Source: OREGON STATE UNIVERSITY submitted to NRP
DISCOVERY OF BIOLOGICAL SIGNATURES FOR CRUCIFEROUS VEGETABLE INTAKE: INTEGRATION OF THE BROCCOLI- AND HOST-DERIVED METABOLOME AND THE MICROBIOME
Sponsoring Institution
National Institute of Food and Agriculture
Project Status
ACTIVE
Funding Source
AFRI COMPETITIVE GRANT
Reporting Frequency
Annual
Accession No.
1021481
Grant No.
2020-67001-31214
Cumulative Award Amt.
$2,293,793.00
Proposal No.
2019-07888
Multistate No.
(N/A)
Project Start Date
Apr 15, 2020
Project End Date
Apr 14, 2026
Grant Year
2020
Program Code
[A1342]- Food Specific Molecular Profiles and Biomarkers of Food and Nutrient Intake, and Dietary Exposure
Recipient Organization
OREGON STATE UNIVERSITY
(N/A)
CORVALLIS,OR 97331
Performing Department
School of Bio & Pop Health Sci
Non Technical Summary
Cancer is the 2nd-leading cause of death in the US, and dietary factors are important cancer risk factors. Cruciferous vegetables (crucifers) and their constituent biologically active food components, including indoles and isothiocyanates (ITCs) such as sulforaphane (SFN), appear to modulate chronic disease states including cancer. Overall there is strong support for a preventive and/or mitigating impact of crucifer-derived ITCs and indoles on cancer in animal models, but observational data to date in humans are inconsistent, and the impacts of cruciferous vegetable consumption on breast and prostate cancer risk remain controversial, as results and conclusions from various studies differ. Such discrepancies arise in part owing to methodological limitations of accurately assessing dietary exposures, which thus constitute a major barrier in the field. Despite widespread use, classical dietary-intake instruments including food frequency questionnaires (FFQ) and other dietary recall methods are subject to well-known limitations. This is coupled with inadequate understanding of the bioavailability and active metabolites of dietary compounds. To circumvent and address these issues, based on our promising preliminary findings and novel technologies, we seek to identify unique and quantifiable biological signatures of consumption of specific foods. Food type-specific candidate phytochemicals that can be measured in plasma or urine as biomarkers of intake have been identified (e.g. alkylresorcinol levels for whole grains), but key barriers persist to identifying food-specific candidate single biomarkers, such as: 1) rarely are such single compounds exclusively specific for a single food; and 2) many such phytochemicals are short-lived and rapidly cleared from the body. For example, the bioactive ITC SFN, derived from the cruciferous vegetable broccoli, is cleared within 12-24 hr from the circulation. Thus, depending on the timing of sample collection, measures of SFN may or may not reflect habitual broccoli intake, making it a problematic biomarker of intake. This example demonstrates that novel approaches are critically needed to characterize individuals' intakes of specific foods to better understand the association of foods with disease risk and identify at-risk populations. To address this need, we will use a metabolomics-based approach, which can identify combinations of food-derived metabolites capable of providing more accurate, food-specific measures of consumption, and combine it with data-driven prediction models and novel machine-learning algorithms. Advantages of this data-driven approach in biological fluids include a) removing recall bias; b) accounting for individual and food composition variables that impact bioavailability, c) accounting for possible food interactions, and d) accounting for metabolism differences, including microbial breakdown products; and e) providing a more precise and accurate assessment of food intake. The overall goal of the proposed study is to identify reliable and accurate molecular biological signatures of cruciferous vegetable intake. Our central hypothesis is that by combining metabolomics with data- and machine learning-based prediction models, we can identify metabolic signatures that estimate intake of broccoli and relate their unbiased measure of dietary intake more meaningfully than do food intake questionnaires. We will utilize deuterium-labeled broccoli sprouts in combination with mass spectrometry (MS)-based metabolomics approaches to quantify and differentiate between broccoli-specific metabolites and their interactions with their molecular targets of action. To demonstrate the feasibility of our approach, we will present data showing that 1) our labelled broccoli sprouts contain deuterium-labeled glucosinolates (precursors of ITCs) and other labeled phytochemicals representing major biosynthetic pathways that are stable, 2) we have identified microbial-derived metabolites in human stool samples following incubation of broccoli digests and identified microbial derived ITCs (SFN-nitrile and erucin-nitrile) in plasma from humans who consumed broccoli sprouts, 3) that these microbial metabolites are indicative of broccoli intake as late as 12-48 h post-consumption and 4) novel machine learning approaches offer advantages over traditional multivariate methods.
Animal Health Component
20%
Research Effort Categories
Basic
80%
Applied
20%
Developmental
(N/A)
Classification

Knowledge Area (KA)Subject of Investigation (SOI)Field of Science (FOS)Percent
70224991020100%
Knowledge Area
702 - Requirements and Function of Nutrients and Other Food Components;

Subject Of Investigation
2499 - Plant research, general;

Field Of Science
1020 - Physiology;
Keywords
Goals / Objectives
Our goals are:Identify signatures of broccoli intake that quantify and discriminate between broccoli-derived vs "metabolic impacts" on host metabolome present in urine and plasma, in a controlled feeding trial. We will utilize deuterium-labeled and unlabeled broccoli sprouts to directly identify plant-specific metabolomic signatures (Brexposome) versus host metabolome in a controlled human trial that combines consumption of deuterium-labeled and unlabeled sprouts with analyses by mass spectrometry (MS)-based metabolomics.Identify microbial-mediated metabolites and associated microbiota following broccoli consumption. We will identify gut microbiota-derived metabolites of broccoli consumption using a human fecal incubation system. Second, in the controlled feeding trial, we will examine the microbiome pre- and post- broccoli consumption and evaluate microbial metabolites. By integrating multi-omics data, specifically 16S rRNA gene sequences and metabolomic measurements, we will predict which gut microbial taxa are responsible for which biotransformations of broccoli compounds and broccoli metabolites.Develop a model for predicting broccoli intake and the effects of Brexposome metabolites on human host metabolism by using machine learning methods integrating the Brexposome, the gut microbiome and the host metabolome. Our working hypothesis is that specific metabolomic signatures associate with the gut microbiome and the host metabolome. To test this hypothesis, we will use machine learning to estimate broccoli sprout consumption and also correlate the signature-based estimates with data obtained from dietary questionnaires. The development of a conceptual model for defining the impact food-derived phytochemicals have on the host metabolome has the potential to dramatically alter the field of nutrition and be broadly applied to the assessment of many other dietary exposures and their host effects.
Project Methods
Deuterium labelling of sprouts: Broccoli and alfalfa sprouts will be grown hydroponically in automatic sprouting chambers (EasyGreen Sprouting Co) in the Moore Family Center for Whole Grain Foods, Nutrition and Preventive Health, under consultation with Dr. James Meyers (Horticulture, Oregon State University) using a nutrient solution enriched with 25% D2O as described in97. Additional batches of broccoli and alfalfa sprouts will be grown without deuterium as unlabeled controls. Sprouts will be harvested as 5-day old sprouts, and glucoraphanin and glucobrassicin content will be measured by LC-QToF MS in ToF mode. We will combine the chromatographic peak areas representing all isotopologues of glucoraphanin and glucobrassicin. The combined peak area will be converted into concentration by using calibration curves for both analytes.Untargeted metabolomics: Plasma and urine samples from human subjects who consumed unlabeled or deuterium-labeled broccoli sprouts will be analyzed by LC-QToF MS in the positive and negative ion mode using our untargeted metabolomics platform50-53. QC standard sample and MS signal correction: We will purchase and analyze the NIST Standard Reference Material (SRM) 1950 Metabolites in Human Plasma98 as a QC sample to monitor analytical variability and to correct for any signal drift across batches of samples. The NIST SRM 1950 sample, prepared in the same way as plasma samples from our feeding trial, will be injected after every fourth sample, which will allow for QC Local Regression (LOESS) signal correction and pooling of data from multiple analytical batches99. Data processing: Raw MS files will be imported and processed by the program MarkerView (Sciex). MarkerView detects spectral features (each defined by a unique chromatographic retention time and accurate mass, aligns and integrates peaks, and performs t-tests to identify differentiating metabolites. We will control the false discovery rate (FDR) by applying the Benjamini-Hochberg algorithm62. In addition, MarkerView creates PCA plots for metabolites detected in both the positive and negative ion mode. All spectral features will be automatically screened against our in-house library of 652 metabolite standards52 using PeakView software (Sciex). Metabolite identities will be established on the basis of accurate mass, MS/MS fragmentation, isotope pattern, and LC retention time. The online databases Metlin and HMDB will be used to identify differentiating metabolites not included in our in-house library.Profiling of the Brexpososme and host metabolome: Raw QToF MS data will be processed by using XCMS software that performs chromatographic peak detection and retention-time alignment. The list of peaks will then be re-processed by X13CMS for the purpose of retrieving groups of metabolite isotopologues that differ by the number of incorporated deuterium atoms. The software links any given set of isotopologues representing a labeled metabolite with the m/z of the corresponding unlabeled metabolite (Fig 3). These pairs of labeled and unlabeled metabolites represent broccoli-derived compounds/metabolites and will be used to define the Brexposome. Next, the deuterium-enriched metabolites will be subtracted from the original XCMS peak list to obtain the host metabolome (Fig 1). Differences in the subtracted metabolomes of broccoli-treated and untreated human subjects reflect metabolome changes induced by broccoli sprout consumption and will be investigated to reveal the molecular targets of broccoli phytochemicals. Brexposome and host metabolome data will be examined for broccoli intake by machine learning techniques in Aim 3.Broccoli and human stool incubation: To simulate oral, gastric, and intestinal digestion, 5-day old broccoli sprouts will be digested in vitro with broccoli sprout digest (BSD), as previously published 109,110. We will obtain fresh stool samples subjects prior to the intervention period. Fecal slurries will be prepared from the stool samples by homogenization with PBS buffer (0.1 M, pH 7). 500 uL of 20 % fecal slurry will be mixed with 10 mL of chopped meat medium, and incubated with or without 1 mL BSD at 37 °C for 24-48 hours in anaerobic conditions 111. After incubation, medium will be extracted and analyzed by UPLC-QToF mass spectrometry.Controlled feeding studies: Stool samples from subjects (design described in Aim1) will be obtained at home using the OMNIgene GUT collection kits112 and EtOH collection kit and brought to clinic sites. Samples will be obtained pre-broccoli consumption and at 24, 48 and 72 hrs post-consumption for the short term study. For the 7-day study, samples will be obtained each day through day 10. Untargeted metabolomics will be performed on aliquots collected from these stool samples16S sequence generation and analysis: Following our prior work 58, we will extract whole genomic DNA from each incubation or subject stool sample using a QIAGEN PowerFecal DNA kit. We will then PCR amplify the V4V5 region of the 16S gene116 and sequence amplicons on an Illumina MiSeq. We will produce 300 bp, paired-end sequences and target a sequence depth of 50,000 reads per sample. The DADA2117 software pipeline will quality filter sequences, assemble contigs between paired ends, resolve unique amplicon sequence variants (i.e., ASVs), and taxonomically annotate these sequences using the RDP database. The Phyloseq118 R package will quantify the alpha- and beta-diversity of microbiome samples. Non-parametric tests (e.g., PERMANOVA) will measure differences in alpha- and beta-diversity and beta-dispersion across patient groups and over time119. We have extensive experience developing and applying methods for the analysis of microbiomeMachine Learning Methods: We consider the problem of assessing broccoli consumption as a binary classification problem where a sample is considered as an instance associated with either a positive (with broccoli consumption) or negative (no broccoli) label.Pharmacokinetic behavior of broccoli exposure markers: For the broccoli markers that are most diagnostic for broccoli consumption, we will construct pharmacokinetics (PK) curves to evaluate their time-dependent concentrations in plasma and urine. In our preliminary machine-learning experiments to assess the diagnostic power of broccoli intake markers (see Aim 3), we found that the hepatic metabolite, SFN-GSH, had greater predictive value during the first 3h following broccoli consumption while the gut microbial metabolite, SFN-nitrile, had greater predictive power at the later time points (12 and 24 h) in comparison with other SFN metabolites. These experiments demonstrate the need to examine PK profiles. PK modelling will be performed by Dr. Mark Christensen, Professor in the OSU College of Pharmacy, who has co-published with Dr. Ho46 and with Dr. Stevens46,100,101 on PK studies of dietary phytochemicals (see support letter). We will quantify broccoli-derived metabolites by LC-MS/MS using approaches outlined in Fig 6.Power and Statistical Analysis: Our preliminary studies have detected 4310 features in plasma, 3 h after broccoli consumption. For example, if 10% of these features are statistically significant between broccoli vs. control group, our feeding study with n=20 in each group will provide 86% power to detect an average effect size of 1.3, using multiple 2-sided two-sample t-test [False Discovery Rate (FDR=0.05)]. The power analysis was computed via PASS 14. The primary outcomes will be levels of features/metabolites. T-test or non-parametric tests will be conducted depending on the distribution of the outcomes comparing the groups with FDR adjustment for control of multiple comparisons.

Progress 04/15/23 to 04/14/24

Outputs
Target Audience:Peer scientists and scientific community Changes/Problems: Nothing Reported What opportunities for training and professional development has the project provided?Emily Ho and Laura Beaver also attended the American Society for Nutrition Annual conference in 2023 to learn the most cutting edge knowledge on food biomarker research and network with other professionals in the nutrition and microbiome fields. The Diet and Optimum Health conference was also attended by all research staff in 2023 to learn about cutting edge research in nutrition and network with other professionals. How have the results been disseminated to communities of interest?See publication and conference presentation list. What do you plan to do during the next reporting period to accomplish the goals?To achieve our objectives we will continue by finishing the analysis of the microbiome and metabolomics samples collected in the feeding trial. To this end an LC- MS/MS experiment is planned to resolve identities of some metabolites and possible expand our metabolomics data set. This work is focused on identifying biosignatures of broccoli sprout consumption. Work is also planned to closely examine interrelationships between microbiome and metabolic features identified in the metabolomics data set. This work will also incorporate additional metadata we captures in the human feeding trial. For example the baseline microbiome 16S samples will be analyzed using new methodologies in conjunction, with the targeted SFN levels, and carbohydrate intakes. By integrating multi-omics data we plan to predict which gut microbial taxa are responsible for which biotransformations of broccoli compounds and broccoli metabolites. We also plan to disseminate results as conference presentations and/or manuscripts.

Impacts
What was accomplished under these goals? Objective #1: Identify signatures of broccoli intake that quantify and discriminate between broccoli-derived vs "metabolic impacts" on host metabolome present in urine and plasma, in a controlled feeding trial. In order to identify signatures of broccoli intake we conducted a controlled feeding trial. To date, 265 people have expressed interest in participating and 86 subjects have enrolled and completed all study activities. We have currently closed recruitment to focus on data collection and analysis but have maintained an active IRB so that we can recruit more participants if needed. We have measured food intake for all participants and completed targeted measurements of broccoli related compounds in urine, plasma, and fecal material. This data has also informed our sample selection for our untargeted measurements of broccoli metabolites which has been run using an LC-MS/MS approach and has been analyzed. This directly supports our goal of discriminating between broccoli-derived vs metabolic impacts of broccoli consumption. We have also confirmed the presence of label in specific plant metabolites using an untargeted LC-MS/MS approach. We have developed a novel approach to the identification of label in urine and plasma following the consumption of broccoli sprouts. Briefly, untargeted metabolomics data is processed using open source R packages. While this method has been successful in identifying labeled plant metabolites, when applied to human data it yields a high number of false positives and false negatives. To address this issue, we developed a machine learning based technique to identify metabolites most likely to be labeled for further validation. Through this technique, we have identified 11 metabolites derived from plants which are incorporated with deuterium atoms, additionally, we have shown that our system can be used on both human urine and plasma and from plants grow in the presence of label ranging from 5 days to 3 months. We have submitted a manuscript with our results which is under peer review and all code will be publicly available and is built off open-source tools. Beyond identification of label, we have also begun work utilizing a data-driven approach to identify metabolites associated with broccoli consumption. Applying our data pipeline to the metabolomics data generated from our human feeding trial, we have applied linear mixed models (LMMs) to our data to identify metabolites which are altered with broccoli consumption and utilized partial least squares discriminant analysis (PLS-DA) to identify metabolites which most discriminate between broccoli and alfalfa consumers. Since many plant-derived metabolites are not well annotated or exist in MS/MS databases, we have been using a novel approach using de novo annotation to predict possible metabolite structures and chemical classes. While this allows us to gain more insight into the identity of metabolites, it also allows us to predict if a metabolite is derived from host or plant as endogenous metabolites are typically well annotated and exist within existing MS/MS databases. We presented these results at the American Society for Nutrition Conference in 2023 and we are writing the paper to share this data broadly. Objective #2: Identify microbial-mediated metabolites and associated microbiota following broccoli consumption In order to identify microbial-mediated metabolism of compounds from broccoli sprouts we completed an in vitro digestion of broccoli sprouts and incubated the resulting digesta with human fecal material for 24 hours under anaerobic conditions. 16S rRNA sequencing was conducted as well as untargeted metabolomics on the fecal cultures. We have published one article in 2022 on the impact of microbiome composition on sulforaphane and iberin nitrile production. We have also looked more globally at the metabolome to identify other cruciferous vegetable- and microbial- derived metabolites. Using data-driven approach we integrated metabolomics and 16S microbiome data and identified specific, treatment-dependent relationships between microbes and metabolites. FAdditionally, we used de novo annotation techniques to annotate our metabolomics data and gain greater insight into alterations in the metabolome. Cumulatively, these results were published and allowed us to bridge the gap between taxonomy and function and generate novel hypotheses regarding the role of the microbiome in cruciferous vegetable phytochemical metabolism. From our human feeding study, we further investigated the role of the gut microbiome in the generation of sulforaphane (SFN) and SFN-Nitrile from cruciferous vegetables. Through this study, we collected 280 fecal samples pre- and post- broccoli consumption. These samples were used to analyze participants' microbiome profile, as well as to examine microbial metabolites of broccoli sprouts present in the fecal samples. All fecal samples were processed for16S rRNA sequencing, and microbiome sequencing and analyses have been completed. We have also established fecal sample preparation and targeted mass spec methods to detect SFN and related metabolites in cryopreserved fecal samples. Results showed that we can detect SFN-related metabolites in the fecal samples in 37 out of 38 subjects who were fed broccoli sprouts, with SFN being the most prominent metabolites detected. Additionally, we measured SFN-NIT in human plasma for the first time finding that it persists in circulation for up to 72 hours while SFN and its downstream metabolites are washed out of circulation in under 24 hours. These findings allowed us to hypothesize on the bioactivity and metabolism of SFN-NIT in vivo, a topic had little knowledge on. From our human study we were able to integrate microbiome data to SFN-metabolite data. While we found that a single serving of broccoli sprouts did not alter microbiome composition, we did identify relationships between individual microbes and SFN-metabolites. Members of genus Bifidobacterium, Dorea, and Roseburia were found to be positive associated with SFN metabolites while members of genera Alistipes and Blautia were found to be negatively associated with SFN metabolites. These findings support other observations in the literature as members of Blautia in particular have been shown to decrease with cruciferous vegetable consumption and members of Bifidobacterium have been shown in vitro to hydrolze glucosinolates to nitriles. Additionally, we found a positive correlation between Bifidobacterium, Roseburia, and Dorea and carbohydrate and fiber consumption. The results of this study were recently published in a manuscript. Additionally, work was completed to integrate microbiome data with untargeted metabolomics data to look more globally at other metabolites of cruciferous vegetables and we expect to finish that analysis this year.

Publications

  • Type: Conference Papers and Presentations Status: Published Year Published: 2023 Citation: American Society of Nutrition Conference, June 2023, Boston, MA, Discovery of Novel Metabolomic Signatures Following the Consumption of Broccoli Sprouts Using Untargeted Metabolomics, Laura M. Beaver, John A. Bouranis, Jaewoo Choi, Lily J. He, Carmen P. Wong, Duo Jiang, Thomas J. Sharpton, Jan F. Stevens, Emily Ho
  • Type: Conference Papers and Presentations Status: Other Year Published: 2023 Citation: Diet and Optimum Health Conference, September 2023, Corvallis OR Interplay Between Cruciferous Vegetables and the Gut Microbiome: A Multi-Omic Approach, , Bouranis, John A; Beaver, Laura M; Jiang, Duo; Choi, Jaewoo; Wong, Carmen P; Davis, Edward W; Williams, David E; Sharpton, Thomas J; Stevens, Jan F; Ho, Emily
  • Type: Conference Papers and Presentations Status: Other Year Published: 2023 Citation: Diet and Optimum Health Conference, September 2023, Corvallis OR, Discovery of Novel Metabolomic Signatures Following the Consumption of Broccoli Sprouts Using Untargeted Metabolomics, Laura M. Beaver, John A. Bouranis, Jaewoo Choi, Lily J. He, Carmen P. Wong, Duo Jiang, Thomas J. Sharpton, Jan F. Stevens, Emily Ho
  • Type: Journal Articles Status: Published Year Published: 2024 Citation: Sulforaphane and Sulforaphane-Nitrile Metabolism in Humans Following Broccoli Sprout Consumption: Inter-individual Variation, Association with Gut Microbiome Composition, and Differential Bioactivity. Bouranis JA, Beaver LM, Wong CP, Choi J, Hamer S, Davis EW, Brown KS, Jiang D, Sharpton TJ, Stevens JF, Ho E. Mol Nutr Food Res. 2024 Feb;68(4):e2300286. doi: 10.1002/mnfr.202300286. PMID: 38143283
  • Type: Journal Articles Status: Published Year Published: 2023 Citation: Sulforaphane Bioavailability in Healthy Subjects Fed a Single Serving of Fresh Broccoli Microgreens. Bouranis JA, Wong CP, Beaver LM, Uesugi SL, Papenhausen EM, Choi J, Davis EW 2nd, Da Silva AN, Kalengamaliro N, Chaudhary R, Kharofa J, Takiar V, Herzog TJ, Barrett W, Ho E. Foods. 2023 Oct 15;12(20):3784. doi: 10.3390/foods12203784. PMID: 37893677


Progress 04/15/22 to 04/14/23

Outputs
Target Audience:peer scientists and scientific community Changes/Problems:With COVID-related delays related to human subjects research - we had significant delay in the start of recruitment of our human subjects studies due to needed revisions to IRB protocols to adapt to evolving COVID protocols. What opportunities for training and professional development has the project provided?Research staff attended the 2021 Center for Qualitative Life Sciences fall conference to learn more about current research methodologies and best practices for microbiome research. Additionally, research staff attended the Metabolomics Association of North America (MANA) annual conference to network with other professionals in the field and stay abreast with current knowledge and best practices for metabolomics data analysis. How have the results been disseminated to communities of interest?We have published one paper within the last year related to our goals. It is: Interplay between Cruciferous Vegetables and the Gut Microbiome: A Multi-Omic Approach. Bouranis JA, Beaver LM, Jiang D, Choi J, Wong CP, Davis EW, Williams DE, Sharpton TJ, Stevens JF, Ho E. Nutrients. 2022 Dec 22;15(1):42. doi: 10.3390/nu15010042. PMID: 36615700 Other papers we have published prior to this year that are directly related to the study goals are:: Metabolic Fate of Dietary Glucosinolates and Their Metabolites: A Role for the Microbiome. Bouranis JA, Beaver LM, Ho E. Front Nutr. 2021 Sep 22;8:748433. doi: 10.3389/fnut.2021.748433. 2021. PMID: 34631775 Composition of the Gut Microbiome Influences Production of Sulforaphane-Nitrile and Iberin-Nitrile from Glucosinolates in Broccoli Sprouts. Bouranis JA, Beaver LM, Choi J, Wong CP, Jiang D, Sharpton TJ, Stevens JF, Ho E. Nutrients. 2021 Aug 28;13(9):3013. doi: 10.3390/nu13093013. PMID: 34578891 Results were additionally distributed at the MANA 2022 conference in Edmonton Alberta. What do you plan to do during the next reporting period to accomplish the goals?We will continue by finishing the analysis of the microbiome and metabolomics (labeled and unlabeled) samples collected in the feeding trial. This work is reaching completion in the near future and we are preparing manuscripts and doing final analysis that is likely to produce 3 papers that focus on 1) the role of the gut microbiome in driving inter-individual variation in SFN metabolites, 2) Identification of potential biomarkers of cruciferous vegetable consumption utilizing global untargeted stable isotope tracing 3) Biomarkers of cruciferous vegetable consumption. Initial discussion have begun on the use of machine learning techniques to identify both biomarkers of cruciferous vegetable consumption and other food components. These discussion will become more formal analyses to identify these components. For example the microbiome samples described above will be analyzed in conjunction with the metabolomics samples. By integrating multi-omics data, specifically 16S rRNA gene sequences and metabolomic measurements, we plan to predict which gut microbial taxa are responsible for which biotransformations of broccoli compounds and broccoli metabolites. Furthermore to identify signatures of broccoli intake we have plans to explore machine learning methods. Additional work is planned to more closely examine interrelationships between microbiome, metabolomic and other metadata in both the fecal incubation study and the human feeding trial. Results will be disseminated as conference presentation and manuscripts.

Impacts
What was accomplished under these goals? Objective #1: Identify signatures of broccoli intake that quantify and discriminate between broccoli-derived vs "metabolic impacts" on host metabolome present in urine and plasma, in a controlled feeding trial. In order to identify signatures of broccoli intake we are conducting an ongoing controlled feeding trial. To date, 265 people have expressed interest in participating and 86 subjects have enrolled and completed all study activities. We have currently closed recruitment to focus on data collection and analysis but have maintained an active IRB so that we can recruit more participants if needed. We have measured food intake for all participants and completed targeted measurements of broccoli related compounds in urine, plasma, and fecal material. This data has also informed our sample selection for our untargeted measurements of broccoli metabolites which has been run using an LC-MS/MS approach and is currently being analyzed. This directly supports our goal of discriminating between broccoli-derived vs metabolic impacts of broccoli consumption. We have also confirmed the presence of label in specific plant metabolites using an untargeted LC-MS/MS approach. We have developed a novel approach to the identification of label in urine and plasma following the consumption of broccoli sprouts. Briefly, untargeted metabolomics data is processed using open source R packages. While this method has been successful in identifying labeled plant metabolites, when applied to human data it yields a high number of false positives and false negatives. To address this issue, we developed a machine learning based technique to identify metabolites most likely to be labeled for further validation. Through this technique, we have identified 11 metabolites derived from plants which are incorporated with deuterium atoms, additionally, we have shown that our system can be used on both human urine and plasma and from plants grow in the presence of label ranging from 5 days to 3 months. We are currently drafting a manuscript with our results and all code will be publicly available and is built off open-source tools. Beyond identification of label, we have also begun work utilizing a data-driven approach to identify metabolites associated with broccoli consumption. Applying our data pipeline to the metabolomics data generated from our human feeding trial, we have applied linear mixed models (LMMs) to our data to identify metabolites which are altered with broccoli consumption and utilized partial least squares discriminant analysis (PLS-DA) to identify metabolites which most discriminate between broccoli and alfalfa consumers. Since many plant-derived metabolites are not well annotated or exist in MS/MS databases, we have been using a novel approach using de novo annotation to predict possible metabolite structures and chemical classes. While this allows us to gain more insight into the identity of metabolites, it also allows us to predict if a metabolite is derived from host or plant as endogenous metabolites are typically well annotated and exist within existing MS/MS databases. We plan to the have the results of this analysis published within a year. Objective #2: Identify microbial-mediated metabolites and associated microbiota following broccoli consumption In order to identify microbial-mediated metabolism of compounds from broccoli sprouts we completed an in vitro digestion of broccoli sprouts and incubated the resulting digesta with human fecal material for 24 hours under anaerobic conditions. 16S rRNA sequencing was conducted as well as untargeted metabolomics on the fecal cultures. We have published one article in 2022 on the impact of microbiome composition on sulforaphane and iberin nitrile production. Next, we were interested in looking more globally at the metabolome to identify other cruciferous vegetable- and microbial- derived metabolites. Using data-driven approach we integrated metabolomics and 16S microbiome data and identified specific, treatment-dependent relationships between microbes and metabolites. Additionally, we used de novo annotation techniques to annotate our metabolomics data and gain greater insight into alterations in the metabolome. Cumulatively, these results allowed us to bridge the gap between taxonomy and function and generate novel hypotheses regarding the role of the microbiome in cruciferous vegetable phytochemical metabolism. From our human feeding study, we further investigated the role of the gut microbiome in the generation of SFN and SFN-Nitrile from cruciferous vegetables. Through this study, we collected 280 fecal samples pre- and post- broccoli consumption. These samples were used to analyze participants' microbiome profile, as well as to examine microbial metabolites of broccoli sprouts present in the fecal samples. All fecal samples were processed for16S rRNA sequencing, and microbiome sequencing and analyses have been completed. We have also established fecal sample preparation and targeted mass spec methods to detect SFN and related metabolites in cryopreserved fecal samples. Results showed that we can detect SFN-related metabolites in the fecal samples in 37 out of 38 subjects who were fed broccoli sprouts, with SFN being the most prominent metabolites detected. Additionally, we measured SFN-NIT in human plasma for the first time finding that it persists in circulation for up to 72 hours while SFN and its downstream metabolites are washed out of circulation in under 24 hours. These findings allowed us to hypothesize on the bioactivity and metabolism of SFN-NIT in vivo, a topic we currently have little knowledge on. From our human study we were able to integrate microbiome data to SFN-metabolite data. While we found that a single serving of broccoli sprouts did not alter microbiome composition, we did identify relationships between individual microbes and SFN-metabolites. Members of genus Bifidobacterium, Dorea, and Roseburia were found to be positive associated with SFN metabolites while members of genera Alistipes and Blautia were found to be negatively associated with SFN metabolites. These findings support other observations in the literature as members of Blautia in particular have been shown to decrease with cruciferous vegetable consumption and members of Bifidobacterium have been shown in vitro to hydrolze glucosinolates to nitriles. Additionally, we found a positive correlation between Bifidobacterium, Roseburia, and Dorea and carbohydrate and fiber consumption. The results of this study are currently being drafted into a manuscript. Additionally, work is on-going to integrate microbiome data with untargeted metabolomics data to look more globally at other metabolites of cruciferous vegetables.

Publications

  • Type: Journal Articles Status: Published Year Published: 2022 Citation: Interplay between Cruciferous Vegetables and the Gut Microbiome: A Multi-Omic Approach. Bouranis JA, Beaver LM, Jiang D, Choi J, Wong CP, Davis EW, Williams DE, Sharpton TJ, Stevens JF, Ho E. Nutrients. 2022 Dec 22;15(1):42. doi: 10.3390/nu15010042. PMID: 36615700
  • Type: Conference Papers and Presentations Status: Other Year Published: 2022 Citation: Interplay between Cruciferous Vegetables and the Gut Microbiome: A Multi-Omic Approach. Bouranis JA, Beaver LM, Jiang D, Choi J, Wong CP, Davis EW, Williams DE, Sharpton TJ, Stevens JF, Ho E; Poster - MANA Conference 2002, Edmonton, AB


Progress 04/15/21 to 04/14/22

Outputs
Target Audience:Scientific community Changes/Problems:With COVID-related delays related to human subjects research - we had significant delay in the start of recruitment of our human subjects studies. We adjusted our timelines and instead focused on fecal incubation studies while we waited for approvals for human subjects research to resume. What opportunities for training and professional development has the project provided?Research staff attended the 2021 Center for Qualitative Life Sciences fall conference to learn more about current research methodologies and best practices for microbiome research. How have the results been disseminated to communities of interest?We have published two papers within the last year related to our goals and presented data at the 2021 Center for Quantitative Life Sciences Fall Conference and with investigators involved in AES Multistate W4002 annual meeting. What do you plan to do during the next reporting period to accomplish the goals?We will continue the analysis of the microbiome and metabolomics (labeled and unlabeled) samples collected in the feeding trial. For example the microbiome samples described above will be analyzed in conjunction with the metabolomics samples. By integrating multi-omics data, specifically 16S rRNA gene sequences and metabolomic measurements, we plan to predict which gut microbial taxa are responsible for which biotransformations of broccoli compounds and broccoli metabolites. Furthermore to identify signatures of broccoli intake we have plans to explore machine learning methods. Additional work is planned to more closely examine interrelationships between microbiome, metabolomic and other metadata in both the fecal incubation study and the human feeding trial. Results will be disseminated as conference presentation and manuscripts.

Impacts
What was accomplished under these goals? Objective #1: Identify signatures of broccoli intake that quantify and discriminate between broccoli-derived vs "metabolic impacts" on host metabolome present in urine and plasma, in a controlled feeding trial. In order to identify signatures of broccoli intake we are conducting a controlled feeding trial with broccoli sprouts and have included a control arm of alfalfa sprout feeding. The trial is currently ongoing with 265 interested participants contacting us. To date 70 subjects have enrolled and completed all study activities. Nine additional participants have withdrawn during the study because of scheduling conflicts, family emergencies, and illness. We are currently taking a break from recruitment to focus on data collection and analysis. If needed we will again recruit more participants in Summer 2022. To accomplish the key goal of identifying signatures of broccoli sprout intake we have measured food intake in all of our participants and completed targeted measurements of broccoli related compounds in the urine of participants who consumed broccoli. This data has informed our sample selection for our untargeted measurements of broccoli metabolites which is ongoing and will directly supports our goal of discriminating between broccoli-derived vs metabolic impacts of broccoli consumption. We have also confirmed the presence of label in specific plant metabolites using an untargeted LC-MS/MS approach. We have optimized our metabolomics pre-processing pipeline to utilize AutoTuner for XCMS parameter selection and HiResTEC for label identification. Using this pipeline, we have successful identified 5 labeled metabolites in human plasma and urine. Work is ongoing to develop a machine learning classifier to aid in the identification of label in our human samples. Identification of labeled metabolites in human plasma and urine will allow for us to differentiate between host- and plant-derived metabolites. Objective #2: Identify microbial-mediated metabolites and associated microbiota following broccoli consumption In order to identify microbial-mediated metabolism of compounds from broccoli sprouts we completed both a review of the literature and an in vitro digestion of broccoli sprouts and incubated the resulting digesta with human fecal material for 24 hours under anaerobic conditions. 16S rRNA sequencing was conducted as well as untargeted metabolomics on the fecal cultures. Two papers have come out of this work and a third is currently being written. In this work we examined the abundance of two broccoli related metabolites, sulforaphane-nitrile and iberin-nitrile, whose production could be influenced by gut microbiome composition and may lower the efficacy of cruciferous vegetable interventions. We showed that the Clostridiaceae family was associated with high levels of nitriles, while individuals with a low abundance of nitriles were more enriched with taxa from the Enterobacteriaceae family. We also established in samples from a human feeding trial that the microbiome may contribute to inter-individual variation in these broccoli related metabolites as sulforaphane-nitrile levels were highly variable in urine of participants who consumed broccoli sprouts. We also found sulforaphane-nitrile had peak excretion at 24 h post broccoli sprout consumption. Work is ongoing to further integrate the metabolomics and 16S rRNA data and communicate results in peer reviewed journal articles. Utilizing a human fecal incubation model, we have developed an analysis pipeline for the multi-omic integration of 16S rRNA and untargeted metabolomics data. This analysis allows us to determine relationships between specific microbial taxa and metabolomic products, elucidating the role of the gut microbiome as a source of inter-individual variation in cruciferous vegetable consumption. The results of the human fecal incubation model will be disseminated as a peer-reviewed research article and our analysis pipeline will be tailored and applied to human data from our clinical trial. We also have collected 280 fecal samples pre- and post- broccoli consumption to evaluate microbial metabolites of broccoli sprouts. We have established fecal sample preparation and targeted mass spec methods to detect SFN and related metabolites in cryopreserved fecal samples. Preliminary results showed that we can detect SFN-related metabolites in the fecal samples in 37 out of 38 subjects who were fed broccoli sprouts, with SFN being the most prominent metabolites detected.

Publications

  • Type: Journal Articles Status: Published Year Published: 2021 Citation: Metabolic Fate of Dietary Glucosinolates and Their Metabolites: A Role for the Microbiome. Bouranis JA, Beaver LM, Ho E. Front Nutr. 2021 Sep 22;8:748433. doi: 10.3389/fnut.2021.748433. 2021. PMID: 34631775
  • Type: Journal Articles Status: Published Year Published: 2021 Citation: Composition of the Gut Microbiome Influences Production of Sulforaphane-Nitrile and Iberin-Nitrile from Glucosinolates in Broccoli Sprouts. Bouranis JA, Beaver LM, Choi J, Wong CP, Jiang D, Sharpton TJ, Stevens JF, Ho E. Nutrients. 2021 Aug 28;13(9):3013. doi: 10.3390/nu13093013. PMID: 34578891
  • Type: Conference Papers and Presentations Status: Other Year Published: 2021 Citation: Composition of the Gut Microbiome Influences Production of Sulforaphane-Nitrile and Iberin-Nitrile from Glucosinolates in Broccoli Sprouts. Bouranis JA, Beaver LM, Choi J, Wong CP, Jiang D, Sharpton TJ, Stevens JF, Ho E. Center for Quantitative Life Sciences Annual conference 2021 - POSTER presentation


Progress 04/15/20 to 04/14/21

Outputs
Target Audience:Scientific community Changes/Problems:With COVID-related delays related to human subjects research - we had significant delay in the start of recruitment of our human subjects studies. We adjusted our timelines and instead focused on fecal incubation studies while we waited for approval for human subjects research to resume. What opportunities for training and professional development has the project provided?To learn more about diet and microbiome interactions, research staff attended the Nutrition 2020 Live Online conference in June 2020. How have the results been disseminated to communities of interest?Results were presented as a virtual posters at the Nutrition 2020 Live Online conference in June 2020. In addition, we preparing 2 manuscripts that described our findings related to cruciferous vegetable metabolism in June 2021. What do you plan to do during the next reporting period to accomplish the goals?We will continue recruit subjects and execute proposed feeding trials. We anticipate completion of sample collection for short term feeding study by April 2022. We will continue microbiome and metabolomics (labeled and unlabeled) analysis of samples. We will also start to plan for long term study. We have plans to explore machine learning methods and perform additional data integration to more closely examine interrelationships between microbiome, metabolomic and other metadata in fecal incubation study. Results will be disseminated as conference presentation and manuscripts.

Impacts
What was accomplished under these goals? Objective #1: Identify signatures of broccoli intake that quantify and discriminate between broccoli-derived vs "metabolic impacts" on host metabolome present in urine and plasma, in a controlled feeding trial. In order to identify signatures of broccoli intake we are conducting a controlled feeding trial with broccoli sprouts and have included a control arm of alfalfa sprout feeding. To initiate the study we amended the IRB protocol to include COVID safety precautions, got approval from our IRB board, and obtained necessary COVID related supplies. The trial, which has a single dose of sprouts feeding, is currently ongoing. At this time, recruitment lead to 71 individuals initiating the on-line screening. Among these individuals, 16 subjects meet all the screening criteria and have consented to participate. Recruitment and data collection is ongoing and will continue throughout the year to accomplish the key goal of identifying signatures of broccoli sprout intake. To reach our goal of discriminating between broccoli-derived vs metabolic impacts of broccoli consumption we have also optimized the growing of deuterium labeled alfalfa and broccoli sprouts and confirmed the presence of label in specific plant metabolites using an untargeted LC-MS/MS approach. These labeled sprouts are also currently being grown and feed to participants as part of the single dose of sprouts study. Furthermore, a pilot study of subjects fed labeled and unlabeled broccoli sprouts was completed and urine was analyzed using an untargeted LC-MS/MS method. Data from the pilot study is being analyzed with a variety of software (X13CMS, DeltaMS, geoRge, HiResTEC) to optimize methods and facilitate the discovery of specific broccoli-derived metabolites which is a key outcome we are working towards. Objective #2: Identify microbial-mediated metabolites and associated microbiota following broccoli consumption In order to identify microbial-mediated metabolism of broccoli sprouts we completed an in vitro digestion of broccoli sprouts and incubated the resulting digesta with human fecal material for 24 hours under anaerobic conditions. 16S rRNA sequencing was conducted as well as untargeted metabolomics on the fecal cultures. Using dada2, we identified 72 ASVs associated with bacteria within our samples. Analysis at both the alpha- and beta-diversity levels indicated little treatment effect from incubation with cruciferous vegetables, however, beta-diversity analysis indicated the presence of two sub-populations amongst our samples. Further analysis utilizing a negative binomial generalized linear model indicated that these two groups differed in abundance of bacteria from the Clostridiaceae and Enterobacteriaceae families. To identify metabolites which differentiated between Clostridiaceae-type and Enterobacteriaceae-type individuals, we built a linear mixed model for our metabolomics data. Of 28,000 metabolomic features, 281 were significantly different (q < 0.05) between the two groups, indicating divergent metabolic capabilities between the two groups. One metabolite of interest, sulforaphane nitrile, had a greater abundance in Clostridiaceae-type individuals, indicating that metabolism of cruciferous vegetable glucosinolates could be directly impacted by the composition of the gut microbiome. Work is ongoing to further integrate the metabolomics and 16S rRNA data and communicate results in peer reviewed journal articles.

Publications

  • Type: Conference Papers and Presentations Status: Published Year Published: 2020 Citation: Effects of Collard Green Consumption on the Human Plasma and Urine Metabolome: An Untargeted Analysis. Current Developments in Nutrition. 2020;4(Supplement_2):372-372. doi:10.1093/cdn/nzaa045_005