Colonisation of the gut microbiome during the neonatal period is crucial for establishing a defence against environmental antigens in the mucosal immune system.1 The development of the gut microbiota during infancy is influenced by various factors, including the host genotype, mode of delivery and feeding type.1–3 Notably, feeding type significantly regulates the diversity and composition of the gut microbiota during the first years of life.4 Disturbances in the gut microbiome during early life have been associated with several non-communicable diseases, including atopic dermatitis (AD).5–7
AD is a common, chronic, pruritic inflammatory skin disease characterised by increased cutaneous reactivity to environmental triggers. It affects at least 15% of children.8 Although AD often manifests in infancy and early childhood, it can persist into adolescence and adulthood.9 Numerous studies have suggested that the intestinal microbiota of infants plays a critical role in the development of AD.10–13 However, the exact aetiology of this disease remains unclear. A recent study indicated that breast feeding results in lower operational taxonomic unit counts and α-diversity while formula feeding induces the opposite.4 Moreover, although specific microbial taxa have been associated with AD, such as Escherichia and Bifidobacterium species,14 the contribution of particular bacterial species to disease onset remains unclear.
Despite observing differences in the gut microbiota before the onset of AD, identifying the causes of these differences between healthy infants and those with AD is challenging because of numerous confounding factors affecting gut microbiome colonisation.15–17 For instance, similar feeding types with different nutritional ingredients, such as breast feeding with or without the introduction of allergenic foods,5 can lead to distinct gut microbial perturbations during infancy.4 Consequently, we hypothesised that breast milk with varying compositions could affect the gut microbial community in infants. However, studies on the complex relationship between the gut microbiome during infancy and breast milk-derived metabolites are limited. Furthermore, it remains unclear how and which breast milk-derived metabolites influence the gut microbiome of infants with AD. In this study, we conducted metagenomic and metabolomic analyses, supported by animal experiments, to explore the potential relationship between breast milk-derived metabolites and gut dysbiosis in infants with or without AD.
Materials and methodsSubject recruitmentA total of 250 healthy, young female volunteers, free of systemic disease and no history of antibiotics treatment in the past months, were recruited at the Department of Obstetrics and Paediatrics, The First People’s Hospital of Foshan, Guangdong, China. Up to the time of sampling, none of infants had received antibiotics treatment.
Sample collectionIn this longitudinal study, faecal samples of infants in different stages were collected. Concurrently, maternal faecal sample in the third trimester of pregnancy (T3) and breast milk samples at 30–42 days postpartum were collected. All samples were immediately frozen at −80℃ on collection.
DNA extraction and metagenomic sequencingDNA was extracted from faecal samples for subsequent use in sequencing library preparation. The prepared DNA libraries were then sequenced using the Illumina HiSeq2500 sequencer. The sequencing process used the HiSeq PE Cluster Kit V.4 and the HiSeq SBS V.4 250 cycle kit (Illumina, San Diego, California, USA).
Metabolomic detectionMetabolites in breast milk samples, collected between 30 and 42 days post partum, were analysed using a Liquid Chromatograph Mass Spectrometer.18 The raw data obtained were processed with Compound Discoverer (V.3.1) to confidently identify metabolites. This identification was facilitated by comparison with a comprehensive database. Only metabolites exhibiting a coefficient of variance <30% were considered for subsequent analysis.
For the analysis of mouse faecal metabolites, the collected stool samples underwent untargeted metabolomic profiling. This was conducted using a Vanquish UHPLC system coupled with an Orbitrap Q Exactive series mass spectrometer (Thermo Scientific, Massachusetts, USA). The Kyoto Encyclopaedia of Genes and Genomes (KEGG) pathway analysis was performed using the MetaboAnalyst platform (V.5.0).19
Metagenomic data analysisMetagenomic data underwent quality control using FastQC20 and Cutadapt.21 Subsequently, potential human reads were removed using HiSAT2.22 The remaining non-human reads were aligned to our custom Refseq microbial genome database, which allowed us to categorise them into distinct taxonomic units. For data normalisation, the metric Reads Per Kilobase per Million mapped reads was employed. Additionally, the abundance of microbial KEGG pathways was assessed using the HMP Unified Metabolic Analysis Network (HUMAnN3, V.3.0).23
Decontamination of metagenomic dataTo further filter out potential contaminants, the R package Decontam (V.4.3.0)24 was used. In this process, blank samples were considered as presumptive contaminants.
Prediction of major histocompatibility complex affinityThe major histocompatibility complex (MHC), also known as human leucocyte antigen (HLA), plays a crucial role in stimulating T cell activity. The MHC affinity of various bacterial genes was predicted using NetMHC (V.4.0) and NetMHCIIpan (V.2.1).
Cell culture and inflammatory stimulationThe human keratinocyte cell line HaCaT was used to study the direct effects of eicosapentaenoic acid (EPA) and AA on immunomodulatory functions of keratinocytes in vitro. Cell cultures were maintained at 37 ℃ in a humidified atmosphere containing 5% CO2. For the experiment, cells were seeded into 96-well plates and subsequently stimulated with a mixture of TNF-α and INF-γ. Cell viability was evaluated using the Cell Counting Kit-8 (CCK8 Kit, Beyotime, China).
Mouse AD model of ovalbumin-induced dermatitisFemale BALB/c mice were purchased from Cyagen, China, and were subsequently housed in an accredited animal facility at the Cyagen Animal Centre. To induce allergen-induced dermatitis, ovalbumin was administered to the mice. Following the induction phase, samples from the mice were collected and used for subsequent experimental analyses.
Flow cytometryTo analyse the proportion of immunocytes following induction by AA or EPA, freshly isolated and stimulated mouse peripheral blood mononuclear cells (PBMCs) were stained with a panel of commercially available antibody.
H&E stainingSkin and intestine tissues were embedded in paraffin and sectioned into approximately 5 µm thick slices. These sections were performed HE staining.
RNA sequencing and transcriptomic data processingTotal RNAs were extracted from skin and intestinal tissues, followed by reverse transcription, library construction and high-throughput sequencing. The transcriptomic data from both skin and intestinal tissues were preprocessed to remove adaptors using Cutadapt. The cleaned data were then aligned to the GRCm39.genome.fa reference genome with Tophap2.25 Subsequently, FeatureCounts26 was used to quantify read counts for each gene. Finally, the expression profiling data were normalised using the TPM (Transcripts Per Kilobase Million) method.
Liquid chromatography/mass spectrometry of polyunsaturated fatty acidThe concentrations of polyunsaturated fatty acids (PUFAs), such as AA, EPA and DHA, in breast milk were quantified using liquid chromatography/mass spectrometry.
Faecal collection and faecal microbiota transplantationFaecal samples were collected from healthy female BALB/c mice (stool-HM) and prepared to create a homogenate. Mice were intragastrically administered with stool-HM with or without two selected bacteria and mannan. The procedure was conducted twice weekly.
The quantitative PCR assaysMessenger RNA was extracted from the samples, and complementary DNAs (cDNAs) were synthesised from the mRNA, the cDNAs were then used as templates to perform triplicate quantitative PCR experiments. The results obtained from these experiments were analysed using the 2−ΔΔCT method to determine the relative gene expression levels.
Statistical analysisData are presented as means±SE. Statistical analysis between groups was analysed according to the data distribution and data type. The threshold for statistical significance was assumed to be p<0.05.
Additional experimental methods and information are provided in online supplemental materials file.
ResultsOverview of the longitudinal metagenomic dataThe flow chart depicts the participant recruitment and sampling stages for this cohort (figure 1A). To evaluate potential contamination during the sampling and library construction periods, RNase-free water was used as a negative control (NC), whereas Mec served as a baseline to determine the quality of infant faecal samples after sequencing. Mec reflects the intrauterine environment, which is generally considered sterile.27 The faecal DNA concentration of Mec was identical to that of the NC but showed a significant difference compared with the first (1st M), third (3rd M), sixth (6th M) months and maternal faecal samples (figure 1B). The content of identified bacterial reads in most of the Mec sequencing data was similar to that of the NC but still showed an obvious discrepancy with others (figure 1C). Consistent with our previous study,28 bacteria were the primary microbiotas, and their composition in Mec remained different from that of the others, except for the NC (online supplemental figure 1A). To further confirm these findings, principal component analysis (PCA) based on normalised relative abundance revealed significant differences among the samples (figure 1D and online supplemental figure 1B). Detailed analysis revealed that Proteobacteria, Firmicutes and Actinobacteria were the primary phyla (figure 1E). Their composition in Mec as well as the microbial pathways showed significant differences from those of the other stages (figure 1E–G). In summary, we conclude that the faecal samples used in this study are suitable for further investigation.
Figure 1Overview of the data in this study. (A) Flow chart depicted the subject’s recruitment and sampling stages of this cohort. The dotted line in the middle arrow represents the sampling stages antepartum, whereas the solid line indicates the stages post partum. (B) DNA concentrations of selected samples. (C) Quality control of sequencing data. (D) Principal coordinate analysis based on relative abundance revealed the significant difference of microbial structure. (E) Relative abundances of microbial composition. Bacteria with abundances lower than 1% at the species level were grouped into category ‘others’. (F) Disease status corresponding to each sample. (G) Relative abundance of microbial pathways. AD, atopic dermatitis; NC, negative control.
Breast milk and its metabolites showed a relationship between the gut microbiome and ADInitial clinical statistical analyses revealed that a decrease in breastfeeding and an increase in formula feeding from the 3rd M were associated with a gradual reduction in the morbidity of infant AD (figure 2A,B and online supplemental figure 2A). During the observation period, which spanned from birth to 6 months, 78 infants (~31.2%) were diagnosed with AD and 27 (~10.8%) without (figure 2D and online supplemental figure 2B). The samples were manually divided into two groups, AD and healthy, based on clinical data from the 6th M for further analysis. Factors, such as delivery mode,2 infant sex29 and maternal body mass index,30 did not influence the infant gut microbiome (online supplemental figure 2C). Although it has been established that feeding type can impact the infant gut microbiome,31 with breast feeding being the predominant daily feeding method (online supplemental figure 2D, low), our analysis of the gut microbiome from randomly selected samples revealed no significant influence from the feeding type (online supplemental figure 2D, upper). Interestingly, we did observe differences between the AD and the healthy groups. PCA results indicated that the infant gut microbiota also exhibited perturbations from the 3rd M (figure 2C). This is consistent with the findings of previous studies suggesting that breastfeeding plays a crucial role in shaping the gut microbiome.4 32 33 These findings suggest that the metabolites in breast milk may be associated with AD onset.
Figure 2Breast milk-derived metabolites affected gut microbiome associating with disease status. (A) Clinical statistical data of infants with and without AD. (B) Statistics of feeding type in daily milk time. P values were calculated using one-way ANOVA based on Brown-Forsythe test. (C) Principal coordinate analysis (PCA) revealed the differences of gut microbiome across sampling times. Upper, sampling time, lower, PCA results based on relative abundance revealed the difference between AD and healthy groups. (D) Venn plot showed a statistics of infants with AD across different ages. (E) PCA result based on the abundance of metabolites revealed differential breast milk-derived metabolites. (F) KEGG pathway of differential metabolites. Negative P refers to the calculation of p value based on the abundance of metabolites detected in the negative ion mode. Positive P indicates that the p values were derived from the abundance of metabolites in positive ion mode. Meta P denotes the use of the overall abundance of all metabolites for the calculation of p values. KEGG was evaluated using MetaboAnalyst. (G) Pairwise correlation between differential bacteria and differential metabolites. Metabolites with significantly positive correlation was showed. (H) A similar molecular structure was found between EPA and 3-aminonon-5-enoic acid. (I) EPA and AA were catalysed into different metabolites by the same enzyme (modified from the review of Ishihara et al 65). (J) The breast milk-derived metabolites showed comparable abundance between AD and healthy grouVertical dotted lines indicated the mean value of abundance. Density curves indicated the data distribution. AUC which was used to show the similar group separation of disease. (K) The hypothesis explored the relationship between breastfeeding and AD onset. The thickness of line in Sankey diagram reflected the morbidity of AD. AA, arachidonic acid; AD, atopic dermatitis; ANOVA, analysis of variance; AUC, area under the curve; DHA, docosahexaenoic acid; EPA, eicosapentaenoic acid; KEGG, Kyoto Encyclopaedia of Genes and Genomes; LTs, leukotrienes; PGs, prostaglandins.
The composition of breast milk-derived metabolites changes over the lactation period.34 Colostrum is richer in protein but has a lower fatty acid content, whereas mature milk contains higher levels of fatty acids and less protein.35 36 To identify the metabolites that may affect the onset of AD, we conducted an untargeted metabolomic analysis using mature milk samples collected from 30 to 42 days post partum (figure 1A). This analysis revealed 78 differential metabolites (figure 2E and online supplemental figure 3A), several of which were derived from EPA and were involved in anti-inflammatory processes (figure 2F). Pairwise correlation analysis between the differential metabolites (online supplemental figure 3B) and differential bacteria (online supplemental figure 3A) indicated that two breast milk-derived metabolites, fucose (which was more concentrated in the AD group) and 3-aminonon-5-enoic acid (which was more prevalent in the healthy group), showed a positive correlation with the gut microbiota (figure 2G). Previous studies have reported that the gut microorganisms metabolise fucose to produce short-chain fatty acids associated with immunomodulatory effects.37–39 3-aminonon-5-enoic acid, which has a molecular structure similar to that of EPA (figure 2H), may also have a predictable anti-inflammatory effect.40 Thus, we propose that breast milk-derived metabolites influence the gut microbiome, and this relationship might be linked to AD.
EPA and docosahexaenoic acid (DHA) are metabolised by cyclooxygenases (COXs) and lipoxygenases (LOXs) to produce substrates with anti-allergic and anti-inflammatory effects (figure 2I). In contrast, arachidonic acid (AA) is metabolised into prostaglandins (PGs) and leukotrienes (LTs), which are known to promote inflammation.41 42 We found that the concentration of AA was higher in the AD group than in the healthy grouConversely, the levels of EPA and DHA, along with their common derivative, 18-hydroxy-eicosapentaenoic acid, were lower in the AD group (figure 2J). In addition, the abundance of AA was approximately 15 times greater than that of EPA or DHA (figure 2J). These findings regarding PUFAs were confirmed by targeted metabolomic analysis of breast milk (online supplemental figure 3D,E). Based on these findings, we hypothesised that AA, a metabolite derived from breast milk, is associated with AD morbidity in infants (figure 2K).
AA affected the gut microbiome of infantsTo further investigate the role of AA in regulating gut microbiome dysbiosis, which may trigger the onset of AD, we categorised breast milk samples into two groups based on their AA, EPA or DHA content. The accuracy of this classification was assessed using receiver operating characteristic curves (online supplemental figure 4A). Distinct separation revealed significant differences in the metabolite composition between the two groups (online supplemental figure 4B). However, high concentrations of EPA, but not of DHA, affected the gut microbial structure at the 6th M (online supplemental figure 4C). In contrast, AA significantly affected the infant gut microbiome in the 1st M, even at low concentrations (online supplemental figure 4C). These findings suggest that AA has a more pronounced effect on gut microbial disturbances during infancy.
Although the gut microbial structure changes, it remains unclear whether the microbial pathways are affected. According to the results showing that the microbial structure in the 6th month was simultaneously affected by PUFAs (online supplemental figure 4C), a comparison of microbial pathways was conducted between high and low concentrations of PUFAs. The result indicated a significant difference in the microbial pathways of AD group when exposed to different concentrations of AA (online supplemental figure 4D). Further investigation using PCA indicated that only AA, but not EPA, significantly influenced microbial functional pathways from the 1st M (figure 3A and online supplemental figure 5A). To elucidate the microbial pathways, we categorised the differential pathways into five groups (online supplemental figure 5B and online supplemental table 1). A consistent pattern was observed in which the metabolism of amino acids and fatty acids shifted between the healthy and AD groups under high concentrations of AA (figure 3B, online supplemental figure 5C and online supplemental figure 6). We employed the fold change (FC) derived from relative abundance to assess the activity of differential microbial functional pathways potentially linked to gut dysbiosis. The findings revealed that the β-(1,4)-mannan degradation pathway, with an FC of 11.07, was the most active among the analysed pathways (FC range: 1.02–1.78, figure 3C). This suggests that certain mannan-degrading bacteria may be associated with gut microbial imbalance when exposed to high concentrations of AA.
Figure 3High concentration of AA influenced gut microbiome and their functional pathways. (A) Comparison of gut microbial pathways under high concentration of metabolites. (B) Microbial pathways shifted between AD and healthy groups under high concentrations of AA. (C) The activity of mannan-related pathways in AD group. Differential microbial pathways were grouped into five categories, which was calculated by LEfSe. Point size represented fold change. Mannose-related microbial pathways were coloured in blue. (D, E) Target bacteria were selected based on a significant correlation between their abundance and the β-(1,4)-mannan degradation pathway. In (D) bacteria that showed a significant correlation were highlighted in in red; the circle sizes in figure E represented the fold change. Mannan-dependent bacteria in figure E were coloured in blue. (F) The relative abundance of mannan-dependent bacteria was compared across different stages. P values were calculated using Student’s t-test. *p < 0.05, **p < 0.01. AA, arachidonic acid; AD, atopic dermatitis; EPA, eicosapentaenoic acid; ns, no significant difference.
To identify the target bacteria potentially involved in mannan degradation, we calculated Pearson correlation between the differential bacteria in the high AA concentration group and the β-(1,4)-mannan degradation pathway. The results indicated that four bacteria, Anaerolactibacter massiliensis, Mobiluncus porci, Olegusella massiliensis and Escherichia coli, were positively correlated with this pathway (figure 3D,E). Their abundance, particularly that of E. coli, was significantly higher in the AD group (figure 3F). Thus, we conclude that AA derived from breast milk plays a significant role in gut microbial imbalance during infancy.
Mannose metabolism promoted lipopolysaccharide biosynthesis and induced gut dysbiosisPrevious studies have reported that mannose promotes the biosynthesis of O-antigens and lipopolysaccharide (LPS).43 44 To determine whether gut dysbiosis in infants with AD was associated with mannose-related metabolism, we initially analysed differential genes in the microbiome to identify the active microbiota (figure 4A and online supplemental figure 7A). The results indicated that the active bacteria clustered into two major genera, Escherichia and Bifidobacterium (figure 4B, inside), as described previously.4 5 32 E. coli, a Gram-negative bacterium, was found to be more active than Bifidobacterium species in the gut microbiome of the AD group (figure 4B,C). As a critical biomarker from the first to the sixth month (online supplemental figure 3C), the abundance of live E. coli, reflected by genome coverage (figure 4D, upper), gradually increased in the AD group with high concentrations of AA (figure 4D, lower and online supplemental figure 7B). In contrast, the abundance of live Bifidobacterium species, including B. faecale (figure 4E, upper panel), significantly decreased in the AD group (figure 4E, lower panel).
Figure 4Differential expression of microbial genes reflected bacterial activity associated with the AD immune state. (A) Principal coordinate analysis based on the abundance of metabolite revealed the differential metabolites. (B) Estimation of bacterial activity. The bacterial activity was indicated by the number of differentially expressed genes detected by HUMAnN3. Bar plot showed the bacterial activity while circle plot showed the composition of microbial genera. Bacterial activity was quantified by dividing the number of genes in individual species by the total number of differentially expressed microbial genes. (C) The heatmap illustrated the abundance and distribution of differentially expressed microbial genes. Colour bar on the heatmap indicated the disease status while the bar on the left denoted the bacterial categories. (D, E) Average abundance of active bacteria. Upper, live bacteria was reflected by genome coverage; lower, average microbial abundance. Each points in the box plot represented individual sample. (F) Position correlation between the abundance of Escherichia coli and GDP-mannose-derived O-antigen biosynthesis pathway. (G) Prediction of MHC affinity based on differentially expressed bacterial genes revealed the disease status in AD patients. Predicted T cell activity was calculated by 1/MHC affinity. (H) The hypothesis suggested that the potential aetiology of AD onset during infancy might be associated with arachidonic acid derived from breast milk. *p < 0.05, ****p < 0.0001. AD, atopic dermatitis; KEGG, Kyoto Encyclopaedia of Genes and Genomes.
For E. coli, a representative Gram-negative bacterium, we inferred that the microbial pathways associated with its unique characteristics, such as O-antigen and LPS biosynthesis, might be closely linked to mannose-related metabolism. As predicted, pathways related to Gram-negative bacteria, including GDP-mannose biosynthesis, GDP-mannose-derived O-antigen biosynthesis and LPS biosynthesis, demonstrated a strong positive correlation (online supplemental figure 7C). Additionally, these pathways, along with β-(1,4)-mannan degradation, also showed a positive correlation with the abundance of E. coli (figure 4F and online supplemental figure 7D). To further confirm the potential relationship between these pathways and E. coli, we selected two pathways: GDP-mannose biosynthesis and GDP-mannose-derived O-antigen biosynthesis, from the HUMAnN3 data of all samples. We then calculated the number of bacteria that corresponding to each pathway. The results revealed that most of the Gram-negative bacteria, especially E. coli, were more active in the microbial pathways (online supplemental figure 8). These findings suggest that mannose metabolism facilitates the growth of Gram-negative bacteria by enhancing the biosynthesis of O-antigens and LPS, leading to gut microbial dysbiosis. This dysbiosis, in turn, could trigger skin inflammation associated with the activation of CD4+ T cells (figure 4G).
In summary, we hypothesised that AA derived from breast milk possibly stimulated mannose metabolism in several mannan-using bacteria via the β-(1,4)-mannan degradation pathway. Some Gram-negative bacteria, such as E. coli, would use GDP-mannose to synthesise O-antigen and LPS through the GDP-mannose-derived O-antigen and LPS biosynthesis pathway.45–48 This process may lead to the disturbance of some Gram-negative bacteria, ultimately resulting in gut dysbiosis. Furthermore, skin inflammation is exacerbated by the biosynthesis of PGs and LTs when the COX and LOX enzymes catalyse AA. The increased production of inflammatory mediators may contribute to the progression of AD (figure 4H).
AA aggravated skin inflammationTo adequately illustrate the role of AA in skin inflammation, we initially stimulated allergen-induced dermatitis in BALB/c mice (see online supplemental methods), which mimics AD. This was followed by the intragastric administration of PUFAs (figure 5A and see online supplemental methods). Similar to humans with AD, mice exhibited skin lesions as well as comparable thickness of the skin granule layer and infiltration of inflammatory cells in the adipose layer (figure 5B). However, AA, but not EPA, significantly increased the area of skin lesions, whereas there was no significant difference in the thickness of the skin cuticle or the infiltration of inflammatory cells in the adipose layer between mice administered AA and those administered EPA (figure 5C).
Figure 5Arachidonic acid exacerbated AD-like skin inflammation in mice. (A) Workflow chart for animal experiments. (B) Pathological detection of AD-like skin lesions. Blue arrow showed the position of incrassate skin cuticle while red arrow indicated the infiltration of inflammatory cells in the skin’s adipose layer. (C) Statistical data were collected on the area of skin lesions, the thickness of the skin cuticle and the number of infiltrated inflammatory cells in skin’s adipose layer. (D) The expressional level of inflammatory factors. (E) The secretory level of serum IgE. (F) The ratio of peripheral CD4+ and CD8+ cells. (G) The proportion of peripheral CD4+ and CD8+ cells. (H) The ratio of peripheral CD4+ and CD8+ cells. (I) The proportion of peripheral Th1, Th2, Th17 and Treg cells detected by flow cytometry. In this figure, the colours in box plots were showed: red, control group; orange, AD group without administration of PUFA; blue, AD group with AA treatment; darkgreen, AD group with EPA treatment. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. AA, arachidonic acid; AD, atopic dermatitis; EPA, eicosapentaenoic acid; ns, no significant difference; PUFA, polyunsaturated fatty acid.
Based on the AD-like skin inflammation observed (online supplemental figure 9 and online supplemental figure 10A), skin transcriptomic analysis revealed that AA induced the expression of several inflammation-related genes, including interleukin-6 (IL-6), intercellular adhesion molecular 1 (Icam1), vascular cell adhesion molecule 1 (Vcam1) and nuclear factor kappa-B (Nf-κB, figure 5D). Additionally, ELISA results indicated that AA could simultaneously promote IgE secretion (figure 5E).
Furthermore, previous studies have reported an imbalance in the peripheral Th1/Th2 cell ratio.49 50 To investigate the influence of AA on peripheral T cells, we performed flow cytometry on PBMCs from mice (online supplemental figure 11). The results revealed that CD4+ T cells were the predominant immunocytes in the inflammatory state (figure 5F,G), which is consistent with our bioinformatic predictions (figure 4G). Further analysis indicated that inflammation disrupted the CD4/CD8 cell ratio (figure 5H). Notably, EPA, but not AA, restored this balance (figure 5H). Additionally, AA was observed to increase the proportion of Th2 cells and enhance IL-4 expression (figure 5I and online supplemental figure 12A), a finding supported by cellular experiments (online supplemental figure 13). In contrast, EPA promoted the proliferation of regulatory T (Treg) cells and upregulation of IL-10 (figure 5I and online supplemental figure 12B). These findings suggest that AA exacerbates inflammation, whereas EPA might alleviate this process, potentially through the action of IL-10.
AA did not result in intestinal inflammation but induced gut dysbiosis associated with ADThe relationship between the AA-induced gut dysbiosis and intestinal inflammation remains unclear. The results of intestinal RNA-sequencing data from animal experiment using PCA revealed no significant differences in gene expression between the AD and control groups, even after the intervention with PUFA (online supplemental figure 9). Additionally, the expression of several pyroptosis-related genes, including caspases, NLRP, IL-1β and IL-18,51 did not differ significantly in intestinal tissues between the AD and control groups (online supplemental figure 10B). HE staining of intestinal tissues from the AD group showed no evidence of inflammatory infiltration compared with the control group (figure 6A). These findings suggested that AA does not induce intestinal inflammation.
Comments (0)