Advertisement

The use of broad-range bacterial PCR in the diagnosis of infectious diseases: a prospective cohort study

Open ArchivePublished:October 12, 2018DOI:https://doi.org/10.1016/j.cmi.2018.10.001

      Abstract

      Objectives

      Broad-range PCR has the potential to detect virtually any bacterial species via amplification and nucleotide sequencing of a DNA region common to all bacteria. We aimed to evaluate its usefulness and clinical relevance when applied to a wide variety of primary sterile materials.

      Methods

      A prospective study including 1370 samples (75 heart valves, 151 joint tissue samples, 230 joint aspirates, 848 whole blood samples and 66 culture-negative cerebrospinal fluid samples) were studied by using a commercial PCR system for detecting 16S rDNA (Molzym). The PCR results were compared with culture and were considered to provide added diagnostic value only if the PCR approach revealed new pathogens that were missed by culture.

      Results

      The added value of PCR was evident in 173 of 555 PCR-positive samples (0.126; 0.109–0.144 (proportion from all tested samples; 95% confidence interval)), most frequently in examinations of heart valves (0.56; 0.448–0.672) and joint tissue samples (0.219; 0.153–0.284). In contrast, the lowest rate of PCR with added value was noted for blood samples, regardless of the patient cohort they had been drawn from (nononcologic patients from intensive care: 0.065; 0.043–0.087, haematooncologic children: 0.048; 0.027–0.070). Moreover, PCR missed up to 7.1% of blood culture findings (0.071; 0.048–0.095) regarded as clinically relevant, which was the second highest failure rate after joint tissue samples (0.099; 0.052–0.147).

      Conclusions

      Broad-range PCR substantially increases detection rate of pathogens, especially from heart valves and joint samples. However, a concurrent risk of false-negative PCR results justifies the need for parallel culture.

      Keywords

      Introduction

      Over the last two decades, we have witnessed an explosion in the use of nucleic acid amplification techniques (NAATs) in microbiologic testing services driven by efforts to overcome the shortcomings of culture-based diagnostics. In principle, NAATs can detect microorganisms faster and with higher sensitivity than culture, and they do not require intact cells. These favourable features of NAATs are particularly valuable in detecting fastidious or uncultivated bacteria [
      • Relman D.A.
      • Schmidt T.M.
      • MacDermott R.P.
      • Falkow S.
      Identification of the uncultured bacillus of Whipple’s disease.
      ] or infectious agents in patients receiving antibiotic therapy [
      • Harris A.M.
      • Bramley A.M.
      • Jain S.
      • Arnold S.R.
      • Ampofo K.
      • Self W.H.
      • et al.
      Influence of antibiotics on the detection of bacteria by culture-based and culture-independent diagnostic tests in patients hospitalized with community-acquired pneumonia.
      ,
      • Rovery C.
      • Greub G.
      • Lepidi H.
      • Casalta J.P.
      • Habib G.
      • Collart F.
      • et al.
      PCR detection of bacteria on cardiac valves of patients with treated bacterial endocarditis.
      ].
      NAAT approaches can be utilized for the detection of a single or a predefined group of pathogens. Use of this pathogen-specific approach was on a steep rise in the 2000s, and many assays were designed to offer rapid and specific detection of various agents [
      • Cybulski Jr., R.J.
      • Bateman A.C.
      • Bourassa L.
      • Bryan A.
      • Beail B.
      • Matsumoto J.
      • et al.
      Clinical impact of a multiplex gastrointestinal PCR panel in patients with acute gastroenteritis.
      ,
      • Greub G.
      • Sahli R.
      • Brouillet R.
      • Jaton K.
      Ten years of R&D and full automation in molecular diagnosis.
      ,
      • Liesman R.M.
      • Strasburg A.P.
      • Heitman A.K.
      • Theel E.S.
      • Patel R.
      • Binnicker M.J.
      Evaluation of a commercial multiplex molecular panel for diagnosis of infectious meningitis and encephalitis.
      ,
      • Ramanan P.
      • Bryson A.L.
      • Binnicker M.J.
      • Pritt B.S.
      • Patel R.
      Syndromic panel-based testing in clinical microbiology.
      ]. The selectivity of species-specific reactions is ensured by targeting DNA sequences that are unique to individual agents, e.g., [
      • Diene S.M.
      • Bertelli C.
      • Pillonel T.
      • Jacquier N.
      • Croxatto A.
      • Jaton K.
      • et al.
      Comparative genomics of Neisseria meningitidis strains: new targets for molecular diagnostics.
      ].
      The alternative to the target-specific NAAT strategy is based on the recognition of a conserved DNA sequence that is common to all bacteria (e.g., the gene for the ribosomal small subunit 16S rRNA [
      • Kolbert C.P.
      • Persing D.H.
      Ribosomal DNA sequencing as a tool for identification of bacterial pathogens.
      ]). Detection and subsequent identification to the genus and/or species level is made by nucleotide sequence analysis of the amplified 16S rDNA fragment [
      • Srinivasan R.
      • Karaoz U.
      • Volegova M.
      • MacKichan J.
      • Kato-Maeda M.
      • Miller S.
      • et al.
      Use of 16S rRNA gene for identification of a broad range of clinically relevant bacterial pathogens.
      ]. This diagnostic approach, largely known as broad-range or universal PCR, is by nature applicable to sterile clinical samples where the cohabitation of more than one bacterial species during infection is very unlikely. It has been evaluated for different specimens, including heart valves [
      • Kuhn C.
      • Disque C.
      • Muhl H.
      • Orszag P.
      • Stiesch M.
      • Haverich A.
      Evaluation of commercial universal rRNA gene PCR plus sequencing tests for identification of bacteria and fungi associated with infectious endocarditis.
      ,
      • Miller R.J.
      • Chow B.
      • Pillai D.
      • Church D.
      Development and evaluation of a novel fast broad-range 16S ribosomal DNA PCR and sequencing assay for diagnosis of bacterial infective endocarditis: multi-year experience in a large Canadian healthcare zone and a literature review.
      ], cerebrospinal fluid (CSF) [
      • Welinder-Olsson C.
      • Dotevall L.
      • Hogevik H.
      • Jungnelius R.
      • Trollfors B.
      • Wahl M.
      • et al.
      Comparison of broad-range bacterial PCR and culture of cerebrospinal fluid for diagnosis of community-acquired bacterial meningitis.
      ], synovial fluid and joint tissue samples [
      • Bemer P.
      • Plouzeau C.
      • Tande D.
      • Leger J.
      • Giraudeau B.
      • Valentin A.S.
      • et al.
      Evaluation of 16S rRNA gene PCR sensitivity and specificity for diagnosis of prosthetic joint infection: a prospective multicenter cross-sectional study.
      ,
      • Fenollar F.
      • Levy P.Y.
      • Raoult D.
      Usefulness of broad-range PCR for the diagnosis of osteoarticular infections.
      ], blood [
      • Wellinghausen N.
      • Kochem A.J.
      • Disque C.
      • Muhl H.
      • Gebert S.
      • Winter J.
      • et al.
      Diagnosis of bacteremia in whole-blood samples by use of a commercial universal 16S rRNA gene-based PCR and sequence analysis.
      ] and pus [
      • Rantakokko-Jalava K.
      • Nikkari S.
      • Jalava J.
      • Eerola E.
      • Skurnik M.
      • Meurman O.
      • et al.
      Direct amplification of rRNA genes in diagnosis of bacterial infections.
      ]. In most studies, the potential of the PCR approach was demonstrated by discordant findings where simultaneous PCR positivity and culture negativity indicated the higher sensitivity of the PCR approach. However, not all studies found clear benefits of broad-range PCR [
      • Blaschke A.J.
      • Byington C.L.
      • Ampofo K.
      • Pavia A.T.
      • Heyrend C.
      • Rankin S.C.
      • et al.
      Species-specific PCR improves detection of bacterial pathogens in parapneumonic empyema compared with 16S PCR and culture.
      ,
      • Borde J.P.
      • Hacker G.A.
      • Guschl S.
      • Serr A.
      • Danner T.
      • Hubner J.
      • et al.
      Diagnosis of prosthetic joint infections using UMD-Universal Kit and the automated multiplex-PCR Unyvero i60 ITI(R) cartridge system: a pilot study.
      ,
      • Simon T.D.
      • Van Yserloo B.
      • Nelson K.
      • Gillespie D.
      • Jensen R.
      • McAllister 2nd, J.P.
      • et al.
      Use of quantitative 16S rRNA PCR to determine bacterial load does not augment conventional cerebrospinal fluid (CSF) cultures among children undergoing treatment for CSF shunt infection.
      ].
      Our aim was to evaluate the clinical utility of broad-range PCR in the context of real-life examinations of various types of clinical materials collected from primary sterile body sites. Special attention was paid to the critical interpretation of cases with only PCR positivity, based on which we were able to define optimal uses for broad-range PCR in diagnostic schemes.

      Methods

       Study population

      This prospective study was run over a period of 4 years (2013 to 2017). We examined 1370 samples from 973 patients who were hospitalized or who visited outpatient clinics at the Motol University Hospital, Prague. Cases of suspected bloodstream infections (BSI) were investigated in children from the Department of Paediatric Haematology and Oncology (DPHO) and in patients from Departments of Anaesthesiology and Intensive Care Medicine and Internal Medicine (termed together ICU); furthermore, patients from the Department of Cardiovascular Surgery were tested for infective endocarditis. Other patients included in the study had suspected joint infections or bacterial meningitis. All subjects provided written informed consent that had been approved by the institutional ethics committee.

       Laboratory methods

      Samples were collected under sterile conditions and subjected to both culture and broad-range PCR. The principle of parallel investigation was applied to all consecutive blood samples, heart valves and orthopaedic material. The only exceptions were joint aspirates with an insufficient volume and additional aliquots of perioperative joint tissue samples, which were not included in this study. CSF samples were checked by PCR only if the primary culture did not recover any agent. Blood culturing was performed in a BACTEC FX (Becton Dickinson, USA) instrument.
      Broad-range pan-bacterial (and panfungal 18S rDNA) PCR detection was carried out using the UMD-SelectNA (Molzym, Germany), which included reagents and protocol for DNA extraction, PCR itself and sequencing. One millilitre of liquid or 0.5 cm2 of the tissue samples were processed according to the manufacturer's instructions, including human cell lysis and degradation of extracellular DNA before bacterial DNA extraction on a SelectNA isolator (Molzym). An optional Add-On10 kit (Molzym), which enables an increase in the starting volume to 10 mL, was applied to all blood samples that had been collected into the K3EDTA tubes.
      SYBR Green–based qualitative PCR assays targeting the V3–V4 domains of the bacterial 16S rRNA gene were run on a CFX96 Real-Time PCR machine (Bio-Rad, USA). Approximately 450 bp long amplicons from the PCR-positive samples were purified and sequenced by an Applied Biosystems 3130xl Genetic Analyser (Applied Biosystems, USA). The sequencing results were compared to the National Center for Biotechnology Information BLAST and SepsiTest BLAST (Molzym). In cases with polymicrobial infections, identification of the individual pathogens was performed by the RipSeq Mixed web-based application (Pathogenomix, USA) [
      • Kommedal O.
      • Lekang K.
      • Langeland N.
      • Wiker H.G.
      Characterization of polybacterial clinical samples using a set of group-specific broad-range primers targeting the 16S rRNA gene followed by DNA sequencing and RipSeq analysis.
      ]. Cutoff values of ≥97 and ≥ 99% similarity were used to assign the resulting sequences to a genus and species, respectively. Duration of all laboratory steps that were necessary for obtaining the pathogen identification by the broad-range PCR approach was approximately 10 hours, and reagent costs per sample equalled about €110.

       Definition of PCR findings

      The accuracy of every PCR result that did not match the culture result was assessed in the context of other clinical and laboratory parameters. PCR positivity in a culture-negative sample was concluded to be of ‘added value’ if at least one of the following conditions was true: (a) the same organism was identified in another sample taken from the same site of infection, (b) the detected organism was well known to be the causative agent of the infection in question or (c) clinical and/or laboratory signs of infection were readily apparent, and the result reassured clinicians to commence or carry on with antibiotic therapy. All other PCR-positive samples were, in principle, of ‘no added value’. This category comprised not only the samples where the PCR-based identification was confirmed by positive culture, but also the samples with ‘not relevant’ PCR detections, i.e., when the bacteria were found by only PCR and did not reach any of the criteria (a) to (c). In addition, we introduced the special category of ‘failure’ if PCR missed any culture result that was sufficient to fulfil at least one of above criteria (a) to (c).

       Groups of microorganisms

      To allow statistical analysis of the types of microorganism detected, we grouped them by their taxonomical and/or clinical relatedness (Supplementary Table S1).

       Statistical analysis

      Logistic regression models were fit to investigate the associations between binary outcomes and sample characteristics, and the likelihoods of binary outcomes were estimated from logistic models. Confidence intervals (CI) for proportions were calculated using asymptotic normal distribution. P ≤ 0.05 was considered statistically significant. For multiple comparisons, the p value was adjusted by the Bonferroni correction according to the number of pairwise comparisons. The distributions of the PCR cycles in subgroups were compared by ANOVA. The Tukey method was implemented for multiple comparisons, and empirical distribution functions were used for the figures.

      Results

      The types of material investigated and the results of their testing are summarized in Table 1. Nearly 70% of all samples had broad-range PCR results identical to the culture results: identical agents were identified by both methods in 188 samples, and 757 samples were double negative. The remaining 425 samples were either positive by only one method (n = 353) or positive by both methods; however, in the latter case, their results differed from each other by the organisms identified (72 samples). As a result of the greater number of unique PCR-positive samples, the overall sensitivity of the broad-range PCR approach was much higher than that of the culture approach (0.816 vs. 0.437). The added value of the PCR approach was apparent in 173 of 555 PCR-positive samples. In contrast, 382 results fell into the category of no added value, including 33 failures, where some relevant organisms identified in double-positive samples were only recovered by culture. After adding 24 more culture-positive samples with false-negative PCR results and four samples where the PCR result was of added value but in which culture concurrently revealed another clinically important microorganism, we reached an overall failure rate of 4.5%.
      Table 1Positive results obtained by PCR and/or culture and breakdown by type of material
      CharacteristicHeart valveCSFJoint infectionBSI infectionTotal
      Joint tissueAspirate fluidBlood DPHOBlood ICU
      No. of samples positive by PCR and/or culture
      Samples tested75661512303724761370
      Positive samples553497135106186613
      Positive by culture and PCR11449733687260
       Agreement (full)8237662451188
       Disagreement (full or partial)32127123672
       ‘Added value’10321714
       ‘No added value’ due to:10446713580246
       Agreement with culture
      Includes samples from categories ‘agreement (full)’ and ‘disagreement (full or partial)’ where any additional finding by culture had no clinical relevance.
      8243682861210
       (Being) not relevant223371936
      Positive only by PCR443040594973295
       ‘Added value’41630411724159
       ‘No added value’ (= not relevant)32410183249136
      Positive only by culture0083212658
      Sensitivity of assays
      Sensitivity of PCR was calculated only from PCR-positive samples that did not miss any organism, divided by all samples positive by at least one method. Similarly, sensitivity of culture did not include culture-positive samples with additional or different finding by PCR.
      PCR (95% CI)0.982 (0.947–1.000)1.000 (NA)0.804 (0.725–0.883)0.933 (0.891–0.975)0.689 (0.601–0.777)0.715 (0.650–0.780)0.816 (0.785–0.846)
      Culture (95% CI)0.145 (0.052–0.239)0.059 (0.000–0.138)0.526 (0.426–0.625)0.526 (0.442–0.610)0.462 (0.367–0.557)0.468 (0.396–0.539)0.437 (0.398–0.476)
      Assessment of PCR: Added Value and Failure
      Added value42633431831173
      % from all samples (proportion); (95% CI)56.0% (0.560); (0.448–0.672)9.1% (0.091); (0.022–0.160)21.9% (0.219); (0.153–0.284)18.7% (0.187); (0.137–0.237)4.8% (0.048); (0.027–0.070)6.5% (0.065); (0.043–0.087)12.6% (0.126); (0.109–0.144)
      % from all positive samples (proportion); (95% CI)76.4% (0.764); (0.651–0.876)17.6% (0.176); (0.048–0.305)34.0% (0.340); (0.246–0.434)31.9% (0.319); (0.240–0.397)17.0% (0.170); (0.098–0.241)16.7% (0.167); (0.113–0.220)28.2% (0.282); (0.247–0.318)
      Failure0015483461
      % from all samples (proportion); (95% CI)0 (0); (NA)0 (0); (NA)9.9% (0.099); (0.052–0.147)1.7% (0.017); (0.000–0.034)2.2% (0.022); (0.007–0.036)7.1% (0.071); (0.048–0.095)4.5% (0.045); (0.034–0.055)
      BSI, bloodstream infection; CI, confidence interval; CSF, cerebrospinal fluid; DPHO, Department of Paediatric Haematology and Oncology; ICU, Departments of Anaesthesiology and Intensive Care Medicine and Internal Medicine.
      a Includes samples from categories ‘agreement (full)’ and ‘disagreement (full or partial)’ where any additional finding by culture had no clinical relevance.
      b Sensitivity of PCR was calculated only from PCR-positive samples that did not miss any organism, divided by all samples positive by at least one method. Similarly, sensitivity of culture did not include culture-positive samples with additional or different finding by PCR.
      More than one microbial species were identified in 104 samples. Culture revealed the coexistence of multiple pathogens in 37 of 61 PCR failure cases. Thus, PCR failures were largely a consequence of an inability to identify multiple organisms within a single sample.

       Types of samples and broad-range PCR

      Blood collected from DPHO children had the least likelihood of being PCR positive. This quality differed significantly even from the likelihood of PCR positivity of the blood from ICU patients (Table 2). However, both blood sample groups shared the trait that their respective PCR positivity was the least likely to be of added value. The highest likelihood of PCR positivity with added value was by far reported for heart valves. Conversely, the likelihood of culture positivity for heart valves was very low, second only to CSF. When we checked the probability of full agreement between positive PCR and culture results, the analysis revealed that the highest likelihood was for orthopaedic aspirates. The least likely agreement was identified for blood samples.
      Table 2Likelihood of PCR positivity, added value, culture positivity and agreement between PCR and culture by clinical material
      Likelihood of…Heart valveCSFJoint tissueAspirate fluidJoint samples in totalBlood DPHOBlood ICUBlood samples in total
      PCR positivity (n = 1370)0.733310.515220.58942,30.57392,30.580120.22851,20.33611,2,30.28891
      Added value of PCR (n = 1370)0.560040.090940.21854,50.18704,50.199540.04844,50.06514,50.05784
      Culture positivity (n = 1370)0.14676,70.06066,7,80.377560.33046,70.349160.15326,7,90.23746,8,90.20056
      Agreement between PCR and culture (n = 256; 4 CSF samples excluded)0.7273NA0.75510.904110,110.84430.6667110.5862100.609810
      Highest values are in bold, lowest values are italic. CSF, cerebrospinal fluid; DPHO, Department of Paediatric Haematology and Oncology; ICU, Departments of Anaesthesiology and Intensive Care Medicine and Internal Medicine.
      p values between two materials that were found statistically different are as follows: 1, <0.0001 between heart valve and blood DPHO, heart valve and blood ICU; 2, <0.0006 between blood DPHO and any other type of material; 3, <0.0001 between any joint samples and blood ICU; 4, <0.0001 between heart valve and any other type of material; 5, <0.0001 between any joint sample and any blood; 6, <0.0009 between joint tissue and any other type of material; 7, <0.003 between aspirate fluid and blood DPHO, CSF and heart valve; 8, 0.0028 between CSF and blood ICU; 9, 0.0025 between blood ICU and blood DPHO; 10, <0.0001 between aspirate fluid and blood from ICU; 11, 0.0031 between aspirate fluid and blood DPHO.

       Microorganisms detected by broad-range PCR

      The recovery frequency of individual species or groups of microorganisms is summarized in Supplementary Table S1. For each agent or group of agents that was detected by culture, we calculated the likelihood that its detection was not missed by PCR (Table 3).
      Table 3Likelihood of PCR or culture positivity and of added value per species/groups of microorganisms
      Likelihood of…Staphylococcus aureusCoNSStreptococciEnterococciEnterobacteriaceaeNFGNBOther microfloraOther environmental bacteria
      PCR positivity (culture positive samples; n = 288)0.9510.551,20.710.510.812,30.640.41,3LF
      LF indicates low frequency of culture positivity (n = 3).
      Culture positivity (PCR positive samples; n = 487)0.6350.44,5,70.374,80.4260.7440.3640.104,5,7,80.044,5,6,7,8
      Added value of PCR (PCR positive samples; n = 487)0.360.1410,130.579,100.50110.2790.110.4812,130.0810,11,12
      Highest values are in bold, lowest values are italic. Samples with multiple organisms detected and groups of organisms with far low frequency were excluded. CoNS, coagulase-negative staphylococci; NFGNB, nonfermenting Gram-negative bacilli.
      p values between two species/groups of organisms that were found statistically different are as follows: 1, <0.0001 between S. aureus and CoNS, S. aureus and enterococci, S. aureus and other microflora; 2, <0.0008 between Enterobacteriacae and CoNS; 3, <0.002 between Enterobacteriacae and other microflora; 4, ≤0.0009 between Enterobacteriacae and any other group of microorganisms (excluding S. aureus and enterococci); 5, ≤0.0015 between S. aureus and CoNS, S. aureus and other microflora, S. aureus and other environmental bacteria; 6, ≤0.0009 between enterococci and other environmental bacteria; 7, 0.001 between CoNS and other microflora, CoNS and other environmental bacteria; 8, < 0.0003 between streptococci and other microflora, streptococci and other environmental bacteria.; 9, ≤0.0015 between streptococci and Enterobacteriacae; 10, ≤0.0001 between streptococci and CoNS, streptococci and other environmental bacteria; 11, ≤0.0015 between enterococci and other environmental bacteria; 12, ≤0.0004 between other microflora and other environmental bacteria; 13, ≤0.0001 between CoNS and other microflora.
      a LF indicates low frequency of culture positivity (n = 3).

       ‘Not relevant’ PCR findings and threshold cycle

      An average threshold PCR cycle (Ct) was found to be significantly higher if PCR was the only method that detected agents that were deemed not relevant (Ct 34.07 vs. Ct 28.26; p < 0.0001). The cumulative distribution of the Ct values for all of the PCR-positive samples is depicted in Fig. 1; the Ct distributions related to individual types of material with a sufficient number of observations are shown in Supplementary Fig. S1. An analysis of the Ct values by quantiles (Supplementary Table S2) allowed the selection of the borderline Ct for results that were not relevant.
      Fig. 1
      Fig. 1Cumulative distribution of all PCR-positive samples based on their PCR threshold cycle. Added value was apparent in 173 PCR-positive samples; 382 samples were of no added value because of agreement with a positive culture (210 samples) or because of lack of clinical evidence for assigning their unique PCR positivity to be of added value (172 samples; ‘not relevant’).

      Discussion

      The main goal of this study was to demonstrate the suitability of PCR with a pan-bacterial breadth of detection for examinations of various types of primary sterile materials. Contrary to most common methodologies that explore parameters of sensitivity and specificity, we evaluated the category of added value of PCR sensu stricto, i.e., whether it can provide useful information in addition to what is provided by culture. Thus, simple PCR confirmation of positive culture findings was not scored as added value, although it had the potential to shorten the turnaround time when compared to conventional culture.
      To justify implementing a novel diagnostic method into routine, it should enable the detection of true pathogens that are invisible to culture. The risks of false-positive and false-negative PCR results are often associated with the detection of contaminating bacterial DNA and with interference from an excessive amount of human DNA, respectively [
      • Miller R.J.
      • Chow B.
      • Pillai D.
      • Church D.
      Development and evaluation of a novel fast broad-range 16S ribosomal DNA PCR and sequencing assay for diagnosis of bacterial infective endocarditis: multi-year experience in a large Canadian healthcare zone and a literature review.
      ,
      • Trung N.T.
      • Hien T.T.
      • Huyen T.T.
      • Quyen D.T.
      • Van Son T.
      • Hoan P.Q.
      • et al.
      Enrichment of bacterial DNA for the diagnosis of blood stream infections.
      ]. In this respect, it is noteworthy to stress that the PCR approach tested here selectively detects DNA from only living microorganisms, while human and extracellular bacterial DNA are enzymatically degraded.
      In line with recent studies, broad-range PCR proved to be extremely beneficial for testing heart valves [
      • Miller R.J.
      • Chow B.
      • Pillai D.
      • Church D.
      Development and evaluation of a novel fast broad-range 16S ribosomal DNA PCR and sequencing assay for diagnosis of bacterial infective endocarditis: multi-year experience in a large Canadian healthcare zone and a literature review.
      ,
      • Shrestha N.K.
      • Ledtke C.S.
      • Wang H.
      • Fraser T.G.
      • Rehm S.J.
      • Hussain S.T.
      • et al.
      Heart valve culture and sequencing to identify the infective endocarditis pathogen in surgically treated patients.
      ]. The investigations of CSF showed a similar benefit in that it never failed; however, its added value was much lower (9%) due to frequent detection of likely contaminating agents. Nevertheless, it definitely helped to resolve difficult-to-diagnose cases, including those where patients were receiving antibiotic therapy at the time of CSF collection (Table 4).
      Table 4Selected cases of unique PCR findings with added value
      Clinical materialSex, age (years)Threshold cycle (Ct)Broad-range PCR result
      Heart valveM, 3712.82Bartonella quintana
      M, 4614.54Staphylococcus lugdunensis
      M, 6818.76Abiotrophia defectiva
      M, 6720.04Streptococcus sanguinis
      CSFF, 125.69Neisseria meningitidis
      M, 3
      Same patient.
      32.03Fusobacterium necrophorum
      F, 1 month36.76Streptococcus agalactiae
      M, 733.62Streptococcus pneumoniae
      Joint aspirate fluidM, 7327.77Finegoldia magna
      F, 7321.23Streptococcus dysgalactiae
      Joint tissueM, 6234.70Parvimonas micra
      F, 8433.90Staphylococcus lugdunensis
      Blood DPHOM, 233.95Candida albicans
      F, 1134.00Enterococcus cecorum
      F, 1534.27Klebsiella pneumoniae
      Blood ICUM, 4534.03Streptococcus pyogenes
      M, 3
      Same patient.
      24.39Fusobacterium necrophorum
      M, 227.18Neisseria meningitidis
      CSF, cerebrospinal fluid; DPHO, Department of Paediatric Haematology and Oncology; ICU, Departments of Anaesthesiology and Intensive Care Medicine and Internal Medicine.
      a Same patient.
      The lowest added value for the PCR approach was found for blood samples. We intentionally tested two heterogeneous patient cohorts to challenge the breadth and interpretation of PCR. While nononcologic patients from ICUs represented a normal category that can develop BSI with conventional BSI pathogens, haematooncologic children can experience virtually any opportunistic infection in the bloodstream, and their clinical signs can be subtle [
      • Penack O.
      • Buchheidt D.
      • Christopeit M.
      • von Lilienfeld-Toal M.
      • Massenkeil G.
      • Hentrich M.
      • et al.
      Management of sepsis in neutropenic patients: guidelines from the infectious diseases working party of the German society of hematology and Oncology.
      ]. Only 4.8% (DPHO) to 6.5% (ICU) of all PCR examinations were unambiguously valuable, as they revealed causative agents of culture-negative BSI episodes. Moreover, we obtained false PCR negativity in 7.1% of the ICU blood samples and poor concordance between PCR and culture positivity. Nieman et al. [
      • Nieman A.E.
      • Savelkoul P.H.M.
      • Beishuizen A.
      • Henrich B.
      • Lamik B.
      • MacKenzie C.R.
      • et al.
      A prospective multicenter evaluation of direct molecular detection of blood stream infection from a clinical perspective.
      ] reported similar figures of additional PCR findings (6.4%) and false-negative PCR results (4.2%) using the same platform as we did in this study. Thus, irrespective of the patient group, the PCR results shared low rates of added value and the presence of failures, indicating universal limits of broad-range PCR to blood samples.
      Failures also occurred in orthopaedic material. Nevertheless, PCR investigation of joint samples resulted in much greater PCR sensitivity and, more importantly, much greater added value. The latter was true particularly in the examination of joint tissues; however, its downside was a failure rate of 9.9%, likely caused by a nonhomogeneous distribution of bacterial cells and by an unequal number of investigated samples per patient. While only a single sample was taken for PCR, multiple ones were usually sent for culture. The issue of nonhomogeneity was less of an issue for joint aspirates. Evidence of a high concordance between positive PCR and culture results shows aspirates to be a more appropriate material for culture.
      Unique PCR results were naturally challenging to interpret and to assign to either the ‘added value’ or ‘not relevant’ categories. In this respect, we recognized that PCR Ct values can simplify this decision making. When a cutoff value of 36 was applied, a minimum of 25% samples were correctly disregarded as being not relevant.
      The broad-range PCR approach worked extremely well for Staphylococcus aureus. In addition, PCR alone identified a substantial number of other microorganisms that were invisible to culture, likely as a result of concurrent antibiotic therapy, suboptimal culture conditions or lower sensitivity. Conversely, failures in PCR detection often happened in cases of suspected polymicrobial infection, which is a known limitation of the 16S rDNA Sanger-based sequencing methodology [
      • Moore M.S.
      • McCarroll M.G.
      • McCann C.D.
      • May L.
      • Younes N.
      • Jordan J.A.
      Direct screening of blood by PCR and pyrosequencing for a 16S rRNA gene target from emergency department and intensive care unit patients being evaluated for bloodstream Infection.
      ], but a great opportunity for emerging next-generation sequencing platforms [
      • Parize P.
      • Pilmis B.
      • Lanternier F.
      • Lortholary O.
      • Lecuit M.
      • Muth E.
      • et al.
      Untargeted next-generation sequencing–based first-line diagnosis of infection in immunocompromised adults: a multicentre, blinded, prospective study.
      ,
      • Peker N.
      • Couto N.
      • Sinha B.
      • Rossen J.W.
      Diagnosis of bloodstream infections from positive blood cultures and directly from blood samples: recent developments in molecular approaches.
      ].
      Besides failures, we shall underline other shortcomings of broad-range PCR we encountered during the study: (a) occasional inability to determine species identity. This happened as a result of a complete match of our sequenced 16S rRNA gene fragment with multiple reference sequences or because of lacking entry in BLAST databases that would have 100% similarity to our query. Nevertheless, this had a minimal clinical impact because identification to the species level failed mostly in bacteria that fell into groups of ‘other microbiota’ or ‘other environmental bacteria.’ (b) The workflow with a number of laboratory steps. This requires not only approximately 90 minutes' hands-on time of a trained staff; it also results in a relatively long turnaround time of 10 hours. (c) Missing information about PCR specificity because we did not test control samples from healthy population. (d) One of the criteria for added value comprised only an opinion of a treating physician, which is by nature subjective and may not be specific enough.
      In summary, we believe that a broad-range PCR approach should become a method of choice for the examination of joint aspirates and heart valves, and should be treated as auxiliary for making a definitive analysis of specimens negative by other microbiologic methods, such as CSF. For testing of blood samples, it poses a risk of false negativity, thus precluding its routine use without waiting for blood culture results.

      Acknowledgements

      The authors are grateful to M. Antuskova (Department of Medical Microbiology, 2nd Faculty of Medicine, Charles University), K. Farkasovska (Department of Medical Microbiology, Motol University Hospital) and M. Lustig (Molzym, Germany) for excellent technical support.

      Appendix A. Supplementary data

      The following is the Supplementary data to this article:

      Transparency Declaration

      Supported by the Ministry of Health of the Czech Republic (grant 15-28157A ). All authors report no conflicts of interest relevant to this article.

      References

        • Relman D.A.
        • Schmidt T.M.
        • MacDermott R.P.
        • Falkow S.
        Identification of the uncultured bacillus of Whipple’s disease.
        N Engl J Med. 1992; 327: 293-301
        • Harris A.M.
        • Bramley A.M.
        • Jain S.
        • Arnold S.R.
        • Ampofo K.
        • Self W.H.
        • et al.
        Influence of antibiotics on the detection of bacteria by culture-based and culture-independent diagnostic tests in patients hospitalized with community-acquired pneumonia.
        Open Forum Infect Dis. 2017; 4 (ofx014)
        • Rovery C.
        • Greub G.
        • Lepidi H.
        • Casalta J.P.
        • Habib G.
        • Collart F.
        • et al.
        PCR detection of bacteria on cardiac valves of patients with treated bacterial endocarditis.
        J Clin Microbiol. 2005; 43: 163-167
        • Cybulski Jr., R.J.
        • Bateman A.C.
        • Bourassa L.
        • Bryan A.
        • Beail B.
        • Matsumoto J.
        • et al.
        Clinical impact of a multiplex gastrointestinal PCR panel in patients with acute gastroenteritis.
        Clin Infect Dis. 2018; https://doi.org/10.1093/cid/ciy357
        • Greub G.
        • Sahli R.
        • Brouillet R.
        • Jaton K.
        Ten years of R&D and full automation in molecular diagnosis.
        Future Microbiol. 2016; 11: 403-425
        • Liesman R.M.
        • Strasburg A.P.
        • Heitman A.K.
        • Theel E.S.
        • Patel R.
        • Binnicker M.J.
        Evaluation of a commercial multiplex molecular panel for diagnosis of infectious meningitis and encephalitis.
        J Clin Microbiol. 2018; 56 (e01927-17)
        • Ramanan P.
        • Bryson A.L.
        • Binnicker M.J.
        • Pritt B.S.
        • Patel R.
        Syndromic panel-based testing in clinical microbiology.
        Clin Microbiol Rev. 2018; 31 (e00024-17)
        • Diene S.M.
        • Bertelli C.
        • Pillonel T.
        • Jacquier N.
        • Croxatto A.
        • Jaton K.
        • et al.
        Comparative genomics of Neisseria meningitidis strains: new targets for molecular diagnostics.
        Clin Microbiol Infect. 2016; 22 (e1–7): 568
        • Kolbert C.P.
        • Persing D.H.
        Ribosomal DNA sequencing as a tool for identification of bacterial pathogens.
        Curr Opin Microbiol. 1999; 2: 299-305
        • Srinivasan R.
        • Karaoz U.
        • Volegova M.
        • MacKichan J.
        • Kato-Maeda M.
        • Miller S.
        • et al.
        Use of 16S rRNA gene for identification of a broad range of clinically relevant bacterial pathogens.
        PLoS One. 2015; 10e0117617
        • Kuhn C.
        • Disque C.
        • Muhl H.
        • Orszag P.
        • Stiesch M.
        • Haverich A.
        Evaluation of commercial universal rRNA gene PCR plus sequencing tests for identification of bacteria and fungi associated with infectious endocarditis.
        J Clin Microbiol. 2011; 49: 2919-2923
        • Miller R.J.
        • Chow B.
        • Pillai D.
        • Church D.
        Development and evaluation of a novel fast broad-range 16S ribosomal DNA PCR and sequencing assay for diagnosis of bacterial infective endocarditis: multi-year experience in a large Canadian healthcare zone and a literature review.
        BMC Infect Dis. 2016; 16: 146
        • Welinder-Olsson C.
        • Dotevall L.
        • Hogevik H.
        • Jungnelius R.
        • Trollfors B.
        • Wahl M.
        • et al.
        Comparison of broad-range bacterial PCR and culture of cerebrospinal fluid for diagnosis of community-acquired bacterial meningitis.
        Clin Microbiol Infect. 2007; 13: 879-886
        • Bemer P.
        • Plouzeau C.
        • Tande D.
        • Leger J.
        • Giraudeau B.
        • Valentin A.S.
        • et al.
        Evaluation of 16S rRNA gene PCR sensitivity and specificity for diagnosis of prosthetic joint infection: a prospective multicenter cross-sectional study.
        J Clin Microbiol. 2014; 52: 3583-3589
        • Fenollar F.
        • Levy P.Y.
        • Raoult D.
        Usefulness of broad-range PCR for the diagnosis of osteoarticular infections.
        Curr Opin Rheumatol. 2008; 20: 463-470
        • Wellinghausen N.
        • Kochem A.J.
        • Disque C.
        • Muhl H.
        • Gebert S.
        • Winter J.
        • et al.
        Diagnosis of bacteremia in whole-blood samples by use of a commercial universal 16S rRNA gene-based PCR and sequence analysis.
        J Clin Microbiol. 2009; 47: 2759-2765
        • Rantakokko-Jalava K.
        • Nikkari S.
        • Jalava J.
        • Eerola E.
        • Skurnik M.
        • Meurman O.
        • et al.
        Direct amplification of rRNA genes in diagnosis of bacterial infections.
        J Clin Microbiol. 2000; 38: 32-39
        • Blaschke A.J.
        • Byington C.L.
        • Ampofo K.
        • Pavia A.T.
        • Heyrend C.
        • Rankin S.C.
        • et al.
        Species-specific PCR improves detection of bacterial pathogens in parapneumonic empyema compared with 16S PCR and culture.
        Pediatr Infect Dis J. 2013; 32: 302-303
        • Borde J.P.
        • Hacker G.A.
        • Guschl S.
        • Serr A.
        • Danner T.
        • Hubner J.
        • et al.
        Diagnosis of prosthetic joint infections using UMD-Universal Kit and the automated multiplex-PCR Unyvero i60 ITI(R) cartridge system: a pilot study.
        Infection. 2015; 43: 551-560
        • Simon T.D.
        • Van Yserloo B.
        • Nelson K.
        • Gillespie D.
        • Jensen R.
        • McAllister 2nd, J.P.
        • et al.
        Use of quantitative 16S rRNA PCR to determine bacterial load does not augment conventional cerebrospinal fluid (CSF) cultures among children undergoing treatment for CSF shunt infection.
        Diagn Microbiol Infect Dis. 2014; 78: 188-195
        • Kommedal O.
        • Lekang K.
        • Langeland N.
        • Wiker H.G.
        Characterization of polybacterial clinical samples using a set of group-specific broad-range primers targeting the 16S rRNA gene followed by DNA sequencing and RipSeq analysis.
        J Med Microbiol. 2011; 60: 927-936
        • Trung N.T.
        • Hien T.T.
        • Huyen T.T.
        • Quyen D.T.
        • Van Son T.
        • Hoan P.Q.
        • et al.
        Enrichment of bacterial DNA for the diagnosis of blood stream infections.
        BMC Infect Dis. 2016; 16: 235
        • Shrestha N.K.
        • Ledtke C.S.
        • Wang H.
        • Fraser T.G.
        • Rehm S.J.
        • Hussain S.T.
        • et al.
        Heart valve culture and sequencing to identify the infective endocarditis pathogen in surgically treated patients.
        Ann Thorac Surg. 2015; 99: 33-37
        • Penack O.
        • Buchheidt D.
        • Christopeit M.
        • von Lilienfeld-Toal M.
        • Massenkeil G.
        • Hentrich M.
        • et al.
        Management of sepsis in neutropenic patients: guidelines from the infectious diseases working party of the German society of hematology and Oncology.
        Ann Oncol. 2011; 22: 1019-1029
        • Nieman A.E.
        • Savelkoul P.H.M.
        • Beishuizen A.
        • Henrich B.
        • Lamik B.
        • MacKenzie C.R.
        • et al.
        A prospective multicenter evaluation of direct molecular detection of blood stream infection from a clinical perspective.
        BMC Infect Dis. 2016; 16: 314
        • Moore M.S.
        • McCarroll M.G.
        • McCann C.D.
        • May L.
        • Younes N.
        • Jordan J.A.
        Direct screening of blood by PCR and pyrosequencing for a 16S rRNA gene target from emergency department and intensive care unit patients being evaluated for bloodstream Infection.
        J Clin Microbiol. 2016; 54: 99-105
        • Parize P.
        • Pilmis B.
        • Lanternier F.
        • Lortholary O.
        • Lecuit M.
        • Muth E.
        • et al.
        Untargeted next-generation sequencing–based first-line diagnosis of infection in immunocompromised adults: a multicentre, blinded, prospective study.
        Clin Microbiol Infect. 2017; 23 (e1–6): 574
        • Peker N.
        • Couto N.
        • Sinha B.
        • Rossen J.W.
        Diagnosis of bloodstream infections from positive blood cultures and directly from blood samples: recent developments in molecular approaches.
        Clin Microbiol Infect. 2018; 24: 944-955