This analysis employed the largest-scale gut microbiome genome-wide meta-analysis to date obtained from the international consortium MibioGen (15). The project studied the genome-wide genotypes and 16S ribosomal RNA gene sequencing of 18,340 participants in 24 cohorts from 11 countries, targeting variable regions V4, V3–V4, and V1–V2 of the 16S rRNA gene to delineate the microbial composition and to conduct classification using direct classification. Microbiome trait loci (mbTL) mapping was performed to identify genetic loci that affect the relative abundance in gut microflora. A total of 131 genera, 35 families, 20 orders, 16 classes, and 9 phyla with an average abundance of over 0.1% were included.

The GWAS summary data of Eczema came from the data released by the IEU Open GWAS project (GWAS ID: ieu-a-996, Trait name: Eczema) (16). The GWAS data included 11,059,641 single-nucleotide polymorphisms (SNPs) and 40,835 samples, including 10,788 in the case group and 30,047 in the control group. The GWAS summary data of Psoriasis (Number of SNPs = 16,380,464; ncase = 4,510; ncontrol = 212,242), Rosacea (Number of SNPs = 16,380,452; ncase = 1,195; ncontrol = 211,139), AD (Number of SNPs = 16,380,443; ncase = 7,021; ncontrol = 198,740), Vitiligo (Number of SNPs = 16,380,442; ncase = 131; ncontrol = 207,482), and Acne (Number of SNPs = 16,380,454; ncase = 1,299; ncontrol = 211,139) came from the data released by FinnGen consortium (GWAS ID: finn-b-L12_PSORIASIS, finn-b-L12_ROSACEA, finn-b-L12_ATOPIC, finn-b-L12_VITILIGO, and finn-b-L12_ACNE, Trait name: Psoriasis, Rosacea, Atopic dermatitis, Vitiligo, and Acne) (17). Age, sex, top 10 major components, and genotyping batches were corrected during the original author’s analysis (18). The details of the data sources used in this study are shown in Table S1 ( Supplementary 1 ).

The bacterial taxa were classified into five hierarchical levels (phylum, class, order, family, and genus) to analyze, and each taxon was considered as a feature. SNPs associated with the gut microbiome were identified and used as instrumental variables. To ensure the authenticity and accuracy of the conclusions on the causal link between gut microbiome and inflammatory dermatoses risk, we selected instrumental variables (IVs) based on the following four criteria. First, the SNP-phenotype association level must reach the locus-wide significance threshold (P< 5 * 10^-8) to select potential IVs. Unfortunately, only a small number of SNPs were selected as IVs. To explore more relationships between inflammatory dermatoses and gut microbiota to obtain more comprehensive results, the second threshold (P< 1 * 10^-5) was used to identify SNPs which were selected as the second IV set. Second, SNPs with minor allele frequency (MAF) ≤ 0.01 were removed. Third, SNPs with R² values<0.001 (clumping window size=10,000 kb) were filtered, and only the one with the lowest p-value was kept. Fourth, when palindromic SNPs existed, the allele frequency data were used to infer the forward-strand alleles.

The principle design of the whole study is shown in Figure 1 . The detailed research flow chart and the three assumptions of Mendelian randomization analysis are presented in Figure S1 of Supplementary 2 . In this study, the following popular MR methods were used to examine whether there was a causal association between gut microbiota and inflammatory dermatoses: Inverse Variance Weighted (IVW) test, Maximum Likelihood (ML), Weighted Mode, MR-Egger regression, Weighted median estimator (WME), and MR Pleiotropy RESidual Sum and Outlier (MR-PRESSO). IVW test obtains the overall estimation of the influence of the gut microbiota on inflammatory dermatoses, by combining the Wald estimates for each SNP with a meta-analytic approach. It ignores the intercept term in regression and uses the inverse of outcome variance (se²) as the weight for fitting. If there is no horizontal heterogeneity, the IVW results would be unbiased (19). The ML method is analogous to IVW, and its standard error is smaller when the same assumptions are met (20). Based on the assumption of instrument strength independent of direct effect (InSIDE), MR-Egger regression is used to evaluate the existence of horizontal pleiotropy with intercept terms (21). If the intercept term is zero, it indicates no horizontal pleiotropy, and the MR-Egger regression result agrees with IVW (21). The WME method can estimate causality more correctly when more than 50% of the instrumental variables are invalid (22). When the InSIDE assumption is not fulfilled, the Weighted Mode has been shown to have a superior ability to detect a causal effect, with less bias and a lower type I error rate than MR-Egger regression (22). We also identified and corrected pronounced outliers with MR-PRESSO tests and MR-Egger regression (23).

Furthermore, we quantified the heterogeneity among the selected SNPs with Cochran’s Q statistic and identified potential heterogeneous SNPs with the “leave-one-out” analysis omitting each instrumental SNP in turn. Lastly, a reverse MR analysis of gut microbiota and inflammatory dermatoses was conducted with the same procedures and parameters as the forward MR.

In addition, we calculated the F statistic performed to assess the strength of the relevance between instrumental variables and exposure by the formula F = R 2 × ( n − k − 1 ) ( 1 − R 2 ) × k . R² is the proportion of the variance of the trait accounted for by the SNP, k is the number of IVs, and n is the sample size (24). An F value over 10 indicates no significant weak instrumental bias. We utilized an online calculator tool available on https://shiny.cnsgenomics.com/mRnd/, developed by Marie-Jo A Brion et al., to determine the power of MR estimates (25, 26).

Furthermore, we utilized the PhenoScanner software to extract genes encompassing all identified SNPs and subsequently conducted a pathway enrichment analysis on those genes linked to SNPs contained within the instrumental variables of gut microbes demonstrating a significant causal association with inflammatory skin disorders (27). The enrichment analysis was executed using the clusterProfiler package (28). We omitted KEGG pathways with a P-value exceeding 0.05 and those with fewer than three mRNAs represented in the enrichment path.

False discovery rate (FDR) correction was conducted by applied fdrtools procedure, with a false discovery rate of FDR< 0.1 (29). All analysis was performed on R (version 4.2.2) and MR Analysis was based on MR-PRESSO (version 1.0) (23), TwoSampleMR (version 0.5.6) (30), and meta (version 6.2.1) packages (31).

According to the selection criteria of IVs and removing the number of SNP repeated in different microorganisms, we identified 27 SNPs associated with gut microbiota at a significance level of p< 5 × 10⁻⁸. The details of 27 SNPs in the exposure and six outcome variables are shown in Table S2 ( Supplementary 1 ). Such few IVs is not enough for high-performance MR analysis. Therefore, under the screening of another threshold p< 1 × 10⁻⁵, we got 2,123 SNPs associated with gut microbiota in psoriasis, AD, rosacea, vitiligo, and acne, and 2,155 SNPs for eczema. The details of 2,155 SNP in the exposure and six outcome variables are shown in Table S3 ( Supplementary 1 ).

In our study, we employed gut microbiota as an exposure and investigated its association with six inflammatory dermatoses using MR analysis. We filtered the IVs based on a threshold of p< 5 × 10⁻⁸. As indicated in Table S4 ( Supplementary 1 ), the limited number of available IVs restricted us to only two MR evaluation methods for analysis. However, the results showed that there was no causal relationship between the other five dermatoses and gut microbiota, except for acne, which had a causal relationship with a few gut microbiota. It is worth noting that MR analysis is not recommended when there are less than 3 IVs, as per the authoritative statement on MR analysis. Therefore, we opted to use IVs filtered by a threshold of p< 1 × 10⁻⁵ for subsequent MR analysis.

All identified gut microbiota with no less than 3 IVs were retained. The causal effects of the remaining 192 gut microbiota on psoriasis, rosacea, atopic dermatitis, vitiligo, and acne are shown in Table S5 ( Supplementary 1 ).

Based on the estimate of IVW and ML, the result showed five, 18, 20, 14, 16, and 12 gut microbiotas had significant causal effects on psoriasis ( Figure 2A ), rosacea ( Figure 2B ), AD ( Figure 2C ), vitiligo ( Figure 2D ), acne ( Figure 2E ), and eczema ( Figure 2F ), respectively. It is worth noting that the common probiotics such as family Bifidobacteriaceae (OR = 0.82, 95%CI: 0.68–0.99, Pvalue = 0.0230, FDR = 0.0499) and genus Bifidobacterium (OR = 0.84, 95%CI: 0.73–0.98, Pvalue = 0.0197, FDR = 0.0289) have a protective effect on AD ( Figure 2C ). Order Lactobacillales (OR = 0.25, 95%CI: 0.06–0.97, Pvalue = 0.0447, FDR = 0.0757) had a protective effect on vitiligo ( Figure 2D ). In addition, family Bifidobacteriaceae (OR = 0.62, 95%CI: 0.43–0.90, Pvalue = 0.0118, FDR = 0.0319), family Lactobacillaceae (OR = 0.70, 95%CI: 0.51–0.96, Pvalue = 0.0159, FDR = 0.0260), and genus Lactobacillus (OR = 0.68, 95%CI: 0.51–0.91, Pvalue = 0.0068, FDR = 0.0103) have a protective effect on acne ( Figure 2E ).

Figure S2 ( Supplementary 2 ) presents a consolidated network highlighting gut microbiota implicated in several inflammatory skin conditions concurrently.

As shown in Table S5 ( Supplementary 1 ), the explaining rate of the total variation (R² values) of the 192 gut microbiota in six inflammatory dermatoses ranged from 0.40% to 10.58%, and the F values ranged from 12.27 to 139.83, excluding the possibility of weak genetic tool variables. Based on Cochran’s IVW Q test, there was no significant heterogeneity among these IVs ( Table S6 in Supplementary 1 ). In addition, according to the results of MR-Egger regression intercept analysis ( Table S7 in Supplementary 1 ), except for phylum Verrucomicrobia (p = 0.029) on rosacea, the other 205 intestinal microorganisms had no significant horizontal pleiotropy. However, further MR-PRESSO analysis ( Table S8 in Supplementary 1 ) did not find the horizontal pleiotropy in phylum Verrucomicrobia (Global test P-value = 0.356) on rosacea.

Based on the selection criteria of IVs, we obtained 38, 14, 58, 7, 12, and 42 SNPs (P< 1 * 10^-5, R²< 0.001) significantly associated with psoriasis, rosacea, AD, vitiligo, acne, and eczema, respectively. Summaries and details of each SNP are presented in Table S9 ( Supplementary 1 ). The results of reverse MR analysis ( Table S10 in Supplementary 1 ) showed that AD was causally associated with genus Eubacterium brachy group (OR = 1.41, 95%CI: 1.07–1.87, Pvalue = 0.0185, FDR = 0.0442, MR Egger). Vitiligo was causally associated with genus Ruminococcaceae UCG004 (OR = 0.97, 95%CI: 0.95–0.99, Pvalue = 0.0197, FDR = 0.0279, IVW), which indicated a bidirectional causal effect between them. No other significant causal relationship was found between six inflammatory dermatoses and the gut microbiota ( Table S10 in Supplementary 1 ). No notable heterogeneity was detected by Cochran’s Q statistics ( Table S11 in Supplementary 1 , P > 0.05). The results of MR-Egger regression intercept analysis indicated a significant horizontal pleiotropy when we evaluated the cause-effect of AD on genus Eubacterium brachy group (p = 0.0217) ( Table S12 in Supplementary 1 ). But, further MR-PRESSO global test ( Table S13 in Supplementary 1 ) suggested no evidence of pleiotropy (Global test P = 0.566).

By investigating the genes aligned with the instrumental variables of gut microbes showing a marked causal association with inflammatory skin disorders (refer to Table S14 in Supplementary 1 ) and conducting KEGG pathway enrichment analysis, we discerned a notable enrichment in inflammatory signaling pathways, including IL-17 signaling pathway, Chemokine signaling pathway and Cytokine-cytokine receptor interaction (refer to Table S15 in Supplementary 1 , Figure S3 in Supplementary 2 ). This underlines the potential mechanisms by which gut microbes might influence the progression of inflammatory skin diseases via pathways like interleukins and cytokines.