Skin Microbiota Diversity Distinguishes Cd And Uc Patients
To find out whether the skin microbiota differs between CD and UC patients and healthy controls (HC), we analyzed the bacterial composition of the skin at the retroauricular crease, the lumbar region, and the inguinal crease using amplicon 16S rRNA gene sequencing.
Separate models for each sampling site and diagnosis revealed that microbiota at the retroauricular crease of CD patients has significantly increased Shannon index (richness and evenness) compared to UC or HC, while the difference between UC and HC was not pronounced. The number of amplicon sequence variants (ASVs) and Chao1 index showed only slightly increased richness and evenness of the skin microbiota at the retroauricular crease in CD patients compared to UC patients and HC (Fig. 1A). Beta diversity of skin microbial communities at the reatroauricular crease, expressed as Unweighted UniFrac and Bray-Curtis distance between CD, UC, and HC showed that samples cluster by diagnosis. As a result, UC and HC were shown to be more similar to each other (Fig. 1B). We confirmed that these differences were not influenced by biologic treatment, sex, timescale of sampling, BMI, or the number of weeks after treatment initiation (data not shown).
Bacterial microbiota of the skin at retroauricular crease was dominated by the phyla Actinobacteria and Firmicutes, and, to a lesser extent, Proteobacteria. The most abundant bacterial genera in all samples were Cutibacterium, Staphylococcus, and Corynebacterium, followed by the species Staphylococcus hominis and Staphylococcus epidermidis. The relative abundance of each taxon differed between CD, UC, and HC, with CD having a relatively more diverse composition than UC or HC (Fig. 1C).
We observed slightly higher alpha diversity in CD compared to UC patients, and significantly higher diversity in CD over HC at the lumbar region, but not at the inguinal crease (Supplementary Figure S2A and S2B, respectively). We again confirmed that these differences were neither influenced by biologic treatment, sex, timescale of sampling, BMI, nor the number of weeks after treatment initiation (data not shown). Beta diversity of skin microbial communities at the lumbar region showed a significantly distinct microbiota profile of IBD patients when compared to healthy individuals. At the inguinal crease, microbial beta diversity of CD patients did not differ from HC but differed from that of UC patients. Beta diversity of UC patients at the inguinal crease was significantly different from that of HC (Supplementary Figure S2C).
Out of the three skin sites we sampled, the retroauricular crease showed the most pronounced differences in skin microbiota composition between CD, UC, and HC. For that reason, we analyzed this particular sampling site in more detail.
CD and UC have their own microbial pattern while retaining the correlation of the shared taxa
To explain patterns distinguishing CD, UC, and HC that we observed by looking at alpha and beta diversity of our samples, we performed Differential abundance analysis (DAA). We revealed that CD patients had significantly higher abundance of Corynebacterium (ASV86) and Pseudomonas (ASV1183), and significantly lower abundance of other species such as Actinomyces (ASV28), Cutibacterium (ASV212), Lawsonella (ASV87), Prevotella (ASV303) or Streptococcus (ASV607) over HC. UC patients showed increased abundance of Corynebacterium (ASV86) and Pseudomonas (ASV1183) over HC, otherwise they followed similar pattern as CD (Fig. 2A).
The DAA regression coefficients for CD vs. HC and UC vs. HC comparisons were closely correlated (Spearman’s correlation: rho = 0.7560, p < 0.0001), suggesting that both IBD conditions promote similar responses in the skin microbiota. At the same time, however, additional DAA comparison between the CD and HC groups showed that each for of IBD possess specific patterns of changes in ASVs abundances. For example, Corynebacterium (ASV86), Bacteroides (ASV278), or Methylobacterium methylorubrum (ASV909) were significantly increased in CD over UC, while Cutibacterium (ASV212), Delftia (ASV1039) or Psychrobacter (ASV1174) were significantly increased in UC over CD (Fig. 2B).
A specific skin microbiota profile might predispose IBD patients for skin side effects manifestation following anti-TNFα treatment
In this section focused on monitoring SkAE following anti-TNFα treatment, we used longitudinally collected samples from 25 IBD patients (17 CD and 8 UC). In this study cohort, SkAE at several skin sites affected a total of 13 CD and 4 UC patients. Different SkAE manifestations on different body sites included drug exanthema, eczema, papulopustular exanthema, herpes, and shingles. Comparing all CD and UC samples regardless of the sampled site and longitudinal nature of their collection, we observed that skin microbiota of CD patients with SkAE tended to have different proportional composition from that of CD patients without SkAE. Specifically, CD patients with SkAE had lower frequency of Cutibacterium (ASV212) and Staphylococcus (ASV636) and a higher frequency of Corynebacterium (ASV86), Micrococcus (ASV179), Enhydrobacter (ASV1167) or Anaerococcus (ASV757) when compared to CD patients without SkAE. There were no pronounced differences between UC patients with and without SkAE (Supplementary Figure S3A).
As evaluating skin microbiota on several body sites together might be misleading, we further focused on SkAE manifestation (and therefore microbiota comparison) at the retroauricular crease, as this site showed the most pronounced differences in Shannon entropy between CD, UC, and HC. Longitudinally collected samples are particularly useful here, as they could help predict shifts in the microbial composition associated with SkAE manifestation at this skin site. We obtained samples from 2 CD patients with drug exanthema, 2 CD patients with eczema, and 4 UC patients with papulopustular exanthema at the retroauricular crease (Fig. 3A). We did not find any apparent skin microbiota changes that would correspond to SkAE manifestation, IBD severity, or response to the biologic treatment (i.e., loss of response during the treatment), regardless of the metric used (Fig. 3B, 3C, and 3D).
On the other hand, comparing the baseline of SkAE patients (CD+/UC+) with the patients who did not go on to develop SkAE (CD−/UC−) indicated a potential to predict the development of SkAE based on specific microbiota composition. Percentual taxonomic differences of ASVs between CD+/UC+ and CD−/UC− patients and HC showed that in CD patients who later developed SkAE (CD+) the most abundant genus was Staphylococcus (ASV636) (29%), whereas in CD patients who did not develop SkAE (CD−), the most abundant genus was Cutibacterium (ASV212) (25%). Furthermore, in contrast to the CD− cohort, the CD+ cohort was characterized by low abundance of Staphylococcus hominis (ASV642) (2% CD+; 14% CD−), higher abundance of Staphylococcus epidermidis (ASV640) (10% CD+; 5% CD−), and presence of Dietzia (4% in CD+, below 1% in CD−). On the other hand, CD+ cohort showed only negligible presence of Anaerococcus (ASV757) (below 1% in CD+, 6% in CD−) and Finegoldia (ASV761) (below 1% in CD+, 2% in CD−) compared to CD− cohort. In UC patients who later developed SkAE (UC+), we observed higher abundance of Anaerococcus (ASV757) (4% in UC+, below 1% in UC−) and Lawsonella (ASV87) (2% in UC+, below 1% in UC−), and lower abundance of S. hominis (ASV642) (1% in UC+, 8% in UC−) compared to UC− patients who did not develop SkAE. Interestingly, genus Corynebacterium (ASV86) was highly abundant in UC− cohort (17%) and it was under the detection limit (set to 1%) in UC+ cohort. Healthy controls were shown to possess high levels of Cutibacterium (ASV212) (53%) and Staphylococcus (ASV636) (15%) and, to a lesser extent, also Corynebacterium (ASV86) (4%). However, Corynebacterium (ASV86) in HC was present to a much lesser extent than in UC− patients who did not develop SkAE (4% in HC vs. 17% in UC−) (Supplementary Figure S3B). MetamicrobiomeR-based differential abundance analysis after multiple testing corrections revealed significant association of Gemella (ASV201), Enhydrobacter (ASV370), Pseudoclavibacter (ASV52), and Kocuria (ASV55) with CD+ cohort (Supplementary Figure S3C). This analysis, however, showed no significant associations for UC cohort after multiple testing corrections.
Incidence of SkAE during anti-TNFα therapy is associated with changes in the serum levels of biomarkers of epithelial barrier function
To gain insight into the pathogenesis of skin side effects of anti-TNFα therapy, we examined different biomarkers related to skin and gut barrier function and to immune response. We analyzed 22 potential serum biomarkers at baseline, during the manifestation of SkAE, and after the healing of SkAE (Supplementary Figure S4). We found that a marker closely associated with epithelial integrity, L-FABP, was lowered during the manifestation of SkAE (Fig. 4A). Conversely, I-FABP increased significantly after SkAE were healed (Fig. 4A). Moreover, we showed that there were no differences in L-FABP, I-FABP, and E-FABP at baseline between the groups of patients with and without SkAE (Supplementary Figure S5). Further correlation analysis of the clinical data together with potential biomarkers of SkAE identified several features associated with the occurrence of SkAE or specific to the absence of SkAE (Fig. 4B, Supplementary Figure S6). Interestingly, in patients suffering SkAE we found a positive correlation of (i) I-FABP with TNFα levels (r = 0.906); (ii) BMI with the levels of TIMP-1 (r = 0.786), MMP-9 (r = 0.786), LBP (r = 0.928), and EG-VGF (r = 0.808); and (iii) osteoprotegerin with IBD severity (r = 0.866), calculated as Spearman correlation coefficient. Furthermore, SkAE affected patients showed a negative correlation of (i) fecal calprotectin with hemoglobin (r = -0.821), (ii) weight with osteoprotegerin (r = 0.786), and (iii) BMI with the levels of IGFII (r = -0.786), which was not observed at baseline or at the study endpoint (Fig. 4B). In contrast, we found the same positive correlation pattern between the levels of FC and IBD severity at baseline and at the study endpoint, but not during SkAE incidence (r = 0.954 and r = 0.845) (Fig. 4B). There was no association of SkAE to blood levels of anti-TNFα (data not shown).