Elevated immunoglobulin G antibodies to the proline-rich amino-terminal region of Epstein–Barr virus nuclear antigen-2 in sera from patients with systemic connective tissue diseases and from a subgroup of Sjögren's syndrome patients with pulmonary involvements

Abstract

Associations of Epstein–Barr virus (EBV) and autoimmune diseases have been hypothesized. We have analysed IgG antibodies to EBV nuclear antigen (EBNA)-2 in sera from Japanese patients with autoimmune systemic connective tissue diseases (CTD), exemplified by systemic lupus erythematosus (SLE), primary Sjögren's syndrome (SS), rheumatoid arthritis (RA), systemic sclerosis (SSc) and secondary SS (classical CTDs complicated with SS). An enzyme-linked immunosorbent assay (ELISA) which uses glutathione-S-transferase polypeptides fused to EBV nuclear antigen (EBNA)-2 and EBNA-1 was developed. Ratios of IgG antibody reactivity to whole IgG concentrations of sera were calculated to normalize EBNA-2 and EBNA-1 antibody levels to the hypergammaglobulinaemia that occurs in CTD. The ELISA optical density OD₄₅₀ readings of IgG antibodies to both the amino-terminal aa 1–116 of EBNA-2 and carboxyl-terminal aa 451–641 of EBNA-1 were elevated significantly in patients with SLE, primary SS, RA, SSc and secondary SS when compared to EBNA-1. The OD readings were divided by serum IgG concentrations to normalize for the hypergammaglobulinaemia. The specific levels of IgG antibodies to the amino-terminal region of EBNA-2 were elevated in patients with SLE, primary SS or RA, as well as those with secondary SS complicated with SLE or RA. The EBNA-2 amino-terminal region contains a polyproline tract and a proline-rich sequence and has considerable amino acid sequence homology with many cellular proline-rich proteins. High ratios of EBNA-2 aa 1–116 to EBNA-1 aa 451–641 IgG antibody levels which probably suggest reactivation of EBV latent infection were associated significantly with pulmonary involvement in SS patients. These results are consistent with the hypothesis that the sequence similarity between the amino-terminal region of EBNA-2 and proline-rich cellular proteins is associated with pathogenesis in a subpopulation of CTD patients, possibly by the molecular mimicry–epitope shift mechanism.

Introduction

The hypothesis that infectious agents are involved in human autoimmune diseases has been a focus of both immunological and microbiological studies [1,2]. Although experiments using laboratory animals show causal relationships between microbial infection and autoimmune diseases [3], the mechanism of pathogenesis is unclear. The sequence homology and immunological cross-reactivity between human cytomegalovirus and cellular HLA-DR2 beta chain [1] has led to the molecular mimicry theory, which has been interpreted widely as a mechanism for pathogenesis of autoimmune diseases by infectious agents [4–6]. The molecular mimicry theory has been strengthened by reports that show epitope spreading via diversification of autoimmune epitopes that may or may not be related to the original sequence and the gradual release of sequestered epitopes [3,7–16]. Epstein–Barr virus (EBV) infection might be among the factors that are involved in both rheumatoid arthritis (RA) [17–23] and in Sjögren syndrome (SS) [24,25].

We have previously reported that geometric mean titres of IgG antibodies to an EBV-encoded nuclear antigen-2 (EBNA-2) are elevated in sera from patients with SS with an enzyme linked immunosorbent assay (ELISA) using fusion polypeptides of β-galactosidase and EBNA-2 [26]. In this report, we have further analysed levels of IgG antibodies that react with EBNA-2 and EBNA-1 in sera from patients with major systemic connective tissue diseases (CTD). The sera in the present study were collected at a single hospital that was different from hospitals where patients in our previous report were diagnosed and sera were collected. In addition to patients with SS and RA, we have also studied individuals with systemic lupus erythematosus (SLE), systemic sclerosis (SSc) and polymyositis/dermatomyositis (PM/DM) [27–30]. Patients with mixed connective tissue disease (MCTD) that is classified by Kasukawa's criteria [31] and secondary SS (classical CTDs complicated with SS) were also studied. We have refined our previously reported ELISA [32] by introducing fusion polypeptides of glutathione-S-transferase (GST) with EBNA-2 or EBNA-1 as antigen. Furthermore, hypergammaglobulinaemia is common in SS [25,33–35]. The immunoglobulins in these patients are often autoantibodies directed against non-organ-specific antigens, such as serum immunoglobulins (rheumatoid factor), antinuclear antibodies, extractable ribonucleoprotein antigens (known as Ro or SS-A and La or SS-B) and organ-specific antigens [35,36]. In this study we have normalized the ELISA readings to total IgG concentrations in sera to correct values for the hypergammaglobulinaemia that occurs in patients.

Materials and methods

Construction of GST fusion polypeptides expressing EBNA-1 or EBNA-2 polypeptides

The EBNA-1 expressing GST vector, pGST-EBNA-1(C) was constructed by inserting the EBNA-1 DNA clone K2 [32] with the carboxyl-terminal aa 451–641 sequence into the EcoRI cloning site of pGEX-5X-1 (Amersham Pharmacia Biotech, Buckinghamshire, UK). The EBNA-2-expressing GST vector, pGST-EBNA-2(N), was constructed by subcloning the EBNA-2 DNA clone Y1 [32] that contains the amino-terminal aa 1–116 region at the EcoRI site of pGEX-5X-3 (Amersham Pharmacia Biotech). To express the GST-fusion proteins the Esherichia coli strain JM109 was transformed with pGST-EBNA-1(C) or pGST-EBNA-2(N). The E. coli JM109 transformants were grown and incubated in the presence of 0·3 mm IPTG at 37°C for 2 h. The cells were disrupted by ultrasonication, centrifuged at 10 000 g for 30 min, and the supernatants were subjected to Glutathione-Sepharose-4B (Pharmacia) affinity chromatography. GST fusion polypeptides were eluted with phosphate buffered saline (PBS) containing 10 mm reduced glutathione (Sigma, Missouri, USA).

ELISA and antigens

A volume of 100 µl per well of 2·0 µg/ml of GST-fusion protein in carbonate buffer (pH 9·6) was incubated at 4°C overnight on polystyrene 96-well microplates (Nunc, Roskilde, Denmark), followed by two washes with PBS. Wells were blocked by adding PBS containing 0·5% bovine serum albumin (BSA) and 5% lactose. Plates were incubated at 4°C overnight and washed with PBS containing 0·05% Tween 20. Serum specimens were diluted with PBS containing 0·5% bovine serum albumin and 0·05% Tween 20 and 100 µl of each dilution was added to duplicate wells. Plates were incubated at 37°C for 1 h, then washed three times with PBS containing 0·05% Tween 20. Plates were incubated with horseradish-peroxidase-conjugated goat antihuman IgG (MBL, Japan) at 37°C for 1 h, and then washed three times. Thirty min after adding 100 µl of 3,3′,5,5′-tetrametylbenzidine (DAKO) per well, the OD₄₅₀ was read with a microplate autoreader (ER-8000, Sanko Junyaku, Japan). The mean OD₄₅₀ ± 3 s.d. obtained by assay of 12 different human sera, which were negative for EBV in both standard immunofluorescence (IF) tests and Western blots using the recombinant antigens, was used as the cut-off value. Background OD₄₅₀ readings of IgG antibodies to GST were subtracted from OD₄₅₀ readings of each serum using GST-EBNA-1(C) or EBNA-2(N) as antigen. Rheumatoid factor in sera was analysed with RF-LATEX ‘Seiken’ (Denka Seiken, Japan). Rheumatoid factor (Denka Seiken) was incorporated into some ELISA to test the interfering potential.

To test sera for antinuclear antibodies against multiple nuclear antigens, we used the ELISA kit MESACUP ANA (MBL, Japan) that contains recombinant proteins of RNP-A, RNP-70 k, 60K SS-A, 52K SS-A, SS-B, CENP-B, Scl-70 and Jo-1, purified Sm and SS-A proteins, and dsDNA and ssDNA of lambda phage.

Patients and sera and immunoglobulin classes

The study groups were Japanese individuals who lived in and around Tokyo. Informed consent was obtained from all participants. The study protocol was approved by the Board for Medical Ethics and Human Investigation of the International Medical Center of Japan (IMCJ) and complied with both institutional and international guidelines for the study of human subjects. Sera from both patients with CTD and matched control healthy adult volunteers were collected at the Connective Tissue Disease Division, IMCJ. There were 102 SLE, 74 primary SS, 70 RA, 15 SSc, 51 secondary SS, seven PM/DM, nine MCTD patients and 120 volunteers in the healthy control group. Patients were examined and diagnosed according to the classification criteria for CTD [27–31,37–39]. Immunoglobulin classes were quantified with IgG-TIA ‘Seiken’, IgM-TIA ‘Seiken’ and IgA-TIA ‘Seiken’ (Denka Seiken, Japan).

Statistical analyses

Statistical differences in ELISA readings were analysed by the non-parametric Mann–Whitney U-test that has been established to be independent of a distribution pattern of a given population and the non-parametric Spearman's rank correlation coefficient test. Statistical differences in the incidence of clinical features and laboratory findings were analysed by either Fisher's exact probability or χ² tests.

Sequence homology searches

Sequence homology searches were performed using the program ‘Protein; the search for short, nearly exact matches’ of blast 2·2.9 [40] and the database SwissProt. Sequence homology searches were also performed using the program genomes/human and the databases ‘Build protein’ and ‘Human ref prot’ of blast 2·2.9 and SwissProt.

Clinical assessment

Data for 134 clinical manifestations or laboratory diagnostic tests such as arthritis, malar rash and leukopenia of patients were compared. Pulmonary examinations were carried out as described previously [41]. Pulmonary function was evaluated with a Collins Pulmonary Testing System (Collins, MA, USA). The restrictive change in pulmonary function was defined when the percentage of vital capacity was = 80%. Carbon monoxide diffusion in the lung was measured by the single-breath method [42]. The decreased diffusion capacity of carbon monoxide was defined when the percentage of diffusion lung capacity of carbon monoxide (DL_CO) was = 70% of the normal clinical value [41].

Results

ELISA background readings to GST are low and interference from rheumatoid factor in ELISA of IgG antibodies to EBNA-2 (aa 1–116) is negligible

The average of ELISA OD₄₅₀ reading at a 1 : 200 dilution of human sera to β-galactosidase was approximately 0·08. In the ELISA using GST as antigen, the average of OD₄₅₀ readings at a single 1 : 200 dilution of sera (n = 515) from the patients with the CTD and healthy control donors was 0·015 and the median was 0·010. The average and median were both smaller than those in the ELISA using maltose-binding protein (MBP) as antigen (average = 0·020, median = 0·012). Therefore, pGST-EBNA-1(C) and pGST-EBNA-2(N) were used in this study for ELISA antigens.

In sera from 180 patients with CTD, the average concentration of the rheumatoid factor (RF), which is a complex of mostly IgM antibodies directed to the Fc portion of IgG, was 130 and the median was 52 U/ml. RF concentration was less than 250 U/ml in 160 patient sera and did not exceed 1500 U/ml in any single sera. To test for a possible effect on the ELISA, RF was added to sera from nine RF-negative healthy individuals. Interference by RF under the concentration of 2000 unit/ml was negligible in ELISA for EBNA-2(N) IgG antibodies (Fig. 1a), but RF impacted the ELISA results for EBNA-1(C) IgG antibodies (Fig. 1b). The results suggest that RF reacted with some of EBNA-1(C) IgG antibodies, thus reducing the effective concentration of EBNA-1(C) IgG antibodies detected in the ELISA.

Fig. 1.

Interference by rheumatoid factor in the ELISA and concentrations of IgG, IgA and IgM proteins in sera from patients with systemic lupus erythematosus (SLE), primary Sjögrens’ syndrome (SS), rheumatoid arthritis (RA), systemic sclerosis (SSc), polymyositis/dermatomyositis (PM/DM) or mixed connective tissue disease (MCTD). (a, b) Rheumatoid factor was added to nine rheumatoid factor-negative sera in the concentration range of 125–2000 U/ml. Each of the nine sera are represented by individual symbols. (a) OD₄₅₀ readings in ELISA for IgG antibodies to EBNA-2(N). (b) OD₄₅₀ readings in ELISA for IgG antibodies to EBNA-1(C).

We tested whether there was a correlation between the presence of autoantibodies in patients’ sera and ELISA OD₄₅₀ readings of IgG antibodies to EBNA-2(N) and EBNA-1(C). There were no significant differences in ELISA OD₄₅₀ readings of IgG antibodies to the EBNA-2(N) and EBNA-1(C) polypeptides between the antinuclear antigen mixture MESACUP ANA antibody-positive and MESACUP ANA-negative subgroups, between SS-A antibody-positive and SSA antibody-negative subgroups, between dsDNA antibody-positive and dsDNA antibody-negative subgroups, and between the ssDNA antibody-positive and ssDNA antibody-negative subgroups of patients with SS, SLE or RA (data not shown). These results suggest that there is no relationship of IgG antibodies to EBNA-2(N) and EBNA-1(C) to the autoantibodies to ANA, SS-A, dsDNA and ssDNA and that the levels of IgG antibodies to EBNA-2(N) and EBNA-1(C) in sera from the CTD patients are not correlated significantly with the presence or absence of these autoantibodies.

ELISA levels of IgG antibodies against EBNA-2 and EBNA-1 in sera from patients with CTD

Levels of IgG antibodies to EBNA-2(N) and EBNA-1(C) are shown in Fig. 2. The frequency of extremely high ELISA OD₄₅₀ readings (>0·40) to EBNA-2(N) was much higher in sera from patients with SLE or RA than from healthy control individuals (Fig. 2a). The frequency of extremely high ELISA OD₄₅₀ readings (>1·0) to EBNA-1(C) was little higher in sera from patients with primary SLE, primary SS, RA, SSc or MCTD than in sera from healthy controls compared to EBNA-2 (aa 1–116) IgG levels (Fig. 2c).

Fig. 2.

OD₄₅₀ readings of IgG antibodies to EBNA-2(N) and EBNA-1(C) in sera from patients with SLE, SS, RA, SSc or MCTD. (a) OD₄₅₀ readings of IgG antibodies to EBNA-2(N) in sera from patients with SLE, SS, RA, SSc, PM/DM or MCTD. (b) OD₄₅₀ readings of IgG antibodies to EBNA-2(N) in sera from patients with secondary SS. (c) OD₄₅₀ readings of IgG antibodies to EBNA-1(C) in sera from patients with SLE, SS, RA, SSc, PM/DM or MCTD. (d) OD₄₅₀ readings of IgG antibodies to EBNA-1(C) in sera from patients with secondary SS.

IgG levels to EBNA-2(N) were significantly higher in sera from patients with SLE, primary SS, RA, SSc or MCTD and from patients with secondary SS than from healthy controls (Fig. 2a,b and Table 1). OD₄₅₀ values of IgG antibodies to EBNA-1(C) in sera from the patients with primary SS, SSc or secondary SS associated with RA were significantly higher than in sera from healthy controls sera (Fig. 2c,d and Table 1). There were no significantly elevated ELISA levels of IgG antibodies against VZV and human B19 parvovirus in sera from patients with CTD (data not shown).

Table 1.

Median values of ELISA OD₄₅₀ readings for IgG antibodies to EBNA-2(N) and EBNA-1(C) in patients with different connective tissue disease

	EBNA-2(N)		EBNA-1(C)
Diseases of patients groups (no. of specimens)	Median	Mann–Whitney test1	Median	Mann–Whitney test1
Systemic lupus erythematosus (SLE) (100)	0·124	P < 0·0001	0·424	n.s.
Primary Sjögrens's syndrome (primary SS) (73)	0·124	P < 0·0001	0·459	P = 0·0007
Rheumatoid arthritis (RA) (59)	0·118	P < 0·0001	0·361	n.s.
Systemic sclerosis (SSc) (13)	0·146	P < 0·008	0·541	P < 0·03
Mixed connective tissue disease (MCTD) (9)	0·123	P < 0·03	0·517	n.s.
Healthy control (120)	0·0525		0·381
Secondary SS
Secondary SS associated with SLE (19)	0·151	P = 0·0003	0·509	n.s.
Secondary SS associated RA (23)	0·197	P < 0·0001	0·499	P = 0·0009
Secondary SS associated with SSc (9)	0·104	P < 0·03	0·308	n.s.
Healthy control	0·065		0·376

¹P-values were calculated by comparing the medians of patient groups to that of normal healthy controls.²n.s.: non-significant.³SS associated with SLE, RA or SSc.

IgG hypergammaglobulinaemia in sera from the patients with SS, SLE, RA, SSc, PM/DM or MCTD

To examine whether IgG antibodies against EBNA-1(C), EBNA-2(N) and EBNA-2(C) are elevated specifically in patients with autoimmune CTD, we measured concentrations of immunoglobulin proteins. Concentrations of IgG antibodies in sera from patients with CTD were significantly higher than those from the healthy control group (Fig. 3a). The differences among the patient groups were significant when analysed by Mann–Whitney test; specifically, SLE, primary SS, RA and SSc (P < 0·0005), PM/DM (P < 0·004) and MCTD (P < 0·002). Correlation coefficients between the ELISA OD measurements of EBNA-2(N) IgG and IgG concentrations in sera were calculated by Spearman's method and Pearson's method. Pearson's correlation coefficients indicated significant correlations between the ELISA OD₄₅₀ measurements of EBNA-2(N) IgG and IgG concentrations in sera from patients with primary SS (r = 0·370, n = 74), SLE (r = 0·308, n = 102) and RA (r = 0·480, n = 58). Spearman's rank correlation coefficients also indicate similar significant correlations in sera from patients with primary SS (r = 0·310, n = 74), SLE (r = 0·330, n = 102) and RA (r = 0·375, n = 58). Correlation coefficients between EBNA-1(C) and IgG concentrations in patients sera were significant in RA sera by the methods of Pearson (r = 0·490, n = 58) and Spearman (r_s = 0·430), but not in primary SS and SLE sera.

Fig. 3.

Immunoglobulin concentrations in sera from patients with SLE, SS, RA, SSc or MCTD. (a) IgG concentration. (b) IgA concentration. (c) IgM concentration. Median values are indicated by the horizontal bars in the scattergrams. Median values of immunoglobulin concentrations are indicated by the horizontal bars in the scattergrams.

Concentrations of IgA antibodies in the sera from patients with SLE, primary SS, RA and SSc, but not with PM/DM, were also significantly higher (P < 0·03) than in the normal controls (Fig. 3b). Levels of IgM in sera from patients with CTD were comparable to healthy controls (Fig. 3b).

Specific levels of IgG antibodies to EBNA-2(N) are elevated in sera from patients with primary SS, SLE or RA

To examine whether levels of IgG antibodies to EBNA-2(N) are elevated specifically, OD₄₅₀ readings of IgG antibodies to EBNA-2(N) were divided by total IgG concentrations (mg/dl). The normalized ELISA values are referred to as specific OD₄₅₀. The medians of the specific OD₄₅₀ values to EBNA-2(N) in sera from patients’ groups were 0·084 in patients with SLE, 0·070 in patients with primary SS, 0·070 in patients with RA and 0·042 in sera from healthy controls (Fig. 4a and Table 2). The differences between the patients’ sera and control sera were all significant (Table 2). Medians of the specific OD₄₅₀ readings of IgG antibodies to EBNA-2(N) in sera from patients with SSc, PM/DM or MCTD were not statistically different from the median of the healthy control group (Table 2). The median of the specific OD₄₅₀ of IgG antibodies to EBNA-2(N) of sera from patients with secondary SS that is complicated with SLE was 0·088 and that from patients with secondary SS that is complicated with RA was 0·075, both of which are higher than the 0·050 value for healthy controls (Fig. 4b and Table 2). These results indicate that serum levels of IgG antibodies to EBNA-2(N), whether or not they are calibrated relative to whole IgG concentration, are significantly higher in sera from patients with SLE, primary SS or RA, as well as those with secondary SS complicated with SLE or RA, when compared to levels in sera from healthy controls.

Fig. 4.

Specific levels of IgG antibodies to EBNA-2(N) and to EBNA-1(C) as defined by ELISA OD₄₅₀ normalized to whole IgG concentration. (a) Specific OD₄₅₀ readings of IgG antibodies to EBNA-2(N) in sera from patients with SLE, SS, RA, SSc or MCTD. (b) Specific OD₄₅₀ readings of IgG antibodies to EBNA-2(N) in sera from patients with secondary SS. (c) Specific OD₄₅₀ readings of IgG antibodies to EBNA-1(C) in sera from patients with SLE, SS, RA, SSc or MCTD. Median values are indicated by the horizontal bars in the scattergrams.

Table 2.

Specific serum levels of IgG antibodies to EBNA-2(N) in sera from patients with different connective tissue diseases.¹

		EBNA-2
Diseases of patients groups (no. of specimens)	Median	Mann–Whitney test2
Systemic lupus erythematosus (102)	0·084	P < 0·0001
Primary Sjögrens's syndrome (74)	0·070	P = 0·0005
Rheumatoid arthritis (59)	0·070	P = 0·002
Systemic sclerosis (13)	0·056	n.s.3
Polyomyositis/dermatomyositis (7)	0·042	n.s.
Mixed connective tissue disease (9)	0·070	n.s.
Secondary SS associated with SLE (19)	0·088	P = 0·004
Secondary SS associated RA (23)	0·075	P = 0·002
Secondary SS associated with SSc (9)	0·050	n.s.
Healthy control	0·042

¹Specific serum levels of IgG antibodies to EBNA-2 were calculated by dividing ELISA OD₄₅₀ readings by whole IgG protein concentrations.

²P-values were calculated by comparing the medians of patient groups to that of normal healthy controls.

³Not significant (n.s.).

The median of the specific OD₄₅₀ of IgG antibodies to EBNA-1(C) in sera from patients with the systemic CTDs were not higher than that of healthy control sera (Fig. 4c).

Amino acid sequence homologies of the amino-terminal region of EBNA-2 with proline-rich cell proteins

Sequence homology searches for cellular proteins with homologies to the EBNA-2 amino-terminal proline-rich region (aa 51–116, -FVGENTGVPP PLPPPPPPPP PPPPPPPPPP PPPPPPPPPP PSPPPPPPPP PPPQRRDAWT QEPSPL-) have shown that there are many cell proteins with considerable homology to this region. Twelve cell proteins with the highest homologies to the EBNA-2 region are: enabled protein homologue (Z score = 105, E-value = 3e-23); neural Wiskott–Aldrich syndrome protein (WASP) (Z score = 105, E-value = 4e-23); WASP family member 2 (102, 4e-22); WASP (101, 8e-22); diaphanous protein homologue 1 (diaphanous-related formin 1) (Z score = 100, E-values = 1e-21); hypothetical acrosin-like protease (Z score = 99, E-value = 3e-21); myeloid/lymphoid or mixed lineage leukaemia protein 4 (Z score = 98, E-values = 8e-21); piccolo protein (aczonin) (Z score = 90, E-values = 2e-18); splicing factor 1 (Z score = 88, E-value = 5e-18); bromodomain-containing protein 4 (Z score = 87, E-value = 2e-17); protocadherin 15 precursor (Z score = 83, E-value = 2e-16); and Huntington's disease protein (Z score = 81, E-values = 8e-16). Similar homology searches using the programs genomes&sol;human, blast 2·2.9, blosum 62 and the database ‘Build protein’ (score ≥ 80 bits, E-value ≤6e-17) added more cellular proteins with the most significant alignments as follows: zinc finger homeodomain 4 (Z score = 91, E-value = 2e-19); formin-like 2 (Z score = 88, E-value = 1e-18); similar to zinc finger protein 469 (predicted; KIAA 1856 protein) (Z score = 85, E-value = 1e-17); KIAA1205 (Z score = 84, E-value = 1e-17) (predicted); formin-like 1 (leucocyte formin) (Z score = 84, E-value = 2e-17); KIAA0339 (Z score = 82, E-value = 6e-17) (predicted); and Wiscott–Aldrich syndrome gene × like protein (Z score = 82, E-value = 6e-17) in addition to myeloid/lymphoid or mixed lineage leukaemia 4 (note that Z score = 86 and E-value 3e-18 when the programs genomes&sol;human, blosum 62 and the database ‘Build protein’ are used) and the other above-shown cellular proteins. This significant amino acid sequence homology is because the EBNA-2 amino-terminal region contains 2 polyproline tracts at aa 63–88 and 90–100.

Association of pulmonary involvements with a subgroup of SS patients who have high EBNA-2 (aa 1–116) IgG and low EBNA-1 (aa 451–641) IgG antibody levels

We analysed whether clinical manifestations or laboratory findings could be assigned to subgroups of SS patients with relatively high IgG antibodies to EBNA-2. Data for clinical manifestations or laboratory findings of the subgroup of patients with primary and secondary SS who had EBNA-2(N) to EBNA-1(C) IgG ratios ≥ 0·8 (referred to as the high EBNA-2/EBNA-1 IgG subgroup) and those that had EBNA-2(N) to EBNA-1 IgG ratios <0·8 (referred to as the medium–low EBNA-2/EBNA-1 IgG subgroup) were compared, because it has been reported that the high EBNA-2 to EBNA-1 IgG ratios are associated with recent primary infection of EBV [43–45], probably with either reactivation of latent EBV infection or superinfection [45], and may be of diagnostic importance in patients with immunodeficiencies [46]. The incidence of restrictive change in pulmonary function test was 44% (7/16) in the high EBNA-2/EBNA-1 IgG subgroup and 13% (15/118) in the medium–low EBNA-2/EBNA-1 IgG subgroup. The difference was significant by Fisher's exact probability test (P = 0·006). The incidence of decreased diffusion capacity of carbon monoxide was 44% (8/18) in the high EBNA-2/EBNA-1 IgG subgroup and 15% (18/118) in the low EBNA-2/EBNA-1 IgG subgroup. The difference was significant by Fisher's exact probability test (P = 0·007). There was no correlation with either clinical or diagnostic laboratory test data between a high EBNA-2(N)/EBNA-1(C) IgG subgroup and a low–medium EBNA-2(N)/EBNA-1(C) IgG subgroup in patients with either SLE or RA (data not shown).

Discussion

EBNA-2 is an early transcription factor that is essential for EBV-induced cell immortalization of human B lymphocytes and EBNA-1 is necessary for the maintenance of EBV plasmid DNA and the EBV latent infection [47–49]. The more frequent presence of antibodies to EBNA-2 in sera from RA patients than in healthy control sera [50,51] has been documented, but the characteristics and significance of the EBNA-2 antibodies are unknown. The elevated levels of EBNA-2 IgG may not be a cause, but a result, of its antigenic cross-reactivity with cellular or other proteins of infectious agents. The results in this study will be useful for serological characterization and diagnosis, and future studies on the mechanism of the CTDs. It is also noteworthy that reactivation of B cells in patients with the CTDs can be polyclonal and that polyclonal B cell activation occurs in the course of infectious mononucleosis caused by primary EBV infection [52]. The low background ELISA in this report has facilitated the quantification of specific levels of EBNA-2 and EBNA-1 IgG antibodies in sera from various patients.

First, our previous observation of elevated levels of IgG antibodies to EBNA-2 (14 kDa) in sera from SS patients [26] has been confirmed. Furthermore, our results show that elevated ELISA OD₄₅₀ readings of IgG antibodies to the EBNA-2(N) are associated with SLE, RA or SSc and with secondary SS. These results have extended early reports that described a high incidence of anti-EBNA-2 antibodies in rheumatoid sera [51].

Secondly, in most earlier analyses of sera from RA patients EBNA-1 proteins containing either the Gly/Ala repeating region (aa 90–328), a major cross-reactive epitope region [53–55], or peptides derived from the Gly/Ala repeat sequence were used as antigen [23,56]. Our data show the statistical association of OD₄₅₀ readings of EBNA-1(C) IgG antibody with sera from primary SS patients (median = 0·459, P < 0·0007) but not RA patients, when compared to healthy controls (median = 0·369). The results have confirmed our previous observation of elevated levels of IgG antibodies to EBNA-1 (K2) in sera from patients with SS but not RA [26].

Thirdly, this is the first report in which ELISA readings to EBV antigens have been normalized to total IgG protein concentrations in patients’ sera. This analysis is important because it specifically quantifies the serum levels of antibodies to EBV antigens and has indicated unambiguously that the rise in antibody levels to EBNA-2 is specific.

Fourthly, our results show that the median levels of the specific OD₄₅₀s of IgG antibodies to EBNA-1(C) in sera from patients with SLE, primary SS or RA were no higher than in sera from healthy controls. Further clarification of an association of EBNA-1 IgG antibodies with autoimmune CTD is necessary.

Fifthly, a possible aetiology of EBV for SLE has been suggested by an increased prevalence of EBV infection in young patients and previous EBV exposure in adults [57,58]. A high proportion of the SLE sera were found to contain antibodies to EBV-encoded EA antigen [59]. In addition, a high degree of homology between the EBNA-2 C-terminal sequence GRGKGKSRDKQRKPGGPWRP (aa 354–373) and the antigenic C-terminal domain GRGRGRGRGRGRGRGGPRR of the SmD1 ribonucleoprotein (aa 101–119), a target of autoantibodies in a portion of SLE patients, has been reported [60]. The elevated levels of specific IgG antibodies to the EBNA-2(N) in sera from patients with SLE in this study support the hypothesis that EBV is involved in the pathogenicity of SLE.

Sixthly, it is intriguing that antibodies from a SLE patient reacted with similar proline-rich sequences in Sm B/B′, U1-RNP C and U1-RNP A peptides [10]. In addition, antibodies to proline-rich native type II collagen have been found in RA and SLE [61,62]. The EBNA-2 aa 1–116 sequence is also proline-rich [63], which is intriguing from the viewpoint of the hypotheses of molecular mimicry and epitope spread that involve diversification of autoantibodies [64,65]. Our data indicate that there are a number of proline-rich cell proteins that have considerable sequence homology with EBNA-2 aa 1–116. Whether such proline-rich cell proteins are associated with the autoimmune disease should be addressed in future studies.

Finally, pulmonary involvement in primary SS has been reported from several laboratories [41,66] and probably indicates the presence of an inflammation accompanied by infiltration of lymphocytes in interstitial tissue of the lung [67]. EBNA-2 IgG antibodies decrease or disappear after infectious mononucleosis caused by EBV primary infection; in contrast EBNA-1 IgG antibodies remain stable long after diagnosis of infectious mononucleosis [43–45]. This study suggests that a subgroup of SS patients with high ratios of EBNA-2(N) : EBNA-1(C) IgG antibodies are significantly more prone to pulmonary involvement when compared to the other SS patients. One interpretation of these data is that SS patients with the high EBNA-2(N) : EBNA-1(C) IgG antibody ratios have experienced reactivations of latent EBV infection. This interpretation is supported by a report that a large fraction of T cells infiltrating affected joints from a patient with chronic RA recognizes two EBV transactivators (BZLF1 and BMLF1) [68]. An alternative interpretation is that cell proteins are aberrantly exposed because of local inflammatory responses that lead subsequently to tissue destruction during pulmonary involvement. The resulting IgG antibodies to polyproline or proline-rich sequences of self-antigens will cross-react immunologically with the EBNA-2 amino-terminal aa 1–116. These two interpretations are not mutually exclusive.

The molecular mimicry and epitope shift mechanisms for autoimmune diseases have been demonstrated experimentally in animal models in which synthetic peptides with similar amino acid sequences were used as antigens [8,69], as well as in patients with RA or multiple sclerosis [7,70]. The genetic associations that have been reported between HLA-DR alleles and antibodies to type II collagen in RA patients implicate autoimmunity to type II collagen as a major component in the multifactorial pathogenesis of RA [71]. Thus, the findings in this report that IgG antibodies reacting with the proline-rich amino-terminal region of EBNA-2(N) are associated statistically significantly with SLE, primary SS, RA and secondary SS suggest that the molecular mimicry between the amino-terminal region of EBNA-2 and proline-rich cellular proteins, probably in combination with the epitope shift mechanism, may be related to pathogenesis in a subpopulation of patients with CTD.