Evaluation of a dual-antigen ELISA test for the serodiagnosis of tuberculosis

Bulletin of the World Health Organization, 66 (2): 203-209 (1988) © World Health Organization 1988 Evaluation of a dual-antigen ELISA test for the serodiagnosis of tuberculosis S. F. RAHEMAN,1 S. WAGNER,2 H. MAUCH,3 N. D. VASUDEVA4, & D. L. INGOLE5 Two antigen preparations from Mycobacterium tuberculosis and M. bovis, respect- ively, were used in an enzyme-linked immunosorbent assay (ELISA) to quantitate the level ofIgG antibodies in samples of169 sera obtained in Indiafrom 75 patients with pulmonary tuberculosis andfrom 94 controls. The results ofthe two ELISA tests were integrated using defined criteria. All the controls were correctly classified by this dual-ELISA approach and only two patients were categorized asfalse negatives. The method has a specificity of 1. 00 and a sensitivity of 0.974, while the gain in the certainty of diagnosis was 1.974, only slightly less than the ideal value of 2. 00. A serological test for tuberculosis, which was a dream of early immunologists, has now become a near possibility. Over the last 10 years there has been extensive worldwide effort to develop such a test. For example, an enzyme-linked immunosorbent assay (ELISA) was first used with encouraging results by Nassau et al. in 1976 for the detection of IgG anti- bodies to Mycobacterium tuberculosis in patients with tuberculosis (1). Subsequently, various workers have attempted to minimize the nonspecificity of the test by using purified mycobacterial antigens. In this way, Daniel et al. used antigen 5 in studies carried out in Argentina, Bolivia, China, and the USA and reported a sensitivity of 70% and a specificity of 97% (2). Also, Grange et al. used FLISA to determine IgG, IgA, and IgM antibodies to M. tuberculosis (3). Gibson et al. reported that better discrimination between tuberculosis patients and healthy controls could be achieved by assaying the four subclasses of IgG antibodies (4). Previously, in a study that compared three different M. tuberculosis antigen fractions, we reported that the occurrence of false negatives in the ELISA test ' Visiting Researcher, Institute for Microbiology and Immunology, Krankenhaus Zehlendorf, Berlin (West) 39. Requests for reprints should be sent to this author at 5 Nelson Square, Chlindwara Road, Nagpur-440013, India. 2 Biochemist, Institute for Microbiology and Immunology, Kranken- haus Zehlendorf, Heckeshorn, Berlin (West). 3 Head, Institute for Microbiology and Immunology, Krankenhaus Zehlendorf, Heckeshorn, Berlin (West). 4 Associate Professor, Department of Preventive and Social Medi- cine, Government Medical College, Nagpur, India. 5 Head, Department of Preventive and Social Medicine, Govermnent Medical College, Nagpur, India. arose because of negative antibody response to specific mycobacterial antigens (5). The mean total IgG level of the false negatives was similar to that of the true positive cases and was significantly higher than the mean total IgG levels of healthy controls. On the basis of this finding, we hypothesized that the false negative patients who exhibited a negative anti- body response to one mycobacterial antigen might be positive to an antigen derived from a different species of Mycobacterium. Use of two different antigens de- rived from two mycobacterial species in the ELISA tests for serodiagnosis of tuberculosis could therefore reduce the number of false results. We describe here the results of a study to test this hypothesis. MATERIALS AND METHODS Samples of sera were collected from 75 cases of pulmonary tuberculosis and from 95 controls. Smear testing of cases revealed that 60 were positive and 15 negative. All cases had recently been diagnosed among patients who had been hospitalized in the District Tuberculosis Centre, Nagpur, India. The controls were divided into two groups: one com- prising 43 healthy individuals without any clinical or radiological evidence of tuberculosis, and the other comprising 52 patients who had been admitted for diseases other than tuberculosis (as determined by clinical and radiological examinations). One of the controls, however, was subsequently confirmed to have tuberculosis of the spine and was, therefore, excluded from the study. Both cases and controls 4873 203- S. F. RAHEMAN ET AL. were residents in or around Nagpur city and were skin-tested with 1 tuberculin unit (TU) of purified protein derivative (PPD) RT-23. Every subject was examined for the presence of a BCG scar and screened with a chest X-ray. Sera were coded and stored at -20 'C before being sent by air to Berlin (West) on solid carbon dioxide for the ELISA tests. Antigens Antigen cytosolic fraction ofM. tuberculosis. This antigen was prepared from M. tuberculosis substrain Ayoama B, obtained from Professor Brehemer, Robert Koch Institute, Berlin (West). The bacilli were disrupted in a Ribi cell fractionator (RF-1) (6) at 206 850 and 241 325 kPa.a The soluble cytoplasmic fraction was separated from the cell walls and re- sidual whole cells by ultracentrifugation at 160 000 g for 3 hours (7), and the supernatant was passed through a 0.45-gm filter and lyophilized. Antigen 60 of M. bovis. This is a cytoplasmic antigen prepared from a BCG strain of M. bovis (8) and was kindly donated for the study.b a Ivan Sorvall Inc., Connecticut, USA. b Anda Biologicals, Strasbourg, France. skAntOn @ysIo MAN Assay technique A quantitative ELISA test was used to determine the level of IgG antibodies to the two mycobacterial antigens. For the cytosolic antigen coating, flat- bottomed Immulon microtitration plates were used.' The antigen was diluted to 100 ug/ml with 0.1 mol/l sodium carbonate buffer (pH 9.4) before being used to coat the plates, which were then incubated overnight at room temperature or for 1 hour at 37 'C. Plates coated with antigen 60 were supplied commercially.d Phosphate-buffered saline (0.15 mol/l) containing 0.05% Tween 20 (pH 7.2) was used to dilute the sera and to wash the plates after each incubation. Sera were simultaneously titrated at a dilution of 1:1000 with antigen cytosolic fraction and antigen 60 and then incubated at 37 'C for 1 hour. After washing the plates three times with phosphate-buffered saline, horseradish-peroxidase-conjugated goat anti-human IgGe was added at a dilution of 1:12 000 and incubated for 1 hour at 37 'C. A mixture of 1% TMB and hydrogen peroxide containing citric acid buffer was used as a substrate for the colour reaction. ' Greiner, Federal Republic of Germany. d See footnote b. ' Tago, Denmark, lot no. 901801. biAnftausn * 0* * 0 0 0 a 0 * * *t **0 s00 ___: * 0 0 0 0 *5 see@ *e0 0*0 * 0 0 - .:B. - _ _ 5055Ct00l000 500550 Controls Fig. 1. Schematic representation of levels of IgG antibodies to (a) cytosolic antigen fraction of Mycobacterium tuberculosis and (b) antigen 60 from M. bovis. Levels are expressed as multiples of adult norm (MAN) values in the 94 controls and 75 cases of pulmonary tuberculosis. Each serum is represented by a dot J True False 204 TUBERCULOSIS SERODIAGNOSIS USING A DUAL-ANTIGEN ELISA TEST 205 LII True False CASES True positives Positives 97.34 % Fa Negatives CONTOlST R 0 True negatives 100 % Fig. 2. Results of the dual-antigen ELISA test based on the multiples of adult norm (MAN) difference criteria. Sera that gave false results are individually represented by stars The colour intensity of each well was recorded at X=450 nm using an automatic microplate reader!f The tests were performed in duplicate and repeated if the end points did not agree within one dilution. The results for the sera on each plate were com- pared with those for two standard sera-one (neg- ative) for sera with low antibody titres and another (positive) for sera with higher antibody titres. The mean value of the titre for the 94 controls was calcu- lated from the absorptions as well as from a pool of all the control sera. All the other values were calculated as multiples of this mean, and the final results were expressed as multiples of the adult norm (MAN) value (9). A value of 3.8 MAN gave the optimum discrimination between cases and controls and this was used as the threshold level in the study. The neg- ative standard serum and the positive standard serum had a value of 1 MAN and 3.8 MAN, respectively. RESULTS The antibody titres to antigen cytosolic fraction and antigen 60 are presented for cases and controls in f Dynatech MK 600. Fig. 1. The performance of cytosolic antigen was better for controls, and gave only two false positives that had, in any case, titres close to the threshold value. However, the performance of antigen 60 was better for cases with four false negatives. Except for the sera from two patients, the same false results were not common to the two antigens used in the study. Sera from eight patients that were negative in the cytosolic ELISA were significantly positive in the antigen 60 ELISA; and two false negatives in the antigen 60 ELISA became significantly positive in the cytosolic ELISA. Similarly, all the false positive sera in the antigen 60 ELISA became negative in the cytosolic ELISA and vice versa (see Tables 3 and 4). The overall results, which demonstrate the enhanced efficacy of the dual-antigen ELISA in discriminating tuberculosis patients from healthy controls, are summarized in Fig. 2. The following criteria were used to classify the sera as positive or negative in the dual-antigen test: -those that had values > 3.8 MAN for both antigens were taken to be positive; -those that had values < 3.8 MAN for both antigens were taken to be negative; S. F. RAHEMAN ET AL. Table 1. Sensitivity, specificity, and gain in diagnostic certainty of the dual-antigen tuberculosis ELISA (DATE) and sputum smear examinationa Gain in Sensitivity Specificity certainty Test (S) (T) (S + T) Dual-antigen ELISA 0.974 1.00 1.974 Sputum smear: Kenya 0.65 0.998 1.648 India 0.574 0.992 1.566 ' Based on reports from Kenya and India (Z 10). - those from asymptomatic individuals that were marginally positive (around 4 MAN) for one antigen and significantly negative (around 1 MAN) for the other were taken to be negative; and -those from symptomatic individuals that were neg- ative for one antigen and significantly positive (around 5 MAN) for the other were taken to be positive. All the control sera were thus correctly classified and only two tuberculosis patients gave false neg- atives. Of the two tests, the cytosolic ELISA was more specific (specificity, 0.979), while that for anti- gen 60 was more sensitive (sensitivity, 0.947). The dual-antigen tuberculosis ELISA (DATE) had a specificity of 1.00 and a sensitivity of 0.974, and Table 2. Correlation between the results of the ELISA tests for the two individual mycobacterial antigens, cytosolic and antigen 60 No. of No. of Outcome cases controls Total Positive to both antigens 63 0 63 (89.35) Negative to both antigens 2 86 88 Positive to one antigen and negative to the other 10 8 18 (10.6 Total 75 94 169 ' Represents the percentage of results for which the diagnosis was unequivocal. b Represents the percentage of results for which the diagnosis was -.' based on the criteria defined in the text. produced a gain in diagnostic certainty of 1.974 (Table 1). Table 2 illustrates the correlation between the results for the two antigens, while Tables 3 and 4, respectively, give characteristics of the study subjects who gave false positive or negative results for either of the two antigens. In an alternative method of classifying the results using the arithmetic means of the two titres (Fig. 3) to categorize the study subjects, all the controls remained true negatives (specificity, 1.00), while Table 3. Study profile of the eight control sera that were false positives in either the cytosolic or antigen 60 (A 60) ELISA test but were true negatives in the dual-antigen tuberculosis ELISA (DATE)' ELISA results (MAN) Age DATE Sample (years) Sex BCG status Chest X-ray Diagnosis Tuberculin test A 60 Cytosolic Mean result TB/PC13 32 M Unimmunized NADb Healthy +ve 4.46c 1.75 3.1 -ve TB/PC15 50 F Unimmunized NAD Healthy -ve 3.89c 1.12 2.5 -ve TB/PC1l0 40 M Unimmunized NAD Healthy -ve 4.64c 2.29 3.5 -ve TB/PC115 14 M Unimmunized NAD Healthy -ve 4.06c 1.35 2.7 -ve TB/PC219 55 M Unimmunized NAD Inguinal hernia -ve 5.06c 0.76 2.91 -ve TB/PC251 30 M Unimmunized NAD Appendicitis - 4.5 c 1.04 2.77 -ve TB/PC127 18 M Immunized NAD Healthy +ve 0.84 3.85c 2.38 -ve TB/PC218 70 F Unimmunized NAD Myocardial -ve 1.00 3.89c 2.44 -ve infarction The threshold level between negative and positive was taken to be 3.8 multiples of adult norm (MAN). b NAD - nothing abnormal detected. c False positives. 206 TUBERCULOSIS SERODIAGNOSIS USING A DUAL-ANTIGEN ELISA TEST Table 4. Study profile of the 12 cases of pulmonary tuberculosis that were negative with either or both of the two mycobacterial antigens and the results of dual-antigen tuberculosis ELISA (DATE) ELISA results (MAN) Age Chest X-ray and DATE Sample (years) Sex BCG status clinical findings Tuberculin test A 60 Cytosolic Mean titre result' TB/PP 8 30 F Unimmunized Bilateral infiltration and -ve (6 mm) 4.96 0.73 b 2.84 + ve cavity on left side TB/PP15 22 M Unimmunized Extensive bilateral -ve (6 mm) 1.43b 1.03b 1.23 -veb infiltration TB/PP33' 21 F Unimmunized Infiltration in right + ve (20 mm) 5.00 2.80b 3.90'c + ve hilar region TB/PP35 50 M Unimmunized Effusion in right lung + ve (20 mm) 5.05 0.76b 2.90 + ve TB/PP39 30 F Unimmunized Extensive infiltration + ve (15 mm) 4.90 1.39b 3.18 + ve in right lung TB/PP48 26 M Immunized Infiltration of left lung - 5.21 0.97b 3.10 + ve TBtPP60' 40 M Unimmunized Bilateral extensive infiltration - 8.27 3.60b 5.93C + ve TB/PP66 60 M Unimmunized Bilateral miliary + ve (16 mm) 5.00 0.98 b 2.99 + ve infiltration TB/PP72 51 M Unimmunized - +ve (16 mm) 1.29b 0.54b 0.91 -veb TB/PP73 18 F Unimmunized - + ve (30 mm) 4.76 1.02 b 2.89 + ve TB/PP 3c 50 M Unimmunized Bilateral cavities +ve (15 mm) 3.3b 5.00 4.15'c +ve TB/PP 6 24 F Unimmunized Hilar infiltration + ve (16 mm) 1.5b 4.90 3.20 + ve ' Classified using the criteria defined in the text. b False negatives. Although negative for one of the antigens, the arithmetic mean titres for these sera were > 3.8 MAN. three of the false negatites (see Table 4) became positive (sensitivity, 0.88). The gain in diagnostic certainty using this analysis of the results is 1.88, a little lower than the first approach discussed above. DISCUSSION The cross-reactivity of M. tuberculosis with other environmental bacteria was long considered to be a major obstacle in developing a specific serological test for tuberculosis. However, with the isolation of specific mycobacterial antigens and the development of improved immunological techniques, the pro- portion of false positives in tuberculosis serology testing has been significantly reduced. Here, we have focused on the proportion of tuberculosis patients who exhibit a negative antibody response to specific mycobacterial antigens, thereby resulting in unavoid- able false negatives in serological tests. We expected that sera which were negative towards one myco- bacterial antigen might exhibit a positive response to an antigen from a different mycobacterial species and that this phenomenon could be used to devise an improved procedure for the serodiagnosis of tuber- culosis (5). Antigen cytosolic fraction from M. tuberculosis and antigen 60 from M. bovis were selected for this purpose. The combined ELISA results for the two antigens had an unprecedented accuracy; 167 out of 169 sera were correctly classi- fied. For 89.35% of the subjects in the study group (Table 2) the diagnosis was unequivocally positive or negative; and for only 10.65% of the subjects did the final result depend on the analysis based on the third and fourth classification criteria described in the results. Ifthe arithmetic means ofthe titres are used as an alternative approach to combine the ELISA results for the two antigens, only nine out of the 169 sera remain as false negatives. The predictive value of a diagnostic test is a widely 207 S. F. RAHEMAN ET AL. True False Fig. 3. Results of the dual-antigen ELISA based on the arithmetic mean criteria. Sera that gave false results are individually represented by stars used indicator of its accuracy. Cornell & Koepsell (10), however, argue that predictive values may not reflect the gain in information provided by a test. Indices of gain in certainty, which depend directly on the sum of the specificity and sensitivity of a given test, express more comprehensively the performance of diagnostic tests. A test that has an index of gain in certainty of 1.00 provides no information, while the maximum gain in certainty is obtained when the index has a value of 2.00. The gain in certainty for the DATE test was estimated as 1.974, which is sig- ficantly higher than that for sputum smear exami- nation calculated on the basis of reports from Kenya and India (1.64 and 1.566 respectively) (11). From these observations, we conclude that the dual-antigen ELISA test described for the serodiag- nosis of tuberculosis offers a potential solution to the problem of serodiagnosis of tuberculosis. ACKNOWLEDGEMENTS We are grateful to Ms I. Sudau and Mr M. P. Mane for their technical assistance and to Dr R. Maes for providing a sample of antigen 60. Our sincere thanks are also due to Mr M. Rafi Sheikh for secretarial assistance. 208 TUBERCULOSIS SERODIAGNOSIS USING A DUAL-ANTIGEN ELISA TEST 209 RESUME EVALUATION D'UN ELISA A DEUX ANTIGtNES POUR LE SERODIAGNOSTIC DE LA TUBERCULOSE Le present article decrit une methode visant a minimiser la proportion des faux resultats dans le s6rodiagnostic de la tuberculose; il s'agit d'un titrage ELISA faisant appel a deux antigenes mycobacteriens. Les auteurs avaient prece- demment rapport6 que les faux negatifs observes dans les 6preuves de serodiagnostic de la tuberculose etaient dus a une suppression de la reponse immunitaire vis-a-vis d'anti- genes mycobact6riens specifiques, ce qui les a conduit a penser que ces faux negatifs donneraient un resultat positif avec un antigene d'une autre esp&ce mycobact6rienne. Un ELISA A deux antigenes, a savoir une fraction cyto- solique de Mycobacterium tuberculosis (souche Ayoama B) et un antigene cytoplasmique de M. bovis (souche BCG), a donc 6t6 mis au point pour quantifier les IgG dans 169 serums provenant de 75 cas de tuberculose pulmonaire et de 94 temoins d'origine indienne. Les r6sultats des deux epreuves ont ete integres selon des criteres bien definis. On n'a trouve que deux faux n6gatifs communs aux deux anti- genes, et pour tous les autres serums, le diagnostic a ete juste. L'indice de sp&cificit6 de cette epreuve a deux anti- genes a ete de 1,00 et l'indice de sensibilite de 0,974; le gain de fiabilite du diagnostic a ete evalue a 1,194 (1'examen des frottis d'expectoration procure un gain de fiabilit6 compris entre 1,566 et 1,64). Si l'on classe les resultats en fonction de la moyenne arithmdtique des titres obtenus avec les deux antigenes, on obtient un gain de fiabilite de 1,88. L'ELISA a deux antigenes ddcrit ici pourrait donc constituer un excellent test serologique pour la tuberculose. REFERENCES 1. NASSAU, E. ET AL. Detection of antibodies to M. tuber- culosis by microplate ELISA. Tubercle, 57: 67-70 (1976). 2. DANIEL, T. M. ET AL. ELISA of IgG antibodies to M. tuberculosis antigen 5 for serodiagnosis of tubercu- losis. Indianjournal ofpediatrics, 52: 349-355 (1985). 3. GRANGE, J. M. ET AL. ELISA: a study of antibodies to M. tuberculosis in the IgG, IgA and IgM subclasses in tuberculosis, sarcoidosis and Crohn's disease. Tubercle, 61: 145-152 (1980). 4. GIBSON, J. A. ET AL. Antibody response to M. tubercu- losis in the four subclasses of immunoglobulin G in patients with smear positive pulmonary tuberculosis. European journal of respiratory diseases, 69: 1-6 (1986). 5. RAHEMAN, S. F. ET AL. A study of humoral immune response in tuberculosis. Indianjournal oftuberculosis (In press). 6. MAUCH, H. & BREHMER, W. Mycobacterial antibodies after tuberculin testing, BCG vaccination, BCG im- munotherapy, and against cross-reacting antigens in a solid-phase radioimmunoassay. Zentralblatt far Bak- teriologie, Mikrobiologie und Hygiene, I. Ableitung originale, A251: 380-388 (1982). 7. RIrKI, K. ET AL. Analysis of the soluble cytoplasmic components ofmycobacterium and nocardia by crossed immunoelectrofocussing. Zentralblatt far Bakterio- logie, Mikrobiologie und Hygiene, I. Abteilung originale, A251: 389-398 (1982). 8. Cocrro, C. & VANLINDEN, F. Preparation and properties of antigen 60 from Mycobacterium bovis (BCG strain). Clinical and experimental immunology, 66: 262-272 (1986). 9. MAUCH, H. ET AL. Analysis of a solid-phase radio- immunoassay for antibodies to cytoplasmic antigen fractions of Candida albicans. Journal ofimmunologi- cal methods, 43: 181-192 (1981). 10. CORNELL, F. A. & KOEPSELL, T. D. Measure of gain in certainty from a diagnostic test. Americanjournal of epidemiology, 121: 744-753 (1980). 11. Tuberculosis prevention trial, Madras. Trial of BCG vaccines in south India for tuberculosis prevention. Indian journal ofmedical research, 72: 1 (1980).