Examining the effect of wound cleansing on the microbiome of venous stasis ulcers

Common treatment for venous leg wounds includes topical wound dressings with compression. At each dressing change, wounds are debrided and washed; however, the effect of the washing procedure on the wound microbiome has not been studied. We hypothesized that wound washing may alter the wound microbiome. To characterize microbiome changes with respect to wound washing, swabs from 11 patients with chronic wounds were sampled before and after washing, and patient microbiomes were characterized using 16S rRNA sequencing and culturing. Microbiomes across patient samples prior to washing were typically polymicrobial but varied in the number and type of bacterial genera present. Proteus and Pseudomonas were the dominant genera in the study. We found that washing does not consistently change microbiome diversity but does cause consistent changes in microbiome composition. Specifically, washing caused a decrease in the relative abundance of the most highly represented genera in each patient cluster. The finding that venous leg ulcer wound washing, a standard of care therapy, can induce changes in the wound microbiome is novel and could be potentially informative for future guided therapy strategies.

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 development 5,6 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][10][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][10][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][13][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 antibioticresistant 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.

| 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 technique 23 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 À80 C. Wounds were washed and then swabs were collected immediately following wash using the same procedure as pre-wash.

| 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.

| 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.

| 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 GreenGenes 25 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.

| 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 = -Σp i Áln(p i ), where p i 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 beforewash 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.

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

| 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 afterwash 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 clinically 27 (Tables 1 and S1).
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 patients 16,28 (Figure 1(A)).
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  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 con- ). However, additional data are required to confirm these observations. When quantifying number of genera in beforeand 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.

| 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 T A B L E 2 Top 10 genera in largest number of patients before and after wash 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 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.

| 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

| DISCUSSION
Our 16S sequencing classification of bacterial genera in VLU wounds illustrates that microbiome differences within and across patients before and after washing are diverse and individual to the patient. We identified that patients' microbiomes in wounds prior to washing ranged from being predominantly characterized by a few genera to many genera and that relative abundances of genera were highly variable across patients. Trends in the number and distribution of genera within patients before and after washing were variable, with both increases and decreases in diversity. Thus, no consistent trend in diversity was observed. Though the number of patients in this study was relatively small, we were able to find trends in VLU microbiome Chronic wound infections have been shown to be polymicrobial, yet biofilms can also be predominantly populated with one bacterial species. Interestingly, before washing of wounds in this patient cohort, an average of 7 bacterial genera, and as many as 15 genera, could be detected. However, two patients had wounds with a single genera colonizing at more than 50% relative abundance. The varying degree to which patient samples were polymicrobial agrees with previous research that found that most wounds are polymicrobial, but a small percentage of wounds were predominantly colonized by one or two species. 20 Bacteria ranging from commensals to known pathogens have been shown to form biofilms in studies of VLU microbiomes. In our study, patient wounds with predominantly one genus were comprised of Proteus. Proteus, an opportunistic pathogen, has been shown to be an infectious agent in soft tissue infections, and was seen as a dominant genera colonizing VLUs in another study, though only in a few patient samples. 28 In diabetic foot ulcers with monoculture wounds, Proteus mirablis was associated with limb loss. In the same study, Morganella, another predominant colonizer in this study, and known contributor to soft tissue infections, was also found associated with limb loss in diabetic foot ulcers. 29 Members of the family Alcaligenacea were also highly represented in two patient samples. Genera within Alcaligenacea are gram-negative rods found in a variety of environments including the human body, and have shown conflicting results in regards to promoting or inhibiting wound healing. Interestingly, one study has implicated Alcaligenes faecalis as a causative agent in skin and soft tissue infections in a small number of patients with vascular disease. 30 In contrast, a more recent study examining the microbiome of diabetic foot ulcers found that Alcaligenes faecalis promoted wound healing in an in vitro skin model suggesting that host, environment, or organism factors determine wound disease outcome in relation to this organism. 31 We also found that a genera identified as Pseudomonas was highly represented in patient samples, particularly before wash.
P. aeruginosa is the most common producer of single organism biofilms and is sometimes associated with poor prognosis for wound healing. 20 However, Pseudomonas representatives have also been shown to be able to colonize wounds without delaying wound healing, suggesting that organisms in this genus may have a diverse role in the wound microbiome. 32 The presence of Proteus, Alcaligenacea, and Pseudomonas in patients in this dataset as predominating colonizers before wash may be suggestive of wounds that have formed biofilms and therefore may be more recalcitrant to therapy.
Previous studies have also found anaerobic bacterial species to predominate in the wound. 8  The stability of the microbiome in chronic wounds over time is associated with poor wound healing. 7 Topical antibiotics and antiseptics are more efficacious when they are able to change the dynamics of the wound microbiome. 7,27 In this study, wounds were washed with chlorohexidine gluconate and hypocholorous acid, each of which is commonly used independently in wound cleansing. 33 Chlorohexidine, when used in isolation, has been shown in a previous study to confound the effects of DNA sequencing through retention of DNA on the skin surface leading to an inability to detect changes in genera or patient microbiome diversity post-wash. The authors hypothesized that the inflammatory reaction induced by chlorohexidine could lead to disruption of the skin barrier surface and promote DNA retention; however, the molecular mechanism of DNA retention was not formally tested in this study. 22 In contrast, although wound washing did not affect overall diversity in our study population, washing was able to destabilize the microbiomes of several patients, as shown by the poor correlation of before and after samples, as well as changes in the relative abundance of bacteria. We hypothesize that we could capture microbiome changes because chlorohexidine and hypochlorous acid were added in combination. increase or decrease in overall diversity of bacteria with standard of care therapy. We did not record debridement as part of standard of care for this patient cohort; however, we recognize that there could potentially be differences in specific genera in patient microbiomes that may be due to debridement in conjunction with washing.
Though diversity of microbiomes was unaffected by washing regimens, individuals in the cohort displayed large shifts in microbiome composition with poor correlation in relative abundances of bacteria before and after wash. Interestingly, wounds with shifting microbiomes may promote wound healing, particularly when a pathogenic bacteria is displaced from the wound niche. 7 Washing, in this patient cohort, had varying effects and it will be interesting to understand the impact of washing and microbiome dynamics on wound healing in future studies.