A circulating antibody panel for pretransplant prediction of FSGS recurrence after kidney transplantation

Identification of Abs associated with rFSGS after renal transplantation

To identify potential autoAbs associated with rFSGS, we used a discovery set of pretransplant sera from 20 patients with biopsy-confirmed diagnosis of primary FSGS as the cause of their end-stage renal disease (ESRD), of whom 10 had progressed to rFSGS within the first year after transplant (mean time to recurrence, 36 days) and 10 had not had recurrence of proteinuria or histological disease after transplantation [nonrecurrent FSGS (nrFSGS)]. At transplant, these two groups of patients were indistinguishable by demographical, disease, or clinical parameters (Table 1). Recurrence was defined as heavy proteinuria with biopsy confirmation of FSGS, with glomerular sclerosis and podocyte fusion and injury, and without evidence of acute rejection, glomerulitis, or allograft glomerulopathy. The patients were monitored for 6 months before transplantation and followed up 12 months after transplantation. Twenty unique pretransplant serum samples were assayed on high-density protein microarrays. Figure 1 summarizes the study. Immunoglobulin G (IgG) Abs against a large panel (n = 789) of unique antigens were significantly increased (P value range, 0.0004 to 0.0433) in the recurrent group at transplant, with a smaller number of IgG Abs (n = 78) showing an increase in the nonrecurrent patients. These significantly different IgG profiles in the sera of rFSGS and nrFSGS patients demonstrated that rFSGS is associated with a signature of increased humoral responses to a variety of non–human leukocyte antigens (non-HLA) after kidney transplantation. The protein array data are available in the Gene Expression Omnibus (GEO) database (GEO ID: GSE57099). The normalized data are presented in table S1, and a volcano plot for the data is presented in fig. S1.

To select the best biomarkers for additional validation of rFSGS prediction from pretransplant sera, we filtered the ProtoArray data to maximize for fold increase (>2) and signal intensity of at least twice the background threshold (>1000), which resulted in a selection of 151 IgG Abs from the larger panel of 789 Abs (table S1). Because the hallmark of histological diagnosis in rFSGS is podocyte/glomerular injury, we additionally filtered the IgG Ab panel to select for antigens that are expressed in the renal glomerulus, using our previously published integrative antibiomic analysis (22), where we performed cross-mapping of autoAb targets with kidney compartment gene expression data (22). We used the kidney- and kidney compartment–specific antigens identified in a previous study by Li et al. (22) and found that there was an enrichment for Ab specifically targeting particular compartments of the kidney, mainly the glomerulus (P = 0.02), the renal outer cortex (P = 0.01), the renal pelvis (P = 0.02), and the renal papillary tips (P = 0.01). Pathway analysis for inferred mechanisms of injury (Ingenuity Pathway Analysis) showed that these Abs were involved in pathways related to antigen presentation (P = 5.97 × 10⁻⁶), inflammation (P = 5.97 × 10⁻⁶), and cell death (P value = 6.03 × 10⁻⁵). We used the filtering criteria described (fold change, signal intensity, inferred glomerular expression, and functional relevance in inflammation and kidney injury) to select a panel of 10 IgG Abs (CD40, SNRBP2, FAS, PTPRO, P2RY11, RXRA, CCL19, MYLK, APOL2, and CGB5) for further validation by ELISA.

ELISA validation of Abs that can predict rFSGS after renal transplantation

Customized ELISA assays were generated on the Meso Scale Discovery (MSD) platform as previously described (21) for the autoAbs shown in Table 2. One hundred forty-one sera collected at different time points from 98 patients were processed for customized ELISA assays for all 10 autoAbs in Table 2. ELISA analyses confirmed statistically significant elevation of all 10 Ab titers [CD40 (P = 0.0002), PTPRO (P = 0.015), FAS (P = 0.0036), CGB5 (P = 0.031), SNRBP2 (P = 0.0044), APOL2 (P = 0.024), P2RY11 (P = 0.019), RXRA (P = 0.01), CCL19 (P = 0.015), and MYLK (P = 0.016)] in pretransplant sera from patients who experienced recurrence of FSGS after transplantation (n = 19) (Table 2 and Fig. 2). The difference in titer of some of these Abs [CGB5 (P < 0.0001), FAS (P < 0.0001), CD40 (P = 0.0002), PTPRO (P = 0.0005), APOL2 (P = 0.0005), P2RY11 (P = 0.0004), and SNRBP2 (P = 0.0035)] remained significant at 1 year after transplantation in rFSGS sera compared to sera from nrFSGS patients (Fig. 2) (table S2). When we performed a subanalysis on changes in Ab titers in the rFSGS group from before transplant to 1 year after, segregating for treatment response (resolution of proteinuria secondary to rituximab and intravenous cyclosporin), the reduction in titers of anti-CD40 Ab approached significance (5.3 ± 1.2 versus 3.8 ± 0.7 arbitrary units; P = 0.087), suggesting that the CD40 axis may be pathogenically relevant in the evolution of rFSGS.

To build a best-fit prognostic Ab panel for the pretransplant selection of primary FSGS patients who have a high risk of recurrence, we conducted receiver operating characteristic (ROC) analysis on log-transformed concentrations of the 10 autoAbs measured by ELISA in an independent set of pretransplant sera and fitted these to logistic regression models. A panel of seven Abs, which included CD40, FAS, PTPRO, P2RY11, CGB5, SNRPB2, and APOL2, predicted rFSGS with an area under the curve (AUC) of 0.90 [confidence interval (CI), 0.81 to 0.99]. A second model was built on three Abs (CD40, PTPRO, and CGB5) and predicted rFSGS with a slightly lower AUC of 0.82 (CI, 0.70 to 0.95). Anti-CD40 Ab was found to exert the maximal impact of any one Ab on the prediction risk of rFSGS, and even when used alone as a predictive biomarker, it had an AUC of 0.77 (CI, 0.63 to 0.92) (Fig. 2H).

Characterizing CD40 immune reactive epitopes in FSGS

Because anti-CD40 Abs were found to be the most effective for predicting rFSGS, we evaluated possible mechanisms for the altered immunogenicity of the CD40 protein in rFSGS. We hypothesized that altered immune reactivity of the CD40 antigen may play a role in rFSGS and result in the increased concentration of anti-CD40 Ab seen in rFSGS both before and after transplantation. To test this hypothesis, we used a peptide array platform to synthesize peptides of 15 amino acids each, spanning the entire CD40 protein. This customized peptide array was hybridized with sera from four rFSGS patients and four nrFSGS control patients to measure the reactivity of different epitopes in the CD40 antigen (Fig. 3) in these two patient groups with similar causes of ESRD and different clinical outcomes. The peptide microarray scan revealed altered immunogenicity of the CD40 protein in the two β strand regions (NSQCC and ESEF) specifically in the rFSGS sera, suggesting that a perturbation in the conformation of the CD40 protein may cause rFSGS.

Staining CD40 in kidney tissue by immunohistochemistry

We used rabbit polyclonal primary Ab against CD40 to perform immunohistochemistry (IHC) on human kidney tissues. No signal was observed in the podocytes in a normal human kidney (samples obtained from the tumor-free part of a total nephrectomy for renal cell carcinoma) (Fig. 4A). However, strong focal staining for CD40 was observed in the podocytes from primary FSGS (n = 2) (Fig. 4B). Strong CD40 staining was also observed in the hyperplastic podocytes (arrows) covering an rFSGS lesion of the not otherwise specified variant (Fig. 4C).

CD40 expression in human podocytes

To analyze the pathogenic relevance of CD40 in podocyte injury, the hallmark type of injury in rFSGS, we evaluated CD40 expression in cultured human podocytes and measured specific changes in CD40 expression after the podocytes were exposed to rFSGS sera. A human podocyte cell line was developed by transfection with the temperature-sensitive SV40-T gene. These cells proliferate at the “permissive” temperature (33°C) and are nondifferentiated. After transferring to the “nonpermissive” temperature (37°C), they enter growth arrest and express markers of differentiated in vivo podocytes, including podocyte proteins such as nephrin, podocin, CD2-associated protein (CD2AP), synaptopodin, and known molecules of the slit diaphragm ZO-1, α-, β-, and γ-catenin, and P-cadherin (25). We found that the CD40 protein is expressed in both nondifferentiated and differentiated cultured human podocytes (fig. S2A). Incubating podocytes with non-FSGS sera, nrFSGS sera, or rFSGS patient sera could not distinguish any changes in podocyte CD40 expression in culture (fig. S2B). Similarly, treatment of these cultured human podocytes with stimulators of podocyte injury, such as lipopolysaccharide (LPS) or tumor necrosis factor–α (TNF-α), did not alter podocyte CD40 expression (fig. S2, C and D).

Disruption of podocyte actin cytoskeleton by purified anti-CD40 Abs from rFSGS

To determine whether anti-CD40 Abs from rFSGS patients have a pathogenic role, we examined the effect of purified anti-CD40 Ab isolated from rFSGS (anti-CD40/rFSGS Ab) and nrFSGS patients on podocyte architecture. We compared this effect to that of other known triggers of podocyte injury such as LPS, puromycin aminonucleoside (PAN), and suPAR, with phosphate-buffered saline (PBS) as the negative control. We found that rFSGS sera and purified anti-CD40 Ab from rFSGS patients caused podocyte depolarization and a reduction in overall cell size, with peripheral reorganization of F-actin as assessed by phalloidin staining (Fig. 5A). Quantification analysis confirmed an overall reduction in F-actin in podocytes treated with purified anti-CD40/rFSGS Ab (Fig. 5A). We next investigated whether the addition of a commercial monoclonal CD40-blocking Ab had any effect on the observed changes in podocyte architecture after previous treatment with the anti-CD40/rFSGS Ab. We observed partial reversal of anti-CD40/rFSGS Ab-induced podocyte depolarization (Fig. 5B) after coculture with the monoclonal CD40-blocking Ab, suggesting that inhibition of binding of anti-CD40/rFSGS Ab to human podocytes may be beneficial for recovery from anti-CD40/rFSGS Ab–induced podocyte injury. suPAR, a recently identified FSGS serological factor, also causes a disruption of podocyte F-actin filaments when used to treat podocytes in culture (Fig. 5A) (8), similar to the changes in podocyte architecture seen after cotreatment of cultured podocytes with anti-CD40/rFSGS Ab. We next evaluated whether the suPAR– β₃ integrin signaling pathway was a mechanistic axis of the podocyte injury seen after treatment with suPAR and anti-CD40/rFSGS Ab. To this end, cultured human podocytes were cotreated with anti-CD40/rFSGS Ab and a monoclonal uPAR-blocking Ab, or with cycloRGDfV, a small molecule that blocks α_vβ₃ integrin activity. We observed amelioration of podocyte injury and reduced podocyte actin depolarization in both coculture experiments (Fig. 5C), suggesting that the suPAR–β₃ integrin signaling pathway is involved in anti-CD40/rFSGS Ab–induced podo-cyte injury in vitro.

Induction of proteinuria in C57BL/6 mice by human anti-CD40/rFSGS Ab

To evaluate the in vivo pathogenicity of human anti-CD40/rFSGS Ab, C57BL/6 mice were injected with either autoAb isolated from patients, such as anti-CD40/rFSGS Ab or anti-CD40/nrFSGS Ab to observe any changes in proteinuria within 8 days after injection. To control for the short half-life of injected human anti-CD40 IgG Ab (26), the Ab injection was repeated 48 hours after the first injection. Urine albumin-to-creatinine ratio (ACR) was calculated daily from day 1 to day 8 after first human anti-CD40/rFSGS Ab injections. A mild but significant increase in proteinuria over baseline was observed on day 8 with anti-CD40/rFSGS Ab injections (the increase of ACR from day 1 to day 8 with anti-CD40/rFSGS Ab injection was 31 ± 7.8 versus 4 ± 2.8 with injection of anti-CD40/nrFSGS Ab; P = 0.037) (Fig. 6A). In a separate experiment, we injected a single dose of full-length (three-domain) suPAR recombinant protein 6 hours after the second dose of the human anti-CD40/rFSGS Ab and observed a significant enhancement of proteinuria in C57BL/6 mice treated with the combination of anti-CD40/rFSGS and suPAR (increase in ACR of 346.8 ± 77.5 with anti-CD40/rFSGS Ab + suPAR versus increase in ACR of 46.3 ± 21.6 with anti-CD40/nrFSGS Ab + suPAR; P = 0.0043; baseline ACR, 102.9 ± 7) (Fig. 6B). This result shows that even three-domain suPAR, which reportedly does not cause proteinuria in wild-type mice (7, 27) when given alone, in contrast to alternative forms of suPAR (7), becomes pathogenic to podocytes in the presence of the autoAb and particularly so in the presence of anti-CD40/rFSGS Ab. Next, we injected a CD40-blocking Ab and found significant reduction in proteinuria in C57BL/6 mice that had received anti-CD40/rFSGS Ab (380.2 ± 94.3 before versus 67.8 ± 2.8 after blocking; P = 0.028) (Fig. 6C).

To further explore whether the effect of human anti-CD40 Ab in rFSGS patients was indeed mediated through CD40, we first examined the cross-species interaction of CD40 Ab and CD40 antigen and found that rFSGS patient-derived CD40 Ab interacted with mouse CD40 antigen, but anti-CD40 Abs/rFSGS or mouse anti-CD40 Abs do not bind to recombinant human CD40 antigen (fig. S3). Thereafter, we used CD40^−/− mice and administered the same amount of anti-CD40 Ab as used in all previous experiments along with human suPAR. There was no significant change in the ACR with injection of anti-CD40 Ab/rFSGS alone (Fig. 6D), but cotreatment with the three-domain suPAR significantly increased ACR in CD40^−/− mice (Fig. 6E; P = 0.048 for anti-CD40 Ab/rFSGS alone versus anti-CD40 Ab/rFSGS + suPAR). In contrast, cotreatment with suPAR and anti-CD40 Ab from nrFSGS patients did not increase ACR in CD40^−/− mice (Fig. 6E). Thus, our data suggest that anti-CD40/rFSGS Ab and suPAR may cooperate in the development of proteinuria.

Study design

A total of 141 available sera from 98 renal transplantation patients were chosen for this study to evaluate IgG Abs specific to rFSGS. All patients with rFSGS had primary FSGS as the cause of ESRD and demonstrated heavy proteinuria (nephrotic range) without evidence of acute rejection, glomerulitis, allograft glomerulopathy, or recurrence of any other primary renal disease. All FSGS cases were biopsy-proven. A panel of seven IgG Abs specific to rFSGS were validated as a biomarker panel for the pretransplant detection of FSGS recurrence. Anti-CD40 Abs isolated from rFSGS and nrFSGS were evaluated for differential antigen binding, Ab pathogenicity in vitro in a human podocyte cell culture model, and Ab pathogenicity in vivo in murine models (C57BL/6 mice and CD40-null mice). The addition of CD40-blocking Ab was evaluated for its ability to abrogate pathogenicity for the above models, in vitro and in vivo. For the ELISA validation, the samples were blinded for the person in charge of conducting the experiment and initial data processing.

Patients and samples

We processed 141 serum samples, obtained from 98 renal transplant patients (64 of whom had FSGS) before renal transplantation and 1 year after. The patients were enrolled at four international transplant centers: Stanford University Medical Center (Stanford, CA, USA); Transplantation Renale Adulte, Hôpital Necker-Enfants Malades (Paris, France); The Johns Hopkins Hospital (Baltimore, MD, USA); and Nephrology and Renal Transplantation, University Hospitals Leuven (Leuven, Belgium). The study was approved by the Institutional Review Boards of the respective institutions. Among the 98 patients, 64 had FSGS as the cause for ESRD requiring transplantation, whereas 34 had other causes of ESRD, such as obstructive uropathy, IgA nephropthy, diabetic nephropathy, hypertensive nephropathy, and glomerulonephritis. Among the 64 FSGS patients, 33 had recurrence of FSGS after renal transplantation (rFSGS), and mean time to biopsy-confirmed recurrence was 2 months ± 16 days. Serum samples were also collected at 12 months after transplantation from 43 patients with FSGS as the cause of ESRD. The detailed demographics of the nrFSGS and rFSGS groups are shown in Table 1.

Immune response biomarker profiling by protein microarrays and ELISA validation

The ProtoArray Human Protein Microarray was used for profiling serum IgG autoAb in 20 pretransplant sera from 10 patients with and without rFSGS (22). Kidney-specific Ab responses in rFSGS were mapped as previously published by our group (22, 43). The MSD technology was used for customized ELISA validation of elevated rFSGS Ab titers against FAS, CD40, CCL19, MYLK, CGB5, SNRBP2, RXRA, P2RY11, PTPRO, and APOL2. One hundred forty-one sera from 98 patients were processed for customized ELISA assays. Sera were obtained from 64 patients with FSGS as a cause of ESRD before transplantation; 33 of these patients had rFSGS within the first year after transplant, and 31 did not. Patients were demographically matched (Table 1) (Fig. 1). In a subset of these patients, follow-up sera samples were available at 1 year after transplantation (17 with rFSGS in the first posttransplant year and 26 patients without recurrence); the customized ELISA assays for all 10 autoAbs were also run on these samples to obtain a longitudinal analysis of Ab titers in the first posttransplant year.

IHC analysis of CD40 expression in human biopsies

Human kidney biopsy tissues were fixed in formalin or AFA (alcohol–formalin–acetic acid) solution and embedded in paraffin. For CD40, antigen retrieval was performed by proteinase K digestion and then hydrolysis by heating in a microwave oven with citrate buffer. The rabbit polyclonal primary anti-CD40 Ab (Anti-CD40 Ab 11E9, Abcam) was diluted at 1:150 and incubated overnight at +4°C. A three-step technique was used with biotinylated anti-rabbit secondary Ab, avidin-biotin-peroxidase complex as the vector, and 3,3′-diaminobenzidine (DAB) as the substrate.

Podocyte cell cultures

Normal human podocytes were cultured in the presence of patients’ sera and lysates for Western blotting (25), using mouse monoclonal Abs against CD40 and GAPDH (glyceraldehyde-3-phosphate dehydrogenase) (7). Cultured human podocytes were treated for 24 hours with rFSGS sera, which contained high concentrations of CD40 autoAb and separately with CD40 Ab from rFSGS patients (n = 4). Cellular F-actin expression and distribution were evaluated by phalloidin staining. For assessment of CD40 expression, some cells were lysed for SDS–polyacrylamide gel electrophoresis and Western blotting with a rabbit polyclonal CD40 Ab (Santa Cruz). Some cells were fixed with 4% para-formaldehyde (PFA) and immunostained with the same primary Ab followed by Alexa Fluor 488–conjugated anti-rabbit IgG (Life Technologies). For FSGS serum treatment, the medium of the cultured podocytes was removed, and pooled recurrent or nonrecurrent pretransplant FSGS serum was added at 4% and incubated for 24 hours. Recombinant suPAR protein (7) and a small molecule that blocks β₃ integrin activity, cycloRGDfV, were each used at 1 µg/ml for additional treatment of cultured podocytes. Monoclonal CD40 Ab (Novus Biologicals) was given at different ratios (1:1 and 3:1) relative to CD40 autoAb to assess for competition between the monoclonal CD40 and the autoAb isolated from rFSGS sera. Twenty-four hours after treatment, the cells were fixed with 4% PFA and immunostained with phalloidin (Life Technologies) and 4′,6-diamidino-2-phenylindole (DAPI). Imaging was performed by Leica deconvolution microscopy.

Examination of the pathogenicity of anti-CD40 Ab in vivo

We investigated the effect of anti-CD40 Ab (CD40 IgG) isolated from patents with rFSGS and nrFSGS in vivo by injecting these Abs into C57BL/6J mice and after the development of proteinuria. We injected anti-CD40 Ab (5 µg/ml) isolated from sera of patients with and without rFSGS into C57BL/6J mice (n = 5 in each group). To control for the short half-life of injected IgG Ab (26), anti-CD40 Ab was injected twice, 48 hours apart. ACR was calculated daily. To examine any additive effect of suPAR on renal injury, on the background of treatment with human anti-CD40 Ab injections from rFSGS and nrFSGS patients, we used female C57BL/6 mice, aged 10 weeks, with body weight ranging from 18 to 20 g. Eight mice were randomly chosen to receive intravenous injections of anti-CD40 Ab isolated from rFSGS patients, and seven were to receive human anti-CD40 Ab isolated from nrFSGS. The dose was inferred from the ratio of CD40 autoAb in rFSGS patients to that in nrFSGS patients, which is about 4:1. The estimated final concentration was 8 µg/ml for anti-CD40 Ab from rFSGS and 2 µg/ml for nrFSGS. Anti-CD40 Ab injection was given every other day, for a total of six doses. Six hours after the last dose of anti-CD40 Ab, 10 µg of recombinant human suPAR (R&D) protein was given intravenously to all mice to analyze the effect of suPAR. Twenty-four hours after the last dose of anti-CD40 Ab, blocking mouse monoclonal CD40 Ab (Santa Cruz Biotechnology Inc., cat. no. sc-59047) was administered intraperitoneally at a dose of 3 µg per mouse. Urine was collected before and every day after the first injection of anti-CD40 Ab to analyze urinary albumin with a mouse albumin ELISA kit (Bethyl Laboratories Inc.) and creatinine concentration with a urine creatinine assay kit (Cayman Chemical Company). Proteinuria is expressed as albumin (mg)/creatinine (g) ratio.

The evaluation of human anti-CD40 Ab and suPAR in CD40^−/− mice (Jackson Labs, stock no. 002928, female, 15 to 20 g, 8 to 10 weeks; n = 3 in each group) was conducted in the same way as detailed above for wild-type C57BL/6J mice.

Epitope profiling of anti-CD40 IgG

The PepStar human peptide microarray (JPT Peptide Technologies), consisting of 15-mer peptides with overlapping by four amino acids, was used to map reactive epitopes of CD40 by probing with serum samples from patients with (n = 4) and without (n = 4) rFSGS. See Supplementary Materials and Methods for additional details.

Statistical analysis

Original data for composite figures are provided in table S2. Statistical analysis of demographic data was performed with Mann-Whitney, one-way ANOVA, and Fisher’s exact tests. ELISA validation results were analyzed with a Mann-Whitney test using GraphPad Prism. Nominal logistic regression modeling was performed to compare expression of Abs between rFSGS and nrFSGS groups analyzed by ELISA. Abs with significant P value (<0.05) were selected by stepwise methods for each Ab. ROC analysis was conducted on the entire panel of autoAb to find the best autoAbs to predict rFSGS. For this purpose, we used three fitted logistic regression models with log-transformed concentrations of Abs; the model selected a panel of 7 of the 10 Abs (against CD40, FAS, PTPRO, P2RY11, CGB5, SNRPB2, and APOL2) for the best discrimination of rFSGS and a minimal set of 3 Abs (CD40, CGB5, and PTPRO) for optimal discrimination of rFSGS. The panel of seven Abs is referred to as FAST.