Skin microbiome in cutaneous T cell lymphoma by 16S and whole genome shotgun sequencing

The development of DNA sequencing methods have widened our conceptions of cutaneous microbial communities (Byrd et al., 2018, Eyerich et al., 2018, Fyhrqvist et al., 2016). Recently, we investigated the cutaneous microbiome in parapsoriasis, but did not find significant differences between healthy skin and diseased skin with the 16S method (Salava et al., 2017). In vitro studies have demonstrated associations between staphylococcal enterotoxins and cutaneous T cell lymphoma (CTCL) (Krejsgaard et al., 2014, Willerslev-Olsen et al., 2013, Blümel et al., 2019). We were thus interested in microbiome characteristics of CTCL and if they could add supplementary information to traditional diagnostics. Patients (n= 20) with mycosis fungoides stages IA-IIB at sampling were included. All patients had cutaneous CTCL lesions and 11 patients were on bexarotene maintenance therapy (Supplementary Table S1). Skin swab samples were obtained from active CTCL lesions and from contralateral healthy-looking skin non-invasively using sterile gloves. The extracted DNA from both lesional and non-lesional skin was analyzed by 16S and whole genome shotgun (WGS) sequencing. The sampling method, DNA extraction and sequence data analysis are outlined in more detail in the supplementary material (Kong et al., 2017, Lu et al., 2017).The 16S data (relative abundances) represented a total of 410 genera, 39 classes and 21 phyla. Four phyla: Actinobacteria (class Actinobacteria, 59.37 %), Firmicutes (predominantly the classes Clostridia and Bacilli, 15.02 %), Proteobacteria (mostly Alpha-, Gamma-, and Betaproteobacteria, 12.46 %) and Bacteroidetes (mainly Bacteroidia, 2.16 %) represented 89 % of all classified sequences and 10.19 % remained unclassified at the phylum level. The most common genera were Staphylococcus (30,0 %), Corynebacterium (22,3 %) Propionibacterium (5,0 %) and Streptococcus (3,0 %). We applied oligotyping of the potential Staphylococcus sequences and the three most common oligotypes were identified as S. epidermidis (48,1 %), S. aureus (33,8 %) and S. capitis (18,0 %) and the rest were unidentified species. DESeq2 (Love et al., 2014) was used to analyze the 16S sequencing data for differences between the sample types (lymphoma-affected vs. healthy-appearing skin) (Grogan et al., 2019). Comparisons based on a model corrected for the subject-specific variation but no significant differences in bacterial taxa or in the Staphylococcus oligotypes were observed (Lindahl et al., 2019).WGS data analyzed with BRACKEN (Lu et al., 2017) represented 1312 identified organisms at genus level, with Cutibacterium (member of the former genus Propionibacterium) (25 %), Corynebacterium (19 %) and Staphylococcus (18 %) as the most abundant genera. We observed no significant differences at the genus level or in microbial diversity (Figure 2). 16S sequencing does not capture propionibacteria very well when compared with WGS (Meisel et al., 2016), therefore explaining the differences in abundances (5% vs. 25%). On the species level most abundant were Cutibacterium acnes (24 %), Staphylococcus epidermidis (7 %), Staphylococcus aureus (7 %) and Corynebacterium kroppenstedtii (6 %) (Figure 1a). WGS data was then analyzed with DESeq2 to test for different abundances of bacterial taxa by comparing sample types (lymphoma vs. healthy skin). The microbiome varied substantially based on the investigated body site and individual patient (Oh et al., 2016). WGS data at species level showed that Staphylococcus argenteus was more abundant in lymphoma skin. However, after removal of reads identified by BRACKEN as additionally mapping to human genome (and previously not identified by BWA), S. argenteus was no longer identified as more abundant in lymphoma-affected skin (adjusted p-value 0.683). Interestingly, S. argenteus has been recently shown to express 4–6 fold higher levels of the alpha-hemolysin exotoxin compared to S. aureus (Johansson et al., 2019), and thus, the pathogenic role of the S. argenteus exotoxin may be more importance than the relative abundance of S. argenteus. This possibility needs to be addressed in future studies. The following 10 bacterial species were identified as more abundant in healthy-appearing skin samples (adjusted p-values < 0.05), regardless of additional reads mapping to human genome: Streptomyces sp. SM17, Bordetella pertussis, Streptomyces sp. PVA 94-07, Methylobacterium oryzae, Serratia sp. LS-1, Burkholderia mallei, Enterobacteriaceae bacterium, Achromobacter ruhlandii, Pseudomonas sp. A214, Pseudomonas sp. st29. The known associations of these species to human diseases or skin disorders are summarized in Supplementary Table S6 and outlined in the supplementary material. The role and significance of the relative reduction of these opportunistic pathogens in MF skin remains to be elucidated.