Comparing culture and molecular methods for the identification of microorganisms involved in necrotizing soft tissue infections

Samples in this study were obtained from NSTI patients by debridement of the infected area, performed at Rigshospitalet (Copenhagen, Denmark). A total of 20 samples from 10 patients were included (Table 1). Disposing factors included diabetes mellitus, adiposity and chronic leg ulcur, inguinal hernia, leukemia, immunosuppression. Limb defects and burn wounds. Several patients experienced severe sepsis or septic shock including organ failure. The debrided tissues were immediately transported to the Department of Clinical Microbiology where each sample was divided into three aliquots for standard culture experiments, molecular analyses, and PNA-FISH experiments. The samples for molecular analyses and the FISH experiments were transferred to tubes containing glycerol or ethanol respectively, and kept frozen until analysis. In addition, other culture results were checked for in the patient’s files (e.g. positive blood cultures).

In all cases the material for molecular diagnostics was leftover debridement material from treatment and diagnosing the patients, which would have been discarded otherwise. No extra sampling from the patients was performed. In addition, patient files were only checked for the purpose of treating the patients and correlating microbial findings to the clinical findings. Data from the patient history is exclusively from internal notes in the Department of Clinical Microbiology for that purpose. Samples sent for further analysis were completely anonymized except for the principal investigator (CM). Samples have been destroyed after the study. Therefore, the present study is considered as quality assessment investigating the potential contribution of novel molecular techniques. Based on this, a written informed consent and ethics committee approval were not needed and Danish law was strictly complied.

All culture analyses of the debrided afflicted area were performed at the Department of Clinical Microbiology at Rigshospitalet. All biopsies were analyzed by Gram-staining and culture. Both aerobic and anaerobic conditions were used. Biopsies were plated on brain heart infusion agar (BHIA, Statens Serum Institut (SSI), Copenhagen Denmark), coagulated agar, and 5 % horse blood agar (SSI) for cultures in 5 % CO₂ atmosphere. Aerobic conditions included plating on modified Conradi-Drigalski (“Blue plates”, SSI), in serum bouillon, in thioglycollate media, and on tellurite agar (SSI) in a normal atmosphere. Colonies were further identified by use of Matrix-assisted laser desorption-ionization time of flight mass spectroscopy (MALDI-TOF MS), (Bruker, Bremen, Germany). Antibiotic resistance patterns were analyzed by disc diffusion test on blood agar (SSI) using Neosensitabs (Rosco Diagnostica, Taastrup, Denmark).

DNA was extracted from samples as described previously [26]. Briefly, the tissue samples were cut into small pieces under sterile conditions. Approximately 1 mm³ of tissue was transferred to a microcentrifuge tube containing tissue lysis buffer (ATL, Qiagen) and 20 mg/mL proteinase K (Qiagen). The sample was incubated at 56 °C until visual inspection indicated that lysis was achieved. 100 μL Zirconia/silica beads mixture (50 μL of 0.1 mm diameter, Biospec, PN: 11079101z and 50 μL of 0.7 mm diameter, Biospec, PN:11079107zx) was added to the microcentrifuge tube and the sample was homogenized for 10 min at 25 Hz using a Qiagen Tissuelyser (Model MM300, cat# 85210). DNA from the lysed samples was extracted using the Qiagen DNeasy Tissue kit, according to the manufacturer’s protocol. The DNA was eluted in 200 μL AE buffer (10 mM Tris · Cl; 0.5 mM EDTA, pH 9.0).

PCR was performed with primers that targeted the first 500 bases of the bacterial 16S rRNA gene or the D2LSU region of the fungal 28S rRNA gene. The resulting PCR products were sequenced using the MicroSeq® 500 kit (Applied Biosystems, Carlsbad, California) according to the manufacturer’s guidelines. The resulting DNA sequences were compared to the sequence library included in the MicroSeq® ID analysis software. In cases where sequencing resulted in mixed chromatograms due to 16S rRNA gene products from multiple species, these chromatograms were analyzed using RipSeq Mixed at www.​ribseq.​com.​

Clone libraries were constructed for all samples except 4A, 6A, 7A, 8A, 8B and 8C (due to insufficient volumes of DNA extract). The libraries consisted of near full length 16S rRNA genes (E. coli position 26–1390) which were obtained as described previously [27, 28]. Briefly, PCR amplicons were cloned using the TOPO TA Cloning ® kit (Invitrogen) according to the manufacturer’s instructions. For each surgical sample, 48 clones were subjected to plasmid purification and sequencing were performed by Macrogen Inc. (Korea) using M13F primer (and M13R primer in some cases). Manual refinement of sequences and construction of consensus sequences were done in CLC Main Workbench (CLC bio, Aarhus, Denmark). Sequences were checked for chimeras using the Mallard software package [29], aligned using SINA Web Aligner [30] and imported into the ARB software package [31] for taxonomic lineage assignment, using the non-redundant (NR) SSU Ref database from SILVA Release 106 as reference database. Operational taxonomic units (OTUs) were constructed across all patient samples for clones having a sequence similarity of more than 97 % since these sequences are typically assigned to the same species. One clone from each OTU was sequenced with both M13F and M13R primers. The resulting consensus sequences and their closest relatives in the database were selected to construct phylogenetic trees using neighbor joining, maximum parsimony and maximum likelihood methods. The non-redundant, near full-length 16S rRNA gene sequences, representing each OTU obtained in this study, were deposited in GenBank under the accession numbers (KP114666-KP114679).

Four hundred fifty four-based pyrosequencing was performed largely as described previously [32]. Briefly, the bar-coded FLX-titanium amplicon pyrosequencing targeted the V1–V2 region of the 16S rRNA gene (using 27 F and 338R primers). The DNA fragments were amplified using Platinum Hi-Fi taq polymerase (Invitrogen) with 800 μM dNTP, 2 mM MgCl₂ and 400 nM of each primer. To each reaction 5 μL of template DNA was added and the volume was adjusted to 50 μL. The PCR incubation conditions were 94 °C for 2 min followed by 30 cycles of 94 °C for 30 s, 55 °C for 30 s, and 68 °C for 60 s, with a final extension at 72 °C for 7 min. Tag-encoded FLX amplicon pyrosequencing analyses utilized the Roche 454 FLX instrument with titanium reagents and titanium procedures (Roche).

Analysis of 16S rRNA gene amplicon sequences was performed using Quantitative Insights Into Microbial Ecology (QIIME v.1.3.0) pipeline [33]. The sequencing data was processed initially with AmpliconNoise [34] to remove noise. Then the QIIME pipeline separated the sequences into individual specimen communities based on the unique 5’ barcode sequence and utilized a suite of external programs to make taxonomic assignments and estimate phylogenetic diversity. These data were used to generate taxonomic summaries. The default settings in QIIME were employed for analysis, except that the sequences were grouped into operational taxonomic units using 99 % sequence similarity for clustering; taxonomic assignments were done using Greengenes taxonomy [35].

An aliquot of each DNA extract was loaded into each of 16 wells of an Ibis 96-well BAC (bacteria, antibiotic resistance genes, candida) detection plates (Abbott Molecular) and processed as described previously [36]. Briefly, PCR amplifications were carried out, and the resulting PCR products were then desalted in a 96-well format and sequentially electrosprayed into the TOF MS as described by the manufacturer. The spectral signals were processed to determine the mass of each strand of the PCR products, which in turn were used to derive the base compositions that were then compared to the Ibis database to obtain species level determinations for all microorganisms [37].

Quantification of S. pyogenes [38] and 16S rRNA genes [39] was performed using hydrolysis probe chemistry. The S. pyogenes assay is commercially available from Biosearch Technologies (Novato, CA). For each sample duplicate 25 μL reactions were run, each containing: 12.5 μL Brilliant® qPCR Master mix (Agilent Technologies, Santa Clara, California), 25 μg BSA (Sigma-Aldrich, Brøndby, Denmark), appropriate concentration of primers and TaqMan® probes (S. pyogenes: 400 nM primers and 100 nM probe, 16S rRNA: 900 nM primers and 200 nM probe), 0.75 μM ROX reference dye (Agilent Technologies) and 2 μL of template DNA. Measurements were obtained by absolute quantification using genomic DNA isolated from S. pyogenes (DSM 20565) and P. aeruginosa (DSM 1253) for total bacteria quantification. The number of isolated genomes was calculated based on DNA concentration (Quant-iT™ dsDNA Assay Kit (Invitrogen)) and genome size estimated to be 1.8 Mbp for S. pyogenes and 6.5 Mbp for P. aeruginosa (http://​img.​jgi.​doe.​gov/​cgi-bin/​pub/​main.​cgi). Dilution series of the genomic DNA covered a range of 10⁶-10⁰ genome copies. Reactions were run on an Mx3005P (Agilent Technologies) with the following program: 10 min at 95 °C, followed by 40 cycles of 30 s at 95 °C, 1 min at 60 °C.

The number of gene copies measured by qPCR was converted to number of CFU per gram sample using \( \raisebox{1ex}{$\mathrm{C}\mathrm{F}\mathrm{U}/\mathrm{g}=\kern0.5em {\mathrm{C}}_{\mathrm{m}\mathrm{easured}}/{\mathrm{C}}_{\mathrm{genome}}*\left({\mathrm{V}}_{\mathrm{total}}/{\mathrm{V}}_{\mathrm{used}}\right)$}\!\left/ \!\raisebox{-1ex}{${\mathrm{m}}_{\mathrm{sample}}$}\right.. \) Here C_measured is the number of copies measured and C_genome is the number of gene copies in the genome of one CFU. The standard deviation of all measurements above the detection limit of the assays was calculated. In cases where both the S. pyogenes and 16S rRNA assay were above the limits of detection, a two tailed t-test was used to provide a hypothesis test of the difference between population means. A statistical value of ≤ 0.05 was considered significant.

The samples were prepared for visualization by imbedding in paraffin, which was sectioned (4 μm) and mounted on microscope slides and stored at room temperature. Before staining or hybridization, the slides were deparafinated by using 2x 5 min xylene, 2x 3 min 99.9 % EtOH, 2x 3 min 96 % EtOH, and washed 3x 3 min in sterile water. The deparafinated NSTI sections were analyzed by FISH using a mixture of Cy3-labeled “Strept probe” targeting Streptococcaceae [40] and a Cy5-labeled broad range bacteria probe (EUB-338) [41]. Hybridization was performed by covering the slide with 12 μL of hybridization buffer containing 0.9 M NaCl, 0.02 M Tris/HCl (pH 8), 0.01 % SDS, and 30 % formamide and probe mix (5 ng/μL of the respective probes) followed by incubation at 46 °C in a humid chamber. After 90 min. slides were rinsed with washing buffer (0.102 M NaCl, 0.02 M TRIS/HCl, 0.01 % SDS) preheated to 48 °C and then incubated in the washing buffer for 15 min at 48 °C. Subsequently, DNA of both bacteria and host cells was stained with 50 μg/ml 4’,6-diamidino-2-phenylindole (DAPI) and incubated for 15 min in darkness at RT after which the slides were washed with dH₂0 and air dried. The slides were mounted with Vectashield (Vector labs) and a cover slip was added. Slides were investigated using a LSM 710 confocal laser scanning microscope (Zeiss, Germany).

The findings by routine culture at the Department of Clinical Microbiology, Rigshospitalet (Copenhagen, Denmark), mostly revealed the presence of one type of pathogenic bacteria in the surgical samples (in 7 of the 10 patients) (Table 2). These monomicrobial infections were primarily caused by streptococci (71 % of the monomicrobial infections), specifically S. pyogenes, S. pneumoniae and non-hemolytic streptococci. The remaining monomicrobial culture findings were identified as Acinetobacter baumanii (patient 3) and fungal infection (patient 7). Furthermore, two patients were found to harbor more than one microorganism (patients 5 and 6), where Bacteroides fragilis with Clostridium paraputrificum and S. pyogenes with Escherichia coli were found. One fourth of the samples investigated by culture did not result in growth of microorganisms. Three of these surgical samples originated from patients where other samples taken from the site of infection resulted in growth of microorganisms. The remaining culture-negative samples originated from a patient, where none of the samples resulted in growth of microorganisms (patient 10).

Generally, the molecular methods confirmed the findings by culture (Table 2 and Table 3). However, using the multiple molecular methods, microorganisms were found in all samples including those that were culture-negative, and in most culture-positive cases additional microorganisms were identified by the molecular methods (Table 3, Fig. 1). Overall, the different molecular methods gave concordant results (although the16S rRNA clone libraries were only constructed for 16 of the samples). There were, however, cases where different microorganisms could only be detected by one molecular method. This was either due to misidentification or identification of additional species (Fig. 1). For patient 10, the results were difficult to interpret since the molecular methods gave differing results (S. pyogenes by Microseq, while Ibis found S. pneumoniae, Clostridium septicum and CoNS, and clone library and 454-pyrosequencing did not give results).

Of the molecular methods applied, only the Ibis T5000 biosensor could identify Candida albicans in sample 7B and Cladosporium cladosporioides in sample 6A.

The findings of S. pyogenes by molecular methods could generally be confirmed by qPCR (Fig. 2). Based on the measurements of bacterial 16S rRNA genes, S. pyogenes was the dominant microorganism in most of the samples, except 6A and 7B. This corresponds with the trend seen in the 454-pyrosequencing data (Fig. 1).

Using FISH it was possible to visualize bacteria in the NSTI samples. The bacteria observed were generally clustered together, and did not appear to penetrate into the lipid droplet of the fat cells, but were instead found in the matrix surrounding the fat cells (illustrated on the representative images obtained for samples from patient 1A, Fig. 3). In a few cases neither of the FISH probes resulted in visualization of cells, including samples 5A, 10A and 10B. Generally, the Streptococcaceae-specific FISH probe confirmed the cases where the bacteria were found by molecular methods.

Microorganisms could be identified by the molecular methods in all samples with the exception of samples from patient 10. Thus molecular methods identified microorganisms in cases where no growth was observed by culture. Overall, there were a total of 17 samples where culture and molecular methods were in agreement, giving either concordant (13 samples) or partially concordant results (four samples). The partial concordance is attributable to the greater diversity found by the molecular methods vs. culture (Tables 2 and 3 and Fig. 1), which is consistent with similar comparative studies of other clinical conditions evaluating microbial detection methods [25, 44‐48]. In the remaining three cases (4B, 7B and 8C) the disagreement between culture and molecular methods were caused by lack of culturability. Antibiotic therapy may be the cause of this discrepancy, since other samples taken from patients earlier in the course of disease did show growth of microorganisms, which were in agreement with the findings by molecular methods (Table 2). Interestingly, these three cases originated from three different patients where multiple samples were taken and only a single sample gave negative results by culture, illustrating that spatial distribution of pathogenic microorganisms may be an issue that must be taken into account during interpretation of results.

Four different molecular methods were used to identify microorganisms: 1) Microseq (direct Sanger sequencing); 2) 16S rRNA gene clone libraries and Sanger sequencing (except for samples 4A, 6A, 7A, 8A, 8B and 8C due to insufficient amounts of DNA sample); 3) the Ibis T5000 biosensor; and 4) 454-based 16S rRNA gene pyrosequencing. Although all of the molecular methods generally provided concordant results, there were some cases where a microorganism was only detected by one of the four methods (Table 3, Fig. 1). Only microorganisms found by at least two molecular methods were considered to be present in the sample. For a single patient (patient 10) the various molecular methods gave discrepant or negative results. The number of sequence reads obtained by 454-pyrosequencing was very low, and it was not possible to construct clone libraries due to negative PCR for patient 10. These findings indicates that the obtained results (Table 3) may be contaminants or background, which is supported by the fact that neither of the FISH probes used in this study gave a detectable signal for the samples obtained from patient 10.

Discrepancies among the molecular methods were primarily associated with the Microseq method, which misidentified or missed the microorganisms that could be found by the other three molecular methods and (most often) culture. In the Microseq reactions the different DNA strands from different species are competing for the same (and limiting) reagents. The method uses only one primer set in contrast to the redundant strategy of the Ibis T5000 biosensor. Furthermore, the Microseq has a limited resolution compared to clone library where different DNA strands are physically picked out and individually sequenced and the 454-pyrosequencing where different DNA molecules are automatically separated before amplification and sequencing. It is therefore possible that the method is oversensitive toward some species of special interest with high affinity for the primers.

In three cases (3A, 5A and 10A) Microseq identified S. pyogenes that were not detected by any other method (except possibly by light microscopy for sample 3A as discussed below). Findings of S. pyogenes in NF can only in really rare cases be interpreted as a false positive result. Concerning the relative sensitivity of the techniques, all applied methods in this study have various biases e.g. differential amplification, primer choice. The chromatograms as well as the intensity of the raw files from Microseq are of sufficient quality to deem the results as correct. There are no indications as to the Microseq results being false positives or cross contaminations, however we do acknowledge that this may be the case.

Interpretation of quantitative results can reveal some interesting aspects of the NSTIs. By qPCR, the presence of S. pyogenes was quantified and related to the total number of bacteria in the sample (estimates of cell numbers based on 16S rRNA gene measurements). When comparing results by qPCR it is important to keep in mind that small variations should not take on assumed relevance, since it has been documented that at best there is a 0.5 log₁₀ variance between repeats of the same template concentrations [49]. Based on the qPCR results (Fig. 2), S. pyogenes generally appears to be the dominant pathogen when it is present in patients, except samples 6A and 7B, where the total number of bacteria seemed to exceed the number of S. pyogenes. These samples contained a number of different species, which supports the findings by qPCR. In the cases where S. pyogenes was dominant (and sample 6A), all the applied methods were able to detect and identify the pathogen. For sample 6A, the relative abundance of S. pyogenes is around 1-2 % according to 454 pyrosequencing and qPCR (Figs. 1 and 2). It is noteworthy that culture identified such a relatively low abundance species, compared to for instance Bacteroides fragilis (detected by both Ibis T5000 biosensor and constituted approximately half of the 454 pyrosequencing reads).

An unexpected finding by qPCR was the presence of S. pyogenes in sample 7B (approximately 7000 CFUs/mg sample). The only other method to detect the pathogen was 454 pyrosequencing (Fig. 1), where the species constituted less than 0.01 % of the reads (and is therefore not reported in Table 2). Seen in this light, the number of CFUs/mg sample quantified by qPCR seems relatively high, and indicates the importance of relating qPCR measurements to other data (here both qPCR measurements of other species and broad-range methods). Without this comparison, it would be easy to mistakenly focus on S. pyogenes, when all the other molecular methods indicate that Mycoplasma species and Fusobacterium necrophorum are the problem.

The use of qPCR support our criteria of detection of bacteria by at least two methods, since S. pyogenes could not be quantified in samples 3A, 5A, 10A and 10B where Microseq had indicated the presence of the pathogen. This is further strengthened by the fact that it was not possible to visualize Streptococcaceae in these samples by FISH, although it is possible that cells are present but not visible due to the high detection limit generally associated with FISH. For the remaining samples where streptococci were detected by multiple methods, it was possible to visualize the organisms by FISH (Fig. 3).

Compared to the aggressive nature of the infection, the relative low number of bacteria generally detected in the debrided tissue samples is somewhat surprising. We speculate that successful antimicrobial treatment is the cause of this, both by reducing the number of pathogens, but possibly also by rendering the pathogens metabolically inactive, which will impede detection by FISH. This seems plausible since the majority of patients in this study survived the NSTI. We cannot, however, rule out that some pathogens were present in the tissue but not detected due to problems during transport and storage of samples for FISH or because they were present in concentrations below the limit of detection for FISH (Fig. 3-A and -B).

The most common finding by molecular methods was S. pyogenes as the sole or dominant pathogen (patients 1, 2, 4 and 9). Some cases of polymicrobial infections were found and included E. coli, streptococci and the anaerobes Bacteroides fragilis, Fusobacterium spp. and Clostridium spp. These findings correspond with bacteria previously reported to be present in polymicrobial NSTIs [13, 42, 50]. Furthermore, fungal NSTI due to Candida albicans has also been reported [6, 42]. The detection of Mycoplasma spp. in polymicrobial NSTI (patient 7 by all molecular methods), is to the best of our knowledge unique. However, animal studies have shown that ulcerative dermal necrosis can be induced in mice by Mycoplasma arthritidis [51]. Mycoplasmas are associated with the mucosa and reside primarily in the respiratory tract and rarely penetrate the submucosa, except in cases of immunosuppression or instrumentation. The lack of cell wall makes the mycoplasmas very sensitive to environmental conditions and isolation of mycoplasmas is complicated due specific nutrient requirements and lack of a single optimal media formulation [52], which may explain why they have not been isolated in NSTI patients before. Interestingly, the localization of infection in patient 7 where mycoplasmas were detected was the neck, but originated and spread from a dental focus, corresponds with the association of mycoplasmas with the respiratory tract.

Infections by species such as A. baumannii (patient 3) and S. pneumoniae (patient 8) as the sole or dominant species are unusual findings. However, A. baumanii is an emergent pathogen and has increasingly been recognized as a prevalent and significant nosocomial pathogen associated with sepsis, wound infections, and pneumonia [53]. A. baumannii and other Acinetobacter sp. have been described as participants in polymicrobial NSTIs [4, 14, 54] and some reports have identified A. baumannii as the sole agent in NSTIs. [53, 55‐57]. Light microscopy of the tissue revealed Gram-positive cocci in chains, which may be involved in the initial phases of the infection. The Gram-positive cocci may either be S. pyogenes as suggested by Microseq (not confirmed by qPCR) or low abundance staphylococci, which were detected by 454-pyrosequencing. The A. baumanii involvement in this case is probably explained by the presence of chronic leg ulcers which could be either colonized or harbor the A. baumanii as biofilms deep in the ulcer [36, 58, 59]. S. pneumoniae, which was dominant in samples from patient 8, is a widespread pathogen that displays enormous heterogeneity with respect to phenotype and pathogenicity, and has been implicated in community-acquired pneumonia, sinusitis, otitis media, orthopaedic infections and meningitis [44, 60‐62]. NSTI due to S. pneumoniae is rare and has primarily been reported in cases where patients were immunosuppressed or had other underlying conditions [21, 63‐67], which does not correspond to the patient history in this case (Table 1). However, serious infections upon septic spread of the S. pneumoniae including to joints and bursas, as in this case, is not unusual.

The realization that many pathogens can cause NSTIs, and that no specific combination of species are found in all cases means that clinicians should be prepared to treat any combination of microbial pathogens [4, 42]. Although appropriate antimicrobial treatment cannot cure NSTI, it can help during the acute phase of the infection [4], which highlights the importance of rapid and comprehensive identification of the pathogens involved.