Interpretation of mycobacterial antibodies in the cerebrospinal fluid of adults with tuberculous meningitis

Authors


Corresponding Author Blanca I. Restrepo, University of Texas Health Science Center Houston-School of Public Health at Brownsville, UTB SPH Bldg, 80 Fort Brown, Brownsville, TX 78520, USA. Tel.: +19568825172; Fax +19568825152; E-mail: blanca.i.restrepo@utb.edu

Summary

Objective  Microbiological identification of Mycobacterium tuberculosis is insensitive and slow, and clinical distinction of tuberculous meningitis (TBM) from other subacute or chronic meningoenchephalitides (SACM) is difficult. Successful use of highly specific M. tuberculosis serological assays on cerebrospinal fluid has been reported, but their performance for diagnosis in a tuberculosis endemic country where they would be of most value is unclear. We sought to determine the biological basis for the uncertainty in interpretation of antibody detection in the CSF of TBM patients.

Methods  We identified prospectively 46 adults with SACM and explored the concordance between TBM diagnosis and detection of highly specific M. tuberculosis antibodies in CSF. The source of antibodies in CSF was explored by evaluating the correlation between antibody titres in CSF with those in serum, or with the albumin quotient. Intrathecal IgG synthesis was assessed by the IgG index.

Results  Positive antibody titres were more frequent among TBM patients (76%), but were also present in individuals with other SACM (59%). A positive correlation between antibody titres in CSF with those in serum, or with the albumin quotient, supported the leakage of antibodies from plasma to CSF through an increased blood–brain barrier permeability. Intrathecal IgG synthesis was only detected in 35% of the TBM cases.

Conclusion  Plasma antibodies likely synthesized in response to previous tuberculosis infections were a major source of mycobacterial antibodies in CSF due to leakage through an impaired blood–brain barrier. Interpretation of mycobacterial antibodies in CSF of adults for TBM, however specific, must take into account the contribution of antibodies from plasma, and hence, has questionable use for diagnosis.

Abstract

Objectifs  L’identification microbiologique de Mycobacterium tuberculosis n’est pas sensible, est lente et la distinction clinique de la méningite tuberculeuse (MTB) des autres méningo-encéphalites subaiguës ou chroniques (SACM) est difficile. L’utilisation avec succès du test sérologique très spécifiques pour M. tuberculosis sur le LCR a été rapportée, mais ses performances ne sont pas claires pour le diagnostic de la tuberculose dans un pays endémique où il serait de plus grande utilité. Nous avons cherchéàétablir les bases biologiques de l’incertitude dans l’interprétation de la détection des anticorps dans le LCR de patients à MTB.

Méthodes  Nous avons identifié prospectivement 46 adultes ayant SACM et avons exploré la concordance entre le diagnostic de MTB et la détection très spécifique d’anticorps M. tuberculosis dans le LCR. La source d’anticorps dans le LCR a été explorée en évaluant la corrélation entre les titres d’anticorps dans le LCR avec ceux dans le sérum ou avec le quotient d’albumine. La synthèse d’IgG intrathécales a étéévaluée par l’indice des IgG.

Résultats  Des titres positifs d’anticorps étaient plus fréquents chez les patients MTB (76%), mais sont également présents chez les individus avec d’autres SACM (59%). Une corrélation positive entre les titres d’anticorps dans le LCR et ceux dans le sérum ou avec le quotient d’albumine, soutenait le passage des anticorps du plasma vers le LCR grâce à un accroissement de la perméabilité de la barrière hémato méningée. La synthèse des IgG intrathécales a été détectée uniquement chez 35% des cas de MTB.

Conclusions  Les anticorps plasmatiques probablement synthétisés en réponse à des infections tuberculeuses préalables étaient une source majeure d’anticorps antimycobactériens dans le LCR due à un passage à travers une barrière hémato méningée déficiente. L’interprétation des anticorps antimycobactériens dans le LCR d’adultes pour la MTB quoique spécifique doit tenir compte de la contribution des anticorps plasmatiques et est par conséquent d’utilisation douteuse pour poser un diagnostic.

Abstract

Objetivos  La identificación microbiológica de Mycobacterium tuberculosis es poco sensible y lenta, y la distinción clínica de la meningitis tuberculosa (MTB) de otras formas subagudas o crónicas de meningoencefalitis (MSAC) es difícil. El uso exitoso de ensayos serológicos altamente específicos en líquido cefalorraquídeo ha sido previamente reportado, pero su desempeño para el diagnóstico de la tuberculosis en un país endémico, en donde serían de mayor utilidad, no está claro. Hemos buscado determinar las bases biológicas de la incertidumbre en la interpretación de la detección de anticuerpos en el líquido cefalorraquídeo (LCR) de pacientes con MTB.

Métodos  Hemos identificado de manera prospectiva 46 adultos con MSAC, y explorado la concordancia entre el diagnóstico MTB y la detección de anticuerpos altamente específicos para M. tuberculosis en el LCR. La fuente de anticuerpos en el LCR fue explorada evaluando la correlación entre títulos de anticuerpos en LCR con aquellos en suero, o con el cociente de albúmina. La síntesis de IgG intratecal se evaluó utilizando el índice IgG.

Resultados  Los títulos positivos de anticuerpos eran más frecuentes entre pacientes con MTB (76%), pero también estaban presentes en otros MSAC (59%). Una correlación positiva en los títulos de anticuerpos en LCR con aquellos en suero, o con el cociente de albúmina, apoyaba la filtración de anticuerpos desde el plasma al LCR por un aumento en la permeabilidad de la barrera sangre-cerebro. Solo se detectó síntesis intratecal de IgG en un 35% de los casos de MTB.

Conclusiones  Los anticuerpos en plasma probablemente sintetizados como respuesta a infecciones previas de tuberculosis eran una fuente importante de anticuerpos anti-micobacteria en el LCR debido a la filtración a través de una barrera sangre-cerebro deteriorada. La interpretación de los anticuerpos anti-micobacteriales en LCR de adultos para MTB, aunque específicos, debe tener en cuenta la contribución de los anticuerpos en plasma, y por lo tanto, tiene una utilidad cuestionable para el diagnóstico.

Introduction

A major difficulty in management of tuberculous meningitis (TBM) is to distinguish it from other subacute or chronic central nervous system (CNS) meningoencephalitides (Thwaites & Tran 2005). Prompt diagnosis and initiation of chemotherapy is essential to prevent neurological sequelae or death (Zuger & Lowy 1997; Norris & Buckley 2003). Identification of Mycobacterium tuberculosis in cerebrospinal fluid (CSF) is difficult: detection of bacilli in direct smears lacks sensitivity (Zuger & Lowy 1997; Thwaites et al. 2004) and M. tuberculosis isolation may take weeks (Zuger & Lowy 1997; Thwaites et al. 2000, 2004). Detecting mycobacterial antibodies in CSF has been extensively documented (Chandramuki et al. 2002), and proposed as a rapid, economic and simple diagnostic alternative. However, it is still unclear whether either presence or titres of antibody indicates active TBM. While there is evidence for intrathecal antibody production in TBM (Kinnman et al. 1981; Park et al. 1993; Cho et al. 1995), antibodies may also transfer from plasma to CSF through the leaky blood–brain barrier provoked by TB meningitis (Prabhakar et al. 1990). Thus M. tuberculosis antibodies may already be present in the plasma of individuals from regions endemic for tuberculosis, where a significant proportion of the adult population will have latent tuberculosis infection or active non-CNS tuberculosis. We explored the source of mycobacterial antibodies in CSF in order to understand their diagnostic value in patients with TBM or other subacute or chronic meningoencephalitides (SACM) from Columbia, a tuberculosis-endemic country (incidence rate 45/100 000 in 2005; WHO 2007).

Materials and methods

Between 2000 and 2003, adult patients admitted to the neurology or infectious diseases wards of the Hospital Universitario San Vicente de Paul, a non-profit private hospital in Medellín that offers highly complex patient care to those who cannot pay, with signs and symptoms suggestive of a SACM (Coyle 1999) were prospectively identified. Before enrolment the study was explained. Participants either provided written consent themselves, or, when potential participants were unconscious, it was given by a legal representative.

Patients were diagnosed with confirmed TBM when M. tuberculosis was isolated from CSF, and with probable TBM when detection of mycobacteria was either inconclusive (direct smear positive, culture negative), or both assays negative but with: (i) history of direct contact with a pulmonary tuberculosis case, (ii) head computer tomography (CT) and/or magnetic resonance imaging (MRI) compatible with tuberculous meningitis or tuberculoma and/or (iii) an adequate response to the anti-mycobacterial therapy (e.g. sustained improvement in mental status or other neurological signs in the first 2 weeks of treatment). The remaining patients with SACM but no evidence of TBM were classified as ‘SACM-controls’. ‘Reference’ CSF (CSF reference control) and in some cases a serum specimen from the same individual (CSF-serum reference control) were obtained from individuals who reported to the physician with neurological symptoms but had normal CSF by cytochemical and biochemical analysis. The Ethics Committees from all participating institutions reviewed and approved this study.

Cerebrospinal fluid was checked to exclude blood contaminated specimens. Specimens were dispensed into aliquots for routine cytochemical analysis, identification of M. tuberculosis (for cases admitted with SACM), or stored frozen at –20 °C for antibody titration, albumin quotient and IgG index measurements. Mycobacteria identification was carried out by direct smear using Kinyoun or Auramine–Rhodamine staining, and cultured on Lowenstein–Jensen medium and in BACTEC MGIT 960 System (BD Diagnostics Systems, Sparks, CA, USA). Enzyme-linked immunosorbent assay for mycobacterial antibody titration in sera and CSF was performed as described with a few modifications (Silva et al. 2003). Namely, (i) antigens included M. tuberculosis recombinant proteins ESAT-6, 14 kDa or alpha-crystallin 19 and 38 kDa purified to near-homogeneity from Escherichia coli (Colangeli et al. 1998), or a semi-crude preparation of culture filtrate from M. tuberculosis (Olsen et al. 2000) as a reference that contains multiple native mycobacterial antigens and (ii) incubation with CSF was performed in blocking buffer [1% non-fat skim milk in PBS plus 0.05% tween-20 (PBS-T)], and serum in blocking buffer plus 250 mm NaCl. Positive antibody titres were arbitrarily estimated to be above the mean OD450 of healthy reference controls plus two standard deviations. ELISA assays to quantify albumin and IgG in paired serum and CSF were performed using the conditions described above for CSF (Silva et al. 2003), with the following modifications: For albumin, (i) plates were coated with known concentrations of human albumin for the standard curve (SIGMA, St Louis, MO,USA) or with dilutions of CSF and serum, (ii) anti-human serum albumin monoclonal was added as secondary antibody (SIGMA Chemical Co.), and (iii) goat anti-mouse conjugated with horseradish peroxidase as conjugate (Southern Biotechnology Associates, Inc., Birmingham, AL, USA). Albumin quotient was calculated as (albuminCSF/albuminserum) × 1000 (1997). For IgG detection, (i) plates were coated with mouse anti-human IgG (Southern Biotechnology Associates) (ii) known concentrations of human IgG were used for the standard curve (SIGMA) and (iii) rabbit anti-human IgG-horseradish peroxidase (Southern Biotechnology) was used as conjugate. The IgG index was calculated as (IgGCSF/IgGserum)/(albuminCSF/albuminserum) (1997). ELISA assays for albumin and IgG were validated by testing 10 randomly selected patient specimens with SACM and 10 CSF-serum reference controls by kinetic nephelometry (Array 360 Beckman Coulter, Brea, CA, USA). Non-parametric methods were used to compare groups, with P < 0.05 considered significant.

Results

Twenty-nine patients were diagnosed with TBM; 24 confirmed and five probable (four responded to treatment and one had positive smear; Table 1). Seven had concomitant pulmonary tuberculosis based on positive sputum smear, and eight had HIV co-infection (none with mixed infections). In the same clinical setting 17 patients were diagnosed with SACM (SACM controls; Table 1). Twelve were co-infected with HIV, and two of them had mixed infections (neurosyphilis and cryptococcosis; toxoplasmosis and cytomegalovirus). Diagnosis of SACM controls included toxoplasmosis (eight cases), neurosyphilis (two cases), cryptococcosis (two cases), vasculitis (one case), cytomegalovirus (one case), neurocysticercosis (one case), systemic lupus erythematosus (one case) or unknown aetiology (three cases).

Table 1.   Characteristics of patients with subacute or chronic meningoenchephalitides at enrollment*,†
 TBM, n = 29SACM controls, n = 17P-value‡
  1. *Data expressed as n (%) unless otherwise specified; †per cent based on specimens with available information; ‡P-value estimated by chi-square or Fishers exact.

Patient characteristic
 Male18 (62)10 (59)0.83
 Median age (IQR)40 (17)33 (13)0.15
 HIV positive8 (27.6)12 (71)0.005
CNS signs and symptoms
 Symptoms duration in days (IQR)15 (7.5)15 (22)0.9
 Headache25 (96)14 (88)0.54
 Seizures6 (24)4 (24)1.0
 State of consciousness
  Stupor9 (31)0 (0)0.002
  Somnolient17 (59)8 (47) 
  Conscious3 (10)9 (53) 
 Fever25 (89)10 (59)0.03
 Papilledema15 (52)3 (18)0.03
 Motor system14 (48)10 (59)0.5
 Deep tendon reflex15 (52)7 (41)0.49
 Cranial nerves12 (43)4 (24)0.21
 Meningeal signs3 (10)9 (53)0.004
 Vomiting/nausea15 (60)4 (27)0.06
 Abnormal chest X-ray14 (54)9 (82)0.15
Abnormal CSF findings
 Glucose (≥40 mg/dl)29 (100)8 (47.1)<0.001
 Leukocytes (≥5/ml)29 (100)14 (82.4)0.02
 Protein (≥45 mg/dl)29 (100)11 (64.7)0.001

Tuberculous meningitis cases had positive anti-mycobacterial titres in CSF against at least one recombinant antigen (18 cases: 62%), culture filtrate (15 cases: 52%), or both (22: 76%; Table 2), with median anti-mycobacterial titres to the 19kDa, 38kDa and culture filtrate antigens higher for TBM patients than for controls (Figure 1). These findings reaffirmed that a proportion of patients with TBM do mount an antibody-mediated immune response to mycobacterial antigens (Kalish et al. 1983; Hernandez et al. 1984; Chandramuki et al. 2002). Despite the use of recombinant antigens specific for M. tuberculosis, some control patients also had positive antibody responses, most frequently among the SACM controls (n = 10; 59%), rather than the (uninfected) ‘reference’ CSF controls (n = 5; 26%; Table 2). The possibility of cross-reactivity with other bacterial or fungal moieties in the CSF cannot be ruled out but was unlikely due to the specific nature of the recombinant antigens used in our study (Colangeli et al. 1998; Lyashchenko et al. 1998).

Table 2.   Proportion of patients with positive anti-mycobacterial titres in cerebrospinal fluid
Study groupsAntigens, n(%)
ESAT-614 kD19 kD38 kDCFAny RC AgAny Ag
  1. Cut-off for positive titre: higher than median CSF reference control plus 2 SDs: ESAT-6, 0.18; 14 kDa, 0.24; 19 kDa,

  2. 0.21; 38 kDa, 0.17; CF, 0.25

  3. Any RC Ag, positive titre with any of the four recombinant antigens; Any Ag; positive titre with any of the five antigens.

TBM (n = 29)5 (17)11 (38)11 (38)8 (28)15 (52)18 (62)22 (76)
SACM control (n = 17)2 (12)6 (35)5 (29)3 (18)6 (35)7 (41)10 (59)
CSF reference control (n = 19)2 (11)1 (5)2 (11)2 (11)1 (5)5 (26)5 (26)
Figure 1.

 Scatter plots of antibody titres in cerebrospinal fluid of tuberculous meningitis cases and controls. The absorbance (OD 450) for antibody titres against ESAT-6, the 14 kDa, 19 kDa, 38 kDa and culture filtrate antigens is shown. Horizontal bar indicates median for each study group. Circles, TBM; squares, subacute or chronic meningoenchephalitides controls; triangles, healthy controls.

We explored whether the possible source of mycobacterial antibodies in CSF among TBM and non-TBM, SACM patients was due to leakage of antibodies from plasma to CSF. This hypothesis was supported by finding a positive correlation between antibody titres in paired serum and CSF specimens against the 14 kDa 19 kDa, ESAT-6 and culture filtrate in TBM cases (rho values between 0.56 and 0.83; P < 0.05) or SACM control patients (rho values between 0.57 and 0.72; P < 0.05), or for all antigens tested for both patient groups combined (Table 3 and Figure 2). To further explore the relationship between serum and CSF antibodies, we determined whether patients with higher titres in CSF had higher blood–brain barrier permeability, assessed by the albumin quotient value (Table 3). The positive correlation between antibody titres in CSF and albumin quotient values for all patients with SACM supported the possible leak of antibodies from serum to CSF (Table 3). TBM patients had higher albumin quotient values than SACM controls or the paired CSF-serum reference controls (Table 4).

Table 3.   Correlation between mycobacterial antibodies in CSF and serum or AQ values for the same patient in all SACM cases tuberculous meningitis (TBM) and non-TBM
Antibodies toCSF Abs vs. serum AbsCSF Abs vs. AQ
rhoP-value95% CIrhoP-value95% CI
  1. r values obtained using Spearmans correlation; CF, culture filtrate. Abs, antibodies; AQ, albumin quotient; CSF, cerebrospinal fluid

ESAT-60.51<0.0010.71, 1.050.420.010.04, 0.69
14 kDa0.76<0.0010.42, 0.900.5<0.0010.08, 0.71
19 kDa0.69<0.0010.59, 1.00.53<0.0010.13, 0.74
38 kDa0.360.020.12, 0.700.370.020.21, 0.79
CF0.7<0.0010.49, 0.950.62<0.0010.38, 0.91
Figure 2.

 Scatter plot illustrating concordance between anti-14 kDa levels in serum (x-axis) and CSF (y-axis) from TBM patients. Paired CSF and serum specimens from TBM patients were evaluated for antibody titres by enzyme-linked immunosorbent assay as described in methods, and absorbance 450 was plotted. Each dot represents one patient; Rho = 0.83; P < 0.001.

Table 4.   Albumin quotient and IgG index in TBM cases, SACM and CSF-serum reference controls
 Albumin quotientIgG index
Median (IQR)Patients above 7.2*n(%)Median (IQR)Patients above 0.96*n(%)
  1. *Highest value for CSF-serum reference control was estimated as median + IQR.

  2. n is lower due to missing serum for one patient with TBM or in SACM controls.

  3. P < 0.05 compared with SACM control.

  4. §P < 0.05 compared with CSF-serum reference control.

TBM (n = 26)†28.5 (15.8)‡,§23 (88)0.53 (0.33)9 (34.6)
SACM control (n = 16)†12 (4)§11 (68.8)0.82 (0.56)§11 (68.8)
CSF-serum reference control (n = 10)4 (3.2)Reference0.5 (0.46)Reference

Discussion

Previous studies have suggested that intrathecal antibody synthesis occurs in TBM patients (Kinnman et al. 1981; Park et al. 1993; Cho et al. 1995) and serodiagnostic assays for TBM are based on this principle: detection of mycobacterial antibodies in CSF and lower or undetectable titres in serum. Compared with CSF-serum reference controls, an increased IgG index indicated intrathecal antibody synthesis in SACM controls (P = 0.002), but not in TBM patients (P = 0.77; Tables 3 and 4). The proportion of patients with values above the median plus IQR of the control group (IgG index of 0.96), and therefore, with evidence of intrathecal IgG synthesis was 34.6% for TBM and 68.8% for SACM controls (Table 3). The proportion of patients with intrathecal IgG synthesis may be indeed higher than these observed values as 34 patients with CNS pathology (TBM plus SACM controls) had increased blood–brain barrier permeability (Table 3), and this alteration can hinder IgG index detection (Anonymous 1991).

We show that in a tuberculosis endemic population a major source of antibodies to specific M. tuberculosis antigens in CSF of adults is likely to originate from serum in both TBM cases and SACM controls. This limits the use of serology for prompt TBM diagnosis in the clinical setting. We assume that pre-existing antibodies in serum reach the CSF as a result of a leaky blood–brain barrier regardless of the current central nervous system pathology. Mycobacterial antibodies may be present in serum in response to a previous symptomatic or latent extra-cerebral tuberculosis infection. This hypothesis is supported by the trend for higher mycobacterial titres in individuals with abnormal chest X-rays (mean rank for anti-culture filtrate titres 10.5; P = 0.07). TBM from reactivation of a latent tuberculosis infection would be more likely in adults (our study population). This is in contrast to children, who would be more likely to present TBM as a complication of primary tuberculosis (Zuger & Lowy 1997).

We cannot exclude contamination of CSF with blood during the lumbar puncture, although we were careful to minimize this by collecting CSF in three consecutive tubes, and using the last one for serological studies. We also rejected for analysis specimens with obvious blood contamination.

Even though our sample size is relatively small (given the low frequency of TBM compared with pulmonary TB), our findings warrant the importance of confirming these findings in a larger cohort of patients. Regardless of the high specificity of the antigens used in this study, the prospective use for serological diagnosis should take into account pre-existing antibodies in serum and degree of blood–brain impairment. Thus, serological discrimination of adult patients with TBM from patients with other SACM should be interpreted with caution in a country where tuberculosis is endemic.

Acknowledgements

We thank Dr Maria Laura Gennaro for providing the recombinant antigens and technical support, Dr Robert Wallis for providing the culture filtrate, Dr Carlos S. Uribe for his clinical advice and Drs Susan P. Fisher-Hoch and Joseph McCormick for comments on the manuscript. This study was supported by a GORGAS research collaboration award to Dr BI Restrepo, and grants from the Comité para el Desarrollo de Investigación de la Universiad de Antioquia and COLCIENCIAS 2213-04-11900 to Dr J Robledo.

Ancillary