INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

Human granulocytic ehrlichiosis (HGE) was first reported in 1994 from Wisconsin and Minnesota (2, 5) and subsequently from other areas of the United States (1, 9, 32, 36) and Europe (17, 25, 34). The etiologic agent of HGE is an obligatory intracellular bacterium that belongs to the Ehrlichia equi- Ehrlichia phagocytophila group on the basis of 16S rRNA gene sequence comparison (5) and serological cross-reactivity (8). Development of a specific, sensitive, and rapid diagnostic method to distinguish HGE from other tick-borne infections is desirable to ensure appropriate antimicrobial therapy, because Ixodes spp. are the vector for the HGE agent, Borrelia burgdorferi, and Babesia microti and the coexistence of the latter two pathogens and the HGE agent in ticks has been demonstrated by using PCR (24, 28, 29, 35). Serologic data suggest human exposure to all three agents in Connecticut, Minnesota, and Wisconsin (18-20), and HGE and Lyme borreliosis have simultaneously been reported (7, 21). At present, there is no single "gold standard" for the diagnosis of HGE. Culture isolation, PCR, and serology each have strengths and weaknesses. Positive results in culture isolation or PCR tests are definitive, but negative results in culture isolation or PCR are not definitive. Negative indirect immunofluorescence assay (IFA) results with samples tested at both acute and convalescent stages may be definitive (if samples are not from immunocompromised individuals and antibiotic treatment is not initiated too early), but positive IFA results may not be definitive because of the subjective nature of evaluation and occasional cross-reactivity with other antigens. Comparison of sensitivity and specificity is, therefore, meaningful when compared among the same type of assay, such as serology, but not when compared between different types of tests. IFA using HGE agent- or E. equi-infected cells as antigen is currently the most widely used method for diagnosis of HGE (3, 6, 22). Although IFA is sensitive and simple, it has several problems. Since the HGE agent is an obligatory intracellular bacterium, it is necessary to culture the HGE agent in eukaryotic host cells to prepare infected-cell antigen. Culturing is labor-intensive and produces batch-to-batch variation in antigens. Furthermore, the use of whole infected cells as antigen may increase the false-positive rate due to antigenic cross-reactivity. The visual evaluation of test results precludes rapid testing of a large number of samples, and the subjective evaluation of test results may cause variation in titers among different laboratories and technical personnel. Moreover, the cutoff titers for positive IFA reactions differ among laboratories, ranging from 20 to 80 (3, 4, 6, 21, 34, 37, 38).

Enzyme-linked immunosorbent assay (ELISA) is desirable for automated testing of large numbers of serum samples. Ravyn et al. (26) reported that an ELISA using native HGE agent cultured in HL-60 cells as antigen is more sensitive than IFA. However, some samples are ELISA positive but Western blot negative. They recommended a two-test method of screening by ELISA and confirmation of specificity by Western blot analysis (26). Recently, an ELISA for HGE using a recombinant surface 44-kDa protein (HGE-44) of the HGE agent (NCH-1 strain isolated from a patient in Massachusetts) fused with maltose binding protein was reported (total molecular mass of approximately 80 kDa) (12). We previously demonstrated that a dot blot assay using the recombinant major surface 44-kDa protein P44 (rP44) of the HGE agent (strain no. 13, isolated from a patient in New York) is useful for serodiagnosis of HGE (39). Further molecular analysis revealed that among approximately 20 P44 proteins of the multigene family, P44-2 is most abundantly expressed by the HGE agent in HL-60 cell culture, and HGE patients' sera reacted with a synthetic peptide specific to P44-2 (40). In the present paper, we compare the usefulness of our rP44 and the recombinant P44-2 hypervariable region (rP44-2hv) as antigen with ELISAs using various sera, including B. microti antibody-positive sera, and investigated the need for either Escherichia coli preabsorption of sera or cleavage of fused vector peptide from the fusion protein.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Sera. A total of 109 sera were collected from 72 patients at Westchester Medical Center in New York State from June 1995 to September 1997 who were suspected of having HGE based on clinical signs and exposure to ticks. Sera were collected once from 45 patients, twice from 17 patients, and three times from 10 patients, with more than 4 days between collection days. Convalescent-phase sera from nine patients from whom the HGE agent was isolated and five sera from patients who were HGE-PCR positive were used as positive controls (11, 38). Nested PCR as described by Chen et al. (5) and Sumner et al. (31) was used to detect the HGE agent DNA in blood specimens. Sera from 20 healthy humans from Westchester, New York, and sera from 53 healthy humans in Japan, kindly provided by Makoto Kawahara, Nagoya City Public Health Research Institute (Nagoya, Japan), were used as negative controls. HGE is endemic in Westchester but has not yet been identified in Japan. Ten sera, which were demonstrated to be seropositive by ELISA and Western blotting analysis for B. burgdorferi (10), were also tested in the present study. Two sera from B. microti-infected patients were supplied by Lily I. Kong, MRL Diagnostic Laboratory (Cypress, Calif.). Five sera, positive for antibodies against Ehrlichia chaffeensis by IFA and Western immunoblotting (33), were provided by MRL Diagnostic Laboratory.

IFA. IFA was performed as previously described (27). Briefly, the HGE agent (isolate no. 13, referred to as New York isolate [27]) was cultured in the human promyelocytic leukemia cell line HL-60. Heavily infected (>80% infected) cultures of cells were suspended in RPMI 1640 medium and were dispensed onto 12-well slides at a concentration of 10⁴ cells/well. A twofold serial dilution of test sera starting at 1:20 was prepared in 2× phosphate-buffered saline (PBS; 19 mM Na₂HPO₄, 12 mM NaH₂PO₄, 300 mM NaCl, pH 7.4). Ten microliters of each dilution of the serum was reacted with antigen at 37°C for 1 h. After washing, the slides were reacted with 10 µl of fluorescein isothiocyanate-conjugated goat anti-human immunoglobulin G (IgG) (Organon Technika, Westchester, Pa.) at a dilution of 1:200. After incubation at 37°C for 1 h, the slides were washed, counterstained with Evans blue (Sigma, St. Louis, Mo.), and examined under the epifluorescent microscope. The serum antibody titer was expressed as the reciprocal of the highest dilution of serum that showed a positive reaction. Serum that had an antibody titer greater than 1:20 was considered positive because all the sera from healthy individuals used in this study, except one false-positive serum, were IFA negative at a serum dilution of 1:20.

rP44 and rP44-2hv antigen. E. coli BL21(DE3)/pLysS (Novagen, Inc., Madison, Wis.) transformed with pEP44, the recombinant pET30a vector (39), was cultured, and rP44 protein was purified by using His-Bind Resin (Novagen) as previously described (39).

ELISA. A 2-µg/ml concentration of each recombinant antigen was used in the present experiment. Fifty microliters of recombinant antigen diluted with 0.05 M carbonate buffer (pH 9.6) was absorbed onto individual wells of Coster 96-well enzyme immunoassay-radioimmunoassay plates (Corning, N.Y.) and was left overnight at 4°C in a humidified box. Excess antigen solution was removed, and the wells were then coated with 100 µl of 5% nonfat dried milk (Kroger, Cincinnati, Ohio) in 2× PBS at 37°C for 1 h to block non-specific protein binding sites. Sera diluted 1:20 with 5% nonfat dried milk in 2× PBS, with or without preabsorption with E. coli, were added to respective wells (50 µl/well) and were incubated at 37°C for 1 h. A twofold serial dilution of positive control sera was prepared with 5% nonfat dried milk in 2× PBS and was reacted with EK-rP44 in the same manner. After three washes with 0.05% Tween 20 in 2× PBS, 50 µl of a 1:1,000 dilution of peroxidase-conjugated goat anti-human IgG (Kirkegaard & Perry Laboratories, Inc., Gaithersburg, Md.) was added to each well and incubated at 37°C for 1 h. The wells were again washed as described above, and 100 µl of a mixture of 0.2 mM 2,2'-azinobis(3-ethylbenzthiazolinesulfonic acid) and 0.004% H₂O₂ in 0.006 M citric acid-0.008 M Na₂PO₄ (pH 4.0) was added. After incubation at room temperature for 20 min, 50 µl of 1 M H₂SO₄ was added to stop the reaction, and the optical density at 405 nm (OD₄₀₅) was measured. The assay was repeated three times for each sample with different batches of rP44 protein-coated plates, and the reproducibility was confirmed. The antibody titer of positive control serum was expressed as the reciprocal of the highest dilution of serum that showed a positive reaction.

Preabsorption of sera. E. coli BL21(DE3)/pLysS transformed with pET30a vector suspended in 2× PBS was disrupted by sonication and dispensed to each well of a 96-well microplate (100 µl/well) and centrifuged at 400 × g for 15 min. The supernatant was removed, sera diluted 1:20 with 2× PBS containing 5% nonfat dried milk was added to each well, and the plate was incubated at 37°C for 30 min and further incubated at 4°C overnight. Then, the plate was centrifuged at 400 × g for 15 min, and the supernatant was used in ELISA.

Western blotting analysis. Sephacryl S-1000 column-purified HGE agent (10 µg) (38) and rP44 protein (3 µg) separated by 10% sodium dodecyl sulfate-polyacrylamide gel electrophoresis were transferred to a nitrocellulose membrane and immersed in Dulbecco's PBS containing 5% nonfat dried milk at 4°C overnight to block nonspecific reactions. The membrane was reacted with sera diluted 1:100 followed by incubation with peroxidase-conjugated goat anti-human IgG at the dilution of 1:2,000. As a positive control, a mouse monoclonal antibody against the 44-kDa protein of the HGE agent, 5C11 (16), and peroxidase-conjugated goat anti-mouse Igs (IgG, IgA, and IgM) (ICN Pharmaceuticals, Aurora, Ohio) were used at dilutions of 1:500 and 1:2,000, respectively. The peroxidase reaction was carried out in 70 mM sodium acetate buffer (pH 6.2) containing 0.3% diaminobenzidine tetrahydrochloride (Nakarai Tesque, Inc., Kyoto, Japan) and 0.03% H₂O₂, and the reaction was stopped by washing the membrane in 0.1 M H₂SO₄.

Protein assay. Protein concentration was measured by using BCA Protein Assay Kit (Pierce, Rockford, Ill.) using bovine serum albumin as standard.

Statistical analysis. The data were analyzed by STATVIEW, version 4, for Macintosh (Abacus Concepts, Berkeley, Calif.) to determine Spearman's rank correlation coefficient.

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

Determination of the cutoff value in ELISA. Sera from 20 healthy humans in the United States and 52 of 53 healthy humans in Japan were IgG antibody negative (<1:20) by HGE-IFA. One IFA-positive serum was further examined by Western immunoblot analysis as described later. ELISA was independently performed three times for each IFA-negative serum using EK-rP44, rP44, or EK-rP44-2hv as antigen. The mean and standard deviation (SD) of OD₄₀₅ of each sample were calculated. Under the assumption of a normal (Gaussian) distribution, the expected true negative rate is 99.9% if the cutoff value selected is equal to the mean of the negative reference serum plus three times the SD (15). When EK-rP44 was used as an antigen, the mean absorbance of these negative sera was 0.142 and the SD was 0.040. Therefore, the cutoff value was 0.262. When rP44 was used as an antigen, the mean absorbance and SD were 0.143 and 0.033, respectively, and the cutoff value was 0.243. The mean absorbance and SD were 0.074 and 0.020 for EK-rP44-2hv, and the cutoff value was 0.134.

ELISA using positive control sera. Convalescent-phase sera from 14 patients from whom HGE agent was isolated and/or who were HGE-PCR positive (11, 38) were used as positive controls in ELISA testing. The results from ELISA using EK-rP44 as an antigen are summarized in Table 1. The tests were independently repeated three times with ELISA plates coated with different batches of recombinant antigen; in every test all the samples were positive. The antibody titers were determined by IFA and ELISA were similar. Coefficients of variation were calculated by dividing the SDs of replicates by the means of replicates. A value of less than 20% indicates adequate reproducibility (15). In the present results, coefficients of variation of all samples were less than 20%, indicating the adequate precision of this system. The results of ELISA using rP44 or EK-rP44-2hv were similar (data not shown).

Reactivity of sera from patients positive for antibodies to E. chaffeensis, B. burgdorferi, or B. microti. To check cross-reactivity in ELISA, 4 E. chaffeensis-, 10 B. burgdorferi-, and 2 B. microti-positive sera were tested. Sera from B. microti-infected patients reacted with rP44 antigen in ELISA, and OD₄₀₅s were 0.858 and 0.786, respectively. However, these reactivities disappeared after digestion of rP44 with EK, and the OD₄₀₅s reduced to 0.147 and 0.122, respectively. These sera from B. microti-infected patients did not react with EK-rP44-2hv antigen (OD₄₀₅ < 0.139). None of E. chaffeensis- or B. burgdorferi-infected sera reacted with rP44 antigen (OD₄₀₅ < 0.243), EK-rP44 antigen (OD₄₀₅ < 0.262), or EK-rP44-2hv antigen (OD₄₀₅ < 0.139).

ELISA of patient sera using EK-rP44 or rP44. One hundred nine sera from 72 patients suspected of having HGE were examined by ELISA testing. All IFA-negative sera were negative by ELISA using EK-rP44 or rP44. Of 39 IFA-positive (>1:20) sera, 35 from 21 patients were positive by ELISA using EK-rP44 or rP44. To verify that an ELISA-positive reaction was not against E. coli proteins that may be present in the affinity-purified rP44 antigen preparation, all patient sera were preabsorbed with E. coli transformed with pET-30a expression vector in order to remove antibodies which might be present and potentially react with E. coli proteins, and the ELISA was repeated. The same results as obtained without preabsorption of sera were obtained (data not shown). We, therefore, conclude that the ELISA reactivity was against rP44, not against E. coli proteins, and that preabsorption of sera with E. coli is not required for this ELISA.

View larger version (36K): [in this window] [in a new window]   FIG. 1.   Western blot analysis of patient sera by using the native HGE agent as an antigen. Samples subjected to sodium dodecyl sulfate-polyacrylamide gel electrophoresis consisted of 10 µg of purified whole-cell preparation of HGE agent. The protein was transferred to nitrocellulose membrane and was incubated with a 1:100 dilution of sera (B66, J14, B20, A30, B02, B14, MoAb, or J01). Numbers at the left are molecular mass standards (in kilodaltons) based on the broad-range prestained standards (Bio-Rad Laboratories, Richmond, Calif.).

Comparison of results of IFA and ELISA using EK-rP44. The results of IFA and ELISA using sera from patients and healthy individuals were compared (Table 3). Usually, a new assay is evaluated by comparison with another serological assay or combination of assays, and relative sensitivity and specificity are calculated by making the previous assay result the standard. In the present study, however, false-positive cases in IFA were observed and ELISA was more specific than IFA. In such a case, the relative diagnostic sensitivity and specificity can be calculated by making the new method the standard of comparison (15). The relative diagnostic sensitivity and specificity of IFA, calculated by this manner, were 100 and 97%, respectively. These results indicate that ELISA using EK-rP44 is as sensitive as IFA and more specific than IFA. The correlation of the results of IFA titers and ODs of ELISA using EK-rP44 is shown in Fig. 2. The rho value calculated by Spearman's rank correlation is 0.740 and is statistically significant (P < 0.001), indicating a positive correlation between IFA titer and OD of EK-rP44 ELISA.

View larger version (13K): [in this window] [in a new window]   FIG. 2.   Correlation between IFA titers and ODs of ELISA by using EK-rP44 as an antigen. Each circle represents one sample. The rho value calculated by Spearman's rank correlation was 0.740 (P < 0.001) (n = 181).

Comparison of results of ELISAs using EK-rP44 and EK-rP44-2hv. All IFA-negative sera were negative by ELISA using EK-rP44-2hv. Of 35 EK-rP44 ELISA-positive sera, 27 sera from 16 patients were positive by EK-rP44-2hv ELISA. Three patients were negative by EK-rP44-2hv ELISA but positive by EK-rP44 ELISA for two different times. The results of ELISAs using EK-rP44 and EK-rP44-2hv are compared in Table 4. The relative diagnostic sensitivity and specificity was calculated by making the results of EK-rP44 antigen the standard of comparison. The relative diagnostic sensitivity of the EK-rP44-2hv antigen was 77% and specificity was 100%. These results indicate that EK-rP44 antigen is more sensitive than EK-rP44-2hv antigen in detecting antibodies against the HGE agent in patients.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

The 38- to 49-kDa proteins of the HGE agent have been shown to be immunodominant antigens in human infection (13, 26, 38). IgG antibodies against 44-kDa protein were detected in all culture-positive and PCR-positive patients' sera (38, 39) and in seven of nine acute-phase patients (1 week after onset of symptoms) and in 10 day sera of mice exposed to Ehrlichia-infected (NCH-1 strain) ticks (13). These proteins are encoded by the p44 multigene family (40). Recently, Zhi et al. cloned several genes belong to this family (39, 40). Zhi et al. demonstrated that Western blot analysis and dot immunoblot assay using rP44 as antigen were as specific and sensitive as IFA (39). Furthermore, dot immunoblot assay using a synthetic oligopeptide specific to the hypervariable region, P44-2hv of one of P44 proteins, P44-2, and convalescent sera from patients with HGE demonstrated that antibody specific to P44-2hv was developed in these patients (40). However, Western blotting analysis and dot immunoblot assay are not convenient for automated testing of a large number of clinical samples. We developed an ELISA system using a recombinant P44 as antigen because ELISA is easily adapted to automation, allowing rapid testing of a large number of patient samples at a relatively low cost. There was significant positive correlation between IFA titers and ODs of rP44 ELISA. Since another report of ELISA using recombinant HGE-44 does not describe ODs (12), it is difficult to compare that data with our data. Although this was not examined in another ELISA study (12), preabsorption of patient sera with E. coli components which might be present in the purified recombinant protein is not required in our ELISA using recombinant proteins. Because it may be important for diagnosis to detect low-titer antibody in sera without false-positive results, we used the serum dilution of 1:20. In another report of an ELISA using recombinant HGE-44 fused with maltose binding protein, cutoff ELISA ODs were chosen at 0.45 (1:160 serum dilution), 0.38 (1:320 serum dilution), and 0.26 (1:640 serum dilution) (12). These cutoff ODs are higher than our cutoff OD, especially considering the serum dilution used. Since the OD ranges of positive sera were not described and ODs of the E. coli control or maltose binding protein alone were not shown in that study, it is difficult to interpret the high background OD of that ELISA. This variation may be due to the difference of HGE agent strains or 44-kDa protein genes cloned or expression vectors used. Amino acid identities between P44 and HGE-44 and between P44-2 and HGE-44 are 75.3 and 80.2%, respectively (40).

In our ELISA systems using EK-rP44 and EK-rP44-2, no sera from the limited number of patients infected with B. burgdorferi, B. microti, or E. chaffeensis had positive reactions. Although B. microti-infected patients' sera reacted with rP44 without EK treatment, the reactivity disappeared after treatment with EK or rP44. The exact reason for this cross-reactivity is unknown. The rP44 used in the present experiments was cloned in the pET-30a expression vector. This vector encodes some affinity tags which are useful for assaying expression levels and purifying proteins. EK treatment is able to separate the affinity tags from the recombinant protein. B. microti-infected patients' sera also reacted with recombinant rP30 of Ehrlichia canis (data not shown), which was prepared by using the same vector system (23). These results suggest that the false reactivity might be directed to the affinity tag region. Another report of an ELISA using rHGE-44 fused with maltose binding protein (12) did not examine whether human babesial infection sera cross-react with the fusion protein. B. burgdorferi, B. microti, and the HGE agent are carried by the same tick vector, Ixodes scapularis (28, 35). Simultaneous infections of patients with these agents have been reported (7, 21). Antibiotics effective for these microorganisms are different and it is important to distinguish among these diseases. Treatment of rP44 by EK had no effect on the results of HGE-ELISA. Therefore, an ELISA system using the EK-rP44 described here may be able to distinguish HGE from human monocytic ehrlichiosis, babesiosis, and Lyme borreliosis.

The five sera, which were positive by IFA testing but negative by ELISAs using EK-rP44 and EK-rP44-2hv, were further examined by Western blotting analysis. Three sera from one patient and one healthy individual reacted only with an approximately 70-kDa protein of the HGE agent. Previously, Ijdo et al. reported that heat shock protein 70 (HSP70) of the HGE agent (an 80-kDa protein by their description) was cross-reactive with B. burgdorferi HSP70 (14). HSP70s of many microorganisms such as E. coli, Mycobacterium tuberculosis, and Plasmodium falciparum share common antigenicity. It has been identified as an immunodominant antigen in these infections (30). Although we did not determine whether the 70-kDa protein that reacted with the sera was HSP70 of the HGE agent or not, it is possible that the protein is HSP70. ELISA using EK-rP44 or EK-rP44-2hv could eliminate false-positive reactions due to cross-reactions caused by common antigens including HSP70 present in many microorganisms. One serum reacted with a single 44-kDa band of the HGE agent but not with rP44 in Western blot analysis. The rP44 used in the present study lacks one-third of P44 C terminus (39). Because the HGE agent expresses multiple P44 homologous proteins encoded by a polymorphic multigene family (40), it is possible that this serum reacted with one or more of these P44 homologous proteins distinct from rP44. However, since P44 homologous genes have highly conserved N-terminal regions (40), mouse polyclonal antibody against rP44 strongly recognizes multiple 44- to 42-kDa proteins in six HGE isolates (39), and sera from patients infected with the HGE agent reacted with not only multiple P44s but also with other proteins in Western blotting analysis using the purified HGE agent (13, 26, 38). Thus, it is unlikely that individuals infected with the HGE agent develop an antibody against only a single band of the HGE agent. Lastly, one serum collected more than 3 months after the first sampling, at a time when the IFA titer was significantly decreased, did not react with any proteins of the native HGE agent or rP44 in Western blotting. The disappearance of antibody against the HGE agent after recovery from disease has been reported (11, 26). We speculate that the antibody against the HGE agent had disappeared in this case and that the reactivity in IFA was considered false positive.

The rP44 used in this study is coded by the N-terminal conserved region and a part of the hypervariable region of p44 homologous genes (39, 40). As one of the membrane proteins, this protein is hydrophobic, thus relatively difficult to handle (39). In contrast, rP44-2hv is encoded by hypervariable region of the p44-2 gene and is hydrophilic (40), thus easy to handle. However, the sensitivity of rP44-2 in ELISA was lower than that of rP44. This means that the protein coded by conserved regions of the p44 gene may be required for sensitive detection of anti-HGE antibodies in patients. Alternatively, since three patients were repeatedly positive with rP44 but negative with rP44-2hv antigen, they may be infected with different strains of the HGE agent which lack or do not express p44-2. Because of the excellent specificity, objectiveness, and ease of the assay, this ELISA system using EK-rP44 as antigen is expected to improve serodiagnosis of HGE.