The Performance of Metagenomic Next-Generation Sequencing In Suspected Central Nervous System Infection

Background: To explore the performance of metagenomic next-generation sequencing (mNGS) technology in patients of suspected central nervous system infection. Methods: From January 2018 to March 2021, 75 cases were enrolled in this retrospective analysis at Hunan Provincial People’s Hospital. The clinical data of patients with suspected central nervous system infection who underwent cerebrospinal uid mNGS were analyzed. The performances of mNGS were compared with the conventional methods. Result: The sensitivity of mNGS, culture and smear in the diagnosis of 75 patients were 55%, 4.4%, 6.7%; theirs’specicity were 54.3%, 100%, 100%; theirs’ positive predictive value (PPV ) were 57.9%, 100%, 100%; theirs’negative predictive value (NPV) were 51.4%, 41.4%, 41.7%, respectively. There was 41(54.6%) cases whose mNGS results were consistent with the nal diagnosis. 22(29.3%) mNGS results were considered as both mNGS positive/Case consistent; 19( 25.3%) mNGS results were considered as both mNGS negative/Case consistent; 18(24%) mNGS results were considered as both mNGS positive/Case inconsistent; 16(21.3%) mNGS results were considered as both mNGS negative/Case inconsistent. mNGS identied 35 irrelavant pathogens in this study. Conclusion: mNGS showed a high sensitivity compared to conventional methods. There are still several challenges in clinical application. It is necessary to establish unied and effective standards for interpreting mNGS results.


Introduction
More than 100 known pathogens can cause central nervous system infection [1]. Viruses, bacteria, fungi, parasites, and amoebae are all the common pathogens [2]. Central nervous system (CNS) infection is a serious neurologic condition, but the etiology remains unknown in more than half patients [3]. So accurate etiology diagnosis is crucial for the successful treatment of central nervous system infection.
Conventional diagnostic test methods for CNS infections include cerebrospinal fuid (CSF) cell count, glucose, and protein measurements, CSF Gram staining, culture, biomarkers such as procalcitonin (PCT), creative reactive protein (CRP), serology and PCR detection. However, in the pathogens detection of CSF infection, traditional microbiological techniques such as culture, staining often have low sensitivity and PCR detection is also limited to the gene sequence of known pathogenic microorganisms [4]. mNGS, is a promising technique, offers a relatively unprejudiced diagnostic tool for all pathogens included in the database library in a single test, regardless of prior suspicions of candidate pathogens and without the need for isolation and culture [5]. As a novel diagnostic tool, mNGS has been used wildly for the identi cation of various pathogens from clinical samples such as tissues, CSF or plasma [6]. mNGS shows signi cant advantages in pathogen detection in meningitis [7], pulmonary infection [8,9], sepsis [10], and other types of diseases [11][12][13][14]. However, numerous challenges remain in the interpretation of mNGS results when applying mNGS test into clinical microbiology laboratory diagnosis [15].
In CNS infection, multiple studies have demonstrated the ability of mNGS for the detection of viruses, bacteria, fungi, parasites, and some uncommon pathogens from CSF or brain tissue, indicating its ability to identify pathogens in CNS infection of unknown etiology [16][17][18][19][20]. Our study further demonstrated the overall ability of mNGS in the rapid diagnosis of CNS infection caused by bacteria, viruses, fungi, and M. tuberculosis.

Study design and participants
The cases of patients with suspected central nervous system infection, who were admitted to Hunan Provincial People's Hospital from January 1, 2018 to March 31, 2021, were reviewed. Each registered patient whose CSF were sent for mNGS was enrolled. The nal diagnosis was adjudicated by a panel discussion following hospital discharge when the results of all tests and patients' responses to the antimicrobial therapy were available.
Meanwhile, clinical data of all enrolled patients, including CSF cell count, glucose, and protein measurements, CSF Gram staining, CSF culture, biomarkers such as white blood cell, procalcitonin (PCT), creative reactive protein (CRP), serology and PCR detection were collected. For bacteria (excluding mycobacteria), viruses, fungi, and parasites, whose read number was in the top 10 in the complete belonging list, mNGS identi ed a microbe(on species level) [21]. The result was considered MTB-positive when at least one speci c read was mapped to the species or genus level [22].

Statistical analysis
Sensitivity, speci city, positive predictive value, and negative predictive value of mNGS for the diagnosis of central nervous system infection were determined. Statistical analyses were performed using SPSS version 20.0 (IBM Corporation, Armonk, NY, USA). A p value less than 0.05 was considered statistically signi cant.
The demographic and clinical characteristics of 75 patients with infectious diseases and non-infectious diseases are summarized in Table 1. In the infectious diseases group, the CSF protein was higher than non-infectious diseases group (1027±165.6 vs 369.6±126.8, P=0.004). The CSF/serum glucose ratio is signi cantly lower than non-infectious diseases group (0.42±0.03 vs 0.59±0.04, P=0.0009). While CSF WBC, chlorine, LDH and CSF pressure showed no signi cant differences between two groups.
Overall, metagenomics sequencing showed satisfying pathogen detection rate compared to conventional methods. Culture reported 2 positive (one cryptococcal meningitis, one bacterial meningitis, smear reported 3 positive (one tuberculosis meningitis, two cryptococcal meningitis ), and mNGS detected 40 positive detection rate of mNGS was signi cantly higher than that of culture and smear in etiology detection.

Discussion
We retrospectively enrolled patients who suspected central nervous system infection during the period from January 2018 to March 2021, and attempted to analyze the performance of metagenomic next-generation sequencing. In this study, cerebrospinal uid were tested by both mNGS and conventional methods. The positive rate of mNGS of 55% was signi cantly higher than that of conventional methods. It is similar with the 27.9-60% that has been reported in the literature [1]. mNGS showed a high sensitivity compared to conventional methods. But the speci city, positive predictive value, and negative predictive value of mNGS were lower. The NPV of mNGS when compared with the nal diagnosis was 51.4%. This does indicate that a negative mNGS result cannot exclude CNS infection. Therefore, clinical evaluation remains important when mNGS results are negative.
Of note is that the speci city of mNGS in the diagnosis of tuberculous meningitis is 100%, which allows a negative mNGS test to be used as one of the diagnostic methods to exclude tuberculous meningitis. However, there was one case who were detected by smear but missed by mNGS in present study.
Despite mNGS has many advantages in detection pathogens, there are several challenges that might hinder its application in a clinical setting. At rst, although mNGS provides microbial information that is di cult to obtain by traditional methods. However, the criteria cannot de nitively determine which microbes are pathogenic because of the presence of contamination, background microorganisms, and the detection of circulating cell-free DNA from nonpathogenic microbes. It is necessary to establish uni ed and effective standards for interpreting the results.
The second challenge of mNGS is that the interpretation of mNGS results be rather confusing presently. It is unable to discriminate the pathogenicity status of the pathogens detected. Large amounts of information could be confusing and even misleading to physicians while making clinical decisions [27]. The third challenge is that samples may easily contaminated by environmental bacteria or human parasitic bacteria during sampling or sequencing, it is di cult to decide which pathogen is involved in infection, colonization, or contamination. Thus the results may interfere with the clinician's judgment. We believe that mNGS reports need to be interpreted very carefully, we have to interpreted it in combination with clinical ndings and all laboratory tests, preferably in a multidisciplinary manner [20,28]. Only those microorganisms consistent with the patient's clinical symptoms will be considered pathogens. Finally, the cost-effectiveness of the broad application of mNGS also needs further investigation [29].
Our study has some limitations. First, this was a retrospective study. As a retrospective study, limited data and data accumulation were not controlled by the researcher. It enrolled small cases. In further studies, we will enlarge the sample size to validate this diagnostic potential in prospective cohort study. Second, we used preliminary data for analysis but still lack of control group simultaneously in our cohort. Besides, the mNGS results and identi cation may be easily in uenced by many factors. At last, lack of a standard comparator for diagnostics, and classi cation bias were also the limitations of this study.
In conclusion, mNGS had a superior advantage in detecting potential pathogens than conventional methods in suspected central nervous system infection. There are several challenges in clinical application for mNGS. It needs establishing uni ed and effective standards for interpreting mNGS results.

Availability of data and materials
The datasets during and/or analysed during the current study are available from the corresponding author on reasonable request.

Ethics approval and consent to participate
All procedures performed in this study involving human participants were conducted in accordance with the ethical standards of the Ethics Committee of The Hunan provincial People's Hospital, the 1964 Helsinki declaration and its later amendments, and with comparable ethical standards.

Consent for publication
Informed consent was waived since the nature of retrospective study. None of the data could be traced back to an identi able patient. Tables   Table 1 Clinical Figure 1 The consistent proportion of mNGS and nal diagnosis