Plasma from patients with anti-glomerular basement membrane disease could recognize microbial peptides

Infection has long been suspected as a trigger of autoimmune diseases, and molecular mimicry mechanism was hypothesized in this study. Microbe originated peptides were searched from the Uniprot database based on a previous defined critical amino acid motif within α3129−150, isoleucine137, tryptophan140, glycine142, phenylalanine 143 and phenylalanine 145. 23826 microbial peptides were identified using our searching strategy, among which seven were related with human infections. Circulating IgG and IgM antibodies against the seven microbial peptides were detected using ELISA in 76 patients with anti-GBM disease. Four peptides were recognized by both IgG and IgM antibodies, and one peptide was recognized by IgG antibodies only. Peptides from Bacteroides, Saccharomyces cerevisiae, and Bifidobacterium thermophilum possessed the highest recognition frequency with the prevalence of 73.7%, 61.8% and 67.1% for IgG, 56.6%, 44.7% and 67.1% for IgM in anti-GBM patients. Patients with antibodies against these microbial peptides showed more severe kidney injury, including higher serum creatinine and higher percentage of crescent formation. In conclusion, antibodies against microbial peptides were identified in the circulation of anti-GBM patients, implying its etiological role in eliciting autoimmune response against α3(IV)NC1 through molecular mimicry.

In our present investigation, we identified seven microbe-derived peptides based on the motif (IxxWxGFxF) from Uniprot database, which could infect human beings. Both IgG and IgM antibodies against most of these peptides were detected in plasma from patients with anti-GBM disease. This result raised the possibility that epitope mimicry by microbial peptides might contribute to the onset of human anti-GBM disease.

General data of patients
The demographic and clinical data of 76 patients with anti-GBM disease are shown in Table 1 Searching microbial peptides using the critical residues on α3(IV)NC1 We searched the Uniprot database (http://www.uniprot.org) for microbial peptide candidates that mimic the critical motif of isoleucine (I) 137 , tryptophan (W) 140 , glycine (G) 142 , phenylalanine (F) 143 , and phenylalanine (F) 145 on α3(IV)NC1. Peptides were selected based on three criteria: 1) peptides with critical residues IxxWxGF; or 2) peptides with critical residues GFxF; and 3) peptides derived from microbes that could infect human beings. 23826 peptides were identified using our searching strategy, among which seven were related with human infections (Table 2), and were synthesized for further study (designated as P1-P7).
In order to confirm the specificity of circulating autoantibodies against microbial peptides in anti-GBM patients, antibodies against 20 linear peptides unrelated to microbes or α3(IV)  NC1 were also detected. Only 2 of the 20 peptides were recognized by 3 of the 76 (3.9%) plasma samples. Among the 39 patients with positive HLA-DRB1 Ã 1501, antibody reactions towards the seven microbial peptides were re-calculated and compared to those 11 patients without this susceptible HLA allele. No significant difference of recognition frequencies was identified between the two groups of patients (P>0.05).
In Cross-reaction between autoantibodies against α3(IV)NC1, P14 and microbial peptides An inhibition ELISA was performed to examine cross-reaction between autoantibodies against α3(IV)NC1 or P14 and the antibodies against microbial peptides. Binding of antibodies against the microbial peptides was strongly inhibited by themselves, as well as α3(IV)NC1 and P14, which indicated that the antibodies against microbial peptides cross reacted with P14 and α3(IV)NC1 (Fig 3).

Discussion
The present study demonstrated that microbe-derived peptides with critical motif on α3(IV) NC1 could be recognized by the majority of anti-GBM patients. Their antibodies showed cross-reaction with those against P14 and α3(IV)NC1. These findings implicate that molecular mimicry might be a link between infection and autoimmunity in human anti-GBM disease.
In the current study, we found that four microbial peptides possessed both circulating IgM and IgG reactions and one possessed IgG reaction in anti-GBM plasma. All of them were derived from opportunistic bacteria. Although they may not induce symptomatic infections themselves, they might elicit immune responses when they gain access to the circulation [36], and the ensuing epitope spreading and somatic hypermutation [37][38][39] might cause autoimmunity towards P14 and α3(IV)NC1. This finding provided the first evidence for the activation of immune responses towards bacteria in the circulation of patients with anti-GBM disease, which is the first step of molecular mimicry-mediated autoimmune responses. The titer of autoantibodies against these microbial peptides presented positive correlation with the severity of kidney damage. However, the level of autoantibodies against α3(IV)NC1 only correlated with that of P1 among the seven microbial peptides. In our previous study, we found that intermolecular epitope spreading happened in patients with severe renal damage, with higher titer of autoantibodies against different α chains [40]. Similarly, in this case, higher titer of autoantibodies against microbial peptides might indicated larger chance of epitope spreading to P14 and α3(IV)NC1. Whether these autoantibodies had direct pathogenic role in kidney damage needs further investigation. 78% of anti-GBM patients included in our study carried HLA-DRB1 Ã 1501 allele, which is consistent with previous reports [29,30,33]. Although no significant difference on the frequency of microbial peptide recognition was found between the patients with and without HLA-DRB1 Ã 1501 allele, closer associations between antibodies against bacteria-derived peptides and kidney injury was observed in patients having HLA-DRB1 Ã 1501. HLA-DRB1 Ã 1501 is the susceptible allele for anti-GBM disease, therefore, epitope spreading from microbial peptides to P14 and α3(IV)NC1 might be easier to happen in patients with HLA-DRB1 Ã 1501 and resulting in more severe renal damage. Using microbial peptides based on critical motifs of α3, this cross reactivity was detected in patients with anti-GBM disease. However, further attempts were needed on how the cross-reactivity happened on T cell and B cell levels [41].
It is interesting that we detected autoantibody towards the microbial peptides both in plasma of patients with ANCA-associated vasculitis and anti-GBM disease, but few were detected in patients with other autoimmune kidney diseases, such as lupus nephritis, IgA nephropathy, or idiopathic membranous nephropathy. It has been reported that almost 30% patients with of anti-GBM disease had coexisting MPO-ANCA [42][43][44][45]. Since no cross reaction was found between MPO and GP antigen [44], cross detection in linear epitopes of MPO and GP autoantigen were explored in our previous investigation, which found that over half of anti-GBM patients had autoantibodies against linear peptides derived from MPO [46]. In our present investigation, there is no significant difference in the titer of IgG autoantibody against P2, P4, P5 and P7 and IgM autoantibody against P7 between AASV and anti-GBM patients. This interesting result suggests that in the initiation of the two diseases, similar microbes generated immune response in the hosts, but different genetic predispositions, especially HLA alleles, lead to different antigen presentation and epitope spreading process subsequently [47,48].
This observational study has several limitations. Firstly, although antibodies towards microbial peptides were found in patients, immunization of these peptides failed to induce crescentic glomerulonephritis in WKY rats. Therefore, further investigations in humanized mice models might be needed to further elucidate the pathogenicity of these microbial peptides. Second, microbe-derived peptide searching was only based on the critical amino acids on α3 136-146 , which may result in missing of other potential microbes mimicking other parts of the whole α3(IV)NC1.
In conclusion, antibodies towards microbe-derived peptides were detected in the circulation of patients with anti-GBM disease. These microbes might etiologically participate in the pathogenesis of human anti-GBM disease through molecular mimicry.

Materials and methods
Patients and plasma 76 patients with anti-GBM disease were included in Peking University First Hospital from 1996 to 2012. Clinical data were collected at the time of diagnosis as well as during follow-up. All the patients had positive circulating anti-GBM antibodies, as detected by ELISA using purified bovine α(IV)NC1 (Euroimmun, Lubeck, Germany) and recombinant human α3(IV)NC1 as solid-phase antigens. 17 out of the 76 patients were also positive for MPO-ANCA, detected by both indirect immunofluorescence assay and antigen-specific ELISA using purified MPO as solid phase ligand (Euroimmun, Lubeck, Germany). Plasma from patients with anti-GBM disease was collected before initiation of immunosuppressive therapy. Among the 76 patients, 50 patients had DNA samples and underwent HLA genotyping. Plasma from 18 patients with membranous nephropathy (MN), 32 patients with primary focal segmental glomerulosclerosis (FSGS), 21 patients with minimal change disease (MCD), 20 patients with IgA nephropathy, 20 patients with lupus nephritis were collected as disease controls.
Plasma from 91 healthy blood donors was collected as normal controls. All the plasma was stored at -20˚C until use. The research was in compliance of the Declaration of Helsinki and approved by the ethics committee of Peking University First Hospital. Written informed consent was obtained from each participant. Authors do not have access to information that could identify individual participants after data collection.
We searched the Uniprot database (http://www.uniprot.org) for microbial peptide candidates based on the critical motif of IxxWxGFxF. Peptides were selected based on three criteria: 1) peptides with critical residues IxxWxGF; or 2) peptides with critical residues GFxF; and 3) peptides derived from microbes were related to infections in human beings.

Preparation of linear peptides
Peptides were synthesized on an automatic peptide synthesizer using Fmoc (9-fluorenylmethyloxycarbonyl) chemistry (Beijing Scilight Biotechnology Ltd Co, Beijing, China), and purified by a reverse-phase CIS column on a preparative HPLC. Purified peptides were analyzed by HPLC for purity and mass spectrometry for correct sequence. Peptides with purity > 98% were used for further tests.

Detection of circulating antibodies against peptides by ELISA
Highly purified synthetic peptides were diluted at 10μg/ml with coating buffer (0.05M bicarbonate buffer, pH 9.6) and plated on half of the wells of a polystyrene microtitre plate (Nunc Immunoplate; Nunc, Roskilde, Denmark), at 4˚C overnight. The other half was coated with coating buffer alone as antigen-free wells to exclude non-specific binding. The plate was washed three times with PBS containing 0.1% Tween-20 (PBST) between steps. 2% bovine serum albumin (BSA) diluted with PBS was used to block non-specific binding sites. The plasma was diluted to 1:100 in PBST containing 1% BSA and were added in duplication. Every plate contained positive, negative and blank controls. The plate was incubated at 37˚C for 60min. The binding was detected with alkaline phosphatase-conjugated goat anti-human IgG or IgM (Fc specific; Sigma, St Louis, MO, USA) at a dilution of 1:5000. P-nitrophenyl phosphate (1mg/ml; Sigma) was used in substrate buffer [1M diethanolamine and 0.5mM MgCl 2 (pH 9.8)]. Color development was measured spectrophotometrically at 405nm (Bio-Rad, Tokyo, Japan) 30min later. Plasma from 91 healthy blood donors were used to build up the cutoff value using mean+2SD. The positive control was chosen as a sample with an OD value of 1.5. Each plate contained this positive control. Samples were re-examined when SEMs >10% were found.
Cross-reaction between autoantibodies against α3(IV)NC1, P14 and microbe-derived peptides Cross-reactions between autoantibodies against α3(IV)NC1 or P14 and microbial peptides were investigated using an inhibition ELISA. Briefly, polystyrene microtiter plates were coated with soluble microbial linear peptides (P1-P7) at a concentration of 10 μg/ml. Diluted plasma was pre-incubated with soluble linear peptides (P1-P7), recombinant human α3(IV)NC1 or P14 at a concentration from 0.5 to 50 μmol/l at 37˚C for 60 min. The mixtures were then transferred to the peptide-coated microtiter plates, and the bound autoantibodies were detected with alkaline phosphatase-conjugated secondary autoantibodies, as described above.

Statistical analysis
Differences in quantitative parameters were assessed using Student's t test or nonparametric test. Differences in qualitative data were compared using chi-squared test or Fisher's exact test. Pearson and Spearman rank correlations were performed to analyze the relationship between antibody levels and clinical data as appropriated. Peptides with recognition <5% were excluded from further statistical analysis. P value <0.05 was considered significant. Analysis was performed with SPSS statistical software package (version 10.0, Chicago, IL, US).