A total of 131 CSF samples were collected during 2006–2008 from adults with unexplained meningitis or encephalitis. The patients in this study were 18–86 years of age (median 39 years); the male-to-female ratio was 2.4:1. Fourteen percent of patients had diabetes, and 8% were HIV positive. In addition to fever and headache, mental status was abnormal for 22% of patients; 2% had paralysis/seizure.

Two patients with unexplained encephalitis died. Samples were analyzed by applying a staged molecular approach, including multiplex MassTag PCR, 16S rRNA-gene PCR, DNA microarray, and high-throughput sequencing (Figure). We identified 31 (24%) pathogens in 131 CSF samples from patients with meningitis and encephalitis of unknown cause. Candidate pathogens were identified for 25 (27%) of 94 patients with unexplained meningitis and 6 (16%) of 37 with unexplained encephalitis (Table 3). EBV (16 cases) accounted for most of the identified pathogens in unexplained meningitis cases, followed by Escherichia coli (5 cases), enterovirus (2), HHV-2 (1), and M. tuberculosis (1). Herpesviruses were identified for unexplained encephalitis cases and includedvaricella-zoster virus (3 cases), HHV-1 (2), and cytomegalovirus (1). We also tested 3 CSF samples from patients with noninfectious CNS disease as negative controls; no agents were identified in these samples.

Five bacterial pathogens were identified by 16S rRNA PCR (E. coli, 5 patients), and 1 HHV-1 was identified by microarray analysis. We pursued unbiased high-throughput pyrosequencing of RNA from CSF by pooling the remaining negative samples (screened previously by MassTag PCR and microarrays); however, no specific pathogen was identified by pyrosequencing. All positive samples detected by MassTag PCR, 16S rRNA PCR, and microarray were verified by targeted singleplex PCR amplification and sequencing.

Of 16 EBV-positive patients, 9 (56%) were HIV infected. To differentiate EBV latent infection from lytic infection, we further measured EBV-specific antibodies in EBV DNA–positive CSF samples; in 8 cases, paired CSF and serum samples were available for analysis (Table 4). IgG against EBV VCA was found in 11 (69%) of 16 EBV DNA–positive CSF samples; however, none of the CSF samples were positive for VCA IgM, and only 1 CSF sample was positive for both EBV early antigen and nuclear antigen, indicating current or recent EBV infection. All 8 serum samples, which were paired with 8 of the CSF samples, were positive for VCA IgG and EBV nuclear antigen but negative for VCA IgM and EBV early antigen, suggesting that latent EBV infections existed in those patients.

We have described results of a comprehensive, staged molecular analysis of unexplained encephalitis and meningitis conducted by using CFS samples obtained over a 3-year period in a major referral hospital in Taiwan. Our investigation confirmed the presence of microbial sequences in 31 (24%) of 131 CSF samples; 25 were identified by MassTag PCR.

We detected EBV in 16 samples. EBV is a ubiquitous HHV that infects 90% of adults worldwide (22). In Taiwan, too, most persons are infected with EBV in early childhood (23). VCA was detected in all EBV-positive CSF samples. This finding is consistent with EBV-related CNS disease. In a study of 5 patients with mononucleosis, all had EBV DNA and EBV-specific antibodies in the CSF during the acute phase of disease in association with neurologic manifestations but not during convalescence (24). In contrast, in the same study, EBV DNA and EBV antibodies were not detected in the CSF of 17 EBV-seropositive patients with other CNS infections, such as mumps, meningitis, and rubella encephalitis (24).

It could be argued that any inflammatory process has the potential to carry EBV into the CSF as a passenger in infiltrating lymphocytes; however, PCR results for EBV in CSF are rarely positive for patients with other CNS infections (25). One study showed that only 11 of 2,233 specimens from 2,162 patients were EBV positive by PCR (26). In another study, EBV was not detected by PCR of CSF from patients with bacterial meningitis, despite the presence of more lymphocytes/monocytes in the CSF of control patients with bacterial meningitis than in the CSF of the EBV-positive patients (27). In accord with these results, we did not detect EBV in the CSF of patients with CNS infections of bacterial origin, such as M. tuberculosis and E. coli.

All of the 16 EBV-positive patients in our study had meningitis; none had encephalitis. EBV is an uncommon cause of meningitis (0.9%) (28) and of encephalitis (2.3%) (29). However, given the data reported here, EBV-associated CNS infections in immunocompromised patients may be underestimated. In our study, 56% of EBV-positive patients were HIV infected. EBV has been reported to be associated with primary CNS lymphomas in HIV-infected persons (30,31), and EBV DNA can be detected in the CSF of 80%–100% of HIV-infected patients with primary CNS lymphomas (32–34); however, EBV also may be detected in the CSF of 12% of HIV-infected patients with neurologic symptoms but without lymphoma (35). None of the EBV-positive patients in our study had evidence of CNS lymphoma by contrast computed tomography or MRI of the brain. Because EBV is a ubiquitous virus, the incidence of CNS infections caused by EBV may be underestimated. Our results suggest that EBV should be considered for all patients with CNS infections, especially in immunocompromised patients.

We used 16S rRNA gene amplification to screen for any bacteria in CSF before analysis by microarray and pyrosequencing. E. coli was detected in 5 samples that had been culture negative in previous analysis, most likely because of prior treatment with antibacterial drugs, which can confound the results of bacterial culture. Indeed, CSF can rapidly become sterile for bacteria after antibacterial treatment, with meningococci and pneumococci becoming noncultivable within 2 and 4 hours, respectively (36). E. coli meningitis is a common cause of neonatal meningitis (37), but it is a rare cause of acute bacterial meningitis in adults (38). In Taiwan, E. coli meningitis has been reported in adults with chronic underlying conditions (39). Among 5 patients identified with E. coli in our study, 3 had underlying clinical conditions, including pulmonary tuberculosis (2 patients) and HIV infection (1).

In our study, most of the candidate pathogens were identified by the multiplex MassTag PCR and 16S rRNA PCR. Although this finding may raise questions about the utility of more expensive and sophisticated techniques, such as high-throughput pyrosequencing for detecting CNS infections, the unique potential of high-throughput pyrosequencing has been demonstrated by its implication of an arenavirus in transplant-associated encephalitis (16) and a novel astrovirus in XLA-linked agammaglobulinemia-associated encephalitis (17). This approach may facilitate pathogen discovery for patients with CNS infections of unexplained cause as it becomes more popular in molecular diagnostics.

Our study has some limitations. First, because samples were collected from only 1 referral hospital, our results may not reflect the case distribution of CNS infections in Taiwan but instead might represent some of the more difficult-to-diagnose cases. Second, CSF samples were stored for 3 years after collection, and nucleic acid degradation, particularly of RNA, might have occurred before laboratory analysis, which may explain the low number of RNA viruses detected. Third, we cannot rule out misclassification of encephalitis with noninfectious causes, such as the newly described immune-mediated anti-N-methyl-D-aspartate receptor encephalitis (40), for which we did not test.

Rapid and accurate identification of the causative agent of a CNS infection can affect clinical management of individual patients. On the scale of populations, agent identification is crucial for determining the incidence of CNS infections caused by specific agents, enabling prioritization of targets for public health intervention and to prevent outbreaks of disease.

Our study confirms the power of multiplex MassTag PCR as a rapid diagnostic tool for identifying pathogens for patients with CNS infections and shows that viral and bacterial pathogens were detected in CSF from patients with CNS infections of unidentified cause. Additionally, the staged molecular approach incorporating complementary tools may enable detection of pathogens for patients with CNS infections of previously unrecognized causes, which would otherwise be missed. This approach may aid in explaining the observed worldwide high proportion of CNS infections of unknown cause.