Comparison of an Established Antibody Sandwich Method with an Inhibition Method of Histoplasma capsulatum Antigen Detection

Abstract

The Histoplasma antigen immunoassay utilizes an antibody sandwich method that provides a rapid and reliable means of diagnosing the more severe forms of histoplasmosis. Inhibition assays have been developed for antigen detection and offer at least one potential advantage, namely, reduced antibody requirements. We have developed an inhibition assay using the polyclonal antibody employed in our standard sandwich assay. Urine and serum specimens from patients with culture-proven histoplasmosis and controls were tested using both methods. The two methods had similar sensitivities for detection of antigen in urine (antibody sandwich = 92.5% versus inhibition = 87.5%, P = 0.500) and serum (82.5% versus 80.0%, P = 1.000). With serum, the specificities of both methods were similar (antibody sandwich assay = 95.0% versus inhibition assay = 92.5%, P = 1.000), and with urine, the specificity of the antibody sandwich method was superior (97.5% versus 80.0%, P = 0.039). While the overall reproducibility of both methods was excellent (with urine, antibody sandwich assay intraclass correlation coefficient = 0.9975 and with serum = 0.9949; correlation coefficient of the inhibition assay with urine = 0.9736 and with serum = 0.9850), that of the inhibition method was only fair to poor for the controls: urine = −0.0152, serum = 0.5595. Reproducibility was good for the controls using the sandwich method: urine = 0.7717, serum = 0.9470. Cross-reactivity was observed in specimens from patients infected with Blastomyces dermatitidis, Paracoccidioides brasiliensis, and Penicillium marneffei. In conclusion, the decreased specificity and inferior reproducibility with control specimens suggest that the inhibition assay has poorer precision toward the lower end of the detection range.

The Histoplasma antigen immunoassay is a well-established method for the rapid diagnosis of histoplasmosis (11, 13, 14). In addition, the assay is semiquantitative and can be used to monitor treatment and identify relapse (6, 9). The method used is an antibody sandwich enzyme immunoassay (EIA) employing anti-Histoplasma capsulatum polyclonal antibody for both the capture and detection steps. The sensitivities in serum and urine for patients with disseminated disease are 82 and 92%, respectively (14), specificity is >98% (12), and reproducibility is excellent (2, 12). However, production of polyclonal antibody of sufficient sensitivity and specificity toward the carbohydrate antigen has had limited success and has proven to be time consuming, laborious, and expensive. Also, a cross-reacting antigen may be detected in samples from patients infected with Blastomyces dermatitidis, Paracoccidioides brasiliensis, and Penicillium marneffei, but a previous study determined that the Histoplasma antigen immunoassay could be used to assist in the diagnosis of these cross-reacting mycoses because the treatment regimens for all are similar and the epidemiological and clinical differences among these mycoses can be used for differential diagnosis (8).

An inhibition assay would enable more precise quantitation of the Histoplasma antigen (HcAg) and reduce the amount of antibody required for testing. An inhibition EIA has been described that uses a murine monoclonal antibody to detect the presence of an H. capsulatum protein antigen in the serum and urine of patients with histoplasmosis (4; L. J. Wheat, Letter, J. Clin. Microbiol. 37:2387, 1999). The overall sensitivity in serum was 71.4% for all clinical forms of histoplasmosis. However, antigen was not detected well in urine (44.0% overall sensitivity) and cross-reactivity occurred in sera from patients with aspergillosis, cryptococcosis, and tuberculosis, features not observed in the Histoplasma antigen immunoassay.

In this paper, we report the development of an inhibition EIA capable of detecting HcAg in both serum and urine specimens. The sensitivity, specificity, reproducibility, and cross-reactivity of the inhibition EIA are compared to those of the standard antibody sandwich EIA.

MATERIALS AND METHODS

Specimens.

All specimens were stored at or below 4°C. Forty serum specimens and 40 urine specimens from patients with culture-proven histoplasmosis, collected from 1992 through 1997, and 40 serum controls and 40 urine controls, originally requisitioned for pregnancy or toxicology screening, were tested concurrently in a blind fashion by the antibody sandwich EIA and the inhibition EIA. Previous studies have shown that antigen levels from histoplasmosis cases are unevenly distributed over the complete result range of negative (<1.0 EIA units [EU]) (2) to highly positive (>10.0 EU) (7, 10). For this study, specimens from culture-proven histoplasmosis cases were selected so that the distribution of the original results mirrored that of untreated patients with disseminated histoplasmosis.

Of the 40 serum specimens from culture-proven cases, 16 were highly positive (>10 EU), 14 were moderately positive (2.1 to 10.0 EU), 4 were weakly positive (1.0 to 2.0 EU), and 6 were negative (<1.0 EU) when originally tested by the Histoplasma antigen immunoassay. Of the 40 urine specimens from culture-proven cases, 20 were highly positive, 14 were moderately positive, 4 were weakly positive, and 2 were nega- tive when originally tested. Serum and urine specimens were tested by separate assays.

Fifty urine specimens from patients with other fungal infections or tuberculosis, collected from 1983 through 1997, were tested separately using both methods. The mycoses evaluated were blastomycosis, paracoccidioidomycosis, coccidioidomycosis, aspergillosis, candidiasis, cryptococcosis, penicilliosis, and sporotrichosis.

Antibody sandwich EIA.

The antibody sandwich EIA method used has been previously described (2).

Inhibition EIA.

The procedures for preparing the inhibition plate and making standards and dilution buffer are outlined below and are based upon procedures described by Gomez et al. (4). The reaction plate procedure was developed in house.

(i) Inhibition plate.

A 340-μl volume of phosphate-buffered saline containing 0.05% Tween 20 and 5% (wt/vol) bovine serum albumin (BSA; Sigma, St. Louis, Mo.) was added to each well of 96-well flat-bottom polystyrene plates (Costar, Ann Arbor, Mich.). The plates were incubated at 37°C for 2 h and then washed three times with phosphate-buffered saline containing 0.05% Tween 20 (PBS-Tween). The wells of these plates were used as basins where antibody was mixed with patient specimens to allow immune complexes to form.

(ii) Standards.

Standards were made by serially diluting a 10.0-mg/ml stock of an NaOH extract of whole H. capsulatum yeast cells (NaOH extract) in either freshly collected normal human urine or commercially available normal human serum (sterile filtered, non-lipid stripped, off the clot; Scantibodies Laboratory, Inc., Calabasas, Calif.). The method used for production of the NaOH extract was that of Reiss et al. (5) and yields a crude polysaccharide extract of H. capsulatum. The standards ranged from 0.005 to 50.000 μg/ml and were stored at −40°C. The same lot of standards was used throughout the study. The standards' values were plotted on a semilog scale with a sigmoid four-parameter regression used to fit the standard curve.

(iii) Dilution buffer.

Standards, patient specimens, and controls were mixed with a dilution buffer consisting of either normal human serum or normal human urine diluted 1:5 in PBS-Tween containing 0.02 M MgCl₂ and 1% (wt/vol) BSA.

(iv) Inhibition.

Standards, patient specimens, and controls were diluted 1:2 with dilution buffer in the wells of the inhibition plates. All samples were tested in duplicate. A 100-μl volume of an optimal dilution of biotinylated rabbit anti-H. capsulatum polyclonal immunoglobulin G in dilution buffer was next added to each well and mixed. The inhibition plates were incubated for 30 min at room temperature with shaking at 500 rpm and then for 16 to 18 h at 4°C without shaking.

(v) Reaction plate.

The wells of 96-well flat-bottom polystyrene plates were coated at 100 μl/well with NaOH extract at 2.0 μg/ml in 0.9% NaCl. These reaction plates were incubated at 37°C for 1 h and then washed extensively with PBS-Tween. A 340-μl volume of 0.01 M Tris-HCl (pH 7.0) containing 5% BSA was then added to each well of the reaction plates. The reaction plates were incubated at 37°C for 1 h and then washed. A 100.0-μl sample from each corresponding well of the inhibition plates was then transferred to the reaction plates. The reaction plates were incubated at 37°C for 1 h and then washed. Horseradish peroxidase-labeled streptavidin (Boehringer Mannheim, Indianapolis, Ind.) in 0.10 M Tris–0.85% NaCl (pH 8.0) containing 5% BSA was then added to the reaction plates. The reaction plates were incubated at 37°C for 1 h and then washed. Each reaction plate was developed by adding TMB Microwell Peroxidase Substrate (Kirkegaard & Perry Laboratories, Gaithersburg, Md.) and incubating it in the dark for 6 min. The reaction was stopped by the addition of 2 N sulfuric acid. Each plate was read on a V-Max reader (Molecular Devices, Sunnyvale, Calif.) at a wavelength of 450 nm.

Statistical analysis.

The reproducibility of each technique was assessed separately with urine and serum samples by calculating the intraclass correlation coefficient (ICC). The ICC is a measure of agreement among repeated tests. The reliability of each assay was also assessed by the true disease state (culture-proven histoplasmosis cases or controls). A logistic regression model was used to predict the true disease state as a function of the inhibition EIA value. This was used to determine a cutoff point for disease classification. The cutoff point was defined as the value of the inhibition assay at which the sum of its sensitivity and specificity is at a maximum (1). The cutoff point for the antibody sandwich EIA has been well established through repetitive testing and experience and was defined as 1.0 EU. Using these cutoff points, the sensitivity and specificity of the tests were compared to assess equality using McNemar's test. All assays for reproducibility were performed within 1 week using the same reagents and standards where applicable. In the inhibition EIA, optical densities at 450 nm falling outside the detectability range set by the standards were assigned values of 0.005 μg/ml for results of <0.005 μg/ml and 50.000 μg/ml for results of >50.000 μg/ml.

RESULTS

Sensitivity and specificity. (i) Antibody sandwich EIA.

The cutoff point of the antibody sandwich EIA was 1.0 EU for serum and urine specimens, as described previously (12). The lower limits of HcAg detection were 0.010 μg/ml in serum and 0.005 μg/ml in urine as determined by the titer of the NaOH extract. HcAg was detected in 33 (82.5%) of 40 serum specimens and 37 (92.5%) of 40 urine specimens from culture-proven histoplasmosis cases (Table 1; Fig. 1A). Seven serum and three urine specimens were falsely negative (Table 2). Of these, four serum and two urine specimens were also originally negative, two serum specimens and one urine specimen were originally weakly positive, and one serum specimen was originally highly positive. All false-negative results were confirmed as negative upon retesting. Two serum case specimens that were originally negative were confirmed as weakly positive.

TABLE 1.

Comparison of the sensitivities and specificities of the antibody sandwich and inhibition EIAs for detection of HcAg

Method	No. (%) of specimens positiveb
	Serum		Urine
	Sensitivity	Specificity	Sensitivity (%)	Specificity (%)
Antibody sandwich EIA	33 (82.5)	38 (95.0)a	37 (92.5)	39 (97.5)a
Inhibition EIA	32 (80.0)	37 (92.5)	35 (87.5)	32 (80.0)

^aFalse-positive specimens had either excessive variability between duplicates (CV, >20.0%) or duplicate values consisting of one positive value and one negative value. All but one were subsequently confirmed as negative with further testing. 

^bForty specimens were tested in each case. 

FIG. 1.

Antibody sandwich EIA and inhibition EIA reactivity. Graphs: A, antibody sandwich EIA; B, inhibition EIA. The dotted line represents the cutoff point (antibody sandwich EIA, 1.00 EU; inhibition EIA, 0.080 [serum] and 0.010 [urine] μg/ml). An open circle indicates a value with excessive duplicate variation (CV, >20.0%). For graph A, n is the number of results contained within the darkened box. For graph B, n is the number of results that are ≤0.005 μg/ml.

TABLE 2.

Comparison of false-negative results of the antibody sandwich and inhibition EIAs

Specimen type	Inhibition EIA		Sandwich EIA		Original EIA/RIAb
	Result (μg/ml)	Interpretationa	Result (EU)	Interpretationa	Result (RUc/EU)	Interpretationa
Serum	0.011	Neg	0.8	Neg	13.3	High pos
	0.066	Neg	5.0	Mod pos	2.8	Mod pos
	0.058	Neg	4.3	Mod pos	1.7	Wk pos
	0.623	Pos	0.6	Neg	1.5	Wk pos
	0.216	Pos	0.8	Neg	1.1	Wk pos
	0.019	Neg	0.7	Neg	0.8	Neg
	0.019	Neg	0.9	Neg	0.8	Neg
	0.005	Neg	0.6	Neg	0.8	Neg
	0.592	Pos	1.3	Wk pos	0.7	Neg
	0.005	Neg	1.4	Wk pos	0.5	Neg
	0.005	Neg	0.6	Neg	0.5	Neg
Urine	0.009	Neg	2.3	Mod pos	3.8	Mod pos
	0.008	Neg	1.8	Wk pos	2.9	Mod pos
	0.005	Neg	0.7	Neg	1.3	Wk pos
	0.005	Neg	0.7	Neg	0.8	Neg
	0.005	Neg	0.7	Neg	0.6	Neg

^aPos, positive; Neg, negative; Wk pos, weakly positive; Mod pos, moderately positive; High pos, highly positive. 

^bRIA, radioimmunoassay. 

^cRU, RIA units (2). 

Thirty-eight (95.0%) of 40 serum controls and 39 (97.5%) of 40 urine controls were negative (Table 1; Fig. 1A). The two control serum specimens and one control urine specimen false positive in the antibody sandwich EIA were weakly positive but had either excessive variability between duplicates (coefficient of variation [CV], >20%) or duplicate values consisting of one positive value and one negative value. Two of these controls were negative upon retesting, and one serum control continued to display unacceptable duplicate variability. Further testing never yielded a reliable result.

(ii) Inhibition EIA.

The cutoff and lower limits of HcAg detection of the inhibition assay were determined to be 0.080 μg/ml in serum and 0.010 μg/ml in urine. All standard curves had coefficients of detection (r²) of greater than 0.986 with a mean and standard deviation of 0.995 ± 0.005.

The sensitivity of the inhibition EIA for detection of HcAg in both the serum and urine specimens of patients with histoplasmosis was similar to that of the antibody sandwich EIA (serum: P = 1.000; urine: P = 0.500). HcAg was detected in 32 (80.0%) of 40 serum specimens and 35 (87.5%) of 40 urine specimens from culture-proven histoplasmosis cases (Table 1; Fig. 1B). Eight serum and five urine specimens were false negative in the inhibition EIA (Table 2). Of these, five serum and two urine specimens were originally negative, one serum and one urine specimen were originally weakly positive, one serum and two urine specimens were originally moderately positive, and one serum specimen was originally highly positive.

The specificity of the inhibition EIA was similar in serum and reduced in urine specimens compared to the antibody sandwich EIA (serum, P = 1.000; urine, P = 0.039). Thirty-seven (92.5%) of 40 serum controls and 32 (80.0%) of 40 urine controls were negative (Table 1; Fig. 1B). All three false-positive serum samples and two of the eight false-positive urine samples were confirmed with retesting.

Reproducibility. (i) Antibody sandwich EIA.

Reproducibility was uniform throughout the complete result range of the antibody sandwich EIA for both serum and urine specimens. The ICCs for serum and urine specimens from culture-proven cases and controls were 0.9949 and 0.9975, respectively (Fig. 2). The ICCs of specimens from culture-proven cases were 0.9913 for serum and 0.9952 for urine specimens. The ICCs of control specimens were 0.9470 for serum and 0.7717 for urine specimens.

FIG. 2.

Antibody sandwich EIA reproducibility with serum and urine specimens. Graphs: A, serum specimens; B, urine specimens. Closed circles represent specimens from culture-proven histoplasmosis cases. Open circles represent control specimens. The dotted line represents the cutoff point (1.00 EU). The ICC of serum specimens is 0.9949. The ICC of urine specimens is 0.9975.

(ii) Inhibition EIA.

Reproducibility was better for specimens from culture-proven cases than for those from controls. The ICCs of serum and urine specimens from cases and controls were 0.9850 and 0.9736, respectively (Fig. 3). The ICCs of specimens from culture-proven cases only were 0.9841 for serum and 0.9708 for urine. The ICCs of specimens from controls were 0.5595 for serum and −0.0152 for urine specimens.

FIG. 3.

Inhibition EIA reproducibility with serum and urine specimens. Graphs: A, serum specimens; B, urine specimens. Closed circles represent specimens from culture-proven histoplasmosis cases. Open circles represent control specimens. The dotted line represents the cutoff point (serum, 0.080 μg/ml; urine, 0.010 μg/ml). The ICC of serum specimens is 0.9850. The ICC of urine specimens is 0.9736. n is the number of reproducible results that are ≤0.005 μg/ml.

Cross-reactivity.

To evaluate the cross-reactivity of the antibody sandwich and inhibition EIAs, 50 urine specimens from patients diagnosed with tuberculosis or a mycosis other than histoplasmosis were tested using each method. In the antibody sandwich EIA, cross-reactivity was seen in specimens from patients infected with B. dermatitidis (88.9%), P. brasiliensis (88.9%), or P. marneffei (85.7%) (Fig. 4A). The same cross-reactivity pattern was seen in the inhibition EIA (B. dermatitidis: 66.7%; P. brasiliensis: 44.4%; P. marneffei: 71.4%) (Fig. 4B). Cross-reactions were not observed with urine specimens from individuals with tuberculosis, coccidioidomycosis, aspergillosis, candidiasis, cryptococcosis, or sporotrichosis for either assay.

FIG. 4.

Cross-reactivity of the antibody sandwich EIA and inhibition EIA. Graphs: A, antibody sandwich EIA; B, inhibition EIA. All of the specimens tested were urine. The dotted line represents the cutoff point (antibody sandwich EIA, 1.00 EU; inhibition EIA, 0.010 μg/ml). The other mycoses are coccidioidomycosis (n = 5), candidiasis (n = 5), cryptococcosis (n = 6), aspergillosis (n = 5), and sporotrichosis (n = 2). An open circle indicates a value with excessive duplicate variation (CV, >20.0%).

DISCUSSION

The antibody sandwich EIA was slightly, but not significantly, more sensitive for detecting HcAg in serum and urine than was the inhibition EIA. The lower detection limits of the inhibition EIA were eightfold higher for serum samples and twofold higher for urine samples compared to the sandwich EIA. In this study, the sensitivities of the antibody sandwich EIA for serum and urine were comparable to previous observations (14).

Both methods failed to detect antigen in some case specimens but detected antigen in others that were originally negative in the Histoplasma antigen immunoassay. While case specimens falsely negative in the antibody sandwich EIA were originally negative or only weakly positive, case specimens falsely negative in the inhibition EIA were originally moderately positive in some cases, illustrating the difference in sensitivity between the methods. One serum specimen, however, originally yielded a highly positive result during clinical testing but was falsely negative in both the antibody sandwich and inhibition EIAs. This discrepancy may be explained in one of two ways: the original result was positive but the specimen lost reactivity during storage, or the original specimen was mislabeling before storage.

The specificity of the sandwich assay for urine specimens of healthy controls was superior to that of the inhibition assay, while its specificity for serum samples was similar. In our clinical testing, all three controls falsely positive in the antibody sandwich assay would not have been classified as positive due to unacceptable duplicate values. Two of the three would have been classified as negative upon retesting. The third did not yield a reliable result with repeated testing. These findings are consistent with our earlier observations that the specificity of the sandwich assay is >98% (2, 12). In the inhibition assay, all three false-positive serum controls and two of the eight false-positive urine controls were confirmed with retesting. All specimens falsely positive in the inhibition assay that were confirmed were within the lower range of positivity (serum, 0.103 to 0.246 μg/ml; urine, 0.013 to 0.038 μg/ml), implying that the inhibition assay cannot clearly differentiate cases with low positive results from control specimens.

Both methods yielded positive results for patients with mycoses known to share cross-reactive antigens with H. capsulatum. This observation confirms our previously published observations (8). Others, using a monoclonal antibody inhibition EIA, reported cross-reactivity for patients with aspergillosis (20.0%), cryptococcosis (10.0%), and tuberculosis (22.2%) (4). Cross-reactivity with these infections has not been observed in the Histoplasma antigen immunoassay (2, 8, 12) and was not observed for either of the methods investigated in this study. Those investigators did not study cross-reactivity in patients with blastomycosis or penicilliosis but did report cross-reactivity in patients with paracoccidioidomycosis (20.0%) (4).

The antigenic targets of the two inhibition assays differ, providing a possible explanation for the differences in cross-reactivity. Also, those investigators used a monoclonal antibody to an H. capsulatum protein antigen whereas in this investigation we used a polyclonal antibody to whole H. capsulatum. While we would have preferred to use a monoclonal antibody, we have been unable to prepare monoclonal antibodies capable of detecting the carbohydrate antigen found in patients (3).

Reproducibility was excellent for both the antibody sandwich EIA and the inhibition EIA when results from cases and controls were analyzed together. When cases and controls were analyzed separately, the reproducibility of the sandwich method remained excellent. However, the reproducibility of the inhibition EIA, although excellent for cases, was poor for serum controls and fair for urine controls, reflecting impaired precision toward the lower end of the detection range. Poor precision in this range obscures the determination of the cutoff value used to discriminate positive from negative results and impedes the assay's ability to clearly distinguish between histoplasmosis cases and controls, adversely affecting the sensitivity and specificity of the assay. In clinical testing, fewer than 5% of specimens are positive for HcAg (unpublished observation). Therefore, at least 95% of results would occur in this area of reduced assay precision and reproducibility, presumably leading to an increased false-positivity rate for patients without histoplasmosis.

In summary, while the inhibition method is capable of detecting HcAg in cases of histoplasmosis with sensitivity and reproducibility similar to those of the sandwich method, its poorer performance at the low end of the detection range may lead to increased false positivity. This shortcoming appears to be an inherent attribute of the inhibition method, not of the polyclonal antibody employed. Therefore, the inhibition method is not a viable antibody-efficient alternative to the established sandwich method utilized in the Histoplasma antigen immunoassay.