Modern molecular tools such as 16S rRNA gene-based pyrosequencing provide powerful means to define chronic wound bacteria. We found that chronic wounds supported complex microbial communities comprised of a wide-range of bacterial taxa including fastidious anaerobic bacteria that were not observed using culture-based methods. The bacterial wound communities characterized in this study were similar in composition to those reported by other groups using 16S rRNA gene-based methods 
. The number and proportion of bacterial taxa ranged greatly in individual wounds. Additional research involving longitudinal sampling is needed to understand the dynamics of bacterial communities in chronic wounds.
Not surprisingly, bacterial diversity was substantially higher when determined by 16S rRNA gene-based pyrosequencing analysis as compared to the culture-based analyses. The limitations of culture-based methods to characterize diverse bacterial communities from environmental and clinical samples have been noted previously; however, many organisms missed by culture-based methods in the current study were theoretically culturable using conventional methods. Some of the organisms that were missed by culture-based methods were proportionally rare and may have been masked by more dominant organisms in the culture media. Other organisms, such as those belonging to the Neisseriaceae and Campylobacteriaceae families, are fastidious and thus require special culture media that are not typically used when culturing wounds in clinical laboratories. Obligate anaerobes, such as Clostridiales family XI, are particularly difficult to grow and were not identified using culture-based methods in the current study. Using 16S rRNA gene-based sequence analysis, we identified bacteria from Clostridiales family XI in 25 of the 32 wounds analyzed. Five genera from this family were identified: Anaerococcus
, and Peptoniphilus
. Complex anaerobic microbiota that include the Clostridiales family XI have been associated with diseases such as bacterial vaginosis 
, diabetic foot ulcers 
, necrotizing fasciitis 
, and periodontal disease 
. Thus, the data presented here highlight the limitations of routine clinical culture to detect potentially important fastidious pathogens.
We compared 16S rRNA gene-based pyrosequencing to culture-based analyses for detecting bacterial taxa that were cultured at least once during the study. In this analysis, we found that detection of these culturable bacteria was consistently greater by 16S rRNA gene-based pyrosequencing as compared to culture-based methods. While culture-negative/pyrosequencing-positive discordant pairs were common, culture-positive/pyrosequencing-negative discordant pairs were rare. All bacterial families identified by culture-based methods were targeted by the amplification primers used in the current study; therefore, insufficient sampling is the most likely explanation for the rare culture-positive/pyrosequencing-negative discordant pairs 
. In contrast, there are several possible explanations for culture-negative/pyrosequencing-positive discordant pairs, including: 1) Molecular detection of viable/non-culturable bacteria.
Viable/non-culturable bacteria may include non-planktonic bacteria existing in biofilms, which are common in chronic wounds; 2) Molecular detection of bacteria that were proportionately rare within the community and masked by more dominant bacteria in culture media.
This is an expected limitation of using non-selective culture media; 3) Molecular detection of DNA from dead (non-viable) bacteria.
Detecting DNA from dead bacteria is a common criticism of using DNA-based molecular methods to characterize microbial communities. While we acknowledge this potential bias, our data suggest that underestimation by culture-based methods is far more likely. Strategies to reduce or eliminate nucleic acids from dead bacteria, such as incorporation of DNA digestion steps prior to cell lysis, may help minimize detection of dead bacteria. Another approach would be to perform RNA (cDNA)-based analysis, which would identify bacteria that are metabolically active.
The V3 hypervariable region is one of the most phylogenetically informative regions of the 16S gene, but this study illustrates the limitations of this region for taxonomic assignment using the RDP classifier. Previous work with the RDP classifier indicated that 83.2% of the bacteria in the Bergey corpus could be accurately assigned to the appropriate genus using 200 base segments of the 16S gene. Three phyla—Firmicutes, Proteobacteria, and Actinobacteria—were the most commonly misclassified 
. The 16S sequences used in the current study were all greater than or equal to 200 bases, with most (89.1%) of the sequences falling into Firmicutes, Proteobacteria, and Actinobacteria phyla. High percentages of the Firmicutes and Actinobacteria sequences were successfully assigned to the genus level, 88.0% and 82.3%, respectively; however, only 53.0% of the Proteobacteria sequences were successfully assigned to the genus level. Most of the Proteobacteria sequences belonged to four families: Pseudomonadaceae, Enterobacteriaceae, Oxalobacteraceae and Neisseriaceae. Varying proportions of all four families were successfully assigned to the genus level: 69.2%, 29.0%, 23.8% and 1.8%, respectively. A combination of sequence homoplasy and inaccurate database assignments (i.e., falsely assigned reference sequences in the RDP database) could have contributed to our inability to assign Proteobacteria sequences to taxonomic groups below the family level.
Recent antibiotic use was associated with increased Pseudomonadaceae colonization in the current study. A similar association was reported previously in a study of tracheal colonization among critically ill, intubated patients 
. In this earlier study, daily endotracheal aspirates were collected from patients after intubation. Bacterial communities from six of the seven patients shifted from relatively diverse communities to Pseudomonas
-dominated communities with the administration of antibiotics. Interestingly, Pseudomonas
isolates collected from these patients were susceptible to the administered antibiotics in laboratory drug-susceptibility tests. It was hypothesized that this paradoxical finding was the result of differential susceptibility of Pseudomonas
growing planktonically versus those growing in biofilms 
. Biofilms are thought to be an important factor contributing to the chronicity of certain non-healing wounds. Administration of antibiotics may select for biofilm-producing organisms such as Pseudomonas
and delay rather than aid wound healing.
One of the limitations of our study was its observational design, which may have resulted in selection bias. Participants were not excluded from the study based on prior therapies and eight patients entered the study having been treated with antibiotics within the previous two weeks. Those participants entering the study with recent exposure to antibiotics may have been treated in response to pre-existing Pseudomonas colonization or infection, which may have biased the observed association between antibiotic use and increased Pseudomonas colonization. Three participants entered the study without recent exposure to antibiotics and were treated during the study. Sub-analyses of these three participants revealed a significant increase in Pseudomonas abundance after antibiotic treatment. These data support the hypothesis that antibiotic use selected for increased Pseudomonas colonization, but additional prospective studies will have to be conducted to confirm these findings.
The 16S rRNA gene-based pyrosequencing analysis confirmed our clinical observations indicating that diabetics were significantly more likely to be colonized with Streptococcus
. Increased Streptococcus
colonization may be an important factor contributing to the disproportionate morbidity associated with chronic wounds among diabetics compared to non-diabetics 
. Antibiotic use was associated with decreased Streptococcus
colonization among diabetics and thus may be a suitable therapeutic option for treating diabetic patients with Streptococcus
infections. Further studies are needed to confirm the association between diabetes and Streptococcus
colonization and to elucidate the biological basis for this association.
Currently, wound management is largely empirical and based on principles of reducing bacterial load and preventing infection 
; however, the complexity of the wound environment makes it likely that antimicrobial therapy could result in unintended consequences. We have little prospective data on the microbiological response to antimicrobial wound therapies. Thus, application of 16S rRNA gene-based pyrosequencing to characterize wound microbial communities with respect to clinical outcomes and therapeutic interventions (particularly antibiotic treatments) will provide critical insights into the roles of microbiota in wound healing and the impacts of wound therapies.