Insights into the detection of AMPA cross-reactivity: comparing cyclic peptide- to protein-based assays

AMPA detection in serum samples

Baseline serum samples from 608 RA patients included in the Leiden Early Arthritis Clinic (EAC) were assessed for ACPA, anti-CarP and AAPA levels as described previously [7, 15, 17, 18]. All recent-onset arthritis (less than 2 years) individuals received their diagnosis within 1 year of follow-up. Briefly, heat-inactivated FCS was modified with either chemical or enzymatic reactions to contain citrulline (Cit), homocitrulline (Hcit) or acetyllysine (Ac). Citrullination was performed as previously described with minor alterations [15]. Protein (1 mg/mL) was incubated overnight at 37 °C in 20 mM Tris–HCl, 1 mM EDTA, 10 mM DTT, and 5 mM CaCl₂. Carbamylation was induced by incubating protein (5 mg/mL) with 1 M KOCN (in PBS) for 12 h at 37 °C. Protein was subsequently desalted with Zeba spin desalting columns (Thermofisher) according to the manufacturer’s protocol. Acetylation was carried out as previously described [19]. Briefly, 20 mL of protein (1 mg/mL in Na₂CO₃) was incubated with 100 µL acetic anhydride and 400 µL pyridine, both added dropwise. The solution was incubated for 18 h at 30 °C under gentle agitation. To stop the reaction, 400 µL of 1 M Tris was added per 20 mL of solution. Proteins were concentrated using concentrator columns (Pierce, 10 K MWCO) and further purified with desalting columns (Zeba, 7 K MWCO, Thermofisher). Protein modifications were validated by ELISAs as described below using monoclonal antibodies specific to one or more of the introduced modifications.

Next, Nunc maxisorb plates (Thermofisher) were coated with modified or unmodified FCS and incubated with samples overnight. Bound immunoglobulin G (IgG) was detected using rabbit anti-human IgG (Dako) and HRP-labeled goat anti-rabbit IgG (Dako). HRP enzyme activity was visualised using ABTS read at 415 nm. For peptide ELISAs, plates coated overnight with streptavidin (Invitrogen) were incubated with biotinylated modified (Cit, Hcit, Ac) or unmodified (Arg, Lys) CXP2 (Patent EP2071335) synthesised at the LUMC facility (Table S1). Following sample incubation, bound antibodies were detected with HRP-labeled goat anti-human IgG (Dako) and visualised with ABTS. Using a standard from pooled RA sera, absorbance values were interpolated into arbitrary units per milliliter (au/mL). The value of the unmodified backbone, defined in this study as the molecular scaffold (either peptide or protein) onto which the PTM was introduced, was subtracted from the value of the corresponding PTM-modified backbone to obtain the specific AMPA signal.

Cut-off calculation and data analysis

To determine a positive response towards a specific PTM backbone, 80 healthy donor samples were measured to calculate a cut-off of average au/mL plus two times the standard deviation of the specific signal (Fig S1). Samples were assigned a positive or negative status for each autoantibody and patients who tested negative on every modification across all backbones were excluded from data analysis. Data from 389 patients were analysed using Rstudio. Correlation analyses on AMPA levels measured on either backbone were performed in GraphPad Prism.

Statistical analysis

The 389 EAC patients were divided into double positive, single positive, or negative groups depending on their PTM reactivity on each backbone. A McNemar’s test was used to calculate whether patient PTM reactivity differed significantly per backbone used. All analyses were performed in Rstudio.

AMPA isolation

AMPAs were isolated from sera of three RA patients and one healthy donor as a control. Nunc maxisorb plates (Thermofisher) were coated with FCS containing either Cit, Hcit or Ac modifications, and incubated with samples (24 µL serum) overnight. For the peptide-based capturing, wells of a streptavidin coated plate (Microcoat) were incubated with biotinylated modified (Cit, Hcit, Ac) CXP2 (Patent EP2071335) or CXP4 synthesised at the LUMC peptide synthesis facility (Table S1). Samples were eluted with 0.1 M acetic acid, followed by neutralisation with 1 M Tris–HCl (pH 9). The samples were then loaded onto Fc-specific beads (CaptureSelect FcXL, Thermofisher) and incubated overnight, followed by the same elution and neutralisation process. Eluted samples were titrated and measured for AMPA reactivity on FCS- or CXP2/4-based ELISAs. AMPA reactivity profiles were generated by normalising ACPA, anti-CarP, and AAPA reactivities per sample, reflecting the epitope recognition profile. This was achieved by combining the area under the curves (AUC) from the titration or absorbances for each of the reactivities and expressing them as a proportion of the total reactivity.

Detection of multiple AMPA reactivities is increased with CXP2 compared to PTM-FCS

To evaluate the impact of PTM backbones on AMPA detection, we measured three autoantibody reactivities in sera of 389 EAC patients using both protein- (FCS) and peptide- (CXP2) based assays. Patients were stratified as positive or negative for each individual assay and then grouped by backbone to be classified as single-positive, double-positive, or negative.

To visualise shifts in individual patient reactivity against PTM-FCS and CXP2, we used Sankey diagrams. When comparing anti-citrullinated versus anti-carbamylated backbone reactivities, nearly half of the single positive ACPA patients on FCS-based backbones shifted to the double-positive category on CXP2 (Fig. 1A). Significantly more patients were found to be anti-CarP single positive on PTM-FCS assays (13.1%) compared to CXP2 (0.8%) (Fig. 1A). When comparing anti-citrullinated versus anti-acetylated backbone reactivity, double-positive patients also increased from 30.1% on PTM-FCS to 48.6% on CXP2 (Fig. 1B). Similarly, in anti-carbamylated versus anti-acetylated backbone reactivity, CXP2 yielded significantly higher detection of samples reactive to both modified peptides (45.8%) compared to PTM-FCS (30.8%) (Fig. 1C). The same trend persisted when comparing all three backbone reactivities for each assay with triple positive patients increasing from 26.0% on PTM-FCS to 45% on CXP2 (Fig. S2).

Despite the increased multi-reactivity on CXP2, the proportion of non-reactive samples was also consistently higher compared to PTM-FCS, highlighting different detection sensitivities between the two assays (Fig. 1A-C). To determine whether the apparent discrepancy observed between the assays might be due to imprecise measurements surrounding the cut-off, we compared AMPA levels across both backbones. While AMPA levels based upon the two assays correlated moderately in many patients, some showed clear backbone-based level differences (Fig. 1D-F). Thus, apparent discrepant findings between the assays did not appear to be largely attributable to borderline values around the cut-off limits. Rather, our results suggest that PTM backbones can significantly influence AMPA reactivity profiles, with peptide-based assays increasing detection of reactivities towards multiple PTM residues.

Modified FCS captures AMPAs with a more restricted epitope recognition profile compared to CXP2

Next, we aimed to investigate whether the enhanced detection of multiple AMPA reactivities with CPX2-based ELISAs compared to FCS-based assays can be attributed to increased detection of cross-reactive AMPA. To this end, we selected three patients with high AMPA values, isolated AMPAs from serum using either modified FCS or CXP2, followed by IgG purification, and assessed the reactivity of the isolated antibodies towards the three different PTMs (Fig. 2A). The starting material contained reactivity to all three PTMs, as determined by both FCS- and peptide-based ELISAs (Fig. 2B and C, respectively). Following AMPA-IgG isolation, samples were titrated and reactivity was reassessed (Fig. 2D and E).

AMPA reactivity profiles were generated to provide a comparative overview of antibody responses against different post-translational modifications (Fig. 2F and G). First, ACPA, anti-CarP, and AAPA reactivities from each sample were quantified by calculating the AUC values from titration assays. These values reflect the strength of each antibody response. The AUCs for ACPA, anti-CarP, and AAPA were then summed to obtain the total AMPA reactivity value for each sample. Finally, each individual AUC was expressed as a proportion of this total, yielding a normalised reactivity profile that illustrates the relative contribution of each antibody specificity (ACPA, anti-CarP, and AAPA) within the overall AMPA response. No AMPA reactivity was detected against unmodified CXP2 (arginine or lysine), and healthy donor serum displayed no reactivity in the assays (Fig. S3). ACPA-IgG isolated from patient 1 with Cit-FCS predominantly exhibited anti-Cit-FCS reactivity (Fig. 2D), whereas ACPA-IgG isolated from the same patient with CCP2 showed a more cross-reactive epitope recognition profile (Fig. 2E). Similar findings were observed across all three patients, where isolation with modified FCS led to enrichment of antibodies against the specific PTM used for isolation, while CXP2-isolated antibodies broadly recognised different PTMs (Fig. 2F and G, respectively). Individual patient PTM-reactivity profiles were confirmed to be highly consistent when using either CXP2 or CXP4 for isolation and detection (Fig. 2G and Fig. S4), indicating that reactivity is maintained across both peptide backbones. Since CXP4 has a defined sequence, it allows assessment of PTM reactivity in a clear sequence context. Validating our findings against the widely used but sequence-undefined CXP2 confirms that the observed reactivity patterns are not limited to a single synthetic peptide. Overall, these findings show that the antigen backbone influences AMPA isolation as modified FCS captures AMPAs with a more restricted, less cross-reactive epitope recognition profile compared to those captured by CXP2/4.

To evaluate whether this difference was due to the detection or isolation procedure, we next conducted a “backbone-switch ELISA”. In these experiments, AMPAs were isolated with one backbone (e.g. FCS) followed by detection in an ELISA with another antigen backbone (e.g. CXP4) (Fig. 3). AMPA-IgG isolated with a peptide (CXP4) and detected by FCS-based ELISA displayed a diverse recognition profile (Fig. 3A). This shows that the CXP4-isolated AMPAs remained highly cross-reactive with various PTMs, even when detected using a protein-based ELISA. Isolation of AMPA-IgG with PTM-FCS and detection by peptide-based ELISA (Fig. 3B) appeared to increase the diversity of the recognition profile compared to that obtained with FCS-based ELISA (Fig. 2F). Overall, these findings emphasise that the antigen backbone influences both the isolation of AMPA and the detection of AMPA cross-reactivity.

Ethics approval and consent to participate

The study was conducted with approval from the regional ethics committee at Leiden University Medical Center. All participants provided written informed consent.

Declaration of generative AI and AI-assisted technologies in the writing process

ChatGTP-4o was used exclusively to improve readability and refine the language in this document. All scientific content was thoroughly reviewed and validated by all authors following each edit, ensuring that the accuracy and integrity of the research remained intact.

Consent for publication

Not applicable.

Competing interests

L.A.T. and R.E.M.T. are mentioned as inventors on a patent on ACPA IgG V-domain glycosylation. The other authors declare no conflict of interest.