Modified Crohn's Disease Exclusion Diet and exclusive enteral nutrition (EEN) resolve oral dysbiosis in pediatric Crohn's disease: a prospective cohort study

Abstract

Background and Aims

Changes to the oral microbiome have been reported in patients with Crohn’s disease (CD). The aim of this study was to determine the characteristics and dynamics of the oral microbiome in children randomized to 1 of 2 nutritional treatments for CD.

Methods

Participants (n = 54) in this randomized controlled trial (NCT02843100) received the Crohn’s Disease Exclusion Diet (CDED) with either partial enteral nutrition (PEN; n = 28) or exclusive enteral nutrition (EEN; n = 26). The oral microbiome was assessed via swabs from the dorsum of the tongue and the buccal gingiva by 16S rRNA sequencing at 0, 2, 8, 14, 24, and 52 weeks.

Results

There were no significant differences in primary or clinical outcomes between the 2 groups. Due to the COVID-19 pandemic, sampling by 24 weeks was limited to 34 participants. At week 0, moderate-severe disease activity (Pediatric Crohn’s Disease Activity Index [PCDAI] > 30) was associated with decreased Porphyromonas, Haemophilus, Alloprevotella, Neisseria, and Bergeyella species and increased Actinomyces. As patients entered remission (PCDAI < 10), we observed a restoration in the abundance of these taxa. A modified oral dysbiosis index (MODI) was generated, capable of distinguishing mild from moderate-severe disease activity based on microbiome profiles. Dysbiosis decreased as treatment continued and patients entered remission. Patients on CDED exhibited more significant dysbiosis index changes at weeks 8-24, compared with the EEN group. Application of the index across published oral microbiome data sets validated its ability to discriminate health from CD.

Conclusion

Oral microbiome changes in pediatric CD reflect disease activity and parallel therapeutic response to CDED and EEN over time. Additional validation of the proposed dysbiosis index should be undertaken in adequately powered future studies.

Summary

The oral microbiome was examined in children randomized to 2 nutritional treatments for Crohn’s disease (CD). At baseline, specific oral microbiome changes were associated with disease severity. Following treatment, these microbiome changes were reversed, and this reversal could predict remission.

Key Messages.

• Changes to the oral microbiome have been described in pediatric CD

• In this study we examined the oral microbiome in an international cohort and examined how CD activity and the oral microbiome responded to nutritional therapy

• We demonstrated that the oral microbiome structure reflects disease activity in this international pediatric CD cohort

• The oral microbiome responds to nutritional therapy and these changes can predict remission

• Future validation may determine if the oral microbiome could be used to monitor disease activity

Introduction

The human microbiome is integral to immune regulation and mucosal homeostasis. Perturbations in microbial communities, termed dysbiosis, have been implicated in the pathogenesis of a range of chronic inflammatory and immune-mediated disorders, including malignancy, atopy, and inflammatory bowel disease (IBD).^1,2 In IBD, dysbiosis is thought to disrupt the host–microbiota equilibrium, leading to inappropriate immune activation and sustained mucosal inflammation.^3–5 Although the precise mechanisms remain incompletely defined, alterations in microbial composition and function are increasingly recognized as central to disease initiation and progression.^6,7 Crohn’s disease (CD), a prototypical form of IBD, has a multifactorial etiology involving changes in environment, host genetics, and possibly oral-gut transit of microorganisms.^4,5,8,9 Recent studies have extended the focus beyond the gut lumen to the oral cavity, proposing a potential role for the oral-gut axis in disease pathogenesis.^9–11 Human studies have demonstrated distinct oral microbial signatures between controls and patients with CD, particularly among those with oral manifestations such as aphthous ulcers and mucosal cobblestoning.^11–13 Interstudy variability in microbial profiling, sampling techniques, and patient cohorts has limited consensus on the precise nature and clinical relevance of oral dysbiosis in CD.^12–19 The oral microbiome nonetheless offers a unique, noninvasive window into mucosal immune activity. Emerging evidence suggests that oral microbial shifts correlate with intestinal inflammation and treatment response, raising the possibility of the of the oral microbiome as a biomarker and therapeutic target in IBD.^8,16,19,20

We previously identified significant dysbiosis of the tongue microbiome in a cohort of treatment-naive pediatric patients with IBD, with alterations correlating closely with disease severity and resolving following induction of remission.¹³ Notably, this study focused on mucosal rather than salivary samples, allowing for a more anatomically and functionally relevant characterization of the oral microbial niche.

Despite growing interest, longitudinal data on the oral microbiome in pediatric CD, particularly in the context of nutritional therapy, are lacking. The present study aimed to characterize oral microbiome dynamics over time in children with CD enrolled in a multicenter, international, randomized controlled trial of the Crohn’s Disease Exclusion Diet (CDED) and exclusive enteral nutrition (EEN). We investigated the temporal evolution of the oral microbiome in pediatric CD following treatment, whether microbial changes mirrored clinical disease activity and assessed the impact of dietary modulation on oral dysbiosis.

Materials and methods

Study design

The current study was undertaken as part of the DIETOMICS study (clinicaltrials.gov: NCT02843100), a randomized trial which was previously described in detail.²¹ Briefly, in this open-label trial, patients aged 8-18 years with either established or newly diagnosed mild-to-moderate CD were randomized to 1 of 2 treatment arms: (a) 2 weeks of exclusive enteral nutrition (EEN; Modulen IBD) followed by 3 phases of the Crohn’s Disease Exclusion Diet (CDED) + partial enteral nutrition (PEN; Modulen IBD) up to week 24 (group 1), or (b) standard care with 8 weeks of EEN followed by free diet + 25% PEN (group 2). Maximal follow-up was at week 52. For eligibility, a diagnosis of inflammatory-phenotype CD (Paris B1) based on clinical, endoscopic and histological features as per established criteria was required with mild-moderately active disease and diagnosis duration of < 36 months.²² Disease activity was defined by the PCDAI score as follows: remission < 10; mild 11-30; moderate 31-50; severe ≥ 51.23 Patients were excluded if the disease was limited to the distal colon, if they had perianal disease, and if they had undergone past/current biologic therapy or current/recent steroid therapy. The study was conducted across sites in Israel, Canada, Spain, Ireland, and Sweden. Comprehensive structured data collection was undertaken at specified timepoints throughout the study using study-specific case report forms. Each site obtained local ethical approval, and written informed consent was required for each participant. The primary end point was steroid-free remission by week 14, and treatment failure was defined as failing to achieve clinical remission by week 8. Immunomodulator maintenance therapy with thiopurines or methotrexate was permissible after 4 or 5 weeks, respectively.

Swabbing protocol

The oral microbiome was sampled using a polyurethane sponge swab (CultureSwab EZ, Becton Dickinson) to sample the dorsum of the tongue and the buccal gingiva in all participants. Written instructions and accompanying audiovisual education were provided for each participating site to standardize the approach to oral mucosal swabbing. Swabs were transferred to −80 °C immediately and processed within 6 months.

Swabs were collected from all patients at baseline, during screening visits, before enrollment and randomization to either treatment arm. Swabs were taken at follow-up time points from patients at 2, 8, 14, 24, and 52 weeks post enrollment (Table S2).

DNA extraction and sequencing

DNA was extracted from mucosal swabs using the MasterPure Complete DNA/RNA Purification Kit (Epicentre Biotechnologies). The manufacturer’s extraction protocol was supplemented with an additional incubation step with Ready-Lyse lysozyme (Epicentre Biotechnologies) and a bead disruption step, as described previously.¹³ The DNA was resuspended in Tris-EDTA buffer (pH 7.5) and stored at −80 °C.

Amplification of the V1-V2 region of the 16S rRNA gene was carried using the KAPA HiFi Hot start system (KapaBiosystems) with the primers 27F-YM and 338R-R (27F-YM: 5′-TCGTCGGCAGCGTCAGATGTGTATAAGAGACAGAGTCAGTCTGTCAGAGTTTGATYMTGGCTCAG; 338R-R: 5′-GTCTCGTGGGCTCGGAGATGTGTATAAGAGACAGTATGGTAATTCATGCTGCCTCCCGTAGRAGT).²⁴ The V1-V2 region was selected based on the finding that approximately 90% of species in the Human Oral Microbiome Database (HOMD) can be correctly identified using these sequences.²⁵ Sample indexing was carried out with the Nextera XT Index Kit (Illumina) and library quantification and purification was carried out according to Illumina protocols. Size and integrity of indexed amplimers was determined using a Bioanalyzer (Agilent Technologies) and samples were normalized to 4 nM. Samples were combined to generate a pooled library, denatured, and combined with PhiX control DNA (5%) and loaded at a concentration of 6 ppm. Paired-end sequencing was performed using the Illumina 500 cycle MiSeq reagent kit. All sequence data have been submitted to the NCBI sequence read archive (SRA).

Data analysis

Bacterial 16S rRNA sequences were processed and filtered using Dada2 with the following parameters: maxN = 0, maxEE = c(2,5), and truncQ = 2.²⁶ The 10 terminal bases were trimmed from reverse reads due to the drop in sequence quality. Following error estimation and correction using the Dada2 algorithm, paired reads were merged and chimeras were removed following their identification with the remove BimeraDenovo command. Taxonomy was assigned using the Human Oral Microbiome Database (HOMD) classifications (eHOMD 16S rRNA RefSeq version 15.1).²⁷

Phyloseq was used to calculate alpha diversity and richness indices for each sample.²⁸ For beta diversity analysis, the Aitchison distance metric was used by applying PCA to the center log-ratio (CLR)–transformed counts using the package “microbiome” in R studio. Further visualizations (Iris plots, biplots) and analysis of community structure, including PERMANOVA, were conducted using the package “MicroViz” in R studio.^29,30 Multivariate analysis was conducted with MaAsLin 2 using normalized counts and the indicated patient metadata variables as “fixed-effects” and site as a “random effect.” ³⁰ Further differential analysis was carried out using a combination of methods including MicrobiotaProcess, which applies a nonparametric Wilcoxon test on relative abundances.^31,32 All P values were adjusted for multiple comparisons by use of the Benjamini-Hochberg correction.

Dysbiosis index

A modified oral dysbiosis index (MODI) was generated based on the analytic approach used to previously describe intestinal dysbiosis in pediatric CD.⁶ The MODI is the log of [total abundance in organisms increased in IBD] over [total abundance of organisms decreased in IBD] for all samples. Based on empirical analysis of the results from MaAsLin 2, the ratio of Bacilli, Actinobacteria, and Flavobacteriia over Clostridia, Gammaproteobacteria, Betaproteobacteria, and Bacteroida was chosen to generate the MODI presented here.

Validation of the index was carried out in independent populations using data from a unified bioinformatic meta-analysis of the oral microbiome in patients with IBD carried out by co-author R.J.W., including previously published paediatric data.^13,33 The analysis included data from the prospective longitudinal MUSIC study (NCT04760964) and 6 additional published studies of the oral microbiome in adult IBD patients (total IBD patients = 259, controls = 224).^11,34–37 The raw 16S rRNA sequencing files from the 7 datasets were downloaded, preprocessed and taxonomically profiled using a unified bioinformatics framework. The nf-core ampliseq version 2.14.0 pipeline was used in combination with DADA2, and data were aggregated into a phyloseq object, after which the MODI was applied.

Results

Oral dysbiosis in CD

The clinical details and the results of the clinical outcomes of the trial were published in detail previously.²¹ A total of 61 patients (aged 8-18 years) from 7 centers (Barcelona, Dublin, Edmonton, Malaga, Stockholm, Tel Aviv, Vancouver) were initially sampled at the study screening stage; of these, 28 were randomized to CDED with PEN (group 1) and 26 to EEN (group 2). Clinical characteristics of participants in the oral microbiome study are presented in Table S1. Unfortunately, it was necessary to halt the trial due to the impact of the COVID-19 pandemic, as intensive in-person visits and oral swabbing were not feasible. This situation negatively impacted adherence to the oral sampling follow-up protocol of already recruited patients, reducing the power and interpretation of the study accordingly. Tongue and gingival swab samples were recovered from all participants and were analyzed in parallel. At baseline, patients from all centers had similar core population structures (genus level) on analysis of relative abundance and principal coordinate analysis (PCoA) analysis (Figure 1). Geography had no significant impact on the tongue microbiome profiles (PERMANOVA P = .154); however, gingival profiles exhibited greater geographic variation (PERMANOVA P = .003). For comparison, swabs from a group of age-matched healthy controls (n = 108) were obtained during the same time period (Figure 2). In keeping with our previous discoveries in CD, we discerned a similar tongue dysbiosis at baseline in CD relative to controls, with significantly increased levels of Actinomyces spp. (padj = 2.9 × 10⁻⁶; Log2 FC 1.17) (Figure 2A-C) and also reduced abundance of Clostridia (Stomatobaculum [padj = 0.029; Log2 FC −2.23], Catonella [padj = 0.006; Log2 FC −1.9], Oribacterium [padj = 0.046; Log2 FC −1.29], and Lachnoaerobaculum [padj = 0.2; Log2 FC −1.0]) (Figure 2A). Patients from Dublin (n = 10) exhibited increased levels of Bacilli (ie, Staphylococcus and Streptococcus spp.), in keeping with our previous findings in Irish IBD patients (Figure S1). However, in contrast, we observed reduced abundance of Bacilli (ie, Streptococcus, Gemella, and Granulicatella spp.) when all centers were analyzed (Figure 2D).

Figure 1.

Structure of the oral microbiome in patients with Crohn disease (CD) from different cities. (A) Bar graph showing relative abundance of genera in tongue samples and associated principal coordinate analysis (PCoA) plot of Bray-Curtis dissimilarity values. (B) Bar graph showing relative abundance of genera in gingival samples and associated PCoA plot of Bray-Curtis dissimilarity values.

Figure 2.

Oral microbiome dysbiosis in tongue (A) and gingival (B) samples from all CD patients compared to healthy controls. Heatmaps show the results of multivariate analysis with MaAsLin2 investigating the impact of diagnosis (Dx) controlled for age and sex. Taxa in red indicate show increased abundance in Crohn disease (CD) (Dx) or with increasing age or male sex, blue indicates reduced abundance in that category. The legend and color intensity reflect effect size (see methods). (C) and (D) show box plots of the relative abundances of Actinomyces spp. and Streptococcus spp. in tongue and gingival samples from CD patients and controls.

Dysbiosis and disease activity

We next examined whether the level of clinical disease activity in patients, measured using the PCDAI, correlated with the oral microbiome. We did not observe a significant trend of increased alpha diversity (P > .5) in samples from children with moderate-severe disease activity compared to mild activity (Figure S2). However, we observed a significant change in community structure in patients with mild and moderate-severe activity in both tongue and gingival samples (PERMANOVA < .01; Figure 3). In our previous study, machine learning techniques were used to identify genera diagnostic for IBD.¹³ Consistent with our previous findings, we again found that many of the same genera were also altered as disease activity increased, including a reduction in the abundance of Porphyromonas (padj = 0.04; Log2 FC −1.2), Haemophilus (padj = 0.07; Log2 FC −0.25), Alloprevotella (padj = 0.025; Log2 FC −0.4), Neisseria (padj = 0.008; Log2 FC −0.4) and Bergeyella (padj = 0.04; Log2 FC −0.8) species (Figure 3). Actinomyces spp. also showed a trend of increasing abundance with disease activity, similar to our previous study (Figure S3).

Figure 3.

Oral microbiome dysbiosis in relation to disease activity (per the Pediatric Crohn’s Disease Activity Index [PCDAI]) in tongue (A) and gingival (B) samples from all patients at week 0. NMDS plots (top panels) were generated from Bray-Curtis dissimilarity values and significant differences in sample distribution were determined using PERMANOVA (P values indicated above). Heatmaps (lower panels) show the results of multivariate analysis with MaAsLin2 investigating the impact of PCDAI score at week 0, controlled for age and sex. Taxa in red indicate show increased abundance in patients with PCDAI score of 3 relative to 2, increasing age or male sex, blue indicates reduced abundance in that category. Legend and colour intensity reflects effect size (see methods).

Longitudinal changes in the oral microbiome

We previously reported that clinical remission was associated with a recovery of taxa found to be depleted prior to starting treatment.¹³ Changes in the oral microbiome over time have not been described following CDED treatment, and only to a limited extent following EEN. To address this, we examined changes in the oral microbiome to week 52 in both treatment arms, to both describe any changes and determine correlates with clinical outcomes. Due to the impact of the COVID-19 pandemic, sampling was limited to a total of 34 participants in week 24 and 18 participants in week 52.

Although we did not observe significant changes in overall community structure between medium-term and later follow-up time-points (Figure S4C and D), comparison of the week 0 and week 52 microbiomes identified significant changes in taxa abundance (Figure 4). In children with moderate disease activity at week 0, their corresponding week 52 findings were characterized by an increase in the abundance of taxa previously to be diminished in CD, including Alloprevotella (padj = 0.1; Log2 FC 0.45), Haemophilus (padj = 0.039; Log2 FC 0.56), Porphyromonas spp. (padj = 0.12; Log2 FC 0.37) (Figure 4). Conversely, taxa more abundant in patients with active disease at week 0 had reduced by week 52, including Actinomyces, (padj = 0.08; Log2 FC −0.25), Ottowia (padj = 0.06; Log2 FC −0.26) and Atopobium (padj = 0.08; Log2 FC −0.45) species (Figure 4). Analysis of taxa dynamics across all available time points showed that changes in abundance generally occurred in the period following induction (ie, after week 14; Figure S5). Some taxa had already shown significant changes by week 8 (Figure S5), including the Absconditabacteria (increased in tongue and gingival samples, P < .01) and Actinomyces spp. (decreased in tongue samples, P < .05).

Figure 4.

Summary of oral microbiome changes at week 52 following the Crohn’s Disease Exclusion Diet (CDED). Heatmaps show the results of multivariate analysis with MaAsLin2 in (A) tongue and (B) gingival samples controlled for age and sex. First column (age) indicates taxa significantly affected by increasing age and second column (week) indicates taxa showing changes in abundance at week 52 compared to week 0. Taxa in red show increased abundance and blue indicates reduced abundance in that category. Legend and color intensity reflects effect size (see methods). Boxplots show supporting results of Kruskall-Wallis tests of taxa abundance in tongue (B) and gingival (D) samples (P < .05) ranked by effect size (right panel, Log₁₀ Linear Discriminant Analysis score [Log₁₀ LDA]). All taxonomic levels are included (s = species, g = genus, f = family, c = class, o = order, p = phyla).

A modified oral IBD dysbiosis index

We previously developed an oral dysbiosis index (ODI) for pediatric IBD in a cohort of Irish patients by calculating a log ratio of IBD increased taxa/decreased taxa (ie, Bacilli + Flavobacteriia/Clostridia + Betaproteobacteria + Negativicutes + Fusobacteriia), which correlated with disease activity status.¹³ In order to develop a measure of oral dysbiosis applicable to patients with CD world-wide, we analyzed the data in Figures 3, 4, and 5 to identify taxa that could improve the resolution of the index. The class Actinobacteria exhibited increased abundance with disease activity in patients from all centers (Figures 2 and 5) and was incorporated in to the “CD increased taxa” group. Additionally, Haemophilus spp. (ie, Gammaproteobacteria), Porphyromonas spp. and Alloprevotella spp. (ie, Bacteroida) all showed decreased abundance in active disease (Figure 5), thus we incorporated these into the “CD decreased taxa” group, replacing the less discriminatory taxa Negativicutes and Fusobacteriia. A modified oral dysbiosis index (MODI) was derived, which could effectively discriminate mild and moderate-severe CD based on tongue and gingival samples (P < .05; Figure 5A). The MODI was decreased significantly in week 52 samples compared to week 0 samples on pairwise analysis (P < .05; Figure 5B). To explore MODI dynamics over time, analysis by timepoint was undertaken and showed that tongue samples exhibited a significant drop in MODI by week 8 (P < .01; Figure 5B), whereas decreases in gingival samples did not reach statistical significance until week 52 (P < .001; Figure 5B). We also analyzed alpha diversity longitudinally to determine if this corresponded to changes in the MODI. At week 52, gingival samples showed reduced biodiversity, with lower Shannon and Inverse Simpson values (P < .05; Figure S4).

Figure 5.

Application of a modified oral dysbiosis index (MODI) to oral microbiome samples from Crohn disease [CD] patients. (A) The index generated from tongue and gingival samples recovered from patients at week 0 was compared in children with mild and moderate disease severities. Significant P values are indicated with asterisk (Mann-Whitney, tongue P = .047; gingival P = .047). (B) The index generated from paired week 0 and week 52 samples (n = 18) were compared using a paired t-test (tongue P = .02; gingival P = .048). Lower panels show MODI changes by week, with red asterisks indicating significant change from week 0 (Wilcoxon test, *=P < .05; **=P < .01; ***=P < .001).

CDED improves oral dysbiosis

The current trial randomized participants to follow either CDED with PEN (group 1) or EEN (group 2).²¹ To characterize the impact of these regimens on the oral microbiome, we also examined the MODI of tongue and gingival samples over time. Gingival samples from both groups exhibited a reduced MODI at week 52, but this finding only reached significance in the CDED arm (P < .01; Figure 6A). A significant and sustained reduction in MODI from the tongue was seen from week 8 onward in the CDED group, but not in the EEN treatment arm (P < .05; Figure 6A). The significant reduction of the MODI values in the CDED group was attributed to more sustained reduction in the levels of Actinobacteria and Bacilli and increased abundance of Clostridia and Bacteroidia over the course of the treatment (Figure S6).

Figure 6.

(A) Changes in MODI over time in tongue and gingival samples from Group1 (Crohn Disease Exclusion Diet [CDED]) and group 2 (EEN) patients. Red asterisks indicate significant change from week 0 (Wilcoxon test, *=P < .05; **=P < .01). (B) Area under the curve-receiver operating characteristic (AUC-ROC) analysis of the ability of the MODI to predict remission using indices calculated from Week 0 and Week 52 tongue samples.

We next examined the utility of the MODI to monitor disease activity by analyzing its ability to predict remission status using area under the curve (AUC)-receiver operating characteristic analysis of disease activity status data from week 0 and week 52. The MODI could predict remission in tongue samples at week 52 with an AUC of 0.759 (Figure 6B) and a sensitivity of 0.66 and a specificity of 0.83 (Table 1). When only CDED-treated patients were analyzed the AUC increased to 0.891 (sensitivity 1.0; specificity 0.8) and in EEN-treated patients only was reduced to 0.577 (sensitivity 0.71; specificity 0.54) (Figure 6B; Table 1).

Table 1.

Sensitivity, specificity, and accuracy of the modified oral dysbiosis index to detect remission at week 52 in group 1, group 2, and all patients.

	Sensitivity	Specificity	Accuracy
Group 1 (CDED + PEN)	0.8	1.0	0.84
Group 2 (EEN)	0.54	0.71	0.58
All patients	0.83	0.67	0.80

Abbreviations: CDED, Crohn’s Disease Exclusion Diet; EEN, exclusive enteral nutrition; PEN, partial enteral nutrition.

Next, we undertook external validation of the MODI based on a unified bioinformatic analysis of the oral microbiome in IBD recently carried out by Whelan et al.,³³ and a pediatric IBD cohort previously collected in Dublin.¹³ Using the aggregated data compiled by Whelan et al., which included 224 healthy adults and 259 IBD patients, a significant increase in the MODI was observed in IBD patients (P = .002; Figure S7). Furthermore, the index was specific for CD patients as we also observed a significant difference in the index when CD and UC patients were compared (P < .001; Figure S7). Application of the index to the pediatric population described by Elmaghrawy et al. (13) also showed a trend toward an increase in MODI in IBD patients, although the significance was weaker in this group (P = .07; Figure S8). However, the association was more significant when only CD patients were analyzed, and those with moderate to severe disease activity exhibited significantly increased MODI levels (P = .032; Figure S8).

Discussion

Oral microbiome profiling offers a promising, noninvasive window into the pathogenesis and clinical course of CD, yet its diagnostic and mechanistic potential remains underexplored. Here, we extended prior observations from a national cohort to determine whether oral dysbiosis signatures in pediatric CD are conserved across geographically diverse populations and could serve as accessible biomarkers for disease monitoring and therapeutic response. Our studies examined the microbiome of oral swabs because tongue swabs exhibit high microbial diversity, are temporally stable, and may better reflect systemic health compared to saliva, in addition to the ease of collection.^38,39 Unfortunately, our assessment of oral dysbiosis signatures as tools to assess treatment success is underpowered due to the high withdrawal rate of participants due to the COVID-19 pandemic.

This study is, to our knowledge, the first to characterize longitudinal changes in the oral microbiome in children or adults with CD undergoing treatment with CDED compared to EEN, revealing robust microbiome signatures associated with disease activity and treatment response. These associations support the hypothesis that CD is a disease of the entire gastrointestinal tract, often manifesting as ulceration in the oral cavity, suggesting that systemic immune responses to translocated oral pathobionts are involved in pathogenesis.⁹ We identified consistent microbial shifts in response to nutritional induction therapy, reinforcing the potential utility of oral dysbiosis as a dynamic biomarker of inflammation. Compared with pharmacological therapies, diet-based treatments directly modulate the microbiome. As the era for precision nutrition therapies dawns, our observation of differing microbiome outcomes between therapeutic diets underscores the importance of interrogating their mechanistic and functional impacts beyond clinical outcomes alone. It would be interesting to determine in future studies whether these changes are directly related to oral inflammation in patients with CD by also examining salivary cytokine levels. Our findings provide significant translational research opportunities for noninvasive monitoring of pediatric CD and represent a substantial advance over previous single-center or cross-sectional studies of the oral microbiome in IBD.

Oral dysbiosis in pediatric CD is characterized by reproducible depletion of Clostridial taxa and enrichment of Actinomyces spp., independent of geography. These patterns were consistent across all international sites and confirmed in comparison to healthy Irish controls (Figure  2). Overall, the tongue microbiome structure was remarkably consistent across all study centers, reinforcing its potential use as a reliable diagnostic tool. Our findings are concordant with previously published dysbiosis signatures—reduced Clostridia and increased Actinomyces—in North American pediatric ileal biopsies.^6,18 However, these results differ somewhat from our previous finding of uniform Streptococcus enrichment in IBD.¹³ Non-Irish CD participants exhibited lower streptococcal abundance than healthy controls, which may reflect differing ethno-geographic colonization patterns rather than disease status per se. High levels of streptococci have been linked to high sugar intake and recent studies have indicated relatively high levels of sugar consumption in young Irish children compared with children in other European countries.^40,41 By integrating dietary context, matched controls, and international recruitment, our study advances the field by decoupling geographic microbial signatures from core disease–associated dysbiosis. This adds critical nuance to prior interpretations and supports a conserved global dysbiosis fingerprint centered on Actinomyces and Clostridial depletion in particular.

We verified and modified an oral dysbiosis index that was capable of discriminating controls from patients and also stratifying disease activity, which has development potential for monitoring therapeutic response. Consistent with our previous findings, moderate disease activity (per PCDAI) was associated with reduced abundance of Porphyromonas, Haemophilus, Alloprevotella, Neisseria, Bergeyella, and Absconditabacteria species and increased Actinomyces spp. (Figure 4). This finding suggests an internationally conserved signature of oral dysbiosis in CD that reversed over time in treatment responders. We modified our original index^3,13 to incorporate Actinomyces, Gammaproteobacteria (including Haemophilus spp.), and Bacteroidia (including Porphyromonas spp.) reflecting the prospectively collected international microbiome data with paired clinical metadata here. The modified index stratified disease activity showing a significant decline from week 0 to 52. Notably, tongue samples reflected treatment response as early as week 8, more significantly in CDED-treated patients, providing a potentially rapid, noninvasive clinical monitoring tool. While it is tempting to speculate that the earlier microbiome improvements seen following CDED indicate an additive benefit over EEN, the study was underpowered to make such a conclusion.

Further validation of the oral dysbiosis index was provided by applying the formula to an aggregated data set incorporating 6 independent studies analyzing the oral microbiome in IBD patients. This index showed good discriminatory power to separate healthy controls and IBD patients and was most effective at identifying patients with CD. Similarly, in previously published pediatric data sets, the index could discriminate CD patients with moderate to severe disease activity from healthy controls. The findings position this validated MODI as a potential next-generation biomarker in pediatric CD, with immediate translational implications for personalised monitoring following dietary therapy.

We acknowledge that this study has a number of limitations which must be taken into consideration when interpreting its results. The COVID-19 pandemic significantly disrupted recruitment and retention, resulting in incomplete longitudinal sampling, suboptimal power, and impaired machine learning analytics and modeling. An alternative study design (superiority or noninferiority design) would be required to compare the 2 treatment regimens using a microbiome-based primary outcome measure. The absence of healthy, geographically matched control cohorts at each recruitment site limits interpretation of region-specific differences, especially with regard to streptococcal carriage. Although oral hygiene was previously shown to have a minimal impact on mucosal microbiota in pediatric IBD, future studies should collect detailed dental assessments, including caries burden and oral manifestations of CD, to refine interpretation. This study highlights the importance of undertaking multisite international microbiome studies in IBD, and while our index performed well, further validation in larger, prospectively followed cohorts is required before clinical deployment.

Conclusion

Pediatric CD is characterized by conserved patterns of oral dysbiosis and dynamic microbiome shifts associated with disease activity and therapeutic response. This study is the to our knowledge to show that oral microbiota profiles can be used to track disease activity longitudinally following CDED or EEN treatment. The impact of the COVID-19 pandemic on the study diminished its power and applicability. Nevertheless, common microbial markers of CD activity were found in geographically disparate groups. Such findings open new avenues for non-invasive biomarker development and highlight the promise of oral microbial surveillance in personalised disease management.

Author contributions

S.H., G.M.–study concept and design; S.H., G.M., A.Mc.Q.–analysis and interpretation of data; drafting of the manuscript; statistical analysis; administrative; patient enrollment. V.M.N.L., S.H., G.P.M., S.L., H.J.R., O.O., J.M.D.C., O.O., E.W., R.S.B.—administration, patient enrollment, clinical data collection, manuscript revision.

Supplementary material

Supplementary data is available at Inflammatory Bowel Diseases online.

Funding

Grant funding for this study was provided by the Children’s Health Foundation, Dublin, Ireland (NCRC c/18/2).

The clinical trial received grant funding from Nestlé Health Science, who also supplied Modulen^® IBD to all enrolled patients throughout the study period. Funders were not involved in the design or conduct of the study.

Conflicts of interest

The following authors declare these potential conflicts of interest: S.H.–Research grant from Takeda and Janssen in the past 3 years. No conflict for this study. V.M.N.-L.–Consultation fee, research grant, royalties, or honorarium from Ordesa, Adacyte, Nutricia, Pfizer, AbbVie, Takeda, and Nestlé Health Science. S.L.–Educational fees from Takeda and McKesson outside the submitted work. H.J.R.–Nutricia, Semper and Adacyte for research support and speaker fee. None of those have any relation to the present study. A.O.–Research support and advisory board for AbbVie, research site for AbbVie, Janssen, Pfizer, Eli Lilly. F.J.M.C.–Honoraria for talks and advisory from Abbvie, Nestlé Health Science, MSD, Adacyte, Janssen. O.O.–Dr. Olén has been PI on projects at Karolinska Institutet financed by grants from Janssen, Takeda, AbbVie, Pfizer, Bristol Myers Squibb, and Ferring, and also report grants from Pfizer, AbbVie, Galapagos, and Janssen in the context of national safety monitoring programs. None of those studies have any relation to the present study. J.V.L.–Nestlé Health Sciences consulting/research support/speaker fee and Janssen (Research Support/Travel). R.S.B.–Received funds for consulting or speaker—from Nestlé Health Science, Takeda, Megapharm and Janssen; serves on advisory board for Evinature. E.W.–Advisory board: AbbVie, Nestlé Health Science, Pfizer, BioJamp; Speaker: AbbVie, Nestlé Health Science, Janssen, Mead Johnson Nutrition. G.P.M., A.McQ., G.H., and R.J.W., no conflicts to declare.

Data availability

All of the patient demographic and clinical details used in this analysis are contained in Table S1. All of the 16S RNA sequence data and associated patient identifiers are also available to download from NCBI, accession number PRJNA1378163.