Study design and patient description
This study was designed to define new circulating auto-antibodies in glomerulonephritis and their potential correlation with the clinical outcome. It was divided in different subsequent steps starting from a discovery approach that utilized a new peptide array build up for this study, followed by a validation step utilizing a second peptide array customized on findings deriving from discovery and ended with the extension of the analysis of circulating levels restricted to specific auto-antibodies (ELISA) derived from the prior phases above. Pathological and functional analysis concluded the work up.
The study comprised a total of 56 healthy donors and 211 patients (170 with MN, 23 with idiopathic focal segmental glomerulosclerosis (FSGS), 6 with genetic FSGS, 12 with IgA glomerulonephritis). They were utilized at different phases of the study as it follows (see Scheme 1): a) the fishing of new antibodies with the peptide array (discovery approach) was initially done in a subset of 4 healthy and 10 MN patients randomly extracted from the whole sample dataset; b) validation of results deriving from step 1 was carried out in 40 healthy donors and 97 MN patients; c) all healthy donors and different pathology subsets (n 267) were utilized for comparison of results obtained with quantitative tests (ELISA) on selected antibodies. Clinical characteristics of healthy, MN, and other categories of patients with glomerulonephritis are reported in Table 1a,b.
Whole proteome peptide discovery approach
The number of samples to be investigated in the discovery approach (10 MN sera, 4 healthy donors) was set out a priori considering the maximal capacity of the peptide array to trace the whole genome panel of 7.499.126 peptides (see Material and Methods section) made up of 16 overlapping aminoacids each (Figure 1A). In the discovery phase, two independent mathematical models were devised that allowed the identification of peptides with a significant probability to make a distinction between MN and healthy sera on the basis of the fluorescence intensity (S-PIE) and to correlate the intensity of reaction with the clinical outcome in MN patients (WGCNA). Specifications of each individual approach are reported in M&M.
S-PIE identified 8,243 peptides (0.1% of the 7.499.126 panel) with R-ratio higher than 106 that is the limit of the noise signal (Figure 1A,B). Subsequently, all peptides with a similar ratio in at least 1 control were removed, resulting in 3,831 remaining peptides. As a last filter, all sequences with low Shannon information (<3.17) were discarded, to keep only 756 peptides (0.01% of the 7.499.126 panel).
In the WGCNA approach, the whole dataset of 7.499.126 peptides filtered for fluorescence intensity (i.e. 8.676 peptides with median intensity MN sera > healthy sera) were clustered into 22 modules characterized by peptides with a similar fluorescence signal intensity to which it was assigned an arbitrary colour (Figure 1C). Among these, red, turquoise and blue modules showed the higher Spearman correlation with the clinical outcome at T12 and T24 (redT12=0.74; redT24=0.92; turquoiseT12=0.53; turquoiseT24=0.67; blueT12=0.63; blueT24=0.8) for a total of 277 peptides (PLA2R1 peptides excluded). Overall, the discovery phase (S-PIE and WGCNA) generated a total of 1.000 peptides (756 S-PIE, 277 WGCNA, 33 common to both) that identified 467 proteins.
Peptide array technology identifies prior known PLA2R1 epitopes
A part of the discovery approach was devoted to match the results of the assay with what already discovered on epitopes of PLA2R1, the major recognized autoantigen of MN 1.
S-PIE recognized 9 peptides within the constitutive 2765 peptides of PLA2R1 with R-ratio larger than the cut-off (>106) that included all the epitopes of the protein (CysR, FNII, CTDL1-8). WGCNA recognized 96 and 108 peptides of the first five CTLD domains of PLA2R1 that correlated with the clinical outcome (proteinuria at 12 or 24 months) (Supplemental Table 2); we also confirmed that CysR was not included in the epitope predictory panel 23. Fisher's enrichment analysis of PLA2R1 epitopes in the 22 modules identified by WGCNA highlighted epitopes associated with proteinuria outcome (Supplemental Figure 1). The overlapping peptides obtained using both analysis (S-PIE and WGCNA) were then modelled on a 3D cryo-EM model 24 (Supplemental Figure 2). Interestingly, all peptides were accessible, i.e. located on the protein surface, and in our study mapped to the CTLD1, CTLD3 and CTLD6 domains respectfully: in position 289 and 307 (CTLD1), or in position 1080 (CTLD6), as postulated by Cui et al. using genome wide association study 25; in position 285 and 1130, or in the CTLD3 domain in position 590, as previously reported by Fresquet et al.; and finally in position 613 using chimeric PLA2R1 constructs 24.
Validation array identified new autoantibodies in MN sera
To localize epitope binding sites for auto-antibodies in the 467 proteins recognized in the fishing phase, a new peptide array was customized utilizing the 1,000 identified peptides of 13 aminoacids, modified at both edges for addition of 15 amino acids each, for an overall sequence of 43 aminoacids. Eliminating redundant peptides, the new array consisted of 18.557 peptides of 16 aminoacids each that covered the whole sequence of the 467 proteins identified above (Supplemental Table 1). Forty normal sera and 97 MN patients were tested by means of the validation array. Fluorescence intensities were analysed by either S-PIE, multidimensional scaling (MDS), T-test analysis (Volcano plot), Support Vector Machine (SVM) and Partial Last squire discriminat analysis (PLS-DA) (Figure 2). In particular, MDS analysis applied to fluorescence intensity of all peptides allowed the clear discrimination between healthy and MN patients (Figure 2A). S-PIE analysis showed a total of 72 peptides with intensities higher than the cut-off, whereas the application of a T-test corrected for multiple interactions, PLS-DA and SVM identified 71, 33 and 19 peptides maximizing the discrimination between healthy and MN, healthy and anti-PLA2R1+ and healthy and anti-PLA2R1- sera respectively (see Volcano plot in Figures 2B, 2C and 2D). Their intensity profiles were visualised as heat map (Supplemental Figure 3). These 105 peptides identified 21 human proteins (Figure 2E and Supplemental Table 3). The mean of the multiple intensity profiles for each protein is reported in Figure 2E. All the 21 proteins are expressed by human and/or mouse podocyte cells (Supplemental Figure 4). In particular, the cellular origins of these proteins were 15% from plasma membranes, 34% from the cytosol, 24% from the nucleus, 12% were secreted, 3% from extracellular matrix and 12% from other organelles such as Golgi apparatus, endoplasmic reticulum or mitochondria 26,27.
Among the 5 proteins out of the 21 described above identified by both methods, FMNL1 had a strong signal intensity with the highest -Log10 P-values in PLA2R1- MN patients (Figure 2D, 2E, Supplemental Table 1 and 3). Moreover, FMNL1 had the minor rank score and major VIP score in SVM and PLS-DA respectively in the discrimination between PLA2R1- MN patients and healthy donors. Based on this result, we decided to focus on FMNL1protein. It is noteworthy that the FMNL1 peptides correspond to the region with the highest B-cell epitope score, according to the BepiPred-2.0 server (Supplemental Figure 5). The region spanning aminoacids 609 to 639 of FMNL1 belong to an unstructured loop region preceding FH2 domain that should be accessible to autoantibodies.
Anti-FMNL1 IgG4 are present in sera of MN patients
Validation of the peptide array results was carried out using custom developed ELISA using FMNL1 recombinant protein. Prior to assessing the levels of anti-FMNL1 in sera of healthy donors or patients affected by MN, isotype characterisation was carried out. The vast majority of anti-FMNL1 antibodies present in patient with MN were, as expected, of IgG4 isotype (Supplemental Figure 6). Levels of anti-FMNL1 IgG4 were then assessed in healthy and MN patient sera at diagnosis, showing a highly statistical difference (P<0.0001). Anti-FMNL1 IgG4 were also more abundant in anti-PLA2R1- MN patients compared to anti-PLA2R1+ and in the former category (anti-PLA2R-), the antibody levels were higher in those patients who did not undergo remission during the follow up (T12) (respectively, anti-FMNL1>2.641 compared to anti-FMNL1< 2.641 RU/ml [(OR) 10; (Cl)3-35; p=0.0002]) (Figure 3A and B). ROC analysis confirmed these differences (Figure 3C).
Anti-FMNL IgG tot are present in sera of other glomerulonephritis
To test anti-FMNL1 specificity for MN, small subgroups of patients affected by other hystologic and genetic forms of glomerulonephritis were tested (23 with idiopathic focal segmental glomerulosclerosis, 6 with genetic FSGS, 12 with IgA glomerulonephritis) (Table 1b). In these cases, total IgG antibodies were determined Serum levels of anti-FMNL1 IgGs were in all cases higher than in normal control, particularly in patients with IgA GN (Figure 4).
MN sera recognize FMNL1 in cell lysates of macrophages.
Cell lysates of macrophages freshly isolated from whole blood of healthy subjects were blotted using sera from healthy subjects, FMNL1 negative or positive MN patients (Figure 5). Western blot showed that FMNL1 positive MN patient sera recognized a band at 120 kDa, at the same molecular weight as recombinant FMNL1, thus confirming the specific presence of circulating autoantibodies against FMNL1, directed against macrophages in patient specific subset of MN patients.
FMNL1 is expressed in macrophages within glomerular capillaries in glomerulonephritis.
We assessed expression FMNL1 localization in renal biopsies (MN and LN) using immunohistochemistry. FMNL1 was expressed mainly in PLA2R1-negative MN biopsies in glomeruli (arrow heads) as well as in some cells in the tubule-interstitial space (arrows, Figure 6). In glomeruli, FMNL1 was detected in circulating cells within the capillaries and was distinctly different from the membranous staining of PLA2R1 and/or IgG4. This glomerular stainingand particularly in patients with PLA2R1-MN compared to PLA2R1+ MN cases. In glomeruli of control patients (normal renal tissue distal from renal carcinoma), no FMNL1 staining could be detected.
Immunohistochemistry using serial sections of patients with lupus nephritis class V and III (Supplemental Table 4) showed cells expressing FMNL1 to be CD68 positive (monocytes/macrophages) and negative for CD3 and CD79a (T and B lymphocyte markersrespectively) (Figure 7A). Double immunostaining for FMNL1 and CD68 showed co-localization of these two proteins in monocytes/macrophages (Figure 7B).