Structure-guided disulfide engineering restricts antibody conformation to elicit TNFR agonism

Abstract

A promising strategy in cancer immunotherapy is activation of immune signalling pathways through antibodies that target co-stimulatory receptors. hIgG2, one of four human antibody isotypes, is known to deliver strong agonistic activity, and modification of hIgG2 hinge disulfides can influence immune-stimulating activity. This was shown for antibodies directed against the hCD40 receptor, where cysteine-to-serine exchange mutations caused changes in antibody conformational flexibility. Here we demonstrate that the principles of increasing agonism by restricting antibody conformation through disulfide modification can be translated to the co-stimulatory receptor h4-1BB, another member of the tumour necrosis factor receptor superfamily. Furthermore, we explore structure-guided design of the anti-hCD40 antibody ChiLob7/4 and show that engineering additional disulfides between opposing F(ab') arms can elicit conformational restriction, concomitant with enhanced agonism. These results support a mode where subtle increases in rigidity can deliver significant improvements in immunostimulatory activity, thus providing a strategy for the rational design of more powerful antibody therapeutics.

Fig. 1. C-S disulfide variants exhibit the same trend in agonistic activity and conformation in anti-hCD40 and anti-h4-1BB mAb.

a Model of disulfide arrangements in hIgG2 mAb affecting agonistic activity and flexibility, derived from studies of the hIgG2 anti-hCD40 ChiLob7/4_. Disulfides shown in yellow, and cysteines involved in disulfide bonds labeled (with the Kabat numbering system). hIgG1 and hIgG2 wildtypes are shown inset above. b NF-κB/Jurkat/GFP reporter cells expressing hCD40 or h4-1BB were stimulated with serially diluted ChiLob7/4 or SAP1.3 mAb, respectively. NF-κB activation triggers GFP expression and was quantified after 24 h by flow cytometry (inset histograms representative from n = 3 at 0.04 μg/ml for ChiLob7/4 and 5 μg/mL for SAP1.3). Graphs show dose-response curves of the percentage of GFP-positive cells. n = 3–7 independent biological experiments; mean ± SEM, taken from technical triplicates for each independent experiment. hIgG1 (gray) and hIgG2 (black) shown as controls. c F(ab’)₂ of anti-hCD40 and anti-h4-1BB C-S disulfide variants were analyzed by SEC-SAXS. Graphs show dimensionless Kratky plots derived from SEC-SAXS data, with the Guinier-Kratky point (√3, 1.103) indicated by the pale gray crosshairs. hIgG1 (gray) and hIgG2 (black) shown as controls. Errors calculated following standard BioXTAS RAW software procedures. Disulfide C-S antibody variants labeled by color: red C232S + C233S, purple C233S κC214S, blue C232S κC214S. Source data are provided as a Source Data file.

Fig. 2. Structure-guided design of anti-hCD40 hIgG2 mAb with additional engineered disulfides.

a Schematic of the parental agonistic hIgG2 C232S κC214S variant (cross-over) showing experimentally determined disulfides as solid yellow lines. Disulfides not resolved in the structure (PDB: 6TKE) are shown as dashed yellow lines. Cysteine amino acid residues are labeled. b Structure-based design and predicted disulfides for C232S + K228C κC214S (cross-over + K228C) and C232S + T222C κE123C + κC214S (cross-over + T222C κE123C) shown in yellow, with disulfides from the parent cross-over variant in gray. c Experimentally determined crystal structures of the new F(ab’)₂ variants shown as surface representation, with disulfides as sticks. d Sulfur single-wavelength anomalous dispersion (Sulfur SAD) crystallography reveals the position of sulfur atoms and confirms disulfides between neighboring chains. The anomalous electron density is shown as green mesh (anomalous difference Fourier map, contoured at 5 σ). Disulfides shown in yellow as sticks. Engineered antibody variants labeled by color: blue C232S κC214S (cross-over), teal cross-over + K228C, orange cross-over + T222C κE123C.

Fig. 3. Conformationally rigid and compact nature of engineered anti-hCD40 variants revealed through SAXS fits to models extracted from MD simulations.

F(ab’)₂ were analyzed by SEC-SAXS. a Graphs show dimensionless Kratky plots derived from the SEC-SAXS, with the Guinier-Kratky point (√3, 1.103) indicated by the pale gray crosshairs. Errors calculated following standard BioXTAS RAW software procedures. b Models for each F(ab’)₂ were extracted from 6 μs of MD simulation, every 1 ns. Agreement of the calculated scattering curves for the extracted models to the experimental scattering data were calculated using CRYSOL, with the best fitting single model shown (calculated scattering shown in color, experimental scattering shown as gray dots). χ² fit with error-weighted residuals plot also shown. c Conformation of the best fitting single model shown, with heavy chain in a darker shade, light chain in a lighter shade. d For variants 3–5, the hinge angle for best fitting single model shown. For variants 1-2, the mean hinge angle (± SD) for the best-fitting GAJOE-selected ensemble shown. e For variants 3–5, R_g for best fitting single model shown. For variants 1-2, mean R_g (±SD) for the best fitting GAJOE-selected ensemble shown. See Supplementary Fig. 15c. f For variants 3–5, D_max for best fitting single model shown. For variants 1-2, the mean D_max (± SD) for the best-fitting GAJOE-selected ensemble shown. See Supplementary Fig. 15d. Engineered antibody variants labeled by color: red C232S + C233S, purple C233S κC214S, blue C232S κC214S (cross-over), teal cross-over + K228C, orange cross-over + T222C κE123C. Source data are provided as a Source Data file.

Fig. 4. Receptor binding and agonistic activity of engineered anti-hCD40 mAb.

a Serially diluted ChiLob7/4 variants were incubated with Jurkat cells expressing hCD40 with binding detected with a secondary PE-conjugated antibody by flow cytometry. Graphs show binding dose-response curves of geometric mean fluorescence intensity (gMFI). Mean ± SEM, n = 2 independent biological experiments, mean taken from technical triplicate for each independent experiment. b NF-κB/Jurkat/GFP reporter cells expressing hCD40 were stimulated with serially diluted ChiLob7/4 variants, and activation quantified after 24 h by determination of GFP expression levels, assessed by flow cytometry. The graph shows dose-response curves of percentage GFP + cells for the variants. Mean ± SEM, n = 2 independent biological experiments, mean taken from technical triplicate for each independent experiment. ** p < 0.01 at 0.008 μg/mL, one-way ANOVA with Tukey’s multiple comparisons tests (for p-values and significance levels for other concentrations, see Supplementary Table 10). c–g ChiLob7/4 variants were incubated with human B cells, and various activation assays performed. c Homotypic adhesion of human B cells measured 48 h after addition of 0.008 μg/mL ChiLob7/4 mAb. Images are representative of technical triplicates from 1 of 3 independent experiments. Scale bar 200 μm. d–f Activation of primary human B cells determined by upregulation of (d) HLA-DR, (e) CD86, (f) CD23, using flow cytometry; measurements taken 48 h after addition of 0.008 μg/mL ChiLob7/4 mAb. g Proliferation of human B cells assessed by ³H-thymidine incorporation 96 h after addition of 0.008 μg/mL ChiLob7/4 mAb. CPM = counts per minute. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001 one-way ANOVA with Tukey’s multiple comparisons tests (for exact p-values, see Supplementary Table 11). For c, data show representative images from technical triplicate from 1 of 3 independent experiments with independent donors. For d-g, data show technical triplicates (mean ± SEM) from 1 of 3 independent experiments with independent donors. Engineered antibody variants labeled by color: blue C232S κC214S (cross-over), teal cross-over + K228C, orange cross-over + T222C κE123C. hIgG1 (gray) and hIgG2 (black) shown as controls. Source data are provided as a Source Data file.