Rapid diagnosis of M.tuberculosis meningitis by enumeration of cerebrospinal fluid antigen-specific T cells

SUMMARY

Setting

Hospital in-patients with suspected tuberculous meningitis predominantly in India.

Objective

To determine whether interferon-γ-secreting Mycobacterium tuberculosis-antigen-specific T cells are present in the cerebrospinal fluid of patients with tuberculous meningitis, and to evaluate the feasibility of cerebrospinal fluid enzyme-linked immunospot for the diagnosis of active tuberculous meningitis.

Design

Prospective, blinded, hospital-based study.

Results

The overnight enzyme-linked immunospot assay detected Mycobacterium tuberculosis-antigen-specific interferon-γ-secreting T cells in cerebrospinal fluid from 9 out of 10 prospectively recruited patients with tuberculous meningitis, and 0 out of 7 control patients with meningitis of other aetiology. This corresponds to a diagnostic sensitivity of 90% (95%CI 56-100%) and specificity of 100% (95%CI 59-100%).

Conclusion

This pilot study demonstrates proof-of-principle for a new T cell-based diagnostic test for tuberculous meningitis which is rapid, sensitive and specific.

INTRODUCTION

Tuberculous meningitis (TBM) has a mortality of 20-41% in the developed world, and 44-69% in developing nations(1-3). The best predictor of outcome is the stage at which treatment is initiated(4), yet diagnosis is a major challenge, as clinical presentation is non-specific and microbiological confirmation is usually lacking. Ziehl-Neelsen staining of cerebrospinal fluid (CSF) smears has very low sensitivity (10-20%)(​3), whilst culture also lacks sensitivity with results not available for several weeks(1). The development of improved diagnostic techniques is therefore a clinical research priority(5).

The enzyme-linked immunospot assay (ELISpot) detects interferon-γ (IFN-γ) secreting T cells specific for two antigens, early secretory antigenic target-6 (ESAT-6) and culture filtrate protein-10 (CFP-10), which are present in Mycobacterium tuberculosis (MTB), but absent from Mycobacterium bovis BCG vaccine and most environmental mycobacteria(6). Applied to blood samples this assay has high sensitivity and specificity for the diagnosis of TB infection(7, 8) but does not distinguish between latent TB infection and active TB(6). Recently it has been shown that ESAT-6-specific IFN-γ-secreting CD4⁺ T cells are concentrated at sites of active tuberculosis (TB)(​9-​11). Using ELISpot these cells can be detected in pleural, ascitic, pericardial and bronchoalveolar lavage fluid(9-11), suggesting that this assay may be a rapid, sensitive and specific marker for active TB at specific anatomical sites. However, detection of functional MTB-antigen-specific T cells in CSF is scientifically challenging given the rapid activation-induced death of CSF T cells in TBM(12).

We aimed to determine whether IFN-γ-secreting MTB-antigen-specific T cells are present in the CSF of patients with TBM, whether the use of ELISpot on CSF samples is feasible, and whether ELISpot has a clinically useful diagnostic accuracy for active TBM. We therefore performed a prospective, blinded, hospital-based study to evaluate the feasibility, diagnostic sensitivity and specificity of CSF ELISpot.

STUDY POPULATION AND METHODS

Patients with clinical features highly suggestive of TBM presenting consecutively to the study physicians (MMT, CG, BM) were recruited prospectively, along with control subjects with meningitis of other aetiologies, from Christian Medical College Hospital, Vellore, India (n=22), from Southampton General Hospital, Southampton, UK (n=1) and from the Krefeld Clinic, Krefeld, Germany (n=1) between March 2004 and November 2006 (Figure 1). Protocols were approved by the institutional review boards and informed consent was obtained from the patients, or if consciousness was impaired, from next of kin. All eligible participants were included. All CSF samples were subjected to estimation of protein content, microscopy of centrifuged CSF and the following microbiological tests: Gram, Ziehl-Neelsen and India ink stains; bacterial, mycobacterial and fungal culture; and PCR for Mycobacterium tuberculosis using IS6110 primers (not available for cases 4, 6 or 11). Heparinized samples of venous blood (10ml) and of CSF (mean 4ml) were obtained and processed for ELISpot prior to commencing anti-tuberculous chemotherapy. CSF lymphocytes were isolated by centrifugation of CSF at 1500rpm for 10 minutes, and cells washed once and resuspended in complete medium (Roswell Park Memorial Institute 1640 media, Sigma, St. Louis, MO, USA) containing 10% foetal calf serum. PBMC were isolated from blood by differential density gradient centrifugation with Ficoll-Paque^PLUS (Amersham Biosciences, Bucks, UK).

Figure 1.

Study flow diagram

The ex-vivo IFN-γ ELISpot assay was performed using 2.5×10⁵ peripheral blood mononuclear cells (PBMC) or CSF cells incubated overnight for 16 hours in 5% carbon dioxide at 37°C using 100 μL aliquots of complete medium. Single wells of pre-coated interferon-γ ELISpot plates (Mabtech AB, Stockholm, Sweden), each contained single pools of overlapping peptides, one spanning ESAT-6 and the other spanning CFP-10 (Research Genetics Huntsville, AL, USA) at a concentration of 20 μg/mL. Postive and negative control wells contained phytohemagglutinin (PHA, MP Biomedicals, Irvine, CA, USA) or no antigen respectively. Plates were developed with preconjugated detector antibody and chromogenic substrate, BCIP/NBTPLUS (Moss Inc, Pasadena, MD, USA), as previously described(6, 9). The commercially available diagnostic test T-SPOT^®.TB is based on the assay used here as developed by Lalvani(6, 7, 9). In 6 cases and 3 controls, where insufficient cellular CSF was available, lower cell numbers were used; in one patient CSF was visibly contaminated by blood.

ELISpot plates were scored by eye in Vellore, using a hand lens, and results were later confirmed in the United Kingdom using an automated ELISpot counter (AID-GmbH, Straßberg, Germany) with predefined intensity and spot size settings. We report absolute numbers of spot forming cells per well (SFCs) after subtraction of the negative control well. Thresholds for a positive response were 5 (for peptides) or 10 (for PHA): positive control) spot forming cells (SFCs) more than, and at least twice the frequency of, the negative control wells, as previously described(6, 7). A response to one or more peptide pools was considered an overall positive ELISpot result.”

Background numbers of SFC in negative control wells were ≤5/well, except for case 11 (20/PBMC well). Those performing and reading the assays (SR, CL) were blind to personal identifiers and clinical and microbiological data.

A clinician (TSCH), blind to the results of ELISpot assays, prospectively assessed case records of each patient according to predefined criteria used in previous studies(4, 13). Specific clinical criteria included fever, headache and neck stiffness for more than 2 weeks. Supporting criteria consisted of 1) CSF findings of lymphocytic pleocytosis, raised protein levels and sterile cultures 2) CT/MRI findings of hydrocephalus, granulomas or basal exudates 3) evidence of extraneural tuberculosis 4) appropriate response to anti-tuberculosis chemotherapy. Patients with specific clinical criteria were classified as “culture-confirmed tuberculosis” where M.tuberculosis could be isolated from CSF, as “highly probable tuberculosis” if 3 supporting criteria were present or “probable tuberculosis” if only 2 were present. Our predefined analytical plan specified inclusion of “probable” and “highly probable” cases in the final analysis along with cases considered to have “active tuberculosis excluded” with a definite alternative diagnosis. It was also specified in the analytical plan that “indeterminate” cases, in whom a final diagnosis of tuberculous meningitis was neither probable, nor reliably excluded, and non evaluable-assays, would not be included in the final sensitivity and specificity calculations.

Statistical analyses were performed using GraphPad Prism 4 (GraphPad Software Inc, CA) and SPSS version 13.0 (SPSS Inc, Chicago, IL).

RESULTS

24 patients were recruited. 4 patients whose diagnoses were considered clinically “indeterminate” were excluded from analysis, as per protocol. Demographic and clinical characteristics of the remaining 20 participants are shown in Table 1. Eleven patients were classified as cases of TBM, comprising 1 “culture-confirmed”, 5 “highly probable” and 5 “probable” cases (Table 2). M.tuberculosis was isolated from case 11 on culture of CSF. The remaining cases were all smear and culture-negative for MTB, all had a combination of clinical, radiological or biochemical features highly suggestive of TBM, and in each case the response to therapy was as expected for the clinical presentation. All cases were also negative for Gram and India ink stains, bacterial and fungal cultures. PCR for MTB was positive only for case 10. CSF analysis revealed a typical lymphocytic pleocytosis in all 11 tuberculosis cases, and a raised total protein in 9 cases (Table 2).

Table 1.

Demographic and clinical characteristics of the study populations

	Cases		Controls
Characteristic	No.	%	No.	%
Total	11		9
Age in years (median, range)	32 (8 to 69)		36 (14 to 64)
Sex
Male	7	64	8	89
Ethnic Origin
Indian	9	82	9	100
African	1	9	0	0
Caucasian	1	9	0	0
HIV Status
Positive	0	0	0	0
Presenting Symptoms
Fever	7	64	6	67
Headache or meningism	7	64	4	44
Impaired consciousness	7	64	5	56
Cranial nerve palsy or diplopia	2	18	1	11
Paresis	1	9	2	22
Seizures	1	9	1	11
Cough	1	9	0	0
Weight loss	1	9	0	0
Dysarthria	1	9	1	11

Table 2.

Clinical, microbiological and immunological characteristics of tuberculous meningitis cases (n=11)

Patient	1	2	3	4	5	6	7	8	9	10	11
Age (years)	36	43	59	8	63	20	28	25	23	69	38
Sex	M	M	M	M	M	M	F	F	F	F	M
Clinical features and duration of symptomsa	Fever Seizure Confusion 3 months	Headache, Fever 6 months	Fever Triplegia Pott's deformity 1 year	Headache Fever Confusion VII nerve palsy	Fever Confusion Meningism 4 days	Cough Headache Vomiting 1 year	Fever Headache Confusion Vomiting	Headache Vomiting Syncope Meningism	Anorexia Fever Headache Vomiting Diplopia 3 months	Headache, anorexia, CSF positive for TB PCR	Confusion. CSF culture positive for MTB
Radiology	CXR: biihilar lymphadeno- pathy MRI: occipital & trigone granulomas	CT & MRI : communicating hydrocephalus multiple infarctions, endarteritis	CXR: normal MRI spine: L3/4 disc collapse, arachnoiditis	MRI & CT: Basal meningeal enhancement, ventriculomegaly, cerebral vasculitis	CXR: pleural effusion CT brain: normal	CXR: pleural effusion CT brain: normal	CXR: normal	CXR: normal MRI: transverse sinus thrombosis	CXR: normal CT brain: normal	CXR: normal MRI: meningeal enhancement	CXR normal CT: multiple ring enhancing lesions; basal meningitis
Response to therapy	Appropriate	Appropriate	Appropriate	Appropriate	Appropriate	Appropriate	Appropriate	Appropriate	Appropriate	Appropriate	Appropriate
BCG vaccination	No	Yes	No	**	No	Yes	**	Yes	Yes	No	**
Mantoux result	Negative	Negative	Negative	Negative	**	Negative	**	Negative	Negative	Negative	Positive
HIV serology	Negative	Negative	Negative	Negative	Negative	Negative	**	Negative	Negative	**	Negative
CSF Analysis
White cell count (mm−3) (<4)	720	300	25	200	130	80	400	240	210	10	597
Lymphocytes (%)	98	96	90	96	98	94	94	97	97	++	++
CSF Protein (mg/dL) (15-45)	360	128	285	194	141	101	320	42	87	30	514
Diagnostic classification	Highly probable TBM	Highly probable TBM	Highly probable TBM	Highly probable TBM	Probable TBM	Probable TBM	Probable TBM	Probable TBM	Probable TBM	Highly probable TBM	Culture confirmed TBM
PBMC ELISpot results b
Negative control	2	0	0	0	0	0	1	0	0	0	20
PHA	22	>100	>100	10	>100	>100	>100	>100	>100	>300	>200
ESAT6 pool	41	2	29	3	17	24	21	0	0	51	221
CFP10 pool	>100	13	34	32.5	7	18	>100	0	0	43	>200
Overall result	Positive	Positive	Positive	Positive	Positive	Positive	Positive	Negative	Negative	Positive	Positive
CSF ELISpot Results
Input number of cells/well	250000	250000	31000	**	250000	125000	250000	250000	116000	125000	150000
Negative control	0	0	0	0	0	0	1	0	5	0	0
PHA	30	9	18	107	16	28	>100	>100	14	0	168
ESAT6 pool	5	7	9	8	20	24	90	6	0	0	134
CFP10 pool	22	6	0	24	15	19	>100	3	0	0	178
Overall result	Positive	Positive	Positive	Positive	Positive	Positive	Positive	Positive	Negative	Failed	Positive

Abbreviations: BCG: bacilli Calmette-Guerin, CFP-10 pool: pooled peptides from culture filtrate protien-10, CSF: cerebrospinal fluid, CT: computed tomography scan, CXR: plain chest X-radiograph, ESAT-6 pool: pooled peptides from early secretory antigen-6, MRI: magnetic resonance image, PBMC: peripheral blood mononuclear cell, PHA: phytohemagluttinin, TB: tuberculosis, TBM: tuberculous meningitis, **: data not available

^aCase 11 was positive on culture, all other cases were negative for acid-fast bacilli on smear and culture. Except for patients 4,6 and 11, PCR was performed for TB, and was positive only in case 10.

^b250,000 cells/well were used in all PBMC assays, but variable numbers were used in CSF assays. The table shows absolute numbers of spot forming colonies (SFC) per well after subtraction of the negative control well (PHA, ESAT-6 and CPF-10). Data are not corrected for total number of cells per well. A minimum of 10 (PHA) or 5 (peptide pools) SFC/250,000 cells was considered positive. A response to one or more peptide pools was considered an overall positive ELISpot result.

ELISpot on blood was positive in 9 of 11 TBM cases, giving a sensitivity of 82% (95% confidence interval 48-98%). ELISpot on CSF was positive in 9 of 10 TBM cases, including all 5 “highly probable” or “culture-confirmed” cases. Based on these small numbers the sensitivity of ELISpot on CSF for TBM is estimated at 90% (95%CI 56-100%). In one TBM case the CSF ELISpot assay failed, with response neither to peptides nor to the positive control, rendering the assay uninterpretable. In this assay CSF cell numbers were 50% less than the standard 2.5×10⁵/well.

Controls comprised 9 patients with meningitis of other aetiologies, comprising: non-tuberculous aseptic meningitis (n=4), partially treated pyogenic meningitis (n=2), carcinomatous meningitis (n=1), lymphoma (n=1) and post traumatic meningitis (n=1). Except one patient who died of bacteraemia secondary to pyelonephritis, all controls either were histologically confirmed or recovered without anti-tuberculous chemotherapy.

ELISpot on blood was positive in 2 of 8 controls with evaluable assays, giving a specificity of blood ELISpot for active TBM of 75% (95%CI 35-97%). ELISpot on CSF was positive in 0 of 7 controls with evaluable assays, giving a specificity of CSF ELISpot of 100% (95%CI 59-100%). The CSF ELISpot assay failed in two controls, with response neither to peptides nor to positive control. In both cases the CSF was not predominantly lymphocytic (23-40% lymphocytes), and contained few white cells. Moreover, in one instance CSF numbers in the assay were 50% less than the standard 2.5v10⁵/well and in the second, CSF was heavily contaminated with blood due to head trauma.

The frequencies of cells responding to ESAT-6 and CFP-10 in CSF ELISpot from TBM patients were of moderate magnitude in most cases despite low numbers of CSF mononuclear cells used in several assays. Frequencies of responding cells in CSF were not significantly different from those observed in the blood (Table 2, Wilcoxon's P=0.09 (ESAT-6) and P=0.74 (CFP-10)).

There was close agreement for both CSF and blood ELISpot results counted by eye with those counted by automated reader: κ=0.94 (95%CI 0.81 to 1.00) for positivity of peptide wells.

DISCUSSION

In this pilot study we found that ELISpot is feasible for detecting MTB-antigen-specific T cells in CSF with a high diagnostic sensitivity of 90% (95%CI 56-100%) and specificity of 100% (95%CI 59-100%). The high sensitivity, specificity and speed of CSF ELISpot suggest it could be a useful tool for assessment of patients with suspected TBM, since early diagnosis and treatment are critical for a good outcome(1, 5).

By comparison, using Ziehl-Neelsen staining, only 52% at best of initial smears are positive, with many authors finding acid fast bacilli in fewer than 20% of cases(1, 3). Culture also lacks sensitivity: 25-70% in some series(1, 14, 15), or as little as 10% in others, especially from the developing world(1). The tuberculin skin test is also frequently negative or anergic in tuberculous meningitis and other severe forms of active tuberculosis and in this study it was negative in 8 of 9 patients with tuberculous meningitis in whom results were available. Other new techniques may perform better. For example preliminary data suggest sensitivity of 80-85% for techniques for detecting antibody or antibody-secreting cells using ELISA or ELISpot (16, 17). Despite high initial estimates of sensitivity of PCR, a large meta-analysis found a pooled estimate for the sensitivity of commercial tests to be 56%(46-66%)(​3). Moreover in a realistic clinical setting, a large study found a sensitivity of only 38% for the nucleic acid-based Mycobacterium tuberculosis Direct Test(5).

A significant weakness of our study is the lack of microbiological confirmation of TB in all but one of our patients, whilst PCR for MTB was positive in only one other case. However, microbiological confirmation in TBM is commonly not achievable, especially in the setting of a routine service laboratory in a developing nation. Negative cultures are the norm in many studies of TBM patients (14), and especially in evaluations of new diagnostic tests, where perhaps less CSF may be available for standard techniques, microbiological confirmation is often obtained in only a small proportion (0-12%) of cases(4, 13, 16).

A bacteriological gold standard is too insensitive to be used alone for evaluation of new diagnostic tests, which may be more sensitive than culture and Ziehl-Neelsen(15). Therefore, like other investigators(4, 5, 14-16) we used composite reference standards for diagnosis that use culture, Ziehl-Neelsen, clinical, radiological and therapeutic outcome criteria. Use of such reference standards did not affect test statistics computed by meta-analysis(3), but may better reflect the clinically relevant population, rather than a selected subpopulation with high bacterial burden. Notwithstanding the lack of microbiological confirmation, our strict case definitions based on validated composite reference standards make it highly likely that our patients did indeed have TBM.

CSF ELISpot gave evaluable results in 4 tuberculosis cases with suboptimal cell numbers (<250,000/well) (cases 3,6,9,11), of whom 3 were positive. However, whilst use of lower cell numbers is possible, it appears to adversely affect assay sensitivity and performance since lymphocyte numbers were suboptimal in each of the failed assays. Limiting CSF lymphocyte numbers would have a limited impact in true TBM cases, as the majority have a lymphocytic pleocytosis in CSF; however, indeterminate assay results may be more common in patients with suspected TBM and low CSF lymphocyte counts. Our population did not include any individuals with positive HIV serology. HIV infection might also lower CD4 lymphocyte numbers in CSF in tuberculous meningitis, potentially increasing rates of indeterminate or false negative results, although interestingly the diagnostic sensitivity of peripheral blood ELISpot is maintained amongst HIV-positive individuals with active TB (18, 19).

Amongst cases, there was general concordance of assay results between blood and CSF, with the exception of case 8 in whom CSF alone was positive. Whilst compartmentalization of antigen-specific effector memory CD8 T cells has previously been described in a single paediatric TBM case (20), in our study, frequencies of responding cells were equivalent in blood and CSF. This is in contrast to the increased concentration of MTB-antigen-specific T cells at the site of disease in pleural, ascitic, pericardial and pulmonary TB(9-11). The reasons for this are unclear but include the possibility that the inflamed blood-brain barrier allows for less efficient lymphocyte migration than inflamed serosal surfaces. Alternatively clinical presentation in TBM may occur earlier in the disease process, prior to significant lymphocyte compartmentalization.

Amongst controls, two participants were positive by blood ELISpot alone, of whom both were native to Vellore, including one with nodular opacities on plain chest radiograph consistent with old TB. These results would be consistent with incidental detection of latent tuberculosis infection, consonant with the known high sensitivity of blood ELISpot for LTBI(7) and suggest that CSF ELISpot might potentially discriminate between latent and active disease.

CONCLUSION

This study has demonstrated proof-of-principle for ELISpot as a potential new diagnostic test for TBM which is rapid, practicable in a resource-poor setting, and appears to have high sensitivity. However, small case-control studies tend to overestimate sensitivity(3). Large prospective studies are now warranted to validate the clinical utility of CSF ELISpot in routine practice. Such studies should have better microbiological confirmation, and include children and more controls. Controls should be patients in whom the test is clinically indicated, and include a wide range of non-tuberculous central nervous system diseases, including individuals with latent tuberculosis infection, to obtain robust estimates of the diagnostic sensitivity and specificity.