Total joint arthroplasty is an effective treatment for osteoarthritis or rheumatoid arthritis of the hip, the knee, and other joints. However, periprosthetic infection after arthroplasty is always a concern. The problem is that there are occasional cases of occult infection that are difficult to detect preoperatively with use of conventional microbiologic culture techniques1-3. If the implant site is infected, it is essential to identify the etiologic agents—i.e., broadly, whether the species is gram-positive or gram-negative and, specifically, whether Staphylococcus is methicillin-resistant or not. Furthermore, when two-stage revision is chosen, it is often difficult to confirm resolution of the infection after implant removal.
Several recent studies have demonstrated the potential of using real-time polymerase chain reaction assays for the rapid and sensitive identification of bacterial DNA in periprosthetic infections4-6. However, although real-time polymerase chain reaction can enable a far more rapid diagnosis of infection during revision surgery, it still takes two to three hours to extract deoxyribonucleic acid (DNA) with use of conventional methods.
In this study, we utilized an ultrasonication approach to facilitate simple DNA release from the tissues, and we subsequently performed simultaneous real-time polymerase chain reaction assays for both methicillin-resistant Staphylococcus and broad-range bacterial infection. The purposes of the study were (1) to clarify any differences in the detection ability of this method as compared with conventional DNA-extraction methods and (2) to validate the sensitivity and specificity of this method for clinical use.
Preliminary Investigation of Cultured Bacterial Strains
As a preliminary investigation, we confirmed the detection limits of real-time polymerase chain reaction using DNA extracted from cultured bacterial strains, including Escherichia coli (ATCC [American Type Culture Collection] 23231), Staphylococcus epidermidis (ATCC 12228), and clinically isolated strains of methicillin-resistant Staphylococcus aureus, methicillin-resistant Staphylococcus epidermidis, methicillin-resistant Staphylococcus hominis-hominis, Streptococcus species, Pseudomonas, and Salmonella. For the universal polymerase chain reaction, the melting peak temperature was confirmed for each species. Methicillin-resistant Staphylococcus-specific polymerase chain reaction and universal polymerase chain reaction were then performed simultaneously with use of a single amplification program for all eight species.
Patients
All patients scheduled to undergo a second operation at the site of a previous total joint arthroplasty from 2006 to 2008 at our institution were enrolled in this prospective study. Twenty-three patients who underwent a total of thirty operations, including implant removal, one-stage or two-stage revision surgery, and débridement without implant removal, were consecutively enrolled in the study. Eleven of these operations involved removal of a hip prosthesis, three involved removal of a knee prosthesis, ten were revision total hip arthroplasties, three were revision total knee arthroplasties, two involved exchange of hydroxyapatite blocks containing antibiotics after the development of an infection following a total knee arthroplasty, and one was a débridement at the site of a total knee arthroplasty. We obtained intraoperative specimens from three different places for each patient (the capsule, the femoral side, and either the acetabular side [in the total hip arthroplasties] or the tibial side [in the total knee arthroplasties]). Each specimen was then divided into three pieces and subjected to real-time polymerase chain reaction, microbiologic culture, and histopathological evaluation. The institutional review board of our institution approved this study.
Intraoperative Real-Time Polymerase Chain Reaction
We transported each sample from the operating theater to our laboratory in a sterile plastic bag. One milliliter of sterile water was added to each specimen, and we then performed ultrasonication (Bransonic 2510 Ultrasonic Cleaner; Branson Ultrasonics, Danbury, Connecticut) at a frequency of 40 kHz for five minutes. Each plastic bag containing the sterile water and the tissue sample was placed separately in a sonicator. The resulting sonicated solutions were collected and applied to a DNA purification column (QIAamp DNA Mini Kit, QIAGEN, Hilden, Germany). This processing is needed to remove impurities such as blood and joint fluids that can inhibit the polymerase chain reaction. After the addition of 200 µL of Buffer AL (aqueous solution of guanidine hydrochloride; QIAGEN) and ethanol, the sample solutions were loaded onto purification columns, washed, and then eluted according to the manufacturer's instructions.
We immediately analyzed each sample with real-time polymerase chain reaction using the LightCycler system (Roche Diagnostics, Mannheim, Germany). Two different primer and probe sets were used simultaneously: a commercially available methicillin-resistant Staphylococcus aureus-specific detection kit (Roche Diagnostics) and a set designed for broad-range detection by universal polymerase chain reaction6 that targeted a part of the 16S rRNA gene. Both polymerase chain reactions were performed with use of the same cycling program comprising a hot start at 95°C for ten minutes followed by forty-five cycles at 95°C for ten seconds, 55°C for ten seconds, and 72°C for twelve seconds. Universal polymerase chain reaction was not available in our laboratory prior to 2008, so only methicillin-resistant Staphylococcus-specific polymerase chain reaction was performed for several of the patients, who were operated on before that time.
If we obtained a positive result from at least one polymerase chain reaction method (methicillin-resistant Staphylococcus-specific or universal) in at least one sample, the case was defined as polymerase chain reaction-positive. Microbiologic cultures and histopathological evaluations were also identified as positive if at least one specimen was positive.
Microbiologic Culture and Histopathological Evaluation
All specimens were submitted for standard microbiologic culture. The direct plating method was performed at first with simultaneous enrichment with use of the broth culture method. Bacteria were then allowed to grow for up to five days. A VITEK 2 Compact instrument (bioMérieux, Durham, North Carolina) was used for the automated identification of organisms according to the manufacturer's protocol. In addition, histopathological evaluation of all specimens was carried out intraoperatively with use of frozen sections and was also performed postoperatively with use of permanent preparations. The microscope slides were considered suggestive of infection if a minimum of ten high-power fields (×400) contained ten or more neutrophils7,8.
The combined results of the microbiologic culture and the histopathological evaluation were considered for the final diagnosis of infection to calculate the sensitivity and specificity of the intraoperative real-time polymerase chain reaction. If either test showed a positive result, a diagnosis of infection was made. The combined results of the histopathological evaluation of the frozen sections and the intraoperative real-time polymerase chain reaction were considered in any subsequent therapeutic decision-making, such as whether implant removal for two-stage revision or one-stage revision was preferable.
Evaluation of Ultrasonication as a DNA Extraction Method
We evaluated the difference between the success of bacterial detection in DNA samples extracted with use of the described ultrasonication method and the success in samples obtained with conventional manual DNA-extraction methods. Ten culture-positive samples with available remaining tissue were selected from the samples obtained during the thirty arthroplasty procedures analyzed in this study. Each sample was then equally divided into two pieces; one was subjected to DNA extraction by ultrasonication and the other was processed with use of a manual DNA-extraction method (QIAGEN DNA extraction kit). Twenty-nine pairs of samples were prepared in this way for real-time polymerase chain reaction analysis of infection. Methicillin-resistant Staphylococcus-specific polymerase chain reaction was used for the samples that were positive for methicillin-resistant Staphylococcus on culture, and universal polymerase chain reaction was used for those that were positive for other species on culture. The differences in the crossing point (the point at which DNA amplification occurs) between these ultrasonication and manual-DNA-extraction samples were then investigated quantitatively.
Source of Funding
In support of this research, one author received grant funding from the Japan Hip Joint Research Foundation.
Our preliminary investigation confirmed that it was possible to perform methicillin-resistant Staphylococcus-specific polymerase chain reaction and universal polymerase chain reaction simultaneously with the same amplification program for all eight bacterial strains. Figure 1 shows the quantification mode of the LightCycler system that we used for methicillin-resistant Staphylococcus aureus, Staphylococcus epidermidis, and Escherichia coli. Methicillin-resistant Staphylococcus-specific polymerase chain reaction detected methicillin-resistant Staphylococcus aureus, methicillin-resistant Staphylococcus epidermidis, and methicillin-resistant Staphylococcus but not other strains. Universal polymerase chain reaction detected all eight strains and correctly differentiated gram-positive from gram-negative species (Fig. 2). The detection limit of methicillin-resistant Staphylococcus-specific polymerase chain reaction was 100 CFU/mL, and that of universal polymerase chain reaction was 1000 CFU/mL. The melting peaks in the universal polymerase chain reaction showed a temperature of approximately 62.9°C for all gram-positive species and a lower temperature of approximately 56.2°C for all gram-negative species (Fig. 2).
The actual time required to perform the whole procedure, from the attainment of the tissue samples to the confirmation of the final result of the polymerase chain reaction, was ninety minutes or less for all operations. As compared with the combined results of the microbiologic culture and histopathological evaluation, this assay—i.e., the combined results of the universal and methicillin-resistant Staphylococcus-specific polymerase chain reaction tests—had a sensitivity for clinical use of 0.87, a specificity of 0.8, a positive predictive value of 0.81, and a negative predictive value of 0.86. The sensitivity and specificity of the universal polymerase chain reaction alone were 0.83 and 0.75, respectively, when compared with the combined results of the microbiologic culture and histopathological evaluation. The sensitivity and specificity of the methicillin-resistant Staphylococcus-specific polymerase chain reaction alone were 0.4 and 1.0 (Table I).
A table in the Appendix provides the results of the intraoperative polymerase chain reaction, microbiologic culture, and histopathological evaluation. Table II lists the numbers of positive and negative results for each assay. Six cases were positive on methicillin-resistant Staphylococcus-specific polymerase chain reaction testing, and thirteen were positive on universal polymerase chain reaction testing; in three of these cases, both of the polymerase chain reaction assays were positive. Microbiologic culture revealed eleven positive cases: two methicillin-resistant Staphylococcus aureus infections, two methicillin-resistant Staphylococcus epidermidis infections, two infections with other methicillin-resistant Staphylococcus species, two methicillin-sensitive Staphylococcus aureus infections, two Streptococcus infections, and one infection with a coagulase-negative Staphylococcus species. Methicillin-resistant Staphylococcus-specific polymerase chain reaction was positive for four of the six samples that were positive for methicillin-resistant Staphylococcus on culture. In the other two samples (Cases 3 and 22; see Appendix), methicillin-resistant Staphylococcus-specific polymerase chain reaction did not show a positive result intraoperatively, but we confirmed acute inflammation histopathologically in one (Case 3). A postoperative polymerase chain reaction assay also showed a positive result for that sample (Case 3), but not for the other (Case 22). Samples from nine of the ten operations involving a revision as a second stage after implant removal and the use of antibiotic-containing hydroxyapatite blocks in the first stage were negative on culture, polymerase chain reaction, and histopathological evaluation.
All five cases that were positive for methicillin-sensitive Staphylococcus aureus on culture were found to be infected with gram-positive species, and all had a positive result of the universal polymerase chain reaction. The average temperature of the melting peak by universal polymerase chain reaction was 62.8°C (Cases 4, 14, 16, 17, 19, 21, 25, and 29), which is compatible with the established melting peak for gram-positive species (Fig. 2).
The differences in the average cycle number at which amplification occurred between the ultrasonication and conventional-extraction DNA samples was 2.8 cycles for the methicillin-resistant Staphylococcus-specific polymerase chain reaction and 0.5 cycle for the universal polymerase chain reaction. In the methicillin-resistant Staphylococcus-specific polymerase chain reaction in particular, we identified a delay in the cycle number at which amplification occurred.
There are several issues to consider during revision surgery for a periprosthetic infection9. The most important is whether the implant site has been truly colonized by bacteria or not. This is difficult to determine in the occasional instances when the infection is low-grade, as common microbiologic cultures are often negative; however, polymerase chain reaction assay may be positive5,10,11 or the use of other enhanced procedures such as ultrasonication may reveal a positive result for infection2,12,13. Tunney et al. reported that the prevalence of infection around prosthetic joints might be considerably underestimated3,10. They applied an ultrasonication method to the dislodgment of biofilm around total joint implants, then used a 16S rRNA universal polymerase chain reaction assay for the detection of bacterial DNA, and showed a far higher detection rate of occult infection that was not detectable with conventional culture.
Another issue that arises during a two-stage revision arthroplasty is how to properly confirm the resolution of an infection. There are few available tests for this, and histopathological evaluation is currently the most reliable7,8,14-17. The specificity of frozen-section analysis is high in general for the testing of bacterial infection, but the sensitivity of this method is still open to question15,18. To obtain a higher specificity rather than sensitivity in this regard, we adopted detection of ten or more neutrophils per high-power field8 as the definition of acute inflammation and thus an indication of infection. It has proved difficult to establish a single gold-standard test for periprosthetic infection. In this study, we made the diagnosis of infection if at least one test of microbiologic culture or histopathological evaluation showed a positive result.
In the current study, we employed a real-time polymerase chain reaction assay combined with a simple ultrasonication method to release DNA from tissues for the identification of bacterial infection during revision arthroplasty procedures. We applied this method to actual clinical cases to facilitate treatment. For instance, an informed selection of a two-stage revision can be made after confirmation of polymerase chain reaction-positive results, and antibiotics for methicillin-resistant Staphylococcus aureus can be administered immediately when methicillin-resistant Staphylococcus-specific polymerase chain reaction shows a positive result.
The most critical issue regarding our proposed DNA-extraction method is the simple DNA-release step performed with ultrasonication. Several previous studies have demonstrated the ability of ultrasonication to improve the sensitivity of detection of bacteria in regions around implants2,3,12,13. Other studies have also shown the positive effects of ultrasonication in terms of rapid and simple DNA release from bacteria19-22, and we applied this principle directly to the tissue samples that we obtained from the surgical sites during revision arthroplasties. We added washing and purification steps to the sonicate solutions with use of a standard DNA column. Although the detection sensitivity of the polymerase chain reaction assay was reduced compared with that for DNA sampled with use of conventional extraction methods, the rapidity of the ultrasonication method (about eighty minutes from sampling to the final result) potentially allows this method to be performed intraoperatively, which is not currently possible with standard techniques. The breakdown of this eighty-minute time period is five minutes for transportation of the sample, five minutes for ultrasonication, ten minutes for purification, and sixty minutes for real-time polymerase chain reaction. This length of time makes intraoperative diagnosis possible with use of this method in most surgical revisions.
A unique feature of the ultrasonication-based method is the possibility of performing simultaneous detection of methicillin-resistant Staphylococcus and pan-bacterial DNA with use of the same amplification conditions. By determining the melting peak in the universal polymerase chain reaction assay, we can also determine whether an infection with gram-positive bacteria is present. Although we confirmed a difference in the melting peak temperatures for only eight bacterial strains, a previous report demonstrated this difference between gram-negative and gram-positive species of most organisms encountered in musculoskeletal infections23. This characteristic is useful for excluding nonspecific amplification when the cycle number at which specific amplification occurs is relatively late—i.e., when relatively low quantities of bacteria are present.
One limitation of polymerase chain reaction is the inability to determine bacterial viability in polymerase chain reaction-positive but culture-negative cases. Birmingham et al.4 reported the possibility of using reverse transcription-polymerase chain reaction to detect live bacteria and thus reduce the probability of false-positive results. Although use of real-time reverse transcription-polymerase chain reaction with our current method could potentially determine bacterial viability, it would be quite difficult to perform in a clinical setting that requires rapid identification. Kobayashi et al.24 investigated the use of propidium monoazide to differentiate viable bacteria from dead bacteria. This approach may be relatively easier for clinical application because the polymerase chain reaction assay and the DNA preparation are basically the same as those used with the current method. Thus, it is necessary to consider the viability of the bacteria when a polymerase chain reaction-based assay is used for the diagnosis of infection. In the current study, seven cases showed polymerase chain reaction-positive but culture-negative results. We confirmed the presence of acute inflammation with histopathological evaluation in four of these cases (8, 18, 23, and 26; see Appendix), indicating the likelihood of infection. In the remaining three cases (13, 21, and 25), it was uncertain whether our results reflected the presence of a low-grade infection that was not detectable on culture or necrotic organisms. The possibility of detecting necrotic organisms with polymerase chain reaction seems to be higher after implant removal and placement of antibiotic spacers in particular. In Case 25 we opted for implant removal on the basis of a positive polymerase chain reaction result despite the negative result of the histopathological evaluation, whereas in Case 21 we selected revision even though the result of the polymerase chain reaction was positive, which might reflect the presence of necrotic organisms. Other drawbacks of this method are the inability to identify organism species and the inability to perform antibiotic sensitivity testing.
In conclusion, we simultaneously applied two different real-time polymerase chain reaction assays to determine the diagnosis of periprosthetic infection intraoperatively. This was made possible by preparing the DNA samples with a simple release method involving ultrasonication. This rapid method showed a 0.87 sensitivity and 0.8 specificity in clinical use. Additional modifications will be needed to obtain higher sensitivity and specificity and improve the rapidity of the technique.