Use of Peptide-Based Enzyme-Linked Immunosorbent Assay followed by Immunofluorescence Assay To Document Ehrlichia chaffeensis as a Cause of Febrile Illness in Nicaragua

Abstract

Ehrlichia chaffeensis, the etiologic agent of human monocytic ehrlichiosis (HME), has been extensively studied as a cause of acute febrile illness and an emerging tick-borne zoonosis in the United States. Limited data suggest its presence in other regions, including Central and South America but not Nicaragua to date. Diagnosis of E. chaffeensis infection by indirect immunofluorescence assay (IFA) is the reference standard due to its presumed high sensitivity and specificity, but IFA is impractical, variably reproducible, and cumbersome for large epidemiologic studies and for clinical diagnosis in resource-poor regions. We evaluated a high-throughput, objective peptide-based enzyme-linked immunosorbent assay (ELISA) for use alone or in combination with IFA. We found that it performed best as a screening test (sensitivity, 100%; specificity, 84%) to reduce the proportion of serum samples that were required by the more cumbersome and subjective IFA testing to <20%. Using a two-step diagnostic approach (IFA is performed if the ELISA is positive), we identified E. chaffeensis or a serologically and antigenically similar organism as a heretofore unrecognized cause of acute febrile illness in humans in Nicaragua and demonstrated the utility of the peptide ELISA as a screening tool for large-scale clinical studies.

INTRODUCTION

Rickettsiae are frequently the cause of acute febrile illness (AFI) in the tropics when rigorously sought; however, they are infrequently identified because of limited diagnostic tools to identify and confirm infections (1, 2). Ehrlichia chaffeensis, the etiologic agent of human monocytic ehrlichiosis (HME), has been extensively studied as a cause of acute febrile illness and an emerging tick-borne zoonosis in the United States (3). Limited data suggest its presence in other regions, including Central and South America but not Nicaragua to date (4). Serologic diagnosis of E. chaffeensis infection by indirect immunofluorescence assay (IFA) is the reference standard due to its presumed high sensitivity and specificity (5, 6). However, the technique is too time-consuming to be practical for large epidemiologic studies, requires the use of a fluorescence microscope, and requires highly experienced operators for reproducibility. Hence, clinical diagnosis and epidemiologic studies of Ehrlichia chaffeensis are not readily done in resource-poor regions. A peptide-based enzyme-linked immunosorbent assay (ELISA) has promise as a high-throughput, objective, and inexpensive diagnostic tool (7). We refined and optimized a peptide-based ELISA for E. chaffeensis infection and applied it to ascertain whether E. chaffeensis is an unrecognized cause of acute febrile illness in humans in Nicaragua.

MATERIALS AND METHODS

Ehrlichia chaffeensis peptide ELISA cutoff optimization.

Convalescent-phase serum samples from patients with PCR-, culture-, and/or blood smear-proven infections with E. chaffeensis (n = 16; E. chaffeensis geometric mean titer [GMT] by IFA of 1,503 [range, 256 to 16,384] and Anaplasma phagocytophilum GMT by IFA of 86 [range, <64 to 640], all with titers that were ≥8-fold higher for E. chaffeensis than for A. phagocytophilum; 2 samples were not tested for A. phagocytophilum antibodies); PCR-, culture-, and/or blood smear-proven infections with Anaplasma phagocytophilum (n = 27; E. chaffeensis GMT by IFA of 82 [range, <80 to 2,560] and A. phagocytophilum GMT by IFA of 640 [range, 160 to 5,120], all but one with titers that are ≥4-fold higher for A. phagocytophilum than for E. chaffeensis; one PCR-positive patient had equivalent IFA titers of 2,560 for E. chaffeensis and A. phagocytophilum); culture-confirmed and serologically confirmed Borrelia burgdorferi (n = 11); and immunohistochemistry and serologically confirmed Rickettsia rickettsii (n = 2) were used for this study. All samples were confirmed to have IgG antibody titers against the respective infectious agents by IFA. Serum samples from healthy individuals who lacked a history and serologic evidence of E. chaffeensis infection by IFA (E. chaffeensis and A. phagocytophilum IgG IFA, <80) of these infections (n = 50) were also used to determine specificity. The peptide ELISA was used according to Luo et al. with modifications (7). Briefly, 96-well microtiter ELISA plates (round bottom; Costar, Corning, NY) were coated with E. chaffeensis synthetic peptide TRP120 alone or in combination with E. chaffeensis synthetic peptide TRP32 (positive peptide wells). Separate wells were coated with Ehrlichia canis peptides TRP36 and TRP200 (control peptide wells). Single peptides were used at a final concentration of 20 μg/ml in phosphate-buffered saline (PBS). When TRP120 and TRP32 were mixed, each was used at a final concentration of 10 μg/ml. Plates were incubated for 1 h at ambient temperature with gentle agitation. After washing 3 times with Tris-buffered saline with 0.2% Tween 20 (TBST), wells were blocked with TBST containing 10% horse serum, incubated at ambient temperature for 1 h, and washed 4 times with TBST. Serum samples were diluted 1:200 in TBST with 10% horse serum, and 50 μl was added to duplicate antigen and control wells. Each plate included 2 positive-control (anti-E. chaffeensis) serum samples and 2 negative-control serum samples. After 1 h, the plates were washed 4 times with TBST, and wells were reacted with 50 μl of horseradish peroxidase-labeled goat anti-human IgG secondary antibody (KPL, Gaithersburg, MD), diluted 1:5,000 in TBST with 10% horse serum, incubated for 1 h at ambient temperature, and again washed 4 times. Thereafter, each well received 100 μl of ABTS [2,2′-azinobis(3-ethylbenzthiazolinesulfonic acid)] peroxidase substrate solution (KPL, Gaithersburg, MD), and the plates were incubated in the dark for 30 min with gentle agitation. Optical densities (ODs) were read at 405 nm for 1 s using a Perkin-Elmer 1420 Victor³ multilabel counter plate reader (Waltham, MA).

To adjust for nonspecific binding in the optimization study, the ODs of duplicate control peptide wells were averaged for each sample tested and used to determine the net individual OD of the duplicate positive peptide wells, which were then averaged. To calculate a cutoff value, the average and standard deviation of the normalized negative serum controls were calculated for each plate. Various cutoffs were examined by adding between 2 and 7 standard deviations (SDs) to the mean of the normalized negative controls. A receiver operator characteristic (ROC) curve was generated using the sensitivity and specificity determined at each cut off. The single-peptide assay using synthetic peptide TRP120 was found to maximize sensitivity while minimizing losses in specificity at 5 SDs above the mean of the negative controls. Thereafter, the single peptide assay was used to analyze the unknown patient samples, with a positive value corresponding to a normalized OD of ≥5 SDs above the mean of the negative controls.

The same approach was used for unknown serum samples, with the following modifications. Each plate used 2 of 3 different positive-control (one high and one low) serum samples and 2 of 4 negative-control serum samples selected from among the negative-control samples used in cutoff optimization. To account for plate-to-plate variations, negative- and positive-control serum ODs were separately averaged, and the difference was used as a divisor for all sample results, normalizing values to high and low references for each plate. To calculate a cutoff value, the average and standard deviation of the normalized negative controls were calculated for each plate. Various cutoffs were examined by adding between 2 and 7 standard deviations (SDs) to the mean of the normalized negative controls. The results of individual plates were accepted if the mean values of the normalized positive and negative controls were within 10% of the coefficient of variation for all plates. Plates that did not meet this criterion were repeated.

IFA to confirm Ehrlichia chaffeensis infections.

An IFA was performed to confirm possible Ehrlichia chaffeensis infections identified by peptide ELISA and in a cohort of 107 randomly selected peptide ELISA-negative samples to determine the frequency of false-negative results. In brief, IFA was performed as previously described (8) using serum samples diluted 1:80 in 0.1 M phosphate-buffered saline (pH 7.4) with 0.5% nonfat dry milk (PBSM), and 10 μl was loaded onto each well of antigen-coated slides. After 1 h of incubation in a humid chamber, the slides were washed 3 times with PBS and air-dried, and 10 μl of fluorescein isothiocyanate-labeled goat anti-human IgG (KPL, Gaithersburg, MD) at a 1:50 dilution in PBSM was added to each well. The slides were incubated for 1 h and washed 2 times in 0.1 M PBS. The antigen slides were then counterstained with 0.005% Evans blue in 0.1 M PBS for 5 min, rinsed with distilled water, mounted using mounting medium (Vectashield mounting medium; Vector Laboratories, Inc., Burlingame, CA), and examined using epifluorescence microscopy. IFAs for R. rickettsii and Rickettsia typhi were performed as per the manufacturer's recommendations (Focus Diagnostics, Cypress, CA). For some samples, IFA titers were provided by contributors at the Centers for Disease Control and Prevention (CDC), Atlanta, GA.

Application of a new high-throughput tool to identify Ehrlichia chaffeensis infections in a large febrile cohort in Nicaragua.

Convalescent-phase serum samples collected from patients who were ≥1 year of age with acute febrile illness (AFI) in Nicaragua between August 2008 and May 2009 (9) were tested using the optimized single-peptide ELISA. Serum samples that were found to be reactive by peptide ELISA were diluted 1:80 and screened for IgG antibodies by IFA using antigen slides containing E. chaffeensis Arkansas strain in DH82 cells (7). Samples that were reactive by IFA at a 1:80 dilution were titrated to endpoint. The acute-phase serum mates of the convalescent-phase samples that were reactive by IFA were also screened and titrated to endpoint if reactive. Convalescent-phase serum samples that were reactive with E. chaffeensis by IFA were also tested for IgG antibodies to Anaplasma phagocytophilum, Rickettsia rickettsii, and Rickettsia typhi. A. phagocytophilum IFA was performed using slides coated with the Webster strain grown in HL-60 cells. IFAs for R. rickettsii and R. typhi were conducted using commercially available slides (Focus Diagnostics, Cypress, CA). All reactive samples were titrated to endpoint, and paired acute-phase serum samples of any reactive convalescent-phase sample were tested similarly. If the convalescent-phase sample was reactive with A. phagocytophilum, the acute-phase sample was also screened and titrated to endpoint.

RESULTS

Ehrlichia chaffeensis peptide ELISA cutoff optimization.

ROC analyses were performed with the results of the single and combined peptide assays (Fig. 1). With the combined TRP120/TRP32 peptide assay, sensitivity ranged from 54% to 77% and was maximized using a cutoff of 2.4 SDs above the mean of the negative controls; specificity ranged from 81% to 92% and was maximized at a cutoff 6 SDs above the mean of the negative controls. Using the TRP120 peptide only, sensitivity ranged from 88% to 100% and was maximized at a cutoff of 5 SDs above the mean of the negative controls, whereas specificity ranged from 71% to 90% and was maximized at a cutoff of 7 SDs above the mean of the negative controls. Since neither the combined TRP120/TRP32 peptide ELISA nor the single TRP120 assay offered an excellent combination of sensitivity and specificity (prespecified as 90% sensitivity and specificity), we opted to develop the peptide assay as an unbiased high-throughput screening tool prior to second-stage IFA confirmatory testing. To do so, we used the more sensitive TRP120 peptide ELISA at a cutoff of 5 standard deviations above the mean of the negative controls where the optimum sensitivity (100%; 95% confidence interval [CI], 77% to 100%) was found with the highest specificity (84%; 95% confidence interval, 77% to 92%) (Table 1).

FIG 1.

Receiver operator characteristic (ROC) curves for E. chaffeensis single TRP120 and combined TRP120/TRP32 peptide ELISAs using a set of validation serum samples from patients who recovered from E. chaffeensis, A. phagocytophilum, B. burgdorferi, and R. rickettsii infections and from healthy subjects. The cutoffs between positive and negative are based on various standard deviations above the mean optical density of E. chaffeensis seronegative samples.

TABLE 1.

Sensitivity and specificity of Ehrlichia chaffeensis TRP120 peptide ELISA at various possible cutoffs

ELISA	Result	Cutoff (SDs above mean of negative controls)
		2	2.2	2.4	2.6	2.8	3	4	5	6	7
E. chaffeensis	No. true positives	16	16	16	16	16	16	16	16	15	14
	No. false negatives	0	0	0	0	0	0	0	0	1	2
	Sensitivity	1.00	1.00	1.00	1.00	1.00	1.00	1.00	1.00	0.94	0.88
Overall specificity	No. true negatives	64	65	66	67	68	69	73	76	77	81
	Specificity	0.71	0.72	0.73	0.74	0.76	0.77	0.81	0.84	0.86	0.90
A. phagocytophilum specificity	No. false positives	8	8	8	8	7	7	7	7	7	6
	No. true negatives	19	19	19	19	20	20	20	20	20	21
	Specificity	0.70	0.70	0.70	0.70	0.74	0.74	0.74	0.74	0.74	0.78
Other serum sample specificity	No. false positives	18	17	16	15	14	14	10	7	6	3
	No. true negatives	45	46	47	48	49	49	53	56	57	60
	Specificity	0.71	0.73	0.75	0.76	0.76	0.78	0.84	0.89	0.90	0.95

Out of the 14 serum samples that were falsely positive at this cutoff, 7 were known to have antibodies against A. phagocytophilum, 2 had antibodies against B. burgdorferi, and 5 were from healthy subjects. Of those A. phagocytophilum samples with E. chaffeensis TRP120 reactivity, 2 had documented E. chaffeensis antibodies by IFA (160 and 640), and the remaining 5 samples lacked E. chaffeensis antibodies by IFA. The B. burgdorferi-reactive samples were not tested for E. chaffeensis (10) or A. phagocytophilum antibodies, and all samples from healthy subjects lacked E. chaffeensis and A. phagocytophilum antibodies. The specificity among non-Anaplasma controls was 89% (95% confidence interval, 81% to 97%). Compared with the true E. chaffeensis patient serum sample peptide ELISA results, the mean normalized net OD for false-positive samples (0.14 ± 0.21 SD) was lower than for true E. chaffeensis samples (0.50 ± 0.59 SD; P = 0.03), but this statistical difference was not observed when false-positive samples from A. phagocytophilum patients (mean, 0.21 ± 0.27 SD; P = 0.22) were compared with those of the true E. chaffeensis group.

Clinical application of peptide ELISA for testing a large cohort.

To document and quantify E. chaffeensis as a cause of undifferentiated fever in Nicaragua, 748 convalescent-phase samples from Nicaraguan patients with acute febrile illness (9) were tested for the presence of E. chaffeensis antibodies by peptide ELISA. Of these, 120 (16%) possible E. chaffeensis infections (positive by peptide ELISA) were identified. Of these, 6 (5%) convalescent-phase serum samples showed reactivity with E. chaffeensis by IFA. Testing these same 6 samples by IFAs for A. phagocytophilum, R. rickettsii, and R. typhi confirmed infection with E. chaffeensis alone. One patient had evidence of acute E. chaffeensis infection, as demonstrated by seroconversion by IFA with a 4-fold increase in titer between paired samples (acute-phase titer of <80 with convalescent-phase titer of 320 and A. phagocytophilum IFA titers of <40). Serologic tests for R. rickettsii and R. typhi were also negative. In 5 patients, titers were positive and stable (≤2-fold titer change between paired acute- and convalescent-phase serum samples), likely indicating previous infection with E. chaffeensis. One patient was seropositive for E. chaffeensis and A. phagocytophilum infections, although insufficient differences in IFA titers (stable E. chaffeensis titers of 320 versus stable A. phagocytophilum titers of 640) precluded a definitive serological diagnosis. Among 828 convalescent-phase serum samples that were nonreactive in the E. chaffeensis peptide ELISA, 107 were randomly selected for E. chaffeensis IgG IFA. Of these, only 1 was positive (categorical agreement, 99.1%; 95% confidence interval, 95% to 100%); the E. chaffeensis IgG titer by IFA for the acute- and convalescent-phase samples was 160, and the two samples were negative by A. phagocytophilum IFA. Screening with ELISA and confirming all positive samples with IFA for ehrlichioses and related infections in the Nicaraguan febrile cohort confirmed a specificity for the ELISA that was nearly identical to that obtained during the cutoff optimization study but with tighter confidence intervals (CI) because of the size of the cohort (specificity of 84% [95% CI, 81% to 86%] versus specificity of 84% [95% CI, 77% to 92%]). Our data suggest that the seroprevalence of E. chaffeensis in this population is 0.9% (95% CI, 0.4 to 1.9) but may be 1.1% (95% CI, 0.5 to 2.1) if two E. chaffeensis infections were missed (if 6 of 8 detected since the lower bound of the 95% CI for sensitivity of the ELISA in the optimization study was 77%).

Acute E. chaffeensis infection in a Nicaraguan patient with AFI.

The sole patient with acute E. chaffeensis infection confirmed by a >4-fold increase in E. chaffeensis IgG titer by IFA was a 4-year-old male who presented with fevers, chills, throat pain, productive cough, a 5-pound weight loss, and a 1-day history of abdominal pain. Rash, myalgias, arthralgias, and headaches were not reported, nor was there a history of a tick bite. He had been given antibiotics prior to being seen, but he had not been given doxycycline since a rickettsial infection was not suspected. He was febrile (39.5°C), had a heart rate of 120, and a blood pressure of 80/60 mm Hg. Physical examination did not reveal conjunctivitis, eschar, rash, or hepatosplenomegaly, but erythema and edema with mild induration were identified in the skin overlying the left lower trunk. Laboratory studies were unremarkable, notably lacking thrombocytopenia, leukopenia, or elevated serum aspartate transaminase/alanine aminotransferase (AST/ALT). The patient was treated with dicloxacillin for suspected cellulitis of the left trunk and discharged. On a follow-up visit 2 to 4 weeks later, he had fully recovered. PCR targeting E. chaffeensis dsbA in acute-phase blood was negative.

DISCUSSION

Ehrlichia chaffeensis is a Gram-negative obligate intracellular bacterium of the order Rickettsiales that is transmitted to humans most commonly by the bite of a lone star tick (Amblyomma americanum) (3). Signs and symptoms, which include fever, headache, myalgias, and rash, can be mild or severe, and untreated infections may result in death (5, 11, 12). Ehrlichia chaffeensis is well documented in the United States, and a literature review of epidemiological studies conducted in Central and South America suggests the existence of E. chaffeensis in Mexico, Costa Rica, Venezuela, Brazil, Peru, Chile, and Argentina (4, 13–21).

To our knowledge, E. chaffeensis infection in humans in Nicaragua has not been documented and studies to determine the prevalence of this infection have not been done. In this study, we document the presence of E. chaffeensis or a serologically similar species as an uncommon cause of undifferentiated fever in humans in Nicaragua.

Identifying E. chaffeensis infection as an uncommon cause of undifferentiated fever in a new locale, namely, Nicaragua, required the testing of a large cohort of patients from whom paired serum samples were available, a remarkable 90% in our large study cohort. Because IFA is inherently subjective and laborious, even for the most skilled reader, the presence of ehrlichioses in humans in Nicaragua might have been overlooked if we had not rigorously evaluated and employed a 2-step algorithm that allowed us to avert IFA in 620 patients for whom the peptide-based ELISA screening assay was negative. It is also important to recognize that cross-reactions with related species (e.g., Anaplasma), although not predicted, are similarly observed with IFA.

We found that peptide ELISA can be employed as a screening tool to identify E. chaffeensis infections in new geographic areas in conjunction with confirmatory IFA, the gold standard for diagnosis of E. chaffeensis (5, 6). Cross-reactions were observed with A. phagocytophilum using the ELISA, as they also occur with IFA; hence, confirmatory testing of ELISA-positive samples should include IFA testing for E. chaffeensis and A. phagocytophilum to compare relative titers. Diagnosis of E. chaffeensis infection with a two-step method has some disadvantages. Due to the low specificity of the peptide ELISA screening stage, confirmation is still reliant on IFA, a technique that requires extensive training to read and interpret test results and is dependent on the availability of a fluorescence microscope, requirements which are impractical in resource-limited regions (22). However, screening with peptide ELISA vastly reduces the workload of IFA. The advantages of the peptide ELISA include high sensitivity, simplicity, ease of use, ability to easily synthesize the peptide antigens, and suitability for use in small laboratories with limited facilities. The most significant advantage, perhaps, is the potential for high-throughput analysis of patient samples and the potential to conduct serosurveillance studies on a large scale that could foster a better understanding of the role of E. chaffeensis in the burden of acute febrile illness globally. We propose that the peptide-based ELISA described herein might be similarly applied by other researchers and reference or public health laboratories to better define the epidemiology and clinical features of emerging and reemerging ehrlichioses worldwide, which similarly require high-throughput diagnostic tools. Prospective evaluation of the assay in a region in which ehrlichioses are more common can confirm the precise diagnostic sensitivity of our 2-step approach.

The peptide ELISA in this study made use of synthetic peptides from two immunoreactive E. chaffeensis proteins: TRP120, a 120-kDa tandem-repeat protein, and TRP32, a variable-length PCR target protein (7, 23). The two proteins are known to play important roles in ehrlichial pathogenesis, and antibodies against linear epitopes in both are protective against E. chaffeensis infection (24, 25). Prior studies identified TRP120 and TRP32 as the two most sensitive peptides for serodiagnosis (7). We combined equal amounts of synthetic TRP120 and TRP32, but found that this decreased sensitivity with a modest increase in specificity versus TRP120 alone; this led us to use the single assay for testing our large febrile cohort. However, it is possible that other combinations/ratios of these two peptides can better maximize sensitivity and specificity as indicated by the statistically significant differences in net OD between E. chaffeensis true- and false-positive samples observed here. In fact, different synthetic peptides representing other ehrlichial proteins (e.g., TRP47) may be useful in combination with TRP120 to maximize sensitivity and specificity (24, 25).

With this study, we were able to confirm the presence of E. chaffeensis or an antigenically similar agent in Nicaragua and to suggest it as an unrecognized cause of acute febrile illness (1 acute infection with point prevalence of convalescent IgG antibody of at least 0.94% [7/748]). Although our data suggest that ehrlichiosis is not a common cause of fever in this region, it is also possible that these data are underestimates. First, we may have missed a small number of E. chaffeensis infections since the 95% CI for sensitivity for the cutoff optimization study was 77% to 100% and we detected one additional E. chaffeensis infection among 107 randomly selected E. chaffeensis peptide ELISA-negative convalescent-phase specimens. Additionally, the median follow-up for convalescent-phase serum was 15 days, yet 10% of patients may seroconvert at ≥30 days after the onset of symptoms (9).

As in many underdeveloped tropical regions, knowledge about the etiologic agents of acute febrile illness in Nicaragua is limited. This study documents the existence of human E. chaffeensis infection in Nicaragua, a tick-borne illness not treated by commonly used antimicrobial agents. If confirmed, these data may encourage studies that lead to a wider recognition of the clinical and epidemiologic importance of ehrlichiosis, its geographic distribution and vectors, and appropriate clinical and public health measures.

Funding Statement

Patient enrollment was supported by the Hubert-Yeargan Center for Global Health. M.E.R. was supported by a Johns Hopkins Center for Global Health Junior Faculty Grant, a Clinician Scientist Career Development Award from Johns Hopkins School of Medicine, and the National Institute of Allergy and Infectious Diseases, National Institutes of Health (K23AIO83931). Testing for E. chaffeensis by ELISA and additional laboratory support were provided through NIH grants R01 AI044102-09 and R21 AI09606 to J.S.D. I.C. was supported through the Department of Pathology, University of Maryland School of Medicine. The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.