Microbiome and metabolome integrated analysis reveals oral microbiota alterations and associated metabolic pathways in allergic rhinitis

Abstract

Background

Allergic rhinitis is a globally prevalent condition characterized by an increase in incidence. This study aimed to characterize differences in supragingival plaque microbial composition between patients with allergic rhinitis and healthy controls, and to explore the differential metabolites and enriched metabolic pathways to provide reference data for subsequent in-depth investigations into the pathogenesis and potential therapeutic targets of allergic rhinitis.

Material and methods

We employed 16S rRNA gene high-throughput sequencing technology and untargeted metabolomics to analyze the supragingival plaque microbiota of 35 healthy individuals and 35 patients with allergic rhinitis.

Results

At the genus level, Abiotrophia, Rothia, and Actinobacillus showed significant inter-group differences. Ten species, including Prevotella melaninogenica and Capnocytophaga sputigena, exhibited markedly differential abundances. Microbial co-occurrence network analysis revealed a simpler topological architecture in the allergic rhinitis group than that in the healthy control group. Metabolomic profiling identified 106 differentially expressed metabolites that were enriched across 13 associated metabolic pathways.

Conclusion

Under the current inclusion criteria, this study identified significant alterations in the microbial composition of supragingival plaques in patients with allergic rhinitis. Additionally, using the current sample size, this study preliminarily characterised differential metabolites and pathways, providing initial clues for identifying therapeutic targets in allergic rhinitis.

Introduction

Allergic rhinitis (AR) is one of the most prevalent chronic inflammatory disorders globally [1], and its complex and multifactorial pathogenesis remains incompletely understood. The disease mechanism is fundamentally characterised by immune dysregulation, particularly involving a Th1/Th2 imbalance and Treg/Th17 axis disruption [2]. Emerging evidence highlights the critical role of neuro-immune–inflammatory interactions in modulating and amplifying AR symptomatology [3]. Moreover, maintenance of microbial homoeostasis has been identified as a crucial regulatory factor that influences both immune responses and inflammatory processes, thereby affecting disease susceptibility and severity [4]. The interplay among these three factors forms a complex network that governs AR pathogenesis and progression. The clinical symptoms of AR are primarily attributed to the inflammatory response of the nasal mucosa [5]. The immune cells involved in this process orchestrate a cytokine-mediated cascade that perpetuates chronic allergic inflammation [6].

Extensive research has demonstrated significant alterations in the composition and structural organisation of both the gut and airway microbiota in AR [7–10]. The delayed maturation of microbial communities has been identified as a characteristic feature of paediatric allergic disorders, underscoring the predictive value of early microbial alterations in allergy development [11]. However, the relationship between the oral cavity, which harbours the body's second-largest microbial ecosystem, and AR remains poorly understood. Current evidence substantiates a robust association between the oral microbiota and the initiation and progression of allergic diseases. Metagenomic investigations have revealed that Prevotella species detected in the gastrointestinal tract of patients with asthma are partially derived from the oral cavity and contribute to asthma pathogenesis through their metabolites [12]. Furthermore, Streptococcus salivarius, identified in the dysbiotic nasal microbiota of AR patients, modulates disease progression by influencing epithelial cell adhesion [9]. Beyond direct microbial interactions, microbial communities substantially influence the pathogenesis and modulation of allergic disorders through their metabolic byproducts [13], which are generated not only in the gastrointestinal tract, but also in the oral microenvironment [14]. Short-chain fatty acids, end products of dietary fibre fermentation, regulate the Th1/Th2 balance and enhance epithelial barrier function, thereby exerting anti-allergic effects [15]. Amino acid metabolites regulate Th17/Treg equilibrium via the aryl hydrocarbon receptor pathway, thereby modulating allergic responses [16]. Bacterial-derived d-tryptophan attenuates Th2-mediated immune reactions [17]. Vitamins exert direct immunomodulatory effects on immune cells and indirectly influence allergic progression through microbial environment modulation [18]. These metabolites collectively constitute a complex interaction network between the microbiota and the host, and their dysregulation is a significant factor in enhancing allergic susceptibility and exacerbating disease severity. Consequently, a comprehensive investigation of the microbial and metabolic contributions to AR pathogenesis through integrated microbiome-metabolome analyses is a crucial research direction.

The advent of next-generation sequencing technology has facilitated comprehensive investigations of bacterial community dynamics [19]. In contrast to conventional phenotypic identification approaches, 16S rRNA gene sequencing demonstrates superior accuracy for bacterial characterisation within clinical specimens by minimising technical artifacts and phenotypic heterogeneity [20]. Concurrently, liquid chromatography-mass spectrometry provides a robust platform for profiling diverse metabolites under pathological conditions, offering critical insights into microbial-host interactions and contributing significantly to the elucidation of disease mechanisms [21]. Despite these advancements, the interplay between the oral microbiota and their metabolic signatures in patients with AR remains poorly characterised. This study aimed to explore alterations in the microbial composition, differential metabolites, and their metabolic pathways within the supragingival plaques of patients with allergic rhinitis, to provide data that may support the pathogenesis and clinical treatment of allergic rhinitis.

Materials and methods

Study population and sample collection

From September 2024 to May 2025, patients with allergic rhinitis who met the diagnostic criteria of the Guidelines for Diagnosis and Treatment of Allergic Rhinitis (2022, Revised Edition) were recruited from the First Affiliated Hospital of Harbin Medical University. Baseline evaluations including standardised questionnaires, physical examinations, and allergen testing were conducted during the initial visit. Using predefined inclusion and exclusion criteria, this study achieved preliminary control of confounding factors such as body mass index and oral health status between the two groups. However, it should be acknowledged that diet, an important confounder influencing oral microbiota, was not adequately accounted for, and the results may be affected accordingly. The detailed inclusion and exclusion criteria are outlined in Supplementary Table S1. A total of 35 patients with AR and 35 healthy controls were included in this study. Following a 2-hour fasting period, supragingival plaque samples were aseptically collected from all molar surfaces using sterile flocked swabs, logged, and immediately stored in cryovials on dry ice, with subsequent transfer to −80 °C storage within 4 hours.

The study protocol was approved by the Institutional Review Board of the First Affiliated Hospital of Harbin Medical University (License No.: [2025JS77]). Written informed consent was obtained from all the participants in compliance with the Declaration of Helsinki.

DNA extraction and 16S rRNA gene sequencing

Total microbial community DNA was extracted from supragingival plaque samples using a FastPure Stool DNA Isolation Kit (MJYH, Shanghai, China). DNA integrity was assessed using 1% agarose gel electrophoresis, and DNA concentration and purity were determined using a NanoDrop 2000 spectrophotometer (Thermo Scientific, USA).

Using the extracted DNA as a template, the V3-V4 region of the 16S rRNA gene was amplified with barcoded primers 338F (5'-ACTCCTACGGGAGGCAGCAG-3') and 806 R (5'-GGACTACHVGGGTWTCTAAT-3'). The PCR reaction mixture (20 μL) contained 10 μL of 2 × Pro Taq, 0.8 μL each of forward and reverse primers (5 μM), 10 ng/μL template DNA, and ddH₂O to reach the final volume. The amplification programme consisted of initial denaturation at 95 °C for 3 min, followed by 27 cycles of denaturation at 95 °C for 30 s, annealing at 55 °C for 30 s, and extension at 72 °C for 45 s, with a final extension at 72 °C for 10 min. Each sample was amplified in triplicates. PCR products from the same sample were pooled in equal amounts, purified by 2% agarose gel electrophoresis, and quantified using a Synergy HTX microplate reader (Biotek, USA).

Purified PCR products were used to construct libraries using the NEXTFLEX Rapid DNA-Seq Kit (BioScientific, USA). After adaptor ligation, magnetic bead-based size selection, PCR enrichment, and magnetic bead purification, paired-end sequencing was performed using a NextSeq 2000 platform (Illumina, USA).

Raw sequencing data were quality controlled using Fastp (v0.19.6). Bases with a quality value below 20 (within a 50 bp sliding window) were filtered out. Reads with a length of less than 50 bp after quality control or those containing ambiguous bases (N) were removed. Paired-end reads were assembled using FLASH (v1.2.11) with a minimum overlap length of 10 bp and a maximum mismatch ratio of 0.2. The samples were demultiplexed based on barcodes (0 mismatches allowed) and primers (2 mismatches allowed), and the sequence orientations were corrected. OTUs were clustered at 97% similarity using USEARCH (v11) and chimeras were removed. Sequences annotated as chloroplasts or mitochondria were excluded from analysis. The sequencing depths for each sample are presented in Supplementary Table 2. The number of sequences per sample was reduced to 20,000, which resulted in an average sequence coverage of 99%. Taxonomic annotation was performed using the RDP classifier (v2.13) against the Silva database (v138, May 7, 2025) with a confidence threshold of 70%, and community composition at each taxonomic level was summarised.

Sample preparation and LC-MS analysis

Considering the cost of untargeted metabolomic analyses, a random sampling strategy was adopted. Of the 70 samples across the two groups that had already undergone 16S rRNA gene sequencing, six were selected from each group for untargeted metabolomic analysis.

A 50 mg solid sample was placed into a 2 mL centrifuge tube, to which one 6 mm grinding bead and 400 μL of extraction solution (methanol: water = 4:1, v/v) containing 0.02 mg/mL internal standard (L-2-chlorophenylalanine) were added. The samples were ground using a frozen tissue grinder for 6 min (−10 °C, 50 Hz), followed by low-temperature ultrasonic extraction for 30 min (5 °C, 40 kHz). The extract was allowed to stand at −20 °C for 30 min and then centrifuged at 13,000 × g for 15 min at 4 °C. The supernatant was collected for injection. Equal volumes of metabolites from each sample were pooled to prepare quality control (QC) samples, which were inserted every 5–15 sample to monitor the analytical reproducibility.

The LC-MS/MS analysis was performed using a UHPLC-Q Exactive HF-X mass spectrometry system (Thermo Fisher Scientific, USA). Chromatographic conditions were as follows: injection volume 3 μL; separation was achieved on an HSS T3 column (100 mm × 2.1 mm i.d., 1.8 μm). Mobile phase A consisted of 95% water and 5% acetonitrile containing 0.1% formic acid, whereas mobile phase B consisted of 47.5% acetonitrile, 47.5% isopropanol and 5% water containing 0.1% formic acid. The flow rate was 0.40 mL/min, and the column temperature was maintained at 40 °C. Mass spectrometry was performed in positive and negative ion-switching modes with a scan range of 70-1050 m/z. The sheath gas flow rate was 50 psi, the auxiliary gas flow rate was 13 psi, and auxiliary gas heater temperature was 425 °C. The spray voltages were 3500 V and −3500 V for positive and negative ion modes, respectively. The ion transfer tube temperature was 325 °C, and the normalised collision energy was set to a cycle of 20–40–60 eV. The resolution was 60,000 for the full MS scans and 7,500 for the MS/MS scans. Data were acquired in a data-dependent acquisition mode.

Raw data were processed using Progenesis QI (Waters Corporation) for peak detection, extraction, alignment, and integration. Compound annotation was performed by matching the in-house HMDB, METLIN, and Majorbio databases (July 30, 2025). The following preprocessing steps were applied to the annotated data: features with a missing rate exceeding 20% in any group were removed, missing values were imputed using the minimum value across all samples, peak intensities were normalised using the sum normalisation method, and variables with a relative standard deviation (RSD) of >30% in QC samples were excluded. The data were then log₁₀-transformed to obtain the final data matrix for subsequent analysis.

Statistical analysis

Clinical baseline characteristics

Comparisons of clinical baseline characteristics were performed using the SPSS software (v 27.0). For continuous variables, normality was assessed using the Shapiro-Wilk test. Variables following a normal distribution were presented as mean ± SD, and group comparisons were conducted using the independent samples t-test. Variables that did not follow a normal distribution were presented as median (interquartile range, IQR), and group comparisons were performed using the Mann–Whitney U test. Categorical variables are presented as counts (percentages), and group comparisons were conducted using the chi-square test or Fisher's exact test. A two-tailed p < 0.05 was considered statistically significant.

Microbiome data analysis

All data analyses were performed using the Majorbio Cloud Platform. Alpha diversity indices were calculated using Mothur software (v1.30.2). Intergroup differences in alpha diversity were assessed using the Wilcoxon rank-sum test with BH correction. PCoA based on the Bray-Curtis distance algorithm was performed to examine the similarity of microbial community structures among the samples. ANOSIM with 999 permutations was conducted to test the differences in the microbial community structure between the groups. Differential microbial taxa were identified using the Wilcoxon rank-sum test with BH correction, and a corrected P < 0.05 was considered statistically significant. Given the limited sample size of this study, a network analysis was performed as an exploratory analysis to preliminarily reveal potential co-occurrence patterns among microorganisms and generate hypotheses for future research. For correlation network analysis, genus-level co-occurrence networks were constructed separately for the AR and healthy control groups based on Spearman's correlation coefficient. Only correlations with |r| > 0.6 and p < 0.05 were retained for network visualisation [22,23]. Spearman's correlation coefficient was used to analyse the associations between the top 50 most abundant genera in the AR group and the clinical symptom data. Spearman correlation coefficients were calculated using R software (version 3.3.1), and heatmaps were generated using the pheatmap package. The colour gradient in the heatmap represents the sign and magnitude of the correlation coefficients (r), with |r| ≥ 0.1 as the threshold for colour display, and P < 0.05 was considered statistically significant.

Metabolomics data analysis

PCA and PLS-DA were performed on the processed data matrixes. A seven-fold cross-validation was applied to evaluate the model stability, with R²Y and Q² serving as the model evaluation metrics. Permutation testing (n = 200) was performed to verify that the model did not overfit. Significant differential metabolites were selected based on variable importance in projection (VIP) > 1 from the PLS-DA model and p < 0.05 from the Wilcoxon rank-sum test. Given the exploratory nature of this analysis, multiple testing corrections were not applied. Differential metabolites were annotated to metabolic pathways using the Kyoto Encyclopaedia of Genes and Genomes database (July 30, 2025). Pathway enrichment analysis was performed using the Python package scipy.stats, and the biological pathways most relevant to the experimental treatment were identified using Fisher's exact test. Spearman's correlation coefficient was used to analyse the relationship between the differentially abundant microbial genera and differential metabolites between the two groups. The method is the same as that described in the heatmap analysis section in Microbiome data analysis.

Results

Clinical characteristics of study participants

The study cohort comprised 70 participants, stratified into two groups: an allergic rhinitis (AR) group (n = 35; 12 males, 23 females; mean age 28.69 ± 6.64 years) and a healthy control (HC) group (n = 35; 14 males, 21 females; mean age 26.37 ± 4.07 years). Comprehensive clinical data, including demographic information and symptom profiles (nasal congestion, nasal itching, sneezing, rhinorrhea, and eye symptoms), were systematically collected using standardised questionnaires following sample acquisition. The comparative demographic and baseline characteristics of the two groups are presented in Table 1. No statistically significant differences in smoking history or alcohol consumption were observed between groups. Nevertheless, as both smoking and alcohol intake represent important confounders that affect the oral microbiome, the failure to adjust for these factors in this study constitutes a recognised limitation.

Table 1.

Demographics and clinical characteristics of study participants.

Characteristic	AR (n = 35)	HC (n = 35)	P value
Age (y) (Mean ± SD)	28.69 ± 6.64	26.37 ± 4.07	0.279
BMI (kg/m2) (Mean ± SD)	23.23 ± 4.63	22.49 ± 3.67	0.698
Sex (n, %)			0.621
Male	12 (34.3%)	14 (40.0%)
Female	23 (65.7%)	21 (60.0%)
Delivery (n, %)			0.597
Spontaneous delivery	26 (74.2%)	24 (68.5%)
Caesarean section	9 (25.7%)	11 (31.4%)
Feed (n, %)			0.780
Breastfeeding	27 (77.1%)	26 (74.2%)
Formula feeding	8 (22.8%)	9 (25.7%)
Cigarette (n, %)			0.076
Yes	6 (17.1%)	1 (2.8%)
No	29 (82.8%)	34 (97.1%)
Wine (n, %)			0.056
Yes	9 (25.7%)	3 (8.5%)
No	26 (74.2%)	32 (91.4%)
Family history (n, %)			<0.01
Yes	16 (45.7%)	2 (5.7%)
No	19 (54.2%)	33 (94.2%)
Total IgE (KU/L) (Mean ± SD)	300.9 ± 60.14	−

Data are expressed as mean ± SD, or number(percentage). P values calculated with independent-sample t test, X²or nonparametric tests as appropriate. AR, allergic rhinitis; HC, healthy control. n = 35 (AR), n = 35 (HC).

Oral microbiome profiling

Sequencing depth assessment

Rarefaction curves demonstrated that samples across all groups reached saturation plateaus, indicating an adequate sequencing depth to capture the majority of microbial community members within each group (Supplementary Figure 1). Rank abundance curves exhibited smooth distributions, reflecting high microbial richness and evenness in both groups (Supplementary Figure 2).

Alpha and beta diversity

Alpha diversity metrics, including the Ace, Chao, Shannon, and Simpson indices, revealed no significant differences between the AR and HC groups (Supplementary Figures 3). Beta diversity analysis conducted using principal coordinate analysis (PCoA) showed no distinct clustering patterns between the two groups (Supplementary Figure 4). This observation was further corroborated by ANOSIM, which confirmed that there were no significant differences in the microbial community structure.

Microbial composition in AR patients vs. healthy controls

Following stringent quality control, 4,274,097 high-quality sequences were generated from 70 supragingival plaque samples, which were clustered into 1,473 OTUs at a 97% similarity threshold. AR and HC groups exhibited 347 and 279 unique OTUs, respectively, with 847 shared OTUs (Figure 1a).

Figure 1.

Classification and composition of microbiota in AR patients versus HC. (a) Venn diagram of oral microbiome in AR and HC groups. (b) Phylum-level composition of oral microbiome in AR and HC groups. (c) Genera-level composition of oral microbiome in AR and HC groups. (d) Differential genera identified between the AR and HC groups. (e) Differential OTUs identified between the AR and HC groups. The wilcoxon rank-sum test with BH correction was applied, and adjusted p < 0.05 was considered statistically significant. AR, allergic rhinitis; HC, healthy controls. n = 35 (AR), n = 35 (HC).

At the phylum level, the predominant taxa in both groups were Bacillota, Pseudomonadota, Actinomycetota, Bacteroidota, and Fusobacteriota (Figure 1b). At the genus level, the core oral microbiota comprised Streptococcus, Haemophilus, Neisseria, Actinomyces, Fusobacterium, Rothia, Leptotrichia, Veillonella, Corynebacterium, and Capnocytophaga (Figure 1c).

The results of the intergroup comparison of microbial differences showed that at the genus level, Abiotrophia was significantly enriched in the AR group, whereas Rothia and Actinobacillus were more abundant in the HC group (adjusted p < 0.05) (Figure 1d). At the species level, Prevotella melaninogenica, uncultured Abiotrophia sp., Capnocytophaga sputigena, etc. were enriched in the AR group. The HC group was enriched with Actinomyces massiliensis, uncultured Haemophilus sp., Actinobacillus sp., Candidatus Gracilibacteria bacterium HOT-871, Actinomyces israelii, Selenomonas sp. oral taxon 137 str. F0430 (adjusted p < 0.05) (Figure 1e).

Microbial co-occurrence network analysis

Co-occurrence networks were constructed based on the 50 most abundant microbial genera in AR and HC groups. The network of the HC group (Figure 2a) exhibited greater overall complexity than that of the AR group (Figure 2b). The HC network contained 36 nodes and 128 edges, whereas the AR network contained 35 nodes and 100 edges. Two key modules were identified in the HC group (module 1: nine nodes, 25 edges; module 2: four nodes, five edges), whereas the AR group featured only one key module with higher connectivity (11 nodes, 51 edges). The overall network was more complex in the HC group, but the key modules in the AR group showed significantly higher node numbers and connection densities than any key module in the HC group (Table 2).

Figure 2.

Network analysis of HC and AR groups and correlation analysis of genus-level microbiota with clinical symptoms in the AR group. Network analysis of HC and AR groups. (a) Network of the HC group. (b) Network of the AR groups. Spearman's |r| > 0.6 and unadjusted p < 0.05 were retained for network visualisation. Nodes represent bacterial taxa, and edges represent significant correlations, with the edge width proportional to the correlation strength. The red and blue edges indicate positive and negative correlations, respectively. Correlation heatmap between clinical symptoms and bacterial taxa in the AR group. (c) Correlations between clinical symptoms of patients with AR and oral bacterial taxa. Spearman’s correlation analysis between the top 50 microbial genera (relative abundance) and clinical symptom data in the AR group. unadjusted P values are shown. *p < 0.05, **p < 0.01,***p < 0.001. Red and blue indicate the positive and negative correlations, respectively. AR, allergic rhinitis; HC, healthy controls. n = 35 (AR), n = 35 (HC).

Table 2.

Descriptive statistics of network components by group.

Module	Node	Edge
HC	36	128
AR	35	100
Key Module 1(HC1)	9	25
Key Module 2(HC2)	4	5
Key Module 3(AR1)	11	51

AR, allergic rhinitis group; HC, healthy control group; HC1, key module 1 for healthy control group; HC2, key module 2 for healthy control group; AR1, key module for allergic rhinitis group. n = 35 (AR), n = 35 (HC).

Association between microbes and clinical features

Spearman’s rank correlation analysis was performed between the top 50 genera and clinical symptoms of AR. The results indicated that Abiotrophia and Granulicatella were positively correlated, whereas Kingella was negatively correlated with nasal congestion. Arachnia was positively correlated with nasal itching, Actinobacillus was negatively correlated with nasal itching. Capnocytophaga, Arachnia, Bergeyella, and Neisseria were positively correlated with sneezing, whereas Actinobacillus, Gemella, and Streptococcus were negatively correlated. Aggregatibacter was positively correlated with eye symptoms (Figure 2c).

Oral metabolomics analysis

Sample quality control analysis

Overlaid base peak ion chromatograms of all quality control samples showed good overlap in both positive and negative ion modes, with minimal fluctuations in retention time and peak response intensity, indicating stable instrument performance throughout the analysis (Supplementary Figures 5a, b). Principal Component Analysis of QC and experimental samples revealed near-overlapping QC sample points, indicating instrument stability and good data reproducibility (Supplementary Figure 5c).

Metabolite identification and classification

A total of 362 and 335 metabolites were identified in the positive and negative ion modes, respectively. Identified metabolites were classified using KEGG and HMDB databases. The recognised metabolites primarily belonged to the categories of bioactive compounds (n = 83; Supplementary Figure 6a), phytochemical compounds (n = 83; Supplementary Figure 6b), and lipids (n = 105; Supplementary Figure 6c). The top 20 KEGG pathways involved 220 metabolites, predominantly ABC transporters, protein digestion and absorption, purine metabolism, D-amino acid metabolism, and lysine degradation (Supplementary Figure 6 d).

Differential metabolites and pathway enrichment analysis

The PCA results showed that the variance contribution rates of the first and second principal components were 35.7% and 22.7%, respectively. Samples from the AR and HC groups exhibited a certain separation trend, suggesting potential differences in the microbial community composition of supragingival plaque between the two groups. QC samples showed tight clustering, indicating that the metabolomic data obtained in this study had good stability and reproducibility (Figure 3a).

Figure 3.

Supragingival plaque metabolites signature in AR patients versus HC. (a) PCA of supragingival plaque metabolomics data from AR, HC, and QC samples. (b) PLS-DA shows distinct metabolite profiles between the AR and HC groups. (c) Volcano plot identifies metabolites that differ between the AR and HC groups. Due to the exploratory nature and small sample size, no multiple testing correction was applied. Metabolites with VIP > 1 (PLS-DA) and unadjusted p < 0.05 (Wilcoxon rank-sum test) were considered significantly differential. (d) Thirteen enriched pathways with the most significant differences between the AR and HC groups were identified based on the KEGG database. P-values were calculated using Fisher's exact test (unadjusted). The size of each point represents the number of metabolites in the corresponding pathway. AR, allergic rhinitis; HC, healthy controls. n = 6 (AR), n = 6 (HC).

PLS-DA results showed a clear separation between the AR and HC groups (Figure 3b). PLS-DA model validation indicated no overfitting (R [2] = 0.984, Q² = 0.758, the intercept of Q² was −1.277) (Supplementary Figure 7a, 7b), supporting the reliability of the results.

By comparing the metabolites identified between the AR and HC groups and selecting those with VIP > 1 and Wilcoxon rank-sum test, p < 0.05, 106 differential metabolites were identified as candidate differential metabolites (67 upregulated and 39 downregulated in AR) (Figure 3c). Among the most significant differential metabolites between the groups, Cadaverine, Putrescine, Spermine, and L-Glutamine were upregulated, while Endomorphin-2 was downregulated in AR (Figure 4a).

Figure 4.

Analysis of differential metabolites between AR and HC groups and differential microorganisms’ association with differential metabolites. (a) Five differential metabolites identified between the HC and AR groups: cadaverine, putrescine, spermine, L-glutamine, and endomorphin-2. p < 0.05 (Wilcoxon rank-sum test). *p < 0.05 For differential metabolites, Wilcoxon rank-sum test was performed without multiple testing correction (exploratory analysis). (b) Association between oral microbiota and differential metabolites. Spearman correlation analysis was performed between differentially abundant microbial genera and the five differential metabolites. Red and blue indicate positive and negative correlations, respectively. *p < 0.05, **p < 0.01,***p < 0.001. For correlation analysis, Spearman correlation was performed without multiple testing correction; unadjusted P values are shown. AR, allergic rhinitis; HC, healthy control. n = 6 (AR), n = 6 (HC).

KEGG enrichment analysis revealed that the differential metabolites were significantly enriched in pathways such as glutathione metabolism, protein digestion and absorption, D-amino acid metabolism, arginine and proline metabolism, and lysine degradation (Figure 3d).

Integrated analysis of microbes and metabolites

An integrated analysis was performed using all differential metabolites and microbes at the genus level in the AR and HC groups (Supplementary Figure 8). Heatmap analysis revealed that Rothia exhibited significant negative correlations with spermine, cadaverine, L-glutamine, and putrescine. Abiotrophia showed significant positive correlations with cadaverine, L-glutamine and putrescine (Figure 4b).

Discussion

This study revealed differences in the supragingival plaque microbiome and metabolome between adult patients with AR and healthy controls. AR, a prevalent chronic respiratory disorder characterised by Th2-type immune responses, represents a substantial global burden [24,25]. Although significant advancements have been made in elucidating the roles of respiratory and gut microbiota in respiratory diseases [26–29], the influence of the oral microbiome on AR remains underexplored. Considering the pivotal role of microbial communities and their associated metabolites in host immune interactions [30], we performed a comparative analysis of the supragingival plaque microbiota in 35 patients with AR and 35 healthy controls to preliminarily explore the potential role of oral microbial changes in AR.

Using 16S rRNA high-throughput sequencing, we observed no significant differences in alpha diversity indices between the AR and HC groups (Supplementary Figure 3). To date, few studies have examined supragingival plaque microbial diversity in adult patients with AR. However, existing research on microbial diversity and richness in both adults and children with asthma or allergic diseases has yielded inconsistent findings; for instance, 7-year-old children with allergic diseases exhibited significantly lower salivary microbial diversity than healthy children [31], whereas African American asthmatic children demonstrated significantly higher salivary microbial diversity than healthy controls [32]. Given the divergent patterns of alpha diversity across studies, this metric may not serve as a reliable indicator for characterising AR disease status or pathogenesis. Furthermore, no significant differences in beta diversity were observed between the AR and HC groups, suggesting that functional differences in the microbial communities may not be contingent on alterations in beta diversity (Supplementary Figure 4).

Although no significant differences in microbial diversity were observed between the HC and AR groups, distinct microbial taxa were identified in the supragingival plaques of both cohorts (Figure 1a). These AR-specific OTUs may serve as potential biomarkers of this disease. Further investigations are required to validate their consistency across diverse AR populations and oral microenvironments as well as to evaluate their therapeutic potential as novel targets for AR management. At both phylum (Figure 1b) and genus (Figure 1c) levels, the predominant microbial composition was comparable between the two groups. However, at the genus level, the AR group exhibited a significant reduction in the relative abundance of Rothia and a marked enrichment of Abiotrophia compared with the HC group (Figure 1d). Rothia, a commensal genus of the upper respiratory tract, has been shown to decrease the risk of asthma in children, and its colonisation in germ-free mice has been demonstrated to mitigate airway inflammation in offspring, underscoring its protective role against asthma and other allergic diseases [33]. Additionally, Rothia species can inhibit respiratory pathogens, such as those associated with asthma, through the competitive secretion of SagA, thereby modulating disease progression [34]. In contrast, Abiotrophia has been implicated in infective endocarditis and the promotion of inflammation [35,36]; its enrichment in the AR group suggests that it may participate in the disease process by influencing systemic inflammation. At the species level, distinct microbial profiles were enriched in AR and HC groups (Figure 1e). Notably, Prevotelal melaninogenica, which was enriched in the AR group, is known to disrupt the oral epithelial barrier and induce inflammation via modulation of IL-36 subfamily cytokines [37], and the IL-36 subfamily plays a pivotal role in the pathophysiology of various inflammatory diseases, including AR. [38]. Furthermore, Capnocytophaga sputigena, also enriched in the AR group, has been associated with periodontitis [39], highlighting the need to explore the potential link between periodontal disease and AR, as well as the impact of periodontal health improvement on AR outcomes.

Network analysis based on single-factor correlations was employed as an exploratory tool, and the results were primarily intended to generate testable scientific hypotheses for subsequent investigations. The microbial interaction network observed in the healthy control group (Figure 2a) exhibited greater topological complexity than that observed in the AR group (Figure 2b), which is broadly consistent with prior reports of reduced microbial co-occurrence network complexity in children with AR and asthma [40,41], potentially suggesting a shared pattern. Distinct core modules within the network may reflect co-occurring or co-regulated bacterial assemblages or indirectly reflect a certain partitioning of community functions. The core modules in the AR group demonstrated increases in both node count and connections compared to those in the control group (Table 2), a phenomenon warranting further attention in future studies. Furthermore, Rothia occupied the keystone species position in the healthy control network, whereas this role was not observed in the AR group core modules, providing a preliminary clue for exploring its potential function in maintaining oral microecological homoeostasis.

Emerging evidence suggests that distinct stressor profiles and mental health symptom manifestations during early pregnancy are associated with specific oral microbiome signatures, with microbial community variations primarily influenced by the nature of the clinical symptoms [42]. In addition, in a cohort of individuals with heterogeneous psychiatric diagnoses and non-psychiatric controls, schizophrenia symptoms were associated with particular bacterial taxa [43]. Extending these findings, we employed correlation heatmaps to examine the associations between the oral microbiota and AR clinical manifestations. The analysis revealed significant correlations between specific bacterial genera and various AR clinical symptoms, indicating the potential role of the oral microbiota in modulating AR symptomatology (Figure 2c).

The metabolomic analysis identified 106 candidate metabolites (Figure 3c). Based on the pathophysiological characteristics of AR, we focused on five candidate differential metabolites that were primarily enriched in glutathione metabolism, lysine degradation, d-amino acid metabolism, arginine and proline metabolism, and neuroactive ligand-receptor interaction pathways (Figure 4a, Figure 3d). These metabolic pathways are involved in immunomodulation, inflammatory responses, and neuroimmune-inflammatory regulatory processes (Figure 5). However, given the limited sample size of this study, the biological significance of these enriched pathways remains to be validated in large-scale cohorts. Lysine is an essential amino acid that cannot be synthesised by the human body, and microorganisms play a critical role in lysine metabolism [44]. Previous studies have demonstrated that the lysine degradation pathway provides energy for bacterial growth under anaerobic conditions [45]. In this study, the abundance of periodontal-associated anaerobic bacteria was elevated in the AR group, concomitant with an enrichment trend in the lysine degradation pathway. Although these two observations co-occurred at the observational level, a causal relationship could not be inferred. Furthermore, cadaverine, a product of lysine degradation, was enriched in the AR group. Literature reports indicate that it can elevate histamine levels through competitive inhibition of diamine oxidase activity [46,47], providing a clue for exploring its potential role in nasal mucosal inflammation. At least one-third of the total human d-amino acid pool is derived from microbial synthesis [48]. The D-amino acid metabolism pathway was enriched in this study, suggesting that this pathway may be involved in local respiratory mucosal inflammation through the activation of inflammatory signalling pathways and the promotion of inflammatory cytokine secretion [49], although this association was not validated in the present study. Arginine metabolism generates polyamines, including putrescine and spermine [50], and existing studies have established associations between these polyamines and airway inflammation and epithelial injury [51,52]. Glutathione is a critical regulatory factor for normal immune system function with protective and immunomodulatory roles [53,54]. Additionally, endogenous opioid peptide endomorphin-2 was enriched in the neuroactive ligand-receptor interaction pathway; literature reports indicate that it can bidirectionally modulate neurogenic inflammation and immune cell activity through activation of μ-opioid receptors [55–57]. Notably, the aforementioned metabolite and pathway analyses represent exploratory findings based on limited samples, and interpretation of their biological significance requires further validation through independent cohort studies with larger sample sizes and in vitro functional experiments.

Figure 5.

Schematic of the immune, inflammatory, and neuro-immune-inflammatory axis in AR pathogenesis. The box on the right displays the differential metabolic pathways affecting the AR mechanism.

Correlation analysis between microorganisms and metabolites revealed significant associations between the differentially abundant genera Rothia and Abiotrophia and candidate metabolites, including cadaverine (Figure 4b). Genome annotation based on the Kyoto Encyclopaedia of Genes and Genomes database indicated that Rothia possesses complete lysine degradation (map00310) and glutathione metabolism (map00480) pathways. Cadaverine, a convergence molecule in these two metabolic pathways, serves as both an intermediate product of lysine degradation and a substrate for glutathione conjugation. Based on these findings, we propose the following hypothesis: the negative correlation between Rothia abundance and cadaverine levels may reflect a covariation relationship between this genus and the metabolite or may suggest an association within shared biological processes. Whether Rothia modulates host metabolism to influence cadaverine levels and subsequently participates in immune or inflammatory responses requires further investigation.

Although this study provides several valuable findings, several factors must be considered. First, the study population was exclusively recruited from northern China, with a limited sample size, which may restrict the generalisability of the findings and introduce regional specificity. Constrained by the cost of metabolomic profiling, the sample size for the metabolomic analysis was small. Therefore, the metabolomics and related integrative analyses in this study are exploratory and require further validation in large-scale independent cohorts. Second, as a cross-sectional study, it can only reveal associations among variables, and causal relationships cannot be inferred. Third, although basic demographic information was recorded, important potential confounders affecting the oral microbiome, including oral hygiene habits, dietary habits, smoking, and alcohol consumption, were not comprehensively assessed; subsequent studies should incorporate these factors to control confounding bias more thoroughly. Fourth, this study employed 97% similarity clustering for OTUs rather than the higher-resolution ASV approach. Although this choice was made to ensure comparability with previous studies, the limited taxonomic resolution is a limitation of this study.

Conclusion

In conclusion, this study observed significant differences in supragingival plaque microbial composition between patients with AR and healthy controls and identified several differential metabolites and associated metabolic pathways. However, constrained by the cross-sectional study design, limited sample size, and uncontrolled confounders, causal relationships between the microbiome-metabolome and AR could not be established, nor could it be determined whether alterations in microbial composition represent a cause or consequence of the disease. These exploratory findings await further validation through longitudinal studies with larger sample sizes and functional experiments, to provide more robust evidence for subsequent in-depth investigations into the pathogenesis and potential therapeutic targets of allergic rhinitis.

Supplementary Material

Revised supplementary figures.pdf

Revised supplementary figures.pdf

Revised supplementary Table.docx

Revised supplementary Table.docx

Funding Statement

This work was supported by the Natural Science Foundation of Heilongjiang Province of China under [Grant numbers: PL2024H061, JJ2024ZL0027]; and Harbin Medical University under [Grant number: YJSCX2024-103HYD].

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability statement

The data that support the findings of this study are openly available in NCBI BioProject at https://www.ncbi.nlm.nih.gov/, reference number PRJNA1470146. The data that support the findings of this study are openly available in MetaboLights at https://www.ebi.ac.uk/metabolights/, reference number MTBLS14582.

Ethical approval

The study was conducted in compliance with the Helsinki Declaration (updated version, 2013) and was approved by the Ethics Committee of the First Affiliated Hospital of Harbin Medical University. Research was carried out under the approved protocols. All participants provided written informed consent.

Supplementary material

Supplemental data for this article can be accessed at https://doi.org/10.1080/20002297.2026.2697535.