Antibody agonists trigger immune receptor signaling through local exclusion of receptor-type protein tyrosine phosphatases

Abstract

Antibodies can block immune receptor engagement or trigger the receptor machinery to initiate signaling. We hypothesized that antibody agonists trigger signaling by sterically excluding large receptor-type protein tyrosine phosphatases (RPTPs) such as CD45 from sites of receptor engagement. An agonist targeting the costimulatory receptor CD28 produced signals that depended on antibody immobilization and were sensitive to the sizes of the receptor, the RPTPs, and the antibody itself. Although both the agonist and a non-agonistic anti-CD28 antibody locally excluded CD45, the agonistic antibody was more effective. An anti-PD-1 antibody that bound membrane proximally excluded CD45, triggered Src homology 2 domain-containing phosphatase 2 recruitment, and suppressed systemic lupus erythematosus and delayed-type hypersensitivity in experimental models. Paradoxically, nivolumab and pembrolizumab, anti-PD-1-blocking antibodies used clinically, also excluded CD45 and were agonistic in certain settings. Reducing these agonistic effects using antibody engineering improved PD-1 blockade. These findings establish a framework for developing new and improved therapies for autoimmunity and cancer.

Keywords: antibody; autoimmunity; cancer; immune checkpoint; immune receptor; immunotherapy; kinetic-segregation model; receptor agonism.

Graphical abstract

Figure 1. Signaling by agonistic antibodies

(A) The kinetic-segregation model-based explanation for antibody signaling. By binding receptors (e.g., CD28) close to the membrane, FcR-engaged strong antibody agonists (e.g., JJ316; left) create small (<22 nm) gaps between apposing cells that locally exclude large RPTPs, which would otherwise oppose receptor phosphorylation by kinases. By contrast, partial agonists (e.g., JJ319; right) that bind to the “tops” of the receptors generate larger (>22 nm) gaps that less efficiently exclude the RPTPs, producing much weaker signaling. (B) IL-2 production induced by indirectly immobilized JJ316 (orange) and JJ319 (green) anti-CD28 antibodies (10 μg/mL), either in the absence (dashed lines) or the presence of increasing amounts of immobilized mitogenic KT3 anti-CD3ε antibody (solid lines), in DO11.10 murine T-hybridoma cells expressing a rat-mouse chimeric form of CD28. The black dashed line indicates the amount of IL-2 produced in the absence of antibody. (C) Titration of the JJ316, JJ319, and KT3 antibodies. (D and E) IL-2 production by Yae5b3k (D) and TCR-expressing (TCR⁺) BW (E) cells induced by indirectly immobilized JJ316 and JJ319 anti-CD28 antibodies (see B). (F) Effects of varying the expression of the TCR (~225 [Lo, low], 500 [Int, intermediate], and 1,160 [Hi, high] receptors/cell) and full-length CD28 (~2,380 [Lo], ~11,030 [Int], and ~38,660 [Hi] receptors/cell), on signaling by BW cells induced by JJ316 (10 μg/mL). BW cells expressing intermediate TCR and CD28 levels were used for experiments in Figures 1G–1J, 2, S1H, and S1I. Primary T cells are estimated to express ~9,000 copies of CD28. (G) Calcium signaling responses of CD28-transduced DO11.10, Yae5b3k, and TCR⁺ BW cell lines loaded with Fluo-4 and stimulated with KT3 directly coated onto glass coverslips, or with JJ316 or JJ319 antibody indirectly immobilized via DAM; 1 μM ionomycin was used to confirm the signaling capacity of the TCR⁺ BW cells. Error bars are standard deviations (SDs). (H) Staining of cells for DAM adsorption from DAM antibody-coated surfaces following culture for 8 h under the conditions described in (B), in the presence of the indicated antibodies at 10 μg/mL, using a FITC-labeled rabbit anti-donkey IgG antibody. (I) Staining of cells for CD69 expression prior to (0 h) and following an 8 h culture as in (H), using a PE-Cy7-labeled anti-mCD69 antibody. (J) Time course of CD69 expression, measured as in (I), during a 24 h culture as in (H); MFI, mean fluorescence intensity. Signaling data were fitted to a binding model as indicated in Table S1. Values for IL-2 produced in the absence of antibodies, indicated by broken lines in (B), (D), and (E), were included in the fitting and analysis. The data are representative of 2–4 independent replicate experiments.

Figure 2. Dependence of signaling by agonistic antibodies on their immobilization and on kinase, receptor, RPTP, and antibody dimensions

(A) IL-2 production by BW cells expressing intermediate levels of CD28 (or tCD28) and the TCR (see Figure 1F), treated with JJ316 antibody following expression of compact or extended forms of hemagglutinin (HA)-tagged Lck (left); Lck expression levels were comparable at MFI values of 4,071 and 3,471 measured with a PE-conjugated anti-HA antibody, respectively. The stippling (right) marks the contribution of CD28 to the increased signaling in the presence of TMLck. Figure S1G explains why there is signaling in the presence of tCD28. (B) Signaling effects of immobilized and soluble forms of intact or Fab′₂ fragments of the JJ316 antibody (at 10 μg/mL), measured with TCR⁺ BW cells expressing intermediate levels of CD28 (HC, heavy chain; LC, light chain; 2°, secondary). Error bars represent SD. (C) Effects of JJ316, JJ319, and KT3 antibodies coupled indirectly at high levels to Ni-NTA-coated plastic via histidine-tagged sFcR, on signaling by CD28-expressing TCR⁺ BW cells. (D) Schematic showing how the dimensions of CD28, the RPTPs CD45 and CD148, and the JJ316 antibody were altered. (E and F) Signaling effects of JJ316 and IC10 (anti-hIgG1 Fc) antibodies on TCR⁺ BW cells expressing intermediate (E) and high (F) levels of a form of CD28 (FcCD28) extended via the insertion of the Fc region of hIgG1 at the junction between the extracellular and transmembrane regions of the receptor. The inter-mediate and high levels of expression of FcCD28 were comparable to those for CD28 in Figure 1F. (G and H) Effect of co-expressing truncated forms of the RPTPs CD45 (G) and CD148 (H) on signaling by CD28-expressing TCR⁺ BW cells. (I) Effects on signaling of extending the hinge region of JJ316 with 30 or 50 residues of mucin-like sequence. Signaling data were fitted to a binding model as indicated in Table S1. Values for IL-2 produced in the absence of antibodies, indicated by broken lines in (E), (F), and (I), were included in the fitting and analysis. The data are representative of 2–4 independent replicate experiments.

Figure 3. Anti-CD28 antibodies locally exclude CD45 from sites of contact

(A) Schematic showing the setting for analyzing the effects of anti-CD28 antibodies on CD45 distribution at contacts of T cells with DAM-coated glass surfaces, using TIRF imaging. (B) TIRF images of tCD28-expressing BW cells labeled with Alexa Fluor 647-tagged JJ316 or JJ319 antibody (red) and Alexa Fluor 488-tagged YW62.3.20 anti-CD45 Fab fragments (blue), interacting with DAM-coated coverslips. Note that antibody fluorescence was strongly correlated with membrane (CellMask) staining (Figure S3A), so dark regions in the antibody channel correspond to parts of the cell outside the evanescent field. (C–E) Average CD45 fluorescence intensities for each cell (C), histograms showing the probability density for the pixelwise CD45 fluorescence intensity for all cells (D), and average CD45 intensities in regions of high antibody fluorescence (E; N = 46 cells [JJ316], N = 49 cells [JJ319]). In the violin plots, dashed lines indicate the median, and dotted lines the quartiles. (F–H) As in (C)–(E), comparing the effects of JJ316 antibody (N = 65 cells) and JJ316 with a 50-residue extension (JJ316_Plus50; N = 69 cells). (I) GVD analysis of CD45 versus antibody distribution. The degree to which the antibodies and CD45 tend to localize in different regions is indicated by the color scale (yellow, high; blue, low). AU, arbitrary units. See Figure S5 for a description and for simulation-based validation of the GVD method. (J) Schematic showing the setting for analyzing the effects of anti-CD28 antibodies on CD45 distribution at contacts of T cells with sFcR-antibody-presenting SLBs. (K and L) Mask-based analysis of CD45 exclusion from regions of antibody accumulation for (K) cells that interacted with JJ316 (N = 142 cells) versus JJ319 antibody (N = 133 cells), and (L) cells that interacted with JJ316 (N = 114 cells) versus JJ316_Plus50 antibody (N = 95 cells), on the sFcR-antibody-presenting SLBs. Scale bars are 5 μm. Data were combined from experiments performed over 3 separate days. A two-sample Student’s t test was used for statistical comparisons. For these experiments, tCD28 expression matched that of TCR^Int-CD28^Hi BW cells (see Figure 1F).

Figure 4. Effects of anti-PD-1 antibodies on CD45 exclusion and signaling in vitro

(A) Positions of the epitopes of the clone 19 (purple), clone 2 (blue), and nivolumab (orange) anti-PD-1 antibodies and of the binding site for PD-L1 (yellow) on PD-1, relative to the membrane. (B and C) Mask-based analysis of CD45 exclusion from regions of high antibody fluorescence intensity for (B) cells treated with clone 19 (N = 204 cells) and clone 2 (N = 212 cells), and (C) cells treated with nivolumab expressed as a mIgG1 antibody (Nivo_mIgG1; N = 58 cells) or as an extended mIgG1 (Nivo_Plus; N = 61 cells). The setting for the imaging experiments was analogous to that in Figure 3J. In the violin plots, dashed lines indicate the median, and dotted lines the quartiles. Scale bar is 5 mm. Data were combined from experiments performed over 3 separate days. A two-sample Student’s t test was used for statistical comparisons. (D) Impact of anti-PD-1 antibodies on IFNγ production by human PBMCs activated with anti-CD3 and anti-CD28 antibodies. Each symbol represents a different healthy donor, with IFNγ levels normalized to the “no antibody” condition for each donor. One-way ANOVA with Dunnett’s multiple comparison follow-up testing was used to compare each group to the mIgG1 isotype control. (E–H) Effects of clone 19, clone 2, Nivo_mIgG1, Nivo_D265A, and Nivo_Plus on T cell activation in a co-culture NFAT reporter system. Jurkat T cells expressing hPD-1 and a luciferase reporter driven by an NFAT response element (NFATrep) were cultured with TCS cells expressing an anti-CD3 (OKT3) scFv antibody construct. Signaling effects of PD-1 antibodies were measured in four settings: Jurkat PD-1 cells cultured with (E) TCS cells expressing mFcγR2b, (F) TCS cells expressing PD-L1, (G) TCS cells expressing PD-L1 and mFcγR2b, and (H) unmodified TCS cells. The dashed horizontal line in each plot indicates the level of luciferase production by resting Jurkat T cells. (I) Effects of hIgG4 isotype PD-1 antibodies, i.e., humanized clone 19 and nivolumab and pembrolizumab biosimilars (bios), on T cell activation in the reporter system with PD-1-expressing Jurkat T cells cultured with TCS cells expressing hFcγR2b. Error bars shown represent SD.

Figure 5. Signaling effects of PD-1 agonists

(A) Schematic showing the setting for analyzing the effects of anti-PD-1 antibodies on CD45 and SHP2 distribution at contacts of T cells with sFcR-antibody- or sPD-L1-presenting SLBs, using TIRF imaging. (B) TIRF images of TCR^- Jurkat T cells expressing a fluorescent form of SHP2 (SHP2Halo) and wild-type PD-1 (PD-1WT), PD-1 lacking its cytosolic domain (PD-1Δcyt), or PD-1 with its cytosolic tyrosine residues mutated to phenylalanine (PD-1Y>F). The cells were incubated with Janelia Fluor HaloTag 646 ligand to label SHP2Halo (green) and Alexa Fluor 555-tagged Gap8.3 anti-CD45 Fab fragments (blue) before interacting with Alexa Fluor 647-tagged clone 19 (presented by sFcR) or sPD-L1 (red) on SLBs. (C) Violin plots of CD45 (blue) and SHP2 (green) mask-based exclusion values comparing different antibody-mediated contacts: WT-clone 19 (N = 101 cells), PD-1Δcyt-clone 19 (N = 122 cells), WT-clone 2 (N = 110 cells), WT-Nivo_mIgG1 (N = 78 cells), and WT-Nivo_Plus (N = 83 cells), for cells expressing either PD-1WT (WT) or PD-1Δcyt (Δcyt). (D) Violin plots of CD45 (blue) and SHP2 (green) exclusion values comparing clone 19 and PD-L1 mediated contacts: WT-clone 19 (N = 193 cells), Y>F-clone 19 (N = 173 cells), WT-PD-L1 (N = 136 cells), and Y>F-PD-L1 (N = 168 cells), for cells expressing either PD-1WT (WT) or PD-1Y>F (Y>F). In (C) and (D), the Kruskal-Wallis test with Dunn’s multiple comparison follow-up testing was used to compare each group to the WT control. (E) Dot plots comparing the CD45 and SHP2 exclusion values for cells expressing PD-1WT (gray), PD-1Y>F (orange), or PD-1Y>F (pink), following clone 19 (top) or PD-L1 (bottom) mediated contact. (F) Structure of the PD-1 (white)-clone 19 Fab (purple) complex (left; PDB: 8eq6) compared with the positioning of nivolumab (orange; PDB: 5WT9) and pembrolizumab (green; PDB: 5JXE) Fab fragments bound to PD-1 (right). The position of PD-1 in the right versus the left panel differs by a 45° anti-clockwise rotation in the plane of the page. (G) Ribbon representation of PD-1 with the clone 19 epitope marked in purple. (H) Structural differences between apo PD-1 (PDB: 3RRQ) and PD-1 in the clone 19 Fab-PD-1 complex, mapped onto PD-1 from the Fab-PD-1 complex. Thickness of the putty cartoon representation corresponds to the distance between equivalent Cα atoms after superposition. Distances vary between 0.08 and 8.5 Å.

Figure 6. In vivo effects of anti-PD-1 antibodies

(A) Effects of clone 19 and nivolumab derivatives on the antigen-specific expansion of humanized vs. WT PD-1 expressing and non-expressing CD4 T cells from OT-II mice following immunization with ovalbumin. A 50:50 mix of cells was injected into recipient mice on day 0, which were then immunized intraperitoneally (IP) with 100 μg of ovalbumin on day 1 and treated IP with 200 μg of isotype control or PD-1 antibodies on day 2. The ratio of cells in the spleens of the mice on day 8, quantified by flow cytometry and normalized to the average ratio in isotype control treated mice (dashed line), is shown. (B) Effects of clone 19 on KLH-induced DTH. Mice were immunized with KLH antigen on day 0, 1 h after treatment with anti-PD-1 or isotype control antibody (10 mg/kg), and then challenged intradermally with KLH in one ear on day 5. The difference in biopsy weight between the challenged and unchallenged ear (dashed line) in different treatment groups measured on day 6 is shown. (C) Effects of clone 19 on the BM12 transfer model of SLE. Splenocytes from huPD-1 mice were transferred IP into BM12 recipient mice, which were then treated with 10 mg/kg anti-PD-1 or isotype control antibody the next day. On day 35, markers of SLE disease severity were assessed including splenomegaly (left), Tfh cell expansion in the spleen quantified by flow cytometry (middle), and serum auto-antibody levels assessed by ELISA (right). Each point represents an individual mouse. Error bars shown represent SD. For each model, data are representative of two independent experiments. One-way ANOVA with Dunnett’s multiple comparison follow-up testing comparing each group to the isotype control, was used for statistical comparisons.