Prospective Study of Use of PCR Amplification and Sequencing of 16S Ribosomal DNA from Cerebrospinal Fluid for Diagnosis of Bacterial Meningitis in a Clinical Setting

Abstract

We have evaluated the use of a broad-range PCR aimed at the 16S rRNA gene in detecting bacterial meningitis in a clinical setting. To achieve a uniform DNA extraction procedure for both gram-positive and gram-negative organisms, a combination of physical disruption (bead beating) and a silica-guanidiniumthiocyanate procedure was used for nucleic acid preparation. To diminish the risk of contamination as much as possible, we chose to amplify almost the entire 16S rRNA gene. The analytical sensitivity of the assay was approximately 1 × 10² to 2 × 10² CFU/ml of cerebrospinal fluid (CSF) for both gram-negative and gram-positive bacteria. In a prospective study of 227 CSF samples, broad-range PCR proved to be superior to conventional methods in detecting bacterial meningitis when antimicrobial therapy had already started. Overall, our assay showed a sensitivity of 86%, a specificity of 97%, a positive predictive value of 80%, and a negative predictive value of 98% compared to culture. We are currently adapting the standard procedures in our laboratory for detecting bacterial meningitis; broad-range 16S ribosomal DNA PCR detection is indicated when antimicrobial therapy has already started at time of lumbar puncture or when cultures remain negative, although the suspicion of bacterial meningitis remains.

Bacterial meningitis is a serious disease with high morbidity and mortality. To reduce death or permanent neurological sequelae as much as possible, a fast and correct diagnosis is of the utmost importance. The current standard for the diagnosis of bacterial meningitis is microscopic examination and subsequent culture of cerebrospinal fluid (CSF). However, this approach might have some disadvantages with regard to the desired rapidity and sensitivity.

Results of culture may only be available after 24 to 48 h and sometimes, for instance, when the number of viable organisms in the CSF is low, it may take even longer. Moreover, the sensitivity of microscopic examination and culture of CSF can be debated.

First, bacterial concentration in the CSF has a profound effect on the results of microscopy. Regardless of the type of organism in the CSF, the percentage of positive microscopic results is only 25% with <10³ CFU/ml and 60% in the range of 10³ to 10⁵ CFU/ml (32). Second, in an extensive study over a period of 27 years, it appeared that culture might miss the diagnosis of bacterial meningitis in at least 13% of cases (16, 55). Acknowledged reasons for this lack in sensitivity are CSF obtained after the start of antibiotic treatment and meningitis due to fastidious or slow-growing microorganisms.

Recently, PCR-based assays have become available to provide an early and accurate diagnosis of bacterial meningitis (2, 10, 12, 14, 22, 30, 33, 35, 39, 42, 45, 46, 51, 54, 60).

Some of these assays are aimed at specific pathogens of bacterial meningitis, such as Neisseria meningitidis, Streptococcus pneumoniae, and Haemophilus influenzae (10, 12, 46, 51, 54), whereas others use broad-range bacterial PCR (2, 14, 22, 30, 33, 35, 39, 42, 45, 60). The use of broad-range bacterial PCR has a great advantage in that it also detects microorganisms that are found less frequently or even unknown causative agents of bacterial origin. Even though broad-range PCR has been available since the early 1990s (5, 58), most clinical laboratories have not yet implemented this technique in their daily clinical practice because its use is hampered by two major problems.

First, this method is more vulnerable to contamination than a species-specific PCR. Taq DNA polymerases are frequently reported as a major source of contaminating bacterial DNA (6, 13, 27, 40, 47). Several approaches overcoming this problem have been reported with success in agarose gel electrophoresis detection systems, such as UV irradiation, 8-MOP treatment, DNase treatment, or restriction endonuclease treatment (9, 21, 25, 26, 28, 44, 52). However, it has been clearly demonstrated that most decontamination procedures also affect the sensitivity of the broad-range PCR when a sensitive detection system, such as TaqMan, is used (13). Amplicon size might have an effect on contamination, since most contaminating DNA derives from nonviable organisms. This DNA is most probably unprotected and therefore susceptible to DNA degradation. This could implicate that bigger amplicons are less sensitive for this kind of contamination than are small amplicons. In fact, for cytomegalovirus DNA it has been demonstrated that amplicon size does have a great effect on final PCR results (4). This would implicate that amplifying the entire 16S rRNA gene would be less contamination sensitive compared to a smaller part of the gene.

Second, at this time no DNA extraction protocol is available that has the same effectiveness for both gram-positive and gram-negative bacteria. Most studies use enzyme treatment (e.g., proteinase K or lysozyme), boiling, phenol-chloroform extraction, ethanol precipitation, commercial kits, or some combination of these methods (12, 14, 30, 33, 42, 45). Bead beating in a guanidinium thiocyanate buffer, followed by organic extraction, has been shown to be effective for extracting nucleic acids from clinical samples containing mycobacteria (29). However, organic extraction procedures that use phenol and chloroform are a hazard for laboratory staff because of the toxic nature of both compounds. Most organic extraction procedures are also relatively laborious and therefore not suitable for processing samples in a clinical laboratory. Recently, it was shown that physical disruption in combination with the use of a commercial extraction kit has the same quality as standard phenol-chloroform extraction (41).

To tackle these problems, we have chosen to combine a widely used silica-guanidinium thiocyanate procedure for extracting nucleic acids (3, 7, 15, 18, 20, 23, 24, 34, 38, 43, 48, 49, 50, 53, 56, 57, 61) with bead beating in order to achieve a uniform extraction procedure. Moreover, to diminish contamination problems as much as possible, primers were chosen so that almost the entire 16S rRNA gene was amplified.

Here we describe the outcome of a prospective study of the performance of our broad-range bacterial PCR with subsequent sequencing on CSF in comparison with conventional procedures in the diagnosis of bacterial meningitis. After amplification with universal bacterial primers, positive amplicons were sequenced directly, and the obtained sequences were compared to database sequences for bacterial determination.

MATERIALS AND METHODS

Bacterial strains.

The following bacterial strains were used in reconstruction experiments: Escherichia coli ATCC 25922 and Staphylococcus aureus ATCC 25923. Bacteria were grown on 5% sheep blood agar plates overnight at 37°C and harvested (∼10⁹ CFU) into 1 ml of saline.

CSF samples and patients.

Between January 2002 and May 2003, 227 CSF samples from 222 patients were collected at the decentralized units of the Regional Public Health Laboratory for Groningen and Drenthe, the Laboratory for Pathology and Medical Microbiology of the Isala Clinics in Zwolle, and the Laboratory for Medical Microbiology of the Academic Hospital of Groningen. All patients included had meningitis as part of their differential diagnosis. These patients were characterized with the following features: mean age, 24.5 years (range, 0 to 87.9 years), and 125 males (56.3%) and 97 females (43.7%). Upon arrival at the laboratory, 200 to 400 μl of each CSF sample was removed under sterile conditions and stored at −20°C until further processing for DNA extraction and PCR. The remaining CSF was used for standard bacterial culture and Gram and methylene blue staining. All included samples had sufficient volume, ensuring that both methods would produce reliable results.

Bacterial culture and direct microscopic examination.

After removal and storage of 200 to 400 μl of CSF for DNA extraction, the remaining CSF was centrifuged for 15 min at 1,500 × g. The supernatant was discarded and the sediment was resuspended in the remaining liquid by vortexing. From this suspension Gram and methylene blue stains were made, and bacterial culture was inoculated. The bacterial culture was inoculated into 5% sheep blood and chocolate agar plates and a thioglycolate enrichment broth with XV-factor and then incubated in a CO₂ incubator at 35°C.

DNA extraction from CSF.

DNA was extracted from CSF samples by a modification of the procedure described by Boom et al. (3). Briefly, 200 μl of CSF was added to 1,000 μl of lysis buffer L6 (5.25 M guanidinium thiocyanate, 50 mM Tris-HCl [pH 6.4], 20 mM EDTA, 1.3% [wt/vol] Triton X-100) and 800 mg of acid washed 0.1-mm-diameter zirconia-silica beads (BioSpec Products, Inc., Bartlesville, Okla.). This mixture was shaken in a Mini-Beadbeater-8 (BioSpec Products, Inc.) for 1 min at 3,200 rpm. After bead beating, the sample was centrifuged for 1 min at 12,000 × g. Then, 1,000 μl of supernatant was transferred to a new screw-cap reaction tube containing 20 μl of size-fractionated silica particles prepared as described previously (3) (Silica, SiO₂, S-5631; Sigma-Aldrich Chemie, Zwijndrecht, The Netherlands). After being vortexed, the sample was left at ambient temperature for 10 min. After this 10-min binding step, the silica-DNA pellets were washed and dried as described previously (3). DNA was eluted in 100 μl of TE buffer (10 mM Tris, 1 mM EDTA [pH 8.0]). If the DNA was not used immediately, it was stored at −20°C until further use.

PCR and sequencing primers.

All primers were obtained from Eurogentec (Eurogentec Nederland b.v., Maastricht, The Netherlands). The primer pair used for amplification consisted of 27F (5′-AGA GTT TGA TC[A/C] TGG CTC AG-3′) and 1492R (5′-G[C/T]T ACC TTG TTA CGA CTT-3′). This primer pair amplifies an ∼1,500-bp fragment of the 16S rRNA gene between positions 8 and 1509 of the E. coli 16S rRNA gene and is considered to be universal for the domain Bacteria (31). For reamplification of samples that yielded insufficient product for a successful sequencing reaction, the following primer pair was used: 357F (5′-CCT ACG GGA GGC AGC AG-3′) and 1221R (5′-CAT TGT AGC ACG TGT GTA GCC-3′). Sequencing was done with the following primers: 27F (5′-AGA GTT TGA TC[A/C] TGG CTC AG-3′) and 515R (5′-TAC CGC GGC TGC TGG CAC-3′). For reamplified samples, the sequencing primers used were: 357F (5′-CCT ACG GGA GGC AGC AG-3′), 797F (5′-CAA AC[A/G] GGA TTA GAT ACC C-3′), 907R (5′-CCG TCA ATT C[A/C]T TTG AGT TT-3′), and 1221R (5′-CAT TGT AGC ACG TGT GTA GCC-3′).

16S ribosomal DNA (rDNA) PCR.

Amplification of the bacterial 16S rRNA gene was carried out in a GeneAmp 9600 thermocycler (Applied Biosystems, Nieuwekerk a/d IJssel, The Netherlands). The final reaction mixture (50 μl) contained 25 μl of DNA eluate; 400 ng of each primer; 2.5 U of AmpliTaq Gold DNA polymerase (Applied Biosystems); 0.5 U of uracil-N-glycosylase (Applied Biosystems); 5 μg of bovine serum albumin (Roche Diagnostics Nederland B.V., Almere, The Netherlands); 20 μg of alpha-casein (C-6780, Sigma-Aldrich Chemie); 10 mM Tris-HCl (pH 8.3); 50 mM KCl; 3.5 mM MgCl₂; dATP, dGTP, and dCTP at concentrations of 200 μM each; and 400 μM dUTP (Applied Biosystems). The thermal cycling profile used was as follows: 2 min at 50°C and 10 min at 95°C, followed by 30 cycles of 20 s at 95°C, 20 s at 50°C, and 3 min at 72°C, followed finally by 10 min at 72°C.

Amplicon detection was carried out by agarose gel electrophoresis of 12 μl of amplification product through horizontal 1% agarose (Agarose MP; Roche Diagnostics Nederland B.V.) gels containing 1 μg of ethidium bromide in 0.5× Tris-borate-EDTA buffer. After electrophoresis, amplicons were made visible by UV light irradiation.

Reamplification of samples that yielded insufficient amounts of product for direct sequencing was carried out in the same reaction mixture of the first-round PCR, except for the presence of uracil-N-glycosylase, which was not included in reamplification reactions. Template DNA was purified with the QIAquick PCR purification kit (Westburg, Leusden, The Netherlands) according to the manufacturer's instructions and subsequently diluted 1,000 times. The thermal cycling profile was also the same as for first-round PCR.

Precautions to prevent false positives and false negatives.

In order to prevent false-positive and false-negative results in the broad-range PCR, our laboratory workflow was organized in the way described by Millar et al. (36), with a few minor exceptions. These exceptions were as follows: (i) all reagents were molecular grade, but not UV irradiated before use, and (ii) separate pipettes were used in sample preparation, PCR setup, and post-PCR handling, but pipettes were not UV irradiated before use. Extra measures used to prevent false positives were as follows: (i) use of the uracil-N-glycosylase system to prevent amplicon carry-over contamination, (ii) a negative extraction control was included for every five clinical samples to check for cross-contamination during extraction and for quality control of extraction chemicals, and (iii) a negative PCR control was included in each run to quality control the PCR chemicals. To prevent false negatives, all PCRs were done in duplicate—one reaction without a spike and one reaction with 40 pg of purified E. coli DNA as a spike to control for PCR inhibition.

Sequencing.

Amplicons were purified with the QIAquick PCR purification kit (Westburg) according to the manufacturer's instructions. The sequencing reactions were performed with the Big Dye terminator cycle sequencing kit v1.0 or v1.1 (Applied Biosystems) according to the manufacturer's instructions. Cycle reactions were purified by either manual sodium acetate-ethanol precipitation or by the use of the DyeEx 2.0 spin kit (Westburg) according to the manufacturer's instructions. The manual procedure was as follows: 20 μl of cycle sequencing product was added to 50 μl of 96% (vol/vol) ethanol and 2 μl of 3 M sodium acetate (pH 6.4), and this mixture was left at ambient temperature for 15 min. The mixture was centrifuged for 30 min at 12,000 × g, and the supernatant was discarded by pipetting. A portion (250 μl) of 70% (vol/vol) ethanol was added, and the mixture was vortexed and centrifuged for 15 min at 12,000 × g. The supernatant was again discarded by pipetting, and the pellet was centrifuged for 5 min at 12,000 × g. After the small amount of supernatant was removed by pipetting, the pellet was dried for 15 min at 60°C under vacuum conditions. This drying step is part of both the manual and the DyeEx procedures. The dried cycle sequencing products were then dissolved in 25 μl of template suppression reagent (Applied Biosystems) and analyzed on an ABI Prism 310 genetic analyzer (Applied Biosystems) according to the manufacturer's instructions.

Data analysis.

Raw data was analyzed with the Sequence Analysis software v 3.3 and the Factura software v 2.2 (Applied Biosystems). Sequences from both strands were aligned by using the AutoAssembler software v 2.1 (Applied Biosystems) and edited, to resolve discrepancies, by evaluation of the electropherograms. The resulting consensus sequence, created from the double-stranded part of the alignment, was used for comparison with sequences stored in GenBank, EMBL, and DDJB by using the basic local alignment search tool (BLAST; http://www.ncbi.nlm.nih.gov/BLAST/) and the similarity matrix tool from the Ribosomal Database Project II (RDP-II; http://rdp.cme.msu.edu/html/analyses.html) (11). For both BLAST and RDP-II analyses, the highest-scoring species was reported. Sequences identified in the present study are deposited in GenBank under accession numbers AY325737 to AY325764, AY332613 to AY332615, and AY429260.

RESULTS

Lower limit of detection of the procedure.

To assess the lower limit of detection, the procedure was performed on 10-fold serial dilutions of E. coli and S. aureus ranging from 10⁷ to 10⁰ CFU/ml of pooled CSF, which were previously proven to be both culture and broad-range 16S rDNA PCR negative. Bacteria used for spiking were quantified by agar dilution. Agarose gel electrophoresis of the amplified products showed a lower limit of detection of 100 and 200 CFU/ml of CSF for E. coli and S. aureus, respectively. Amplicons obtained from the dilutions that contained 100 and 1,000 CFU/ml needed reamplification for a successful sequencing identification. All amplicons yielded the correct identification of the spiked organism after sequencing (data not shown).

Laboratory findings for CSF samples from conventional procedures.

A total of 227 CSF samples were analyzed with both conventional and broad-range 16S rDNA PCR and the results are summarized in Table 1. Microbial growth in bacterial culture was detected for 37 CSF samples. In 21 samples the direct microscopic examination was confirmed by culture, but for the remaining 16 culture-positive samples no bacteria were seen. None of the culture-negative samples had any evidence of bacteria in the direct microscopic examination, but in four samples yeasts were seen. Bacterial species isolated during culture were the following: N. meningitidis (14 isolates), coagulase-negative staphylococci (CNS) (8 isolates), S. pneumoniae (8 isolates), Haemophilus influenzae (2 isolates), E. coli (2 isolates), viridans group streptococci (1 isolate), Listeria monocytogenes (1 isolate), and a Pantoea species. (1 isolate).

TABLE 1.

Overall results obtained by 16S rDNA PCR compared to culture

Method	Finding	No. of CSF samples found to be culture:		Total no. of specimens
		Positive	Negative
16S rDNA PCR	Positive	24	6	30
	Negative	13	184	197
Total		37	190	227

Laboratory findings from broad-range 16S rDNA PCR.

Broad-range 16S rDNA PCR resulted in a total of 30 positive and 197 negative specimens. Sequencing of these positive amplicons resulted in the following identifications: S. pneumoniae (11 times), N. meningitidis (10 times), H. influenzae (2 times), E. coli (2 times), Streptococcus salivarius (1 time), Streptococcus agalactiae (1 time), Prevotella bivia (1 time), and Fusobacterium nucleatum (1 time). One sample resulted in a mixed sequence.

Concordant PCR-positive and culture-positive samples.

Table 2 shows the results for 22 CSF samples for which conventional and molecular methods were in complete concordance. Pathogens identified by both methods included most of the common causative agents of bacterial meningitis. N. meningitidis was identified most often, with a total of nine cases, followed by S. pneumoniae (eight cases), E. coli and H. influenzae (two cases each), and finally S. salivarius (viridans group streptococci in culture) with one episode. In this group the broad-range 16S rDNA PCR showed supplementary value in one sample. This sample yielded a strain of S. pneumoniae with very poor growth characteristics, making biochemical identification not possible, even after 8 days. Identification of the isolated strain, resulting in S. pneumoniae, was only possible by 16S rRNA gene sequencing. Broad-range 16S rDNA PCR, followed by sequencing of the original CSF sample, resulted in S. pneumoniae within 48 h.

TABLE 2.

Results of concordant PCR-positive and culture-positive specimens

Lab no.	Broad-range 16S rDNA method			Conventional method
	PCRa	BLAST (% identity)	RDP-II (% similarity)	Culture resultb	Microscopy resultc
1489	+	E. coli (100)	E. coli (1.000)	E. coli (+)	(−)
1495	+++	S. salivarius (99.9)	S. salivarius (1.000)	Viridans group streptococci (+++)	GPC (++++)
1530	+++	S. pneumoniae (100)	S. pneumoniae (1.000)	S. pneumoniae (+++)	GPC (++++)
1532	+++	N. meningitidis (100)	N. meningitidis (0.998)	N. meningitidis (+++)	GNC (++++)
1563	+++	S. pneumoniae (100)	S. pneumoniae (0.998)	S. pneumoniae (+++)	GPC (++++)
1577	++	N. meningitidis (100)	N. meningitidis (1.000)	N. meningitidis (++)	GNC (+++)
1682	+++	N. meningitidis (99.8)	N. meningitidis (0.995)	N. meningitidis (+++)	GNC (+++)
1713	++	N. meningitidis (100)	N. meningitidis (0.998)	N. meningitidis (++)	GNC (++)
1826	+++	N. meningitidis (99.8)	N. meningitidis (0.993)	N. meningitidis (+)	GNC (+)
1835	+++	S. pneumoniae (99.8)	S. pneumoniae (0.998)	S. pneumoniae (++)	GPC (++++)
1836	+	S. pneumoniae (99.9)	S. pneumoniae (0.999)	S. pneumoniae (+)	GPC (+)
1863	+	E. coli (99.5)	E. coli (0.998)	E. coli (+++)	GNR (++)
1898	+++	N. meningitidis (99.8)	N. meningitidis (0.993)	N. meningitidis (++)	GNC (++++)
1908	+++	N. meningitidis (99.8)	N. meningitidis (0.993)	N. meningitidis (+++)	GNC (++)
1963	+	H. influenzae (99.7)	H. influenzae (0.999)	H. influenzae (++)	(−)
2055	+++	S. pneumoniae (99.6)	S. pneumoniae (0.994)	S. pneumoniae (+)d	(−)
2108	++	H. influenzae (99.7)	H. influenzae (0.997)	H. influenzae (+++)	GNR (++++)
2109	++	S. pneumoniae (99.8)	S. pneumoniae (0.998)	S. pneumoniae (+++)	GPC (++++)
2302	++	N. meningitidis (100)	N. meningitidis (0.999)	N. meningitidis (++)	GNC (+)
2452	++	S. pneumoniae (100)	S. pneumoniae (0.998)	S. pneumoniae (+)	GPC (++)
2457	+++	S. pneumoniae (99.8)	S. pneumoniae (0.998)	S. pneumoniae (+++)	GPC (++++)
2471	++	N. meningitidis (100)	N. meningitidis (1.000)	N. meningitidis (+++)	GNC (++)

^aThe PCR result is reported as follows: +, weakly positive; ++, moderately positive; and +++, strongly positive.

^bSymbols in parentheses are relative numbers of colonies on the primary plate: +, few or enrichment broth positive; ++, moderate; and +++, many.

^cGPC, gram-positive cocci; GNC, gram-negative cocci; GNR, gram-negative rods. Symbols in parentheses are relative numbers of bacterial cells seen in direct microscopy: −, none; +, sporadic; ++, few; +++, moderate; and ++++, many.

^dEnrichment broth yielded a positive result; however, subculturing of the strain did not yield enough growth for identification procedures. The strain was identified by 16S rDNA sequencing.

Concordant culture-negative and PCR-negative samples.

Bacteria were not detectable in the vast majority, i.e., 81.1% (184 of 227 specimens), of analyzed CSF samples by both conventional and molecular methods, as illustrated in Table 1. Retrospective analysis of these 184 CSF samples revealed a virus or yeast by culture, microscopy, or specific viral PCR in 7 samples: enterovirus RNA and herpes simplex virus DNA, respectively, were detected in 1 sample each. In the other five samples yeasts were detected by microscopy and/or culture. The low number of viral agents that were finally detected is due to the fact that only a minority of the samples was analyzed for the presence of viruses.

PCR-negative and culture-positive discrepant samples.

For 13 CSF samples the positive culture result obtained could not be confirmed by the broad-range 16S rDNA PCR. Eight of these CSF samples resulted in the isolation of CNS, as is shown in Table 3. From the remaining five samples, N. meningitidis was isolated three times, and L. monocytogenes and a Pantoea sp. were both isolated once each. The CNS and Pantoea sp. culture results were regarded as contamination of bacterial culture and therefore as false positives. In the other four patients the diagnosis of bacterial meningitis was confirmed clinically.

TABLE 3.

Results of discordant PCR and culture specimens

Discrepancy type and lab no.	Broad-range 16S rDNA method			Conventional method		Comment(s)g
	PCR resulta	BLAST (% identity)	RDP-II (% similarity)	Microscopy resultb	Culture resultc
PCR-negative and culture- positive discrepancies
1572	−	NDf	ND	(−)	CNS (+)	Contamination*
1705	−	ND	ND	(−)	CNS (+)	Contamination*
1751	−	ND	ND	(−)	CNS (+)	Contamination*
1818	−	ND	ND	(−)	N. meningitidis (+)	Bacterial meningitis*
1824	−	ND	ND	(−)	CNS (+)	Contamination*
1961	−	ND	ND	(−)	N. meningitidis (+)	Bacterial meningitis*
2061	−	ND	ND	(−)	L. monocytogenes (+)	Bacterial meningitis*
2106	−	ND	ND	(−)	CNS (+)	Contamination*
2225	−	ND	ND	(−)	CNS (+)	Contamination*
2228	−	ND	ND	(−)	CNS (+)	Contamination*
2251	−	ND	ND	(−)	Pantoea species (+)	Contamination*
2263	−	ND	ND	(−)	CNS (+)	Contamination*
2266	−	ND	ND	GNC (+)	N. meningitidis (+)	Bacterial meningitis*
PCR-positive and culture- negative discrepancies
1750	+	S. pneumoniae (99.9)	S. pneumoniae (0.999)	(−)	(−)	Antibiotic treatment†
2125	+++	S. agalactiae (99.7)	S. agalactiae (0.997)	(−)	(−)	Antibiotic treatment†
2145	+++	P. bivia (98.5)	P. bivia (0.998)	(−)	(−)	Anaerobic culture not performed†
2264	++	N. meningitidis (99.8)	N. meningitidis (1.000)	(−)	(−)	Antibiotic treatment†
2304	+++	S. pneumoniae (100)	S. pneumoniae (0.999)	(−)	(−)	Antibiotic treatment†
2403	+++	S. pneumoniae (100)	S. pneumoniae (0.999)	(−)	(−)	Antibiotic treatment†
PCR and culture identifi- cation discrepancies
1936d	+	Lactococcus sp. (87.8)	Neisseria sp. (0.925)	(−)	N. meningitidis (+)	SSEARCH result N. meningitidis
2253e	+++	F. nucleatum (99)	F. nucleatum (0.998)	GNC (+)	N. meningitidis (+)	PCR contamination

^aThe PCR result is reported as follows: +, weakly positive; ++, moderately positive; and +++, strongly positive. Numbers in parentheses are identity percentages.

^bGNC, gram-negative cocci. Symbols in parentheses reflect relative numbers of bacterial cells seen in direct microscopy: −, none; and +, sporadic.

^cSymbols in parentheses reflect relative numbers of colonies on the primary plate: +, few or enrichment broth positive; −, none.

^dAn electropherogram showed clear signs of a mixed sequence and resulted in discrepant sequence results between BLAST and RDP-II. A third analysis with the SSEARCH algorithm identified N. meningitidis.

^eThe sequencing result of F. nucleatum did not match the clinical presentation of the patient, although all controls were valid. Reanalysis of the original CSF and the original DNA isolate identified both N. meningitidis. The first result was therefore reported as contamination of the PCR reaction.

^fND, not done.

^g*, Clinical interpretation; †, possible reason for negative culture.

PCR-positive and culture-negative discrepant samples.

For six culture-negative CSF samples the broad-range 16S rDNA PCR did yield a positive result. In three of these samples S. pneumoniae was identified, and in the other three N. meningitidis, S. agalactiae, and P. bivia were detected once each (Table 3). For five of these CSF samples antimicrobial therapy had already started at the time of lumbar puncture.

PCR-positive and culture-positive discrepant samples.

For two samples sequence data did not match the organism isolated by culture (Table 3). The first sample, from which N. meningitidis was isolated in culture, showed clear signs of a mixed sequence in its electropherograms obtained after broad-range 16S rDNA PCR. When this sequence was analyzed by using the BLAST algorithm, the sequence was most related to a Lactococcus species. However, when the sequence was analyzed with the similarity matrix algorithm from RDP-II it resulted in a Neisseria species. Another sample, from which also a N. meningitidis strain was isolated in culture, resulted in a sequence identification of F. nucleatum.

Overall performance of the broad-range PCR assay compared to culture.

For calculation of the overall performance of the assay compared to bacterial culture, the data from Table 1 were used, with the exception that the nine false-positive culture results were excluded from the analysis. This results in an overall sensitivity of 86%, a specificity of 97%, a positive predictive value (PPV) of 80%, and a negative predictive value (NPV) of 98%.

DISCUSSION

In the present study we evaluated broad-range 16S rDNA PCR and subsequent sequencing in diagnosis of bacterial meningitis in a clinical setting. Our results show that the molecular findings were in good agreement with those obtained from direct microscopy and bacterial culture. The results were in complete concordance for 206 CSF specimens (91%), including 184 culture-negative and PCR-negative samples and 22 culture-positive and PCR-positive samples.

Our broad-range 16S rDNA PCR proved to be extremely valuable in detecting bacterial meningitis when antibiotics had been started prior to lumbar puncture. For five CSF samples obtained after the start of antimicrobial therapy, a significant bacterial pathogen (S. pneumoniae [three times], S. agalactiae [one time], and N. meningitidis [one time]) was identified by broad-range 16S rDNA PCR, whereas the culture results remained negative. These data confirm the findings of earlier studies (10, 30, 42, 45).

Our broad-range 16S rDNA PCR also proved of value in the case of a 2.5-month-old boy with clinical symptoms of sepsis and meningitis. P. bivia was found in the CSF of this child, while conventional methods remained negative, since anaerobic culture is not part of the standard operation procedure for detecting bacterial meningitis in most microbiological laboratories. The boy was treated empirically with ampicillin and ceftazidime and recovered completely. Recently, isolation of P. bivia from CSF in a critical ill young patient has been described (8).

In two samples, both PCR and culture results were positive, but the microorganism found was different for each method. The first discrepant identification involved a CSF sample from which N. meningitidis was isolated. The sequence obtained after broad-range bacterial PCR showed clear signs of a mixed sequence in its electropherograms. This sequence was analyzed with the algorithms from the both BLAST and RDP-II databases. The BLAST search resulted in a Lactococcus species, most closely related to L. lactis (87.8%), with N. meningitidis (86.5%) as the next possible match, whereas the RDP-II search resulted in a Neisseria species (percent similarity values, 0.925 to 0.923). Upon manual comparison, >99% of the sequences from both L. lactis and N. meningitidis could be retrieved within the electropherogram. Finally, to confirm the previous findings, the original sequence was also subjected to a third search algorithm, SSEARCH from the Software and Tools for Genome Analysis website (http://www-btls.jst.go.jp/), resulting in N. meningitidis. Although lactococci have recently been associated with human disease (1, 17, 19, 37, 59), we concluded that, based upon the clinical evaluation, the lactococcal DNA was most likely a contaminant introduced at the time of lumbar puncture. The second discrepant result was the detection of F. nucleatum in the CSF of a 13-year-old girl presenting with mild symptoms of bacterial meningitis. Culture of the CSF yielded N. meningitidis, and the child was treated accordantly. The detection of F. nucleatum did not match the clinical presentation of the patient, even though all controls included were valid. Therefore, the original CSF sample, as well as the original DNA isolate, was retested, resulting in the identification of N. meningitidis for both samples. Resequencing the original PCR product again yielded F. nucleatum as in the first analysis, confirming that the first PCR had to be contaminated. This case emphasizes that all laboratory findings should always be placed in perspective with the clinical symptoms, even when all controls to prevent false-positive results make a run valid.

One of the shortcomings of our approach seems to be the sensitivity in the lower range of bacterial loads encountered in bacterial meningitis. A total of 13 culture-positive CSF samples remained negative by broad-range 16S rDNA PCR. In four of these samples it was evident that a bacterial pathogen was present, whereas the others were considered to be contaminants of bacterial culture on clinical grounds. All of these samples, which had low numbers of colonies on primary plates, were estimated to have bacterial loads below our limit of detection. However, it should be emphasized that the sensitivity of our assay (100 CFU/ml of CSF) should be sufficient for detecting the majority of bacterial meningitis cases, since La Scolea et al. showed that only 16% of culture-positive CSF samples in bacterial meningitis have bacterial loads of <1,000 CFU/ml (32).

The overall performance of our assay, compared to bacterial culture, is in line with other published broad-range PCR methods (30, 33, 42), with an overall sensitivity of 86%, a specificity of 97%, a PPV of 80%, and an NPV of 98%. It should be emphasized that the relatively low PPV is at least partly explained by the inadequacy of bacterial culture: in six PCR-positive, culture-negative CSF samples a significant bacterial pathogen was identified in a patient with all of the signs of bacterial meningitis (Table 3).

The molecular method validated here is not only suitable for the detection of bacteria in CSF but can also be used for the detection of bacteria in many other clinical samples from anatomic sites that are considered to be sterile in healthy subjects. We have demonstrated that we can detect clinically significant bacteria in other clinical samples, such as blood culture fluids, a brain abscess, a liver abscess, pus from a vertebra, synovial fluids, and pericardial fluid, for which conventional methods could not (data not shown).

In conclusion, our broad-range 16S rDNA PCR with subsequent sequencing has proven to be a valuable supplementary test in daily clinical practice. Bacterial culture of CSF remains the cornerstone in the diagnosis of bacterial meningitis. However, especially when antimicrobial therapy has already started at the time of lumbar puncture, conventional methods should always be accompanied by molecular detection, since the sensitivity of direct microscopic examination and bacterial culture drops substantially when therapy has started. We are currently adapting the standard procedures in our laboratory for the detection of bacterial meningitis; broad-range 16S rDNA detection is indicated when antimicrobial therapy has already been started at the time of lumbar puncture or when cultures remain negative, although the suspicion of bacterial meningitis remains.