Abstract

Background and Aims

   The Mediterranean diet pattern (MDP) is associated with
   health-associated gut microbes and metabolites. However, the impact of
   the MDP on the fecal metabolome in ulcerative colitis (UC) remains
   unclear. We characterized the fecal metabolome of patients with UC with
   high adherence to the MDP compared to the Canadian habitual diet (CHD).
   Furthermore, we explored potential differences in the fecal metabolome
   between dietary responders and nonresponders to the MDP.

Methods

   Utilizing untargeted metabolomics on a subset of fecal samples obtained
   from a randomized controlled trial, adult patients with quiescent UC
   underwent a 12-week intervention following either the MDP (n = 8) or
   CHD (n = 8). Liquid chromatography-tandem mass spectrometry was
   employed to profile endogenous fecal metabolites, while 16S amplicon
   sequencing was utilized to profile the fecal microbiota.

Results

   A total of 701 human metabolites were detected, with 35 exhibiting
   significant differential expression between the MDP and CHD groups.
   Noteworthy, folate biosynthesis, sphingolipid biosynthesis, and steroid
   biosynthesis were identified as major pathways affected. Moreover,
   microbial analysis showed that individuals with increased levels of the
   class Bacteroidia (Bacteroides vulgatus [B. vulgatus], B. uniformis,
   and B. acidifaciens) in their stool at baseline were more likely to
   respond to the MDP.

Conclusion

   High adherence to an MDP is associated with beneficial metabolite
   changes associated with reducing inflammation in UC. In addition,
   fiber-degrading microbes abundant before dietary intervention played a
   role in the responsiveness to the MDP. This work lays the groundwork
   for developing a metabolic signature associated with the MDP to develop
   personalized nutrition strategies for UC prevention and treatment.
   [33]ClinicalTrials.gov Number: [34]NCT03053713.

   Keywords: Mediterranean Diet, Inflammatory Bowel Disease, Metabolome,
   Ulcerative Colitis

Graphical abstract

   [35]graphic file with name ga1.jpg
   [36]Open in a new tab

Introduction

   Inflammatory bowel disease (IBD) is a global health issue. The chronic
   relapsing and remitting pattern of IBD, including ulcerative colitis
   (UC) and Crohn’s disease, imposes a lifelong burden, including
   metabolic and neuropsychiatric comorbidities and extraintestinal
   manifestations, leading to challenges in disease
   management.[37]^1^,[38]^2 The financial burden of these diseases is
   immense and has been ranked as one of the fifth most costly
   gastrointestinal conditions in annual health-care expenditures.[39]^3
   This is concerning since recent trends show a steady increase in the
   incidence of IBD in India, China, Africa, and South America.[40]^4

   The modern Western diet, characterized by higher consumption of animal
   and dairy products, refined sugars, processed foods, and a reduction in
   plant-based fibers, is frequently implicated as an environmental risk
   factor for IBD.[41]^5 Processed foods in Western diets often contain
   additives, emulsifiers, and bulking agents, which are not typically
   present in traditional diets.[42]^6 This diet significantly differs
   from the traditional diets of earlier generations, which were locally
   sourced and consumed shortly after harvest. Another major difference is
   the addition of xenobiotics, mainly derived from environmental
   pollution or pesticides.[43]^7 Diet should be included as a tool to
   optimize the treatment of IBD, especially considering that it is a
   major predictor of the gut microbiome. The gut microbiome is critical
   in modulating immune responses and metabolic functions and maintaining
   intestinal barrier integrity. The Westernized microbiome is
   distinguished by the decline of Volatile and/or Associated Negatively
   with Industrialized Societies of Humans taxa, commonly found in
   populations adhering to traditional lifestyles, whereas the rise of
   Bloom or Selected in Societies of Urbanization/Modernization taxa are
   prevalent in modern, industrialized populations.[44]^8^,[45]^9 In IBD,
   evidence supports a lack of diversity and general instability in the
   bacteriome, often with a dysbiosis characterized by the relative
   depletion of Faecalbacterium praustnitzii and Roseburia hominis and the
   enrichment of Escherichia coli, Ruminococcus torques, and R.
   gnavus.[46]^10 The metabolomic profile is marked by diminished
   concentrations of short-chain fatty acids like butyrate, alterations of
   amino acids—particularly branched-chain amino acids—and alterations in
   fatty acid esters (acylcarnitines). Additionally, there are variations
   in the profiles of secondary bile acids.[47]10, [48]11, [49]12 Despite
   these advances, discrepancies among various studies have made it
   challenging to validate microbiome and metabolome signatures across
   diverse populations, diminishing their diagnostic value in IBD.
   Therefore, it is crucial to understand the complex interactions between
   diet, the gut microbiome, the metabolome, and the host immune system to
   effectively utilize diet as a therapeutic tool for managing IBD.

   One dietary approach gaining attention in IBD is the Mediterranean diet
   pattern (MDP).[50]13, [51]14, [52]15 The diet is characterized by
   increased consumption of legumes, whole grains, vegetables, fruits,
   nuts, and seeds, and low intake of red meat with olive oil as the main
   culinary fat.[53]^16 The synergy of the high levels of dietary fiber,
   phytochemicals, and the fat blend (high monounsaturated fat) confer
   beneficial anti-inflammatory and antioxidant properties.[54]^17
   Observational studies demonstrate that adherence to an MDP is
   associated with reductions in all-cause mortality[55]^18 and lower
   levels of fecal calprotectin (FCP).[56]^13^,[57]19, [58]20, [59]21
   Results from clinical trials suggest that the MDP is associated with
   improvements in biomarkers of inflammation (C-reactive protein and FCP,
   reduction of disease activity indices [partial Mayo score]),
   alterations in the bacteriome and improved quality of
   life.[60]^12^,[61]^14^,[62]^15^,[63]^22^,[64]^23 Given the benefits of
   the MDP to overall health and the tolerability of the diet in IBD, the
   American Gastroenterology Association recommends that all patients with
   IBD follow an MDP.[65]^24

   In this study, we characterized the fecal metabolome in patients with
   quiescent UC following an MDP compared to those following the CHD.
   Additionally, we aimed to determine whether baseline gut microbiota
   could predict the response level to an MDP.

Materials and Methods

Study Design and Sample Collection

   The present study is a post hoc exploratory analysis based on data
   obtained from a 12-week randomized control trial (RCT) (intention to
   treat analyses) that investigated whether the MDP vs CHD would affect
   disease activity, inflammation, and the gut microbiome in patients with
   mild, quiescent UC.[66]^22 Briefly, following informed consent,
   participants were randomized to either an MDP or CHD. The participants
   randomized to the MDP received a series of one-on-one online coaching
   sessions from a Master’s trained registered dietitian (RD; NH) with
   expertise in nutrition and IBD. The participants were counseled on how
   to adapt their diet based on the Mediterranean Diet Pyramid.[67]^16 The
   CHD participants met with the dietitian at a similar intervention
   intensity to the MDP and received a nutritional analysis comparing
   their intakes to the dietary reference intakes; however, they were not
   provided tailored dietary advice. The Mediterranean Diet Serving Score
   (MDSS) assessed adherence to the MDP.[68]^25 The MDSS score ranges from
   0 to 24 points, with greater than 13.5 points indicating adherence to
   the MDP. Moreover, the ASA24-Canada 2016[69]^26 data were used to
   validate the MDSS and enhance compliance. Comprehensive information on
   participant recruitment, the dietary intervention, collection of
   dietary data, baseline and week-12 diet details, and MDP dietary
   adherence scores are described elsewhere.[70]^27

   Untargeted metabolomics was completed on a subset of the fecal samples
   MDP (n = 8) or CHD (n = 8). As this was an exploratory analysis, we
   chose to examine samples in the MDP group that were most adherent to
   the MDP (median MDSS = 21), whereas those with lowest adherence scores
   to the MDSS scores (median 13 points) were chosen from the CHD group
   ([71]Table 1).

Table 1.

   Demographic Description of the Cohort of the Participants at Week 12
   (n = 16)
   Factors Canadian habitual diet (n = 8) median (IQR) MDP (n = 8) median
   (IQR)
   Demographics
    Male, n (%) 3 (38) 2 (25)
    Age (y) 42 (34–55) 56 (35–64)
    Body Mass Index (kg/m^2) 24 (22–28) 22 (20–29)
   Disease activity
    Partial Mayo score 0 (0–1) 0 (0)
    SCCAI 1 (0–1) 0 (0–1)
    Δ in fecal calprotectin over 12 wk (mcg/g)[72]^a 293 ([–19]–458)
   (n = 7) 3 ([–95]–75)
   Diet and quality of life
    Healthy eating index 55 (51–62) 85 (83–91)
    MDSS 13 (8–15) 21 (16–23)
    SIBDQ 62 (58–64) 63 (54–65)
   Medications[73]^b
    No medications, n (%) 0 1 (13)
    5-ASA, n (%) 8 (100) 6 (75)
    Immunomodulator, n (%) 0 1 (13)
    Biologic, n (%) 0 1 (13)
   [74]Open in a new tab

   5-ASA, 5-aminosalicylic acids; IQR, interquartile range; MDP,
   Mediterranean Diet Pattern; SCCAI, Simple Clinical Colitis Activity
   Index; SIBDQ, Short Inflammatory Bowel Disease Questionnaire.
   ^a

   Δ in fecal calprotectin = change from baseline to week 13.
   ^b

   Participants were presribed one or more therapy.

   To identify which patients responded more effectively to the MDP
   intervention, we subset the microbiome and metabolome data from the MDP
   group for more refined analysis. To distinguish responders (RS) from
   nonresponders (NS), we defined RS as those achieving or sustaining
   remission, characterized by a 50% reduction in FCP from baseline to
   week 12, or sustained remission (FCP <50) from baseline to week 12
   accompanied by a decrease or no change in disease activity according to
   the Simple Clinical Colitis Activity Index.[75]^28

Metabolomics Analysis by Ultra-High Performance Liquid Chromatography-Tandem
Mass Spectrometry

   Fecal samples from week 12 (final samples) were analyzed by Agilent
   1290 UPLC attached to a Q-TOF (Agilent 6530B) (Agilent, CA, USA) with
   dual Agilent Jetstream electrospray source. Analyte separation was
   achieved using a Waters BEH-C18 column (2.1 × 100 mm and 1.7 um;
   Milford, MA, USA). For dual-ion data acquisition, a gradient-elution of
   0.05% aqueous formic acid with 5-mM ammonium formate (Solvent A) and
   acetonitrile with 0.05% formic acid (Solvent B) was used. The linear
   gradient of positive electrospray ionization (ESI+) mode was described
   as follows: 0 minutes: 4% B; 2 minutes: 4% B; 8 minutes: 40% B; and 16
   minutes: 96% B; 24 minutes: 100% B; 38 minutes: 100% B; 39 minutes: 4%
   and negative (ESI-) mode was: 0 minutes: 4% B; 2 minutes: 4% B; 8
   minutes: 40%; 16 minutes: 90%; 24 minutes: 100%; 38 minutes: 100%; 39
   minutes: 4% B. For MS analysis, the scan range was set to full MS—we
   selected ion monitoring with a scan range between 85–950 m/z in ESI+
   and 85–1050 m/z in ESI-. The Jetstream electrospray source was run in
   both ESI+ and ESI- ionization mode with a sheath gas temperature of
   325 °C, capillary voltage is set as 3000 volts for ESI+ and 4000 volts
   for ESI-, nozzle voltage was set as 1000 volts in ESI+ and 2000 volts
   in ESI-, the fragmentor voltage was 120 volts for both ion modes. Mass
   correction was done automatically through MassHunter software using the
   standard Agilent reference masses of 121.050873 and 922.009798 in ESI+
   and 112.985587 and 1033.988109 in ESI-.

Data Processing and Annotation

   Data (ESI+ and ESI-) were preprocessed with the online XCMS
   ([76]https://xcmsonline.scripps.edu/) platform for peak annotation. The
   parameters were selected as: 10 ppm, mzdiff = 0.01, retention time
   correction method = obiwarp with all other parameters set as default.
   The data matrix containing variables (retention time, m/z, peak
   intensity) was obtained for further annotation against the human
   metabolome database.[77]^29

Pathway Enrichment Analysis

   For pathway analysis, we utilized the human and prokaryote Kyoto
   Encyclopedia of Genes and Genomes (KEGG) database integrated within
   MetaboanalystR 4.0[78]^30 to map metabolites from both host and
   microbial communities. For pathways showing statistically significant
   differences between groups, the annotated compounds within these
   pathways were extracted and visualized using GraphPad Prism v10.0
   (GraphPad Software, Boston, Massachusetts, USA, [79]www.graphpad.com)

Weighted Gene Coexpression Analysis (Weighted Correlation Network Analysis
[WGCNA])

   Correlation analysis of metabolites was performed using WGCNA,[80]^31
   with a soft thresholding power range from 1 to 20. The scale-free
   topology model fit (Rˆ2) of 0.6 was achieved at a power of 14. The
   features were clustered into modules by hierarchical clustering with
   Euclidean distance and a minimum module size of 10. Metabolites within
   each cluster were then analyzed for KEGG pathway enrichment using
   MetaboanalystR 4.0.[81]^30

Fecal Microbiome Analysis

   The sequencing and bioinformatics analyses were previously
   described.[82]^22 Briefly, fecal DNA was extracted using the QIAamp
   PowerFecal Pro DNA kit (Qiagen, #51804). Sequencing was performed by
   the Gut4Health Microbiome Core Facility (BC Children’s Hospital
   Research Institute, Vancouver, BC), targeting the V4 hypervariable
   region of the 16S rRNA gene. Amplification was conducted on the
   Illumina MiSeq platform using the following primers: Forward
   5′-AGTCAGTCAGCCGGACTACNVGGGTWTCTAAT and Reverse
   5′-AGTCAGTCAGCCGGACTACNVGGGTWTCTAAT, which included the Illumina
   adapter overhang.

   After sequencing, analyses were carried out using the QIIME 2 platform
   (version 2021.4).[83]^32 Demultiplexed reads from 2 MiSeq runs were
   imported into QIIME 2, and primers were removed using the q2-cutadapt
   plugin.[84]^33 Quality control involved filtering, dereplication,
   chimera removal, denoising, and merging paired-end reads for each run
   separately using the DADA2 plugin with default parameters.[85]^34 The
   resulting amplicon sequence variant (ASV) tables were merged for
   downstream analysis. A phylogenetic tree was constructed using the
   SATé-enabled phylogenetic placement technique via the
   q2-fragment-insertion plugin,[86]^35 utilizing a backbone tree based on
   the Greengenes reference database (version 13.8).[87]^36

   For taxonomic classification, a classifier trained on the full length
   of the 16S region was applied, incorporating environment-specific
   abundance weights tailored to human fecal samples from the readytowear
   tool.[88]^37 This weighted approach enhanced classification accuracy
   compared to standard Naive Bayes methods.[89]^38 Alpha diversity
   metrics were calculated, including ASV richness, Shannon’s diversity
   index, and Faith’s phylogenetic diversity.[90]^32 Beta diversity was
   assessed using compositional tensor factorization to account for
   interindividual variation over time. ANOVA was applied to the
   compositional tensor factorization results to assess differences
   between dietary groups.[91]^38

   To account for the longitudinal design of the study (baseline and week
   12 fecal samples, with subject-level variation), the BIRDMAn tool
   [[92]https://birdman.readthedocs.io/en/stable/index.html] was employed
   to rank ASVs most associated with each group. These rankings were
   visualized using Qurro,[93]^39 and the top 10 ASV differentials for
   each group were exported into R[94]^40 via the qiime2R package
   [[95]https://github.com/jbisanz/qiime2R] for additional custom
   visualization and statistical analysis. BugBase
   [[96]https://doi.org/10.1101/133462] was used to predict high-level
   phenotypes, assessing the proportions of Gram-positive, Gram-negative,
   aerobic, anaerobic, facultative anaerobic, biofilm-forming, and mobile
   element-containing bacteria.

Statistical Analysis

   Samples were normalized by median to correct for systematic
   differences, and then logarithmic transformation (log10) was applied to
   normalize for data distribution. Pareto scaling was applied to mitigate
   the influence of differing scales among features. Next, annotated
   metabolites were analyzed using a significance threshold of fold change
   (FC) > 2 and P value <.05. Unsupervised principal component analysis
   (PCA) and supervised partial least squares discriminant analysis
   (PLS-DA) were used to visualize data variances by R 4.2.0 and ggplot2.
   PLS-DA cross-validation showed the top three components. Scores were
   plotted for each pairwise comparison and statistical differences were
   determined by the Wilcoxon rank-sum test using MetaboanalystR
   4.0.[97]^30

Results

   In liquid chromatography-mass spectrometry, ESI is used to detect and
   analyze metabolites. There are two common modes, positive (ESI+) and
   negative (ESI-), with each mode able to detect different metabolites
   based on their chemical properties. A total of 4279 peaks were detected
   in the (ESI+) mode and 7841 peaks in the ESI negative (ESI-)
   ([98]Table A1). In ESI+, 242 features were upregulated (FC >2), and 279
   (FC <0.5) were downregulated in the MDP compared to control.
   Statistical analyses using a t-test (FC >2 and FC <0.5, P <.05)
   identified 156 significant features (19 upregulated and 45
   downregulated features in the MDP) ([99]Table A2). In the ESI- mode,
   324 features were upregulated, and 208 were downregulated in the MDP.
   T-test identified 72 significant features, with 41 upregulated and 31
   downregulated features in MDP based on FC and t-test criteria
   ([100]Table A3).

   While PCA did not show clear separation between the 2 groups, PLS-DA
   revealed clear stratifications between groups ([101]Figure 1A and B). A
   volcano plot (FC >2, P < .05) revealed 27 upregulated peaks and 59
   downregulated peaks ([102]Table A4, [103]Figure 1C). The top 50
   variable peaks are shown in [104]Figure 1D and E shows an upregulation
   of organic acids while nucleosides, nucleotides, and analogu are among
   the downregulated molecules. Metabolites with a variable importance in
   projection (VIP) score cutoff above 1.25 are shown in [105]Figure A1.

Figure 1.

   [106]Figure 1
   [107]Open in a new tab

   MDP dietary intervention alters the fecal metabolome. PCA (A) and
   pl-SDA (B) analysis of the metabolites from patients with either CHD or
   MDP. (C) Volcano plot demonstrates altered peak features between MDP
   and CHD (P < .05, FC >2). (D) Top 50 significantly altered peak
   features between CHD and MDP group. (E) Chemical classification of
   upregulated and downregulated metabolites in the MDP group.

Identification of Metabolites Associated with Consumption of the MDP

   A detailed analysis using the human KEGG database revealed a total of
   701 human metabolites. Of these, 35 showed significant differential
   expression between the MDP and CHD groups ([108]Figure 2B). The top 25
   differential pathways are shown in [109]Figure 2A, with the two six
   showing significance (false discovery rate <0.05). By segregating the
   upregulated compounds (FC >2) and downregulated compounds (FC <0.5)
   into distinct subsets regardless of the P value, notable findings
   emerged. The upregulated and downregulated compounds exhibited
   significant coverage in distinct pathways as shown in [110]Figure 2C.

Figure 2.

   [111]Figure 2
   [112]Open in a new tab

   Pathway analysis of metabolites altered in patients receiving the MDP
   intervention. (A) Overview of enriched metabolite pathways. (B) Volcano
   plot depicts the significantly altered host metabolites (P < .05). (C)
   Bubble plot illustrating the regulation of metabolic pathways. The
   x-axis represents the direction of regulation (downregulation or
   upregulation), while the y-axis lists the various metabolic pathways.
   The bubble size corresponds to the coverage (%), and the color gradient
   indicates the false discovery rate with darker shades representing
   lower values. MDP decreased the major metabolites in sphingosine
   biosynthesis (D), steroid hormone biosynthesis (E) and increased folate
   biosynthesis (F).

   Sphingosine-1-phosphate (S1P) mediates the inflammatory response by
   binding to a family of S1P receptors, which are involved in various
   cellular processes, including lymphocyte trafficking, endothelial
   barrier function, and cascade of inflammatory
   responses.[113]^41^,[114]^42 It is worth noting, compounds such as
   4-hydroxy sphinganine, phytosphingosine, sphingosine,
   L-threo-sphingosine, and dihydro-sphingosine, and 7-hexadecenoic acid
   were significantly downregulated in the MDP as shown in [115]Figure 2D.
   The decreased S1P pathway parallels the observed reduction in stress
   steroid hormones including cortisol, 21-deoxycortisol, corticosterone,
   and aldosterone in MDP group, while 11-dehydrocorticosterone was
   upregulated ([116]Figure 2E). Steroid hormones, in particular cortisol,
   are produced locally in the intestinal mucosal tissues in response to
   immune cell activation and in turn, contribute to balancing these
   immune responses to avoid tissue damage.[117]^43

   Folate biosynthesis is the top enriched pathway across all detected
   metabolites ([118]Figure 2F). While folate itself did not significantly
   differ (P = .3861), downstream metabolites like sepiapterin (P = .0064)
   and 7,8-dihydroneopterin (P = .0064) showed significant increases in
   those following the MDP. These results are relevant not only for
   inflammatory conditions but also for the gut–brain–microbiome
   axis.[119]^44 [120]Figure 2C shows an upregulation in tyrosine and
   tryptophan metabolism. However, serotonin, kynurenine, and dopamine did
   not show significant differences across dietary patterns.

   Additionally, we observed environmental chemicals and food additives in
   our cohort. [121]Figure 3 highlights the top 10 metabolites (Log2 FC
   >1, P < .05). Among these compounds, several are commonly found in
   various food products and environmental sources. The potential health
   effects of these specific compounds remain unclear, and there are
   limited reports of their detection in human tissue.

Figure 3.

   [122]Figure 3
   [123]Open in a new tab

   Environmental chemicals or food additives that are upregulated with
   consumption of the MDP. Log2fold change of environmental chemicals are
   shown in red, and food additives are shown in blue.

Differences in Metabolites Associated with Consumption of the Mediterranean
Diet Pattern Between Responders and Non-Responders

   [124]Table A5 show the characteristics of the responders (RS) and
   non-responders (NS) to the MDP. The PCA analysis ([125]Figure 4A)
   demonstrates a clear separation between the two groups, while PLS-DA
   reveals an even more distinct differentiation ([126]Figure 4D),
   indicating a distinct metabolic profile in the context of MDP
   treatment. Further investigation into RS and NS to the MDP intervention
   revealed differences in 109 annotated metabolites ([127]Table A4), with
   19 unique peaks identified in the human KEGG database ([128]Figure 4B).
   Upregulated metabolites are shown in [129]Figure 4C. VIP scores for the
   top 15 metabolites contributing to the variation in metabolic profiles
   between NS and RS samples are shown in [130]Figure 4E. The metabolites
   that meet the VIP score cutoff above 1.25 are shown in [131]Figure A2.

Figure 4.

   [132]Figure 4
   [133]Open in a new tab

   Patients who respond to the MDP have a distinct fecal metabolomic
   profile. (A) PCA plot demonstrates a clear separation between RS and
   NS. (B) Volcano plot highlights the statistically significant
   metabolites between the two groups. (C) Pathway enrichment analysis of
   the differentially expressed metabolites, showing the biochemical
   pathways that influence RS and NS. (D) PLS-DA analysis further confirms
   the separation of the RS and NS groups based on their metabolic
   profiles. (E) VIP plot from the PLS-DA analysis identified the top 15
   important metabolites driving the group separation.

   Further WGCNA was analyses identified 28 modules, grouped into 3
   clusters ([134]Figure 5A and B). Cluster 1 increased in RS treatment
   compared to NS, cluster 2 decreased, and cluster 3 remained unchanged
   ([135]Figure 5C). Metabolites from clusters 1 and 2 were separately
   analyzed for pathway enrichment. Significantly upregulated and
   downregulated pathways (false discovery rate <.05) are shown in
   [136]Figure 5E. The chemical superclasses of upregulated and
   downregulated metabolites are shown in [137]Figure 5D.

Figure 5.

   [138]Figure 5
   [139]Open in a new tab

   WGCNA analysis of metabolic differences between RS and NS following the
   MDP. (A) WGCNA dendrogram classifies the metabolites into distinct
   modules based on their expression patterns. (B) Eigengene network
   analysis illustrates the correlation of each module with the RS and NS
   groups, with modules categorized into 3 clusters based on their module
   alteration patterns between RS and NS, summarized in (C). (D) Chemical
   categorization of upregulated metabolites from cluster 1 and
   downregulated metabolites from cluster 2. (E) Pathway enrichment
   analysis showing the biochemical pathways associated with the
   upregulated and downregulated metabolites from cluster 1 and 2,
   respectively.

Non-responders to the MDP Have Less Bacteroidia in Their Fecal Microbiome

   We investigated the microbial differences between RS and NS to the MDP.
   Our findings indicate that individuals with reduced levels of the class
   Bacteroidia in their stool at baseline were less likely to respond
   positively to the MDP, as illustrated in [140]Figure 6A. Moreover, we
   focused on individual species to investigate variations within these
   microbial groups at baseline between RS and NS. By filtering out taxa
   with extensive zero values across samples, we identified 6 relevant
   taxa for comparison ([141]Figure 6B). Among these taxa, known fiber
   degraders like Bacteroides acidifaciens, B. vulgatus, B. uniformis,
   were observed in higher abundance levels in patients who responded
   favorably to the dietary intervention before the treatment.

Figure 6.

   [142]Figure 6
   [143]Open in a new tab

   Microbiome composition as a predictor of MDP responsiveness. (A)
   Comparison of relative abundance of Bacteroidia class between RS and NS
   at baseline. (B) Relative abundance analysis of all species and genus
   within the Bacteroidia class.

Discussion

   To our knowledge, this is the first published report to profile the
   fecal metabolome from a UC RCT following an MDP. We uncovered
   metabolites linked to 3 crucial pathways—sphingosine, steroid, and
   folate biosynthesis— which correlated with high adherence to an MDP.
   Increasing observational data from population-based research has
   revealed significant alterations in IBD patients' gut microbial
   structure and metabolic profiles.[144]^45^,[145]^46 Recently, a study
   in a pediatric cohort with UC found alterations in sphingolipid and
   folate biosynthesis metabolic pathways,[146]^46 and a large
   observational cohort observed a marked increase in the fecal levels of
   sphingolipids, including several sphingomyelins and ceramides.[147]^45
   It is important to note that these studies, being observational in
   nature, did not control for dietary variables or patterns, unlike our
   study, which accounted for these factors. Metabolomic studies in IBD
   reveal alterations in tryptophan metabolism among patients.[148]^47
   Although we initially detected a signal in our dataset, further
   investigation did not uncover significant differences between our
   groups.

   We noted a significant decrease in fecal sphingolipids (4-hydroxy
   sphinganine, phytosphingosine, sphingosine, L-threo-sphingosine, and
   dihydro-sphingosine, and 7-hexadecenoic acid) in the MDP. Sphingolipids
   are integral components of the intestinal cell membrane, synthesized
   either through the de novo condensation of serine with palmitoyl-CoA or
   via the absorption of endogenous and dietary sphingolipids.[149]^48 S1P
   serves as a central player in orchestrating the inflammatory cascade.
   It achieves this by binding to a family of S1P receptors, which govern
   diverse cellular processes, including lymphocyte trafficking,
   endothelial barrier function, and the initiation of inflammatory
   responses.[150]^41^,[151]^48 The upregulation of the S1P signaling
   pathway is implicated in the pathophysiology of IBD, contributing to
   both inflammation and barrier dysfunction.[152]^49 In the context of UC
   treatment, S1P inhibitors may be used in select cases to impede the
   trafficking of lymphocytes from lymph nodes, diminishing the pool of
   inflammatory cells available to migrate to the intestinal mucosa and
   facilitating gut inflammation resolution.[153]^49 We hypothesize that
   the MDP could potentially participate in S1P inhibition, and may be one
   mechanism of how the diet exerts its anti-inflammatory effects.
   Remarkably, the MDP responders experienced an additional downregulation
   of metabolites of the S1P pathway (ie, 3-dihydrosphinganine and
   sphingosine).

   Stress hormones, cortisol and corticosterone, have been associated with
   intestinal barrier dysfunction and can trigger flares in IBD.[154]50,
   [155]51, [156]52 Stress directly activates the hypothalamus to release
   corticotrophin-releasing factor (CRF). This induces the anterior
   pituitary gland to secrete adrenocorticotrophic hormone, further
   stimulating the adrenal cortex to secrete cortisol.[157]^53 CRF can
   induce mast cell degranulation and increase mucosal
   permeability.[158]^54 In our study, participants adhering to the MDP
   had a reduction in corticosterone and cortisol levels. The reduction in
   metabolites associated with steroid biosynthesis suggests the MDP
   modulates cortisol, supporting its anti-inflammatory effects. Notably,
   corticosterone and its downstream inactive metabolite,
   11-dehydrocorticosterone, exhibited additional reductions in patients
   who responded positively to the MDP.

   Patients with IBD are at a higher risk of folic acid deficiency, as
   they frequently avoid food products, such as fresh fruits and
   vegetables, which are the main sources of folic acid. Additionally, the
   use of sulfasalazine by patients may result in folic acid
   deficiency.[159]^55 On the contrary, high levels of folate reduce the
   risk of IBD and colon cancer.[160]^56^,[161]^57 The increase in folate
   biosynthesis may be linked to increased consumption of folate-rich
   foods in the MDP, measured by increased consumption of vegetables,
   fruit, and legumes.[162]^22^,[163]^27 Folate plays a crucial role in
   DNA methylation, red blood cell formation, and reduction in cancer
   risk.[164]^58^,[165]^59 One of the key downstream metabolites of the
   folate pathway that upregulated in the MDP was sepiapterin. In a rodent
   model of colitis, sepiapterin administration significantly reduced the
   number of infiltrating inflammatory macrophages and neutrophils and the
   expression of proinflammatory cytokines interleuckin-1β,
   interleuckin-6, and interleuckin-17A.[166]^60 Additionally, sepiapterin
   treatment was associated with a decrease in the number of tumors. The
   anti-inflammatory and antitumor effects of sepiapterin must be further
   investigated in humans.

   We saw several unexpected metabolites upregulated in the stool of the
   MDP group, which included pesticides (tert-butyl carbamate), herbicides
   (dimethyl glutarate, 2,2-dimethylglutaric acid), environmental
   chemicals (mycotoxins; Austalide A and Austalide A, propyleneglycol
   diacetate, N-vinylimidazole) and food additives and flavorings
   (ascorbyl palmitate, anisole, gluconic acid). It is important to
   consider that while the MDP is generally associated with positive
   health outcomes, these metabolites indicate that even healthy dietary
   patterns can introduce certain chemicals into the body. A recent study
   that compared urinary pesticide metabolite levels between conventional
   and organic foods revealed statistically significant disruptions in
   metabolic pathways associated with inflammation, oxidative stress, and
   the requirements of xenobiotic detoxification.[167]^61 The consumption
   of pesticides has been linked to increased levels of oxidative stress
   and inflammation, potentially contributing to various diseases,
   including cancer.[168]^61 Despite the growing body of literature about
   the adverse health consequences of pesticide and food additive exposure
   in recent years, studies focusing on human exposure through the
   environment and dietary intake remain limited.

   There is considerable inter-individual variation in response to
   nutritional interventions. Therefore, some interventions may benefit
   certain individuals or population subgroups more than others, depending
   on their genotype, phenotype, and environment.[169]^62 Here, we
   identify microbiome-derived biomarkers (taxon) that differ between RS
   and NS. At the class level, we observed a reduction in Bacteroides. At
   the genus and species level, responders appeared to have an increased
   abundance of Bacteroides vulgatus (B. vulgatus), B. uniformis, and B.
   acidifaciens. B. vulgatus, B. uniformis and B. acidifaciens, all of
   which have been shown to have a protective effect on colitis, including
   alleviating the colitis phenotype, reducing inflammatory response, and
   improving intestinal barrier function in animal models.[170]63,
   [171]64, [172]65 Bacteroidetes (in particular, B. acidifaciens) have an
   important role in gut microbiota by producing high levels of
   short-chain fatty acids, including acetate, propionate, or butyrate.
   Many experimental animal data suggest that their high levels can
   discontinue the progress of several inflammatory, autoimmune and
   allergic diseases.[173]^66 NS had an increased abundance of
   Parabacteroides merdea and Ordoribacter splanchnicus. On the contrary,
   Parabacteroides merdae was increased in human patients with
   hypertension and polycystic ovary syndrome[174]^67 Ordoribacter
   splanchnicus has been identified as an opportunistic pathogen with
   potential significance for diagnosing and treating UC.[175]^68 Although
   these findings need to be confirmed in a larger sample, these microbes
   provide a potential hypothesis as to why individuals respond or do not
   respond effectively to the MDP.

   Our study has several strengths. Firstly, we employed a nontargeted
   approach to provide a comprehensive and detailed view of the fecal
   metabolome. This methodology enabled us to illustrate the relative
   contribution of each pathway to the overall metabolic alterations.
   Despite the small sample size, our findings are biologically plausible,
   and the identified biological pathways and functions align with those
   described in previous studies. As a hypothesis-generating study, the
   sample size restricts our ability to adjust for all potential
   confounders. such as medication use variations, other chronic
   comorbidities, and different environmental influences, whose impacts on
   metabolic profiles remain unknown. Furthermore, the cross-sectional
   design prevented us from observing longitudinal changes in metabolomic
   profiles and limited us to reporting associations between diet and
   omics measurements at a single time point. Future studies with larger
   sample sizes and a longitudinal approach will be necessary to identify
   alterations in UC more definitively.

   In conclusion, this study offers valuable insights into an adult UC
   population’s gut microbiome and fecal metabolome that responds to an
   MDP. We identified several key metabolites and pathways associated with
   the diet's anti-inflammatory effects. Future research involving larger,
   longitudinal studies is essential to further validate these results and
   explore the therapeutic potential of dietary interventions in managing
   IBD.

Acknowledgments: