different from baseline at P 0.05 and between-group differences, i.e. differ- ences in changes between Fruit fl ow and placebo group at P 0.05. N ¼ 22.
5
A. Rehman et al.
The Journal of Nutrition xxx (xxxx) xxx
concentrations at the end of the intervention when compared with baseline in the Frui fl owgroup ( 5.3ng/mL, P 0.05) but not in the placebo group (Figure 5). Plasma BAs. There were no signi fi cant between-group differ- ences in plasma BA. Within groups, taurochenodeoxycholic acid (TCDCA) and taurodeoxycholic acid (TDCA) increased slightly with both Fruit fl ow and placebo, from baseline to the end of intervention, whereas chenodeoxycholic acid (CDCA), glyco- cholic acid (GCA), glycochenodeoxycholic acid (GCDCA), and glycodeoxycholic acid (GDCA) only increased in the Fruit fl ow group ( P 0.05, respectively, Supplemental Table 2). Plasma SCFA and other organic acids. There were no signi fi cant between-group differences (Supplemental Table 2). Within- groups, we observed a slight but signi fi cant increase in pyru- vate in the Fruit fl ow group, as well as increases in acetate in both the groups ( P 0.05, respectively). Fecal metabolites Fecal BAs. There were no signi fi cant between-group differences in fecal BA (Supplemental Table 3). We observed only one sig- ni fi cant within-group change; fecal cholic acid (CA) increased from baseline to the end of intervention in the Fruit fl ow group but not in the placebo group ( P 0.05, Supplemental Table 3). Fecal SCFA and other organic acids. There were no between- group differences in fecal SCFAs; only valerate increased from baseline to end of the intervention in the placebo group ( P 0.05, respectively, Supplemental Table 3). Stool consistency and gastrointestinal symptoms There were no signi fi cant between- or within-group differ- ences in stool consistency or gastrointestinal symptoms ( datanot shown ). Adverse events and blood safety pro fi les There were no serious adverse events (SAEs), or withdrawals due to adverse events (AEs) observed during the study. All AEs were of mild or moderate intensity, and none were deemed to be related to the investigational products by the assigned study clinician ( data not shown ). Standard hematology and biochemistry assessment showed no signi fi cant within and between-group ef- fects, including glucose, hsCRP, and bilirubin. Data not shown .
comparing both groups. In addition, there were several signi fi cant changes in relative abundance of microbial taxa with Fruit fl ow, such as decreases in Bacteroides, Ruminococccus, and Hungatella related OTUs, as well as increases in Alistipes which are all known for the involvement in TMA/TMAO metabolism. There were no between-group differences in SCFAs and BAs in both feces and plasma but several signi fi cant changes within groups such as an increase in CA in feces or plasma pyruvate with Fruit fl ow. TMAO has been established as an independent risk factor for promoting atherosclerosis by stimulating foam cell formation, deregulating enterohepatic cholesterol metabolism, and impairing macrophage reverse cholesterol transport [13, 15, 16, 38]. Given that the production of TMAO from dietary choline is dependent on metabolism by the intestinal microbiota [13, 15, 38], gut microbiota-based interventions have been suggested as a novel strategy for preventing and treating cardiovascular diseases (CVD). Several natural products rich in polyphenols have been shown recently in animals and humans to lower plasma TMAO levels [9 – 11]. For example, Chen et al. [10] fi rst demonstrated in mice that resveratrol attenuates TMAO-induced atherosclerosis by regulating TMAO synthesis via remodeling gut microbiota. This was con fi rmed in other studies with experimental rodents [11] as well as in a pilot study in 20 normal-weight subjects showing a signi fi cant decrease in serum TMAO (1.87 to 0.66 μ M) after 4 wk of supplementation with a 300 mg of polyphenol-rich grape pomace [9]. In contrast, a recent study in overweight and obese subjects revealed that raspberry consumption of 280 mg/d increased plasma TMAO levels, although the data revealed a large interin- dividual variability, with 6 subjects showing decreased plasma TMAO levels and 11 subjects with increased TMAO levels [39]. FIGURE2. Effects of 4-wk of supplementation of Fruit fl ow or placebo (maltodextrin) on gut microbiota beta diversity in overweight and obese adults. Principal component analysis (PCoA) is based on A) Bray – Curtis and B) Jaccard distance matrixes of Fruit fl ow and placebo groups at baseline and the end of the intervention. Ellipses represent an 80% con fi dence interval. Lines connect samples from the same participants. Density plots show the projection of PCoA points onto the PC1 and PC2 axis. N ¼ 22.
Discussion
We found that Fruit fl ow when administered over 4 wk at 2 150 mg Fruit fl ow per d, but not placebo, signi fi cantly reduced fasting plasma ( 1.5 μ M) and urine ( 19.1 μ M) TMAO as well as plasma LPS ( 5.3 ng/mL) from baseline to the end of interven- tion. However, these changes were signi fi cant only for urine TMAO when comparing the changes between groups. An untar- geted metabolomic analysis revealed a clear distinction between plasma samples collected after Fruit fl ow interventions and control samples, with TMAO being the top-ranking feature driving this distinction. When analyzing fecal microbiota, we found changes in microbial beta, but not alpha, diversity with a signi fi cant dif- ference in Jaccard distance-based principal component when
6
A. Rehman et al.
The Journal of Nutrition xxx (xxxx) xxx
FIGURE 3. Effects of 4 wk of supplementation of Fruit fl ow or placebo (maltodextrin) on gut microbiota composition in overweight and obese adults. Data represent a volcano plot of differential abundant OTUs between groups at (A) species level and (B) genus level and within groups at species level in (C) the Fruit fl ow and (D) the placebo group. The X-axis position of each point represents effect size differences at the end of the intervention. The horizontal line represents the unadjusted p-value cut-off at 0.05. Points above the effect size cut-off (0.20) and p-value cut-off are color-coded and annotated with text. Black dots represent non-signi fi cant taxa; green dots represent taxa above the effect size cut-off; blue dots represent taxa above p-value cut-off; Red dots represent taxa above effect size and p-value cut-off. N ¼ 22.
FIGURE 4. Effects of 4 wk of supplementation of Fruit fl ow or placebo (maltodextrin) on metabolomic pro fi ling in overweight and obese adults. Untargeted metabolomics showing (A) PLS-DA 2D plot of the multivariate analysis of plasma samples collected after Fruit fl ow interventions and control samples (R2 ¼ 0.432, Q2 ¼ -0.766) and (B) the 15 most bins driving the discrimination between Fruit fl ow and control samples.
7
A. Rehman et al.
The Journal of Nutrition xxx (xxxx) xxx
the Frui fl ow group and Bacteroides acidifaciens and related spe- cies Parabacteroides goldsteinii whenFrui fl ow was compared with placebo. Cruden and Galask et al . [42] identi fi ed that Bacteroides sp. could reduce choline to TMA [43]. Moreover, Bacteroides sp. has been suggested to reduce TMAO to TMA [44]. Thus, it is reasonable to hypothesize that a reduction in Bacteroidetes via Fruit fl ow is one mechanism explaining the observed effects on TMAO. We also observed a reduction in Hungatella hathewayi and Rumonococcus faecis within the Frui fl ow group. Hungatella hathewayi is known to be a TMA producer [45], whereas Rumo- nococcus was positively correlated with plasma TMAO in a study in rats [46]. Moreover, we found an increase in Alistipes with Fruit fl ow when comparing between groups , another member of the Bacteroidetes phylum. This is in line with a recent clinical trial investigating the effect of short period lifestyle changes on car- diovascular disease markers and gut microbiome in which TMAO levels were inversely correlated with the relative abun- dances of Alistipes [47]. Finally, we observed an increase in Clostridium carnis when Frui fl ow was compared with placebo. Although the Clostridium group is generally known to produce TMA, which would contradict our observations, it is thought to be related to C. asparagiforme , C. hathewayi , and C. sporogenes [48, 49]. On the other hand, the Clostridium group is also known to produce large amounts of SCFA from several nutrients and therefore often cited for its health bene fi ts [50], although we did not fi nd an increase in fecal SCFA in our study. Other studies have also reported a broad range of antimicrobial mechanisms by which polyphenols modulate gut microbial communities as well as prebiotic-like properties favoring the growth of key bene fi cial species such as A. muciniphila, B. thetaiotaomicrom, F. prausnitzii, Bi fi dobacteria, and Lactobacilli [51]. However, under our experimental conditions, we did not observe similar effects. As mentioned above, TMAO may promote atherosclerosis by several mechanisms, including inhibiting hepatic BA synthesis [13, 15, 16, 38]. In the study in mice by Chen et al. [10], resveratrol attenuated TMAO-induced atherosclerosis not only by inhibiting microbial TMA production but also by increasing hepatic BA syn- thesis. This was thought to be due to resveratrol that increased the relative abundance of Lactobacillus and Bi fi dobacterium , which resulted in an increased bile salt hydrolase activity, BA deconju- gation, and subsequently increased fecal excretion of bile. The decrease in ileal BA content and repression of the enterohepatic farnesoid X receptor (FXR)- fi broblast growth factor 15 (FGF15) axis increased hepatic BA synthesis. We found that plasma CDCA, GCA, GCDCA, and GDCA rose signi fi cantly in the Fruit fl owgroup but not in the placebo group which may suggest an increase in hepatic BA synthesis. However, we did not observe changes in Lactobacillus or Bi fi dobacterium or reduced fecal BA contents. Thus, Fruit fl ow or its metabolites may alter FXR receptor activity and plasma BA pro fi les through other mechanisms as has been reported for metformin or acarbose, and this may further depend on gut microbiota compositions before treatments [52, 53]. Clearly further data are required for a better understanding of these effects. There was no signi fi cant difference in plasma LPS when comparing changes in both groups; however, we observed a signi fi cant within-group decrease in plasma LPS with Fruit fl ow but not placebo. LPS is a gut microbiota-derived factor and has been suggested to drive onset and progression of chronic in fl ammation-related diseases such as obesity, T2DM, or non- alcoholic fatty liver disease (NAFLD) [14, 54 – 56] Normally,
Our fi ndings are in line with the studies showing an overall reduction in plasma TMAO [10, 11, 19]. The mean decrease of 1.51 μ M in plasma TMAO within the Fruit fl ow group is similar to what was observed in the study by Annunziata et al. [9] (mean reduction of 1.2 μ M) and suggests reproducible effects with polyphenol-rich extracts. Although the reduction in our study was not signi fi cantly different between groups, most likely because of a lack of power in the per-protocol analysis, the observed effects size may be clinically relevant. A recent cross-sectional study including 377 patients with acute ischemic stroke and 50 healthy controls found that plasma TMAO levels were higher in patients with ischemic stroke than in healthy controls (median 5.1 vs. 3.0 μ mol/L; P < 0.001), and every 1- μ mol/L increase in TMAO was associated with a 1.13-point in- crease in the National Institutes of Health Stroke Scale and 1.69-mL increase in infarct volume after adjustment for vascular risk factors. We also investigated postprandial changes in plasma TMAO in a subgroup of participants that completed an acute egg chal- lenge. Fasting plasma TMAO levels were recently suggested to underestimate the TMAO-producing function of gut microbiota because of host ’ sef fi cient renal clearance of TMAO [40, 41]. As a result, oral challenge tests using egg or carnitine have been used to investigate postprandial increases in plasma TMAO [40, 41]. However, we did not observe an effect of Fruit fl ow on post- prandial plasma TMAO concentrations which may be related to the large variation in postprandial responses between in- dividuals, limited power ( n ¼ 9 subjects) and sampling time of only 8 hours which in turn requires independent research. Our observations support the notion that changes in plasma and urine TMAO are dependent on changes in the gut micro- biota. We found signi fi cant changes in several microbial taxa, although these were not re fl ected in microbial alpha and global beta diversity indices but a signi fi cant shift in Jaccard PC1 indicating that Fruit fl ow consumption can affect gut microbial composition. One strategy to alter plasma TMAO is to selectively decrease the proportion of bacteria that produce TMA and/or increase the proportion of microbes metabolizing TMA ef fi - ciently. We observed a signi fi cant decrease in the relative abundance of Bacteroides uniformis and Bacteroides ovatus within FIGURE5. Effects of 4 wk of supplementation of Fruit fl ow or placebo (maltodextrin) on plasma lipopolysaccharides (LPS) in overweight and obese adults. Boxes represent median and interquartile range (IQR), and the diamond shape, symbol mean values of absolute changes in plasma LPS from baseline to end of the intervention in each group. There were within-group changes, i.e. different from baseline at P 0.05. N ¼ 22.
8
A. Rehman et al.
The Journal of Nutrition xxx (xxxx) xxx
employees of Atlantia Clinical Trials Ltd., which was contracted by DSM to carry out the clinical trial. AKDR is a member of the Scienti fi c Advisory Board of Provexis PLC, UK
LPS concentrations are high in the gut lumen where gut bacteria reside and low in plasma because the intestinal epithelial layer creates an effective barrier against LPS penetration. The observed reduction in plasma LPS after Fruit fl ow consumption, therefore, suggests improvements in intestinal barrier function, although we did not directly measure this. There are a few limitations of this study that require consid- eration. 1) The small sample size allows only limited interpreta- tion of data. Although we powered the study based on available effect size data, many participants had to be excluded from the analysis which resulted in only 22 participants for the per- protocol analysis. 2) Given the limited power and the explor- atory nature of the analysis of secondary outcomes, we did not control for multiplicity of testing. 3) Diet is a key modulator of the gut microbiota. Although this trial followed a cross-over design, there was no dietary monitoring during the intervention and therefore changes in habitual diet over time may have in fl uenced the outcome. In addition, we did not analyze dietary choline as a potential confounder, nor did we control for foods like marine fi sh which can contain variable levels of TMA and TMAO [43, 57, 58]. In fact, diets rich in long-chain omega 3 fatty acids of marine origin have been shown to increase plasma TMAO concentration suggesting TMAO levels as a biomarker of these healthy foods but not a universally valid biomarker of cardiometabolic risk [59]. In conclusion, 4 wk of supplementation with 2 150 mg Fruit fl ow per day, a watery tomato extract rich in secondary metabolites including polyphenols, signi fi cantly reduced fasting plasma and urine TMAO as well as plasma LPS from baseline to the end of intervention. However, these changes were signi fi cant only for urine TMAO when comparing the changes between groups. Given that both, plasma TMAO and LPS are suggested as biomarkers of cardiometabolic risk and intestinal permeability and low-grade in fl ammation, respectively, these effects can be considered host health bene fi ts. Changes in microbial beta, but not alpha, diversity paralleled these effects with a signi fi cant difference in Jaccard distance-based Principal Component be- tween groups as well as changes in microbial composition with Fruit fl ow, such as decreases in Bacteroides, Ruminococccus, and Hungatella and increases in Alistipes related OTUs which are all known for the involvement in TMA/TMAO metabolism [43 – 47]. There were no between-group differences in short-chain fatty acids and BA in both faces and plasma but several signi fi cant changes within groups such as an increase in CA in faces or plasma pyruvate with Fruit fl ow. In addition, an untargeted metabolomic analysis revealed TMAO as the most discriminant plasma metabolite between Fruit fl ow and placebo plasma sam- ples. Together, our results support earlier fi ndings that polyphenol-rich extracts can lower TMAO concentrations and that this may be related to a targeted modulation of the gut microbiota along the de fi nition of the prebiotic concept. Funding
Acknowledgments
We thank Julia Bird for assisting in clinical trial management, Estel Canet-Martinez and Stephane Etheve for coordinating an- alytics, Alex Karagiannis-Voules for statistical support, Niamh O ’ Kennedy at Provexis Plc, London, UK for coordinating the production of the material and the staff of Atlantia Clinical Trials Ltd. Atlantia for their work in the completion of this clinical trial. RES, SF, RD conceived and designed the experiments, GDG and SP completed the statistical analysis and RES, AR, SMT, and GDG wrote the manuscript. AKDR reviewed the manuscript. All au- thors helped interpreting the data, read and approved the fi nal manuscript. All data described in the manuscript, code book, and analytic code will be made available upon request. Appendix A. Supplementary data
Supplementary data to this article can be found online at http s://doi.org/10.1016/j.tjnut.2022.11.009.
References
[1] Y. Fan, O. Pedersen, Gut microbiota in human metabolic health and disease, Nat Rev Microbiol 19 (2021) 55 – 71. [2] Z.Y. Kho, S.K. Lal, The human gut microbiome – a potential controller of wellness and disease, Internet, Front Microbiol (2018) 9. [3] A.M. Valdes, J. Walter, E. Segal, T.D. Spector, Role of the gut microbiota in nutrition and health, BMJ 361 (2018) k2179. [4] R.W. Hutkins, J.A. Krumbeck, L.B. Bindels, P.D. Cani, G. Fahey, Y.J. Goh, et al., Prebiotics: why de fi nitions matter, Curr Opin Biotechnol 37 (2016) 1 – 7. [5] L.B. Bindels, N.M. Delzenne, P.D. Cani, J. Walter, Towards a more comprehensive concept for prebiotics, Nat Rev Gastroenterol Hepatol 12 (2015) 303 – 310. [6] G.R. Gibson, R. Hutkins, M.E. Sanders, S.L. Prescott, R.A. Reimer, S.J. Salminen, et al., Expert consensus document: The International Scienti fi c Association for Probiotics and Prebiotics (ISAPP) consensus statement on the de fi nition and scope of prebiotics, Nat Rev Gastroenterol Hepatol 14 (2017) 491 – 502. [7] M.N. Clifford, Diet-derived phenols in plasma and tissues and their implications for health, Planta Med 70 (2004) 1103 – 1114. [8] M. Due ~ nas, I.Mu ~ noz-Gonz alez, C. Cueva, A. Jim enez-Gir on, F. S anchez- Pat an, C. Santos-Buelga, et al., A survey of modulation of gut microbiota by dietary polyphenols, BioMed Res Int 2015 (2015), 850902. [9] G. Annunziata, M. Maisto, C. Schisano, R. Ciampaglia, V. Narciso, G. Tenore, et al., Effects of grape pomace polyphenolic extract (Taurisolo ® ) in reducing TMAO serum levels in humans: preliminary results from a randomized, placebo-controlled, cross-over study, 11, Nutrients Multidisciplinary Digital Publishing Institute, 2019, 139 – 139. [10] M. Chen, L. Yi, Y. Zhang, X. Zhou, L. Ran, J. Yang, et al., Resveratrol attenuates trimethylamine-n-oxide (TMAO)-induced atherosclerosis by regulating TMAO synthesis and bile acid metabolism via remodeling of the gut microbiota, mBio American Society of Microbiology 7 (2016) e02210 – e2215. [11] T. Lim, J. Ryu, K. Lee, S.Y. Park, K.T. Hwang, Protective effects of black raspberry (Rubus occidentalis) extract against hypercholesterolemia and hepatic in fl ammation in rats fed high-fat and high-choline diets, Nutrients 12 (2020) E2448. [12] C. Sim o, V. García-Ca ~ nas, Dietary bioactive ingredients to modulate the gut microbiota-derived metabolite TMAO. New opportunities for functional food development, Food Funct Royal Society of Chemistr 11 (2020) 6745 – 6776. [13] Z. Wang, E. Klipfell, B.J. Bennett, R. Koeth, B.S. Levison, B. DuGar, et al., Gut fl ora metabolism of phosphatidylcholine promotes cardiovascular disease, Nature 472 (2011) 57 – 63. [14] J.-H. Yuan, Q.-S. Xie, G.-C. Chen, C.-L. Huang, T. Yu, Q.-K. Chen, et al., Impaired intestinal barrier function in type 2 diabetic patients
DSM Nutritional Products funded the study.
Author disclosures
RES, RD, AR, and SF are/were employees of DSM Nutritional Products Ltd., Kaiseraugst, Switzerland, which is involved in the commercialization of Fruit fl ow. SMT, GDG, and TGD are
9
Page 1 Page 2 Page 3 Page 4 Page 5 Page 6 Page 7 Page 8 Page 9 Page 10Powered by FlippingBook