Eight-Plex PCR and Liquid-Array Detection of Bacterial and Viral Pathogens in Cerebrospinal Fluid from Patients with Suspected Meningitis

Abstract

We here report on the development of a novel multiplex PCR with product detection in a Luminex 100 suspension array system. The assay covers the nine most important bacterial and viral pathogens found in Danish meningitis patients. The microorganisms include Neisseria meningitidis, Streptococcus pneumoniae, Escherichia coli, Staphylococcus aureus, Listeria monocytogenes, Streptococcus agalactiae, herpes simplex virus types 1 and 2, and varicella-zoster virus. The study was based on 1,187 samples, of which 55 were found to be positive by PCR. The assay was found to have an excellent sensitivity and an excellent specificity compared to the results of a “gold standard,” defined by routine laboratory tests, for the two most important pathogens, S. pneumoniae (95 and 99.1%, respectively) and N. meningitidis (100 and 99.7%, respectively). The method provides a valuable supplement to the traditional microscopy and culture of cerebrospinal fluid (CSF) samples in a routine diagnostic setting, and results can be available within 1 workday. The method is suitable for use for the initial screening and identification of nine important microorganisms in CSF samples from patients with suspected meningitis. Compared to microscopy and culture of CSF, this rapid and sensitive method will support physicians with the selection of the appropriate antimicrobial agents and the initiation of timely treatment in the absence of live microorganisms in the CSF.

Bacterial meningitis is an infectious disease with a high rate of mortality and many cases of severe sequelae (4, 5, 23). Since the early administration of antibiotics correlates with reduced rates of morbidity and mortality, it is of crucial importance to initiate relevant treatment as soon as possible (10, 11, 13, 19, 22). The administration of antibiotics to patients with suspected meningitis before hospital admission and the collection of a cerebrospinal fluid (CSF) sample have become common practice (19). This practice may correspondingly impair microbiological diagnosis by culture of the bacteria responsible for the infection. The focus in recent years has therefore been on the development of alternative methods that can be used to obtain an etiological diagnosis, e.g., PCR. The use of broad-range 16S rRNA gene PCR with subsequent sequencing has been reported (1, 20, 24). This approach can be difficult to handle in a routine laboratory due to the presence of interfering bacterial or viral DNA, which may easily be added during sample handling, or, importantly, due to the presence of traces of bacterial DNA in the enzymes and mixtures used for PCR. The method also lacks speed because of the requisite sequencing step. Alternatively, species-specific real-time PCR (3, 21) has the advantage that rapid and high throughput is possible, but due to the limited number of detection channels of most real-time instruments, only a very limited number of different targets can be analyzed at a time.

We chose to develop a system that uses a species-specific multiplex PCR focused on the bacterial and viral microorganisms most frequently found in CSF and that subsequently detects the PCR products on the Luminex 100 detection system, which is able to detect up to 100 different targets at a time. The leading causes of bacterial meningitis in Denmark are Streptococcus pneumoniae and Neisseria meningitidis, which accounted for nearly 70% of all registered cases of bacterial meningitis in Denmark in 2002 and 2003 (50.5% and 18.2%, respectively) (12). An assay covering N. meningitidis, S. pneumoniae, Escherichia coli, Staphylococcus aureus, Listeria monocytogenes, Streptococcus agalactiae (group B streptococci), herpes simplex virus (HSV) type 1 (HSV-1), HSV-2, and varicella-zoster virus (VZV) will cover the etiology of 86.9% (185/213) of reported cases of bacterial meningitis (12). The development and validation of the use of an eight-plex PCR assay targeting these species are described in this article.

MATERIALS AND METHODS

CSF samples.

From 11 November 2004 to 11 November 2005, 1,607 CSF samples were sent to the Department of Clinical Microbiology, Aarhus University Hospital, Skejby, Denmark. Of these, 1,031 samples sent for bacterial analysis and 156 samples sent for viral analysis (in total, 1,187 samples) were included in this study. Four hundred twenty samples were excluded because they were (i) doublet samples from the same patient, in which one sample was sent for bacterial analysis and one sample was sent for viral analysis on the same day (n = 128); (ii) samples sent from the forensic medical department (n = 54); (iii) samples with insufficient volumes (n = 70); or (iv) samples that were not collected for this project (n = 168).

Handling of samples.

The samples included were received marked for analysis for either bacterial microorganisms (bacteriological specimens) or viral microorganisms (viral specimens). The bacteriological specimens were routinely analyzed by CSF microscopy and culture of the precipitate obtained by centrifugation. PCR-Luminex analysis was performed with the supernatant. As part of the “gold standard” definition, blood cultures positive ±7 days from the time of sampling of the CSF were also registered.

Viral specimens were analyzed for HSV-1 and -2 and for VZV by an in-house real-time PCR (9, 14) and by PCR-Luminex analysis; both methods used uncentrifuged specimens. The bacteriological specimens were also analyzed for HSV and VZV by use of the supernatant. The viral specimens were not analyzed for bacteria since culture and microscopy were not routinely performed for these samples submitted for viral examination only.

Viral and bacterial strains.

Six bacterial and three viral reference strains were used in the study for establishment of the assay, including N. meningitidis serogroup B strain ATCC 13090, S. pneumoniae strain ATCC 6301, S. aureus strain ATCC 27217, E. coli strain ATCC 25922, S. agalactiae strain ATCC 12927, L. monocytogenes strain VDL 148 (National Veterinary Institute, Denmark), HSV-1 strain McIntyre-B, HSV-2 strain MS, and the VZV QCMD test strain. Additional strains were analyzed to ensure the positive detection of other reference strains and serogroups, including N. meningitidis serogroup A strain HF96; N. meningitidis serogroup C strain ATCC 13102; N. meningitidis serogroup E strains HF44 and HF113; N. meningitidis serogroup W135 strain HF136; N. meningitidis serogroup X strains HF42 and HF29; N. meningitidis serogroup Y strain HF13; N. meningitidis serogroup Z strains HF18 and HF37 (6); S. aureus strains ATCC 25178, ATCC 25923, ATCC 29213 (methicillin susceptible), and ATCC 43300 (methicillin resistant); S. pneumoniae strain ATCC 49619; S. pneumoniae strains with intermediate resistance to penicillin (strains I-460, I-533, I-602, I-761, and I-896), and 14 clinical isolates of S. pneumoniae (of serogroups 9N, 14, 9V, 6B, 7F, 19F, 4, 18C, 23F).

Primer and probe design.

Eight pairs of primers and probes were designed by using the Primer3 web program (17). The GenBank reference sequence GIs applied for the primer and probe design were GI 21902494 for N. meningitidis, GI 153693 for S. pneumoniae, GI 46736 for S. aureus, GI 304909 for E. coli, GI 124495038 for L. monocytogenes, GI 840865 for S. agalactiae, GI 59860 for HSV-1 and -2, and GI 30575486 for VZV. All primers and probes were manufactured by MWG-Biotech AG (Ebersberg, Germany), and are shown in Table 1.

TABLE 1.

Primers and probes designed for use in multiplex PCR with Luminex analysis

Species	Oligonucleotide namea	Oligonucleotide sequence (5′-3′)	Target gene	Positionb
N. meningitidis	Nme-s	GTGATGGTGCGTTTGGTGCAGAATA-biotin	ctrA	504-528
	Nme-as	CACATTTGCCGTTGAACCACCTACC		644-620
	Nme-Pas	C12-CAACACACGCTCACCGGCTGCCGTCAGCGGCATAC		605-571
S. pneumoniae	Spn-s	CGCAATCTAGCAGATGAAGCAGGTT-biotin	lytA	911-935
	Spn-as	AAGGGTCAACGTGGTCTGAGTGGTT		1034-1010
	Spn-Pas	C12-ACTCGTGCGTTTTAATTCCAGCTAAACTCCCTGTA		986-952
S. aureus	Sau-s	AGCTGCACCCATGCCGACAC	coa	2681-2700
	Sau-as	GAATGTGAATGGTGGCGCTATTGCT-biotin		2831-2807
	Sau-Ps	C12-CAATACACATCGTAACCATGCCGTAACGGCTAT		2701-2733
E. coli	Eco-s	CGTGGTGGTCGCTTTTACCACAGAT	metH	133-157
	Eco-as	TCCACTTTGCTGCTCACACTTGCTC-biotin		239-215
	Eco-Ps	C12-GCGTTTATGCCAGTATGGTTTGTTGAATTTTTATT		158-192
L. monocytogenes	Lmo-s	TGTAAACTTCGGCGCAATCAGTGAA	hly	712-736
	Lmo-as	GCTTTGCCGAAAAATCTGGAAGGTC-biotin		831-807
	Lmo-Ps	C12-GGGAAAATGCAAGAAGAAGTCATTAGTTTTAAACA		737-771
S. agalactiae	Sag-s	GCCCAGCAAATGGCTCAAAAGC	cfb	449-470
	Sag-as	TCAGGGTTGGCACGCAATGAA-biotin		597-577
	Sag-Ps1c	C12-TATCAAAGATAATGTTCAGGGAACAGATTATGAAAAAACGG		499-535
	Sag-Ps2	C12-TATCAAAGATAATGTTCAGGGAACAGATTATGAAAAACCGG		499-535
HSV-1 and HSV-2	HSV-s	GGTGTTCGACTTTGCCAGCCTGTAC	DP	2482-2506
	HSV-as	CCACCTCGATCTCCAGGTAGTCC-biotin		2612-2590
	HSV-Ps	C12-CCCAGCATCATCCAGGCCCACAACCTGTGCTTCAG		2507-2541
VZVd	VZV-s	GATGGTGCATACAGAGAACATTCC-biotin	DP	742-765
	VZV-as	CCGTTAAATGAGGCGTGACTAA		880-859
	VZV-Pas	C12-AAAGTTCCGCGCTGCAGGTTCCAGTAATGCTCTA		858-825

^aS sense; as, antisense; Ps, sense probe; Pas, antisense probe.

^bNucleic acid positions on reference genomes (see Materials and Methods).

^cThe difference between Sag-Ps1 and Ps2 is A/C at base pair position 38 (boldface and underlined).

^dThe primers were adapted from Kimura et al. (9).

DNA extraction.

DNA was extracted by using a MagNAPure compact automatic system and total nucleic acid isolation kit I (Roche, Penzberg, Germany), according to the manufacturer's recommendations. The Total_NA_Plasma_100_V (version MPC3.1) purification protocol was used. The starting volume varied from 100 to 400 μl of CSF or CSF supernatant, and DNA was eluted in 50 μl elution buffer and was subsequently stored at −20°C.

PCR amplification.

The multiplex PCR amplification was performed in a total volume of 10 μl containing the following: 4 μl template, 200 nM of each primer, and 1× multiplex master mix (Qiagen, Ballerup, Denmark). A touchdown PCR (i.e., with the annealing temperature reduced 1/2°C for every new cycle, starting with 64.5°C and ending with 58°C) was performed on an MWG thermocycler (Aviso GmbH, Greiz, Germany) under the following conditions: 95°C for 15 min; 14 cycles of 94°C for 30 s, 65°C −1/2°C for 3 min, and 60°C for 30 s; 36 cycles of 94°C for 10 s, 58°C for 30 s, and 60°C for 30 s; and 1 cycle of 72°C for 5 min.

Principle of suspension array and microsphere coupling.

The principle of the Luminex suspension array system is available at the Luminex Corporation website. In brief, species-specific PCR products labeled with biotin are mixed with microspheres that have been coupled to gene-specific probes. PCR products are subsequently marked with streptavidin-R-phycoerythrin (SAPE), which binds to the biotin on the PCR products. Each probe-microsphere set captures its specific PCR product if the product is present in the sample. Subsequently, each microsphere is analyzed with a red laser that identifies the color of the microsphere and a green laser that analyzes the surface content of the SAPE bound to the hybridized PCR products. Detection tests are run in a microtiter plate, and the results are available within 2 h.

The coupling of eight sets of probes to Luminex xMAP microspheres was performed according to the manufacturer's protocol. The concentrations of microspheres were determined with a Bürker-Türk hemocytometer.

PCR product detection and identification.

Five microliters of the PCR product was analyzed in a 50-μl hybridization assay mixture containing 3 M tetramethyl ammonium chloride (Sigma, Denmark), 50 mM Tris-HCl (pH 8.0), 4 mM EDTA (pH 8.0), 0.1% Sarkosyl (Sigma), and 5,000 of each of the eight probe-coupled microspheres. The PCR products were denatured for 5 min at 95°C and hybridized to the microspheres for 15 min at 55°C. A 25-μl hybridization solution containing 300 ng SAPE (Molecular Probes, Invitrogen, Taastrup, Denmark) was added to the hybridization assay mixture, and the mixture was subsequently incubated at 55°C for another 5 min on the Luminex 100 analyzer (Ramcon A/S, Birkerød, Denmark). The reactions were analyzed with Luminex 100 IS software (version 2.1.26). One hundred microspheres were analyzed per sample for each species except S. agalactiae, for which 200 microspheres were analyzed per specimen to improve the signal-to-noise ratio.

Gold standard definition.

The bacteriological specimens were defined as true positive if they were positive by CSF microscopy and/or CSF culture. Likewise, specimens were defined as true positive if the PCR-Luminex analysis detected the same microorganism found in a positive culture of blood from the patient ±7 days from the time of CSF sampling. A standard clinical microbiological interpretation of the same pathogen found in two samples independently taken from the same patient within a short period of time strongly suggests a link. The viral specimens were defined as true positive if the in-house real-time PCR analyses were positive.

Determination of cutoff values.

The cutoff values for the eight sub-PCRs were based on the mean fluorescence intensities of 108 no-template controls (NTCs) run during the study period. The cutoff values chosen were five times the 95th percentile for N. meningitidis, S. pneumoniae, E. coli, S. aureus, L. monocytogenes, and VZV and two and three times the 95th percentile for HSV and S. agalactiae, respectively.

Analytical sensitivity for detection of bacterial species.

The concentration of DNA from the type or the reference strains was determined with the Quant-iT PicoGreen double-stranded DNA reagent (Molecular Probes, Invitrogen), according to the manufacturer's protocol, and a Fl600 photometer (BioTek Instruments, Inc., Winooski, VT). The limit of detection was determined with a 10-fold dilution series. The mean value of three independent experiments was calculated.

Analytical sensitivity for detection of viral species.

Analytical sensitivities for the detection of HSV and VZV were obtained by analyzing and comparing the results for 10-fold dilution series of the same sample obtained by both the quantitative in-house real-time assay and the qualitative Luminex assay (the mean values of three independent experiments).

Analytical specificity.

To test the analytical specificity of the assay, the following type and reference strains were analyzed: Streptococcus anginosus group G strain NCTC 10713, Streptococcus mitis strain NCTC 12261/SK142, S. mitis lytA⁺/lys⁺ strain SK564, S. mitis lytA⁺ strains SK609 and SK322 (both of which contained a lytA gene homologous to the S. pneumoniae lytA gene used as the target in this assay) (8), Streptococcus oralis strain NCTC 11427/SK23, Streptococcus sanguinis strain SK1778, Streptococcus salivarius strain SK2187, Streptococcus intermedius strain ATCC 27335/SK54, Enterococcus faecalis strains ATCC 29212 and ATCC 51299, Streptococcus pyogenes strain ATCC 12385, beta-hemolytic streptococcus serogroup C strain ATCC 12388, beta-hemolytic streptococci serogroup G strain ATCC 12394, Klebsiella pneumoniae strains ATCC 8045 and ATCC 13883, Staphylococcus epidermidis strain ATCC 49134, Pseudomonas aeruginosa strain ATCC 27853, Acinetobacter sp. strain BANCO 90 (Bacterium anitratum CO culture), Haemophilus influenzae strains ATCC 49247 and ATCC 49766, Bacteroides fragilis strain ATCC 25285, and a clinical isolate of Morganella morganii (isolate B3406II).

Confirmatory PCR.

For E. coli and S. aureus, our in-house routine PCR assays were applied for confirmation of positive PCR-Luminex results (data not shown).

RESULTS

Analytical sensitivity for detection of bacterial species.

The analytical sensitivity of the bacteriological assay was calculated as the minimum number of genomes detectable per ml of bacterial suspension. Analysis for N. meningitidis had the highest sensitivity, with a detection limit of 23 genomes per ml. For S. pneumoniae, S. aureus, E. coli, and L. monocytogenes, the limiting numbers were 105, 33, 58, and 26 genomes per ml, respectively. The lowest sensitivity was found for S. agalactiae, for which 5,998 genomes per ml was required for a positive result.

Analytical sensitivity for detection of viral species.

The analytical sensitivities of the HSV PCR-Luminex assay were 30 viral genomes per ml for HSV-1 and 71 viral genomes per ml for HSV-2. The limit of detection for the VZV PCR-Luminex assay was 28 genomes per ml.

Analytical specificity.

No cross-reactions with the bacterial species tested by our multiplex PCR-Luminex assay were found. Notably, the S. mitis lytA⁺ strains were negative by our test.

Clinical sensitivity, specificity, and positive predictive value (PPV) and negative predictive value.

The cutoff values chosen resulted in positive PCR results for 55 CSF specimens. Of these, 28 were positive for S. pneumoniae, 9 were positive for S. aureus, and 8 were positive for N. meningitidis. E. coli accounted for seven positive specimens and S. agalactiae accounted for two, and a single specimen infected with L. monocytogenes was found. The distribution of the test results for bacteria obtained by microscopy, culture, and PCR is shown in Table 2.

TABLE 2.

Distribution of bacterial test results for samples positive by one or more assays

Organism	Resulta					Total no. of patient samples with a positive result
	CSF microscopy	CSF culture	Blood culture	PCR-Luminex assay	Gold standard
S. pneumoniae	+	+	+	+	+	6
	+	+	−	+	+	6
	−	+	−	+	+	3
	+	−	+	+	+	1
	+	−	−	+	+	1
	−	−	+	+	+	2
	−	−	−	+	−	9
	−	+	+	−	+	1
No. of samples positive	14	16	10	28	20	29
N. meningitidis	+	+	−	+	+	4
	−	+	−	+	+	1
	−	−	−	+	−	3
No. of samples positive	4	5	0	8	5	8
S. aureus	−	+	−	+	+	2
	−	−	−	+	−	7
	−	+	+	−	+	2
No. of samples positive	0	4	2	9	4	11
S. agalactiae	−	−	+	+	+	1
	−	−	−	+	−	1
	0	0	1	2	1	2
E. coli	−	+	−	+	+	1
	−	−	+	+	+	1
	−	−	−	+	−	5
No. of samples positive	0	1	1	7	2	7
L. monocytogenes	−	+	−	+	+	1
No. of samples positive	0	1	0	1	1	1

^a+ and −, samples positive or negative by the analysis applied.

According to the gold standard, 19 of 28 S. pneumoniae PCR-positive specimens were true positives and 9 were false positives. The PCR-Luminex assay failed to identify one specimen that was positive by CSF and blood culture. For N. meningitidis, all five true-positive specimens were detected by the PCR-Luminex assay, which also detected three false-positive specimens. Two of four CSF specimens with true-positive results for S. aureus were detected by the PCR-Luminex assay, leaving two specimens with false-negative results. Seven specimens were false positive by the S. aureus subassay, and all of these were found to be negative by the confirmatory PCR. One CSF specimen with a true-positive result for S. agalactiae, as confirmed by blood culture, was found by the PCR-Luminex test, which also detected one specimen with a false-positive result. The E. coli subassay detected the two true-positive specimens plus a further five false-positive specimens, all of which were found to be negative by the confirmatory PCR. The one specimen with a positive result by the L. monocytogenes subassay was true positive, as confirmed by CSF culture. Table 3 shows the results of the viral PCR-Luminex assays. All six specimens with true-positive results were detected by the VZV subassay, and three specimens had false-positive results. The VZV-positive samples were tested by an independent VZV PCR assay in another Danish laboratory, where one of the three samples with false-positive results were also found to be positive (data not shown). The HSV subassay identified six specimens with true-positive results and one specimen with a false-positive result.

TABLE 3.

Distribution of viral test results

Virus	Resulta		Total no. of patient samples with a positive result
	Real-time PCR (gold standard)	PCR-Luminex assay
VZV	+	+	6
	−	+	3
No. of samples positive	6	9	9
HSV	+	+	6
	−	+	1
No. of samples positive	6	7	7

^a+ and −, samples positive or negative by the analysis applied.

The sensitivities, specificities, and PPVs for the PCR-Luminex subassays are shown in Table 4. The sensitivities were 100% for most assays; The exceptions were the assays for S. pneumoniae and S. aureus, for which the sensitivities were 95 and 50%, respectively. The specificities were all high and ranged from 99.1 to 100%. The PPVs varied from 22 to 86% for the assays for which several positive specimens could be included in the calculations. The numbers of specimens positive for S. agalactiae and L. monocytogenes were too small to allow for a realistic calculation of the PPVs. The negative predictive values were, in practice, 100% for all subassays.

TABLE 4.

Sensitivity, specificity, and PPV for the PCR-Luminex subassays

Organism	Sensitivity		Specificity		PPV
	%	No. of samples positive/total no. tested	%	No. of samples positive/total no. tested	%	No. of samples positive/total no. tested
S. pneumoniae	90.5	19/21	99.1	1,157/1,166	67.9	19/28
N. meningitidis	100.0	6/6	99.7	1,178/1,181	66.7	6/9
S. aureus	33.3	3/9	99.3	1,170/1,178	27.3	3/11
E. coli	(100)a	2/2	99.3	1,177/1,185	20.0	2/10
S. agalactiae	(100)	3/3	100.0	1,184/1,184	(100)	3/3
L. monocytogenes	(100)	1/1	100.0	1,186/1,186	(100)	1/1
HSV-1 and HSV-2	100.0	6/6	99.9	1,180/1,181	85.7	6/7
VZV	100.0	6/6	99.7	1,178/1,181	66.7	6/9

^aThe numbers in parentheses are based on numbers of samples too small to perform a valid calculation.

DISCUSSION

We describe the development and validation of a screening assay for meningitis consisting of an eight-plex PCR with subsequent detection by use of a Luminex 100 liquid-array system that covers 86.9% of the bacterial pathogens responsible for meningitis in Danish patients (12). Additionally, tests for HSV-1, HIV-2, and VZV were included in the assay, since these viruses can cause serious cases of encephalitis or meningitis and can be treated with antiviral compounds.

The strength of the assay is that it can test for the presence of a broad panel of both bacterial and viral agents and it can provide fast and reliable results; therefore, it can be used as a screening test. The results show that the assay has a high analytical sensitivity and specificity and that the cutoff values chosen, since we favored specificity, results in a high negative predictive value.

A total of 1,187 CSF samples collected over 1 year were examined, and we found that 55 samples were positive by PCR-Luminex assay analysis. Of these, 30 (55%) could be confirmed to have true-positive results. The significance of the positive results is strengthened by the fact that aliquots of the same sample were used for both the conventional and the molecular biological assay methods.

The main weakness of the assay is that, like all other PCR-based assays, investigators can find only what they are looking for. In 14 specimens, additional pathogens were found by culture, including Moraxella species (n = 1), nonhemolytic streptococci (n = 2), S. pyogenes (n = 1), micrococci (n = 1), S. epidermidis (n = 1), coagulase-negative staphylococci (n = 5), Candida albicans (n = 1), Candida dubliniensis (n = 1), and Clostridium baratii (n = 1). The seven micrococci and coagulase-negative staphylococci, however, were probably contaminants. Notably, no cases of Haemophilus influenzae infection were found. This microorganism was previously a common cause of bacterial meningitis, but after the vaccination of all children against H. influenzae type b was introduced in Denmark in 1993, the rate of meningitis caused by this organism has decreased by 97% (7). Therefore, a probe for H. influenza type b was not included in the assay design.

More than 90% of cases of viral meningitis are caused by enteroviruses, a large group of viruses comprising echovirus, coxsackievirus, enterovirus types 68 to 72, and poliovirus (15, 16, 18). Probes for this group of viruses were not included in the assay design, mainly because the course of disease caused by these viruses is mild compared to the course of the infections caused by the two virus species for which probes were included: HSV and VZV. Moreover, HSV and VZV can and should be targeted by antiviral therapy, which is not the case with enteroviruses. A further obstacle is that enteroviruses have RNA genomes that demand the transcription of RNA to cDNA before PCR can be performed. Thus, the speed of diagnosis would be delayed at the cost of information about the presence of a relatively benign infecting agent.

Procedures involving opened (unsealed) PCR plates are included in the assay, and we have not succeeded in applying dUTP-uracil glycosylase decontamination with the Qiagen multiplex PCR kit. This may be the reason why we observed a relatively high rate of false-positive results for E. coli and S. aureus. These two species also represented the majority of positive outliers in the NTC samples. To confirm the positive results for S. aureus and E. coli obtained by the PCR-Luminex assay, we analyzed the positive specimens by established alternative PCR tests that target different sequences in the coa and metH genes, respectively (data not shown). After this retesting, nine specimens false positive for S. aureus and five specimens false positive for E. coli compared to the results of the gold standard test were found. The false-positive results were probably due to contamination, although the common precautions against laboratory contamination were taken (2). The handling of unsealed specimens for testing by PCR and the inability to apply decontamination procedures, such as treatment with dUTP-uracil glycosylase, are major concerns regarding the assay; and more effort will be made in the future to include a dUTP-uracil glycosylase treatment step in the PCR-Luminex assay. Notably, our routine use of the assay after the study period has not revealed any contamination, which indicates that the low-throughput use of the system can be safely applied. We strongly believe that this screening assay is of value in its present form. Results may be interpreted in much the same way that the results of culture of any specimen type are interpreted, taking into account both clinical and paraclinical information to reach a true diagnosis.

The rate of false-negative results was relatively high for the S. aureus PCR-Luminex assay, which showed a sensitivity of 33% (Table 4). This may be due to the gram-positive bacterial cell membrane, which is notoriously difficult to open for DNA extraction. It was tested whether boiling of the bacteria before DNA isolation would improve the DNA isolation and thus heighten the sensitivity, but no difference was observed (data not shown). The number of true-positive results may be lower for this organism because none of the 10 culture-positive CSF specimens were detected by CSF microscopy, which may indicate that some of the S. aureus organisms found by culture could have been contaminants.

The S. agalactiae PCR-Luminex subassay showed a very low analytical sensitivity, but the one true-positive case of S. agalactiae meningitis found by culture of the blood of the patient was also detected by the PCR and not by CSF microscopy or culture. Thus, despite a suboptimal analytical sensitivity, the clinical sensitivity of the particular S. agalactiae PCR-Luminex subassay chosen improved the etiological diagnosis of meningitis. The low analytical sensitivity may have been caused by a suboptimal PCR efficiency, possibly combined with weak hybridization of the product to the probe, or it was possibly due to problems opening the gram-positive bacterial cell wall, as was seen for S. aureus. Thus, the S. aureus and the S. agalactiae PCR-Luminex subassays should be further optimized, if they are intended for routine use.

Gold standard definitions may cause false low specificities in cases in which a newly developed test is more sensitive than the test(s) used for the gold standard definition. In this study, we defined the gold standard for the diagnosis of bacterial meningitis as a positive CSF culture and/or microscopy result or a positive PCR-Luminex subassay result in combination with a positive culture result for blood obtained ±7 days from the time of CSF sampling. Correspondingly, the gold standard for the diagnosis of viral meningitis was defined as a positive in-house real-time PCR assay result. The scores for the true-positive samples were thus based on these laboratory data alone. If clinical and paraclinical information had been included, the specificities and PPVs of the different PCR subassays may have been improved. We are testing this hypothesis in an ongoing study.

Receiver operating characteristic curves for each subassay should ideally have been applied for the determination of the cutoff values. For all species except perhaps S. pneumoniae, too few positive samples were available to make a relevant receiver operating characteristic curve. The cutoff values were instead set by a practical approach on the basis of the values for the NTCs and on the basis of positive results for the samples by the gold standard assay. Over time, when more samples have been tested, these cutoff values may be subject to change.

A drawback of the evaluation of the method was that the CSF supernatant had to be used for all bacterial analyses. This may have led to an impaired sensitivity, since both bacterial and human cells are concentrated in the pellet after centrifugation. The sensitivity would presumably be enhanced if untreated CSF had been used. Here we had to accept the use of the supernatant, since it would have been unethical to compromise routine analyses.

Compared to the culture results, which were ordinarily available several days after the PCR results, 3 positive samples among 23 samples with true-positive results (13%) were missed (1 sample positive for S. pneumoniae and 2 samples positive for S. aureus). On the other hand, four extra samples (17%) with true-positive results were detected by the PCR-Luminex assay (two samples positive for S. pneumoniae, one sample positive for S. agalactiae, and one sample positive for E. coli).

In conclusion, the assay described here can be used as a valuable supplement to the traditional microscopy and culture of CSF specimens in a routine diagnostic setting. The assay is suitable for an initial screening for the most important microorganisms found in Danish patients with meningitis. Use of this rapid and sensitive method will enable physicians to start treatment with appropriate antimicrobial agents in the absence of live microorganisms in the CSF in a more timely manner than is possible by the use of microscopy and culture. In this study, we have focused on assay validation. In an ongoing study, we will include clinical and paraclinical data for the same patient group, which will broaden the basis of the gold standard definition. The conclusions for true- or false-positive samples made here may then be subject to change, and new knowledge regarding the clinical performance of the PCR-Luminex assay described here will be made available.