Multimodel transcriptomics narrows the gap between mouse colitis and human IBD
Inflammatory bowel disease has long exposed one of translational medicine’s most stubborn problems: mouse models can reproduce pieces of intestinal inflammation, but rarely the full human syndrome. In a multimodel analysis published in Gut, investigators compared 13 widely used IBD mouse models against patient datasets to identify which molecular programs persist across species and which remain model-specific. The goal was not simply to rank one mouse model above another. It was to define conserved inflammatory circuits that can be trusted when preclinical work is used to justify human biology, target selection, or biomarker development.
The study focused on transcriptomic relationships at both bulk and single-cell levels, using curated and a priori statistical correlative methods to align preclinical colitis and ileitis with human IBD. Across the model set, the analysis found shared pathways, ontologies, and cellular compositions that translated between mice and patients. A core inflammatory gene signature emerged, anchored in T-cell homing, innate immunity, and the epithelial barrier. Those three axes matter because they sit at the center of mucosal immune traffic, microbial sensing, and barrier failure—the biological triangle that drives chronic intestinal injury in human disease.
A conserved inflammatory core emerged across species
The most striking result was not that mouse models differed from one another. It was that some of those differences still converged on a common molecular backbone of IBD. Investigators describe a conserved core inflammatory signature of IBD-associated genes that was strong enough to be distilled into a new mouse gut Molecular Inflammation Score, or mMIS. The score provides a compact readout of inflammatory intensity across models, based on the shared transcriptional features that track with intestinal inflammation rather than with any single experimental platform.
This kind of score has practical value because intestinal inflammation is not a monolith. Some models may lean harder on epithelial injury, others on lymphocyte recruitment, and others on innate immune activation. A composite molecular readout helps separate biological signal from the noise created by different induction strategies. It also offers a more disciplined way to compare models when the question is not whether a mouse is inflamed, but whether it is inflamed in a way that resembles the human disorder under study.
A second major finding sharpened that point further.
Transcriptomic relatedness across models tracked with the mode of induction, not the canonical Th1, Th2, or Th17 label.
Induction method mattered more than Th labeling
The analysis found that the similarity between IBD models was driven by how inflammation was induced, rather than by the usual immunotype shorthand often used to classify them. In other words, the experimental trigger shaped the transcriptomic outcome more strongly than whether the model was traditionally considered Th1-, Th2-, or Th17-skewed. For translational researchers, that is an important correction. Immunologic labels can be useful shorthand, but they do not fully capture the molecular architecture of disease once tissue-level inflammation is underway.
The implication is straightforward: model selection should be based on the biological question, not just the historical label attached to a mouse line. A study aimed at epithelial barrier failure, for example, may require a model with a different transcriptomic profile than one designed to interrogate lymphocyte trafficking or innate immune activation. By tying relatedness to induction mode, the analysis offers a more realistic framework for choosing the right preclinical system for a specific human IBD question.
Cell-type specificity separated the models from one another
Beyond the conserved core, the compendium also identified distinct signatures for B cells, T cells, and enteric neurons in specific mouse IBD models. Those findings matter because they show that individual models are not interchangeable replicas of a single inflammatory state. Each can preserve its own cellular bias, and those biases may be useful when a study is designed around a particular arm of intestinal pathology.
The enteric neuron signal is especially notable in an intestinal disease context, because neural involvement is often overshadowed by immune-centric analyses. Yet motility, secretion, and neuroimmune communication all shape the clinical expression of IBD. A model that preserves neuronal signatures may therefore be more informative for work that extends beyond mucosal cytokines and into gut function, symptom generation, or tissue-level cross-talk.
The same logic applies to B-cell and T-cell signatures. These immune compartments are not interchangeable, and their transcriptional footprints can influence how a model behaves under different inflammatory triggers. By mapping those differences explicitly, the study gives investigators a clearer way to align experimental design with mechanism.
What the mouse gut Molecular Inflammation Score can and cannot do
The new mMIS is best viewed as a translational filter rather than a surrogate for all of IBD biology. It captures a conserved inflammatory program that spans model systems and patient datasets, but it does not erase the limitations of animal work. It can help identify which models reproduce key molecular features of human intestinal inflammation, yet it cannot make every mouse model a stand-in for the full complexity of clinical disease. Genetic diversity, microbial exposure, treatment history, and chronicity all remain difficult to mirror in a controlled experimental system.
Still, the score provides something preclinical research has needed for years: a structured way to measure whether a model shares the transcriptional logic of human IBD. When paired with single-cell and bulk transcriptomic comparisons, it can help narrow the distance between exploratory biology and clinical relevance. That is particularly useful in a field where misaligned preclinical models can slow therapeutic development and cloud interpretation of mechanistic studies.
A shared resource for the IBD research community
The compendium database created through this work is available as a web explorer at http://trr241.hosting.rrze.uni-erlangen.de/SEPIA/. Its availability turns the analysis from a static publication into a working reference for investigators comparing models, pathways, and cell states. For a field built on iterative experimentation, that kind of shared resource can help standardize how molecular relevance is assessed before a model is moved into downstream studies.
The broader message is precise rather than sweeping. Human IBD biology can be approximated in mice, but the approximation is selective, not total. Conserved circuits involving T-cell homing, innate immunity, and epithelial barrier biology appear to be robust enough to bridge species in this analysis, while other signatures remain model-specific. That combination should guide how future experiments are framed, especially when the goal is to infer something about human disease rather than about murine inflammation alone.
What this work does not settle is how these transcriptomic relationships will perform across every therapeutic context or across the full spectrum of Crohn’s disease and ulcerative colitis phenotypes. It also does not eliminate the need for careful validation in human tissue. It does, however, give the field a more disciplined molecular map for deciding which mouse models deserve translational confidence and which should be used more narrowly.