Microbial predictors of healing and short-term effect of debridement on the microbiome of chronic wounds

Abstract

Chronic wounds represent a large and growing disease burden. Infection and biofilm formation are two of the leading impediments of wound healing, suggesting an important role for the microbiome of these wounds. Debridement is a common and effective treatment for chronic wounds. We analyzed the bacterial content of the wound surface from 20 outpatients with chronic wounds before and immediately after debridement, as well as healthy skin. Given the large variation observed among different wounds, we introduce a Bayesian statistical method that models patient-to-patient variability and identify several genera that were significantly enriched in wounds vs. healthy skin. We found no difference between the microbiome of the original wound surface and that exposed by a single episode of sharp debridement, suggesting that this debridement did not directly alter the wound microbiome. However, we found that aerobes and especially facultative anaerobes were significantly associated with wounds that did not heal within 6 months. The facultative anaerobic genus Enterobacter was significantly associated with lack of healing. The results suggest that an abundance of facultative anaerobes is a negative prognostic factor in the chronic wound microbiome, possibly due to the increased robustness of such communities to different metabolic environments.

Conflict of interest statement

The authors declare no competing interests.

Figures

Fig. 1. Taxonomic composition of skin and wound samples.

a Average relative abundance of genera within each sample type (only genera with average relative abundance >1% are shown). Staphylococcus taxa are labeled at the species level. Gradient color scale is for visualization purposes only. b Relative abundance of genera in each sample (bar graph limited to the 20 most abundant taxa overall; “NA” indicates OTUs without taxonomic classification; Staphylococcus taxa are labeled at the species level).

Fig. 2. Association of abundant OTUs with pre-debridement wound samples or skin samples, inferred by DESeq2 or BGLMM.

OTUs (with average relative abundance > 0.1%) found to be significant (criteria described in Methods) in at least one of the models with enrichment in wound samples (red) or enrichment in skin samples (blue). OTUs found to be not significantly enriched in that model are shown as gray. For DESeq2, the log2 fold-change in variance-stabilized abundance is shown with error bars indicating the estimated 95% confidence interval (1.96× standard error, n = 19). For BGLMM, the median of estimated βj1 (pre-debridement effect for OTU j, see Methods for details) with 95% credible interval error bars are reported (n = 19). The heatmap shows the log10 (relative abundance in wound minus relative abundance in skin) of each OTU of each patient for a visual comparison. OTUs are labeled by their genus name or lowest available taxonomy assignment if applicable; otherwise, the original OTU label from QIIME open OTU picking is used. Note that multiple OTUs may belong to the same genus.

Fig. 3. Comparison of pre- and post-debridement samples.

Pre- and post-debridement samples have similar numbers of exclusive OTUs (a); lower and upper bounds of the boxes correspond to the first and third quartiles, center lines indicate the median, and whiskers extend up to 1.5× interquartile range; any points beyond the whiskers are outliers. Shared OTUs account for a large majority of microbiota (b). c Analysis of statistically significant enrichment of individual taxa in pre- vs. post-debridement samples by DESeq2 and BGLMM; OTUs are sorted by descending average relative abundance. Note that Sphingopyxis was only found to be abundant in patient 15. d Coarse-grained differential abundance analysis of aerobes, anaerobes, and facultative anaerobes using DESeq2 shows no significant difference immediately after debridement. “Mixed” indicates taxa that were not annotated due to: low relative abundance (<0.1% on average), no taxonomic annotation, or ambiguous oxygen requirements. For all BGLMM and DESeq2 inferences, error bars indicate 95% confidence interval, or 1.96× standard error, respectively, and n = 19.

Fig. 4. Comparison of healed and nonhealing wounds.

Average relative abundance of taxa classified by oxygen requirements (anaerobic (a), aerobic (b), and facultative anaerobes (c)) suggests facultative anaerobes may be predictive of healing outcome. Plots were filtered to show taxa with >0.5% average relative abundance within each sample type and outcome. Cumulative relative abundance of aerobes, anaerobes, facultative anaerobes, and unassigned taxa in wound samples that did or did not heal (d). Healed wounds are ordered by estimated wound age when known; unhealed wounds are ordered by treatment time up to the point of medical record data collection. Differential abundance analysis of healing outcomes for taxa with different oxygen requirements using DESeq2 indicated substantial enrichment of facultative anaerobes in nonhealing wounds (e). Error bars indicate estimated 95% confidence interval (1.96× standard error, n = 19). Taxonomic associations (OTU with average relative abundance > 0.1%) identified by BGLMM or DESeq2 with healed or unhealed wounds, comparing pre-debridement or post-debridement samples from each outcome, indicates significant enrichment of Enterobacter in nonhealing wounds (f). Error bars indicate 95% confidence intervals for BGLMM inference (n = 19) and estimated 95% confidence interval (1.96× standard error, n = 19) for DESeq2.