Diagnosis by real-time polymerase chain reaction of pathogens and antimicrobial resistance genes in bone marrow transplant patients with bloodstream infections

During October 2008 to July 2011, 160 BCs from Bactec bottles were analyzed from 33 immunocompromised patients (31 adults and two children) with hematological malignancies who underwent bone marrow transplantation (seven acute myeloid leukemia, six multiple myeloma, six non-Hodgkin’s lymphoma, four acute lymphoblastic leukemia, two Hodgkin’s disease, two chronic myeloproliferative disorders, one chronic lymphoblastic leukemia, one adenovirus dystrophia, one Ewing syndrome, one pineal carcinoma and one testicular cancer). Patients were admitted to the Institute of Pediatric Oncology (GRAACC) and the Adult Hematology Unit of São Paulo Hospital, Federal University of São Paulo, Brazil. Informed consent was obtained from patients and the study was approved by the Clinical Research Ethics Committee of the hospital.

Total DNA was extracted from the positive or negative 500-μl BC broth after incubation on Bactec 9240, using the phenol–chloroform method (Brazol; LGC Biotechnologia, Cotia, Brazil) with 300 mg glass beads (0.3 mm diameter; Scientific Industries, Bohemia, NY, USA) and processed in a disruptor Genie (Scientific Industries) for 10 min to achieve cell lysis.

The primers used for detection of the resistance genes, blaSHV, blaTEM, blaCTX-M, blaIMP, blaSPM, blaVIM, blaKPC, vanA, vanB and mecA, have been described previously [6]. The TaqMan probes for multiplex real-time polymerase chain reaction (PCR) detection of Gram-positive and Gram-negative bacteria have been previously described, as well as species-specific TaqMan probes for detection of Enterococcus spp., CoNS, S. aureus, Streptococcus pneumoniae, Escherichia coli, Klebsiella pneumoniae, Enterobacter cloacae, Proteus mirabilis, Salmonella spp., Serratia marcescens, Acinetobacter baumannii, P. aeruginosa, Stenotrophomonas maltophilia, Mycobacterium tuberculosis, Aspergillus spp., Candida spp. and Fusarium spp. [7]. All the primers and probes were selected from the National Center for Biotechnology Information website (NCBI; http://​www.​ncbi.​nlm.​nih.​gov) and synthesized by Applied Biosystems (Foster City, CA, USA). The primers were checked for specificity in a BLAST search available through the NCBI website (http://​blast.​ncbi.​nlm.​nih.​gov/​Blast.​cgi). A primer set for the hemochromatosis gene was designed to be used as an internal control.

Detection of bacterial DNA was screened using universal primers of 16S rDNA gene. Differentiation between Gram-positive and Gram-negative bacteria was done by TaqMan multiplex real-time PCR [7]. Differentiation of the other genes was performed by monoplex real-time PCR. Amplification for TaqMan probes reactions was performed in a 20-μl reaction volume, using 10 μl TaqMan Universal Master Mix 2X (Applied Biosystems), 5 μl template DNA, 0.5 μM each primer, and 0.3 μM probe. The real-time PCR conditions were as follows: 50°C for 2 min and 95°C for 10 min; 40 cycles of 95°C for 15 s and 60°C for 60 s. Resistance genes amplification by SYBR Green monoplex real-time PCR was performed in a 25-μl reaction mixture containing 12.5 μl Platinum SYBR Green qPCR SuperMix (Invitrogen, Carlsbad, CA, USA), 0.5 μM each primer, and 5 μl purified DNA extracted from samples. The real-time PCR conditions were as follows: 50°C for 2 min and 95°C for 10 min; 40 cycles of 95°C for 15 s and 60°C for 60 s; and a melting curve step (from 68°C to 95°C in increments of 0.5°C/s). The metallo-β-lactamase amplification was performed by multiplex real-time PCR [7]. The ABI 7500 real-time PCR System (Applied Biosystems) instrument was quantified online and at the endpoint with the sequence detection system software (version 2.0; Applied Biosystems).

Data collection and statistical analysis were performed using SSPS version 17.0 software (SPSS Inc., Chicago, IL, USA) and Microsoft Office Excel 2007 (Microsoft, Redmond, WA, USA). The area under the receiver operating characteristic curve (AUC) was determined for binary diagnostic test results for all genes. Comparison of the BC identification by Phoenix versus real-time PCR results were evaluated by χ ² tests. Discordance between the results from the two methods was assessed using McNemar’s test (statistical significance was set at a two-tailed exact test, based on the binomial distribution with p = q = 0.5). The κ statistic was calculated to measure the level of agreement between BC by Phoenix and real-time PCR results.

A total of 160 blood samples collected from 33 patients were evaluated for BSI. Thirty-three samples, representing 15 febrile episodes, tested positive by BC. BC identified 23 Gram-positive bacteria (15 CoNS, four Enterococcus faecium, two Enterococcus faecalis and two Str. pneumoniae) and 10 Gram-negative bacteria [one Pseudomonas putida, two Acinetobacter spp., three Ent. cloacae, one K. pneumoniae, one E. coli, one non-fermenting Gram-negative bacilli (NFGNB) and one Citrobacter freundii].

Real-time PCR detected 37 positive samples (Table 1). The data for negative samples by the two methods are not shown. Twenty-three of 33 samples identified by phenotypic testing were concordantly positive by BC and real-time PCR (Table 1). Twenty-one samples were discordant by both methods. Nine BC-positive samples were negative by real-time PCR and 12 BC-negative samples were positive by PCR.

Table 1 shows only positive samples that were isolated by BC and/or real-time PCR. The other 21 negative results are not shown.

For Gram-positive bacteria, 15 CoNS were positive by BC and 13 of these were detected by real-time PCR. Three CoNS were detected by real-time PCR but did not grow in culture. Three of six BCs that were positive for Enterococcus spp. were negative by real-time PCR, while whereas BC did not detect one sample that was positive by real-time PCR. In two samples, BC identified Str. Pneumoniae, while real-time PCR was not positive for one of these samples.

For Gram-negative bacteria, all Ent. cloacae and E. coli samples were concordant by both methods (Table 1). In three samples, growth of C. freundii, Acinetobacter spp. and P. putida was detected by BC and by Gram-negative probe; however, specific primers and probes were not designed for these pathogens, therefore they were not considered discrepant. Two samples were positive for K. pneumoniae by real-time PCR and negative by BC, whereas one BC-positive sample was negative by real-time PCR. Three samples were identified as A. baumannii by real-time PCR and one of these was positive by BC, one was BC-positive for NFGNB, and one was not detected by BC.

The overall concordance level was 72.7% for phenotypic and real-time PCR methods with a Cohen κ coefficient of 0.7 (95% CI: 0.61–0.80).

For resistance genes, real-time PCR detected 12 CoNS with mecA genes. Nine samples were concordant for detection of the mecA gene compared with phenotypic resistance to methicillin. The vanA gene was detected in four Enterococcus spp. (Enter. faecalis and Enter. faecium) and none were positive for the vanB gene. The concordance was 100% between two vancomycin resistance methods (Table 1).

For Gram-negative bacteria, five samples detected the ESβL gene by real-time PCR. The blaSHV gene was detected in three samples and blaCTX-M in two (Table 1). Only two of the five positive ESβL real-time PCR samples were identified by the phenotypic method as ESβL producers; the other two were identified as ESβL producers by the phenotypic method but were negative by real-time PCR (Table 1).

The blaKPC and metallo-β-lactamase genes were not detected. The A. baumannii samples showed carbapenem phenotypic resistance; however, the carbapenemase genes were not detected.

Three Candida spp. were detected from two patients by real-time PCR but not by BC. In one patient, two samples were positive by real-time PCR. Two samples were positive for Aspergillus spp. by real-time PCR with negative galactomannan antigenemia. Fusarium spp. were not detected.

Real-time PCR performance for bacterial identification was adjusted as follows: sensitivity, 78%; specificity, 93%; negative predictive value (NPV), 95%; positive predictive value (PPV), 72% when compared with the phenotypic method.