The human gut microbiota has been linked with incidence and progression of noncommunicable diseases and their risk factors. Moreover, diet has been identified as an important modulator of microbiota composition and function, but responses vary across individuals.

DiGuMet aimed to explore how gut microbiota interacts with diet and the role of such interactions for cardiometabolic disease risk and to dissect underlying mechanisms and finding biomarkers reflecting such interactions, through extensive molecular phenotyping e.g., microbiomics and metabolomics combined with lifestyle data. We hypothesized that: (i) gut microbiota x diet interactions are a major determinant of metabotypes linked with cardiometabolic disease risk; (ii) distinct metabotypes can be identified in free-living subjects based on comprehensive assessments of health-related status and multi-OMICs; and (iii) these metabotypes could be used as the basis for tailored dietary interventions to maximize health benefits for groups of individuals.

To address hypotheses i-ii, we conducted a validation study, MAX, (n ~500) a sub-cohort of the Danish Diet, Cancer and Health- Next Generations cohort (DCH-NG; n = 39,554), where biological samples (including saliva), clinical and anthropometric data were collected at 0, 6 and 12 months across gender, age and fasting strata. Complementary untargeted and targeted metabolomics analyses of blood and urine samples were undertaken and gut microbiota analyses performed. We investigated the stability of both metabolites and gut microbiota over 1 year (0, 6 and 12 months). Results can guide on which signals in the metabolome and microbiome that may be stable enough to be studied in relation to cardiometabolic risk factors. Moreover, we explored the role of different determinants including diet on the metabolome and gut microbiome.The data was used to mine for novel dietary intake biomarkers and to validate biomarkers of dietary fibre x gut microbiota interactions, intake of berries and healthy food indices.

For metabotyping, we applied a large set of approaches, including machine learning and intergrative clustering methods, on antropometric and clinical data as well as metabolomics and gut microbial data in the MAX study. In contrast to our hypothesis, we did not find any groupings into metabotypes in free-living participants that were deemed sufficiently distinct or differentially related to cardiometabolic risk factors to guide the development into novel metabotype-based prevention strategies. Instead, we discovered metabotypes in postprandial data from two pervious intervention studies and developed new methods for metabotyping using dynamic (time-series) data and diet challange tests. Results suggest that postprandial data is better suited for metabotyping-strategies and that metabotypes reflecting differential trajectories in for example postprandial amino acid levels could be differentially related to health-related outcomes and that such approach could be useful to guide dietary interventions for improved cardiometabolic disease prevention.

To adress to hypothesis iii, we conducted a randomized controlled cross-over trial, FiFerm, to evaluate inter-and intraindividual postprandial responses to two different fermentable cereal fibers and a cellulose control, in 20 individuals with high risk of cardiometabolic disease. SCFA response was time-dependent and results suggested that the different fibers can lead to increases in specific SCFA. Moreover, we found differential response to the fermentable fibers in terms of postprandial SCFA response (acetate, propionate and butyrate), allowing us to define responders and non-responders and "butyorogenic" and "propiongenic" metabotypes. Ongoing analysis into the gut microbiome, plasma lipids and postprandial incretins is providing insight on diet x microbiota responses in the context of SCFA-metabotypes. The SCFA-metabotype challange test shows that it is possible to development a robust and reproducible method to identifying fermentable fiber responders and non-responders. This method is now being opimized and validated in other studies and is ultimately expected to contribute to new tailored dietary strategies, including to prevent cardiometabolic diseases using fermentable dietary fibre.

Taken together we have evaluated two different approaches to metabotyping; metabotyping based on static multiomics data in a free living population and metabotyping based on postprandial response and time series data and developed a diet challange test. The two conducted studies are great resources- multi-omics data, dietary, lifestyle and health data, and fecal, plasma, urine and saliva samples- and have in DiGuMet been used to e.g., expand a database to include polyphenols and exposome biomarkers, optimize a fecal freeze-drying method and develop a food-biomarker ontology and online tools for omics analysis (see highlights and WP 2-4)

The DiGuMet project concludes having conducted one 12-month observational study with >500 participants at three time points generating multi-omics data as well as clinical, antropometric, diet and other lifestyle data and one randomized controlled trial investigating SCFA-based metabotypes basd on inter- and intra-individual response to two different fermentable fibers and their relation to cardiometabolic risk factors measured.

We highlight:

* The establishment of the MAX validation study of the DCH-NG cohort with ~500 Danish participants with data collected at baseline, 6 and 12 months. This includes collection of urine-, blood-, fecal-, and saliva samples along with data collected on clinical and anthropometric variables, disease history, physical activity, socioeconomic factors and detailed dietary data that includes repeated 24h dietary recalls and food frequency questionnaires. The analysis conducted comprise untargeted- and targeted plasma and targeted urine metabolomes and the gut microbiota. This study represents a goldmine for biomarker discovery and validation. It has enabled a thorough investigation into the presence of metabotypes and studies of their relation to health-related outcomes. Furthermore, studies of the stability of metabolomes and the gut microbiota over time and their determinants as well as diet-cardiometabolic risk factor relationships as well as the concentrations and presence of bioactive metabolites, pesticides, and contaminants in the diet. The samples and generated data will continue to be a useful resource for numerous studies to come including metabolite biomarkers for exposures, namely dietary exposures, diet x gut microbiota relationships and cardiometabolic endpoints. Gut microbiota analysis and plasma metabolomics (exploratory outcomes) are used to explore and unravel if the gut microbiota composition and plasma metabolites are associated with the plasma SCFA response after the fiber challenge test and with potential improvements in cardiometabolic parameters, including plasma glucose, insulin, blood lipids, blood pressure and free fatty acid levels. 

 * Investigations into the stability of both metabolites and the gut microbiota over one year (0, 6 and 12 months) and the role of different determinants including diet on the metabolome and gut microbiome. Results can guide on which signals in the metabolome and microbiome that may be stable enough to be studied in relation to cardiometabolic risk factors.

* The thorough investigation into the discovery of metabotypes in the MAX study, based on clinical and anthropometric variables, metabolomics and gut microbial data using machine learning and integrative clustering methods. In contrast to our hypothesis, we did not observe any clusters in free-living participants that we deemed sufficiently distinct to guide the development into novel metabotype-based prevention strategies. Instead, we have observed metabotypes in postprandial data from two intervention studies (Skantze et al. 2022) and unpublished. These represented differential postprandial responses to protein rich meals in key amino acids that related differently to plasma creatinine (Skanze et al. 2022) and differential plasma SCFA AUCs, in response to developed dietary fibre challenges. Taken together, these results suggest that postprandial- and times series data may be more suited for metabotyping-strategies than static data and it remains to validate if it has a meaning for cardiometabolic disease prevention. Our work resulted in novel methodological developments of metabotyping of time-series data reported in Skanze et al. (2022).

* The randomized controlled cross-over trial, Fiferm, conducted to evaluate the presence of metabotypes based on postprandial sort-chain fatty acid (SCFA) response to two fermentable cereal fibers (solubilized arabinoxylans and wheat bran) and cellulose control in 20 individuals with high risk for cardiometabolic disease. The dietary fibers were first used to develop new food products and provided to participants as part of breakfast meals. The concentrations of the short chain fatty acids were time dependent and varied across individuals, allowing for identification of butyratogenic and propiongenic metabotypes (responders and non-responders). Moreover the two fermentable fibers led to differential effects: the intake of solublized arabinoxylans increased plasma levels of propionate by 44% and those of butyrate by 19%, corresponding increases for wheat bran were 7% and 21%, compared to cellulose. Arabinoxylans and wheat bran did not affect postprandial glycaemia in the 8 hours following their intake on group level, but had a differential impact on insulin levels in the first few hours after consumption. The gut microbiota and plasma metabolome (exploratory outcomes) are explored to unravel if the gut microbiota composition and function and plasma metabolites are associated with the SCFA response to the fiber challenge test and with potential improvements in cardiometabolic parameters, including plasma glucose, insulin, blood lipids, free fatty acids as well as with postprandial incretin levels. In respect to methodology development of a challenge test, we have also investigated the number of time points required to identify fermentable fiber responders and non-responders. Taken together, this study has provided support for a novel metabotyping strategy to differentiate individuals based on SCFA response in their plasma. This approach is now being validated in external studies to investigate wherther differential SCFA-metabotypes are differentially related to CMD risk factors after long-term interventions.

In addition, we have expanded our work to also include spin-off projects. We highlight:

* Investigations into gut microbiota x diet interactions and their associated effects on cardiometabolic risk factors, as an explorative analysis of data from a 12-week intervention study on fermented high-fibre cereal products (Liu et al. 2021). This work was in collaboration with partners in Shanghai using novel food products developed by Lantmännen and Barilla. The results suggest that the intake of high-fibre rye products could modify specific gut bacterial genera and that these effects are paralleled with favorable modifications of short-chain fatty acid concentrations and associated with altered glycemic traits. (Liu et al. 2021)

* Establishment of the dietary intake of polyphenols in the MAX study population as well as the DCH-NG cohort. Consortium partners from Spain and Denmark collaborated closely to link an in-house food composition database on polyphenols (Barcelona) to the comprehensive dietary data collected by the Danish partner. The amount of polyphenols consumed for each individual and day (mg/day) has been estimated from the 1670 food items and >5000 ingredients from the 24h dietary recall and will next be estimated using the 475 food items and >1200 ingredients from the food frequency questionnaire. Phenolic intake will be related to cardiometabolic risk factors and metabolomics data.

 * A portfolio of additionally 43 xenobiotic compounds, including 10 pesticides, 2 organoarsenic compounds and 31 household chemicals and environmental pollutants was included in the targeted metabolomics methodology developed by the UB team for the analysis of biological samples from the MAX study. Many of these xenobiotics were detected and quantified in plasma and urine samples, normally at very low concentration levels (sub-ppb). In particular for pesticides, their accurate quantification was below the limits of detection and limits of quantification of our methodology, due to the low levels usually found in the biological samples. Furthermore, due to their very high polarity, most of these pesticides were detected eluted in the void volume of the chromatographic profile, which consequently hinder their unequivocal detection and separation from potential coeluting interferences. The data on contaminants is also planned to be linked with organic vs. conventional food consumption and associated with cardiometabolic diseases.

We have also conducted substantial method development efforts. We have:

* Developed a challenge test with fermentable fibers to detect diet response metabotypes (manuscript in preparation)

* Developed new methods for metabotyping using dynamic (time series) data. This method was developed and applied to identify metabotypes from postprandial metabolite date from diet challenge setting. Metabotypes were differently associated with plasma creatinine (Skantze et al. 2022)

* A method for untargeted metabolomics of fecal samples was developed and optimized for large scale freeze-drying of fecal samples to precede gut microbiome analysis and metabolomics (Ken et al. Metabolomics, 2020)

* Conducted a comprehensive method investigation to derive general population metabotypes by combining anthropometric, biochemical and multiomics data. In addition, investigated approaches to improve metabotyping by adjusting Omics data to better reflect metabolic regulation, similar to adjusting dietary data for total energy intake using the residual method (manuscript).

* Expanded databases to include extensive polyphenol and exposure data. As part of this work, we developed and used a method to estimate the intake of dietary polyphenolic compounds based on software developed by the Spanish partner.

 * Expanded an in-house database with targeted metabolite biomarkers. The UHPLC-MS metabolomics platform can identify and quantify over 1000 metabolites commonly found in biological samples such as urine, serum and plasma. These include about 500 endogenous metabolites, 450 food-derived metabolites, 50 common pollutants, 50 drugs as well as microbiota-derived metabolites and biomarkers related to lifestyle habits e.g., smoking and alcohol consumption. This comprehensive method provides excellent opportunities to detect diet, gut microbiota and disease related biomarkers in the MAX study and beyond. This targeted methodology was published in the Journal of Agricultural and Food Chemistry (JAFC) in 2020 and was awarded the Best Research Article of the Year Award in 2021 for the Division of Agricultural and Food Chemistry.

 * Developed a web-based tool, POMAShiny, with a user-friendly workflow for the visualization, exploratory and statistical analysis of mass spectrometry data and a R/Bioconductor package, POMA, that integrates several statistical methods, including widely applied omics methods. Both POMAShiny and POMA are freely available online.

 * Developed FOBI (Food-Biomarker Ontology), a novel ontology that describes food and their associated metabolite entities in a hierarchical way (Spanish partner, Database Journal, 2020). This ontology has enabled the analysis and integration of the two different datatypes and uses a formal naming system, category definitions, properties and relations between both types of data. FOBI is composed of two interconnected sub-ontologies; a ’Food Ontology’ consisting of raw foods and ‘multi-component foods’ and a ‘Biomarker Ontology’ containing food intake biomarkers classified by their chemical classes. This allows data and information regarding foods and food biomarkers to be visualized in a bidirectional way, going from metabolomics to nutritional data or vice versa. Potential applications of this ontology include the annotation of foods and biomarkers using a well-defined and consistent nomenclature, the standardized reporting of metabolomics workflows or the application of different enrichment analysis approaches to analyze nutrimetabolomic data. FOBI is freely available in both OWL (Web Ontology Language) and OBO (Open Biomedical Ontologies) formats in the project’s Github repository.

 * Developed a novel open-source software (QualiMon – LaMa) for real-time quality monitoring in untargeted LC-MS metabolomics (submitted). This tool fills an important gap since manual curation is largely inadequate and quality issues are frequently not detected until later stages in the bioinformatics pipeline resulting in poor resource economy. QualiMon assesses data quality from automated extraction of thousands of individual quality characteristics. Transgressing samples activate real-time warnings through web services, so measures can immediately be taken. QualiMon is freely available under MIT licence at https://github.com/MetaboComp/QualiMon.