Correspondence: Mitja Rak, Faculty of Health Sciences, University of Primorska, Polje 42, 6310 Izola, Slovenia. Tel.: +386 5 6626467; fax: +386 5 6626480; e-mail: firstname.lastname@example.org
Diagnosis of prosthetic joint infection with culture technique can be problematic since the causative agent(s) are not possible to cultivate in all cases. Molecular methods had been evaluated in many studies but their inclusion in routine diagnostics is still controversial. The purpose of our prospective study was to compare the diagnostic accuracy of broad-range (BR)-PCR and culture technique. Intraoperative samples of periprosthetic tissue were retrieved in 67 patients undergoing revision arthroplasty. Samples were analyzed with culture technique, immunohistochemistry and BR 16S rRNA gene PCR. Bacteria in PCR-positive samples were identified using two different methods: direct sequencing of PCR products and specific TaqMan assays. In 63 cases, full concordance was found between BR-PCR and culture technique. Specific TaqMan assays failed to identify bacteria in four culture- and BR-PCR-positive cases and therefore had a lower sensitivity in comparison with BR-PCR. Molecular methods detected bacteria with the same accuracy as culture; however, identification of bacteria was inferior to culture. Further development of species-recognition techniques is required to improve identification of causative microorganisms.
The quality of life of patients with degenerative diseases or joints injuries has significantly improved in the last decades due to the successful development of joint arthroplasty. However, a small proportion of patients, about 8–14% (Norwegian Arthroplasty Register, 2010), develop joint-related complications that require surgical revision. The most common condition requiring revision is aseptic loosening followed by infection (Gallo et al., 2008; Trebse, 2012). Distinguishing between the two complications is critical because they require different treatment. However, this is not always a straightforward procedure, especially in cases of late or chronic infection, in which clinical presentation is similar to aseptic loosening (Trampuz et al., 2003; Trampuz & Zimmerli, 2005). Reliable diagnostics is necessary for successful treatment of prosthetic joint infections (PJI). Unfortunately, no such routine diagnostic method exists, so a combination of different methods is used to rule in or rule out infection. Among them, culture of periprosthetic tissue is presently the gold standard for diagnosis of PJI. A high rate of false negative results of culture raises doubts about the reliability of the culture. Antimicrobial therapy before sampling, slow and fastidious growth of microorganisms, and changes in metabolism due to growth in biofilm are among the reasons for a high rate of false negative results of culture (Trampuz et al., 2003; Dempsey et al., 2007; Gallo et al., 2008; Vandercam et al., 2008). New methods that can obviate culture-related problems have been developed to improve diagnostic gain, among which polymerase chain reaction (PCR) is the most encouraging due to its promising sensitivity and specificity.
The first case report of using PCR for diagnosis of PJI was published in 1995. The article described PCR as a method with great potential that could change the diagnostic evaluation of the failed prosthetic joints (Levine et al., 1995). Numerous studies have since been published on the same topic (Levine et al., 1995; Mariani et al., 1995, 1996; Tunney et al., 1999; Ince et al., 2004; Panousis et al., 2005; Fenollar et al., 2006; Dempsey et al., 2007; Fihman et al., 2007; Gallo et al., 2008; Kobayashi et al., 2008; Vandercam et al., 2008; De Man et al., 2009; Achermann et al., 2010; Bjerkan et al., 2012; Marin et al., 2012), but there is still debate about whether PCR should be included in routine diagnostics of PJI. Published studies present conflicting evidence about PCR sensitivity and specificity in diagnostics of PJI. Sensitivity ranges from 50% to 96%, and specificity from 74% to 100% (Panousis et al., 2005; Fenollar et al., 2006; Dempsey et al., 2007; Fihman et al., 2007; Gallo et al., 2008; Vandercam et al., 2008; De Man et al., 2009; Achermann et al., 2010).
The aim of our prospective study was to compare culture to broad-range (BR)-PCR using two different approaches for bacteria identification in PCR-positive samples: direct sequencing of PCR products and species- and genus-specific TaqMan assays.
Materials and methods
Sixty-seven patients undergoing revision operation of knee or hip joint endoprosthesis at the Orthopaedic Hospital Valdoltra, Slovenia, during January and May 2011 were included in this prospective study. The control group consisted of seven patients undergoing primary knee or hip arthroplasty in the same time period. Routine intraoperative samples of the periprosthetic tissue were retrieved for standard culture and histological analysis. The Slovenian National Medical Ethics committee approved the study protocol (approval number 40/06/11).
Definition of PJI
At our institution, PJI is defined when at least one of the following criteria is fulfilled: (1) a sinus tract communicating with the implant; (2) purulence around the joint, confirmed cytologically, histologically or macroscopically; (3) microorganisms isolated from the liquids or tissues around the joint implant or from the implant itself. For the purpose of this study the inclusion criteria were changed to be able to determine the accuracy of standard culture and new molecular methods. Our modified clinical definition (MCD) of PJI was thus considered when at least one of the following criteria was present: (1) sinus tract communicating with the prosthesis; (2) macroscopic purulence surrounding the prosthesis; and (3) positive pathohistological result of periprosthetic tissue samples.
Six tissue samples were retrieved routinely at revision operation. Samples were transported in sterile containers to the diagnostic laboratory of the Institute of Public Health Koper (Medical Microbiology Department), Slovenia, where they were immediately processed. The average time for transportation to the microbiological laboratory was 2.5 h. Samples were homogenized with a homogenizer for 90 s (Homogenisator Masticator Digital IUL Instrumnets GmbH, Königswinter, Germany) after 10 mL of liquid thioglycollate medium (TYO) was aseptically added. Aliquots of homogenized tissue (0.1 mL for agar-based medium and 1 mL for liquid medium) were inoculated on appropriate aerobic [Columbia blood agar (CA) and TYO] and anaerobic [Brucella blood agar (BCA) and cooked meat medium (CMM)] culture media. CA and TYO medium were incubated in ambient air for 2 days and 7 days, respectively, at 35 °C (CMM and BCA were incubated for 7 days at 35 °C). Aerobic and liquid media were inspected for growth each day of incubation; anaerobic media were inspected on the 2nd and 7th day of incubation. If growth was observed, identification was done with Gram staining and VITEK 2 Compact (bioMérieux, France). Culture was considered positive if the same species (with the same susceptibility pattern) was identified in at least two samples of periprosthetic tissue (Trampuz et al., 2007; Gallo et al., 2008; Bori et al., 2009; Gomez et al., 2012). In the case of polymicrobial growth, culture was considered positive if two or more species with the same susceptibility pattern grew in two or more samples.
An aliquot of homogenized periprosthetic tissue was used for molecular analysis. Separate designated work benches and pipetting devices were used for DNA extraction and preparation of the PCR mixture. All equipment was cleaned with bleach and UV-treated before use.
Isolation of DNA
Total DNA was extracted from 200 μL of sample with a PureLink™ Genomic DNA Mini Kit (Invitrogen) according to the manufacturer's instructions. Negative control of isolation (NCI) was included with every six clinical samples, for which nuclease-free water (DEPC-treated) was used in place of the clinical sample.
Detection of bacteria
The presence of bacteria was confirmed by amplification of the 16S rRNA gene with BR primers: forward primer (510-FP) 5′-CCAGCAGCCGCGGTAATA-3′ and reverse primer (1066-RP) 5′-CACGAGCTGACGACARCCAT-3′. Amplification of the 16S rRNA gene gave an expected amplicon of around 560 bp. DNA extracted from Staphylococcus aureus (ATCC 25923) was used as a positive control in each run. NCI and non-template control (NTC) were also included in each run. Additionally, amplification of the human glyceraldehyde-3-phosphate dehydrogenase (GADPH) gene with forward primer (G1) 5′-TCCCTGAGCTGAACGGGAAG-3′ and reverse primer (G2) 5′-CGCCTGCTTCACCACCTTCT-3′ (Hmeljak et al., 2010) was used as a control of DNA isolation and PCR inhibition.
All PCR mixtures were conducted in a total volume of 15 μL and consisted of 1.5 μL of isolated DNA, 0.5 μM of each primer, 7.5 μL of QantiTect® SYBR® Green PCR Master Mix kit (Qiagen) and 5.25 μL of nuclease-free water (DEPC-treated). Amplification was carried out on 7300 Real Time PCR (Applied Biosystems), with the following conditions: 50 °C for 2 min, 95 °C for 10 min and 40 cycles at 95 °C for 15 s, 55 °C for 30 s and 72 °C for 60 s. Dissociation curve analysis followed immediately after amplification and served as a control of product specificity. The PCR reaction was considered positive if the difference between the Ct value of the specimens and negative controls was > 1.
Identification of bacteria
PCR products were sequenced with the forward primer at Macrogen Europe (The Netherlands) using a standard sequencing service. Sequence similarity was searched in a public sequence database using blast software at the website of the National Center of Biotechnology Information. If a mixed electropherogram was observed, the sequences were analyzed with the RipSeq Mixed web-based application, which enables species identification in cases of polymicrobial infections (Kommedal et al., 2008, 2009).
All samples that were positive with BR primers or where their melting curve profile indicated the presence of bacterial DNA, were also identified with TaqMan chemistry in combination with species- and genus-specific assays that included the validated primer pair and a fluorogenic labeled probe. These assays were designed for the most common bacteria found in periprosthetic tissue infection: S. aureus, Staphylococcus epidermidis, Staphylococcus sp., Streptococcus sp. and Pseudomonas aeruginosa. Sequences of all the primers and probes are given in Table 1. Assays were checked for cross-reactivity with the following species: S. aureus ATCC 25923, S. epidermidis ATCC 12228, Streptococcus pneumoniae ATCC 49619 and P. aeruginosa ATCC 27853, Staphylococcus capitis (clinical isolate), Staphylococcus warneri (clinical isolate), Staphylococcus haemolyticus (clinical isolate), Staphylococcus hominis (clinical isolate), Streptococcus agalactiae (clinical isolate), Pseudomonas fluorescens (clinical isolate), Pseudomonas putida (clinical isolate) and DNA isolated from periprosthetic tissue by aseptically loosening of the prosthesis. The results of molecular methods were considered positive if the same species was identified in at least two samples of periprosthetic tissue. For polymicrobial identification, BR-PCR was considered positive if two or more species were identified in two or more samples.
Table 1. Nucleotide sequences for species identification with TaqMan chemistry
Tissue samples for immunohistochemistry were fixed in 4% buffered formalin and embedded in paraffin. Microsections were prepared from each paraffin block (Morawietz et al., 2009). Monoclonal antibody CD 15 (Dako, Denmark), diluted 1: 200, was used for staining the polymorphonuclear neutrophils (PMN; Morawietz et al., 2009). Stained cells resembling PMN morphology were counted, excluding those in the lumen of blood vessels. A patient was considered positive if at least one tissue specimen contained on average ≥ 5 PMN in 10 high power microscope fields (Mirra et al., 1982; Kobayashi et al., 2006; Nunez et al., 2007; Gallo et al., 2008; Bori et al., 2009).
Statistical measures of the performance of the postoperative diagnostic methods used in this study were calculated based on definition of PJI (Table 2). Comparison between diagnostic tests was done with McNemar test using sigmaplot 11.0 (USA).
Table 2. Evaluated culture and molecular results based on our MCD of PJI
Values in square brackets are 95% confidence intervals.
Patients included in the revision group consisted of 21 (31.3%) males and 46 (68.7%) females, with a mean age of 71 years (range 47–88 years). Revision operations were performed on the hip joint in 50 (74.6%) cases and knee joint in 17 (25.4%). The control group consisted of four (57.1%) males and three (42.9%) females, with a mean age of 67 years (range 60–68 years).
According to our MCD of PJI, culture was positive in 13 patients (Fig. 1). The most common isolated bacteria were S. aureus (five patients). In other eight cases the following bacteria were identified: S. epidermidis (two patients), Escherichia coli (two patients), Serratia marcescens (one patient), Finegoldia magna (two patients) and Candida parapsilosis (one patient). Monomicrobial infection was detected in 11 patients and polymicrobial in two patients. Culture was positive in three of 51 patients which did not fit our MCD of PJI. Staphylococcus epidermidis was isolated in all three cases and additionally S. hominis was identified in one case. Growth in culture was observed in an additional nine patients that did not fit MCD of PJI, but the isolates were considered to be contaminants (same species in only one sample). The etiology in the nine patients with only one positive culture is presented in Table 3.
Table 3. Single tissue positive specimens in patients that did not fit our modified clinical criteria of PJI
BR-PCR detected bacterial DNA in 12 patients that fit the MCD of PJI (Fig. 1). Monomicrobial infection was detected in nine cases and polymicrobial in three cases. Bacterial DNA was also detected with BR-PCR in five of the 51 patients who did not fit the MCD of PJI. In three patients, the same bacteria was identified in at least two specimens (S. epidermidis), whereas in the other two cases bacterial DNA was present only in single specimen (S. epidermidis and Micrococcus luteus).
According to the second method (species- and genus-specific assays) bacteria were identified in seven of 13 patients who fit the MCD of PJI. Monomicrobial infection was detected in six patients and polymicrobial in one. In addition, bacteria were identified in three of 51 patients who did not fit the MCD of PJI. Concordance in identification between BR-PCR and TaqMan assays was achieved in six cases. No cross-reactivity between the assays was observed.
Concordant detection of bacteria with culture and BR-PCR was achieved in 12 of 16 patients who fit the MCD of PJI. In 10 patients, both methods identified the same species. Culture identified additional species in one patient and BR-PCR also identified additional species in one patient. In addition, in all three of 51 patients who did not fit MCD of PJI, concordant detection was achieved.
In the control group, culture and PCR method results were considered negative. In two patients, coagulase-negative staphylococci (CNS) were found in one sample. In the first patient, both methods detected and identified CNS, whereas in the second patient only PCR detected and identified CNS. In both cases, results were considered to be the result of contamination.
Molecular methods have properties that, in theory, should enable higher sensitivity and specificity in comparison with culture. However, in diagnostics of PJI, the statistical measures of the performance of molecular methods are not always higher than with culture. In studies where the sensitivity of molecular methods is higher, the specificity is lower in comparison with culture (Panousis et al., 2005; Moojen et al., 2007; Kobayashi et al., 2008; Vandercam et al., 2008), or the sensitivity is lower and specificity higher in comparison with culture (Fihman et al., 2007; De Man et al., 2009; Marin et al., 2012). However, in some studies, both sensitivity and specificity are higher in molecular methods than in culture-based diagnostics (Fenollar et al., 2006; Gallo et al., 2008; Achermann et al., 2010).
It is difficult to evaluate and compare similar studies because of the differences in criteria used for the definition of PJI, the number of included subjects and the sample type. There are also important differences in the interpretation of positive results and protocol design for bacterial DNA detection. Making conclusions is thus very difficult.
In our prospective study, we compared molecular and culture methods on the same specimens of periprosthetic tissue. The results of culture and molecular methods were evaluated using the same criterion – the identification of the same species in at least two specimens. Using this criterion, the best concordance between culture and molecular methods was achieved.
Sensitivity (Table 2) of our BR-PCR method was lower than the sensitivity of culture (culture 81%, vs. PCR 75%) but there was no significant difference between the two methods (P =0.317). If we exclude the patient in whom C. parapsilosis was identified with culture, then the sensitivity of molecular methods would change from 75% to 80% and would be identical to the sensitivity of culture.
Culture and BR-PCR identified bacteria in three patients who did not fit the MCD of PJI. In all three patients, bacteria were present in at least two specimens. Therefore, it is hard to claim that these results were the result of contamination during sample processing. It is more probable that these patients were misclassified as aseptic due to criteria selection for our definition of PJI. Indeed, in two of the three patients in whom CNS were identified, PMN were present but did not meet our study cut-off value. This is in agreement with observation that the histology may be false negative in low virulence microorganisms (Bori et al., 2009).
Concordance between the culture and molecular methods was very good, but it took an average of 7 days to achieve final results with microbiological diagnostics. In some cases, final results were only available after 10 days of incubation. In contrast, molecular methods required a maximum of 2 days to obtain final results and, in some cases, the results were known on the same day of sample retrieval. The molecular methods shortened the time needed for an etiologic diagnosis, which is particularly important during treatment of patients with a potential PJI. Until the causative agent or agents are known, treatment of a failed total joint replacement is empirical and is based on combination of broad-spectrum antibiotics that act against most potential germs. The selection usually relies on the institutional guidelines and local susceptibility patterns. The considerable shorting of the time to the diagnosis provided by the molecular methods allows the suboptimal broad antibiotic targeting to be switched to more focused targeting, earlier than with culture method.
In two patients who fit the MCD of PJI, a discrepancy between the results of culture and molecular methods was observed. In both cases, culture identified pathogens, whereas BR-PCR followed by sequencing failed to detect and/or identify the causative microorganism. In the first patient, culture identified F. magna in two specimens and anaerobic Gram-positive cocci and non-sporulating anaerobic Gram-positive bacilli in four specimens, whereas molecular methods identified only F. magna in one specimen. In the remaining five positive specimens, ambiguous electropherograms hindered identification of bacteria. However, RipSeq Mixed web-based application enabled identification of two species: F. magna and Solobacterium moorei. This approach can eliminate the need to further process the PCR product with cloning or other sequence resolving techniques and thereby reduces the cost and time to final results. Furthermore, application of this tool in samples of sonicate fluid, in which bacterial diversity is greater than in periprosthetic tissue (Dempsey et al., 2007), could improve our knowledge about bacterial communities in biofilms that are present on endoprostheses (Gristina & Costerton, 1985; Tunney et al., 1999; Trampuz et al., 2003) and other medical implants. However, RipSeq Mixed web-based application can distinguish up to three species, which in some cases (Fenollar et al., 2006) may not suffice. In the second patient, S. epidermidis was identified in one specimen by both methods. However, culture also identified C. parapsilosis in the remaining five specimens, whereas the molecular method did not detect any fungal DNA because conserved regions complementary to our primers are not present in fungi.
In two patients (one who fit and one who did not fit the MCD of PJI) culture showed greater bacterial diversity than BR-PCR. The lower diversity shown by molecular methods could be due to a high mixture ratio of different template DNA, which may resulted in amplification of the dominant species in the sample.
Identification of bacteria with species- and genus-specific assay gave good results in comparison with sequencing. However, due to the lack of the specific assay, we failed to identify the causative agent in four patients that fit the MCD of PJI, with whom S. marcescens, E. coli and F. magna were identified by culture and BR-PCR followed by sequencing. Although we used a small cluster of assays, the majority of culture and BR-PCR positive cases were identified. With the introduction of more assays, species resolution could be increased.
Our study showed excellent concordance between culture and molecular methods; however, identification of pathogen organisms, in cases of polymicrobial infections, can be hindered and delayed if using standard analytic methods and software. Species-specific assays can be of great use when multiple species are present in a sample, but they still fail to identify bacteria for which assays are not designed. New tools, such as RipSeq Mixed web-based application, can improve species resolution in cases of ambiguous electropherograms and therefore obviate the need for time consuming techniques for species identification, such as cloning of PCR products. In our opinion, BR-PCR could be used in addition to culture method as a screening test to rule out bacterial infection in a much shorter time than culture.
We thank Dr Maria Eugenia Portillo from the Reference Microbiology Laboratory of Catalonia, Barcelona, for critical review of the manuscript and useful comments. This work was supported by the Slovenian Research Agency (program grant no. J3-2218). The authors declare that they have no conflict of interest.