We conducted a cross-sectional study including adult patients with psoriasis seen in our department. Exclusion criteria were the use of antibiotics within the previous 3 months, consumption of probiotics or adherence to a specific diet within the last 6 months, as well as the presence of cirrhosis, neoplasms, inflammatory bowel disease, or other autoimmune diseases. Healthy controls met the same criteria.

After informed consent was obtained, clinical data were collected along with stool samples for gut microbiota analysis and blood samples for standard biochemical tests, complete blood counts, and interleukin measurements.

Serum enzyme-linked immunosorbent assays (ELISA) were performed for TNF-α, IL-17, IL-22, IL-23, IL-31, IL-33, IL-36, IFN-γ, and TGF-β (Invitrogen™ kits, Thermo Fisher Scientific, Waltham, MA, USA). Samples were analyzed in triplicate and read using a Sunrise microplate reader (Tecan, Männedorf, Switzerland). The lower limit of detection was 5pg/mL. Standard curves were generated for each plate, and the optical densities of the zero standard were subtracted from all other standards, controls, and samples to obtain corrected concentrations.

Genomic DNA was extracted from stool samples using the QIAamp PowerFecal DNA Kit (Qiagen, Hilden, Germany), and gut microbiota composition was analyzed by sequencing the hypervariable V3–V4 region of the 16S rRNA gene on an Illumina MiSeq platform.

A mean 141,200 paired Illumina reads per sample was obtained. Quality control was performed using fastqc8 and MultiQC.9 The DDAA2 pipeline (v1.16.0) was then used to generate the amplicon sequence variant (ASV) table from raw reads. Duplicates, noise, and chimeras were removed using default settings. Taxonomic annotation of ASVs was performed using the SILVA database (v138.1).10 A rarefaction curve was calculated using the R package vegan and inspected to assess taxon-count saturation. A phylogenetic tree of ASVs was constructed using Phangorn, and this tree was used to compute UniFrac and weighted UniFrac distances.11 The ASV count matrix was rarefied to the minimum read count per sample.

Alpha-diversity measures capture taxon diversity within each sample. Using the phyloseq package (v1.44.0),12 observed richness and Chao1 were calculated, as well as Shannon, inverse Simpson, and Fisher's alpha diversity indices. To test statistically significant differences across samples from controls and patients with psoriasis, the Wilcoxon rank-sum univariate test was first applied. To control for potential confounders, linear regression models including psoriasis and one or more covariates (body mass index [BMI], sex, age, smoking, alcohol intake, and number of reads) were used.

Beta diversity refers to differences in community composition between samples. The phyloseq package (v1.44.0)12 was used to compute weighted UniFrac distances11 and to ordinate data points in a lower-dimensional space using principal coordinates analysis (PCoA). To evaluate shifts in the overall distribution of ASV abundance, PERMANOVA was applied using the adonis2 function of the vegan package, based on weighted UniFrac distances. Separate tests were run for each variable (psoriasis, BMI, sex, age, smoking, alcohol intake, and number of reads), followed by an additional test combining confounders identified as significant in individual models.

To display taxonomic composition at the phylum level, the plot_bar function of phyloseq was used. ASV counts belonging to the same genus were then aggregated, and the 15 genera with the highest mean abundance were plotted using the fantaxtic R package.

To test for taxa with differential abundance between conditions, ASVs with <1 count in at least 3 samples (prevalence <6.25%) were removed, and DESeq2 with the Wald test was applied. Furthermore, DESeq2 models that included psoriasis status plus covariates (sex, BMI, age, smoking, or number of reads) were also fitted. The likelihood ratio test (LRT) was used to compare these models with models that included only confounding variables. Due to limited statistical power, only a few ASVs had a p-value <0.05 and none had an adjusted p-value <0.05.

A total of 36 patients with psoriasis and 24 controls were included (Table 1). The mean age in the psoriasis group was 46.6 years (SD, 11.7), higher than in the control group (40.7 years, SD, 17.0). Sex distribution was similar across groups (52.8% men in the psoriasis group vs 45.8% in the control group). The mean BMI was comparable between groups (26.6±5.5 and 26.3±5.4kg/m2, mean±SD for psoriasis and control groups, respectively). The prevalence of smokers was significantly higher in the psoriasis group (55.6% vs 20.8% in controls). In addition, the incidence rate of metabolic syndrome was higher in patients with psoriasis (30.6% vs 12.5% in controls). The median Psoriasis Area and Severity Index (PASI) for the psoriasis group was 3.6 (IQR, 4.35). Similarly, 27.8% of the psoriasis group exhibited psoriatic arthritis (PsA), and 94.4% reported prior systemic treatment, 47.2% oral and 47.2% biologic.

Observed richness and Chao1 were computed, as well as Shannon, inverse Simpson, and Fisher's alpha diversity indices. No significant differences were found for any diversity index between groups using the Wilcoxon test (Fig. 1A). No significant differences were found when using linear regression models including confounding variables (sex, smoking, alcohol intake, BMI, and age). Among these variables, only BMI showed a marginal yet significant association with observed richness and Chao1 (R2=0.07, p=0.04) (Supplementary Fig. 1). This suggests that microbial diversity within individuals was similar regardless of whether they had psoriasis. No differences in alpha diversity were found between patients with and without PsA (Fig. 1B).

Fig. 1.

Alpha diversity results based on diversity indices using the Wilcoxon test. (A) Alpha diversity according to psoriasis vs healthy controls. (B) Alpha diversity according to the presence of psoriatic arthritis.

The PCoA of weighted UniFrac distances showed that samples did not cluster according to psoriasis diagnosis (Fig. 2A), nor, among patients, according to the presence of PsA (Fig. 2B). Consistently, the PERMANOVA was not significant, and cutaneous psoriasis explained 1.9% of the variation in microbiota composition (Fig. 2A). Sex and smoking also did not explain a significant portion of variation in microbiota composition (1.4% and 2.5%, respectively; p>0.05) (Supplementary Figs. 2A and B). BMI, on the other hand, explained 4% of the variation in microbiota composition (p<0.05). A BMI gradient can be observed along the 2nd axis of the PCoA, with high-BMI individuals clustering at the bottom of the plot and low-BMI individuals at the top (Supplementary Fig. 2C). Age explained 4.9% of the overall variation in composition (p<0.05) (Supplementary Fig. 2D).

Fig. 2.

Beta diversity based on principal coordinates analysis (PCoA). (A) Beta diversity according to the presence (red) or absence (blue) of psoriasis. No clustering according to psoriasis diagnosis is observed. (B) Beta diversity according to the presence of psoriatic arthritis. No segregation of cases is observed according to this variable.

Within patients with psoriasis, the relationship between alpha-diversity indices and BMI did not persist, likely due to sample size limitations, but the relationships with beta diversity and BMI and age did persist, explaining 5% and 4.8% of variation, respectively (PERMANOVA, p<0.05).

Furthermore, the PASI score showed a significant relationship with beta diversity, explaining 5% of the variation within the psoriasis group, although it was not associated with any alpha-diversity index. The presence of a family history did not show significant relationships with either alpha or beta diversity (Supplementary Fig. 3). Finally, prior treatments, whether classic systemic or biologic, did not appear to be associated with either alpha or beta diversity (Supplementary Fig. 4).

Analysis of absolute abundances revealed differences in the phylum Synergistota, which was less represented in patients with psoriasis (p<0.05) (Fig. 3A), although the total abundance of this phylum was low in both groups. A summary of species found in both groups is shown in Fig. 3B.

Fig. 3.

Total abundance by phylum, relative abundance by species, and volcano plot for differential abundance analysis. (A) Total abundance by phylum. This panel shows total phylum-level abundance in both groups: psoriasis and control. The phylum Synergistota appears less abundant in psoriasis, although overall abundance is low in both groups. (B) Relative abundance by species. This panel displays the relative abundance of the different species in both groups: control (top) and psoriasis (bottom). (C) Volcano plot for differential abundance analysis. This plot shows the results of the differential abundance analysis. ASVs with no differences across groups are shown in green. ASVs with differential abundance in psoriasis are shown in red – lower abundance on the left and higher abundance on the right.

The differential abundance analysis (DAA) identified increased levels of Subdoligranulum sp. (ASV188), Lactobacillus sp. (ASV207), and Bacteroides plebeius (ASV45), and decreased levels of Senegalimassilia anaerobia (ASV225) and Ruminococcus sp. (ASV250) in patients with psoriasis vs controls. After adjusting for confounding factors including BMI, age, sex, and smoking, only Subdoligranulum sp. remained significant (Fig. 3C), indicating a strong positive association between the abundance of this genus and psoriasis.

Additionally, a DAA was conducted considering only patients with psoriasis. No differences in any ASV were identified according to PsA status, family history, PASI, or prior systemic/biologic treatments, likely due to sample size limitations.

Measurement of inflammatory interleukins showed significantly elevated levels of all measured cytokines in patients with psoriasis (Fig. 4A) (p<0.05). When assessing the relationship between ASVs and inflammatory interleukins, only a few differentially abundant ASVs showed group-specific correlations with certain cytokines (Fig. 4B).

Fig. 4.

Interleukin levels in peripheral blood samples and heatmap of correlations between species abundance and interleukin levels. (A) Interleukin levels in peripheral blood samples. This panel shows mean interleukin levels with standard deviation for both groups: control (blue) and psoriasis (red). All interleukins show statistically significant differences. (B) Heatmap of the correlation between species abundance and interleukin levels. This heatmap represents the correlation between ASV abundance and interleukin levels. The top bar represents the p-value for each correlation (gray for p<0.05 and black for adjusted p<0.05). The 2nd bar indicates whether ASVs are more (red) or less abundant (blue) in psoriasis. The top rectangle includes all samples; the one in the middle represents the control group; the bottom one represents the psoriasis group. Colors in each square reflect whether interleukin levels are higher (red) or lower (blue). Red squares for more abundant ASVs represent positive correlations; blue squares represent negative correlations. For less abundant ASVs, blue squares indicate positive correlations and red squares indicate negative correlations.

Focusing on the 5 ASVs with significantly different abundances between groups (Subdoligranulum sp., Lactobacillus sp., B. plebeius, S. anaerobia, and Ruminococcus sp.), ASVs with greater abundance in psoriasis showed positive correlations with cytokines, whereas those with lower abundance showed negative correlations (Table 2). Specifically, a significant positive correlation was found between Subdoligranulum sp. and all cytokines (TNF-α, IL-17, IL-22, IL-23, IL-31, IL-33, IL-36, IFN-γ, and TGF-β) (p<0.05), as well as between Lactobacillus sp. and IL-17, IL-23, IL-36, TNF-α, and TGF-β (p<0.05). Moreover, a significant correlation was found between B. plebeius and IL-22 (p<0.05).

Among ASVs with reduced abundance, a significant negative correlation was found between Ruminococcus sp. and IL-22 and TGF-β (p<0.05), and between S. anaerobia and elevated IL-17 levels in healthy controls only (p<0.05).