1 INTRODUCTION

Venous leg ulcers are chronic wounds affecting approximately 1% of Americans, and are more prevalent in the elderly population (aged 65 years and older).1 They are difficult to treat and can cause significant morbidity in patients who are not able to modify the causative factors for the lesions or who are not surgical candidates. VLUs are recurrent in 72% of patients diagnosed and have the worst healing prognosis of all leg ulcers.2 Typically, VLU wounds remain open for more than a year highlighting the need for prolonged management and more effective treatment strategies.3 VLUs are a late response of venous hypertension in the lower extremity. This hypertension can be caused by inadequate muscle pump function, incompetent valves or venous thrombosis. When venous pressure is elevated, it is difficult to achieve arterial perfusion and thus oxygen delivery, which can both cause the ulcer and retard wound healing. Due to the resultant poor arterial circulation in the area of these ulcers, many of the cellular and proteinaceous elements needed for wound healing are not able to access the wound bed.4 Interestingly, even when venous hypertension is mitigated, either surgically or by modifications in comorbidities, VLUs may persist or recur. This suggests that more research is needed to fully understand the pathophysiology of these ulcers.

Recently, the interplay between the microbiome and the wound environment has been shown to play a role in both chronic wound development5, 6 and wound healing.7, 8 The precise impact of the microbiome on wounds can be tied to microbial diversity of the host skin, where microbial composition varies depending on moisture content of the skin as well as dermal versus epidermal colonization.9-11 Highly represented epidermal skin bacterial phyla include Actinobacteria, Firmicutes, Proteobacteria, and Bacteroidetes, whereas subepidermal compartments typically contain higher proportions of Proteobacteria and Actinobacteria,9-11 Interestingly, unlike the gut and other microbiome systems in which increased diversity of host commensal organisms correlates with improved outcomes in dysbiotic conditions, the role of bacterial diversity in chronic wounds is less straight-forward. Studies examining diabetic ulcers, VLUs, and other chronic wounds show increases or decreases in the diversity of skin bacteria in wounds accompanying positive healing outcomes.12-14 In fact, treatments that disrupt the wound microbiome and cause either increases or decreases in diversity may result in faster healing.7, 8 The variability of the wound microbiome suggests that successful treatment is in part dependent on the individual's microbial ecology and treatment regimen.8

Chronic wounds are colonized by a diverse array of bacterial genera, and one mechanism that is hypothesized to contribute to chronicity is the presence of bacterial biofilms.15 In ulcer environments, commensal skin bacteria may contribute to the formation of biofilms leading to persistent infection, particularly when host immune function is impaired.16 One study found that the most abundant genera in diabetic foot ulcers included Staphylococcus and Corynebacterium.17 Also, anaerobes are found in high abundance due to low oxygen presence in the wound yet are not generally detectable by traditional clinical culturing methods.18 If a biofilm is detectable in VLUs, usually by biopsy, the presence of Staphylococcus aureus and Pseudomonas accounts for a large percentage of the bacteria in the biofilm, and these bacteria are believed to initiate biofilm formation.19 Biofilms in wounds are mostly polymicrobial; however, monomicrobial biofilms are seen as well in a much smaller percentage. Finally, wound depth can determine the presence or absence of particular bacteria.20 Various wound types sampled in different anatomical locations across patients showed a predominance of Staphylococcus residing in superficial wounds, while Pseudomonas persisted in deeper wound layers.6

Treatment of chronic VLUs consists of topical ointments, creams, foams, hydrogels, and other dressing products developed to promote a positive wound healing environment. Chlorhexidine-based antiseptic regimens of the skin have been shown to promote a decrease in bacterial burden and infection in surgical wounds.21 The effects of antiseptics on resident skin epithelial inhabitants have been characterized as short-term with an increase in predominant genera and loss of less abundant genera; effects are highly personalized to skin site following antiseptic treatment.22 Treatment of VLUs follows a regimen of topical antiseptic to the site of the wound followed by application of a secondary dressing that allows for compression. These dressings are changed at intervals of 7–10 days. At dressing change, the wounds are assessed, debrided, and then cleansed or washed with an antimicrobial solution before new dressings are applied. Systemic antimicrobial treatments may be used in certain circumstances but have little efficacy for promoting chronic wound healing due to antibiotic-resistant bacteria present in large numbers in biofilms.3

Although modulation of the wound microbiome is becoming a promising modality for therapy in wound treatment, a comprehensive understanding of microbiome dynamics before and after therapy in VLUs is currently lacking. Therefore, in this study, we undertook a survey of 11 patient microbiomes by performing 16S amplicon sequencing and in vitro culture using a selection of bacteria-specific media on skin swabs isolated from VLUs before and after wound washing. Bacterial composition of wound samples before and after wash were quantified and compared to determine effects of washing on bacterial communities. Results from this study may have implications for the success of therapeutic strategies in treating this disease.

2 METHODS 2.1 Sample collection

Eleven patients with persistent VLUs established for >1 year were included in this study. All patients received the same primary dressing and compression as well as dressing change interval. Wound cleansing was performed with chlorhexidine gluconate and hypochlorous acid solutions for all patients. Information regarding wound debridement was not collected as part of this study and may or may not have been performed as part of standard of care treatment for this patient cohort. Swabs of the wound were collected immediately before and immediately after washing. Swabs were obtained using the Levine technique23 by rotating the swab 360 degrees in a 1 cm square area for 5 s using gentle pressure to release tissue exudate, and samples were placed in cryovials containing 1.5 ml of tryptic soy broth or placed in a cryovial and covered in AllProtect Tissue Preservation Reagent (Qiagen). Tips were broken off into the tubes. The swab tips in broth were processed within approximately 30 min of collection in the laboratory for plating and assessment of bacterial growth on selective media. The swabs in AllProtect Reagent to be used for molecular work were stored at −80°C. Wounds were washed and then swabs were collected immediately following wash using the same procedure as pre-wash.

2.2 Wound characterization and treatment

All wounds were located overlying the medial malleolus. Wounds were treated with a primary dressing that contained ACTICOAT (Smith Nephew) followed by Drawtex (Urgo). A secondary dressing for compression was the 3M™ Coban™ 2 Layer Compression System. This dressing regimen remained in place for a week.

2.3 Bacterial growth on selective growth plates

Swabs from before-wash and after-wash were thoroughly vortexed in media. The tip of the swab was then removed from the media. Media was serially diluted, and plated on tryptic soy agar, MacConkey agar, Streptococcus selective agar, and mannitol salt agar growth selective plates and allowed to grow 24 h at 37°C. Samples were plated in duplicate. Bacterial growth was quantified as present or absent for growth following 24 h.

2.4 DNA extraction and 16S rRNA sequencing

Genomic DNA was extracted from 22 swab samples using the DNeasy PowerWater Kit (Qiagen) and normalized on an Eppendorf epMotion 5075 Liquid Handling Workstation to 0.2 ng/ml. Primers were designed to target the 16S V3 and V4 regions (forward primer: 5′-TCGTCGGCAGCGTCAGATGTGTATAAGAGACAGCCTACGGGNGGCWGCAG-3′, reverse primer: 5'-GTCTCGTGGGCTCGGAGATGTGTATAAGAGACAGGACTACHVGGGTATCTAATCC-3′). The targeted regions were amplified via PCR, purified, attached to Illumina sequencing adapters using the Nextera XT Index Kit Set A, and purified again before library pooling. After quality control, final libraries were sequenced on the Illumina MiSeq using v3 reagents with a 600 bp kit (paired 300-bp reads) in a single 65-h run.

2.5 16S amplicon data processing and bacterial identification

Reads underwent quality filtering using Trimmomatic (version 0.35), where adapter sequences were eliminated, and reads were cut at points of low quality using a sliding window of 4 and a minimum PHRED score of 20. After read quality control, paired-end reads were joined using QIIME's join_paired_ends.py script with default settings. Unjoined reads were discarded and assembled reads were assigned to samples from barcodes using QIIME's split_libraries.py.24 OTUs were identified by open reference OTU picking using the GreenGenes25 13_5 97% database and QIIME's pick_open_reference_otus.py script. For downstream analysis, OTUs present >1% in at least one sample were included.

2.6 Diversity analysis

All diversity analyses utilized the percent abundance table where OTUs present >1% in at least one sample were included. The percent abundance table was analysed using the Bray–Curtis dissimilarity metric as a measure of β-diversity between before- and after-wash samples within a patient, before-wash samples across patients, and after-wash samples across patients. Cluster analysis of before-wash samples was performed using hierarchical clustering of the Bray–Curtis dissimilarity metric. The optimal cluster number of 4 was determined by calculating the consistency of clusters with the silhouette technique, a quantification of how similar an object is to its assigned cluster compared to neighbouring clusters. The maximum average silhouette is considered the appropriate number of clusters. Clustering arrangements 1–10 were tested using the factoextra R package (Figure S1B). Statistical significance of hierarchical clusters was tested using pvclust boostrap analysis of clustering with threshold of au >90% significance.26 To quantify α-diversity, Shannon diversity (H) was calculated for all samples using H = –∑pi∙ln(pi), where pi represents the normalized population fraction of species i. The number of unique OTUs represented in a sample was also used as a metric of α-diversity. To test whether α-diversity changed between before-wash and after-wash samples across all patients, a paired Student's t-test was used on Shannon diversity quantifications and total unique OTUs on before and after samples.

2.7 Data availability

Raw sequencing data are available from the Short Read Archive (SRA) with accession number PRJNA704944.

3 RESULTS 3.1 Characterization of patient venous stasis ulcer microbiome

We investigated the wound bacterial microbiome ecology of 11 patients with established VLUs by collecting skin swabs from wounds following removal of dressings and compression bandages. Swabs were collected prior to and after washing with either a hypochlorous acid- or a chlorohexidine-based soap solution. We performed 16S amplicon sequencing on these before-wash and after-wash swab samples, obtaining libraries with an average sequencing depth of approximately 46,000 reads per sample. From these data, we were able to identify 44 different genera present in any sample at >1% relative abundance, a commonly employed threshold to distinguish low abundance genera from common or high abundance genera clinically27 (Tables 1 and S1).

TABLE 1. Taxonomic classification of taxa >0.01 relative abundance Phylum Class Order Family Genus Actinobacteria Actinobacteria Actinomycetales Actinomycetaceae Arcanobacterium Trueperella Brevibacteriaceae Brevibacterium Corynebacteriaceae Corynebacterium Nocardiaceae Rhodococcus Propionibacteriaceae Propionibacterium Bacteroidetes Bacteroidia Bacteroidales Bacteroidaceae Bacteroides Porphyromonadaceae Porphyromonas Prevotellaceae Prevotella Flavobacteriia Flavobacteriales Flavobacteriaceae Myroides Firmicutes Bacilli Bacillales Staphylococcaceae Staphylococcus Bacilli Gemellales Gemellaceae Bacilli Lactobacillales Aerococcaceae Facklamia Bacilli Enterococcaceae Bacilli Streptococcaceae Streptococcus Clostridia Clostridiales [Tissierellaceae] 1–68 Anaerococcus Finegoldia Peptoniphilus Tissierella Clostridiaceae Clostridium Lachnospiraceae Peptococcaceae Peptococcus Peptostreptococcaceae Peptostreptococcus Proteobacteria Alphaproteobacteria Rhizobiales Rhizobiaceae Agrobacterium Sphingomonadales Sphingomonadaceae Sphingomonas Betaproteobacteria Burkholderiales Alcaligenaceae Achromobacter Alcaligenaceae Comamonadaceae Delftia Oxalobacteraceae Ralstonia Epsilonproteobacteria Campylobacterales Campylobacteraceae Campylobacter Gammaproteobacteria Enterobacteriales Enterobacteriaceae Klebsiella Morganella Proteus Providencia Pseudomonadales Moraxellaceae Acinetobacter Pseudomonadaceae Pseudomonas Xanthomonadales Xanthomonadaceae Stenotrophomonas

Previous studies have shown VLU microbiomes comprised a diverse bacterial population that varies depending on patient. To examine diversity of the bacterial populations in VLUs in this patient cohort, we quantified the relative abundance of bacterial genera in each patient before wash. We found the top 10 most ubiquitous genera to contain mostly gram-negative bacteria with some representation of gram-positive bacteria. Organisms included aerobes, anaerobes, and facultative anaerobes (Table 2). The top five genera in order of most to least abundant were Proteus, Pseudomonas, Morganella, Providencia and Finegoldia. Of these, Morganella was unique to this patient population compared to several studies examining microbiome in VLU patients16, 28 (Figure 1(A)).

TABLE 2. Top 10 genera in largest number of patients before and after wash Number of patients >0.01 OTUs Average abundance Metabolism Gram classification Before After Before After Proteus 9 7 0.24 ± 0.35 0.15 ± 0.20 Facultative anaerobe Negative Pseudomonas 7 9 0.09 ± 0.15 0.16 ± 0.24 Mostly aerobic, some anaerobes Negative Morganella 5 5 0.08 ± 0.12 0.04 ± 0.07 Facultative anaerobe Negative Providencia 5 4 0.03 ± 0.05 0.06 ± 0.10 Facultative anaerobe Negative Finegoldia 4 3 0.04 ± 0.10 0.03 ± 0.09 Anaerobe Positive Streptococcus 3 3 0.03 ± 0.06 0.03 ± 0.07 Facultative anaerobe Positive Anaerococcus 3 4 0.02 ± 0.03 0.01 ± 0.02 Anaerobe Positive Unidentified Alcaligenaceae 3 6 0.12 ± 0.24 0.04 ± 0.06 Aerobe, facultative anaerobe, anaerobe Negative Unidentified Enterobacteriaceae 3 4 0.04 ± 0.10 0.09 ± 0.19 Facultative anaerobes Negative Corynebacterium 2 2 0.01 ± 0.02 0.01 ± 0.03 Aerobe, facultative anaerobe Positive Note: The average abundance data are expressed as mean relative abundance ± SD. Each genera are identified at the genus level and are contributed by a single OTU.

The relative abundance of VLU genera prior to wash. (A) Distribution of relative abundance of the top 5 genera with highest average relative abundance in patient VLU samples across patients before wash. (B) Hierarchical cluster analysis of Bray–Curtis dissimilarity metric of relative percent abundance >1% across patient samples. Samples clustered into 4 groups. (C) Principal coordinate analysis of the Bray-Curtis index using the top 3 most variable axis. (D) Alpha diversity of each sample in each cluster as measured by Shannon H diversity metric. (E) Relative abundance of the top 10 genera across patients

Cluster analysis based on the Bray-Curtis index of relative abundance of shared genera across patients revealed that patients formed four unsupervised clusters (Figure 1(B)) as determined by comparison of intra- and inter-cluster distances through silhouette analysis (Figure S1B). Significance of hierarchical clusters was determined using a multi-scale bootstrap resampling technique on all clusters (4 clusters = pval >90%, Figure S1C). Principal coordinate analysis of the Bray-Curtis distance indices also displayed four groups when the top three most variable axes were measured (total variance captured 60.8%) (Figure 1(C)). To determine patterns in community composition of each of the four clusters, we examined alpha diversity, a measure of community richness and evenness within clusters (Figure 1(D)). We found that cluster 3 had reduced diversity compared to other clusters. We next examined the number and composition of genera in each patient (Figure 1(E)). We found that each patient varied in the community members present in samples before wash, yet a few taxonomic features were consistent within each cluster (Table 3). Cluster 1 contained a single patient (P-42) predominantly colonized by four genera, the most abundant of these genera being an unidentified genus from the family Oxalobacteraceae. Cluster 2 patients (P-33, P-41, and P-43) were colonized mainly by an unidentified genus from Alcaligenaceae (18%–64%) and had 7–9 genera total. Cluster 3 patients (P-28, P-30, and P-40) were colonized by Proteus as the dominant genera (45%–90%) and were colonized by fewer genera overall (2–4) than other patients in this cohort, suggesting reduced diversity of the microbiome in this patient subset. All Cluster 4 patients exhibited a large abundance of Morganella. These data suggest that there may be underlying characteristics in the wound environment promoting growth of particular genera in some patients in this cohort.

TABLE 3. Relative abundance before and after wash of highest ranked genera (single OTUs) within clusters Patient Genera Rank abundance (before) Percent abundance Before After P-42 Unidentified Oxalobacteraceae 1 0.45 0.09 P-43 Unidentified Alcaligenaceae 2 0.18 0.05 P-33 Unidentified Alcaligenaceae 1 0.64 0.19 P-41 Unidentified Alcaligenaceae 1 0.54 0.05 P-30 Proteus 1 0.45 0.08 P-28 Proteus 1 0.79 0.63 P-40 Proteus 1 0.97 0.30 P-26 Morganella 2 0.31 0.22 P-31 Morganella 2 0.23 0.01 P-27 Morganella 4 0.11 0.07 P-44 Morganella 2 0.22 0.07 3.2 Washing venous stasis ulcers does not change diversity of the ulcer microbiome

Previous studies examining VLU microbiomes report contradictory results regarding microbiome diversity in cases of healed or unhealed wounds. In some cases, diversity is either unchanged or higher in wounds that did not heal, and in other cases, no correlation is found regarding length of time of wound healing or treatment with antibiotics.14, 29 To understand the effect of washing of VLUs on microbiome diversity, we quantified patient samples for metrics of microbiome diversity within sample before and after washing across this patient cohort (Figure 2(A), (B)). We first examined the number of genera >1% in each patient's set of samples before and after washing and compared whether there was effect of washing on species richness across the patient cohort (Figure 2(A)). We found no difference overall in species richness before and after washing (p-value 0.42, Student's paired t-test). We also examined the relative proportion of each species within a sample as a measure of diversity using the Shannon diversity index and compared samples before and after washing across the dataset. No statistical difference was detected across the dataset in evenness of the distribution of genera (p value 0.43, Student's paired t-test) (Figure 2(B)). Interestingly, when examining genera richness and genera diversity within individual patients, some notable differences in alpha diversity could be detected (Figure 2(C), (D)). However, additional data are required to confirm these observations. When quantifying number of genera in before- and after-wash samples by patient (present at >1% relative abundance), we found 3/11 patients lost detected genera, 5/11 patients gained detected genera, and 3 patients experienced no change in detected genera (Figure 2(C)). Also, comparison of Shannon indices before and after washing for each patient revealed an approximately even split between patients with increased diversity and patients with decreased diversity following washing (Figure 2(D)). The lack of significant difference in alpha diversity metrics including richness and diversity suggests that washing does not affect diversity of patient microbiome samples in this cohort.

Diversity of VLU samples before and after wash. (A) Distribution of number of genera of OTUs greater than 1% in patient VLU samples before and after wash. (B) Distribution of Shannon indices of OTUs greater than 1% in patient VLU samples before and after wash. Statistics in (A) and (B) were calculated using a paired Student's t-test. (C) Quantification of number of OTUs greater than 1% in each patient's sample before and after wash. (D) Shannon index of OTUs greater than 1% in each patient's sample before and after wash

3.3 Modulation of community structure of venous stasis ulcers following washing

Though overall species richness and evenness were unchanged by washing in our patient population, specific community composition of wounds may be affected by the washing procedure itself, particularly if washing affects the ability of certain bacteria to adhere to the wound. Specific community structure of VLUs across samples in this patient cohort before wash revealed some similarities within clusters of patients' VLU microbiomes (Figure 1(B)–(D)). Therefore, we next examined changes in community structure associated with washing within each patient and each patient cluster. First, we assessed interpersonal variation of all samples before washing and after washing using Bray-Curtis index and found that samples had a high level of variation in community structure in both before and after samples across patients (Figure 3(A)). When comparing intrapersonal variation of samples before and after washing within each patient, community structure variation was reduced. This finding suggests that community structure is more similar within a patient than across patients, supporting the high diversity of VLU microbiomes. Previous studies surveying the healthy skin microbiome9, 10 have similarly found that interpersonal variation is greater than intrapersonal variation.

Bacterial composition changes before and after wash. (A) Bray-Curtis pairwise measurements were made between all samples in before wash across patients, after wash across patients, and before-after samples within patients. Before, After, and Before-After groups were compared using an ANOVA with Bonferroni adjusted pair-wise p-values and an * denotes p-values <0.05. (B) Relative abundance of top 10 genera across patients

To address whether particular bacteria were more or less affected by the washing procedure, we compared the relative abundance of each bacterial genera across patients before and after washing (Figure 3(B)). We found that no bacterial genera had increasing or decreasing trends following wound washing across all patients. However, particular patients exhibited increases or decreases in specific genera. Moreover, strikingly, the salient genera defining each cluster decreased after wash in all patients. Namely, Oxalobacteraceae decreased in P-42 defining Cluster 1, Alcaligenaceae decreased in all Cluster 2 patients, Proteus decreased in all Cluster 3 patients, and Morganella decreased in all Cluster 4 patients (Table 3). Interestingly, 2 of the 3 patients in Cluster 3 exhibiting the most pronounced decrease in Proteus following washing displayed concomitant increases in Pseudomonas (P-30, P-40), though in other clusters, increases in Proteus following washing did not correlate to a decrease in Pseudomonas (P-42). We also found no correlation in number of genera with Bray–Curtis dissimilarity measure (Figure 3(A)) before and after wash, suggesting that the degree of change in community structure following wash is not tied to the number of genera present before wash.

3.4 Culture-based analysis reveals decrease in abundance of viable bacteria

We qualitatively measured the presence or absence of bacteria using standard culturing methods as complimentary analysis to our 16S rRNA gene sequencing results (Figure 4). We found that samples before wash unanimously had bacterial growth on tryptic soy agar (TSA). Most samples excluding P-33 contained gram-negative bacteria in their pre-wash samples as seen by presence of bacteria on MacConkey growth media plates. All samples except P-33 showed growth on Streptococcus selective plates in prewash samples. Six of 11 samples showed growth on MSA plates indicative of Staphylococcus or Micrococcus. After wash, one patient showed an absence of bacterial growth on TSA. Most patients' samples contained gram-negative bacteria. Interestingly, the majority of patients showed an overall loss in Streptococcus and Staphylococcus/Micrococcus species after washing. These results suggest specific bacteria genera are affected by the washing process as several genera were newly detected or lost due to washing.

Bacterial growth on selective media quantified as presence of at least 1 bacterial colony in 2 replicates. Black represents growth in 2/2 replicates, grey represents growth in 1/2 replicates, and white represents growth in 0/2 replicates