Preclinical evaluation of antigen-specific B-cell depletion for phospholipase A2 receptor membranous nephropathy with chimeric autoantibody receptor T-cells

Cells

293T and Jurkat cells were obtained from the University of Pennsylvania. Primary human T-cells were provided by the Human Immunology Core (UPenn). PBMCs from MN patients were collected with informed consent. Anti-PLA2R hybridomas were obtained from Genscript, and Nalm-6 cells from ATCC.

Stimulation, transduction, and expansion of T-cells

Primary T-cells and Jurkat cells were cultured in RPMI with 10% FBS and supplements. T-cells were stimulated with anti-CD3/CD28 beads and IL-2, then transduced with CAAR or control lentivirus and expanded for 10–14 days. CAAR expression was confirmed by flow cytometry.

Animal experiments

NSG mice (6–8 weeks, female) were injected with luciferase+ anti-PLA2R Nalm-6 cells and treated with C17- or C178-CAART (3.5–7×10⁶ cells). Tumor burden was monitored by bioluminescence imaging, and T-cell persistence assessed in spleen and bone marrow. A total of 39 mice were used across three independent experiments. All procedures were approved by IACUC.

Additional details for all methods are provided in the Supplementary Methods.

Preliminary Design and Validation of PLA2R-CAART Against Anti-PLA2R B-cells

Phospholipase A2 Receptor (PLA2R) is a transmembrane protein belonging to the mannose receptor superfamily. It contains cysteine-rich (CysR), fibronectin type II (FnII), and eight C-type lectin-like domains (CTLD1–8) in its extracellular region (Figure 1A). The CysR domain comprises the major immunodominant epitopes in PLA2R MN¹⁵, whereas CTLD1 and CTLD7 immunoreactivity are more likely to be observed in refractory cases with poor renal prognosis^16,17, The CysR domain interacts with the CTLD3 domain to form a ring-like structure¹⁵ (Figure 1A), raising the possibility that these sequences may adopt a compact conformation important for PLA2R immunoreactivity. However, a smaller PLA2R fragment consisting of only the CysR, FnII, and CTLD1 domains was sufficient to adsorb all anti-PLA2R immunoreactivity in PLA2R MN patient sera¹⁸.

CAAR design should incorporate immunodominant epitopes while avoiding domains that could lead to undesirable off-target interactions. Additionally, the size of the CAAR extracellular domain may affect cytotoxic activity, as larger intercellular distances between T-cells and target cells do not exclude CD45 phosphatase from the immune synapse, thereby limiting T-cell receptor signaling¹⁹. We therefore evaluated three initial CAAR constructs: CF1237 (comprising CysR, FnII, CTLD1, CTLD2, CTLD3, and CTLD7), CF17 (CysR, FnII, CTLD1, and CTLD7), and C17 (CysR, CTLD1, and CTLD7), the latter designed to exclude the FnII domain due to its potential interactions with collagen and integrin β1 and its lack of reported epitope activity in PLA2R MN. Lentiviral transduction of CAAR constructs into primary human T-cells resulted in robust cell surface expression (Figure S1A).

Anti-PLA2R B-cell receptor (BCR) sequences from PLA2R MN patients have not previously been reported. To generate anti-PLA2R B-cells targeting immunodominant PLA2R epitopes, we immunized mice with a recombinant human PLA2R protein fragment containing CysR, FnII, CTLD1, and CTLD7 domains. Nine hybridoma cell lines with reactivity to human PLA2R by ELISA were identified (Figure S1B). We generated Jurkat cell lines expressing CysR, CF1, and C17 CAAR extracellular domains and deduced the epitope reactivity of each hybridoma clone by flow cytometry. Three clones (2B2, 12G10, and 13D10), each targeting distinct PLA2R domains (CTLD7, CTLD1, and CysR, respectively), were sequenced and selected for subsequent cytotoxicity assays (Figure S1C).

Preliminary cytotoxicity assays revealed that both C17-CAART and CF17-CAART induced cytolysis of anti-PLA2R hybridomas (Figure S1D). In contrast, CF1237-CAART demonstrated limited cytotoxicity associated with no IFN-γ release, comparable to negative control co-cultures (Figure S1E). Given the potential risk of off-target binding of the FnII domain and the lack of cytotoxic activity from the CF1237 construct, we selected C17 for further development based on its preserved cytotoxicity across a range of soluble anti-PLA2R antibody concentrations (Figure S1F).

During the course of this study, a report identified that 16% of PLA2R MN patient sera demonstrate CTLD8 reactivity²⁰, although two prior epitope mapping studies did not identify CTLD8 reactivity in PLA2R MN sera^15,16, and CTLD8 reactivity does not correlate with the total anti-PLA2R antibody level²⁰, suggesting that anti-CTLD8 antibodies, if present, comprise a minority of the anti-PLA2R antibodies within each patient. Nevertheless, we designed and evaluated an additional CAAR construct, C178, which comprises the CysR, CTLD1, CTLD7, and CTLD8 domains of PLA2R (Figure 1B).

Both C17- and C178-CAART efficiently expressed their respective PLA2R epitopes on the T-cell surface, with C17-CAART showing more consistent surface expression during T-cell expansion (Figure 1C–D). We assessed their cytotoxic activity against hybridomas expressing BCRs specific for CysR, CTLD1, or CTLD7 domains. Both C17- and C178-CAART demonstrated antigen-specific killing of all three hybridoma targets, accompanied by comparable IFN-γ secretion (Figure 1E–1F), with C17-CAART consistently exhibiting slightly higher cytolytic potency across all targets.

Soluble Autoantibody Effects on PLA2R-CAART Cytotoxicity

Given that MN patients demonstrate circulating anti-PLA2R IgG that could potentially interfere with CAART cytolysis, we further compared the cytotoxic activity of C17- and C178-CAART in the presence of soluble monoclonal IgG targeting distinct PLA2R domains (Figure 2A), a situation in which the soluble IgG would be neutralizing, as well as patient-derived MN plasma IgG (Figure 2B–2C). The physiologic range of anti-PLA2R IgG observed in MN patients was below 100 μg/mL in 118 of 120 reported patients, with a mean level of 9–16 μg/mL^21,22,15. A recent study reported mean anti-PLA2R IgG levels of 210 μg/mL in MN patients compared to 120 μg/mL in healthy controls²³, the latter being higher than most of the previously reported MN patients. We therefore tested CAART function in the presence of soluble monoclonal IgG at concentrations up to 100 μg/mL. After 4 hours of co-culture with target cells, a modest reduction in cytotoxicity was observed when PLA2R-CAARTs were exposed to soluble monoclonal anti-PLA2R IgG, especially when the specificity of the soluble IgG matched that of the monoclonal BCR target. However, after 20 hours of incubation, decreased inhibition of cytotoxicity was observed. Negligible cytotoxicity was observed with negative control co-cultures (NTD T-cells with anti-PLA2R target cells, or PLA2R-CAART with anti-desmocollin 1 (Dsc1) hybridoma cells) (Figure 2A). Co-incubation of polyclonal MN plasma IgG in PLA2R-CAART killing assays at a physiologic concentration of up to 10 mg/mL IgG caused a modest reduction in cytotoxicity at the 4-hour time point for both C17- and C178-CAART at an E:T ratio of 30:1, with more pronounced inhibition at a lower E:T ratio of 3:1. After 20 h of co-incubation, cytotoxicity at 30:1 was largely restored, whereas at 3:1 it was partially restored, indicating that longer co-incubation time can partially overcome the soluble autoantibody blockade (Figure 2B–2C).

PLA2R-CAARTs Effectively Eliminate Anti-PLA2R B-Cells and Engraft In An In Vivo Xenograft Model

CAART therapy aims to selectively eliminate B-cells producing pathogenic autoantibodies. However, establishing an in vivo model for CAART targeting anti-PLA2R in membranous nephropathy (MN) is challenging. Mouse PLA2R is not expressed in the murine podocyte²⁴ and human anti-PLA2R IgG does not cross-react with mouse PLA2R²⁵, limiting translational studies using human PLA2R MN plasma samples to induce clinical disease. Another study developed a human PLA2R transgenic mouse model using the adenoviral EIIa ubiquitous promoter²⁶. We performed immunohistochemical analysis of kidneys of these mice, but did not observe expression of human PLA2R on the podocyte membrane (Figure S2A).

We therefore used an anti-PLA2R Nalm-6 xenograft model for the focused purpose of comparing the in vivo cytotoxic activity and evaluating the engraftment capability of PLA2R-CAART compared to a clinically-approved anti-CD19 CART (CART19) in a well-established model. Nalm-6 cells were engineered to express anti-PLA2R BCRs. In initial in vitro cytotoxicity assays in which CART19, C17-CAART, and C178-CAART demonstrated similar levels of surface CA(A)R expression (Figure 3A), C17-CAART specifically killed Nalm-6 cells expressing anti-PLA2R BCRs (12G10, 13D10, and 2B2), comparable to CART19 positive control cells (Figure 3B). Kinetics of C178-CAART cytolysis were delayed compared to CART19 after 4 hours of target cell co-culture, but were comparable after 20 hours.

Having validated the in vitro cytolytic activity of C17- and C178-CAARTs in vitro, we next compared the ability of C17- and C178-CAART to eliminate anti-PLA2R Nalm-6 cells and engraft in vivo, in comparison to anti-CD19 CARTs. On Day 0, an equimolar mixture of Nalm-6 cells expressing anti-PLA2R BCRs (2B2, 12G10, and 13D10, 1 × 10⁶ total) were injected into mice, followed by treatment with either C17- or C178-CAARTs, CART19 (positive control), or NTD T-cells (negative control) on Day 4 (7 × 10⁶ total T-cells) (Figure 4A). Serial bioluminescence imaging (BLI) was conducted to monitor target cell burden in vivo. Both C17- and C178-CAART, along with CART19 positive control, significantly reduced Nalm-6 target cell burden compared to the NTD control group. Target cell outgrowth, measured by total flux (photons/second), was significantly lower in the CAART-treated groups, with C17-CAART showing slightly higher reduction in target cell burden compared to C178-CAART (Figures 4B–4C). Comparable results were observed in two additional independent experiments (7 × 10⁶ T-cells/mouse, Figure 4B-experiment 2, and 3.5 × 10⁶ T-cells/mouse, Figure 4B-experiment 3).

To evaluate in vivo persistence, we analyzed the frequency of CD3⁺ and CAAR⁺ T-cells in bone marrow and spleen at the time of study termination. CD3⁺ T-cell frequencies were comparable across C17-, C178-CAART, and CART19 groups in bone marrow, while CART19 showed significantly higher CD3⁺ frequencies in the spleen (Figure 4D, S3). As expected, CAAR⁺ cells were only detected in the C17- and C178-CAART groups, with no CAAR⁺ population in the CART19 group. Notably, C17-CAART, but not C178-CAART, showed a significantly higher proportion of persisting CAAR⁺ cells than the non-CAAR expressing CART19 control. Although differences in persistence of C17- and C178-CAART in both bone marrow and spleen were not statistically significant, these trends suggest that C17 may exhibit improved in vivo persistence or surface CAAR stability, supporting its selection as the final candidate for further development.

C17- and C178-CAART Adsorb Anti-PLA2R Antibodies and Reduce anti-PLA2R B-Cells in MN Patient Samples In Vitro

We next evaluated C17- and C178-CAART against PLA2R MN samples. We first assessed whether C17- and C178-CAART could adsorb the majority of plasma anti-PLA2R antibodies, which would indicate that they target the majority of anti-PLA2R BCRs in MN patients. Plasma levels of anti-PLA2R antibodies were measured in eight MN patient samples before and after incubation with C17- and C178-CAART, with NTD T-cells serving as a negative control. C17- and C178-CAART reduced anti-PLA2R antibody levels by 57% to 96% (median 87.9% adsorption with C178-CAART compared to 84.3% median adsorption with C17-CAART, Figure 5A). The comparable immunoadsorption profiles suggest that CTLD8-reactive antibodies do not comprise a substantial portion of total anti-PLA2R antibodies within each patient.

We next assessed the antigen-specific cytotoxic efficacy of C17- and C178-CAART against primary B-cells from PLA2R MN patients, the intended target cells in a therapeutic setting. Anti-PLA2R B-cells were identified in B-cell samples from eight MN patients, with their proportion among IgG⁺ B-cells found to be 0.08% to 0.54% (median 0.32%, Figure 5B, S4A). Co-culture experiments with IgG+ B-cells from MN patients revealed anti-PLA2R B-cell killing by both C17- and C178-CAART, 79.3– 93.2% reduction with C17-CAART and 67.9– 81.1% reduction with C178-CAART, with C17-CAART showing a slight advantage in killing efficiency (Figure 5C–D, S4B). These results suggest that C17- and C178-CAART can bind the vast majority of anti-PLA2R BCRs and demonstrates cytotoxic activity against primary anti-PLA2R B-cells in PLA2R MN patients.

PLA2R-CAAR did not demonstrate specific off-target cytotoxicity

To evaluate potential off-target cytotoxic effects of C17- and C178-CAART in vivo, we performed comprehensive histopathological analysis in comparison to donor-matched CART19-treated controls. Mononuclear cell infiltration scores, including large pleomorphic cells possibly reflecting activated T-cells or Nalm-6 cells, were generally similar across most tissues between PLA2R-CAART– and CART19-treated mice (Figure 6A–B). Lymphoid organs such as lymph nodes and spleen showed consistently high infiltration in all groups without cytotoxicity, reflecting engraftment in these tissues. C17-treated mice exhibited slightly higher infiltration in the liver, salivary/lacrimal glands, and pancreas compared to both C178 and CART19. Despite these tissue-specific differences, the overall infiltration burden did not differ significantly between groups (p=0.76), although a mild increase in liver cell injury was observed in C17-CAART compared to C178-CAART and CART19-treated groups, generally reflecting the level of liver cell infiltration (Figure 6A versus Figure 6C). This finding was attributed to xenogeneic graft-versus-host reaction. Mild, sporadic degenerative changes were observed in renal tubules, pancreatic acini, and liver biliary tracts, with no lesion uniquely enriched in the PLA2R-CAART-treated groups (Figure 6C). Sternal fracture or rib dislocation in 2 mice treated with C178-CAART or CART19, respectively, were attributed to trauma from mouse handling (Figure 6D). Other histopathologic findings and organ weights were also comparable across treatment groups, suggesting no substantial organ inflammatory response or immune cell infiltration associated with PLA2R-CAART treatment (Figure 6D–E).

As an unbiased method to identify unexpected off-target interactions against human proteins, we performed a high-throughput membrane proteome array (MPA) screen of approximately 6,000 human membrane proteins, using a recombinant chimera of C17 or C178 linked to the human antibody Fc domain (hFc), and the anti-PLA2R antibody 10G7 as a positive control. Both C17-hFc and C178-hFc constructs demonstrated specific binding to Fc receptors FCGR2B and FCGR1A, which are known to bind human Fc, with no significant additional off-target interactions detected. The binding profiles of C17 and C178 indicated minimal differences in target engagement between the two constructs (Figure 6F).