Epitope specificity of myeloperoxidase antibodies: identification of candidate human immunodominant epitopes

Abstract

Anti-neutrophil cytoplasmic autoantibodies (ANCA) are a common feature of systemic vasculitides and have been classified as autoimmune conditions based, in part, on these autoantibodies. ANCA are subdivided further based on their primary target: cytoplasm (c-ANCA) or perinuclear region (p-ANCA). p-ANCAs commonly target myeloperoxidase (MPO), an enzyme with microbicidal and degradative activity. MPO antibodies are non-specific for any single disease and found in a variety of vasculitides, most commonly microscopic polyangiitis. Despite their prevalence, their role in human disease pathogenesis remains undefined. We sought to characterize the sequential antigenic determinants of MPO in vasculitis patients with p-ANCA. Of 68 patients with significant levels of p-ANCA, 12 have significant levels of MPO antibodies and were selected for fine specificity epitope mapping. Sequential antigenic targets, including those containing amino acids (aa) 213–222 (WTPGVKRNGF) and aa 511–522 (RLDNRYQPMEPN), were commonly targeted with a prevalence ranging from 33% to 58%. Subsequent analysis of autoantibody binding to the RLDNRYQPMEPN peptide was assessed using a confirmatory enzyme-linked immunosorbent assay format, with six patients displaying significant binding using this method. Antibodies against this epitope, along with four others (aa 393–402, aa 437–446, aa 479–488 and aa 717–726), were reactive to the heavy chain structure of the MPO protein. One epitope, GSASPMELLS (aa 91–100), was within the pro-peptide structure of MPO. B cell epitope prediction algorithms identified all or part of the seven epitopes defined. These results provide major common human anti-MPO immunodominant antigenic targets which can be used to examine further the potential pathogenic mechanisms for these autoantibodies.

Introduction

The use of indirect immunofluorescence has identified two major types of anti-neutrophil cytoplasmic antibodies: cytoplasmic ANCA (c-ANCA) and perinuclear ANCA (p-ANCA). The ANCA-associated vasculitides (AAV) vary in clinical presentation, yet all of them share the same central pathology: inflammation of vessel walls. AAV are serious diseases with an extremely high mortality rate when left untreated. Since the discovery of ANCAs more than two decades ago, the definite claim of their pathogenic role in the disease process of systemic vasculitis has been confounded by variations not only in the distribution of ANCA-positive individuals in relation to actual disease but also in the inconsistencies they present in terms of disease severity, activity and progression.

The primary antigenic target of p-ANCA is the lysosomal enzyme myeloperoxidase (MPO). Anti-MPO antibodies can be found in a variety of immune-mediated disorders, including Churg–Strauss syndrome (40–60%), crescentic glomerulonephritis (64%), Wegener's granulomatosis (24%) and most commonly in microscopic polyangiitis (MPA), wherein these antibodies are detected among 80% of affected individuals [1–3]. Strong evidence also exists from animal experiments showing that p-ANCA directed against MPO can cause vasculitis that resembles human vasculitic disease [4].

Direct pathogenic roles of MPO-ANCA have been demonstrated by their binding to target antigens expressed on the surface of primed neutrophils and monocytes, leading to the induction and release of oxygen metabolites, which trigger vascular injury [5–7]. Knowledge about the target epitopes of autoantibodies can provide valuable insight into the mechanisms that initiate and regulate the autoimmune response. Epitope mapping can identify molecular mimics and elucidate the relationship between an alloantigen and autoimmune disease. The analysis of changes in these target epitopes over time in an individual patient may also provide insight into whether relapses are associated with reactivity to a new epitope or reactivation of an antibody response to the same epitope. The purpose of this study is to examine the fine specificity of autoantibodies targeting MPO. This continuing effort could define epitopes that have pathogenic implications, provide insight into the initiation of this autoimmune response and identify potential therapeutic targets.

Materials and methods

Patient sera selection

The Oklahoma Clinical Immunology Serum Repository (Oklahoma City, OK, USA) contains more than 120 000 coded samples from 70 000 individuals. Sixty-eight samples from patients that tested positive for p-ANCA, and had adequate sera stored within the repository, were obtained for further analysis. Frequency matched healthy controls were selected to run in parallel experiments. This work was conducted with appropriate Institutional Review Board approval from the Oklahoma Medical Research Foundation and the University of Oklahoma Health Sciences Center (OUHSC).

Autoantibody screening

Patient sera were tested for ANCA using indirect immunofluorescence (IIF) following the protocol provided by the manufacturer (Inova Diagnostics, Inc., San Diego, CA, USA). Patient samples with a positive p-ANCA titre by IIF were also tested for MPO antibodies by enzyme-linked immunosorbent assay (ELISA) from the same manufacturer to verify the presence of antibodies to myeloperoxidase.

Solid-phase peptide synthesis and anti-peptide assay

The published sequence of MPO (Accession number: PO5164) was used to construct 369 decapeptides of the 745 amino acid protein overlapping by eight amino acids. The peptides were synthesized on the ends of polyethylene pins using f-moc side-chain protection chemistry and arranged in the format of 96-well microtitre plates (Chiron Mimotopes Pty Ltd, Clayton, Victoria, Australia), as described previously [8,9]. Positive control peptides were synthesized on each plate using a peptide with known positive reactivity by a patient serum sample.

Solid-phase peptides were then tested for antibody reactivity using a modified enzyme-linked immunosorbent assay (ELISA) procedure described previously in detail [8,9]. Assay steps were executed by lowering the pins into microtitre plate wells and incubations were carried out in sealed plastic containers. The peptides were blocked in a 3% low-fat milk phosphate-buffered saline (PBS) solution and then incubated with sera containing primary antibodies. The solid-phase supports were washed with PBS with 0·05% Tween and then incubated with anti-human immunoglobulin (Ig)G as a secondary antibody (Jackson Immunoresearch Laboratories, West Grove, PA, USA). Following another wash period, the peptides were incubated in a para-nitrophenyl phosphate solution in order to induce a colour change if an antibody–peptide interaction was present. The colour change was measured using a micro-ELISA plate reader (Dynex Technologies, Chantilly, VA, USA) at 410 nm and the absorbance values were recorded. Positive controls were developed and normalized to an optical density (OD) of 1·0 to standardize results across plates and assays.

MPO confirmatory anti-peptide ELISA

The sequence of amino acids (aa) numbers 511–522 (RLDNRYQPMEPN) was synthesized on an 8-mer branching polylysine backbone (MAP™) to be used as an antigen for further evaluation, using a method described previously [10,11]. The peptide was constructed by the OUHSC Molecular Biology Proteomics Facility at 1 µg/well coated on a 96-well polystyrene plate overnight at 4°C, washed with PBS, and then blocked with 0·1% bovine serum albumin (BSA) for 1 h at room temperature. Serum samples were diluted at 1:100 and 1:1000 in 0·1% BSA-Tween solution, added to the coated plates and incubated for 3 h at room temperature. Following another wash, alkaline phosphatise-conjugated anti-human IgG (Jackson Immunoresearch Laboratories) was diluted 1:10 000 and added for 3 h incubation at room temperature. The plate was washed again following conjugation and then incubated with para-nitrophenyl phosphate tablets (Sigma Chemical Co., St Louis, MO, USA) dissolved in glycine buffer. Plates were read at a wavelength of 410 nm (Dynex Technologies Inc) and standardized to a common positive control at an OD of 1·0.

Structural analysis of MPO epitopes

Identification of antigenic determinants from continuous epitopes such as MPO utilizes empirical methods by measuring several parameters such as hydrophilicity, flexibility, accessibility, turns, exposed surface, polarity and antigenic propensity of polypeptide chains. Amino acids that build up a protein carry a charge once in a solution and together give an isoelectric point (pI) which enables protein separation. The average pI of the identified epitopes were computed and compared to the non-antigenic decapeptides and the Protein Data Bank was used to identify the coordinates for the crystal structure of MPO (PDB code 1CXP), as defined by Fiedler et al. [12]. These coordinates were used to calculate secondary structure solvent exclusion surface areas by using the BALL View version 1.1.1 program [13] and surface areas were calculated using a solvent probe radius of 1·5 Å. We then identified the location and surface availability of our defined epitopes.

The Immune Epitope Database and Analysis resource (http://www.immuneepitope.org) was accessed to determine B cell epitope predictions for the published sequence of MPO. All prediction calculations are based on propensity scales for each of the 20 amino acids found among humans and, in general, 5–7 amino acid residues is appropriate for finding regions that may potentially be antigenic. Four algorithms were utilized in this study: the Bepipred linear epitope prediction tool which predicts the location of linear B cell epitopes using a combination of a hidden Markov model and a propensity scale method [14]; the Kolaskar and Tongaonkar antigenicity, a semi-empirical method that has been shown to predict 75% antigenic determinants accurately on proteins by using physicochemical properties of amino acid residues and their frequencies of occurrence in experimentally known segmental epitopes [15]; the Emini surface accessibility prediction that uses a formula-based calculation that defines a hexapeptide sequence with a surface probability greater than 1·0 as having an increased probability for being found on the surface [16]; and the ElliPro algorithm that predicts antibody epitopes based on a protein antigen three-dimensional (3D) structure [17].

Results

Twelve patients were identified on the basis of p-ANCA reactivity, detectable anti-MPO antibodies (>20 units of reactivity) and serum availability for fine specificity analysis. Of these patients, 58% were male and the average age of individuals within the cohort was 60·5 (±15·6 years of age). All patients were referred for serological evaluation of a clinical systemic vasculitis, with all but one having evidence of significant renal involvement. Healthy control sera displayed no significant binding when tested by anti-MPO ELISA.

Overlapping decapeptides representing the MPO protein were tested against the 12 patient samples and frequency matched control samples. The patients displayed significant reactivity to multiple sections of the protein, including seven major significant epitopes (Fig. 1). Significant epitopes are defined as being those sequences for which at least 33% of patients exhibited an average reactivity ≥3 standard deviations (s.d.) above the normal mean. These major significant epitopes include epitope 1: GSASPMELLS (aa 91–100); epitope 2: WTPGVKRNGF (aa 213–222); epitope 3: SARIPCFLAG (aa 393–402); epitope 4: WDGERLYQEA (aa 437–446); epitope 5: YRSYNDSVDP (aa 479–488); epitope 6: RLDNRYQPMEPN (aa 511–522); and epitope 7: IFMSNSYPRD (aa 717–726) (Table 2). Epitopes 2 and 6 were bound by the highest percentage of patients, having been bound by 41·7% and 58·3% of tested patient sera, respectively. Epitopes 1, 3, 4, 5 and 7 were all bound by 33·3% of patients. While these epitopes were found to be most common among the patients, the overall response was highly variable (Table 1). An example of this in Fig. 1 shows binding patterns from two patients (Fig. 1a,b) that exhibit a response against various MPO decapeptides, with the only similarity found at decapeptides 256–257 (epitope 6). Males displayed a more diverse repertoire of antibody specificities than females, on average targeting 3·7 specificities compared with 1·2 in females. None of the defined epitope sequences displayed significant binding by control samples.

Fig. 1.

Sequential humoral antigenic determinants of myeloperoxidase. Humoral epitope binding profiles of two representative patient samples are presented in the top panels. A representative control serum sample binding is presented in the lower panel.

Table 2.

Common, major humoral antigenic targets of myeloperoxidase (MPO) antibodies

Epitope	Sequence	Amino acids	% Patients	pI
1	GSASPMELLS	91–100	33·3%	3·29
2	WTPGVKRNGF	213–222	41·7%	11·49
3	SARIPCFLAG	393–402	33·3%	9·01
4	WDGERLYQEA	437–446	33·3%	3·83
5	YRSYNDSVDP	479–488	33·3%	3·88
6	RLDNRYQPMEPN	511–522	58·3%	6·96
7	IFMSNSYPRD	717–726	33·3%	6·69

pI: .isoelectric point.

Table 1.

Patient demographics and epitope specificities of individuals tested for anti-myeloperoxidase (MPO) humoral epitopes

							Epitope
Patient number	Age	Race	Sex	pANCA	MPO units	Epitope 6 ELISA (OD)	1	2	3	4	5	6	7
1	67	Other	M	1620	100	0·420	–	+	+++	–	+++	+++	–
2	62	European American	F	60	20	N/A	–	–	–	–	–	+	–
3	73	Other	F	60	81	0·252	–	–	++	–	–	–	–
4	63	European American	M	20	21	0·245	++	–	–	+	+++	–	–
5	75	Other	F	180	100	0·089	+	–	–	–	–	–	+++
6	62	European American	M	180	100	0·100	+++	–	–	+++	+	+++	–
7	64	Eastern Indian	F	540	124	0·066	–	–	–	–	–	–	+++
8	60	European American	M	60	40	1·000	+	+	++	+++	++	+++	–
9	76	Unknown	M	180	94	0·061	–	–	–	–	–	+++	+++
10	65	European American	M	80	33	0·341	–	+++	+	++	–	+	+
11	38	European American	F	60	90	0·118	–	++	–	–	–	–	–
12	22	African American	M	60	24	0·295	–	++	–	–	–	+	–

The targeting of epitopes by each patient in terms of standard deviations (s.d.) above the normal average binding. The binding of each epitope is defined by a scale where (–) represents binding <3 s.d. above the control response (+) represents binding between 3 s.d. and 5 s.d. above controls (++) binding between 5 s.d. and 7 s.d., and (+++) binding >7 s.d. ELISA: enzyme-linked immunosorbent assay; OD: optical density; pANCA: perinuclear region anti-neutrophil cytoplasmic autoantibodies; M: male; F: female.

The RLDNRYQPMEPN (aa 511–522) sequence representing epitope 6, which is the most common antigen target with the highest intensity of binding compared to the other defined epitopes, was used for confirmatory analysis of the solid-phase peptide results. The samples were screened using a peptide ELISA format with the peptide constructed on a polylysine (MAP) backbone. Of the 12 samples (excluding one with insufficient sera), six patients displayed significant levels of this antibody specificity (Table 1), providing 100% concordance with the solid phase epitope mapping.

A detailed analysis of the sequences that comprise the seven epitopes defined by this study was performed for the purpose of trying to identify features that could be used to further differentiate antigenic regions of the protein from those that are non-antigenic. The decapeptides that make up the defined epitope sequences had an average pI of 6·45 (Table 2), while the average pI for the remaining decapeptides equalled 7·11. There was also no significant difference between the amino acid usage within the sequences for antigenic and non-antigenic regions.

To visualize the location of the seven significant and common epitopes, to determine surface availability of these epitopes and to assess the proximity of these epitopes to functional regions of the protein we referred to the crystal structure model of MPO determined by Fiedler et al. [12]. Epitope 1 is located within the pro-peptide region of the protein and is therefore not identified in the processed, mature form of the protein represented in the 3D model. Using this model, epitope 3 is the only epitope within close proximity to the active site of the protein (His261, Arg405 and Gln257) (Fig. 2). Both epitopes 6 and 7 share close proximity within the structural model of the protein, even though they are separated by 195 amino acids within the linear sequence. Interestingly, 11 of the 12 patients target one or both of these two epitopes, suggesting that this commonly targeted region of the protein could be an important feature in identifying immunodominant epitopes in the pathogenesis of AAV.

Fig. 2.

Myeloperoxidase (MPO) epitope delineation and characterization using three-dimensional (3D) structure modelling. (a) Standard deviations above normal control binding for the mean of all 12 patient samples. Epitopes 2–7 are assigned a reference colour for distinction in part B. (b) 3D structural representation of the MPO dimer illustrates the surface location of epitopes 2–7 and their relation to the catalytic sites (light blue).

Comparing our identified epitopes from the Bepipred linear epitope prediction tool we have identified four predicted epitopes (AEYEDGFSLPYGWTPGVKRNG, YRSYNDSVDPR, RYQPMEPNPRVP, SYPR) containing all or part of the amino acid sequences identified in our study (epitopes 2, 5, 6 and 7). Further comparisons with other antibody epitope prediction methods identified epitope 3 containing one predicted epitope (RIPCFLA) by Kolaskar and Tongaonkar antigenicity and epitope 7 containing the last predicted epitope (NSYPRD) by Emini surface accessibility prediction. Using the ElliPro algorithm, we have found epitope 1 embedded in the predicted first epitope and epitope 2 beginning in the second predicted epitope sequence. Thus, utilizing multiple B cell epitope prediction algorithms, similarities were seen between predicted epitopes and all seven identified epitopes in our study.

Discussion

The purpose of this study was to use fine specificity epitope mapping to identify common antigenic targets of MPO that could provide insight into pathomechanisms involving anti-MPO autoantibodies. The pathogenic potential of MPO-ANCA in vasculitis and glomerulonephritis has been demonstrated through murine passive transfer experiments [18]. MPO-ANCA also have the ability to interfere with ceruloplasmin inhibition of MPO [19,20]. This failed regulation of MPO would result in increased MPO activity, including increased production of hypochlorous acid and other proinflammatory mediators, posing a significant risk for damage to the surrounding vasculature [21,22]. MPO-ANCA have been found to be directed against unique MPO epitopes for vasculitis as well as for different secondary complications of vasculitis [23–25]. Thus, examining immunodominant humoral target regions of the MPO molecule is vital and can provide insight into the MPO-ANCA immune response.

Other evaluations of MPO epitope specificity were able to identify broad characteristics of the protein's antigenic potential, both through analysis of epitope restriction [26,27] and through the use of recombinant deletion mutants of the protein [25,28–30]. One study generated multiple human–mouse MPO chimera to examine regions of antibody specificity, while another found that MPO-ANCA recognize epitopes on native human MPO and that 30% of MPO-ANCA do not bind recombinant versions of the human protein [26,31,32]. Studies of competitive binding of antibodies to their target antigen are helpful in determining the relative number of epitopes, but they generally fail to identify the location (target amino acids) of these epitopes. Seta et al. found that at least three independent T cell epitopes exist on the MPO molecule by using recombinant MPO fragments to detect autoreactive CD4⁺ T cells to multiple MPO epitopes [33].

Our experiment has identified successfully seven humoral epitopes among several members of our cohort. The antigenic sequences identified include aa 91–100 (GSASPMELLS), aa 213–222 (WTPGVKRNFG), aa 393–402 (SARIPCFLAG), aa 437–446 (WDGERLYQEA), aa 479–488 (YRSYNDSVDP), aa 511–522 (RLDNRYQPMEPN) and aa 717–726 (IFMSNSYPRD). In studies identifying disease inducing epitopes in anti-glomerular basement membrane (GBM)-associated disease, the majority of patients react to a single, well-defined epitope [34]. With MPO-ANCA, several immunodominant epitopes are proposed to be involved in the disease process of p-ANCA associated vasculitis.

Erdbrugger et al. demonstrated a restriction of antibody reactivity to two intertwined target regions corresponding to the C or D regions of the carboxyl terminus of the heavy chain [31]. In our study, all but one reactive epitope were found on the heavy chain of the mature MPO protein structure (epitopes 2–7), including the most antigenic (epitope 6). Epitopes 4 and 7 were included in the amino acid sequence reported by Fujii et al. [25]. This further highlights the importance of the heavy chain of the MPO protein in disease pathogenesis. They were able to demonstrate that most MPO-ANCA reacted with up to three epitope regions on the heavy chain part of MPO, while none of the MPO-ANCA reacted with the light chain [25,28,31,34]. Crescentic glomerulonephritis also correlates with a particular epitope (Ha epitope) of MPO-ANCA, recognizing the N terminus of the MPO heavy chain [29]. The more frequent and more severe renal involvement reported in the aforementioned study reinforces the notion that the presence of MPO-ANCA, especially those binding at the heavy chain, may determine disease severity as well as its organ involvement.

Epitope specificity in terms of proximity to the active site (His261, Arg405 and Gln257) in the conformational structure of the mature MPO protein has been suggested, but not clearly supported to date. Previous work suggests that it is unlikely that the effects of MPO-ANCA are the result of interference with the active site of the protein, as the enzymatic activity of MPO is mostly unaffected by the presence of MPO-ANCA [35]. Our study validates this hypothesis by showing that the amino acids forming the centre of the active site are not located within any of the defined epitopes of our study, either in the linear sequence of the protein or as indicated by correlation of epitopes with crystallographic structure analysis. Epitope 3 SARIPCFLAG (aa 393–402) shares the closest proximity with the active site of the protein, but with the relatively protected location of the active site within a 10 Å-wide channel on the surface of the protein it is unlikely that antibodies targeting this epitope would interfere with the catalytic activity of the active site. Interestingly, this is the opposite of those seen with other studies, including our parallel experiment studying proteinase 3 (PR3)-ANCA interaction wherein the functional epitopes are located on the surface and proximal to the active sites of the protein structure [36–39]. The important and common finding with our PR3 study is the recognition of a potential immunodominant epitope found in the pro-peptide region (epitope 1) of these enzymes.

Different epitope recognition might lead to different functional influence on native MPO molecules by anti-MPO antibodies, and thus may contribute to the different disease expressions. This explains the highly variable response seen between individuals that recognized the immunodominant antigenic epitopes identified in our study. Only epitopes 6 and 7 have been shown to bind to most of the patient sera. However, we cannot dismiss the importance of the other recognized epitopes, as there is no absolute reactivity found among the normal controls. This difference in immunological characteristics of MPO-ANCA might contribute to the more diverse types of systemic vasculitis seen in this group compared to the PR3-ANCA associated vasculitis. The titres of MPO-ANCA have also been shown not to reflect disease activity at all times [29]. A prospective analysis of multiple serum samples from a large group of patients to determine a clear correlation between the antibody-binding profile and specific disease manifestations or levels of activity or changes thereof is ideal in this setting [11,40].

Anti-MPO autoimmune responses are directed against a limited number of immunodominant epitopes on MPO and the same epitopes are targeted during disease onset and relapse [28]. The identification of these human epitopes in our study, the majority of which belong to the heavy chain of the MPO protein structure, provides a restricted number of significant and common antigenic targets that can serve as a stronger foundation for such experiments.

Disclosures

The authors have no financial disclosures related to this manuscript.