Atopic dermatitis (AD) is a common chronic inflammatory skin disease characterized by recurrent episodes of intense itching, eczematous lesions, and compromised skin barrier function (1, 2). Recently, the global prevalence of AD has significantly increased, imposing substantial economic and psychosocial burdens on patients and their families (3) Although genetic predisposition, environmental exposures, immune dysregulation, and impaired skin barrier function have been strongly associated with AD pathogenesis (4, 5), its precise pathophysiological mechanisms remain incompletely understood, and existing treatment approaches demonstrate limited efficacy.

Accumulating evidence underscores the pivotal role of the gut microbiota in the pathogenesis of atopic dermatitis (AD) (6, 7). Microbial dysbiosis can lead to increased intestinal permeability and immune dysregulation, thereby promoting chronic inflammatory responses (8, 9). However, studies specifically addressing how dietary factors shape the composition and metabolic functions of gut microbiota in AD patients are still limited. In particular, there is limited understanding of the effects of different types of food intake or components on gut microbial diversity and function in AD populations. Most existing research has relied on cross-sectional human studies or isolated animal experiments, lacking integrative approaches that combine animal models with human data. This gap limits our understanding of the causal relationships among diet, gut microbiota, and AD. Given that diet exerts a profound influence on the gut microbial ecosystem—unhealthy patterns, such as high-sugar and high-fat diets, promote dysbiosis, while fiber-rich diets increase beneficial bacteria and support gut homeostasis (10, 11) — there is a clear need for comprehensive investigations into the specific dietary determinants of microbiota changes in AD.

Recent studies have further demonstrated that habitual dietary patterns significantly influence AD risk and severity. For example, high estimated total fat intake has been associated with enhanced allergy sensitization and increased prevalence of atopic diseases in young adults (12). Conversely, frequent consumption of plant-based foods has been shown to reduce the risk of AD exacerbation (13). Other findings suggest that a high intake of trans and saturated fatty acids may increase susceptibility to AD flare-ups (14). In contrast, diets rich in fiber and probiotics are associated with a lower risk of AD and reduced house dust mite allergy (15). These findings support the hypothesis that dietary modulation can shape immune responses and gut microbial dynamics, aligning with the current study’s objective to elucidate the diet–microbiota–AD axis.

Based on the above research gaps, this study adopted a combination of animal experiments and population studies to systematically analyze the effects of dietary factors on the diversity, composition, and function of intestinal microbiota, and further comprehensively explore their relationship with the pathogenesis of AD. Specifically, we first established a DNCB-induced mouse model of AD to investigate changes in skin histopathology, serum IgE levels, and the typical gut microbiota composition. Subsequently, we expanded our exploration to the human population, examining the diversity, composition, and function of the gut microbiota in AD patients, and further investigating the correlation between different dietary patterns and microbiota composition. We hypothesize that specific nutritional patterns can causally alter gut microbiota composition and functional capacity, which in turn modulate immune responses and influence the risk and severity of AD. This design allows us to investigate both the mechanistic pathways in a controlled animal model and the corresponding associations in human populations, thereby providing an essential theoretical basis and scientific evidence for dietary intervention and microbiota modulation strategies in the prevention and treatment of AD.

In this study, animal experiments combined with human population surveys comprehensively explored the relationships between dietary factors, gut microbiota, and atopic dermatitis (AD), highlighting the potential mechanisms through which gut dysbiosis may contribute to AD development.

Results from the animal experiments demonstrated that DNCB successfully induced a mouse model exhibiting typical AD features, including significant pathological skin changes (epidermal thickening, inflammatory infiltration, hyperkeratosis, and necrotic cell debris) and dose-dependent increases in serum IgE levels. These findings are consistent with previous studies, as elevated IgE is widely recognized as a key immunological hallmark of AD (21). Simultaneously, significant alterations were observed in gut microbial diversity and composition in these mice, particularly with notable changes in the proportions of Firmicutes and Bacteroidota. Furthermore, specific microbial taxa, such as Bacilli, Clostridia, and Ruminococcus, were notably enriched in the high-dose group, aligning well with existing theories that propose gut dysbiosis may exacerbate inflammatory immune responses (22, 23).

In the human population study, we validated and expanded the findings from the animal model. AD patients exhibited significantly reduced gut microbiota diversity, consistent with previous research indicating that decreased microbial diversity often correlates with immune dysfunction and increased inflammation (24). Moreover, significant compositional changes in gut microbiota were observed at both phylum and genus levels among AD patients, characterized by decreased abundances of Firmicutes and Bacteroidota and increased abundances of Actinobacteriota and Bifidobacterium, with a notable decrease in Blautia. These results suggested that microbial restructuring might directly contribute to AD pathogenesis, particularly the increased abundance of Bifidobacterium, which could correlate with immune dysregulation and enhanced inflammatory responses in patients.

Interestingly, we observed an increased abundance of Bifidobacterium in AD patients, which appeared counterintuitive given its traditional role as a beneficial genus associated with improved gut health. However, recent studies suggest that the effects of Bifidobacterium may be highly strain-specific and context-dependent. Certain strains may exert pro-inflammatory activity under dysregulated immune conditions, while others are protective. Moreover, host-related factors such as genetic background, immune status, and dietary environment may influence whether Bifidobacterium acts beneficially or potentially contributes to disease progression. Thus, the observed enrichment in AD patients may not necessarily contradict its probiotic reputation but instead highlight the complex host–microbiota interactions that warrant further mechanistic investigation. Given the pronounced strain heterogeneity within Bifidobacterium, our interpretation is hypothesis-generating. Definitive confirmation of functional relevance will require strain-resolved shotgun metagenomics (e.g., read- or MAG-level profiling), complementary metatranscriptomics/metabolomics, and culture-based isolation followed by functional assays and/or gnotobiotic transfer experiments.

Importantly, this study systematically investigated, for the first time, the association between dietary patterns and gut microbiota composition in AD patients. Our findings revealed that the dietary habits of AD patients deviated significantly from those of healthy controls, characterized by an increased intake of fiber-rich and anti-inflammatory foods, such as vegetables and fruits, coupled with a reduced consumption of refined grains. This dietary shift could reflect either proactive adjustments made by patients to alleviate symptoms or passive changes in dietary preferences. Another possible explanation for the observed higher intake of fruits and vegetables and lower intake of refined grains among AD patients was reverse causality. Specifically, patients may have consciously adjusted their dietary habits after the onset of the disease in an attempt to alleviate symptoms. This behavioral change could introduce bias, as the observed dietary patterns may not necessarily represent pre-disease exposures but rather reflect secondary modifications driven by health awareness. Moreover, given that our FFQ relied on retrospective self-reporting, recall bias could not be ruled out, further limiting the causal interpretation of these findings. Future longitudinal or prospective dietary assessments are needed to disentangle whether such dietary patterns are a cause or consequence of AD. Further analysis through Spearman correlation and RDA/CCA confirmed close associations between specific dietary factors (refined grains, vegetables, and fruits) and microbial taxa (such as Firmicutes, Bifidobacterium, Ruminococcus, and Roseburia). These findings align with previous studies that highlight diet as a critical modulator of gut microbial composition (25, 26). Additionally, functional prediction analyses demonstrated significant disruptions in carbohydrate, amino acid, and energy metabolism pathways within the gut microbiota of AD patients. This suggests that despite AD patients adopting healthier dietary habits, metabolic dysfunction within the gut microbiota may impair the effective utilization of nutrients, further perpetuating inflammation and immune dysregulation.

Functional predictions of gut microbiota revealed markedly reduced activity in metabolic pathways involving carbohydrates, amino acids, and energy production in AD patients. Prior research indicated that abnormal gut microbial metabolism could exacerbate gut barrier dysfunction and immune imbalance, thereby promoting AD development and progression. Thus, significant alterations in microbial metabolic function further support the hypothesis of gut microbial dysbiosis as an important factor in AD pathophysiology.

Despite providing comprehensive data support through animal and human studies, this research has certain limitations. Firstly, due to the cross-sectional design, causal relationships between dietary factors and microbiota alterations remain unclear and require confirmation through prospective or interventional studies. Secondly, the lack of significant changes in microbial functional profiles observed in animal experiments may be attributed to insufficient sample size or exposure duration. Future studies should consider longer exposure periods or increased sample sizes. Moreover, the absence of significant functional alterations despite compositional shifts may also reflect methodological limitations. For instance, the 4-week exposure duration may not have been sufficient to induce measurable metabolic reprogramming, and the sequencing depth and sample size could have reduced the sensitivity to detect subtle changes. This study’s functional predictions of microbial communities were performed using PICRUSt2, as described in the Methods section. However, PICRUSt2 relies on reference genomes and predictive algorithms rather than direct metagenomic sequencing, which might have limited the accuracy of functional inferences. This methodological limitation may partly explain the absence of significant functional alterations in the animal experiments despite compositional changes. Despite compositional restructuring, the absence of significant functional changes in mice underscores the limitations of inference-based tools such as PICRUSt2 and the relatively short exposure and sequencing depth. Therefore, these functional outputs are preliminary, best viewed as hypothesis-generating signals rather than definitive evidence of pathway perturbation. To overcome this limitation, future studies should consider integrating direct metagenomic or metabolomic profiling, alongside extended exposure periods and larger cohorts, to more accurately evaluate the functional consequences of gut microbiota alterations. Additionally, although our FFQ was adapted from the validated CHNS instrument, the shortened version used in this study has not been independently validated. This may introduce recall bias and dietary misclassification, which could influence the observed associations between diet and microbiota. Therefore, the results should be interpreted cautiously, and future studies are encouraged to use fully validated FFQs or multiple 24-hour dietary recalls to improve dietary assessment. Furthermore, the FFQ in this study assessed dietary intake over the past 12 months, whereas stool samples for microbiota analysis were collected simultaneously. This temporal mismatch may weaken the precision of observed diet–microbiota associations. Future research should incorporate repeated dietary assessments alongside longitudinal microbiota sampling to more accurately capture dynamic diet–microbiota interactions. Another limitation is that our FFQ did not separately categorize probiotic-containing foods and beverages (e.g., yogurt, cultured drinks). Although such items were grouped within the general dairy category, this decision likely biased our results by underestimating or obscuring the specific contributions of probiotic foods to gut microbiota composition and AD pathophysiology. Given probiotic foods’ strong and well-documented effects on microbial diversity and immune regulation, future research should incorporate dietary assessments that specifically and independently capture probiotic intake to more accurately evaluate their role in modulating the diet-microbiota-AD axis. We did not adjust for potential confounders such as BMI, medication use, socioeconomic status, and allergic comorbidities. Furthermore, our study did not systematically collect detailed clinical information such as disease severity scores (e.g., SCORAD, EASI), disease duration, comorbid allergic conditions (asthma, allergic rhinitis, food allergy), prior treatment history, anthropometric parameters (BMI, weight, height), perinatal and early-life factors (birth mode, feeding type), and lifestyle/environmental factors (urban versus rural origin, exercise habits). The absence of these data limited our ability to perform stratified analyses or adjust for potentially essential modifiers of gut microbiota composition. We acknowledge this as a significant limitation, and emphasize that future studies should incorporate standardized clinimetric scales and comprehensive demographic, clinical, and lifestyle data to strengthen the reliability and interpretability of host–microbiota associations in AD. Finally, detailed sequencing statistics, such as raw read counts and post-filtering valid reads, were not archived and cannot be reported in this manuscript. This represents a substantial methodological weakness, as rarefaction curves, while useful, cannot fully substitute for actual sequencing depth metrics in ensuring reproducibility. Nevertheless, sequencing reliability was evaluated using rarefaction curves based on Shannon and Simpson indices, which reached saturation across all groups. This indicates that sequencing coverage was sufficient to capture the majority of microbial diversity present in the samples, supporting the robustness of subsequent compositional and functional analyses. However, the absence of archived raw read data may limit reproducibility; therefore, future studies should retain complete sequencing metrics to enhance methodological transparency further.

In summary, despite certain methodological limitations, including the cross-sectional design, limited collection of clinical covariates, and reliance on predictive functional inference, our study underscores the integration of animal experiments and human population data to elucidate the interplay between diet, gut microbiota, and AD pathogenesis. By demonstrating that dietary shifts and microbial dysbiosis converge on impaired microbial metabolic pathways, our work provided novel insights into how nutritional and microbial factors jointly shape AD onset and progression. These findings extend existing knowledge and offer a conceptual framework for dietary interventions and microbiota-targeted therapies. Future longitudinal and multi-omics studies with more comprehensive clinical characterization will be essential to validate and expand our observations. However, the present study already provides an important empirical basis and a forward-looking direction for the field.

This research has systematically examined the association between dietary factors, gut microbiota, and atopic dermatitis (AD) through a combined approach of animal experiments and human population studies. The DNCB-induced mouse model has successfully exhibited classical AD pathological characteristics and immune responses, accompanied by decreased gut microbial diversity and structural remodeling of specific microbial populations. Human studies have further confirmed significantly reduced gut microbial diversity in AD patients, characterized by decreased abundances of key taxa such as Firmicutes and Bacteroidota, and increased abundances of Actinobacteriota and Bifidobacterium. Analysis of dietary patterns has revealed a preference among AD patients for fiber-rich and anti-inflammatory foods, which is closely linked to observed shifts in microbiota. Furthermore, gut microbiota functional prediction has indicated notable dysfunction in carbohydrate, amino acid, and energy metabolism in AD patients. These findings generate hypotheses regarding diet–microbiota–AD relationships; definitive functional and causal inferences will require longitudinal designs and multi-omics validation. However, given the observational design, the results should be interpreted as hypothesis-generating, and further prospective and mechanistic studies are needed to confirm causality and inform dietary or microbiota-targeted interventions in AD management.