INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion Appendix References

Lyme disease is a potentially multisystem inflammatory disease caused by Borrelia burgdorferi (56). In the absence of erythema migrans (EM), no symptoms and signs are uniquely diagnostic of Lyme disease. Culturing B. burgdorferi, finding its antigens by immunohistochemistry, or identifying its DNA by PCR in biopsies are all problematic, so indirect markers are used to corroborate infection, usually measurement of antibodies by enzyme-linked immunosorbent assay (ELISA) and immunoblotting. Serum immunoglobulin M (IgM) anti-B. burgdorferi antibodies are often detected in early disease; by 6 to 8 weeks IgG is detected in the majority of untreated patients. Criteria for immunoblot interpretation are widely accepted (11, 16, 17), but testing is not standardized (3, 14, 26, 35, 50) and may be falsely positive due to IgM rheumatoid factor (5, 34) in other diseases (30, 37, 39, 60) and in otherwise healthy individuals (12). Clinical features of Lyme disease can develop even before the elaboration of the humoral immune response to B. burgdorferi. Early in the evolution of the humoral response, seronegativity might be due to the fact that specific antibodies to B. burgdorferi antigens are bound up in circulating immune complexes (a period of "antigen excess"), rendering the antibodies immeasurable by standard techniques (8, 48). Elevated levels of circulating immune complexes were one of the earliest described immunologic phenomena in Lyme disease (24, 25).

Overuse of testing (33, 44, 51) contributes to the misdiagnosis of Lyme disease (54, 55, 58; L. H. Sigal, Editorial, J. Infect. Dis. 171:423-424, 1995), based on the common mistaken belief that a "positive test" is synonymous with "active infection." Persisting seropositivity may be incorrectly interpreted as ongoing infection. Seropositivity is difficult to interpret in patients with posttreatment residual or new symptoms (9, 54, 55), in whom persisting infection is a concern (1, 2, 27, 47, 55, 62). The frequency of false-positive (FP) ELISA results dictates a two-tier strategy (immunoblot confirmation of positive or equivocal ELISA [2-4]).

Other assays include indirect immunofluorescence (IFA) (36, 43), borreliacidal activity (10), and PCR (45, 58). Identification of specific immunoreactivity at the site of inflammation (e.g., antibodies in synovial or cerebrospinal fluid compared to serum) is useful but often cumbersome in identifying local infection (53, 57, 59). A simple assay is needed that can reproducibly confirm early and/or active B. burgdorferi infection. Attempts to improve ELISA have included antibody capture (6, 23, 32); new antigenic preparations (20, 38), including flagellin enhancement (22, 29), recombinant proteins (18, 40), or individual epitopes (28, 31, 61); and testing of antibodies contained within polyethylene glycol (PEG) precipitates of serum (13, 48, 49). Our previous studies were the first to find both B. burgdorferi antigens and IgM antibodies to these antigens within disrupted PEG precipitates from the sera of patients with Lyme disease (8). We were able to prove that the PEG precipitates contained immune complexes (ICs) by purifying the antigens only through their firm binding to serum antibodies (8). We used several of these improvements to develop an assay that, in the serum banks tested, was sufficiently sensitive to confirm early infection, sufficiently specific to obviate immunoblot confirmation, and capable of differentiating active infection from persisting seropositivity in patients with successfully treated disease. In our studies the enzyme-linked, IgM capture, IC, biotinylated antigen assay (EMIBA) fulfills these expectations.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion Appendix References

Growth and preparation of B. burgdorferi sonicate. High-passage B. burgdorferi strain B-31 (54) was grown in Barbour-Stoenner-Kelly medium (4, 46) (made or purchased from Sigma, St. Louis, Mo.) supplemented with 6% normal rabbit serum (Gemini Bio-Products, Calabasas, Calif.), grown in T flasks (Corning Glass, Corning, N.Y.) at 32°C. Four hundred milliliters of late-log-phase culture was harvested by centrifugation (RC5C; Sorvall-DuPont, Wilmington, Del.) at 9,000 × g for 15 min and was washed three times with cold phosphate-buffered saline (PBS), pH 7.2. The final pellet was resuspended in 2 ml of PBS and sonicated (Braun-Sonic 2000) medium setting for four 30 pulses with a 1-min rest between pulses. Approximately 9 mg of protein/ml was obtained, assayed with bicinchoninic acid protein (Pierce, Rockford, Ill.) and stored at 70°C until use. In certain studies a low-passage N40 (provided by Stephen Barthold, Yale University) was propagated and processed as described above (data not shown). The results shown in the paper utilize the B-31 strain, since this is the organism used in the commercial ELISA and immunoblot kits used for comparison. In side-by-side studies, results gained by using N40 and B-31 were identical.

Biotinylation of sonicate. Different long-arm biotin hydroxysuccinamide esters, including biotinamidocaproate N-hydroxysuccinimide (NHS) ester (Sigma) and NHS-LC-biotin II (Pierce) were equally effective; studies described in this paper used biotinamidocaproate NHS. One milliliter of a 9-mg/ml protein sonicate solution was adjusted to pH 9 by adding 0.1 ml of 0.5 M carbonate-bicarbonate buffer, pH 9, immediately before biotinylation. Optimal biotinylation was achieved using 50 mg of biotin ester/ml in dimethylformamide. To the pH-adjusted protein sonicate, 20 µl of biotin solution was added in a 16-by 100-mm glass tube, covered, and slowly rotated for 1 to 2 h at room temperature, pipetting up and down every 15 min. The reaction was stopped by adding 0.1 ml of 1 M Tris-HCl, pH 7.5. Biotinylated sonicate (Bb-bio) was dialyzed against 3 2-liter changes of PBS, pH 7.2, containing 0.02% sodium azide in the cold using a Slide-A-Lyzer (Pierce) with a 10-kDa cutoff. The resulting suspension was assayed for protein bicinchoninic acid, aliquoted, and frozen at 70°C with typical yields of at least 75%.

Serum and plasma collection. For serum, blood was drawn and allowed to clot at room temperature for 1 h; for plasma, heparinized blood was left at room temperature for 1 h. Both types of specimens were then centrifuged in a Sorvall RC5C HS-4 rotor at 769 × g for 10 min. Clear serum or plasma was drawn off by pipette. After serum and plasma samples were utilized in the serologic tests under study, the samples were stored at 70°C until needed for possible further testing. Freeze-thaw slightly decreased the reactivity of some samples but did not reduce any positive samples into the negative range.

IC precipitation. The PEG method was used to isolate ICs (15, 49). Samples (100 µl) were placed in a microfuge tube (Eppendorf) and precipitated with an equal volume of a 7% PEG (average molecular weight, 8,000; Sigma) and 0.44% NaCl in 0.1 M sodium borate buffer, pH 8.4. Tubes were vortexed, left at 4°C for at least 4 h (15), and centrifuged at 10,000 rpm (8,320 × g) for 15 min in the cold. Supernatants were carefully removed with a pipette. The pellet was resuspended and washed twice with 200 µl of 3.5% PEG solution in the same buffer. After the second spin, samples were resuspended in 100 µl of 0.1 M sodium borate buffer, pH 10.2; high pH buffer is better than PBS for dissociating ICs and does not affect antibody stability (P. Coyle, M. Brunner, and S. Schutzer, unpublished data). Dissociated ICs were kept in buffer at 4°C until use. There was no discernible loss of reactivity after 2 weeks of storage (M. Brunner, unpublished data).

EMIBA. One hundred microliters of affinity-purified goat anti-human IgM (mu chain specific) (KPL; Gaithersburg, Md.) per well at 10 µg/ml in 0.04 M-0.0357 M carbonate-bicarbonate coating buffer, pH 9.6, was added to Immulon 4 (Dynatech) microtiter plates. The plates were rotated slowly at room temperature for 2 h and were stored covered at 4°C overnight. The plates were warmed to room temperature and were washed three times with 10 mM PBS, pH 7.5, containing 0.1% bovine serum albumin (Sigma) and 0.05% Tween 20 (PBS-BT) using an automated plate washer (Bio-Tek ELP 35). After the final wash, 0.35 ml of blocking buffer (PBS-BT containing 5% nonfat dry milk)/well was added, and the plates were covered with Mylar. Their contents were incubated for 1 h at 37°C and were washed twice with PBS-BT. One hundred microliters of dissociated immune complexes or serum per well at 1:100 dilution in PBS-BT with 3% fish skin gelatin (Sigma) and 1% heat-inactivated normal goat serum (Vector Labs, Burlingame, Calif.) were added in duplicate. Plate contents were incubated for 2 h and washed three times with PBS-BT, and biotinylated B. burgdorferi was added. The plate was covered and rotated slowly for 1/2 h at room temperature.

Statistical analysis. Based on active or prior status of Lyme disease at the time of phlebotomy, test results were designated true positive (TPhad active LD and a positive test result), true negative (TNdid not have active LD had a negative test result), FP (did not have active LD but had a positive test result), or false negative (FNhad active LD but had a negative test result). Sensitivity and specificity were calculated as follows:
Confidence intervals of 95% for sensitivity and specificity were calculated using the Fleiss correction (19). The significance of differences in the sensitivity and specificity of the different assays tested were assessed by McNemar's chi-square test (52). Also provided are 95% confidence intervals of differences between the sensitivities and specificities of the different assays. As collections were stratified into active and prior groups, the number of samples within some cells was too small to calculate confidence intervals.

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion Appendix References

Comparison of EMIBA with standard serologic assays in patients with Lyme disease and non-Lyme disease controls. At the time of evaluation and phlebotomy, 64 of the 131 patients from the Lyme Disease Center were clinically designated as having active disease (diagnosed with Lyme disease but receiving no prior antibiotic treatment), 28 had prior Lyme disease (had previously undergone antibiotic treatment, with no evidence of active Lyme disease at the time of phlebotomy), and 39 had never had Lyme disease. The sensitivity and specificity of each assay were calculated (Table 1). We then determined if the sensitivity and specificity of the EMIBA and the free-antibody assay were superior to the sensitivity and specificity of each of the comparator assays and if these differences were statistically significant (Table 2).

EMIBA detection of serologic reactivity in EM (early Lyme disease) compared with serodetection utilizing standard assays. Twenty-seven of 131 specimens from the Lyme Disease Center were from patients with EM, 24 before treatment. The other three had been successfully antibiotically treated 3 to 9 months before phlebotomy; none had evidence of infection at the time of phlebotomy. These three constitute the TN and FP groups in Table 4 (see Appendix). EMIBA and the free-antibody assay were more sensitive than comparison assays and more specific than all but the IgG ELISA, to which they were equivalent. Thus, EMIBA and the free-antibody assay were superior to other assays in corroborating early Lyme disease.

EMIBA is able to differentiate between seropositivity in patients with ongoing infection and persisting seropositivity related to past infection (active versus prior disease). Of 131 Lyme Disease Center patients, 64 had active infection and 28 had previous Lyme disease without evidence of active disease at the time of phlebotomy, i.e., previously cured Lyme disease (Table 6). Table 6 presents serologic results in the comparator assays with reference to results in EMIBA and the free-antibody assay: (i) free-antibody assay (+), EMIBA (); (ii) free-antibody assay (), EMIBA (+); (iii) free-antibody assay (+), EMIBA (+); and (iv) free-antibody assay (), EMIBA ().

EMIBA results in control subjects: multiple sclerosis, syphilis, and rheumatic diseases. One (8%) of the 12 multiple sclerosis patients' sera tested was positive for both free antibodies and IC by EMIBA; the sample was negative by IgM immunoblotting.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion Appendix References

Although the diagnosis of Lyme disease should be based solely on appropriate historical and objective clinical findings, serologic evidence favoring the diagnosis may be useful or necessary in some circumstances. In the presence of EM, serologic confirmation is not necessary. However, lacking an EM lesion, an isolated virus-like syndrome following a tick bite cannot be attributed to early B. burgdorferi infection without laboratory confirmation. The early features of Lyme disease may occur too soon after tick bite for specific humoral immune responses to be measurable by current immunoassays. Only with highly sensitive and specific serologic or microbiological tests can unusual or atypical clinical features of B. burgdorferi infection be correctly included within the spectrum of Lyme disease and can illusory associations be identified as such and excluded.

Serologic tests are commonly misused as if they were "diagnostic." We favor the term "anti-B. burgdorferi antibody test" as preferable to the more euphonious but also more prejudicial term "Lyme disease test." Even in circumstances where the clinical likelihood of Lyme disease is very high, this can, at best, be a "seroconfirmatory," not a "serodiagnostic" test. Many serologic assays are currently available to detect anti-B. burgdorferi antibodies, but all share two limitations: (i) in early disease the tests may not detect low levels of specific antibody, and (ii) none differentiates persistent seropositivity in active infection from clinically irrelevant persisting antibodies, i.e., seropositivity without evidence of active infection. We sought to develop an assay capable of satisfying both needs. We started with an IgM capture format (6, 23, 32), which diminishes high background levels and possible confounding serum components, e.g., competing IgG. We knew that the sera of many Lyme disease patients contained IgM antibodies within materials obtained by PEG precipitation, thought to be IC, and that the IgM bound within the precipitate might be undetectable in assays using unprocessed serum (13, 48, 49). Use of the IgM obtained from disrupted PEG precipitates enhanced the sensitivity of immunoblot in Lyme disease (49); these studies all used PEG precipitation, an established method of purifying IC (15), at a 1:10 dilution. Our previous studies demonstrated that this procedure did not merely concentrate serum IgM in a nonspecific fashion (8). These were the first studies to prove that the PEG precipitate actually satisfied the definition of an IC: the PEG precipitates contained antibodies specific for the B. burgdorferi antigen OspA, and the OspA protein was isolated from the putative IC only by its interaction with antibody within the PEG precipitate (8). The rigorous method used to prepare the material subjected to Western blot analysis assured that the OspA was within IC and not nonspecifically PEG precipitated in a high-molecular-weight complex other than a specific IC. In addition PEG precipitates contained IgM against B. burgdorferi proteins, in some cases antigens not recognized by the free antibodies in untreated serum (8), in agreement with previous, small studies (48, 49).

The present studies represent the largest and broadest application of IC technology to the seroconfirmation of Lyme disease and the only comparison of IC- with free-antibody-based assays and currently available immunoassays (IFA, ELISA, flagellin-enhanced ELISA, and Western blotting). Previous studies have shown that anti-B. burgdorferi IgM may persist in the sera of patients with Lyme disease after treatment and apparent cure (1, 27). In order to detect even very low IgM levels in later disease, we made use of the increased sensitivity intrinsic to EMIBA; we were able to further enhance sensitivity by biotinylating the sonicate, giving the same amplification benefits as in biotin-enhanced immunoblotting (49).

Lyme disease sera in this study were from three sources, and in all cases the testing was done blinded to clinical information. The first group of sera was from patients evaluated for Lyme disease at our regional referral center. This population is the most relevant to a serologic trialpatients from an area of endemicity whose symptoms and signs prompted the patients and/or their physicians to consider Lyme disease. The second group was from the CDC collection, an established serum bank used in previous serologic studies (29). The third group, from the Marshfield Clinic, represents an incontrovertible "gold standard" for early Lyme disease-culture-positive EM patients, with healthy controls interspersed.

Studies using all three serum banks showed that EMIBA can serologically confirm early Lyme disease. EMIBA and, to a lesser degree, the free-antibody assay were superior to the comparator assays. In the Lyme Disease Center sera, EMIBA was better able to differentiate between persisting seropositivity indicating active infection (thereby warranting antibiotic therapy) and seropositivity of no clinical significance, relating to prior cured infection; as noted, seropositivity can persist long after clinical cure (1). EMIBA was somewhat less effective in the CDC sera. This was most apparent in all the assays' ability to differentiate active from prior disease. We were able to assign disease status in only 38 CDC samples; clinical information was less detailed than available for the Lyme Disease Center population, especially concerning the timing of previous treatment and time between treatment and phlebotomy. Differences in EMIBA's correlation with disease status may relate to improper assignment of patient samples to active or prior groups based on incomplete information and/or the fact that a short time had elapsed from successful treatment to phlebotomy. In the latter case, FP EMIBA results might be due to the persistence of IC following treatment, a phenomenon noted subsequently in unblinded samples tested in our laboratory (M. Brunner and L. H. Sigal, unpublished observations). It is also possible that the smaller size of the CDC bank may have obscured differences in the assay's predictive power.

Sera from patients with syphilis contain antibodies that also bind to B. burgdorferi. In this control group EMIBA was superior to the free-antibody assay, in that fewer of the luetic sera were positive. FP results were present in other assays, as wellsix of the seven EMIBA-positive samples were positive in IgM immunoblotting. A small proportion of multiple sclerosis patients also had antibodies reacting with B. burgdorferi proteins. We are currently exploring ways to improve the assay to minimize FP tests, especially in these groups (7).

Our results suggest that EMIBA is slightly superior to our free-antibody assay, superior to the standard assays in seroconfirmation of Lyme disease, effective in detecting antibodies in early disease, better than standard assays and comparable to IFA in the small group of sera from patients with positive skin biopsy cultures, and helpful in determining the clinical significance of seropositivity. The concentration of anti-B. burgdorferi antibodies within IC has been demonstrated by comparative immunoblotting (8).

Future studies will include sequential samples before, during, and after antibiotic treatment of Lyme disease and sera from patients with other nonluetic spirochetal infections and from larger numbers of subjects with other rheumatologic and neurologic diseases. EMIBA utilizes IgM from IC; assays measuring the IgA and IgG within IC might also be helpful in the seroconfirmation of active infection with B. burgdorferi. Finally, differently grown organisms may provide a more advantageous antigen pool for testing clinically relevant seroreactivity (21). We are currently working on a quantitative method of comparing levels of free antibodies with levels of antibody within IC to improve the clinical predictive value (i.e., active versus previously cured infection) of our assay. If our current observations are borne out, EMIBA may be able to replace the current two-tiered seroconfirmation approach in Lyme disease with a single assay.

Thus, IC-based serologic assays are valuable in seroconfirmation of the diagnosis and clinical status of B. burgdorferi infection and might be useful in other diseases as well. In early disease, a state of antigen excess, the IgM of the humoral response to a new pathogen may be sequestered in IC, bound to circulating pathogen- or transformed cell-derived antigen targets and undetectable by standard serum-based assays. In such circumstances free-antibody levels would be undetectable. The ongoing production of ICs is dependent on and limited by the ongoing production of pathogen- or transformed-cell-derived antigens, so IC-based assays might be useful in determining if persistent seropositivity is a marker of disease persistence following treatment for infectious or malignant diseases. Our results suggest that further exploration of IC-based seroconfirmatory assays is warranted.