3.1 Characteristics of the study participants
287 patients were screened for inclusion in this study (Fig. 1). 11 patients were initially thought to have CNS infections and mNGS was performed. However, these cases were ultimately excluded following the final diagnosis of a non-infectious disease. Of these 11 excluded patients, 10 had negative mNGS results, and 1 patient receiving immunosuppressive therapy was positive for BK polyomavirus. The final cohort included 276 patients in the study. 176 (63.8%) were male and the median age was 42 years (IQR: 26–54 years). The median time from disease-onset to CSF sampling was 10 days (IQR: 5–25 days). The median white blood cell count in CSF was 80/mm3 (IQR: 19–220/mm3). The median CSF mononuclear cell count was 36/mm3 (IQR: 10–127/mm3). During a follow-up period of 30 days, nine patients died.
3.2 Performance of mNGS for diagnosing CNS infections
276 CSF samples were tested by mNGS and conventional microbiological studies. 122 samples were positive by mNGS (110 true positive, 12 false positive), 126 were positive by conventional microbiological tests, and 114 total positive results were considered ‘aetiologic pathogens’ (Table 2). All mNGS results were obtained in less than 48 hours and 101 CSF samples considered ‘aetiologic pathogens’ were positive by mNGS.
Of the patients diagnosed by mNGS, 16.3% of infections were caused by bacterial, 15.2% by viruses, 2.9% by fungi, and 2.2% by parasites (Fig. 2A).
In total, mNGS detected 11 bacterial species, of which M. tuberculosis (14 cases, 13.9%) and L. monocytogenes (7.9%) were the most common (Fig. 2B), 7 viral species (BK polyomavirus was not the etiologic pathogen), of which VZV (16.8%) and HSV1 (11.9%) were the most common, 2 fungal species, both of which were Cryptococcus (7.9%), and 2 parasitic species, of which T. solium (5.0%) was the most common. Nine co-infections with Epstein-Barr virus (EBV) (three with HSV1, two with Brucella, one with Cryptococcus, one with S. haemolyticus, one with P. aeruginosa, and one with M. tuberculosis), two co-infections with cytomegalovirus (CMV) (one with M. tuberculosis, and one with Cryptococcus), and one co-infection with BK polyomavirus (with HSV1) were detected. The EBV and BK polyomavirus did not appear to be consistent with the clinical manifestations in these two instances of co-infections.
3.3 False positive results of CSF mNGS
In the present study, false positives occurred in 12 (4.3%) patients and were primarily associated with bacterial infections (n=12; Table 2), including E. coli, E. faecium, A. baumannii, S. maltophilia, and P. aeruginosa, and a false positive for Brucella was also seen. Of note, the false-positive samples contained numerous other bacteria, that could be detected simultaneously by NGS. Using our proposed criteria, there were no false positives for viruses, fungi, or parasites. Although EBV was not the etiologic pathogen in most cases, it was present in the CSF of some patients. Additionally, there was some background contamination in most CSF samples (Supplementary Table 2) but these organisms did not meet the criteria for a positive result.
3.4 False negative results of CSF mNGS
In the present study, false negatives occurred in 16 (5.8%) patients (Table 2) and were associated with bacterial, viral and fungal infections. The false negative cases of bacterial infection were all treated with antibiotics prior to sequencing. In the false negative cases of viral infection, 1 or 2 SSRNs were detected in the samples but did not satisfy the proposed criteria for a positive mNGS result. If the criteria for a positive result was relaxed to a SSRN ≥ 1 (RPM ≥ 0.05), there were no false negative cases of HSV1 or VZV or false positive cases of HSV1, HSV2 or VZV. In this study, if we adopted the alternate criteria SSRN ≥ 1 (RPM ≥ 0.05) for viruses and Mycobacterium tuberculosis, there would be additional potential false positives, including 30 EBV, 7 CMV and 5 Mycobacterium tuberculosis infections. It should be pointed out that the possibility of Mycobacterium tuberculosis infection in the 5 cases cannot be ruled out based on the clinical and paraclinical manifestations, because the conventional microbiological methods might fail to detect the Mycobacterium tuberculosis. Of note, there were three false negative cases of Cryptococcus infection.
3.5 Illustrating cases of the study
Case 1: An old male with a medical history of gastric cancer presented with cognitive decline, psychiatric symptoms and seizures for 1 month. Brain MRI revealed a lesion in the left frontal lobe, and the first lumber puncture revealed no pleocytosis. Therefore, ’autoantibody-negative autoimmune encephalitis’ was suspected with the negative results of the antibodies panel. After the treatment with steroids and IVIg, the symptoms aggravated, and the repeated brain MRI showed diffuse cortical lesions in the left hemisphere. After referred to our centre, NGS of CSF was performed and detected abundant HSV 1, which was confirmed with PCR. He was diagnosed with herpes simplex encephalitis and responded well to acyclovir.
Case 2: An old male presented with right eye pain, memory deficits and psychiatric symptoms. Brain MRI showed a lesion in the right mesial temporal lobe. TORCH panel of CSF was positive for CMV IgM. Encephalitis caused by CMV was suspected. However, VZV was detected with NGS of CSF and confirmed with PCR. He was diagnosed with VZV encephalitis.
Case 3: A young male experienced hyperpyrexia and headache for 3 times in the past 10 years. He recovered quickly after the treatment with antibiotics every time without identifying the exact pathogen. Mollaret meningitis was suspected by the local hospital. When he presented with hyperpyrexia, headache and drowsiness 1 months ago, NGS of CSF was performed and detected Streptococcus pneumoniae. Further examinations were performed for the cause of recurrent purulent meningitis. At last, cerebrospinal fluid rhinorrhoea was found out and fixed.