Antibodies to citrullinated α-enolase peptide 1 are specific for rheumatoid arthritis and cross-react with bacterial enolase

Authors


Abstract

Objective

To map the antibody response to human citrullinated α-enolase, a candidate autoantigen in rheumatoid arthritis (RA), and to examine cross-reactivity with bacterial enolase.

Methods

Serum samples obtained from patients with RA, disease control subjects, and healthy control subjects were tested by enzyme-linked immunosorbent assay (ELISA) for reactivity with citrullinated α-enolase peptides. Antibodies specific for the immunodominant epitope were raised in rabbits or were purified from RA sera. Cross-reactivity with other citrullinated epitopes was investigated by inhibition ELISAs, and cross-reactivity with bacterial enolase was investigated by immunoblotting.

Results

An immunodominant peptide, citrullinated α-enolase peptide 1, was identified. Antibodies to this epitope were observed in 37–62% of sera obtained from patients with RA, 3% of sera obtained from disease control subjects, and 2% of sera obtained from healthy control subjects. Binding was inhibited with homologous peptide but not with the arginine-containing control peptide or with 4 citrullinated peptides from elsewhere on the molecule, indicating that antibody binding was dependent on both citrulline and flanking amino acids. The immunodominant peptide showed 82% homology with enolase from Porphyromonas gingivalis, and the levels of antibodies to citrullinated α-enolase peptide 1 correlated with the levels of antibodies to the bacterial peptide (r2 = 0.803, P < 0.0001). Affinity-purified antibodies to the human peptide cross-reacted with citrullinated recombinant P gingivalis enolase.

Conclusion

We have identified an immunodominant epitope in citrullinated α-enolase, to which antibodies are specific for RA. Our data on sequence similarity and cross-reactivity with bacterial enolase may indicate a role for bacterial infection, particularly with P gingivalis, in priming autoimmunity in a subset of patients with RA.

Rheumatoid arthritis (RA) is a chronic inflammatory joint disorder that is considered to be autoimmune, although the autoantigens that trigger and sustain the immune response remain unknown. Over the last 10 years, investigations have shown that an essential feature of many autoantigens in RA is the posttranslational conversion of peptidyl arginine to peptidyl citrulline. Anti–citrullinated protein antibodies (ACPAs) are highly specific (98%) and sensitive (up to 80%) for RA (1, 2), making citrullinated proteins strong candidates for driving the autoimmune response in this disease (for review, see ref.3). Kuhn et al, for example, recently demonstrated that administration of anti–citrullinated fibrinogen antibodies enhanced disease severity in an experimental mouse model of arthritis (4). In the majority of patients with RA, the presence of ACPAs antedates disease onset (5, 6), and ACPA-positive patients present with a more severe and erosive form of arthritis (7–16). A commercial enzyme-linked immunosorbent assay (ELISA) based on cyclic citrullinated peptides (CCPs) detects ACPAs and is now routinely used in the diagnosis of RA.

Citrullination has a physiologic role in the generation of structural tissue such as skin, hair follicles, and the myelin sheaths of nerve fibers. In addition, the accumulation of citrullinated proteins has been described at sites of inflammation, including the joints of patients with all forms of arthritis (17), the brains of patients with multiple sclerosis (18) or Alzheimer's disease (19), and in the muscle fibers of patients with myositis (20). Hence, citrullinated proteins are present in the setting of both health and disease, while tolerance to citrullinated proteins appears to be selectively lost in patients with RA. Thus, in RA, it is the antibody response rather than the expression of antigen that is specific to the disease.

Genes and the environment interact in the development of this complex and heterogeneous disorder. Recent studies have demonstrated that the HLA–DRB1 shared epitope (SE) alleles, the best known genetic risk factor for RA, are associated with only anti-CCP antibody–positive RA, not anti-CCP antibody–negative RA, indicating that HLA SE alleles may be a specific risk factor for the production of ACPAs rather than the for RA itself (21, 22). Furthermore, a strong gene–environment interaction between HLA SE alleles and cigarette smoking is present in anti-CCP antibody–positive patients but not in anti-CCP antibody–negative patients (21, 23). Infectious agents, both bacterial and viral, have also been proposed as potential environmental stimuli (24–27), although to date, no single organism has survived as a compelling candidate for the etiology of the disease. Given that citrullinated proteins are target autoantigens in RA, the pathogen Porphyromonas gingivalis, which expresses the citrullinating enzyme peptidyl arginine deiminase (PAD) (28), could be an environmental trigger of RA in a manner similar to that proposed for smoking (21).

It is not clear which proteins harbor the epitopes targeted by ACPAs, although several candidates, including citrullinated fibrinogen (29), vimentin (30), and type II collagen (31), have been suggested, and we recently identified citrullinated α-enolase as another potential autoantigen (32). Alpha-enolase is abundantly expressed in the rheumatoid joint, antibodies targeting only the citrullinated form of the protein are specific for the disease, and our group recently demonstrated the presence in vivo of citrullinated α-enolase in synovial fluid from patients with RA (33). The molecule is also highly conserved throughout eukaryotes and prokaryotes and could therefore be a candidate for molecular mimicry between bacterial and host proteins (34). In the present study, we mapped the epitope of this anti–citrullinated α-enolase antibody response, using several citrullinated α-enolase peptides (CEPs), with the aim of examining cross-reactivity with bacterial enolase, which could prime autoimmunity in a subset of patients with RA.

PATIENTS AND METHODS

Serum samples from patients and control subjects

Serum samples were obtained, with informed consent and ethics approval from the Regional Research Ethics committee, from 102 consecutive patients with RA who were attending the Rheumatology Clinic at Charing Cross Hospital, London. The disease control group comprised 110 patients with other rheumatic diseases, including systemic lupus erythematosus (n = 32), Sjögren's syndrome (n = 31), Behçet's syndrome (n = 18), psoriatic arthritis (n = 5), and miscellaneous other rheumatic diseases (n = 24). Ninety-two control serum samples were obtained from healthy volunteers. Serum samples from 20 patients with spondylarthritides who were attending the Rheumatology Clinic at the Karolinska University Hospital in Stockholm, Sweden, were also collected, with informed consent and local ethics approval. In addition, serum samples were collected from an independent US cohort of 81 patients with RA obtained at baseline in clinical trials conducted by the Rheumatoid Arthritis Investigational Network (RAIN; coordinating center, University of Nebraska, Omaha) and from 82 age- and sex-matched healthy volunteers. All US samples were obtained with ethics approval from the local institutional review board. All patients with RA met the American College of Rheumatology (formerly, the American Rheumatism Association) 1987 revised criteria for the classification of RA (35).

Peptides

Fifteen cyclic 15–23-mer peptides were synthesized at Cambridge Research Biochemicals (Billingham, Cleveland, UK). The peptide sequences corresponded to amino acid sequences in human α-enolase (Swiss-Prot accession no. P06733) or P gingivalis enolase (Swiss-Prot accession no. AAQ66821), with the addition of cysteine residues at the amino and carboxy termini and the exchange of arginine for citrulline residues at certain positions (Table 1).

Table 1. Alpha-enolase peptide sequences and the antibody response to these peptides in patients with RA and controls*
Peptide nameSequenceRA (n = 102)Disease controls (n = 110)Healthy controls (n = 92)
  • *

    Values are the percent positive, using an optical density at 450 nm cutoff value of 0.1. RA = rheumatoid arthritis; X = citrulline; R = arginine; ND = not done.

  • A total of 130 disease controls were screened for reactivity with human and Porphyromonas gingivalis citrullinated α-enolase peptide 1 (CEP-1).

1A (CEP-1)ckiha-X-eifds-X-gnptvec3732
1A P gingivalisckiig-X-eilds-X-gnptvec341ND
1Bckiha-R-eifds-X-gnptvec40NDND
1Cckiha-X-eifds-R-gnptvec20NDND
1Dckiha-R-eifds-R-gnptvec523
2cvdlftskglf-X-aavpsgc1002
3cyealel-X-dndkt-X-ymgkgvskc1520
4cekgvply-X-hiadlagnsec1622
5cnf-X-eam-X-igaevyhnlknc1143
6cf-X-sgkydldfkspddps-X-yc1250
7cangwgvmvsh-X-sgetedtc422
8cakynqll-X-ieeelgskac740
9ckfag-X-nf-X-nplakc1511
10ciqvvgddltvtnpk-X-iakac6ND1
11cnpk-X-iakavnekscncllc18ND0

ELISAs

Ninety-six–well plates (MaxiSorp; Nunc, Roskilde, Denmark) were coated with α-enolase peptides at 10 μg/ml (diluted in a 50-mM carbonate buffer, pH 9.6) or with carbonate buffer alone and incubated overnight at 4°C. Wells were washed with phosphate buffered saline (PBS)–Tween (0.05%) and blocked with 2% bovine serum albumin (BSA) diluted in PBS for 3 hours at room temperature. Sera were diluted 1:50 in radioimmunoassay (RIA) buffer (10 mM Tris, 1% BSA, 350 mM NaCl, 1% Triton X-100, 0.5% sodium deoxycholate, 0.1% sodium dodecyl sulfate) supplemented with 10% fetal calf serum (FCS), added in duplicates, and incubated for 1.5 hours at room temperature. Plates were washed as described above and incubated with peroxidase-conjugated mouse anti-human IgG (Hybridoma Reagent Laboratory, Baltimore, MD) (diluted 1:1,000 in RIA buffer, 10% FCS) for 1 hour at room temperature. After a final wash (PBS–Tween, 0.05%), bound antibodies were detected with tetramethylbenzidine substrate (KPL, Gaithersburg, MD). The reaction was stopped by the addition of 1M H2SO4, and absorbance was measured at 450 nm in a Multiscan Ascent microplate reader (Thermolab Systems, Franklin, MA). A control serum was included on all plates to correct for plate-to-plate variation. The value for background optical density at 450 nm (OD450) (wells coated with carbonate buffer alone) was subtracted from the peptide OD450 value. OD values above 0.1 were considered to be positive. ELISA of the serum samples obtained from the US cohort was performed in a similar manner but with 2% BSA in carbonate buffer as background and with an OD450 cutoff of 0.2 for positive samples. Anti-CCP antibody status was analyzed using the CCP2 kit (Eurodiagnostica, Malmo, Sweden), according to the manufacturer's instructions.

Inhibition assays

Inhibition experiments were performed in liquid phase using serum samples obtained from 6 anti–CEP-1–positive patients with RA or in solid phase using serum samples obtained from 9 double-positive (anti–CEP-1 positive and CCP positive) patients with RA. Sera were preincubated for 2 hours in RIA buffer containing increasing concentrations (0, 1, 10, and 100 μg/ml) of CEP-1, control peptide 1D, peptide 11, 4, or 9 (representing the second, third, and fourth most reactive peptides), or peptide 7 (representing the peptide with the lowest degree of reactivity). Alternatively, serum specimens were preabsorbed on plates coated with CEP-1 or carbonate buffer alone or were preabsorbed on commercial CCP2 plates. The liquid phase mixtures were centrifuged at 16,200g for 15 minutes before the supernatants were transferred to CEP-1–coated plates and assessed as described above for ELISAs. Inhibition of binding to CEP-1, for serum samples preabsorbed to CCP, was calculated in relation to the maximum inhibition, which by definition was set to 100% for samples preincubated with CEP-1.

Generation of anti–CEP-1 antibodies

A rabbit polyclonal anti–CEP-1 antibody was generated at Cambridge Research Biochemicals. Briefly, 2 rabbits were immunized subcutaneously every 2 weeks, for 10 weeks, with 200 μg keyhole limpet hemocyanin–conjugated peptide 1A in Freund's incomplete adjuvant per boost. Blood was collected 7 days after each injection, and sera were analyzed for the presence of anti–CEP-1 antibodies. When antibody titers reached significant levels, the animals were killed and their blood was harvested. The crude antisera were depleted of cross-reactive antibodies by chromatography on a thiopropyl–Sepharose column conjugated to peptide 1D. The unbound fraction was affinity-purified on a second thiopropyl–Sepharose column conjugated to peptide 1A. Anti–CEP-1–specific antibodies were eluted and further depleted of nonspecific antibodies by 3 subsequent passages through the depleting column. Human anti–CEP-1 antibodies from a patient with RA were purified with high titers of anti–CEP-1 antibodies. The serum was passed through a thiopropyl–Sepharose column conjugated to peptide 1A. Bound antibodies were eluted, after repeated PBS washes, using 3M GuHCl, and dialyzed against PBS. Purified rabbit anti–CEP-1 antibodies (0.39 mg/ml) and human anti–CEP-1 antibodies (0.6 μg/ml) were stored at −20°C until used further.

ELISA to assess CEP-1 specificity of affinity-purified anti–CEP-1 antibodies

Affinity-purified anti–CEP-1 antibodies (rabbit and human) were tested for their anti–CEP-1 specificity in serial dilutions (starting at 1:50 for the rabbit anti–CEP-1 antibody and at 1:2 for the human anti–CEP-1 antibody) on plates coated with peptide 1A or peptide 1D and assayed as described above for ELISAs.

Cloning and expression of human and bacterial enolase

Full-length human enolase and P gingivalis enolase were amplified by polymerase chain reaction (PCR), from vitamin D3–differentiated HL-60 cells and P gingivalis strain W83 (no. BAA-308D-5; American Type Culture Collection, Rockville, MD), respectively. PCR products were ligated between Bam HI and Xho I restriction sites in a pGEX 6P3 expression vector (GE Healthcare, Bucks, UK), 3′ to the glutathione transferase (GST) coding site. GST–enolase protein expression was induced by isopropyl thiogalactose in the protease-deficient BL21 strain of Escherichia coli. Protein was purified using glutathione–Sepharose 4B (GE Healthcare), and the GST moiety was cleaved using PreScission Protease (GE Healthcare). Purified enolase, as determined by Coomassie staining and tandem mass spectometry analysis, was dialyzed against PBS and stored at −20°C until used further.

In vitro deimination

Recombinant human enolase and bacterial enolase were diluted to a concentration of 0.3 mg/ml in PAD buffer (0.1M Tris HCl, pH 7.6, 10 mM CaCl2, 5 mM dithiothreitol) and incubated with rabbit skeletal PAD (Sigma, St. Louis, MO) at a concentration of 7 units/mg protein, for 3 hours at 50°C. Citrullination was terminated by the addition of 20 mM EDTA. Control proteins were treated similarly, apart from the addition of PAD. All samples were stored at −20°C until used further.

Silver staining and immunoblotting

Recombinant human enolase and P gingivalis enolase were electrophoresed on 4–12% NuPAGE Bis-Tris gels (Invitrogen, Paisley, UK) before silver staining, using a standard protocol, or before transfer to nitrocellulose membranes for immunoblotting. Briefly, membranes were blocked with 5% nonfat milk and incubated with rabbit anti–CEP-1 antibody (diluted 1:25) or human anti–CEP-1 antibody, diluted 1:2. Proteins were detected using peroxidase-conjugated secondary antibody (goat anti-rabbit IgG, diluted 1:2,000; Dako, Glostrup, Denmark) or mouse anti-human IgG, diluted 1:500 (Hybridoma Reagent Laboratory, Baltimore, MD). Membranes were developed using the enhanced chemiluminescence technique (Amersham Biosciences, Little Chalfont, UK). Citrullinated proteins were detected using the Anti-Citrulline (Modified) Detection Kit (Upstate Biotechnology, Lake Placid, NY), in accordance with the manufacturer's instructions.

Statistical analysis

All statistical analyses were performed using the Mann-Whitney U test for independent groups.

RESULTS

Detection of a disease-specific antibody response to CEPs in patients with RA.

Eleven citrullinated peptides (peptides 1–11) (Table 1), covering 15 of the 17 arginine residues within α-enolase, were selected on the basis of containing >1 potential citrulline residue within a 20–amino acid sequence, or by the demonstration (by mass spectrometry) in our previous study (32) that arginines had been deiminated to citrulline in vitro (32). Each peptide synthesized was tested for reactivity in 102 patients with RA, 110 disease control subjects, and 92 healthy control subjects. The results showed that 64% of patients with RA and 15% of the control subjects had IgG antibodies to 1 or several CEPs. The pattern of reactivity in patients with RA varied, with the majority of patients having an antibody response to multiple CEPs. In contrast, the antibody response in the control subjects was mainly restricted to one of the peptides, and the IgG antibody levels were significantly lower than those observed in patients with RA (Figure 1A).

Figure 1.

A, IgG response to citrullinated α-enolase peptides in patients with rheumatoid arthritis (RA; n = 102), disease controls (n = 110), and healthy controls (n = 92). B, Antibody response to the immunodominant peptides 1A, 1B, 1C, and 1D in patients with RA. The response to peptide 1A is dependent on the presence of citrulline, as demonstrated by the low antibody levels to the arginine-containing control peptide 1D. The second citrulline residue, rather than the first, is more important for antibody binding (compare peptides 1B and 1C). IgG was measured by enzyme-linked immunosorbent assay, and data are presented as the peptide optical density at 450 nm (OD450) value with the background (carbonate buffer) OD450 value subtracted. Each dot represents the reactivity of 1 serum sample with the indicated peptide. Bars indicate the mean. ∗∗ = P < 0.01; ∗∗∗ = P < 0.0001. ND = not done.

Identification of an immunodominant B cell epitope within citrullinated α-enolase.

Peptide 1 (1A) was the immunodominant peptide, which reacted with 37% of the RA serum samples compared with 2% of the healthy control samples and 3% of the disease control samples (Table 1). To map the antibody epitope further and to investigate the citrulline dependence of this antibody response, another 3 peptides (peptides 1B, 1C, and 1D) were synthesized (Table 1). The low reactivity of control peptide 1D, which does not contain any citrulline residues, confirms the importance of citrulline in this epitope (Figure 1B). The proportion of patients with RA whose sera reacted with peptide 1B (40%) was similar to the proportion whose sera reacted with peptide 1A but was higher than the proportion whose sera reacted with peptide 1C (20%). This, together with the higher antibody levels to peptides 1A and 1B compared with peptide 1C (Figure 1B), indicates that it is the second citrulline residue, rather than the first, that is more important for antibody recognition. Peptide 1A, containing both citrullinated residues, was chosen for further study and is referred to as CEP-1.

Anti-CCP antibodies in the same serum samples gave a diagnostic sensitivity of 71%, with a specificity of 98%. Seven (23%) of the anti-CCP antibody–negative patients with RA had positive results on the anti–CEP-1 ELISA (data not shown). Hence, combining the anti-CCP with the anti–CEP-1 ELISA results increased the overall sensitivity of ACPAs in this cohort to 78%.

Confirmation of the diagnostic sensitivity and specificity of the anti–CEP-1 ELISA in an independent cohort of patients with RA.

To confirm that the high levels of antibodies to CEP-1 were not a peculiarity of the patients in our study, we used the anti–CEP-1 antibody ELISA to test 81 patients with RA and 82 healthy control subjects from a US cohort. To compare the data with the UK cohort, the OD value for the ninety-eighth percentile of control subjects was used as the cutoff point for positive samples. The sensitivity was increased to 62% in the US RA population, with a specificity of 98% (Figure 2). Disease controls were not examined in this part of the study.

Figure 2.

Anti–citrullinated α-enolase peptide 1 antibodies in the US cohort. Data were generated by enzyme-linked immunosorbent assay and are presented as the peptide optical density at 450 nm (OD450) value with the background OD450 value subtracted. Broken line indicates the cutoff value for positive samples, based on the OD value for the ninety-eighth percentile of control subjects. Bar indicates the mean. RA = rheumatoid arthritis.

Anti–CEP-1–specific as well as CCP cross-reactive antibodies in patients with RA.

The epitope specificity of the anti–CEP-1 antibody response was tested in 2 separate inhibition assays. In the first experiment, inhibition of binding to CEP-1 was evaluated in sera from 6 anti–CEP-1–positive patients with RA, using liquid phase. There was a dose-dependent inhibition by the homologous peptide, while there was no inhibition by the arginine-containing control peptide 1D or by the nonreactive citrullinated α-enolase peptide 7. Also, there was no inhibition when using citrullinated α-enolase peptides representing other reactive epitopes on the molecule, i.e., peptides 4, 9, and 11 (Figures 3A–F). In the second experiment, in which we examined cross-reactivity to CCP, it was necessary to use the commercially available CCP2 plates in a solid phase assay, because the sequences of the peptides used in the assay have not been published. In 9 double-positive (anti–CEP-1 positive and anti–CCP positive) RA samples, preabsorption on CCP plates showed inhibition of anti–CEP-1 binding, varying from 0% in 2 of the sera and increasing to a maximum of 53% in 1 sample, suggesting variable cross-reactivity of the antibodies (Figure 3G).

Figure 3.

A–F, No cross-reaction of anti–citrullinated α-enolase peptide 1 (anti–CEP-1) antibodies with other citrullinated α-enolase epitopes. A dose-dependent decrease in binding to CEP-1–coated plates was seen in 6 anti–CEP-1–positive serum samples from patients with rheumatoid arthritis (RA), preincubated with increasing concentrations (0, 1, 10, and 100 μg/ml) of CEP-1 (A), while preincubation with the control peptide 1D (B), peptide 4 (C), peptide 7 (D), peptide 9 (E), or peptide 11 (F) did not result in decreased antibody binding. G, Differing degrees of cross-reactivity with cyclic citrullinated peptide (CCP) in individual serum specimens from 9 double-positive (anti–CEP-1 positive and anti-CCP positive) patients with RA following preincubation with CCP (shaded bars), as demonstrated by a reduction in CEP-1 reactivity that ranged from 0% to 53%. Results are expressed as the percentage of maximum inhibition, which by definition was set to 100% for preincubation with CEP-1 (solid bars). OD450 = optical density at 450 nm.

High sequence similarity between human and bacterial CEP-1.

Human enolase and P gingivalis (Swiss-Prot accession no. AAQ66821) enolase were 51% identical at the amino acid level, across the whole protein. However, when comparing the sequence for peptide 1 (amino acids 5–21), the sequence identity increased to 82%, and the 9 amino acids spanning the immunodominant epitope on peptide 1 (amino acids 13–21) were 100% identical.

Antibodies to P gingivalis CEP-1 in patients with RA, and cross-reaction of purified anti–CEP-1 antibodies with bacterial enolase.

We tested the 102 RA patients from the UK cohort for reactivity with the P gingivalis version of CEP-1, and 34% had positive results (Table 1); IgG antibody levels were similar to those observed for the human version of this peptide (data not shown). Sera from the 110 disease control subjects and from 20 patients with spondylarthritides were also tested for reactivity with P gingivalis CEP-1, and only 1 was positive (Table 1). In addition, serum samples from the US cohort were analyzed for anti–P gingivalis CEP-1 reactivity, and 54% of the patients with RA and 2% of the healthy control subjects had positive results. This antibody response correlated strongly (r2 = 0.803, P < 0.0001) with the antibody response directed to the human version of CEP-1 (Figure 4).

Figure 4.

IgG antibodies to the Porphyromonas gingivalis version of CEP-1 in patients with RA (n = 81) and healthy control subjects (n = 82), as measured by enzyme-linked immunosorbent assay. Fifty-four percent of patients with RA were positive for anti–P gingivalis CEP-1 antibodies (circles), compared with 2% of healthy controls (diamonds). The antibody response to P gingivalis CEP-1 (y-axis) correlates strongly with the antibody response to the human version of this peptide (x-axis) (r2 = 0.803, P < 0.0001). See Figure 3 for definitions.

To investigate whether antibodies directed to the immunodominant epitope of human α-enolase also recognized epitopes in P gingivalis enolase, recombinant P gingivalis enolase was in vitro treated with rabbit skeletal PAD or was left untreated, before immunoblotting with affinity-purified anti–CEP-1 antibodies. These antibodies were generated in rabbits or purified from RA sera, and their specificity for CEP-1 was analyzed by ELISA. Serial dilutions of the rabbit polyclonal antibody (Figure 5A) showed a higher CEP-1 specificity than affinity-purified human immunoglobulins (Figure 5B), which also showed reactivity with the arginine-containing control peptide.

Figure 5.

A and B, Enzyme-linked immunosorbent assays. A, CEP-1 specificity for the rabbit polyclonal antibody and B, cross-reactivity with the control peptide for the affinity-purified human antibody were demonstrated. IgG responses to CEP-1 (solid circles) and to the control peptide (open circles) were measured in serial dilutions. CF, Silver staining and immunoblotting. C, Silver staining showed recombinant Porphyromonas gingivalis enolase (lanes 1 and 2) migrating at a slightly lower molecular weight than recombinant human α-enolase (lanes 3 and 4). D, Successful in vitro citrullination by rabbit skeletal peptidyl arginine deiminase (PAD) was demonstrated for P gingivalis enolase (lane 1) and human α-enolase (lane 3). E and F, Noncitrullinated P gingivalis enolase (lane 2) and human α-enolase (lane 4) were run in parallel as negative controls. Affinity-purified rabbit and human anti–CEP-1 antibodies cross-reacted with citrullinated P gingivalis enolase (lane 1), while a lower degree of reactivity was seen with the noncitrullinated form of the protein (lane 2). Citrullinated (lane 3) and noncitrullinated (lane 4) human α-enolase served as controls. Proteins were subjected to sodium dodecyl sulfate–polyacrylamide gel electrophoresis. ∗∗ = P < 0.01. AMC = anti–modified citrulline (see Figure 3 for other definitions).

Silver staining (Figure 5C) demonstrated that P gingivalis enolase migrated at a slightly lower molecular weight than human α-enolase, consistent with the theoretical molecular weights of 45.8 kd for P gingivalis enolase and 47.2 kd for human α-enolase.

Western blots, using the Anti-Citrulline (Modified) Detection Kit, also confirmed that P gingivalis enolase had been successfully citrullinated in vitro by eukaryotic PAD (Figure 5D). The purified anti–CEP-1 antibodies demonstrated cross-reactivity with citrullinated P gingivalis enolase. Both the rabbit (Figure 5E) and the human (Figure 5F) anti–CEP-1 antibody preferentially bound the citrullinated form of the protein but also showed cross-reactivity to the noncitrullinated form. Control blots, using goat anti-rabbit IgG, mouse anti-human IgG, the anti–modifed citrulline antibody on partially modified membranes, or the goat anti-rabbit IgG secondary on modified membranes were all negative (data not shown).

DISCUSSION

In this report, we describe the identification of a dominant B cell epitope in citrullinated α-enolase, CEP-1, which is reactive with 37–62% of RA sera and 2% of healthy control sera. This epitope shows high sequence identity to bacterial enolase, and affinity-purified antibodies, specific for this epitope, react with both human and bacterial forms of the molecule. These data provide a new model for bacterial infection priming the autoimmune response in RA.

RA is diagnosed according to a set of clinical criteria and probably comprises different disease entities, with different pathogenic pathways leading to a similar clinical outcome. For example, 2 subgroups can clearly be distinguished based on the presence or absence of ACPAs (36). It is also possible that there is a range of disease-driving antigens within the ACPA-positive group of patients with RA (37). In a previous study, we characterized citrullinated α-enolase as one such candidate (32). Using immunoblotting of whole protein, we observed that serum samples from 46% of patients with RA reacted with citrullinated α-enolase, of which 7 samples (13%) also recognized the noncitrullinated protein. Sera from 15% of healthy control subjects reacted with both forms of the molecule, and we speculated that this was attributable to reactivity with noncitrullinated epitopes. In the present study, we attempted to identify RA-specific antibodies by testing cyclic peptides consisting of α-enolase sequences surrounding citrulline residues. In doing so, we have identified a dominant epitope, CEP-1, with a diagnostic sensitivity of 37% and a specificity of 98%.

To confirm that this sensitivity and high specificity were not a peculiarity of the sera used at our institution, we also tested 81 RA samples and 82 healthy control samples obtained from a separate unit and found that the sensitivity was higher (62%), with a specificity of 98%. This difference may be attributable to the fact that the US cohort was derived from patients participating in clinical trials; therefore, patients in the US cohort may have had more active disease than did patients in the UK cohort, which was derived from patients attending a regular clinic, many of whom were receiving treatment. Antibodies to native α-enolase have also been demonstrated in patients with SLE and those with Behçet's disease (38, 39), although in our study, patients with these diagnoses were uniformly negative for anti–CEP-1 antibodies, again supporting the concept that this epitope has specificity for patients with RA.

The importance of citrulline in the anti–CEP-1 antibody response was demonstrated by the low degree of reactivity to the arginine-containing control peptide. However, it was also clear that neighboring amino acids constitute important antigenic determinants, because the response to the other citrullinated α-enolase peptides was much weaker. The 2 amino acids, serine and glycine, flanking the second citrulline residue in CEP-1 have previously been reported to enhance ACPA recognition and binding (37, 40). Thus, it was not surprising to observe that peptide 1B, containing the Ser-Cit-Gly motif, had a higher sensitivity than peptide 1C, which lacks this motif. Our inhibition experiments demonstrate that anti–CEP-1 antibodies do not cross-react with other immunoreactive citrullinated α-enolase peptides. This suggests that antibodies to citrullinated α-enolase react independently with multiple epitopes, which characterizes an antigen-driven immune response. This in turn supports the concept that citrullinated α-enolase is a true autoantigen in RA. The concordance between anti–CEP-1 and anti-CCP antibodies, as well as the results from our inhibition experiments showing a variable degree of cross-reactivity between the 2 antigens, suggests that citrullinated α-enolase may be one of a family of citrullinated autoantigens, for which antibodies are screened, at least in part, by the anti-CCP assay.

At this stage, we do not claim that the anti–CEP-1 ELISA is a new diagnostic assay for RA. Modification of coating conditions, the use of different configurations of peptides, and a standard curve to ensure accurate quantification and reproducibility may increase the sensitivity of the assay. These refinements, together with further analysis of large cohorts, may show that antibodies to CEP-1 define a specific clinical or immunogenetic subset of RA. Thus far, our data using 2 independent cohorts of RA patients and appropriate normal controls and disease controls support the concept that citrullinated α-enolase is a true autoantigen, and that the sequence of the immunodominant peptide representing CEP-1 may be important in the pathogenesis of the disease in the proportion of patients who have the antibody.

In light of the hypothesis that RA may be precipitated by infection, it was intriguing to observe that the sequence of 9 amino acids spanning the immunodominant epitope on CEP-1 was 100% identical to that encoded by P gingivalis. P gingivalis is a gram-negative bacterium that causes adult periodontitis. Like RA, adult periodontitis is a chronic inflammatory disorder in which the accumulation of immune cells leads to local production of proinflammatory cytokines such as tumor necrosis factor α and interleukin-1β, metalloproteinases, and prostaglandins, which results in tissue swelling and degradation (for review, see refs.41 and42). RA is 4 times more common in patients with periodontitis than in the normal population (43), and higher levels of antibodies to P gingivalis (44), as well as a higher prevalence of advanced forms of periodontal destruction, have also been reported in patients with RA compared with control subjects. Finally, both rheumatoid factor production and anti-CCP antibody production have been associated with periodontal disease (41, 45).

These data could imply a common underlying dysregulation of the host immune response in adult periodontitis and RA. However, there is also a striking genetic similarity between the 2 diseases in that both are associated with the HLA SE alleles (46, 47). This, together with the fact that P gingivalis is the only bacterium known to synthesize its own PAD enzyme (28), lends support to another hypothesis: that P gingivalis may be the “septic stimulus” in RA, as proposed by Rosenstein et al (42). Therefore, individuals with periodontal infection may already be exposed to citrullinated antigens, including citrullinated bacterial enolase, generated by host PAD during the inflammatory response or by bacterial PAD produced as a virulence factor to evade host defense (28). In the genetic context of the HLA SE alleles, and in the presence of danger signals, this could result in a pathologic immune response, with the formation of ACPAs. A subclinical arthritis could develop at a later time point, perhaps due to trauma or a viral infection, resulting in citrullination of synovial proteins. Under normal circumstances, this arthritis would be self-limiting. However, the presence of anti–CEP-1 antibodies could lead to cross-reactivity with citrullinated proteins in the joint and amplification of the inflammatory process, with progression to chronic RA.

We were, in fact, able to demonstrate such antibody cross-reactivity between human and bacterial enolase. Not only did patients with RA have antibodies reactive with the P gingivalis version of CEP-1, but anti–CEP-1 antibodies (raised in rabbits or purified from a patient with RA) bound to the immunodominant epitope in P gingivalis enolase. Hence, based on our data, we hypothesize that autoimmunity in the subset of RA patients with a humoral immune response to CEP-1 could be primed by bacterial infection, and that tolerance could be broken by citrullinated bacterial enolase. Cross-reactivity with citrullinated human α-enolase within the joint could then lead to the chronic destructive inflammation that characterizes RA.

AUTHOR CONTRIBUTIONS

Dr. Lundberg had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study design. Lundberg Kinloch, Venables.

Acquisition of data. Lundberg, Kinloch, Fisher, Wegner, Wait, Charles, Mikuls, Venables.

Analysis and interpretation of data. Lundberg, Kinloch, Fisher, Wait, Venables.

Manuscript preparation. Lundberg, Kinloch, Mikuls, Venables.

Statistical analysis. Lundberg.

Acknowledgements

We wish to acknowledge Professor Dorian Haskard, Imperial College London, for providing serum samples from patients with Behçet's syndrome, and Dr. Vivianne Malmstrom and Professor Lars Klareskog, Karolinska Institutet, Stockholm, Sweden, for providing serum samples from patients with spondylarthritides. We would also like to thank James R. O'Dell, MD, as well as investigators and patients from the RAIN Network, for their contributions.

Ancillary