The effects of challenging the healthy skin microbiota with a B. subtilis formulation were studied using the inner ear skin of laboratory mice. This experimental model highly resembles human skin sites in terms of morphology and microbiota and has been successfully used in the assessment of host-microbe interactions [33, 34]. Twenty-four 8-week old C57BL/6 female mice were randomly assigned to one of four groups: Pluronic hydrogel containing B. subtillis, B. subtillis (in LB medium), plain Pluronic hydrogel, and a no-treatment control group (Fig. 1A). Each group was administered with the corresponding formulation twice a day for seven days. The effects of the various treatments were analyzed by determining the bacterial composition of the skin before the first application (day 0) and on days 2, 4, 8, 11, and 14 (Fig. 1B).
After skin sampling, genomic DNA was extracted and the V3-V4 hypervariable regions of 16S rRNA gene were amplified and sequenced using the Illumina technology. A total of 29.5 million high-quality 16S rRNA gene sequences were obtained, each containing between 0.36% and 1.5% of the data. Noise was removed according to the Remove Unwanted Variation (RUV) strategy , using the untreated control group and day 0 samples (before treatment) for normalization (Supplementary Table 1).
The dynamics of the abundance of B. subtilis, in particular, and of the Bacillus genus, in general, before, during, and after administration of B. subtilis formulations was assessed (Fig. 2A and B) and compared with the average abundances of B. subtilis and Bacillus (dashed lines), respectively. It was found that plain Pluronic hydrogel did not impact the abundance of B. subtilis, whose counts remained similar to those recorded for the untreated control group (Fig. 2A). B. subtilis in Pluronic, conversely, had the highest influence on B. subtilis counts on days 2 and 4, which were considerably higher than for the B. subtilis treatment. Nevertheless, four days after the last administration, all groups presented counts that were similar to that of the untreated group. For the Bacillus genus (Fig. 2B), counts were significantly higher for all groups compared with the untreated control group: The group that received B. subtilis showed a 2-fold increase while the two groups that received Pluronic (with and without B. subtilis) exhibited a 6-fold increase. This trend changed on day 4, when all groups presented a two-fold increase compared with the untreated group. Post-challenge, on day 8, Bacillus counts for the two groups that received B. subtilis increased 4-fold, while counts for the group that received plain Pluronic decreased to the control group level. From day 11, i.e. 4 days after the last administration, control levels were attained for all treatment groups with insignificant differences compared with the untreated group.
Principal components analysis (PCoA) was then used to assess skin microbiota dynamics on both the genus (Fig. 3A) and the species (Fig. 3B and Supplementary Fig. 1) levels. Our data indicated a clear clustering according to treatment day and nature as explained by PC1 and PC2, respectively (i.e., data organization according to treatment nature along PC2 axis can be observed in Fig. 3B). Treatment nature and time point influenced the observed clustering to the same extent, as evidenced by the similarity in effect size between PC1 and PC2 (6.97%-6.96% and 5.57%-5.86%, respectively). The B. subtilis in Pluronic group exhibited an enhanced microbial shift compared with the pure B. subtilis group, which showed a very similar pattern but with a slight delay, prominent on day 8 (Fig. 3A). One week after ceasing treatment administration, however, all groups presented a microbiota composition similar to that of the untreated control group.
We further investigated the microbiota shift upon B. subtilis intervention and after its cessation by analyzing the intervention's effects on the relative abundance of the most represented skin bacterial genera (Fig. 4A and Supplementary Table 2). Consistent with the PCoA results, altered bacterial abundance was observed for all treatment groups along the experiment. For Corynebacterium, the most abundant genera in the inner ear skin microbiota, application of B. subtilis in Pluronic resulted in a sharp increase from day 2 to 4, followed by a plateau for the remainder of the experiment. Application of pure B. subtilis caused an abundance increase only on day 8, which remained until day 11 and then decreased to initial values. No significant variations in abundance were observed following the application of pure Pluronic. Interestingly, the relative abundance of Staphylococcus exhibited an inverse trend compared with Bacillus for both B. subtilis-containing formulations: when Bacillus abundance increased, Staphylococcus counts decreased, and vice versa. To obtain a broader view of these bacterial changes, we mapped the differential representation of bacterial genera along the treatment period (Fig. 4B). Only statistically significant differences in bacterial abundance (-1 < log2FC > 1; padj < 0.05, Wald-test) were considered for the analysis (Supplementary Table 3). Following the administration of B. subtilis in Pluronic, several highly related genera were observed to cluster, being either underrepresented or overrepresented compared with the untreated control group (Fig. 4B). For instance, the Lentibacillus, Gemella, Marinococcus, and Virgibacillus genera, of the Bacilliales order, were overrepresented on day 4. The Bacillus genus, on the other hand, was overrepresented on days 2 and 8, while no statistically significant difference to the control was presented on day 4, consistent with the trend observed in the PCoA (Fig. 2B). Staphylococcus abundance decreased during the application of B. subtilis in Pluronic formulation (days 2 and 4) and increased on day 11, three days after the last administration.