Detection of Delta-like 1 ligand for the diagnosis of tuberculous meningitis: An effective and rapid diagnostic method

Tuberculosis is a devastating disease that causes over two million deaths every year.^1,2 Tuberculous meningitis (TBM) is its most severe manifestation and carries serious, long-term consequences.³ The case fatality rate for untreated TBM is almost 100% and delay in treatment often leads to permanent neurological damage.⁴ Therefore, prompt diagnosis and early treatment are crucial for successful TBM management.^5–8 However, due to the complex manifestations of TBM and the considerably high use of antibiotics in its management, many cases of TBM are atypical. In such cases, laboratory results can be negative: the cerebrospinal fluid (CSF) cytology of infection of the central nervous system (CNS) is often nonspecific^9,10 and pathology is rarely detected. Thus, there are difficulties with diagnosis and high rates of misdiagnosis when using traditional CSF and serological tests in patients with suspected TBM.^10,11 A faster, more sensitive and more specific test for the diagnosis of TBM is urgently needed in clinical practice, and developing improved diagnostic techniques has become a clinical research priority.^12–15

Mycobacterium tuberculosis is the causative agent of tuberculosis in humans. The M. tuberculosis cell wall is constructed primarily of peptidoglycan, which is bound to arabinogalactan and mycolic acid in a covalent fashion. Research findings have indicated that mycolic acid and fatty acid metabolism are closely related to M. tuberculosis survival, virulence and avoidance of attacks by the immune system.^16–20 These factors demonstrate the importance of M. tuberculosis to host-fat metabolism.^20–23

Delta-like 1 ligand (DLL1) is an L-type transmembrane protein consisting of 723 amino acids. It is a member of the Notch ligand family, which is a group of transmembrane receptors that is involved in the development of many organs and cellular differentiation. Of particular interest to this work, the Notch1 signal influences the formation of fat cells by regulating the activation of fatty acids and, thus, plays a crucial role in fat metabolism.^24,25 In M. tuberculosis infection, there is an accelerated exfoliation of the DLL1 ligand, which helps the Notch1 signal promote the differentiation of fat cells.

Thus, because DLL1 participates in fat metabolism,^26,27 it could be used as an indicator of M. tuberculosis infection. To explore DLL1 as a potential new tool for the diagnosis of TBM, the present study adopted an enzyme-linked immunosorbent assay (ELISA) for the quantitative measurement of DLL1 in CSF and serum samples from different patient groups: those with infection of the CNS, those with intracranial metastatic tumour and those with no diagnosis. The present study aimed to estimate the diagnostic accuracy of DLL1 as an indicator of M. tuberculosis infection, and the processes followed adhered to the reporting requirements of the Standards for Reporting of Diagnostic Accuracy initiative.²⁸

This study enrolled 173 patients, aged between 2 and 74 years. There were 62 patients with TBM (23 male and 39 female aged 7–56 years, 22.1 ± 14.8 years), 38 patients with viral meningitis/encephalitis (16 male and 22 female aged 6–60 years, 17.9 ± 12.9 years), 26 patients with bacterial meningitis (15 male and 11 female aged 2–37 years, 14.9 ± 8.4 years), 17 patients with intracranial metastatic tumour (10 male and seven female aged 40–74 years, 57.4 ± 9.7 years) and 30 patients with no diagnosis (15 male and 15 female aged 4–52 years, 20.6 ± 14.8 years).

The DLL1 levels in CSF are presented in Table 1. The TBM group exhibited the highest mean DLL1 concentration and, in this group, the majority of patients had CSF DLL1 concentrations > 1.0 ng/ml. Consequently, 1.0 ng/ml was chosen as the cut-off value (sensitivity = 0.871, specificity = 0.991, positive predictive value[PPV] = 0.982, negative predictive value [NPV] = 0.922; ROC curve analysis). There was a highly statistically significant difference between the TBM group and the other groups, when analysed separately and together (P ≤ 0.001). There were no other significant between-group differences.

The DLL1 levels in serum are shown in Table 2. The TBM group and the tumour group exhibited the highest DLL1 concentrations and the differences between these two groups were not statistically significant. In the TBM group, the majority of patients had serum DLL1 concentrations >6.0 ng/ml. Consequently, 6.0 ng/ml was chosen as the cut-off value (sensitivity = 0.823, specificity = 0.910, PPV = 0.836, NPV = 0.902; ROC curve analysis). There was a highly significant statistical difference between serum DLL1 concentrations in the TBM group and the other groups (excluding the tumour group) analysed together (P ≤ 0.001). There were no other statistically significant differences.

We performed a receiver operating characteristic curve analysis and determined that a cut-off of 1.0 ng/ml for CSF DLL1 could be used to distinguish TBM patients from viral meningitis/encephalitis, bacterial meningitis and nondiagnosed patients (Figure 1), with 69% of all CSF DLL1 levels > 1.0 ng/ml. For serum DLL1, this analysis determined that a cut-off of > 6.0 ng/ml could be used to distinguish TBM patients from viral meningitis/encephalitis and bacterial meningitis, with 71% of all serum DLL1 levels > 6.0 ng/ml (Figure 2).

The CSF DLL1 content and the serum DLL1 content were positively correlated with each other (r = 0.393, P ≤ 0.01). The CSF DLL1 content was not correlated with cell density, levels of CSF protein, glucose or chloride, or CSF pressure (r ≈ 0; data not shown).

Delta-like 1 ligand is the human autoploid of Delta, the Notch ligand in Drosophila, and is an L-type transmembrane protein consisting of 723 amino acids.³⁴ As a Notch ligand, it is formed by one short C-terminal region in cytoplasm and a C-Delta/Serrate/LAG-2 region, following a series of eight repeated sequences of epidermal growth factor.³⁵ The gene has been cloned using polymerase chain reaction amplification with a pair of primers based on a sequence of the Delta gene.²⁷ The Notch family is a group of transmembrane receptors that function in the development of multiple organs and cells, and determine the direction of cellular differentiation.³⁶ Specifically, inhibition, gene mutation or abnormal activation of the Notch signalling pathway may correlate with the development of CNS diseases such as Alzheimer’s disease, Parkinson’s disease, stroke, brain tumour and cerebral ischaemia–reperfusion injury.^24,37 Moreover, the Notch1 signal influences the formation of fat cells by regulating the activation of fatty acids, and plays a crucial role in fat metabolism.^24,25 In the context of tuberculosis, which is a chronic wasting disease that leads to disordered fatty acid metabolism and excessive activation of fat cells, accelerated exfoliation of DLL1 as the Notch ligand helps the Notch signal to promote fat-cell differentiation.²⁵ Consequently, the present study adopted an ELISA method for the quantitative measurement of DLL1 content in CSF and serum from patients with CNS infectious diseases, to explore whether DLL1 represented a potential new tool for the diagnosis of TBM.

Results of the present study indicate that, in the TBM group, DLL1 levels in CSF were significantly higher than in other groups: 54/62 patients (87%) in the TBM group displayed levels > 1.0 ng/ml, while no individual exhibited levels > 1.0 ng/ml in the viral meningitis/encephalitis, purulent meningitis or control groups. DLL1 levels in CSF had no correlation with cell density, levels of CSF protein, glucose or chloride, or CSF pressure; consequently, we believe that the DLL1 content of CSF has potential value for the diagnosis of atypical cases of TBM. In the present study, in the TBM group, the serum DLL1 level was also markedly higher than the level observed in other CNS infections. DLL1 measurement in serum was conducted in 62 TBM patients, among whom 51 cases (82%) exhibited levels >6.0 ng/ml, while one case from the viral meningitis/encephalitis group, one case from the purulent meningitis group and no case from the nondiagnosed group displayed DLL1 levels in serum >6.0 ng/ml. We therefore propose that a cut-off value for DLL1 of 6.0 ng/ml in serum could be used in the diagnostic confirmation of TBM (Figure 2).

The present study findings showed that DLL1 levels in both CSF and serum increased significantly in TBM, and that the DLL1 content in CSF correlated positively with the DLL1 content in serum. DLL1 levels did not correlate with cell density or levels of protein, glucose, chloride or CSF pressure. Again, this indicates that DLL1 could be valuable for the differential diagnosis of atypical cases of TBM. Moreover, this diagnostic tool could also differentiate between TBM and viral meningitis/encephalitis or bacterial meningitis. Although patients with intracranial metastatic tumours had higher serum DLL1 levels than those with viral meningitis/encephalitis or bacterial meningitis (indeed, serum DLL1 levels did not differ significantly between cases of TBM and patients with intracranial metastatic tumours), the mean DLL1 level in the tumour group was lower than that in the TBM group, but this difference was not found to be statistically significant. The authors do not consider there would be clinical confusion between the diagnosis of TBM and intracranial metastatic tumours, given that these tumours are diagnosed most effectively using cytology, CT and MRI.³⁸ However, establishing the DLL1 level would be useful in situations when a differential diagnosis is needed.

In nondiagnosed patients in the present study, DLL1 concentrations were very low. Given that these patients by definition had not been diagnosed with TBM, it is reassuring that their DLL1 levels also did not suggest the presence of TBM.

The present study indicates that DLL1 levels in CSF and serum increase significantly in the context of TBM, and that the DLL1 content of CSF is positively correlated with the DLL1 content of serum. However, the mechanism by which DLL1 is increased is unclear. Although excessive exfoliation of DLL1 may help the Notch signal to promote the differentiation of fat cells, it may involve either or both of the following mechanisms. First, the Notch signal pathway is highly important for T-cell differentiation.^39–42 In the event of a deficiency in the Notch receptor, lymphocytes would develop as B cells rather than T cells,⁴³ and activation of the Notch signal pathway could lead to the development of a T-cell lineage rather than a B-cell lineage: DLL1 exfoliation mediated by ADAM (a disintegrin and metalloprotease) seems to contribute to the downregulation of Notch signalling in adjacent cells, which limits the production of T cells.⁴⁴ Secondly, DLL1 is located in the adherens junctions in neurocyte processes, where it functions as a scaffolding protein. When adherens junctions are damaged by TBM, DLL1 can be released, which increases DLL1 content.⁴⁵

The strengths of the present study are that the analyses of CSF and serum samples were blinded to the diagnosis of the patient and that it enrolled a total of 173 patients who belonged to different diagnostic groups. A limitation is the narrow range of the DLL1 calibration standard (0.125–8 ng/ml). Also, this study did not investigate further the non-TBM patients with CSF DLL1 > 1 ng/ml or serum DLL1 > 6 ng/ml, to explain why they had high DLL1 concentrations. These limitations should be considered in future investigations.

In conclusion, the present study showed that DLL1 levels in CSF and serum appear to be high in patients with TBM, and are very low or undetectable in those with viral meningitis/encephalitis or bacterial meningitis. In patients with intracranial metastatic tumours, the DLL1 content in serum increases, whereas DLL1 levels in the CSF remain relatively low. Elevated concentrations of DLL1 in serum and the CSF may indicate TBM. The measurement of DLL1 in CSF and serum represents a potential tool for TBM diagnosis.