Tuberculous Meningitis: Pathogenesis, Immune Responses, Diagnostic Challenges, and the Potential of Biomarker-Based Approaches

Tuberculous meningitis (TBM) is the most devastating form of tuberculosis (TB), causing high mortality or disability. Clinical management of the disease is challenging due to limitations of the existing diagnostic approaches. Our knowledge on the immunology and pathogenesis of the disease is currently limited. More research is urgently needed to enhance our understanding of the immunopathogenesis of the disease and guide us toward the identification of targets that may be useful for vaccines or host-directed therapeutics.

T uberculosis (TB) is the leading cause of death from a single infectious agent (Mycobacterium tuberculosis) and killed nearly 1.5 million people in 2018 (1). TB mostly manifests as a pulmonary disease but also affects other body sites, causing extrapulmonary TB (EPTB). About 5% of all EPTB cases are tuberculous meningitis (TBM), which results from the spread of M. tuberculosis into the meninges and cerebrospinal fluid (CSF) (2). It is unclear what proportion of all TB cases are TBM, as it varies across studies by local TB prevalence, with high proportions (about 10%) suggested in high TB burden settings compared to low TB prevalence settings (around 1%) (3). It is estimated that at least 100,000 individuals develop TBM annually (3). TBM is the most devastating form of TB and continues to cause high morbidity and mortality (4), with an estimated 50% of patients dying or suffering neurological sequelae and complications (5,6). TBM is mostly common in young children (2 to 4 years old) and individuals infected with HIV (4,7). Besides TBM, infectious meningitis is also commonly caused by viruses, bacteria, and fungi, which are often challenging to differentiate from meningitis caused by TB (8). In both children and adults, viral meningitis is more common, followed by bacterial and fungal meningitis (9)(10)(11). Streptococcus pneumoniae is the most common cause of bacterial meningitis worldwide in both adults and children, followed by Neisseria meningitidis (8,12). TB (22%) was reported as the most common form of bacterial meningitis in children from a high TB burden setting, followed by Streptococcus pneumoniae (4%) and Klebsiella pneumoniae (3%) (11). The diagnosis of TBM is challenging and often delayed, with deleterious outcomes for patients. These challenges are even more serious in very young infants (13).
The methods currently used for diagnosing TBM in children are unreliable. Symp-In children, clinical signs may include initial apathy or irritability that progresses to meningism, signs of raised intracranial pressure (such as abducens nerve palsy), and focal neurological signs (29). In adults, clinical signs may include neck stiffness, cranial nerve palsies (cranial nerve III, IV, VI, and VIII), confusion, and coma (29). Clinical motor deficits (monoplegia, hemiplegia, or paraplegia) occur in about 10% to 20% of cases (29,30). Death is invariably inevitable if TBM is not treated.
Immune response. The host inflammatory response plays an important role in TBM pathology (31). Within the CNS, microglia are resident macrophages and are arguably the most prominent immune effector cells responsible for recognition and internalization of M. tuberculosis (32). Microglial cells and migrated infected neutrophils and macrophages become rapidly activated and can proliferate and increase the expression of different molecules and secrete cytokines and chemokines, which in turn modulate immune responses within the CNS (33). Such cytokines and chemokines released by infected microglial cells include TNF-␣, IFN-␥, interleukin-6 (IL-6), IL-1␤, CCL2, CCL5, and CXCL10 (34). The cytokines and chemokines may disrupt the BBB, allowing the influx of other uninfected cells (monocytes, neutrophils, and lymphocytes) (30). Infected cells (microglial cells and migrated cells) together with migrating uninfected cells lead to the formation of Rich focus.
Following the rupture of the Rich focus, release of M. tuberculosis into the subarachnoid space elicits a local T cell-mediated response characterized by caseating granulomatous inflammation (35). The role of T cells in TBM has been supported by several studies. The predominance of ␣␤T cells and NK cells in the CSF of TBM patients was associated with better survival (36). Whole blood transcriptome analysis of children with TBM at different time points demonstrated that reduced T cell proliferation and immune responses is associated with disease progression (37). A recent study reported a similar reduction pattern; however, this study did not assess the changes over time (31).
Inflammatory mediators (cytokines and chemokines) including TNF-␣, IFN-␥, IL-1␤, IL-6, IL-8, and IL-10 are increased in the CSF of patients with TBM (17,(38)(39)(40). TNF-␣ has been linked to a protective role against M. tuberculosis, through the formation of granulomata. Studies on rabbit models of TBM have demonstrated that high levels of TNF-␣ in CSF were associated with worse outcomes (41). The use of TNF-␣ antagonists in combination with antibiotics improved survival and outcomes in rabbits (42). In contrast, CSF TNF-␣ levels of children treated for TBM did not show a significant decline over a 4-week period (43).
Other host mediators implicated in the pathology of TBM include matrix metalloproteinases (MMPs) and vascular endothelial growth factor (VEGF). Elevated levels of MMP-9, MMP-2, tissue inhibitor of metalloproteinase 1 (TIMP-1), and TIMP-2 were reported in CSF samples of pediatric patients with TBM (44). MMP-9 levels decreased significantly early in treatment; however, there was an increase during hospital stay, which was associated with better outcomes (44). MMP-2 and MMP-9 may be involved in the pathology of TBM due to their key role in the disruption of BBB and the BCSF barrier by breaking down the extracellular matrix that constitutes the barriers, which may in turn cause brain edema, tissue damage, and migration of blood-derived inflammatory cells (45). Therefore, the inhibition of MMPs, as proposed in one study, may be a potential strategy for management of complications seen in TBM (45). VEGF is a potent factor of vascular permeability and angiogenesis (46), is vasculotoxic, is prothrombotic, reduces cerebral blood flow, and produces nitric oxide as well as oxygen free radicals (47). In TBM, VEGF has been associated with brain edema and the disruption of BBB (48), with its induction being mediated by TNF-␣ (49).
In a recent study, transcriptional responses of pediatric TBM at a systemic level revealed upregulated innate cell populations, such as neutrophils, macrophages, resting dendritic cells, and plasma B cells, in the blood (31). Similarly, van Laarhoven et al. (36) observed increased numbers of mature neutrophils and classical monocytes in the blood of adult TBM patients versus healthy controls. Gene expression associated with CD4 and CD8 T cells was more predominant in healthy controls, while genes associated with T cell activation and their signaling were significantly downregulated in the TBM cases (31). Furthermore, TBM was also associated with proinflammatory inflammasome signaling pathways (31). This could suggest that innate and inflammasome responses play an important role in TBM and that reduced T cell response is associated with the disease. Although the current findings suggest a role for neutrophils and other innate cells in TBM immunopathogenesis, the roles of different immune cell subpopulations, including the mucosa-associated invariant T (MAIT) cells, a T cell subset that displays innate-like characteristics and which have only recently been described in the CSF of TBM patients (36), remain unclear. Further investigations are needed to assess the frequencies, characteristics, and responsiveness of immune cells in patients with TBM in order to understand the host responses underlying the disease pathology.

CURRENT DIAGNOSTIC APPROACHES FOR TBM
Clinical diagnostic criteria. Owing to the inadequate performance of microbiological tests, there have been attempts to establish clinical diagnostic criteria for the diagnosis of TBM based on a combination of all the evidence from medical history, clinical assessment, and other relevant investigations (including CSF investigations and neuroimaging) (13). Despite numerous efforts to create clinical prediction rules to differentiate TBM from other meningitis based on these tests, standardized diagnostic criteria are still lacking. Thwaites et al. proposed a scoring system for the diagnosis of TBM in adults on the basis of clinical and basic laboratory findings (18 (17). This uniform research case definition classifies patients as "definite," "probable," "possible," and "not TBM" and is based on a composite score of clinical findings, CSF findings, neuroimaging, evidence of TB elsewhere, and exclusion of alternative diagnosis (17). It is important to note that the uniform research case definition criteria were not designed for use in clinical practice. Hence, caution is needed when applying the research case definitions for TBM patient care (50).
Microbiological diagnosis. Smear microscopy is the most widely used rapid and inexpensive diagnostic test for TB; however, staining of CSF smears for acid-fast bacilli has poor sensitivity (about 10% to 15%) (51). Smear microscopy is therefore not reliable for the diagnosis of TBM. Mycobacterial culture, the gold standard for the diagnosis of TB, is recommended by the World Health Organization (WHO) for use in both adults and children, including for TBM. Although culture has a higher sensitivity (about 50% to 60%) for the diagnosis of TBM than that of other TB tests, its turnaround time (up to 8 weeks with solid media such as Lowenstein-Jensen [25,51]) is a limitation. Although automated systems such as Bactec MGIT 960 have shown reduced average time to yield results (18 days versus 38 days), clinicians cannot afford to wait for culture results before treating patients, as death is a distinct possibility if empirical therapy is not initiated (30,52). That notwithstanding, M. tuberculosis culture is still important for recovery of the bacilli needed for downstream phenotypic drug susceptibility testing (DST), as well as epidemiologic and sequencing-based studies.
Molecular tests. To overcome the limitations of the conventional laboratory diagnostic approaches, commercial nucleic acid amplification tests (NAATs) have emerged. These tests have the advantage of rapidity while simultaneously detecting drug resistance. In a recent systemic review and meta-analysis of 18 studies, NAATs were shown to provide better performance with pooled sensitivity and specificity of 96% and 92%, respectively (53). However, the diagnostic accuracy of NAATs is different depending on the specimen type, with respiratory specimens associated with better accuracy (53). These tests are therefore not reliable for ruling out TB from nonrespiratory specimens due to the lower sensitivity. For the diagnosis of TBM specifically, a systematic review and meta-analysis on NAATs reported a pooled sensitivity of 82% and specificity of 99% against culture (54) and sensitivity and specificity of 68% and 98%, respectively, against a composite reference standard (54). In line with these results, another meta-analysis reported a lower sensitivity of 64% but high specificity of 98% against CSF M. tuberculosis culture for commercial NAATs (55).
The Xpert MTB/RIF test (Cepheid, Sunnyvale, CA, USA), arguably the game changer, was developed for the rapid diagnosis of TB. It is an automated closed-cartridge system that allows the rapid (within 2 h) detection of both M. tuberculosis and rifampin (RIF) resistance simultaneously. The WHO recommends the GeneXpert test for the diagnosis of EPTB, including TBM with CSF specimen in both adults and children. The sensitivity of the GeneXpert test for TBM ranges from approximately 50% to 60% (56), with various performances reported. A study from Uganda reported that sensitivity improved from 28% to 72% when larger volumes (6 ml) of concentrated CSF were used, compared with 2 ml of uncentrifuged CSF (56). In another study, the GeneXpert test showed an overall sensitivity of 59.3% compared with clinical diagnosis (based on uniform case definition [17]) in adult TBM suspects (57), with another meta-analysis (1 retrospective study and 4 prospective studies) reporting pooled sensitivity and specificity values of 70% and 97%, respectively (55).
Xpert MTB/RIF Ultra (Xpert Ultra) was developed to overcome some of the shortcomings of the initial Xpert test, including the inadequate sensitivity. In a recent prospective cohort study, Xpert Ultra demonstrated a sensitivity of 70% for probable or definite TBM (diagnosed based on uniform case definition [17]) compared with 43% obtained by either Xpert or culture (58). Compared with either Xpert or culture, Xpert Ultra diagnosed TBM in HIV-infected adults with a sensitivity of 95% (21/22 cases) (58). Although Ultra demonstrated improved performance, it does not appear to be adequate to rule out TBM due to concerns over low negative predictive value (NPV).
Another promising NAAT which will potentially be suitable for use in resourcelimited settings, as it does not require expensive instruments or expertise and yields results in 60 minutes, is the loop-mediated isothermal amplification (LAMP) test (59). When investigated as a diagnostic test for TBM, LAMP showed potential with sensitivity between 88% and 96% and specificity of 80% to 100% (59,60). Another NAAT that is commercially available (the Amplicor TB PCR test) has a sensitivity of ϳ40% and specificity of ϳ90% to 100% in the diagnosis of TBM, as reported in a recent review (14). The requirements of trained laboratory staff and high costs limit the wide use of the test (14). The Gen-probe amplified M. tuberculosis direct test (MTD), another NAAT, was initially intended for the detection of M. tuberculosis in respiratory specimens. When CSF samples were used in the test, it showed potential in the diagnosis of TBM, with pooled sensitivity of 86% and specificity of 99% (54). Another commercial NAAT, the Genotype MTBDRplus is a molecular line probe assay that targets specific genes for M. tuberculosis complex detection, as well as rifampin (RIF) and isoniazid (INH) susceptibility (61). When evaluated in a few TBM cases, the sensitivity of the test was 33%, with specificity of 98% against a clinical reference standard (61). While most NAATs have shown potential as diagnostic tests for TBM, more data on the performance of the tests are still required. Overall, commercial NAATs have demonstrated high specificity but, generally, suboptimal sensitivity for TBM, whereas Xpert Ultra, although promising, can still miss up to 30% of TBM patients. Despite improved diagnostic performance, both the GeneXpert and Xpert Ultra tests cannot rule out TBM due to their low negative predictive value (51,62). Although NAATs are a major diagnostic advance, they are still inadequate to replace culture methods. The CRISPR-M. tuberculosis and metagenomic next-generation sequencing (mNGS) technologies (reviewed in reference 63), may improve the detection of M. tuberculosis in CSF samples with low bacillary load. However, further investigations are required to ascertain the performance of these methods.
Brain imaging. Brain imaging techniques, such as computed tomography (CT) and magnetic resonance imaging (MRI), are part of the clinical diagnostic assessment of TBM. Contrast-enhanced CT imaging reveals basal meningeal exudates specific for TBM and predicts poor outcomes (30). However, neurological signs or features (such as infarcts and hydrocephalus) revealed by CT imaging lack diagnostic specificity for TBM, mainly because similar features are seen in other infectious and noninfectious diseases (30). MRI has been found to have superior diagnostic abilities compared with CT (64). This includes better detection of basal meningeal enhancement and infarcts (especially in the brain stem) and early infection (64). The main common limitation to these brain imaging techniques is that CT scans are normal in about 30% of individuals at an early stage of TBM, while MRI scans are normal in about 15% (30). Furthermore, both CT and MRI evaluations are usually carried by medical experts in a tertiary care setting and are mostly not available in primary care settings or resource-limited settings (MRI more so than CT) (65).
Immune response-based diagnosis. Immunodiagnostic tests, such as interferon gamma release assays (IGRAs), are primarily used for the diagnosis of M. tuberculosis infection. IGRAs measure the IFN-␥ produced by lymphocytes when stimulated with M. tuberculosis-specific antigens. As these tests cannot differentiate latent TB infection (LTBI) from active TB disease, their use in the diagnosis of active TB is discouraged (66). The WHO recommends the use of IGRAs for LTBI testing in individuals who are at risk, including people living with HIV and infants and children aged 5 years and younger who are household contacts of pulmonary TB patients in both low-and high-TB burden settings (66). In TBM, a moderate diagnostic accuracy was reported for CSF IGRA, with sensitivity of 77% and specificity of 88% (67). The most commonly used immunodiagnostic modality in TBM management is the measurement of adenosine deaminase (ADA). ADA is an enzyme that is released by lymphocytes and plays an important role in the proliferation and differentiation of T cells. As the release of ADA from T cells has been associated with cell-mediated immune responses to tubercle bacilli, the measurement of levels of the enzyme in CSF is being done as an approach for the diagnosis of TBM (68)(69)(70). A meta-analysis of 20 studies on the accuracy of ADA reported a pooled sensitivity of 89% and specificity of 91% in the diagnosis of TBM (71). However, evidence about the clinical usefulness of CSF ADA is contradictory. Ekermans and colleagues reported that at an optimal cutoff of 2.0 U/liter, the sensitivity and specificity of CSF ADA was 85.9% and 77.7%, respectively (72). The study further showed that an optimal cutoff value for the routine diagnosis of TBM could not be established, as many cases were missed (72). Furthermore, high numbers of false positives and limited utility were reported for CSF ADA in another study on HIV-infected individuals (73). The main concern with the use of ADA in practice is the fact that similar levels of the protein have been documented in CSF from patients with other infective pathologies, including bacterial meningitis and ventriculitis, thereby making interpretation of the results difficult. ADA is therefore not useful in settings where the differential diagnosis is broad (72). As CSF ADA results may be misleading, clinicians should be aware of its limitations when making TBM diagnostic decisions. The poor standardization of ADA assays and the fact that ADA results are dependent on the integrity of the specimen (72) are further concerns. In summary, none of the currently available methods is adequate for use as a stand-alone test for the diagnosis of TBM. New and improved diagnostic methods are therefore urgently needed.

POTENTIAL OF NOVEL BIOMARKER-BASED APPROACHES
In the search of better TB diagnostic tools, recent studies have investigated several alternative approaches, including the measurement of protein concentrations in biological fluids, transcriptional molecules, and metabolites as biomarkers for TB. Several attempts are being made at detecting such biomarkers in easily obtainable specimens, such as blood, urine, and saliva, among others, with the need for such nonsputumbased approaches being deemed a high priority by the WHO (74).
Host protein biomarkers. Several studies have proven that the measurement of inflammatory proteins, such as cytokines, chemokines, acute phase proteins, and growth factors, can differentiate TB from other infections (reviewed in reference 75). Earlier studies evaluated the value of alternative proteins other than IFN-␥ that were detected in supernatants following the stimulation of blood cells with M. tuberculosisspecific antigens using multiplex immunoassays, mainly the Luminex platform (76-78).
As such studies were based on overnight stimulation assays, most of the recent studies have focused on the evaluation of host markers in unstimulated specimens, including serum (79), plasma (80), urine (81), and saliva (82), given that such biomarkers may be more easily translated into point-of-care tests.
Despite the promise shown by the CSF host inflammatory biomarker-based studies described above, the procedure for the collection of CSF (lumbar puncture) may be a limitation in the implementation of CSF-based tests in resource-limited settings. Bloodbased biomarkers, which have shown potential and are being developed into pointof-care tests for the diagnosis of pulmonary TB (79,89), may be alternatives in TBM diagnosis. When evaluated as a tool for the diagnosis of TBM in children, a modified version of an adult serum seven-marker signature (C-reactive protein [CRP], IFN-␥, IP-10, CFH, Apo-A1, SAA, and NCAM) diagnosed TBM with moderate accuracy (sensitivity of 73.9% and specificity of 66.7%) (90). In the same South African study, a new threemarker serum biosignature (adipsin, A␤42, and IL-10), which diagnosed childhood TBM with sensitivity of 82.6% and specificity of 75.0%, alongside several individual candidate biomarkers were identified (90). Given the potential shown in these studies, there is a need for further discovery and validation of similar biomarkers, followed by incorporation of the most promising candidates into point-of-care tests. The added benefit of blood-based biomarkers is the possibility of detecting them in fingerprick blood, as is currently being done for new pulmonary TB-based prototype tests (https://www .triagetb.com/).
Host transcriptional biomarker-based signatures. Transcriptomics has become a popular approach for biomarker discovery, with several infectious diseases, including TB biosignatures, being discovered using techniques such as RNA sequencing, quantitative real-time PCR, and microarrays. It is suggested that quantifying the shifts in RNA abundances triggered by diseases could help in identifying diagnostic, disease-associated, and treatment response biomarkers. Recent studies have identified gene signatures that predict the onset of active TB several months before the onset of symptoms (91), signatures for the prediction of progression from latent TB infection to active TB in household contacts (92), diagnosis of TB (93,94), and monitoring of TB treatment response (95). Most of these investigations have been adult, pulmonary TB-based studies, with one study reporting on the up-, or downregulation of 796 genes (398 and 398, respectively) in brain tissues of TBM patients who were coinfected with HIV (96). Of importance, four gene products, namely, glial fibrillary acidic protein (GFAP), serpin peptidase inhibitor clade A member 3 (SER-PINA3), thymidine phosphorylase (TYMP/ECGF1), and heat shock 70 kDA protein 8 (HSPA8), were confirmed to be abundant in TBM patients with HIV coinfection (96). As this study compared TBM patients with individuals who succumbed to road traffic accidents, the utility of these genes as candidate TBM diagnostic biomarkers is unknown. In addition to evaluating the usefulness of these genes as TBM diagnostic candidates, further work is required for the identification and validation of TBM-specific biosignatures in well-designed TBM diagnostic studies. Given that prototype fingerprick blood-based mRNA signature tests currently exist (97), validated TBM transcriptomic biosignatures could be further incorporated into such platforms, followed by large-scale field trials against acceptable reference standards.
Host miRNA biosignatures. MicroRNAs (miRNAs) are a class of conserved noncoding small RNAs (21 to 25 nucleotides long), which play an important role in the regulation of gene expression and other biological processes, including cell proliferation, cell differentiation, organ development, apoptosis, immune response, angiogenesis, and onset of disease (98,99). Altered expression of miRNAs has been associated with TB (100). In a study including 112 children with TBM and 130 healthy controls, miR-29a expression in peripheral blood mononuclear cells (PBMCs) showed potential in the diagnosis of TBM, with a sensitivity of 67.2% and specificity of 88.5 and a sensitivity of 81.1% and specificity of 90% when evaluated in CSF (101). When used in combination, CSF plus PBMC miR-29a expression diagnosed pediatric TBM with sensitivity of 84.4% and specificity of 95.4% (101).
Metabolic biosignatures. Metabolomics is an emerging powerful and advanced omics platform, which may be useful in the identification of novel diagnostic biomarkers. The technique is used to identify metabolites that are associated with certain physiological or pathological conditions. Several studies have demonstrated significant differences in amino acids and energy metabolism in CSF samples of TBM patients compared with other groups, including patients with viral, bacterial, and cryptococcal meningitis (103)(104)(105)(106). However, the diagnostic accuracy of the metabolites identified was not reported. In a study that investigated urine metabolic biomarkers in 12 children with TBM and 29 controls, a host biosignature (SUM-4) generated from the sum of urinary concentrations of methylcitric acid, 2-ketoglutaric acid, quinolinic acid, and 4-hydroxyhippuric acid, separated TBM from other groups with an area under the receiver operator characteristic (ROC) curve (AUC) of 96.6% (107). These largely small, proof-of-concept studies provide evidence that host metabolomic biomarkers may be useful in the diagnosis of TBM. However, further work is required in this field, coupled with work focusing on the development of end user-friendly detection devices for the measurement of any candidate metabolites, preferably at the point of care.

CONCLUSIONS
The diagnosis of TBM remains challenging, mainly due to difficulties in the direct detection of M. tuberculosis bacilli in CSF and other specimens from patients who are suspected of having the disease. The recently introduced Xpert MTB/RIF Ultra has shown promise in detecting paucibacillary TB, including diagnosing more TBM cases than the previous version. However, it may still fail to rule out TBM due to inadequate negative predictive value. Furthermore, its implementation in resource-limited settings will be hampered by the same issues that hindered the successful roll out of the GeneXpert MTB/RIF test in such settings (108,109). None of the currently available diagnostic tools is adequate as a stand-alone method for the definite diagnosis of TBM. Therefore, CSF microscopy, mycobacterial culture, and molecular tests, such as GeneXpert, Xpert Ultra, and other NAATs, should all be performed for the diagnosis of TBM, in settings where this is possible. The TB field has recently seen much activity in the discovery, validation, and development of novel biomarker-based tests, but most of this activity is for the management of pulmonary TB, especially in adults. The few host biomarker-based projects that have focused on TBM have shown that the targets described in the WHO target-product profiles (TPPs) for a nonsputum biomarker test for the diagnosis of extrapulmonary TB (sensitivity of at least 80% in CSF samples for microbiologically confirmed TB and specificity of 98%, or as specific as the Xpert MTB/RIF) (74) may be achievable or approachable given the recent promising findings (88,90,99,101,102). However, much work is still required in the refinement and validation of the proposed biomarkers. Furthermore, such biomarker-based approaches will make the most impact only if further developed into easy-to-use diagnostic tests, especially tools that are implementable at the point of care, in resource-limited settings.

DIRECTIONS FOR FUTURE RESEARCH
• Traditionally, TB diagnostic tests have been developed and largely validated in adults with pulmonary TB. Research focusing on the development and validation of tools in children and especially tests that may be useful in the diagnosis of EPTB, including TBM, should be encouraged.
• Assessment of biomarkers produced after stimulation of blood cells with M. tuberculosis antigens (110,111) has been shown to possess diagnostic potential in other extrapulmonary forms of TB (112)(113)(114). Such approaches should be explored in other EPTB types, including TBM.
• Approaches that have resulted in potentially useful pulmonary TB signatures in noninvasive and easily obtainable specimens, including saliva (82,115,116) and urine (81,117), require evaluation in TBM and should be encouraged.
• Further work needs to be done to validate the different inflammatory host biomarker signatures reviewed in the current study (Table 1). Such future work should focus on the following: X Evaluation of the signatures in large independent cohorts of both adults and children that are recruited after clinical suspicion of having meningitis, prior to the confirmation of TBM or no TBM.
X Evaluation of the influence of HIV infection and other comorbidities on biomarker accuracy. X Inclusion of participants from different geographical areas in the evaluation and validation of biomarkers, such as to assess global applicability.
X Incorporation of the most promising globally relevant biosignatures into pointof-care tests, followed by field trials of the tests in multiple settings.
• While validation of the few transcriptomic, metabolomic, and miRNA candidate biomarkers that have so far been identified (Table 1) is ongoing, further work, including new biomarker discovery in new, well-designed TBM studies in which controls are individuals suspected of having TBM as would be obtained in routine clinical practice, is encouraged.
• If the goal of developing a useful TBM biomarker-based point-of-care test remains elusive, the inclusion of different biomarker-based tests in a uniform research omics-based case definition may be beneficial. A uniform research omics-based case definition could be based on different blood and CSF validated biomarker signatures/omics-based tests, which may diagnose TBM with optimal accuracy when combined.
• To further enhance our knowledge of the immunology and pathogenesis of TBM ( Fig. 1), further investigations of the immune cell populations, characteristics, and responses at both the site of disease (CSF) and peripheral (blood) in TBM patients and appropriate controls should be encouraged. Such knowledge may shed light on new potential vaccine and host-directed therapeutic targets.
• Following the development of new tools for TBM, evaluation of the accuracy of the tools against appropriate benchmarks, e.g., the TPPs proposed by the WHO, is encouraged (74).

ACKNOWLEDGMENTS
This review was written as part of an EDCTP2 project, supported by the European Union (grant number TMA2018SF-2470-TBMBIOMARKERS). The project was also made possible through funding by the South African Medical Research Council through its Division of Research Capacity Development under the Internship Scholarship Program, from funding received from the South African National Treasury.
The views and opinions of authors expressed herein do not necessarily state or reflect those of the EDCTP and the SAMRC.