Journal of Immunological Methods, 132 (1990) 91-101 Elsevier
91
JIM 05661
A new ELISA for the detection of double-stranded D N A antibodies W o o d r u f f Emlen, Pisamai Jarusiripipat and Gregory Burdick Division of Rheumatology, University of Colorado Health Sciences Center, 4200 E. 9th Avenue, Denver, CO 80262, U.S.A. (Received 21 August 1989, revised received 30 April 1990, accepted 1 May 1990)
The accurate detection of antibodies to double-stranded DNA is important in the diagnosis and management of patients with systemic lupus erythematosus (SLE). We have developed an ELISA in which plasmid dsDNA is biotinylated and bound to streptavidin coated wells. The biotinylated DNA did not lose is antigenicity, and the DNA which bound to the plate remained double-stranded. Only small amounts of DNA were required to coat the plates, and binding was highly reproducible. After defining the normal range of the assay, sera from patients with various conditions were examined; 0 out of 97 controls were positive, 0 out of 36 patients with RA were positive, 0 out of 14 patients with scleroderma, and two out of 18 patients with chronic liver disease were positive. In contrast, 52% of unselected SLE patients were positive, and there was a good correlation between level of DNA antibodies and disease activity. Comparison of this assay with other DNA assays (the Farr, Millipore, filter, and a commercial ELISA) showed a high degree of correlation between this ELISA and each of the othe r assays. Furthermore, screening of randomly selected ANA positive patients with this assay failed to show a high false positive rate as has been reported with other anti-DNA ELISAs. This assay is simple to perform, requires small amounts of DNA to coat the plate, is highly reproducible, and is specific for IgG antibodies to dsDNA. Key words: Systemic lupus erythematosus; Anti-DNA antibody; ELISA
Introduction
The measurement of antibodies to DNA in patients with systemic lupus erythematosus (SLE) has been shown to be clinically useful both in diagnosis of disease and in following the activity
Correspondence to: W. Emlen, Division of Rheumatology, University of Colorado Health Sciences Center, 4200 E. 9th Avenue, Denver, CO 80262, U.S.A. Abbreviations: SLE, systemic lupus erythematosus; dsDNA, double-stranded DNA; ssDNA, single-stranded DNA; PBS, 0.01 M phosphate, 0.15 M NaC1, pH 7.4; PBST, PBS 0.05% Tween.
of disease after diagnosis has been established (Tan et al., 1966; Koffler et al., 1967; Koffler, 1974; Emlen et al., 1986). A number of studies have shown that IgG antibodies to doublestranded DNA (dsDNA) are most specific for the diagnosis of SLE, and clinically most useful in following disease (Rothfield and Stollar, 1967; Pincus et al., 1969; Minter et al., 1979). Antibodies to single-stranded DNA (ssDNA) and IgM DNA antibodies are found in a number of other connective tissue diseases, liver diseases, as well as in some normal individuals (Notman et al., 1975; Locker et al., 1977). A large number of assays to measure DNA antibodies have been developed. Each of these assays has been plagued with the
0022-1759/90/$03.50 © 1990 Elsevier Science Publishers B.V. (Biomedical Division)
92 problem of purity of dsDNA, as well as problems unique to each assay. The Farr assay, in which radiolabelled DNA is bound to antibody and then precipitated with ammonium sulfate (Pincus et al., 1969) remains the standard method of measurement, although this assay detects IgM as well as IgG antibodies to DNA, and the use of radioactivity makes this assay undesirable in some clinical laboratory settings. Problems with single-stranded contamination of DNA preparations have also resulted in decreased specificity of the Farr assay, yielding positive results in a small number of patients with diseases other than SLE (Hasselbacher and LeRoy, 1974; Bell et al., 1975; Notman et al., 1975; Locker et al., 1977; Chubick et al., 1978). In an attempt to develop an assay specific for dsDNA, Aarden and colleagues (1975) developed an immunofluorescence assay in which antibodies bind to the kinetoplast of Crithidia luciliae. This organelle contains circular dsDNA, and by varying the specificity of the fluorescent antibody, this assay can be made IgG or IgM specific. However, this assay is only semiquantitative, and recent data has shown that the Crithidia kinetoplast also contains histones, resulting in false positive reactivity in, patients with anti-histone antibodies (Deng et al., 1985). A number of solid phase assays have been developed to detect DNA antibodies over the past 10 years. Early on it was noted that ssDNA bound well to plastic, but dsDNA bound poorly (Engvall, 1976; Smeenk, 1986). To facilitate dsDNA sticking to plastic, plates have been UV-irradiated (Zouali and Stollar, 1986), or have been precoated with methylated bovine serum albumin (Rubin et al., 1983) or with positively charged molecules such as poly-L-lysine or protamine sulfate which bind negatively charged DNA (Aotsuka et al., 1979; Klotz et al., 1979; Fish and Ziff, 1981; Eaton et al., 1983; Smeenk, 1986). These ELISAs have been effective for measuring DNA antibodies, but antigen binding to the plates has been variable, and a number of false positive results in non-SLE patients have been reported (Fish and Ziff, 1981; Eaton et al., 1983; Smeenk, 1986). These false positive results may occur because ELISA assays detect low avidity antibodies which are missed in the Farr assay, and which may not be clinically relevant in the pathogenesis of SLE
(Smeenk et al., 1982; Emlen et al., 1986; Smeenk, 1986), or because these assays detect DNA-antiD N A immune complexes (Brinkman et al., 1989). False positives may also result from antibodies to ssDNA since most ELISAs reported to date have used heterogeneous DNA (calf thymus DNA), which is likely to contain single-stranded regions which would preferentially bind to plastic (Engvail, 1976). In an attempt to develop a more dsDNAspecific ELISA, we have used plasmid DNA as an antigen source, and bound dsDNA to plastic via the specific streptavidin-biotin interaction. In this way, dsDNA is bound stably to the plastic independent of charge, and only small amounts of D N A are required to coat wells. This assay has proved to be reproducible, and the results correlate well with other standard DNA assays. When used to examine patients with SLE in varying stages of clinical disease activity, this assay was as sensitive and specific as other commercially available assays. Furthermore, when tested on unselected ANA positive patients, the false positive rate for high DNA antibodies in non-SLE patients was no higher with this ELISA than with the Farr assay.
Materials and methods
Preparation of DNA pGEM-3 plasmid (Promega, Madison, WI) was grown in E. coli and plasmid DNA was isolated by cell lysis followed by centrifugation over a cesium chloride gradient. Plasmid DNA was cut at a single site with endonuclease Sma-1 (Bethesda Research Labs, Gaithersburg, MD), yielding a linear, blunt-end dsDNA molecule of 2867 bp; analysis of this D N A on 0.8% agarose gels demonstrated a single homogeneous band. After ethanol precipitation, DNA was brought to a concentration of 1 m g / m l in water, and mixed with photoprobe biotin (Vector Labs, Burlingame, CA) (1 m g / m l ) at varying D N A / p h o t o p r o b e ratios in a constant volume of 100/xl. These mixes were UV irradiated using an H2 filter (295-400 nm) on ice through glass (to exclude UVB) for a total of 278 J / c m 2. This corresponded to 5 rain at a distance of 10 cm with a Dermalight 2001 Sol 3 solar
93 simulator. Following labeling, an equal volume of 0.1 M Tris p H 9.0 was added, and D N A was extracted x 3 with butanol to remove unreacted photoprobe. D N A was ethanol precipitated and resuspended in PBS (0.01 M phosphate, 0.1 M NaC1, pH 7.4) for use. The degree of biotinylation was determined by measuring the 00260 (DNA) and 00473 (biotin). D N A was brought to 10 m M E D T A and stored at 4 ° C as a concentrated solution (300-1000 /~g/ml) and diluted just prior to coating plates. Tritiated p G E M D N A was obtained by transforming E. coli strain W3110 thymidine with p G E M DNA. Tritiated thymidine was added to the cultures and plasmid D N A isolation and biotinylation procedures were carried out as above.
Standard ELISA procedure Preparation of plates. Immulon-2 96 well flatbottom microtiter plates (Dynatech, Alexandria, VA) were coated with a 1 /~g/ml solution of streptavidin (ICN Biochemicals, Cleveland, OH) in PBS at 150 /d/well overnight at 4°C. Wells were then washed 3 X with PBS/0.05% Tween (PBST) and post-coated with 200 /~1 PBST/0.2% gelatin (Baker, Phillipsburg, N J) for 1 h at 37 ° C. Wells were washed 3 x with PBST and 150 /~1 biotinylated DNA, diluted to 400 n g / m l in PBS/0.2% gelatin was added to even numbered columns; D N A diluent alone (PBS/0.2% gelatin) was added to odd numbered columns. Plates were incubated overnight at 4 ° C and washed 3 X with PBST. Plates were ready to use at this stage, or could be dried and used with no loss of binding activity for up to 6 weeks (longest time tested). ELISA assay. Serum was diluted 1/1000 in serdil (PBST/0.1% gelatin/0.5% bovine gammaglobulin (Sigma, St. Louis, MO)) and 150 lal was added to each well and incubated at room temperature for I h. Every serum to be tested was run in both an odd numbered (control) and an even numbered (DNA coated) well, so that each serum could serve as its own control. Following incubation, wells were washed x 3 with PBST, and 150 /~1 of a 1/2000 dilution (diluted in serdil) of peroxidase linked goat a-human IgG (Kirkegaard & Perry, Gaithersburg, MD) was added for 1 h at room temperature. Wells were washed x 5 with PBST, and 150 /~1 of ABTS solution (2,2'-azino-
bis(3-ethylbenzothiazoline-6-sulfonic acid) (Aldrich, Milwaukee, W I ) + H202 in McIlvain's buffer) was added. Absorbance at 405 nm (OD) was read using a Dynatech Micro ELISA Autoreader (MR580) at a fixed time (30 min), or when the OD of a standard serum had reached a given level. Background OD (non-DNA coated well) was subtracted from the OD of the D N A coated well, yielding the net signal OD. Optical density (OD) was converted to international units (IU) by calibrating the assay with a standard DNA antibody source, serum W o / 8 0 , obtained from the World Health Organization in Amsterdam (Feltkamp et al., 1988). On each plate, the following controls were run: serum diluent alone (used to blank the ELISA reader), normal human serum as a negative control, lupus pool (pool of five SLE sera) as a positive control, and a serum containing antibodies only to ssDNA to insure that the DNA on the plate was double-stranded. S1 nuelease digestion. To ascertain that only dsDNA was exposed on the plates, DNA-coated wells were digested for 60 min at 3 7 ° C with increasing amounts of a single-stranded specific $1 nuclease (BRL) in 10 m M ZnSO 4 at p H 4.6 (Vogt, 1980). To demonstrate that $1 nuclease was active against ssDNA, plates were made in which some wells were coated with heat denatured D N A (column 3), while other wells (column 2) were coated with intact dsDNA and column I with diluent alone. The effect of S1 nuclease digestion on each of these wells was tested using sera reactive with ssDNA or dsDNA. Sera. 97 control sera were obtained from normal blood bank donors at the Belle Bonfils Blood Center, Denver, CO. Rheumatoid arthritis (RA) and progressive systemic sclerosis (PSS) sera were obtained from patients fulfilling the ACR criteria for these diseases who were followed at the University of Colorado Health Sciences Center (UCHSC) Arthritis Clinic. Sera from patients with chronic liver disease (CLD) were obtained from the University of Colorado Health Sciences Gastroenterology Unit, and included six patients with chronic autoimmune hepatitis, five patients with primary biliary cirrhosis, and seven patients with alcoholic liver disease. Sera from SLE patients were obtained from three sources: (1) patients with SLE (by ACR criteria) who were seen in the
94 UCHSC Lupus Clinic; these sera were drawn on all SLE patients regardless of disease activity (n = 28); (2) patients with SLE who were seen in the Tulane University Lupus Clinic (gift of Dr. D. Boulware); these sera were drawn regardless of disease activity (n = 5); (3) sera taken from the UCHSC Serum Bank which had been selected for high D N A binding by screening Millipore filter assay (n = 58). Clinical information was available via retrospective chart analysis on 56 of these SLE patients. Disease activity in these 56 patients was determined according to the criteria of the lupus activity criteria count (LACC) (Urowitz et al., 1984), and patients were divided into three groups: active renal = active urinary sediment defined as RBC casts a n d / o r > 5 R B C / h p f ; active non-renal = arthritis/rash/pleuropericarditis/CNS disease/vascuhtis but no active renal disease; and inactive d i s e a s e - - n o manifestations of SLE. Because data was often available at only one point in time, changes in creatinine or proteinuria could not be considered in patient classification. Patients with active renal disease usually had other manifestations of active disease. In addition, sera from 121 consecutive ANA positive patients (titer > 1 : 64) from the UCHSC Clinical Immunology Lab were screened by ELISA and Farr assay for D N A antibody, and clinical information was obtained by retrospective chart review on all patients with increased D N A antibody levels by either assay. The diagnosis of SLE was made if these patients fulfilled four or more ACR criteria for SLE. Commercial DNA antibody assays. Sera were tested for D N A binding by a Farr binding assay using calf thymus D N A (Amersham Corp., Arlington Heights, IL) and an ELISA assay (Sigma). These assays were performed exactly according to manufacturer's protocols and specifications. Some sera were also tested in a Millipore filter assay (Ginsberg and Keiser, 1973) using 125I-DNA (Electronucleonics, MD) as an antigen source. Results
Assay conditions To establish the assay, initial experiments were performed to determine optimal coating and
post-coating conditions at each step. For these experiments a standard pool of lupus sera, derived from patients with known active SLE, was used as an antibody source (lupus pool). Initial coating of the plates with high concentrations of streptavidin (1 m g / m l ) resulted in a poor signal; therefore a concentration of 1 g g / m l was arbitrarily chosen. To determine the optimal conditions of DNA coating, D N A preparations were biotinylated to varying degrees, and wells were coated with 150 #1 of increasing concentrations of each of these preparations. A constant amount of antibody was added and the resultant signal (OD) was plotted as a function of D N A added to the wells. All biotinylated D N A preparations showed increasing signal with increased D N A coating; however, more heavily biotinylated D N A preparations gave a lower signal, particularly at high D N A concentrations. Less biotinylated D N A gave an adequate signal, but the signal had not reached a maximum at concentrations in excess of 1000 n g / m l (data not shown). Based on these data, we chose standard conditions of biotinylation corresponding to one biotin per 30 bp of D N A and a standard D N A concentration of 400 n g / m i for coating plates. This concentration of D N A was on the plateau portion of the curve, such that minor errors in dilution did not significantly alter results. Because of the possibility that some sera might contain small amounts of antibody reactive with streptavidin or gelatin in the post-coat, ELISA plates were set up so that each sample could act as its own control. Odd numbered columns were coated with streptavidin and post-coat only, while even numbered columns were coated with streptavidin, post-coat, and biotinylated DNA. Each serum sample was run both in the D N A coated well and in a control well, and background absorbance was subtracted from the signal for each serum. A 1/1000 dilution of serum was arbitrarily chosen for screening sera because it gave an good net signal (signal minus background) with background O D values (non-DNA coated well) less than 0.1. With increasing amounts of serum, net signal increased, but background OD values also increased. Dilution of second antibody and time of reading were chosen arbitrarily. We next examined the effects of different DNA substrates and serum dilution on the signal gener-
95
2.0 1.5 v
.Q
25 IU (normal mean+ 3SD). Sera included are 91 SLE patients, 41 normal controls, 16 patients with CLD, and 121 consecutiveunselected ANA positivepatients. Farr Positive Negative
ELISA Positive 71 12 83
Negative 26 160 188
97 172
called negative, and the ELISA was positive for 12 sera that the Farr assay called negative. Review of clinical data of the patients in whom the assays were discordant showed that 15 of 26 Farr positive-ELISA negative patients had SLE; and eight of 12 ELISA positive-Farr negative patients had SLE. Of the total of 97 Farr positive patients, 12 did not have SLE (12.4% false positives); of 83 ELISA positive patients, five did not have SLE (6% false positives). The ELISA was then compared with a commercial ELISA (Sigma) and a standard Millipore filter assay using sera from SLE patients and from normal patients. The ELISA correlated well with both assays (Sigma, n = 74, r = 0.74, p < 0.001; Millipore filter n = 192, r = 0.58, P < 0.001) (data not shown). Using a 2 × 2 table of positive and negative results, the ELISA agreed with the Sigma assay in 74% of the samples, and with the Millipore filter assay in 80% of the samples tested.
Discussion We have developed an ELISA for the detection of lgG dsDNA antibodies which takes advantage of the strong interaction between streptavidin and biotinylated DNA. By pre-coating wells with streptavidin, we were able to efficiently and reproducibly bind dsDNA to the wells. By using plasmid DNA, which is entirely double-stranded, and by blocking the wells prior to the addition of DNA, we have minimized the possibility of nonspecific sticking of D N A which might contain
single-stranded regions. Performance of this assay required only very small amounts of serum (1/1000 dilution of serum routinely), was highly reproducible, and showed a low frequency of false positive results. In order to justify the use of biotinylated DNA as an antigen in this assay, it was necessary to insure that biotinylated D N A was not significantly altered antigenically. We therefore tested 3H-plasmid D N A in a Farr assay, and compared it with the behavior of biotinylated 3H-plasmid D N A in the same assay. Our results (Fig. 2) show that biotinylation of D N A to the degree of one biotin per 30 bp pairs had no effect on antigenicity of the DNA. Higher degrees of biotinylation showed a decreased signal on the ELISA assay, suggesting some possible loss of antigenicity with higher degrees of biotinylation. Although these experiments certainly do not rule out small changes in the antigenicity of the biotinylated DNA, the similarity of the Farr binding curves strongly suggests that binding of antibodies to D N A is not significantly altered by D N A biotinylation, The use of the streptavidin-biotin interaction to bind D N A to the wells has several advantages over other methods. Previously, positively charged molecules such as poly-L-lysine or protamine have been used to facilitate dsDNA binding (Aotsuka et al., 1979; Klotz et al., 1979; Fish and Ziff, 1981; Eaton et al., 1983; Smeenk, 1986). Since these molecules are positively charged and may mimic histones, antibodies reactive to histone or D N A anti-DNA immune complexes (Brinkman et al., 1989) may bind to the positively charged molecules on the plate. The p l of streptavidin is 6.2, such that the binding of dsDNA to the plates in our system is not charge dependent. Furthermore, the biotin-streptavidin interaction is extremely stable, making binding very efficient and reproducible. Indeed, significantly lower amounts of D N A were required to coat the wells in our ELISA than in many other systems (Aotsuka et al., 1979; Klotz et al., 1979; Fish and Ziff, 1981; Eaton et al., 1983; Rubin et al., 1983; Smeenk, 1986). Since it is known that ssDNA sticks non-specifically to plates, the possibility must be raised that in other ELISA systems, small amounts of contaminating ssDNA, which would preferentially bind to the well (Engvall, 1976), might make up a significant
99 fraction of the total D N A bound. In the present ELISA, plates were blocked after the coating of streptavidin and before the addition of DNA, to minimize any non-specific sticking of D N A which was not dependent upon the streptavidin-biotin interaction. Control experiments (Fig. 1) showed that non-biotinylated D N A gave no signal in the ELISA, confirming that non-specific sticking of D N A had not occurred. Furthermore, no signal in the ELISA was generated when biotinylated D N A or non-biotinylated D N A was coated directly onto plates from which the streptavidin had been omitted (Fig. 1). Thus, the use of the streptavidinbiotin linkage system insures that the D N A which binds to the plate accurately represents the D N A which is put into the wells. Analysis of the D N A used to coat wells on agarose gels showed a homogeneous band 2867 bp in size. To insure that the D N A on the plate remained double-stranded, DNA-coated wells were digested with S1 nuclease to remove any ssDNA or D N A containing single-stranded regions. Under conditions that completely removed ssDNA, there was no diminution of antibody binding to dsDNA (Fig. 3), indicating that no antigen had been removed from the plate with S1 nuclease treatment. Furthermore, testing of sera and a monoclonal antibody with ssDNA specificity showed a > 10 : 1 ss/ds signal ratio. These data strongly support the conclusions that biotinylated plasmid D N A is adsorbed onto the streptavidin intact, and remains completely double-stranded. Another important feature of this ELISA is the use of a control well coated with streptavidin alone in parallel with the DNA-coated well for each serum. Serum reactivity to the streptavidin coated well was unusual, occurring in --5% of normal sera and in = 10% of SLE sera. However, if no control well had been run, these sera would have given false positive results. The reason for high binding on the control wells is not yet understood, but is currently under investigation. Calibration of the assay presented some problems, since the international standard for antiD N A is defined by the Farr assay, and contains both IgG and IgM D N A antibodies. To circumvent this problem, we isolated IgG from a lupus serum pool and determined the binding activity in IU of this IgG preparation in the
Amersham Farr assay (as defined by the WHO; Feltkamp et al., 1988). By comparing the absorbance of this IgG in the ELISA with its binding in the Farr, we were able to express ELISA results as IgG IU equivalents. To test the performance of this assay, normal sera and sera from patients with RA, PSS, chronic liver disease, and SLE were tested. No normal patients and none of the sera from RA patients or PSS patients had values greater than three standard deviations above the normal mean; two of 18 patients with chronic liver disease were slightly positive. 52% of unselected patients with SLE gave positive results (Fig. 4). This is in close agreement with the percentage positivity of unselected SLE patients in a number of other anti-DNA assays (Minter et al., 1979; Eaton et al., 1983; Miller et al., 1981; Tzioufas et al., 1987). In 56 of the SLE patients, clinical data was available based on retrospective chart analysis. The ELISA detected 16/18 patients with active renal disease, and patients with active non-renal and inactive disease showed lower levels of positivity (Fig. 5). To compare the ELISA with the Farr assay, all sera which had been tested by both assays were pooled; there was strong correlation between the two assays. This may be somewhat surprising, since the Farr assay detects IgM and IgG antiDNA, while the ELISA is IgG specific. Indeed, many of the most descrepant sera showed higher Farr values than ELISA values, suggesting the presence of IgM anti-DNA. When the two assays were compared by a 2 x 2 table analysis (Table I), 26 patients were positive by Farr assay and negative in the ELISA. Of these, 15 had SLE, and 11 did not have SLE; 12 patients were positive in the ELISA and negative in the Farr; eight of these had SLE and four did not. Results in this ELISA also correlated closely with results from the Sigma ELISA assay and a Millipore filter assay. The use of solid phase assays for the detection of D N A antibodies has been criticized since it may detect low avidity antibodies which may be present in diseases other than SLE, and which may not be relevant to the pathogenesis of SLE (Smeenk et al., 1982; Emlen et al., 1986; Smeenk et al., 1988). However, the use of this ELISA did not generate increased numbers of anti-DNA positive sera when compared to the Farr assay in
100 any of the patient groups studied. To better define the rate at which this assay detected D N A antibodies in non-SLE patients (false-positives), 121 consecutive A N A positive sera were tested in the ELISA and the Farr assay. The ELISA was positive in 17 patients, of whom 14 had definite SLE and two probable SLE (three A C R criteria). The Farr assay was positive in 23 patients, of whom 17 had SLE and two probable SLE. Thus the Farr assay detected four additional SLE patients, all of whom had inactive disease and high levels of IgM anti-DNA (data not shown). However, the Farr assay detected more false positive (4/23, 18.4%) than the ELISA (1/17, 5.8%). We conclude that this ELISA does not result in a greater number of false positive tests than the Farr assay. This lack of false positive results in the ELISA is in contrast to the data of Smeenk et al. (1988) and others who have concluded that the D N A ELISAs detect low avidity, cross-reactive antibodies which may not be representative of pathogenic D N A antibodies in SLE. There are several possible reasons why the present ELISA appears to be relatively insensitive to the detection of low avidity d s D N A antibodies. (1) By avoiding the use of positively charged molecules to pre-coat the plate, we have eliminated the detection of D N A - a n t i - D N A complexes and antibodies reactive with these positively charged substrates. (2) By using only dsDNA, we have minimized detection of ssDNA antibodies, many of which may be low avidity and positive in diseases other than SLE. (3) By only measuring IgG D N A antibodies, low avidity IgM anti-DNA are not detected. (4) By measuring OD at a single serum dilution, rather than using a limiting dilution method, the present ELISA decreases the sensitivity for detection of low avidity antibodies (Steward and Lew, 1985). (5) We have avoided the use of signal amplification in the ELISA since this has been reported to enhance detection of low avidity antibodies (Butler et al., 1978). Although we cannot at present delineate which of these mechanisms is most important, our data suggest that the presently described ELISA is not as susceptible to the detection of low avidity D N A antibodies as previous ELISA assays. This ELISA may in fact be less sensitive for the detection of SLE than the Farr assay, presumably since it cannot detect IgM anti-DNA seen in inactive SLE. However, because
of its specificity for IgG anti-DNA, it may be more specific for SLE than the Farr assay as demonstrated by the lower number of false positive results in an unselected ANA-positive population. In summary, we have developed an ELISA which employs streptavidin-biotin interaction to fix plasmid D N A to the plate and which is specific for the detection of IgG antibodies to dsDNA. The assay correlates closely with the Farr assay and does not show an increased frequency of false positive results when compared with the Farr assay. The technology of this ELISA circumvents many of the problems which have previously plagued anti-DNA ELISAs.
References Aarden, L.A., De Groot, E.R. and Feltkamp, T.E.W. (1975) Immunology of DNA. III. Crithidia luciliae, a simple substrate for the determination of anti-dsDNA with the immunofluorescence technique. Ann. N.Y. Acad. Sci. 254, 505. Aotsuka, S., Okawa, M., Ikebe, K. and Yokohari, R. (1979) Measurement of anti-double-stranded DNA antibodies in major immunoglobulin classes. J. Immunol. Methods 28, 149. Bell, C., Talal, N. and Schur, P.H. (1975) Antibodies to DNA in patients with rheumatoid arthritis and juvenile rheumatoid arthritis. Arthritis Rheum. 18, 535. Brinkman, K., Termaat, R.M., De Jong, J., Van den Brink, H.G., Berden, J.M. and Smeenk, R.J. (1989) Cross-reactive binding patterns of monoclonal antibodies to DNA are often caused by DNA/anti-DNA immune complexes. Res. Immunol. 140, 595. Butler, J.E., Feldbush, T.L., McGivern, P.L. and Stewart, N. (1978) The enzyme-linked immunosorbent assay (ELISA): a measure of antibody concentration or affinity? Immunochemistry 15, 131. Chubick, A., Sontheimer, R.D., Gilliam, J.N. and Ziff, M. (1978) An appraisal of tests for native DNA antibodies in connective tissue diseases. Ann. Int. Med. 89, 186. Deng, J.S., Sontheimer, R.D., Lipscomb, M.F. and Gilliam, J.N. (1985) The binding of antihistone antibodies to Crithidia lueiliae kinetoplasts is growth cycle-dependent. Arthritis Rheum. 28, 163. Eaton, R.B., Schnneider, G. and Schur, P.H. (1983) Enzyme immunoassay for antibodies to native DNA. Arthritis Rheum. 26, 52. Emlen, W., ,Pisetsky, D.S. and Taylor, R.P. (1986) Antibodies to DNA. Arthritis Rheum. 29, 1417. Engvall, E. (1976) Determination of antibodies to DNA by ELISA. Lancet ii, 1410.
101 Feltkamp, T.E.W., Kirkwood, T.B.L., Maini, R.N. and Aarden, L.A. (1988) The first international standard for antibodies to double-stranded DNA. Ann. Rheum. Dis. 47, 740. Fish, F. and Ziff, M. (1981) A sensitive solid phase microradioimmunoassay for antidouble stranded DNA antibodies. Arthritis Rheum. 24, 534. Ginsbert, B. and Keiser, H. (1973) A Millipore filter assay for antibody to native-DNA in sera of patients with SLE. Arthritis Rheum. 16, 199. Hasselbacher, P. and LeRoy, E.C. (1974) Serum DNA binding activity in healthy subjects and in rheumatic disease. Arthritis Rheum. 17, 63. Klotz, J.L., Minami, R.M. and Teplitz, R.L. (1979) An enzyme-linked immunosorbent assay for antibodies to native and denatured DNA. J. Immunol. Methods 29, 155. Koffler, D. (1974) Immunopathogenesis of systemic lupus erythematosus. Ann. Rev. Med. 25, 149. Koffler, D., Schur, P.H. and Kunkel, H.G. (1967) Immunological studies concerning the nephritis of systemic lupus erythematosus. J. Exp. Med. 126, 607. Lafer, E.M., Rauch, J., Andrzejewski, Jr., C., Mudd, D., Furie, B., Furie, B., Schwartz, R.S. and Stollar, B.D. (1981) Polyspecific monoclonal lupus autoantibodies reactive with both polynucleotides and phospholipids. J. Exp. Med. 153, 897. Locker, J.D., Medof, M.E., Bennett, R.M. and Sukhupunyaraksa, S. (1977) Characterization of DNA used to assay sera for anti-DNA antibodies; determination of the specificities of anti-DNA antibodies in SLE and non-SLE rheumatic disease states. J. Immunol. 118, 694. Miller, T.E., Lahita, R.G., Zarro, V.J., MacWilliam, J. and Koffler, D. (1981) Clinical significance of anti-doublestranded DNA antibodies detected by a solid phase enzyme immunoassay. Arthritis Rheum. 24, 602. Minter, M.F., Stollar, B.D. and Agnello, V. (1979) Reassessment of the clinical significance of native DNA antibodies in systemic lupus erythematosus. Arthritis Rheum. 22, 959. Notman, D.D., Dwata, N. and Tan, E.M. (1975) Profiles of antinuclear antibodies in systemic rheumatic diseases. Ann. Intern. Med. 83, 464. Pincus, T., Schur, P.H., Rose, J.A., Decker, J.L. and Talal, N. (1969) Measurement of serum DNA-binding activity in systemic lupus erythematosus. New Engl. J. Med. 281,701. Rothfield, N.F. and Stollar, B.D. (1967) The relation of immunoglobulin class, pattern of antinuclear antibody, and
complement-fixing antibodies to DNA in sera from patients with systemic lupus erythematosus. J. Clin. Invest. 46, 1785. Rubin, R.L., Joshn, F.G. and Tan, E.M. (1983) An improved ELISA for anti-native DNA by elimination of interference by anti-histone antibodies. J. Immunol. Methods 63, 359. Smeenk, R. (1986) Detection of antibodies to DNA by enzyme-linked immunosorbent assay (ELISA). Immunoassay Tech. 2, 121. Smeenk, R. and Aarden, L. (1980) The use of polyethylene-glycol precipitation to detect low-avidity anti-DNA antibodies in systemic lupus erytbematosus. J. Immunol. Methods 39, 165. Smeenk, R., Van der Lelij, G. and Aarden, L. (1982a) Measurement of low avidity anti-dsDNA by the Crithidia luciliae test and the PEG assay. Clin. Exp. Immunol. 50, 603. Smeenk, R., Van der Lelij, G. and Aarden, L. (1982b) Avidity of antibodies to dsDNA: comparison of IFT on Crithidia luciliae, Farr assay, and PEG assay. J. Immunol. 128, 73. Smeenk, R.J.T., Brinkman, K., Van den Brink, H.G. and Westgeest, A.A.A. (1988) Reaction patterns of monoclonal antibodies to DNA. J. Immunol. 140, 3786. Steward, M.W. and Lew, A.M. (1985) The importance of antibody affinity in the performance of immunoassays for antibody. J. Immunol. Methods 78, 173. Tan, E.M., Schur, P.H., Carr, R.I. and Kunkel, H.G. (1966) Deoxyribonucleic acid (DNA) and antibodies to DNA in the serum of patients with systemic lupus erythematosus. J. Clin. Invest. 45, 1732. Tzioufas, A.G., Manoussakis, M.N., Drosos, A.A., Silis, G., Gharavi, A.E. and Moutsopoulos, H.M. (1987) Enzyme immunoassays for the detection of IgG and IgM antidsDNA antibodies: clinical significance and specificity. Clin. Exp. Rheumatol. 5, 247. Urowitz, M.B., Gladman, D.D., Tozman, E.C.S. and Goldsmith, C.H. (1984) The lupus activity criteria count (LACC). J. Rheum. 11, 783. Vogt, V.M. (1980) Purification and properties of S1 nuclease from Aspergillus. Methods Enzymol. 65, 248. Zouali, M. and Stoilar, B.D. (1986) A rapid ELISA for measurement of antibodies to nucleic acid antigens using UVtreated polystyrene microplates. J. Immunol. Methods 90, 105.
91
JIM 05661
A new ELISA for the detection of double-stranded D N A antibodies W o o d r u f f Emlen, Pisamai Jarusiripipat and Gregory Burdick Division of Rheumatology, University of Colorado Health Sciences Center, 4200 E. 9th Avenue, Denver, CO 80262, U.S.A. (Received 21 August 1989, revised received 30 April 1990, accepted 1 May 1990)
The accurate detection of antibodies to double-stranded DNA is important in the diagnosis and management of patients with systemic lupus erythematosus (SLE). We have developed an ELISA in which plasmid dsDNA is biotinylated and bound to streptavidin coated wells. The biotinylated DNA did not lose is antigenicity, and the DNA which bound to the plate remained double-stranded. Only small amounts of DNA were required to coat the plates, and binding was highly reproducible. After defining the normal range of the assay, sera from patients with various conditions were examined; 0 out of 97 controls were positive, 0 out of 36 patients with RA were positive, 0 out of 14 patients with scleroderma, and two out of 18 patients with chronic liver disease were positive. In contrast, 52% of unselected SLE patients were positive, and there was a good correlation between level of DNA antibodies and disease activity. Comparison of this assay with other DNA assays (the Farr, Millipore, filter, and a commercial ELISA) showed a high degree of correlation between this ELISA and each of the othe r assays. Furthermore, screening of randomly selected ANA positive patients with this assay failed to show a high false positive rate as has been reported with other anti-DNA ELISAs. This assay is simple to perform, requires small amounts of DNA to coat the plate, is highly reproducible, and is specific for IgG antibodies to dsDNA. Key words: Systemic lupus erythematosus; Anti-DNA antibody; ELISA
Introduction
The measurement of antibodies to DNA in patients with systemic lupus erythematosus (SLE) has been shown to be clinically useful both in diagnosis of disease and in following the activity
Correspondence to: W. Emlen, Division of Rheumatology, University of Colorado Health Sciences Center, 4200 E. 9th Avenue, Denver, CO 80262, U.S.A. Abbreviations: SLE, systemic lupus erythematosus; dsDNA, double-stranded DNA; ssDNA, single-stranded DNA; PBS, 0.01 M phosphate, 0.15 M NaC1, pH 7.4; PBST, PBS 0.05% Tween.
of disease after diagnosis has been established (Tan et al., 1966; Koffler et al., 1967; Koffler, 1974; Emlen et al., 1986). A number of studies have shown that IgG antibodies to doublestranded DNA (dsDNA) are most specific for the diagnosis of SLE, and clinically most useful in following disease (Rothfield and Stollar, 1967; Pincus et al., 1969; Minter et al., 1979). Antibodies to single-stranded DNA (ssDNA) and IgM DNA antibodies are found in a number of other connective tissue diseases, liver diseases, as well as in some normal individuals (Notman et al., 1975; Locker et al., 1977). A large number of assays to measure DNA antibodies have been developed. Each of these assays has been plagued with the
0022-1759/90/$03.50 © 1990 Elsevier Science Publishers B.V. (Biomedical Division)
92 problem of purity of dsDNA, as well as problems unique to each assay. The Farr assay, in which radiolabelled DNA is bound to antibody and then precipitated with ammonium sulfate (Pincus et al., 1969) remains the standard method of measurement, although this assay detects IgM as well as IgG antibodies to DNA, and the use of radioactivity makes this assay undesirable in some clinical laboratory settings. Problems with single-stranded contamination of DNA preparations have also resulted in decreased specificity of the Farr assay, yielding positive results in a small number of patients with diseases other than SLE (Hasselbacher and LeRoy, 1974; Bell et al., 1975; Notman et al., 1975; Locker et al., 1977; Chubick et al., 1978). In an attempt to develop an assay specific for dsDNA, Aarden and colleagues (1975) developed an immunofluorescence assay in which antibodies bind to the kinetoplast of Crithidia luciliae. This organelle contains circular dsDNA, and by varying the specificity of the fluorescent antibody, this assay can be made IgG or IgM specific. However, this assay is only semiquantitative, and recent data has shown that the Crithidia kinetoplast also contains histones, resulting in false positive reactivity in, patients with anti-histone antibodies (Deng et al., 1985). A number of solid phase assays have been developed to detect DNA antibodies over the past 10 years. Early on it was noted that ssDNA bound well to plastic, but dsDNA bound poorly (Engvall, 1976; Smeenk, 1986). To facilitate dsDNA sticking to plastic, plates have been UV-irradiated (Zouali and Stollar, 1986), or have been precoated with methylated bovine serum albumin (Rubin et al., 1983) or with positively charged molecules such as poly-L-lysine or protamine sulfate which bind negatively charged DNA (Aotsuka et al., 1979; Klotz et al., 1979; Fish and Ziff, 1981; Eaton et al., 1983; Smeenk, 1986). These ELISAs have been effective for measuring DNA antibodies, but antigen binding to the plates has been variable, and a number of false positive results in non-SLE patients have been reported (Fish and Ziff, 1981; Eaton et al., 1983; Smeenk, 1986). These false positive results may occur because ELISA assays detect low avidity antibodies which are missed in the Farr assay, and which may not be clinically relevant in the pathogenesis of SLE
(Smeenk et al., 1982; Emlen et al., 1986; Smeenk, 1986), or because these assays detect DNA-antiD N A immune complexes (Brinkman et al., 1989). False positives may also result from antibodies to ssDNA since most ELISAs reported to date have used heterogeneous DNA (calf thymus DNA), which is likely to contain single-stranded regions which would preferentially bind to plastic (Engvail, 1976). In an attempt to develop a more dsDNAspecific ELISA, we have used plasmid DNA as an antigen source, and bound dsDNA to plastic via the specific streptavidin-biotin interaction. In this way, dsDNA is bound stably to the plastic independent of charge, and only small amounts of D N A are required to coat wells. This assay has proved to be reproducible, and the results correlate well with other standard DNA assays. When used to examine patients with SLE in varying stages of clinical disease activity, this assay was as sensitive and specific as other commercially available assays. Furthermore, when tested on unselected ANA positive patients, the false positive rate for high DNA antibodies in non-SLE patients was no higher with this ELISA than with the Farr assay.
Materials and methods
Preparation of DNA pGEM-3 plasmid (Promega, Madison, WI) was grown in E. coli and plasmid DNA was isolated by cell lysis followed by centrifugation over a cesium chloride gradient. Plasmid DNA was cut at a single site with endonuclease Sma-1 (Bethesda Research Labs, Gaithersburg, MD), yielding a linear, blunt-end dsDNA molecule of 2867 bp; analysis of this D N A on 0.8% agarose gels demonstrated a single homogeneous band. After ethanol precipitation, DNA was brought to a concentration of 1 m g / m l in water, and mixed with photoprobe biotin (Vector Labs, Burlingame, CA) (1 m g / m l ) at varying D N A / p h o t o p r o b e ratios in a constant volume of 100/xl. These mixes were UV irradiated using an H2 filter (295-400 nm) on ice through glass (to exclude UVB) for a total of 278 J / c m 2. This corresponded to 5 rain at a distance of 10 cm with a Dermalight 2001 Sol 3 solar
93 simulator. Following labeling, an equal volume of 0.1 M Tris p H 9.0 was added, and D N A was extracted x 3 with butanol to remove unreacted photoprobe. D N A was ethanol precipitated and resuspended in PBS (0.01 M phosphate, 0.1 M NaC1, pH 7.4) for use. The degree of biotinylation was determined by measuring the 00260 (DNA) and 00473 (biotin). D N A was brought to 10 m M E D T A and stored at 4 ° C as a concentrated solution (300-1000 /~g/ml) and diluted just prior to coating plates. Tritiated p G E M D N A was obtained by transforming E. coli strain W3110 thymidine with p G E M DNA. Tritiated thymidine was added to the cultures and plasmid D N A isolation and biotinylation procedures were carried out as above.
Standard ELISA procedure Preparation of plates. Immulon-2 96 well flatbottom microtiter plates (Dynatech, Alexandria, VA) were coated with a 1 /~g/ml solution of streptavidin (ICN Biochemicals, Cleveland, OH) in PBS at 150 /d/well overnight at 4°C. Wells were then washed 3 X with PBS/0.05% Tween (PBST) and post-coated with 200 /~1 PBST/0.2% gelatin (Baker, Phillipsburg, N J) for 1 h at 37 ° C. Wells were washed 3 x with PBST and 150 /~1 biotinylated DNA, diluted to 400 n g / m l in PBS/0.2% gelatin was added to even numbered columns; D N A diluent alone (PBS/0.2% gelatin) was added to odd numbered columns. Plates were incubated overnight at 4 ° C and washed 3 X with PBST. Plates were ready to use at this stage, or could be dried and used with no loss of binding activity for up to 6 weeks (longest time tested). ELISA assay. Serum was diluted 1/1000 in serdil (PBST/0.1% gelatin/0.5% bovine gammaglobulin (Sigma, St. Louis, MO)) and 150 lal was added to each well and incubated at room temperature for I h. Every serum to be tested was run in both an odd numbered (control) and an even numbered (DNA coated) well, so that each serum could serve as its own control. Following incubation, wells were washed x 3 with PBST, and 150 /~1 of a 1/2000 dilution (diluted in serdil) of peroxidase linked goat a-human IgG (Kirkegaard & Perry, Gaithersburg, MD) was added for 1 h at room temperature. Wells were washed x 5 with PBST, and 150 /~1 of ABTS solution (2,2'-azino-
bis(3-ethylbenzothiazoline-6-sulfonic acid) (Aldrich, Milwaukee, W I ) + H202 in McIlvain's buffer) was added. Absorbance at 405 nm (OD) was read using a Dynatech Micro ELISA Autoreader (MR580) at a fixed time (30 min), or when the OD of a standard serum had reached a given level. Background OD (non-DNA coated well) was subtracted from the OD of the D N A coated well, yielding the net signal OD. Optical density (OD) was converted to international units (IU) by calibrating the assay with a standard DNA antibody source, serum W o / 8 0 , obtained from the World Health Organization in Amsterdam (Feltkamp et al., 1988). On each plate, the following controls were run: serum diluent alone (used to blank the ELISA reader), normal human serum as a negative control, lupus pool (pool of five SLE sera) as a positive control, and a serum containing antibodies only to ssDNA to insure that the DNA on the plate was double-stranded. S1 nuelease digestion. To ascertain that only dsDNA was exposed on the plates, DNA-coated wells were digested for 60 min at 3 7 ° C with increasing amounts of a single-stranded specific $1 nuclease (BRL) in 10 m M ZnSO 4 at p H 4.6 (Vogt, 1980). To demonstrate that $1 nuclease was active against ssDNA, plates were made in which some wells were coated with heat denatured D N A (column 3), while other wells (column 2) were coated with intact dsDNA and column I with diluent alone. The effect of S1 nuclease digestion on each of these wells was tested using sera reactive with ssDNA or dsDNA. Sera. 97 control sera were obtained from normal blood bank donors at the Belle Bonfils Blood Center, Denver, CO. Rheumatoid arthritis (RA) and progressive systemic sclerosis (PSS) sera were obtained from patients fulfilling the ACR criteria for these diseases who were followed at the University of Colorado Health Sciences Center (UCHSC) Arthritis Clinic. Sera from patients with chronic liver disease (CLD) were obtained from the University of Colorado Health Sciences Gastroenterology Unit, and included six patients with chronic autoimmune hepatitis, five patients with primary biliary cirrhosis, and seven patients with alcoholic liver disease. Sera from SLE patients were obtained from three sources: (1) patients with SLE (by ACR criteria) who were seen in the
94 UCHSC Lupus Clinic; these sera were drawn on all SLE patients regardless of disease activity (n = 28); (2) patients with SLE who were seen in the Tulane University Lupus Clinic (gift of Dr. D. Boulware); these sera were drawn regardless of disease activity (n = 5); (3) sera taken from the UCHSC Serum Bank which had been selected for high D N A binding by screening Millipore filter assay (n = 58). Clinical information was available via retrospective chart analysis on 56 of these SLE patients. Disease activity in these 56 patients was determined according to the criteria of the lupus activity criteria count (LACC) (Urowitz et al., 1984), and patients were divided into three groups: active renal = active urinary sediment defined as RBC casts a n d / o r > 5 R B C / h p f ; active non-renal = arthritis/rash/pleuropericarditis/CNS disease/vascuhtis but no active renal disease; and inactive d i s e a s e - - n o manifestations of SLE. Because data was often available at only one point in time, changes in creatinine or proteinuria could not be considered in patient classification. Patients with active renal disease usually had other manifestations of active disease. In addition, sera from 121 consecutive ANA positive patients (titer > 1 : 64) from the UCHSC Clinical Immunology Lab were screened by ELISA and Farr assay for D N A antibody, and clinical information was obtained by retrospective chart review on all patients with increased D N A antibody levels by either assay. The diagnosis of SLE was made if these patients fulfilled four or more ACR criteria for SLE. Commercial DNA antibody assays. Sera were tested for D N A binding by a Farr binding assay using calf thymus D N A (Amersham Corp., Arlington Heights, IL) and an ELISA assay (Sigma). These assays were performed exactly according to manufacturer's protocols and specifications. Some sera were also tested in a Millipore filter assay (Ginsberg and Keiser, 1973) using 125I-DNA (Electronucleonics, MD) as an antigen source. Results
Assay conditions To establish the assay, initial experiments were performed to determine optimal coating and
post-coating conditions at each step. For these experiments a standard pool of lupus sera, derived from patients with known active SLE, was used as an antibody source (lupus pool). Initial coating of the plates with high concentrations of streptavidin (1 m g / m l ) resulted in a poor signal; therefore a concentration of 1 g g / m l was arbitrarily chosen. To determine the optimal conditions of DNA coating, D N A preparations were biotinylated to varying degrees, and wells were coated with 150 #1 of increasing concentrations of each of these preparations. A constant amount of antibody was added and the resultant signal (OD) was plotted as a function of D N A added to the wells. All biotinylated D N A preparations showed increasing signal with increased D N A coating; however, more heavily biotinylated D N A preparations gave a lower signal, particularly at high D N A concentrations. Less biotinylated D N A gave an adequate signal, but the signal had not reached a maximum at concentrations in excess of 1000 n g / m l (data not shown). Based on these data, we chose standard conditions of biotinylation corresponding to one biotin per 30 bp of D N A and a standard D N A concentration of 400 n g / m i for coating plates. This concentration of D N A was on the plateau portion of the curve, such that minor errors in dilution did not significantly alter results. Because of the possibility that some sera might contain small amounts of antibody reactive with streptavidin or gelatin in the post-coat, ELISA plates were set up so that each sample could act as its own control. Odd numbered columns were coated with streptavidin and post-coat only, while even numbered columns were coated with streptavidin, post-coat, and biotinylated DNA. Each serum sample was run both in the D N A coated well and in a control well, and background absorbance was subtracted from the signal for each serum. A 1/1000 dilution of serum was arbitrarily chosen for screening sera because it gave an good net signal (signal minus background) with background O D values (non-DNA coated well) less than 0.1. With increasing amounts of serum, net signal increased, but background OD values also increased. Dilution of second antibody and time of reading were chosen arbitrarily. We next examined the effects of different DNA substrates and serum dilution on the signal gener-
95
2.0 1.5 v
.Q
25 IU (normal mean+ 3SD). Sera included are 91 SLE patients, 41 normal controls, 16 patients with CLD, and 121 consecutiveunselected ANA positivepatients. Farr Positive Negative
ELISA Positive 71 12 83
Negative 26 160 188
97 172
called negative, and the ELISA was positive for 12 sera that the Farr assay called negative. Review of clinical data of the patients in whom the assays were discordant showed that 15 of 26 Farr positive-ELISA negative patients had SLE; and eight of 12 ELISA positive-Farr negative patients had SLE. Of the total of 97 Farr positive patients, 12 did not have SLE (12.4% false positives); of 83 ELISA positive patients, five did not have SLE (6% false positives). The ELISA was then compared with a commercial ELISA (Sigma) and a standard Millipore filter assay using sera from SLE patients and from normal patients. The ELISA correlated well with both assays (Sigma, n = 74, r = 0.74, p < 0.001; Millipore filter n = 192, r = 0.58, P < 0.001) (data not shown). Using a 2 × 2 table of positive and negative results, the ELISA agreed with the Sigma assay in 74% of the samples, and with the Millipore filter assay in 80% of the samples tested.
Discussion We have developed an ELISA for the detection of lgG dsDNA antibodies which takes advantage of the strong interaction between streptavidin and biotinylated DNA. By pre-coating wells with streptavidin, we were able to efficiently and reproducibly bind dsDNA to the wells. By using plasmid DNA, which is entirely double-stranded, and by blocking the wells prior to the addition of DNA, we have minimized the possibility of nonspecific sticking of D N A which might contain
single-stranded regions. Performance of this assay required only very small amounts of serum (1/1000 dilution of serum routinely), was highly reproducible, and showed a low frequency of false positive results. In order to justify the use of biotinylated DNA as an antigen in this assay, it was necessary to insure that biotinylated D N A was not significantly altered antigenically. We therefore tested 3H-plasmid D N A in a Farr assay, and compared it with the behavior of biotinylated 3H-plasmid D N A in the same assay. Our results (Fig. 2) show that biotinylation of D N A to the degree of one biotin per 30 bp pairs had no effect on antigenicity of the DNA. Higher degrees of biotinylation showed a decreased signal on the ELISA assay, suggesting some possible loss of antigenicity with higher degrees of biotinylation. Although these experiments certainly do not rule out small changes in the antigenicity of the biotinylated DNA, the similarity of the Farr binding curves strongly suggests that binding of antibodies to D N A is not significantly altered by D N A biotinylation, The use of the streptavidin-biotin interaction to bind D N A to the wells has several advantages over other methods. Previously, positively charged molecules such as poly-L-lysine or protamine have been used to facilitate dsDNA binding (Aotsuka et al., 1979; Klotz et al., 1979; Fish and Ziff, 1981; Eaton et al., 1983; Smeenk, 1986). Since these molecules are positively charged and may mimic histones, antibodies reactive to histone or D N A anti-DNA immune complexes (Brinkman et al., 1989) may bind to the positively charged molecules on the plate. The p l of streptavidin is 6.2, such that the binding of dsDNA to the plates in our system is not charge dependent. Furthermore, the biotin-streptavidin interaction is extremely stable, making binding very efficient and reproducible. Indeed, significantly lower amounts of D N A were required to coat the wells in our ELISA than in many other systems (Aotsuka et al., 1979; Klotz et al., 1979; Fish and Ziff, 1981; Eaton et al., 1983; Rubin et al., 1983; Smeenk, 1986). Since it is known that ssDNA sticks non-specifically to plates, the possibility must be raised that in other ELISA systems, small amounts of contaminating ssDNA, which would preferentially bind to the well (Engvall, 1976), might make up a significant
99 fraction of the total D N A bound. In the present ELISA, plates were blocked after the coating of streptavidin and before the addition of DNA, to minimize any non-specific sticking of D N A which was not dependent upon the streptavidin-biotin interaction. Control experiments (Fig. 1) showed that non-biotinylated D N A gave no signal in the ELISA, confirming that non-specific sticking of D N A had not occurred. Furthermore, no signal in the ELISA was generated when biotinylated D N A or non-biotinylated D N A was coated directly onto plates from which the streptavidin had been omitted (Fig. 1). Thus, the use of the streptavidinbiotin linkage system insures that the D N A which binds to the plate accurately represents the D N A which is put into the wells. Analysis of the D N A used to coat wells on agarose gels showed a homogeneous band 2867 bp in size. To insure that the D N A on the plate remained double-stranded, DNA-coated wells were digested with S1 nuclease to remove any ssDNA or D N A containing single-stranded regions. Under conditions that completely removed ssDNA, there was no diminution of antibody binding to dsDNA (Fig. 3), indicating that no antigen had been removed from the plate with S1 nuclease treatment. Furthermore, testing of sera and a monoclonal antibody with ssDNA specificity showed a > 10 : 1 ss/ds signal ratio. These data strongly support the conclusions that biotinylated plasmid D N A is adsorbed onto the streptavidin intact, and remains completely double-stranded. Another important feature of this ELISA is the use of a control well coated with streptavidin alone in parallel with the DNA-coated well for each serum. Serum reactivity to the streptavidin coated well was unusual, occurring in --5% of normal sera and in = 10% of SLE sera. However, if no control well had been run, these sera would have given false positive results. The reason for high binding on the control wells is not yet understood, but is currently under investigation. Calibration of the assay presented some problems, since the international standard for antiD N A is defined by the Farr assay, and contains both IgG and IgM D N A antibodies. To circumvent this problem, we isolated IgG from a lupus serum pool and determined the binding activity in IU of this IgG preparation in the
Amersham Farr assay (as defined by the WHO; Feltkamp et al., 1988). By comparing the absorbance of this IgG in the ELISA with its binding in the Farr, we were able to express ELISA results as IgG IU equivalents. To test the performance of this assay, normal sera and sera from patients with RA, PSS, chronic liver disease, and SLE were tested. No normal patients and none of the sera from RA patients or PSS patients had values greater than three standard deviations above the normal mean; two of 18 patients with chronic liver disease were slightly positive. 52% of unselected patients with SLE gave positive results (Fig. 4). This is in close agreement with the percentage positivity of unselected SLE patients in a number of other anti-DNA assays (Minter et al., 1979; Eaton et al., 1983; Miller et al., 1981; Tzioufas et al., 1987). In 56 of the SLE patients, clinical data was available based on retrospective chart analysis. The ELISA detected 16/18 patients with active renal disease, and patients with active non-renal and inactive disease showed lower levels of positivity (Fig. 5). To compare the ELISA with the Farr assay, all sera which had been tested by both assays were pooled; there was strong correlation between the two assays. This may be somewhat surprising, since the Farr assay detects IgM and IgG antiDNA, while the ELISA is IgG specific. Indeed, many of the most descrepant sera showed higher Farr values than ELISA values, suggesting the presence of IgM anti-DNA. When the two assays were compared by a 2 x 2 table analysis (Table I), 26 patients were positive by Farr assay and negative in the ELISA. Of these, 15 had SLE, and 11 did not have SLE; 12 patients were positive in the ELISA and negative in the Farr; eight of these had SLE and four did not. Results in this ELISA also correlated closely with results from the Sigma ELISA assay and a Millipore filter assay. The use of solid phase assays for the detection of D N A antibodies has been criticized since it may detect low avidity antibodies which may be present in diseases other than SLE, and which may not be relevant to the pathogenesis of SLE (Smeenk et al., 1982; Emlen et al., 1986; Smeenk et al., 1988). However, the use of this ELISA did not generate increased numbers of anti-DNA positive sera when compared to the Farr assay in
100 any of the patient groups studied. To better define the rate at which this assay detected D N A antibodies in non-SLE patients (false-positives), 121 consecutive A N A positive sera were tested in the ELISA and the Farr assay. The ELISA was positive in 17 patients, of whom 14 had definite SLE and two probable SLE (three A C R criteria). The Farr assay was positive in 23 patients, of whom 17 had SLE and two probable SLE. Thus the Farr assay detected four additional SLE patients, all of whom had inactive disease and high levels of IgM anti-DNA (data not shown). However, the Farr assay detected more false positive (4/23, 18.4%) than the ELISA (1/17, 5.8%). We conclude that this ELISA does not result in a greater number of false positive tests than the Farr assay. This lack of false positive results in the ELISA is in contrast to the data of Smeenk et al. (1988) and others who have concluded that the D N A ELISAs detect low avidity, cross-reactive antibodies which may not be representative of pathogenic D N A antibodies in SLE. There are several possible reasons why the present ELISA appears to be relatively insensitive to the detection of low avidity d s D N A antibodies. (1) By avoiding the use of positively charged molecules to pre-coat the plate, we have eliminated the detection of D N A - a n t i - D N A complexes and antibodies reactive with these positively charged substrates. (2) By using only dsDNA, we have minimized detection of ssDNA antibodies, many of which may be low avidity and positive in diseases other than SLE. (3) By only measuring IgG D N A antibodies, low avidity IgM anti-DNA are not detected. (4) By measuring OD at a single serum dilution, rather than using a limiting dilution method, the present ELISA decreases the sensitivity for detection of low avidity antibodies (Steward and Lew, 1985). (5) We have avoided the use of signal amplification in the ELISA since this has been reported to enhance detection of low avidity antibodies (Butler et al., 1978). Although we cannot at present delineate which of these mechanisms is most important, our data suggest that the presently described ELISA is not as susceptible to the detection of low avidity D N A antibodies as previous ELISA assays. This ELISA may in fact be less sensitive for the detection of SLE than the Farr assay, presumably since it cannot detect IgM anti-DNA seen in inactive SLE. However, because
of its specificity for IgG anti-DNA, it may be more specific for SLE than the Farr assay as demonstrated by the lower number of false positive results in an unselected ANA-positive population. In summary, we have developed an ELISA which employs streptavidin-biotin interaction to fix plasmid D N A to the plate and which is specific for the detection of IgG antibodies to dsDNA. The assay correlates closely with the Farr assay and does not show an increased frequency of false positive results when compared with the Farr assay. The technology of this ELISA circumvents many of the problems which have previously plagued anti-DNA ELISAs.
References Aarden, L.A., De Groot, E.R. and Feltkamp, T.E.W. (1975) Immunology of DNA. III. Crithidia luciliae, a simple substrate for the determination of anti-dsDNA with the immunofluorescence technique. Ann. N.Y. Acad. Sci. 254, 505. Aotsuka, S., Okawa, M., Ikebe, K. and Yokohari, R. (1979) Measurement of anti-double-stranded DNA antibodies in major immunoglobulin classes. J. Immunol. Methods 28, 149. Bell, C., Talal, N. and Schur, P.H. (1975) Antibodies to DNA in patients with rheumatoid arthritis and juvenile rheumatoid arthritis. Arthritis Rheum. 18, 535. Brinkman, K., Termaat, R.M., De Jong, J., Van den Brink, H.G., Berden, J.M. and Smeenk, R.J. (1989) Cross-reactive binding patterns of monoclonal antibodies to DNA are often caused by DNA/anti-DNA immune complexes. Res. Immunol. 140, 595. Butler, J.E., Feldbush, T.L., McGivern, P.L. and Stewart, N. (1978) The enzyme-linked immunosorbent assay (ELISA): a measure of antibody concentration or affinity? Immunochemistry 15, 131. Chubick, A., Sontheimer, R.D., Gilliam, J.N. and Ziff, M. (1978) An appraisal of tests for native DNA antibodies in connective tissue diseases. Ann. Int. Med. 89, 186. Deng, J.S., Sontheimer, R.D., Lipscomb, M.F. and Gilliam, J.N. (1985) The binding of antihistone antibodies to Crithidia lueiliae kinetoplasts is growth cycle-dependent. Arthritis Rheum. 28, 163. Eaton, R.B., Schnneider, G. and Schur, P.H. (1983) Enzyme immunoassay for antibodies to native DNA. Arthritis Rheum. 26, 52. Emlen, W., ,Pisetsky, D.S. and Taylor, R.P. (1986) Antibodies to DNA. Arthritis Rheum. 29, 1417. Engvall, E. (1976) Determination of antibodies to DNA by ELISA. Lancet ii, 1410.
101 Feltkamp, T.E.W., Kirkwood, T.B.L., Maini, R.N. and Aarden, L.A. (1988) The first international standard for antibodies to double-stranded DNA. Ann. Rheum. Dis. 47, 740. Fish, F. and Ziff, M. (1981) A sensitive solid phase microradioimmunoassay for antidouble stranded DNA antibodies. Arthritis Rheum. 24, 534. Ginsbert, B. and Keiser, H. (1973) A Millipore filter assay for antibody to native-DNA in sera of patients with SLE. Arthritis Rheum. 16, 199. Hasselbacher, P. and LeRoy, E.C. (1974) Serum DNA binding activity in healthy subjects and in rheumatic disease. Arthritis Rheum. 17, 63. Klotz, J.L., Minami, R.M. and Teplitz, R.L. (1979) An enzyme-linked immunosorbent assay for antibodies to native and denatured DNA. J. Immunol. Methods 29, 155. Koffler, D. (1974) Immunopathogenesis of systemic lupus erythematosus. Ann. Rev. Med. 25, 149. Koffler, D., Schur, P.H. and Kunkel, H.G. (1967) Immunological studies concerning the nephritis of systemic lupus erythematosus. J. Exp. Med. 126, 607. Lafer, E.M., Rauch, J., Andrzejewski, Jr., C., Mudd, D., Furie, B., Furie, B., Schwartz, R.S. and Stollar, B.D. (1981) Polyspecific monoclonal lupus autoantibodies reactive with both polynucleotides and phospholipids. J. Exp. Med. 153, 897. Locker, J.D., Medof, M.E., Bennett, R.M. and Sukhupunyaraksa, S. (1977) Characterization of DNA used to assay sera for anti-DNA antibodies; determination of the specificities of anti-DNA antibodies in SLE and non-SLE rheumatic disease states. J. Immunol. 118, 694. Miller, T.E., Lahita, R.G., Zarro, V.J., MacWilliam, J. and Koffler, D. (1981) Clinical significance of anti-doublestranded DNA antibodies detected by a solid phase enzyme immunoassay. Arthritis Rheum. 24, 602. Minter, M.F., Stollar, B.D. and Agnello, V. (1979) Reassessment of the clinical significance of native DNA antibodies in systemic lupus erythematosus. Arthritis Rheum. 22, 959. Notman, D.D., Dwata, N. and Tan, E.M. (1975) Profiles of antinuclear antibodies in systemic rheumatic diseases. Ann. Intern. Med. 83, 464. Pincus, T., Schur, P.H., Rose, J.A., Decker, J.L. and Talal, N. (1969) Measurement of serum DNA-binding activity in systemic lupus erythematosus. New Engl. J. Med. 281,701. Rothfield, N.F. and Stollar, B.D. (1967) The relation of immunoglobulin class, pattern of antinuclear antibody, and
complement-fixing antibodies to DNA in sera from patients with systemic lupus erythematosus. J. Clin. Invest. 46, 1785. Rubin, R.L., Joshn, F.G. and Tan, E.M. (1983) An improved ELISA for anti-native DNA by elimination of interference by anti-histone antibodies. J. Immunol. Methods 63, 359. Smeenk, R. (1986) Detection of antibodies to DNA by enzyme-linked immunosorbent assay (ELISA). Immunoassay Tech. 2, 121. Smeenk, R. and Aarden, L. (1980) The use of polyethylene-glycol precipitation to detect low-avidity anti-DNA antibodies in systemic lupus erytbematosus. J. Immunol. Methods 39, 165. Smeenk, R., Van der Lelij, G. and Aarden, L. (1982a) Measurement of low avidity anti-dsDNA by the Crithidia luciliae test and the PEG assay. Clin. Exp. Immunol. 50, 603. Smeenk, R., Van der Lelij, G. and Aarden, L. (1982b) Avidity of antibodies to dsDNA: comparison of IFT on Crithidia luciliae, Farr assay, and PEG assay. J. Immunol. 128, 73. Smeenk, R.J.T., Brinkman, K., Van den Brink, H.G. and Westgeest, A.A.A. (1988) Reaction patterns of monoclonal antibodies to DNA. J. Immunol. 140, 3786. Steward, M.W. and Lew, A.M. (1985) The importance of antibody affinity in the performance of immunoassays for antibody. J. Immunol. Methods 78, 173. Tan, E.M., Schur, P.H., Carr, R.I. and Kunkel, H.G. (1966) Deoxyribonucleic acid (DNA) and antibodies to DNA in the serum of patients with systemic lupus erythematosus. J. Clin. Invest. 45, 1732. Tzioufas, A.G., Manoussakis, M.N., Drosos, A.A., Silis, G., Gharavi, A.E. and Moutsopoulos, H.M. (1987) Enzyme immunoassays for the detection of IgG and IgM antidsDNA antibodies: clinical significance and specificity. Clin. Exp. Rheumatol. 5, 247. Urowitz, M.B., Gladman, D.D., Tozman, E.C.S. and Goldsmith, C.H. (1984) The lupus activity criteria count (LACC). J. Rheum. 11, 783. Vogt, V.M. (1980) Purification and properties of S1 nuclease from Aspergillus. Methods Enzymol. 65, 248. Zouali, M. and Stoilar, B.D. (1986) A rapid ELISA for measurement of antibodies to nucleic acid antigens using UVtreated polystyrene microplates. J. Immunol. Methods 90, 105.
