JOURNAL OF CLINICAL MICROBIOLOGY, Mar. 1997, p. 691696 0095-1137/97/$04.

00 0 Copyright 1997, American Society for Microbiology

Vol. 35, No. 3

Aseptic Meningitis and Encephalitis: the Role of PCR in the Diagnostic Laboratory
STEVEN J. READ,1 KATIE J. M. JEFFERY,1,2
AND

CHARLES R. M. BANGHAM2*

Oxford Public Health Laboratory, John Radcliffe Hospital, Oxford OX3 9DU,1 and Department of Immunology, Imperial College School of Medicine at St. Marys, London W2 1PG,2 United Kingdom
Received 19 August 1996/Returned for modication 29 October 1996/Accepted 12 December 1996

In this study we have devised a simple and robust PCR strategy to detect a wide range of viruses, bacteria, and parasites, all of which are capable of causing aseptic meningitis and encephalitis. The techniques developed have been used in a routine diagnostic virology laboratory to test prospectively 2,233 cerebrospinal uid specimens. A virus was detected in 147 specimens of cerebrospinal uid from 143 patients. Four sets of primers were sufcient to detect the virus in 135 (94%) of the PCR-positive patients. We conclude that with appropriate primers, PCR can be systematically and economically applied to test for a range of organisms in a routine diagnostic laboratory. In our opinion, PCR will soon become the gold standard test for viral infections of the central nervous system. Conventional techniques for identifying the causative organisms in aseptic meningitis and encephalitis have had limited success. PCR has obvious potential for increasing the rate of positive diagnosis, but the present lack of large-scale, prospective studies makes clinical interpretation of PCR results difcult. The main reasons to identify the causative organisms in cases of aseptic meningitis and encephalitis are (i) to provide a rational basis for chemotherapy and prognosis, (ii) to limit unnecessary investigations, and (iii) to provide epidemiological information. Conventional techniques of virus detection in cerebrospinal uid (CSF) are often unsatisfactory. Virus isolation from CSF in cell culture is inefcient for most of the viruses that cause disease in the central nervous system (CNS). The exceptions are members of the enterovirus group, the echoviruses, polioviruses, and coxsackie B viruses. Virus isolation may, however, take up to 7 days. Circumstantial evidence of enteroviral infections can be gained from culture of throat swabs and stool specimens and from a rise in specic serum antibody titer in cases of herpes simplex virus (HSV) infection of the CNS. However, excretion of enteroviruses in stool specimens can continue for many weeks following an enteroviral infection, and the majority of these infections do not cause CNS or any other symptoms. Brain biopsyperhaps the only existing gold standard in diagnosing viral encephalitisis now rarely justied because of its invasive nature. Laboratory diagnoses of the other organisms included in this study, for example, JC virus (JCV) and toxoplasma, are also difcult by conventional means or are achieved slowly by serological response or isolation in infections with lymphocytic choriomeningitis virus (LCMV), borreliae, and mycobacteria. The slowness and the low sensitivities and (in some instances) specicities of these techniques severely limit their usefulness. Nucleic acid amplication techniques, such as PCR, have obvious advantages compared to conventional techniques: their sensitivities and specicities are unparalleled. Several studies in which PCR was applied to the diagnosis of CNS infections, especially those with HSV (14, 18) and enteroviruses (20, 23), have been undertaken. However, the usefulness of a selected range of PCR tests as a routine screen in a diagnostic laboratory has not been examined previously. This study was therefore designed to develop methods for the routine application of PCR to CSF specimens. It was not designed to measure the clinical sensitivity and specicity of PCR as a diagnostic test, because full clinical information was not available on all 2,162 subjects in the study. A detailed statistical analysis of the results and correlation with clinical data for 410 patients from Oxford hospitals, on whom full clinical details were obtained, will be reported elsewhere (13a).
MATERIALS AND METHODS Specimen selection. The regional diagnostic virology laboratory in Oxford receives CSF specimens from three hospitals in Oxford and eight hospitals in the surrounding region. In addition, samples are referred from other clinical centers in the United Kingdom. We chose to test, without selection, all consecutive CSF specimens received by the laboratory between May 1994 and May 1996. In total, 954 patients were in Oxford city hospitals, 426 patients were in the Oxford region, and 782 patients were in hospitals elsewhere. Six hundred sixty patients were aged 16 years or under at the time of hospitalization. Seventy-two specimens (3.3%) were from patients known to be infected with human immunodeciency virus (HIV). Clinical information included on the request card (which was often very limited) was known to the personnel involved with the project. To enable us to decide on a rational approach to testing CSF by PCR, the rst 200 patient specimens were tested for a range of nine viruses (HSV, varicella-zoster virus [VZV], Epstein-Barr virus [EBV], cytomegalovirus [CMV], human herpesvirus 6 [HHV-6], adenovirus, enteroviruses, measles virus, and mumps virus). Based on the data obtained by testing these 200 specimens, an algorithm (Fig. 1) was constructed and incorporated into the protocol for testing the remaining specimens in this study. Primers specic for LCMV, JCV, Mycoplasma pneumoniae, and Toxoplasma gondii were used in addition if the CSF contained 5 leukocytes/ l or if the patient was immunocompromised. When requested by clinicians directly involved with patient care, PCR was also carried out for Mycobacterium species, Listeria monocytogenes, Borrelia burgdorferi, and Leptospira interrogans. In the initial testing of 200 specimens, we included as controls 28 specimens of CSF taken from patients whose differential diagnoses did not include a CNS infection and in whom a lumbar puncture was done for myelography, CSF electrophoresis, relief of benign intracranial hypertension, etc. Specimen processing. RNA was extracted from 100 l of CSF by the method of Boom et al. (2) with the omission of the guanidinium isothiocyanate washes after the extraction step. In some cases, only 50 l was tested. CSF was used without extraction for amplication of viruses with a DNA genome, as we could detect no inhibition by nonbloodstained CSF in our nested PCR protocol (results not shown), in agreement with other workers (10, 11) but in contrast to one report (8). Experiments to determine the inhibitory effects of CSF on the PCR included titration of virus in samples of CSF to determine endpoint dilutions, which were then compared to dilution endpoints of virus titrated in distilled water. These CSF samples included some with a high protein level and pleocytosis. Specimens that were visibly bloodstained were a particular concern, as the

* Corresponding author. Mailing address: Department of Immunology, Imperial College School of Medicine at St. Marys, Norfolk Pl., London W2 1PG, United Kingdom. 691

692

READ ET AL.

J. CLIN. MICROBIOL.
tures were subjected twice to freeze-thaw cycles. Supernatants were claried by centrifugation at 5,000 rpm for 5 minutes in a Denley BS400 benchtop centrifuge and aliquoted. Two 10-fold dilution series of the same claried culture supernatants were prepared and inoculated onto MRC-5 monolayers. From these same dilutions, 100- l samples were extracted and tested by the enteroviral RT-PCR; the CMV and HSV PCRs were used on 10- l samples of the dilutions directly. The endpoints of virus detection by cytopathic effect (expressed as 50% tissue culture infections doses [TCID50s] per milliliter) and by dilutions at which 50% of PCRs are positive (PCRD50s) were determined for each dilution series. A concentration in PCRD50 per milliliter is calculated for the highest dilution of a sample that gives 50% positive PCRs and is corrected per milliliter. E.g., if 10 l of a 10 4 dilution is PCR positive as the highest dilution, the solution contains 106 PCRD50/ml. In addition, virus particles in tissue culture supernatants were counted by electron microscopy. An aliquot of the supernatant was mixed with an equal volume of latex particles (diameter 90 nm; concentration, 1.37 1010 per ml; Balzers Union, Balzers, Liechtenstein) and the numbers of latex particles and numbers of virus particles (NVPs) were counted in 10 microscope elds (data not shown). The NVP per milliliter was then computed by direct proportionality (Table 2). For measles and mumps viruses, the lyophilized virus preparation was resuspended in 1 ml of sterile distilled water (giving 500 TCID50 of measles virus/ml and 10,000 TCID50 of mumps virus/ml) and a 100- l aliquot was used for extraction to determine the sensitivity of the RT-PCR for these viruses. PCR was performed with a solution containing 20 mM (NH4)2SO4, 75 mM Tris-HCl (pH 9.0 at 25 C), 0.01% (wt/vol) Tween 20 (buffer IV; Advanced Biotechnologies Ltd., Leatherhead, United Kingdom), 1.5 mM MgCl2, 0.25 mM each deoxynucleoside triphosphate (Advanced Biotechnologies Ltd.), 0.1 M each relevant oligonucleotide (R & D Systems Europe Ltd.), and 0.625 U of Taq polymerase (manufacturers units; Advanced Biotechnologies Ltd.). The optimum magnesium ion concentration was determined for each primer set separately by standard titration experiments. RT was performed, where appropriate, in a single reaction mixture as described above, except that 3 mM MgCl2 was used and 0.1 U of Moloney murine leukemia virus reverse transcriptase (manufacturers units; Advanced Biotechnologies Ltd.) was added. RT was carried out by incubating the reaction mixtures for 15 min at 37 C before starting the PCR cycling. CSF (10 l) was added to each PCR mixture containing primers specic for DNA viruses, and 20 l of the extracted, dissolved RNA, equivalent to 50 l of CSF, was added to reaction mixtures containing primers for RNA viruses. In each case, the nal PCR mixture volume was 50 l. After initial denaturation for 3 min at 94 C, amplication was performed with 35 cycles comprising denaturation for 30 s at 94 C, annealing for 20 s at the appropriate temperature, and polymerization for 30 s at 72 C. Amplication with internal (nested) primers was performed in a reaction mixture identical to that described above, except that the total reaction mixture volume was 25 l and that 2 l of the initial reaction mixture was added as the template. Amplication was then performed with 35 cycles as follows: 20 s of denaturation at 94 C, 20 s of annealing, and 30 s of polymerization at 72 C. All thermal cycling was carried out in an Omnigene thermal cycler (Hybaid). Standard procedures to minimize and detect contamination were observed. Amplication products were detected by electrophoresis of 10 l of the secondary reaction mixture through an ethidium bromide-stained 2% agarose gel with transillumination by UV light. From receipt of a specimen, the procedure took about 6 h for the initial screen. All PCR-positive specimens were tested at least twice to conrm the original reaction. At the end of the study, primers were designed for the differentiation of HSV-1 and HSV-2 DNAs, and this differentiation was done retrospectively when CSF was still available.

FIG. 1. Algorithm for PCR with CSF samples for a virology lab. pos, positive; neg, negative; wbc, leukocytes.

inhibitory effect of hemoglobin on Taq polymerase is well known. These specimens were therefore tested (i) undiluted and untreated, (ii) undiluted and untreated but spiked with the respective control nucleic acid, and (iii) after extraction of the nucleic acid by the Boom method. For successful amplication of Mycobacterium tuberculosis from CSF, we found in control experiments with culture-positive CSF that it was necessary to extract DNA from the sample by the Boom method. All specimens originating from Oxford city and region hospitals and containing sufcient volumes of CSF were cultured on MRC-5 broblasts and primary Rhesus monkey kidney cells (Department of Cell Resources, Centre for Applied Microbiology and Research, Porton Down, United Kingdom) to detect infectious viruses. PCR. Nested PCR was performed for all organisms studied. A list of the primer sequences designed specically for use in this study and not published elsewhere is given in Table 1; references are given for the remaining primers. To economize on materials, and after ensuring that sensitivity of detection was unaffected (Fig. 2), primers specic for certain viruses were combined in the reaction mixtures, and the reaction products were distinguished by their electrophoretic mobilities. Thus, primers for HSV and VZV were combined, as were primers for EBV and HHV-6 and primers for the enterovirus group and echovirus types 22 and 23, which show sequence divergence from the other members of the echovirus group (22). Nested PCR products were highly specic and easily distinguished by electrophoresis with reference to their molecular weights. All other organisms were amplied separately. A positive control reaction mixture comprising a dilution of virus grown in tissue culture, T. gondii grown in a mouse peritoneal cavity, or a culture-grown bacterium (Mycoplasma pneumoniae, L. monocytogenes, M. tuberculosis, Leptospira interrogans, or B. burgdorferi) was included in each PCR batch. Where applicable, the control nucleic acid was extracted in parallel with the samples of CSF. The sensitivities of representative tests by PCR and reverse transcription (RT)-PCR were determined with tissue culture-grown CMV, HSV type 1 (HSV1), and poliovirus type 2 and with measles-mumps-rubella (MMR) vaccine (Merck & Co., Inc.) for measles virus and mumps virus. MRC-5 broblast monolayers were infected at a low multiplicity of infection with the AD169 laboratory strain of CMV, a clinical isolate of HSV-1 and a vaccine strain of poliovirus type 2. When complete cytopathic effect had been reached, the cul-

RESULTS In the magnesium ion optimization titration, all the primer sets performed optimally in the range of 1 to 3 mM MgCl2, and no difference in sensitivities of the amplications could be detected within this range after nested PCR for any primer sets (data not shown). It was therefore possible to use a standard buffer containing 1.5 mM MgCl2 for all the primer sets employed in this study. In control experiments to determine the effects of multiplex primers on the sensitivity of the PCR, no loss of sensitivity could be detected compared to the sensitivities of reactions with single primer pairs (Fig. 2). The sensitivity of the PCR was determined for CMV, HSV-1, measles virus, mumps virus, and poliovirus type 2 (Table 2). For poliovirus type 2, 11 virus particles represented 1 TCID50 and 34 virus particles equalled 1 PCRD50, resulting in 3 TCID50 per PCRD50; for HSV-1, 50 virus particles equalled 1 TCID50 and 158 virus particles equalled 1 PCRD50, again resulting in 3 TCID50 per PCR-detectable DNA. The NVP/ infectivity ratios were relatively low for both viruses (7), indicating a highly productive viral replication in vitro. The sensi-

VOL. 35, 1997

ROLE OF PCR IN ASEPTIC MENINGITIS AND ENCEPHALITIS TABLE 1. Primer sets designed for use in this study and not published elsewherea

693

Microorganism

Region amplied

Primer orientationb

Primer coordinates

Primer sequence

HSV-1 and -2 HSV-1 HSV-2 CMV

Glycoprotein D Glycoprotein D Glycoprotein D Glycoprotein B

Enterovirus

5 Nontranslated region

Echovirus type 22

5 Nontranslated region

LCMV

Glycoprotein C

Mumps

Measles B. burgdorferi

SH protein HN genec SH protein SH protein Fusion protein Flagellin gene

o, s o, as o, s o, as o, s o, as o, s o, as i, s i, as o, s o, as i, s i, as o, s o, as i, s i, as o, s o, as i, s i, as o, s o, as i, s i, as i, s i, as o, s o, as i, s i, as

716732 899915 619638 899915 619638 899915 18211837 21302146 18381856 21192138 6585 450470 168185 415432 3353 342363 113133 278296 508527 880898 529548 791810 5069 2648 7594 239261 521540 859881 271290 541551 301320 491510

TGCTCCTACAACAAGTC CGGTGCTCCAGGATAAA CGAAGACGTCCGGAAACAAC CGGTGCTCCAGGATAAA GGACGAGGCCCGAAAGCACA CGGTGCTCCAGGATAAA TGAGGAATGTCAGCTTC TCATGAGGTCGTCCAGA CCAGCCTCAAGATCTTCAT TCGTCCAGACCCTTGAGGTA CGGTACCTTTGTACGCCTGTT GGACACCCAAAGTAGTCGGT CAAGCACTTCTGTTTCCC AGGCTCTTCACACCATGT GCCTTATAACCCGACTTGCT TCCCACATCTCACCAGACAT CATAGTAACCAACACCTAAGA CGTGGTATCCCACTAATAGA CTTCATACTCCCTCGAAGCT GCTACAGCCAGACAATGCTT TATTCTTGTAGTCCAGAAGC ACCGCTCCTACATGGATTGA TCGTAACGTCTCGTGACCCT CTGTGTTGTATTGTGATCCA TGCACTATGCCGGCAATCCA GATCTTTCTAGAGTGATTGATC AGGAGATGATATTGGCTGTT GACAGCGTCGGATAGGCTATAC CTTTCTAGTGGGTACAGAAT TCAGCAATTCTATTAATTTC TCTGATGATGCTGCTGGCAT AGAACCTCTGTCTGCATCTG

a Previously published primers are as follows: HSV-1 and -2 (internal) (18), VZV (24), EBV (4), HHV-6 (25), adenovirus (1), JCV (11), measles virus (external) (9), T. gondii (16), Mycobacterium species (17), L. monocytogenes (13), Mycoplasma pneumoniae (6), and L. interrogans (19). b o, outer primer; s, sense primer; as, antisense primer; i, internal primer. c The mumps HN gene is downstream of the SH gene.

tivity of the PCR was close to that of viral infectivity for which only 11 to 50 virus particles were needed. The ndings for CMV were different in that 12,000 virus particles represented 1 TCID50 but only 38 virus particles equalled 1 PCRD50, indicating that PCR is over 100 times more sensitive than testing for infectivity. For measles and mumps viruses, infectivity was found to be 20 times less and 10 times more sensitive, respectively. Although we did not nd measles or mumps RNA in our

FIG. 2. Comparison of the sensitivities of amplication of HSV and VSV with single and combined primer sets. Lane 1, PCR markers (Promega) containing DNA fragments of 50, 150, 300, 500, 750, and 1,000 bp; lane 2, PCR mixture without addition of template; lanes 3 to 8 and lanes 9 to 14, 10-fold dilution series (diluted in a control CSF) of tissue culture-grown HSV-1 (141 bp) and VZV (208 bp) (amplied with single primer pairs), respectively; lanes 15 to 20, coamplication of tissue culture viruses amplied in lanes 3 to 14 (amplied with combined primer sets). The starting dilution amplied in each case was 1:104 of the tissue culture-grown virus supernatant. The endpoint dilution for HSV-1 and VZV amplication in this dilution series was therefore 1:108 and 1:107, respectively.

clinical materials, the sensitivity comparison suggests that PCR would have been more sensitive in detecting the presence of measles virus and only slightly less sensitive in detecting the presence of mumps virus compared to the sensitivity of infectivity testing. The specicities of the primers designed as part of this study were tested in each case to ensure that the genomes of related viruses or bacteria were not amplied (data not shown). For example, primers for CMV failed to amplify sequences from HHV-6, EBV, VZV, HSV-1, or HSV-2; consensus primers for enteroviruses amplied specic sequences from poliovirus, coxsackie A and B virus, and echovirus isolates (but not echovirus types 22 and 23) and additionally rhinovirus isolates but not isolates from other members of the Picornaviridae. No inhibition of amplication from CSF by high levels of protein or high numbers of leukocytes was detected in control experiments (data not shown). Inhibition by hemoglobin was encountered with specimens taken by traumatic lumbar punctures. However, no inhibition could be detected by the presence of a small number of erythrocytes, a number not sufcient to stain the sample visibly red. The results from samples from the rst 200 patients were used to construct an algorithm (Fig. 1) to provide a logical order for testing further specimens. In this series of samples, nine CSFs were positive by PCR for an enterovirus, six were positive for HSV, four were positive for VZV, and one was positive for EBV. A further 2,033 consecutive CSF specimens

694

READ ET AL.

J. CLIN. MICROBIOL.

TABLE 2. Calculation of the sensitivities of PCR and RT-PCR for the detection of CMV, HSV-1, measles virus, mumps virus, and poliovirus type 2
Virus Log infectivity (TCID50/ml)a Log NVP/mlb Log PCRD50/mlc Log NVP/TCID50 (antilog) Log NVP/PCRD50 (antilog) Log TCID50/PCRD50 (antilog)

CMV HSV-1 Measles virusd Mumps virusd Poliovirus type 2

a b

4.5 6.5 2.7 4.0 7.0

8.58 8.2 8.03

6.0 6.0 4.0 3.0 6.5

4.08 (12,023) 1.7 (50) 1.03 (11)

2.58 (380) 2.2 (158) 1.53 (34)

1.5 (0.03) 0.5 (3) 1.3 (0.05) 1 (10) 0.5 (3)

TCID at which 50% of inoculated monolayers become infected (determined by the manufacturer of the MMR vaccine for the measles and mumps viruses). The NVPs were calculated from electron microscopic counts. c The reciprocal of the highest dilution positive for nucleic acid that was detectable by PCR adjusted to concentration per milliliter. E.g., if 10 l tests positive at a dilution of 10 4 the suspension contains 106 PCRD50. d The MMR vaccine was used, and NVP were not determined.

were tested from 1,962 patients (874 of the patients were male). In total, 147 CSF specimens from 143 patients gave a positive result by PCR (6.6% of all patients) (Table 3). The most common group of viruses identied was the enteroviruses (77 patients). The most common single virus found was HSV-1 (20 patients), followed by VZV (16 patients), EBV (11 patients), untyped HSV (7 patients), HSV-2 (6 patients), CMV (3 patients), JCV (2 patients), and HHV-6 (1 patient). The results of PCRs with the 28 control CSF specimens and PCRs for other organisms were uniformly negative. The PCR results were highly reproducible: in each case, a positive result was obtained in an independent PCR with the original specimen. This was to ensure that amplication was not due to random, sporadic contamination or technical error. Where possible, we corroborated the evidence from PCR by serology or virus culture from CSF, a throat swab, or a stool specimen (13a). In the detailed study of 410 patients in Oxford (13a), 192 of 410 (47%) CSF specimens were cultured. In our series, enteroviruses were cultured from throat swabs or stool specimens from two patients with a typical clinical picture of viral meningitis, but enteroviruses were not detected by RT-PCR in the CSF. In each case the CSF contained abnormal numbers of leukocytes. In these cases, when the stool suspension or throat swab was tested by RT-PCR, the viral sequence was reliably amplied, demonstrating that failure to amplify from the CSF was not due to sequence differences at the primer binding sites. In addition, two CSF specimens, one from an Oxford city hospital and another sent at ambient temperature from outside our region, failed to amplify an enterovirus sequence, but an enterovirus was isolated in tissue culture from the CSF. Both of these isolates were of viruses

that should have been amplied by the general enterovirus primers that we used. Apart from these four cases, we did not obtain other corroborative evidence of a viral infection in any PCR-negative patient. DISCUSSION A particularly important clinical application of PCR is in the diagnosis of infections of the CNS; existing diagnostic investigations are frequently unsatisfactory because they are insensitive or slow or provide only circumstantial evidence of CNS infection. We have compared the PCR endpoints (in PCRD50s per milliliter) with those of infectivity testing in TCID50s per milliliter) and virus particle counts. We are satised that the PCR procedures employed were very sensitive, sometimes more so than the infectivity tests (Table 2). In general, very small numbers of specic nucleic acid molecules per milliliter were detected (Table 2). We are also condent that we have not missed cases of CNS infection with measles virus and mumps virus, as the detection limit of infectious virus by PCR was very low (Table 2). In clinical samples, PCR detection is, however, likely to be more robust than virus culture, because in PCR detection, there is no requirement to maintain the replication competence of the virus and because certain viruses do not grow in tissue culture. In this study, enteroviruses accounted for the majority of viruses detected. This nding is in agreement with the results of previous studies (3, 12, 15). The detection rate for enteroviruses in this study did not have a marked seasonal distribution, in agreement with detection by isolation in tissue culture in this laboratory during the period of this study.

TABLE 3. CSF cell counts, protein amount, and HIV status by virus as detected by PCR
Virus detected No. of patients No. known to be HIV positive CSF leukocyte count/ la Rangea Mean CSF protein (g/ml) Rangea Mean

Enterovirus HSV-1 HSV-2 Untyped HSV VZV EBV CMV HHV-6 JCV
a b

77 20 6 7 16 11 3 1 2

0 0 0 0 1 2 2 0 1

02,500 (54) 0575 (14) 1161,188 (5) 2291 (5) 0310 (9) 198 (7) NAb NA NA

269 102 553 58 125 29 10 1,600 0

0.122.66 (42) 0.301.30 (7) 0.861.70 (5) 0.640.91 (4) 0.652.21 (6) 0.202.50 (6) NA NA NA

0.79 0.98 1.10 0.73 1.24 1.13 0.89 NA 0.5

Numbers in parentheses are the numbers of datum sets available for calculations of ranges and means. NA, data not available.

VOL. 35, 1997

ROLE OF PCR IN ASEPTIC MENINGITIS AND ENCEPHALITIS

695

Coxsackieviruses of types A1 to A6 cannot be grown in cell culture; PCR may therefore be particularly useful in diagnosing CNS infections with these viruses. Several recent papers have reported the use of PCR in the diagnosis of enteroviral meningitis (20, 23). In each study, enteroviral RNA was detected in CSF by RT-PCR, but as in the present study, corroborative evidence from virus culture was not obtained in every positive case. The two CSF PCR false-negative results for enterovirus conrm that at very low virus concentrations the PCR can be negative when culture is positive. It is likely that the sensitivity of PCR with clinical specimens is lower than its true molecular sensitivity for two main reasons. First, technical problems, such as degradation of viral RNA, will lower the sensitivity of the PCR. Second, it is possible that certain infections of the CNS are not always accompanied by the presence of the organism in the CSF. Accurate measures of the clinical usefulness of CSF PCR are difcult to obtain because of the lack of a satisfactory gold standard. We have addressed this question in more detail in a study of the clinical features of 410 patients for whom detailed clinical information was available (13a). We were concerned by the possibility that certain viruses which are often present in peripheral blood mononuclear cells, such as EBV, might be carried into the CSF, where they could be detected by PCR and give a spurious viral diagnosis. However, EBV was amplied from only 11 specimens; 4 were from HIV-infected immunocompromised patients with ring-enhancing lesions on a computerized tomography scan and a presumptive diagnosis of primary intracerebral lymphoma, 2 were from EBV immunoglobulin M-positive cases of primary EBV infection, and 5 were from HIV antibody-negative, immunosuppressed transplant recipients with diverse neurological symptoms and fever. EBV was not amplied from the other 2,222 CSF specimens. These results suggest that EBV PCR of CSF is of clinical use, even in lymphocytic specimens. It is now known that most cases of benign recurrent lymphocytic meningitis (Mollarets meningitis) are caused by HSV, especially HSV-2. Three patients in this study had experienced more than one episode, with clinical features typical of Mollarets meningitis, and indeed HSV DNA was amplied from their CSF specimens. Two of these specimens were amplied with type-specic primers, and HSV-2 was detected. In addition to these patients, three patients without previous episodes of meningitis presented with symptoms and CSF characteristics typical of those seen in aseptic meningitis. The patients received no specic antiviral treatment and were discharged home within a day or two of hospital admission. Specimens of CSFs, stools, and throat swabs from two of these patients were cultured and did not grow an enterovirus. PCRs of the CSF specimens from each patient were negative for enteroviruses but repeatedly positive for HSV. We had sufcient CSF to further identify the virus in two of the cases, and in both cases HSV-2 was amplied. In two of these patients, there was no evidence of concurrent genital herpes. These three patients had a clinical presentation very similar to those of three patients reported previously (21) who also had HSV-2 detectable in their CSFs. Meningitic symptoms are common with primary genital herpes (5). We detected HSV-2 in the CSF specimens of two patients with primary genital herpes who had signs and symptoms of meningitis. In the series reported here, four sets of PCR primers were sufcient to detect 95% of the PCR-positive samples. These primers were specic for enteroviruses, HSV, VZV, and EBV. We have demonstrated that these agents can be detected by PCR with three PCRs per specimen (employing multiplex PCR arrangements), thus making the screen highly efcient in

cost and time. Our reagent costs for multiplex PCR were similar to those of commercially available enzyme immunoassays for immunoglobulin M antibody, for example, those for CMV and EBV. As a consequence of our experience with a large number of specimens, we now test initially for enteroviruses, HSV, and VZV. If the patient is immunocompromised, we also test for EBV, CMV, HHV-6, JCV, and T. gondii. In other countries it may be necessary to include a search for nucleic acids of other organisms (e.g., Japanese B encephalitis virus). On the basis of this simple algorithm, our method is simple, easily performed within a single working day, and robust. We believe that such an approach will replace other methods as the gold standard of laboratory investigation for cases of suspected viral infection of the CNS.
ACKNOWLEDGMENTS We thank Ulrich Desselberger for valuable comments and suggestions on the manuscript. Support for this work was obtained from the Public Health Laboratory Service, London, United Kingdom.
REFERENCES 1. Allard, A., B. Albinsson, and G. Wadell. 1992. Detection of adenoviruses in stools from healthy persons and patients with diarrhoea by two-step polymerase chain reaction. J. Med. Virol. 37:149157. 2. Boom, R., C. J. A. Sol, M. M. M. Salimans, C. L. Jansen, P. M. E. Wertheimvan Dillen, and J. van der Noorda. 1990. Rapid and simple method for purication of nucleic acids. J. Clin. Microbiol. 28:495503. 3. Chonmaitree, T., C. D. Baldwin, and H. L. Lucia. 1989. Role of the virology laboratory in diagnosis and management of patients with central nervous system disease. Clin. Microbiol. Rev. 2:114. 4. Cinque, P., M. Brytting, L. Vago, et al. 1993. Epstein-Barr virus DNA in cerebrospinal uid from patients with AIDS-related primary lymphoma of the central nervous system. Lancet 342:398401. 5. Corey, L., H. G. Adams, Z. A. Brown, and K. K. Holmes. 1983. Genital herpes simplex virus infections: clinical manifestations, course and complications. Ann. Intern. Med. 98:958972. 6. de Barbeyrac, B., C. Bernet-Poggi, F. Febrer, H. Renaudin, M. Dupon, and C. Bebear. 1993. Detection of Mycoplasma pneumoniae and Mycoplasma genitalium in clinical samples by polymerase chain reaction. Clin. Infect. Dis. 17(Suppl. 1):S83S89. 7. Davis, B. D., R. Dulbecco, H. N. Eisen, and H. S. Ginsberg. 1980. Microbiology, 3rd ed., p. 879. Harper & Row Publications, Philadelphia, Pa. 8. Dennett, C., P. E. Klapper, G. M. Cleator, and A. G. Lewis. 1991. CSF pretreatment and the diagnosis of herpes encephalitis using the polymerase chain reaction. J. Virol. Methods 34:101104. 9. Esolen, L. M., B. J. Ward, T. R. Moench, and D. E. Grifn. 1993. Infection of monocytes during measles. J. Infect. Dis. 168:4752. 10. Espy, M. J., J. Aslandzadeh, and T. F. Smith. 1993. PCR detection of herpes simplex virus DNA sequences in cerebrospinal uid, p. 332336. In D. H. Persing, T. F. Smith, F. C. Tenover, and T. J. White (ed.), Diagnostic molecular microbiology. American Society for Microbiology, Washington, D.C. 11. Gibson, P. E., W. A. Knowles, J. F. Hand, and D. W. G. Brown. 1993. Detection of JC virus DNA in the cerebrospinal uid of patients with progressive multifocal leukoencephalopathy. J. Med. Virol. 39:278281. 12. Hammer, S. M., and K. J. Connolly. 1992. Viral aseptic meningitis in the United States: clinical features, viral etiologies, and differential diagnosis. Curr. Top. Clin. Infect. Dis. 12:125. 13. Jaton, K., R. Sahli, and J. Bille. 1992. Development of polymerase chain reaction assays for detection of Listeria monocytogenes in clinical cerebrospinal uid samples. J. Clin. Microbiol. 30:19311936. 13a.Jeffery, K. J. M., S. J. Read, T. E. A. Peto, R. T. Mayon-White, and C. R. M. Bangham. Diagnosis of viral infections of the central nervous system: clinical interpretation of PCR results. Lancet, in press. 14. Lakeman, F. D., R. J. Whitley, and the National Institute of Allergy and Infectious Diseases Collaborative Antiviral Study Group. 1995. Diagnosis of herpes simplex encephalitis: application of polymerase chain reaction to cerebrospinal uid from brain-biopsied patients and correlation with disease. J. Infect. Dis. 171:857863. 15. Meyer, H. M., R. T. Johnson, I. P. Crawford, H. E. Dascomb, and N. G. Rodgers. 1960. Central nervous system syndromes of viral etiology. Am. J. Med. 29:334347. 16. Ostergaard, L., A. K. Nielsen, and F. T. Black. 1993. DNA amplication on cerebrospinal uid for diagnosis of cerebral toxoplasmosis among HIV-

696

READ ET AL.
positive patients with signs or symptoms of neurological disease. Scand. J. Infect. Dis. 25:227237. Perosio, P. M., and T. S. Frank. 1993. Detection and species identication of mycobacteria in parafn sections of lung biopsy specimens by the polymerase chain reaction. Am. J. Clin. Pathol. 100:643647. Powell, K. F., N. E. Anderson, R. W. Frith, and M. C. Croxson. 1991. Non invasive diagnosis of herpes simplex encephalitis. Lancet 335:357. Savio, M. L., C. Rossi, S. Fusi, S. Tagliabue, and M. L. Pacciarini. 1994. Detection and identication of Leptospira interrogans serovars by PCR coupled with restriction endonuclease analysis of amplied DNA. J. Clin. Microbiol. 32:935941. Sawyer, M. H., D. Holland, N. Aintablian, J. D. Connor, E. F. Keyser, and N. J. Waecker. 1994. Diagnosis of enteroviral central nervous system infection by polymerase chain reaction during a large community outbreak. Pediatr. Infect. Dis. J. 13:177182. Schlesinger, Y., P. Tebas, M. Gaudreault-Keaner, R. S. Buller, and G. A. Storch. 1995. Herpes simplex virus type 2 meningitis in the absence of genital

J. CLIN. MICROBIOL.
lesions: improved recognition with use of the polymerase chain reaction. Clin. Infect. Dis. 20:842848. Stanway, G., N. Kalkkinen, M. Roivainen, F. Ghazi, M. Khan, M. Smyth, O. Meurman, and T. Hyypia. 1994. Molecular and biological characteristics of echovirus 22, a representative of a new picornavirus group. J. Virol. 68:8232 8238. Thoren, A., and A. Widell. 1994. PCR for the diagnosis of enteroviral meningitis. Scand. J. Infect. Dis. 26:249254. Wakeeld, A. J., J. D. Fox, A. M. Sawyer, J. E. Taylor, C. H. Sweenie, M. Smith, and V. C. Emery. 1992. Detection of herpesvirus DNA in the large intestine of patients with ulcerative colitis and Crohns disease using the nested polymerase chain reaction. J. Med. Virol. 38:183190. Wilborn, F., V. Brinkmann, C. A. Schmidt, F. Neipel, H. Gelderblom, and W. Siegert. 1994. Herpes virus type 6 in patients undergoing bone marrow transplantation: serologic features and detection by polymerase chain reaction. Blood 83:30523058.

17. 18. 19.

22.

23. 24.

20.

25.

21.