Reading EoE in Blood and Biopsy: Multi-omics Profiling Across Mucosal and Circulating Compartments

From patchy lesions to systems biology in a pediatric disease

Eosinophilic esophagitis (EoE) is a chronic, food-triggered, mucosal immune disorder that presents with dysphagia, food impaction, reflux-like symptoms, and chest discomfort in children. Diagnostic practice still hinges on endoscopic biopsy demonstrating a defined threshold of eosinophils per high-power field after exclusion of alternative causes of esophageal eosinophilia. Pediatric and international consensus statements emphasize that disease can be segmental, which complicates histologic capture and motivates better systemic readouts. The challenge is practical as well as biological: families and clinicians need tools that minimize repeat anesthesia and endoscopy while preserving diagnostic confidence. Multi-omics designs promise to bridge that gap by linking local molecular events in the esophageal mucosa to accessible signals in blood. Recent pediatric work has operationalized this strategy, pairing unbiased transcriptomics and cytometry in biopsies with immune profiling and untargeted plasma metabolomics to search for a circulating EoE signature.

A central insight from transcriptome-first studies is that EoE signatures replicate across geography and platforms, suggesting a conserved disease program in the esophagus. Early U.S. cohorts defined a robust epithelial–immune gene set and later produced a quantitative diagnostic panel, while new pediatric cohorts in Europe reproduced the same directional changes and pathway enrichments. This reproducibility matters because it underwrites attempts to anchor blood signals to an accepted esophageal molecular ground truth. When lesions are patchy, a single biopsy may under-sample biology; a cross-tissue signature can mitigate that sampling error. Systems-level comparisons therefore frame the question not as “is blood enough,” but rather “which blood features best map onto validated esophageal axes.” That framing guides the selection of downstream cytokine panels, immunoglobulin isotypes, and metabolite classes for integrative modeling.

The pediatric multi-omics approach begins by defining the local state with untargeted gene expression in esophageal biopsies. After rigorous quality control and unsupervised analyses, the disease group separates from controls and reveals high-confidence networks that include epithelial barrier programs, chemokine landscapes, and innate–adaptive cross-talk. These networks are not generic airway or gut allergy proxies; they contain esophagus-relevant barrier and cornification elements and enrich for chemical stimulus perception pathways. Such terms, while broad, point to epithelial sensing and response modules that interact with immune circuits to shape tissue tone. Because the transcriptome aggregates both immune infiltration and epithelial state, it provides a stable anchor for cross-modality correlation. That anchor is essential for testing whether circulating molecules truly reflect mucosal biology or merely systemic atopy.

Clinically, the implications are direct: any blood-based candidate must be interpretable against this mucosal reference and robust to confounders like proton-pump inhibitor therapy. Pediatric guidelines remind us that medication status and coexisting reflux can blur clinical impressions, and cohort work has shown that some metabolites track with acid suppression. Integrative pediatric studies therefore stratify by medication and re-validate discriminants in untreated subsets to avoid mechanistic misattribution. This attention to design elevates the credibility of what follows: a prioritized set of cytokines, immunoglobulins, and small molecules that together classify EoE. The goal is neither to replace histology overnight nor to chase a single “silver bullet” biomarker, but to assemble a composite that is faithful to tissue biology and feasible in practice. That composite view is where esophageal omics meets venipuncture.

Epithelial barrier failure meets multilineage immunity

Esophageal transcriptomics in children points to coordinated down-regulation of barrier-supporting genes and up-shifts in chemokines and immune effectors. The barrier signals span protease inhibitors, mucins, and differentiation cues that together maintain epithelial integrity under mechanical and dietary stress. Converging with these epithelial themes are immune pathways that include interferon-linked modules, alarmins, and chemokine axes known to recruit granulocytes and lymphocytes. The spatial reality of EoE—focal erythema, furrows, and rings—fits a model in which barrier dysfunction seeds inflammatory neighborhoods rather than a uniform, pan-esophageal storm. That topology helps explain why transcriptomes from macroscopically uninvolved sites may appear deceptively quiet. It also explains why a multi-parametric signature outperforms any one marker that might rise and fall with sampling location.

Among immune actors, group-2 innate lymphoid cells (ILC2s) emerge as consistent participants in pediatric EoE tissue. These lineage-negative cells are poised to secrete IL-5 and IL-13 upon epithelial alarmin cues and have been shown to be enriched in active disease biopsies. Their presence complements, rather than replaces, Th2 cells, and aligns with chemokine patterns that favor eosinophil trafficking and residency. Beyond type-2 immunity, pediatric biopsies also show signatures consistent with Th17-skewing and interferon activity, challenging any single-axis explanation. The cellular landscape therefore resembles a multiplexed immune ecology rather than a lone Th2 cascade. That ecology requires composite readouts that can capture parallel streams of activation.

Soluble mediators in tissue supernatants reinforce this multiplexed view. Elevated eotaxin family members align with eosinophil biology, while matrix metalloproteases and structural proteins hint at remodeling pressure within the lamina propria. Interferon-linked cytokines and chemokines populate the same network diagrams as epithelial sensing terms, suggesting bidirectional signaling rather than a simple top-down script. Immunoglobulin patterns in tissue include increases in IgE, which track with local eosinophil counts and innate lymphoid cell abundance, pointing to microenvironmental class-switching or local retention. The absence of robust Th2 cytokine protein detection in some pediatric samples—even when mRNA is present—further suggests spatially restricted synapses and rapid consumption. Together these observations argue for reading tissue through integrated panels instead of single analytes. PMC

These insights also re-contextualize why purely histologic counts, though necessary, fail to capture disease tone or trajectory by themselves. EoE histology is the end-result of many upstream choices made by epithelium, innate lymphoid cells, T cells, and recruited granulocytes. Transcriptomics recovers that upstream ledger, while tissue cytometry and soluble mediator profiling quantify its cellular and chemical executors. Each layer adds a different unit to the same equation: barrier integrity, immune composition, and effector chemistry. When considered together, they portray a pediatric disease that is organized, dynamic, and chemically articulate rather than merely “eosinophilic.” That portrait provides the template for what blood must learn to mirror.

Circulating clues: the pediatric immunome and metabolome in blood

Peripheral blood does not simply echo the tissue; it offers a filtered view shaped by compartmentalization, clearance, and dilution. In pediatric cohorts, broad cytokine panels often fail to produce a single discriminant that cleanly separates patients from controls across centers. Yet composite patterns emerge: immunoglobulin class ratios shift, select chemokines move in directions consistent with tissue recruitment, and a small set of cytokine receptors report on systemic tone. These immune features are most valuable when interpreted with the esophageal transcriptome as reference rather than as standalone diagnostics. As classification candidates, they contribute to multi-marker models rather than serving as solo readouts. This model-based thinking changes the question from “is marker X elevated” to “does marker X improve mapping to the tissue program.”

Untargeted plasma metabolomics adds a complementary, pathway-level lens that has proven surprisingly discriminative in pediatric EoE. In medication-stratified analyses, amino-acid metabolism and vitamin B6 vitamers repeatedly surface, alongside bile acid-related signals that can be confounded by proton-pump inhibitors. Within untreated subsets, pyridoxine and its catabolite 4-pyridoxic acid distinguish cases from controls and correlate with families of amino acids, suggesting coordinated control of transamination and one-carbon–linked flux. Tryptophan and related indole chemistry also appear, cohering with literature that connects aromatic amino-acid handling to atopic inflammation. These signals do not assert a deficiency narrative; rather, they suggest that B-vitamin handling and amino-acid routing are responsive to the inflammatory state in EoE. That responsiveness makes them attractive members of a composite blood signature.

Mechanistically, the vitamin B6 axis intersects directly with immune organization through sphingolipid biology. Pyridoxal phosphate is a cofactor for enzymes that regulate sphingosine-1-phosphate (S1P) turnover, a lipid signal that guides lymphocyte egress and innate activation. Experimental systems demonstrate that B6 availability modulates S1P accumulation via the lyase step and, by extension, influences inflammatory amplitude. This is relevant to pediatric EoE because dysregulated epithelial cues and lymphoid traffic are core features of the disease microenvironment. A systemic B-vitamer signal may therefore encode information about immune cell routing and activation thresholds without implying causality. The metabolome thus becomes a compact reporter of multi-cellular dynamics.

Medication status remains a critical covariate in blood-based discovery and must be handled explicitly. Proton-pump inhibitors alter gastric pH and can shift bile acid pools and xenobiotic handling, with downstream consequences for measured metabolites such as methylxanthines and conjugated bile acids. Pediatric multi-omics studies consequently construct stratum-specific models and confirm that key discriminants retain directionality in untreated patients. This practice reduces the risk that a pharmacologic footprint is misread as disease biology. It also sharpens the estimated contribution of immunoglobulins, chemokines, and B-vitamers to the final classifier. A rigorous signature is one that survives this stress test and stays aligned with esophageal molecular axes.

Integrating layers: cross-tissue models that bind plasma to mucosa

The analytical heart of multi-omics is not the individual assay but the integration framework that relates them under supervised constraints. In pediatric EoE, integrative modeling aligns biopsy transcriptomes with plasma cytokines, immunoglobulins, and annotated metabolites to identify the smallest cross-modal set that cleanly separates cases from controls. The optimal solutions often include a handful of genes from epithelial and chemokine programs, a short list of cytokines and receptors, targeted immunoglobulin ratios, and select metabolites such as pyridoxine, 4-pyridoxic acid, and amino-acid derivatives. Correlation structures within these solutions are as informative as the features themselves, revealing tight links between epithelial stress modules and plasma small molecules, and somewhat looser links with bulk antibody levels. This topology makes biological sense: metabolites and cytokines respond on faster timescales than class-switched immunoglobulins and thus track more closely with transcriptional state. The result is a composite that is intelligible mechanistically and tractable clinically.

Two properties signal that such cross-tissue models are on the right track. First, they preserve separation between cases and controls when any single block is projected alone, indicating that no component is carrying an artificial burden. Second, the within-block feature sets are enriched for elements previously implicated in EoE biology, such as ILC2-linked cytokine circuits, eotaxin axes, epithelial barrier genes, and interferon-associated chemokines. Blood features like CXCL12 and immunoglobulin ratios sit naturally in this context because they reflect trafficking and activation states tied to esophageal events. Metabolites slot in as reporters of cofactor and amino-acid routing connected to lymphocyte dynamics and epithelial stress. This coherence across layers reduces the risk of spurious classification by hidden confounders. It also builds confidence that a compact panel could be transferred between centers.

Importantly, integrative pediatric models do not claim that every child’s EoE is identical; rather, they show that shared axes can be read in blood even when endoscopic appearances vary. That trait is invaluable for a disease with documented endotypic diversity and focality. When biopsies undersample an inflamed segment, the blood-anchored composite can still register activity through synchronized shifts in cytokines and metabolites that mirror the dominant tissue programs. Conversely, when blood looks quiescent, the integrated read can prompt a closer look at biopsy site selection or sampling depth. This bidirectional feedback loop between tissue and plasma is the practical value of multi-omics. It transforms both specimen types into mutual validators instead of rival gold standards.

As datasets expand, cross-validation against established, clinically deployed tools becomes essential. The esophageal gene expression diagnostic panel pioneered in earlier cohorts offers one such benchmark and demonstrates how quantitative signatures can complement histology in algorithms. Pediatric multi-omics can use these tools not as competitors but as scaffolds on which to hang circulating extensions. The test of success will be whether a pared-down plasma panel improves triage decisions, reduces unnecessary procedures, and flags relapse risk between endoscopies. That requires not only analytical rigor but also prospective, medication-aware validation in independent pediatric populations. The road is iterative, but the direction is clear.

Toward noninvasive diagnostics and mechanism-informed care

A near-term clinical application of pediatric multi-omics is a two-stage workflow: tissue-anchored diagnosis with a validated transcriptomic or histologic standard, followed by longitudinal, blood-based monitoring using a compact composite panel. In such a design, the initial biopsy supplies the esophageal “truth,” while blood samples track the mechanistic signals most predictive of that truth. Immunoglobulin ratios and selected chemokines provide immune-system context; small molecules such as B-vitamers and amino-acid derivatives report on immunometabolic tone; and optional gene scores calibrate the link back to mucosa. The promise is fewer procedures without surrendering pathophysiologic insight. Pediatric guidelines already anticipate the need for practical monitoring frameworks as children transition to adult care. A multi-omics panel situated within those frameworks is a logical next step.

Therapeutically, mechanism-aware biomarkers can do more than classify; they can stratify response trajectories. If epithelial barrier programs dominate a child’s signature, intensifying barrier-directed care and environmental counseling becomes rational, whereas a signal weighted toward type-2 or interferon-linked chemokines might prioritize immunomodulation. The rising literature on ILC2 biology underscores the relevance of innate lymphoid circuits to both disease activity and remodeling pressures, opening routes to track or modulate these axes over time. Metabolic markers tied to B-vitamin handling and amino-acid routing could also act as dynamic reporters of immune set-points under therapy. None of these applications require numeric thresholds in isolation; they require pattern recognition anchored to tissue reference states. That is precisely what integrative pediatric models are designed to deliver.

Methodologically, the field should continue to prioritize medication-aware designs, external validation, and transparent feature selection. Pediatric studies that explicitly stratify by proton-pump inhibitor use and re-estimate discriminants in untreated patients set a high bar for interpretability. Harmonized pre-analytics, shared spectral libraries for metabolite annotation, and consensus panels for cytokines and isotypes will accelerate reproducibility. Cross-platform replication—microarray to RNA-seq, bead-based assays to alternative immunoassays, Orbitrap-based metabolomics to orthogonal platforms—further secures generalizability. Crucially, feature sets should be chosen for biological coherence as well as classification gain, preserving interpretability for clinicians and families. The result will be panels that illuminate mechanism while guiding care.

Finally, pediatric EoE should be viewed as a proving ground for cross-tissue immunology that can generalize to other focal, food-linked mucosal diseases. The strategy of anchoring blood interpretation to validated mucosal programs, adjusting for medication footprints, and triangulating with multi-layer correlations is portable. As cohorts grow, genetic context and microbiome measures can join the panel to refine endotypes and anticipate flares. Established diagnostic frameworks will remain, but they will be complemented by mechanistically informed, minimally invasive tools that reflect the realities of pediatric care. In this vision, the esophagus teaches the blood what to say, and the blood carries the message back to clinic. That is the trajectory from omics to outcomes.

Study DOI: https://doi.org/10.3389/fimmu.2023.1108895

Engr. Dex Marco Tiu Guibelondo, B.Sc. Pharm, R.Ph., B.Sc. CpE

Editor-in-Chief, PharmaFEATURES