Protease-activation using anti-idiotypic masks enables tumor specificity of a folate receptor 1-T cell bispecific antibody

Abstract

T-cell bispecific antibodies (TCBs) crosslink tumor and T-cells to induce tumor cell killing. While TCBs are very potent, on-target off-tumor toxicity remains a challenge when selecting targets. Here, we describe a protease-activated anti-folate receptor 1 TCB (Prot-FOLR1-TCB) equipped with an anti-idiotypic anti-CD3 mask connected to the anti-CD3 Fab through a tumor protease-cleavable linker. The potency of this Prot- FOLR1-TCB is recovered following protease-cleavage of the linker releasing the anti-idiotypic anti-CD3 scFv. In vivo, the Prot-FOLR1-TCB mediates antitumor efficacy comparable to the parental FOLR1-TCB whereas a noncleavable control Prot-FOLR1-TCB is inactive. In contrast, killing of bronchial epithelial and renal cortical cells with low FOLR1 expression is prevented compared to the parental FOLR1-TCB. The findings are confirmed for mesothelin as alternative tumor antigen. Thus, masking the anti-CD3 Fab fragment with an anti-idiotypic mask and cleavage of the mask by tumor-specific proteases can be applied to enhance specificity and safety of TCBs.

Fig. 1. Mode of action of protease-activated FOLR1-TCB.

On the left panel, the anti-CD3 Fab is blocked by an anti-idiotypic anti-CD3 scFv and thus cannot activate T cells against FOLR1-expressing cells. On the right panel, the linker containing a tumor-specific protease site has been cleaved and the anti-CD3 moiety is active leading to lysis of FOLR1-expressing cells. The figure shows an idealized representation of the Prot-FOLR1-TCB before (left) and after (right) cleavage at the matriptase cleavage site (cyan). The model is based on the full-length IgG crystal structure with PDB ID 1hzh. The protected anti-CD3ε Fab was modeled based on the crystal structure of an idiotype-anti-idiotype Fab complex structure with PDB ID 1iai. The catalytic domain of matriptase (crystal structure with PDB ID 1eax) is shown for reference (gray). Visualized with Biovia Discovery Studio 17R2 and arranged with GIMP. TME tumor microenvironment.

Fig. 2. Design of protease-activated anti-CD3 in IgG and TCB format.

“Knobs-into-holes” technology was used for the generation of heterodimeric molecules and PG LALA mutations were inserted to prevent FcγR binding. a Protease-activated monovalent anti-CD3 IgG (Prot-mαCD3 IgG) with N-terminal fusion of anti-idiotypic anti-CD3 scFv (disulfide-stabilized) and linker containing a protease cleavage site, monovalent anti-CD3 IgG with an anti-idiotypic anti-CD3 scFv N-terminally fused by a noncleavable glycine-serine linker and monovalent anti-CD3 IgG. b Prot-FOLR1-TCB with N-terminal fused anti-idiotypic anti-CD3 scFv and linker containing a protease cleavage site, Prot-FOLR1-TCB with N-terminal fused anti-idiotypic anti-CD3 scFv and a noncleavable GS linker and FOLR1-TCB. c Prot-MSLN-TCB with N-terminal fused anti-idiotypic anti-CD3 scFv and linker containing a protease cleavage site, Prot-MSLN-TCB with N-terminal fused anti-idiotypic anti-CD3 scFv and noncleavable GS linker and MSLN-TCB. For FOLR1 and CD3 binders, a common light chain could be used. For MSLN, the CrossMab technology was used to foster correct pairing of chains.

Fig. 3. Masking of CD3 binder reversibly impairs binding to CD3ε antigen.

Single particle analysis of Prot-mαCD3 IgG. a Negative stain transmission electron microscopy (NS-TEM) of the Prot-mαCD3 IgG containing a matriptase cleavable linker. 2D class-averages, as they result from multivariate statistical analysis of raw micrographs, are displayed in boxes (box size 36 nm). The Fc parts have triangular shapes with a central hole and Fabs are more elongated. At the distal end of the Fab the protective scFv appears as a single small additional density (see white arrow). The complex formation between antibody and ligand is identified by the presence of two Fc-regions linked together by an elongated structure corresponding to Fab and bound ligand. Due to the flexibility of the complex, one of the Fc displays less details in most of the class-averages. b Time-lapsed tapping-mode AFM data of individual Prot-mαCD3 IgG molecules have been recorded before (left), during activation with matriptase (middle), as well as after complexation with CD3 antigen (right) in a qualitative manner. The process is accompanied by length and shape changes which are noticed in the morphology maps. The treatment of Prot-mαCD3 (left) with matriptase results in an activated and shorter molecule (middle). The complexation of the activated molecules with Fc-CD3 gives the expected elongated and larger molecule (right). The lengths changes are compared with cross-section profiles starting with the individual masked molecule (green) undergoing activation (red) and after been complexed (blue), the profiles match the particles dimensions measured with NS-TEM, as depicted in (a). Representative AFM images of n = 3 experiments shown.

Fig. 4. T-cell activation is reversibly impaired by blocking anti-CD3 binder.

T-cell/Jurkat NFAT activation is dependent on crosslinking of mα-CD3. a Monovalent anti-CD3 IgG is able to activate luciferase in a Jurkat NFAT luciferase reporter cell line after crosslinking by plate-coated anti-human Fc (red curve) whereas in the absence of crosslinking no luciferase activity can be detected (blue curve). The dotted line indicates the luminescence for Jurkat NFAT-cells without mα-CD3 IgG on plate-coated anti-human Fc. Each value represents the mean of triplicates, standard deviation is indicated by error bars (representative experiment of n = 3). b Monovalent α-CD3 IgG is able to activate CD8-positive T cells, measured by quantification of the early activation marker CD69, after crosslinking by plate-coated anti-human Fc (red curve) whereas in the absence of crosslinking no CD8-positive T-cell activation can be detected (blue curve). The median fluorescence intensity (MFI) for CD69 of CD8 T cells is shown. Each value represents the mean of triplicates, standard deviation is indicated by error bars (representative experiment of n = 3). c, d mαCD3 IgGs were bound to plate-coated anti-human Fc antibody before Jurkat NFAT reporter cells or PBMCs were added. Jurkat NFAT activation is measured in relative luminescence units (RLU) and T-cell activation was assessed by quantification of CD69 by FACS analysis. Cleavage of the Prot-mαCD3 IgG containing a matriptase cleavable linker was performed by incubation of Prot-mαCD3 IgG with purified recombinant human matriptase for 24 h at 37 °C. c Jurkat NFAT activation mediated by Prot-mαCD3 IgG. Cleaved Prot-mαCD3 IgG, blocked Prot-mαCD3 IgG, mαCD3 IgG and mαCD3 IgG with an N-terminal fusion of a nonspecific fusion (anti-CEA Fab) are shown. The dotted line indicates the luminescence for Jurkat NFAT-cells without any CD3 IgG. Each value represents the mean value of triplicates, standard deviation is indicated by error bars (representative experiment of n = 3). d The median fluorescence intensity (MFI) for CD69 of CD8 T cells is shown. Each value represents the mean value of triplicates, standard deviation is indicated by error bars (representative experiment for three different human PBMC donors).

Fig. 5. Prot-FOLR1-TCB is efficiently blocked while its potency can be fully restored upon linker cleavage.

Dose–response curves for T-cell killing of FOLR1-positive tumor cells after 48 h mediated by TCB using PBMCs as effector cells with an E:T of 10:1. a, b Comparison of Prot-FOLR1-TCB (noncleavable and precleaved with matA site) for tumor cells with different FOLR1 expression levels (a HeLa cells, b Skov-3 cells). The cytotoxicity induced by the antibodies is shown. Each point represents the mean of triplicates. Standard deviation is indicated by error bars (representative experiment for one PBMC donor, n = 2 different human PBMC donors). c, d The cytotoxicity induced by the Prot-FOLR1-TCB containing MMP-matA site is shown. The Prot-FOLR1-TCB containing an MMP-matA site (gray circles) or a noncleavable linker (blue triangles pointing up) is shown. FOLR1-TCB (red triangles, pointing down), precleaved Prot-FOLR1-TCB (orange squares) and a nontargeted TCB (black circles) are used as controls. The cytotoxicity of the Prot-FOLR1-TCB with a noncleavable linker was only measured for four concentrations using Skov-3 cells as the target cells. The nontargeted TCB was only used at the three highest concentrations for both cell lines. The pretreatment of the Prot-FOLR1-TCB (orange squares) was done by incubation of Prot-FOLR1-TCB with recombinant human matriptase for 24 h at 37 °C. Each point represents the mean of triplicates. Standard deviation is indicated by error bars (representative experiment for one PBMC donor, n = 3 different human PBMC donors). e, f Granzyme B release was quantified by FACS analysis after incubation of PBMCs with 10 nM of Prot-FOLR1-TCB (cleavable linker vs. noncleavable linker) and FOLR1-positive target cells for 48 h. Each point represents the mean value of triplicates. Standard error is indicated by error bars (representative experiment for one PBMC donor, n = 3 different human PBMC donors). g FOLR1-positive tumor cell (CellPlayer™ MDA-MB-231 NucLight Red) growth inhibition mediated by Prot-FOLR1-TCB containing MMP-matA site and human PBMCs. Each point represents the mean of triplicates. Standard deviation is indicated by error bars (representative experiment for one PBMC donor, n = 2 different human PBMC donors).

Fig. 6. Cell cytotoxicity of FOLR1-TCB is abolished by blocking of anti-CD3 for primary cells with low FOLR1 expression.

T-cell killing of primary cells, expressing low levels of FOLR1, mediated by Prot-FOLR1-TCB with MMP-matA site using PBMCs as effector cells and an E:T ratio of 10:1. a, b Human bronchial epithelial cell (HBEpiC) (a) and human renal cortical epithelial cell (b) killing assessed after 72 h. Each point represents the mean of triplicates (one human PBMC donor shown, n = 3 different human PBMC donors). Standard deviation is indicated by error bars. c, d Median fluorescence intensity of CD69 of CD8-positive T cells is shown after incubation with TCB and HBEpiC (c) and HrcEpiC (d) cells. The dotted line indicates the MFI for CD8-positive T cells incubated without any TCB. Each point represents the mean value of triplicates, standard deviation is indicated by error bars (one human PBMC donor shown, n = 3 different human PBMC donors).

Fig. 7. Prot-FOLR1-TCB can be activated by patient-derived ovary cancer explants.

Jurkat NFAT reporter assay was used to analyze activation of Prot-FOLR1-TCB (matA or MMP-matA linker) ex vivo by undigested human tumor explants. a Benign tumor of the ovary. b Cancer of the ovary. c Cancer of the ovary. Explants were mechanically cut and then incubated with TCBs and analyzed for CD3 activation using Jurkat NFAT cells. Jurkat NFAT activation is measured in relative luminescence units (RLU). Each symbol indicates the value measured for one biological sample incubated with Jurkat NFAT cells and Prot-FOLR1-TCB MMP-matA site (gray bar), matA site (purple bar), noncleavable site (blue bar) or FOLR1-TCB (red bar). a Each data point shows the mean of technical duplicates measured for one well (n = 2 biological replicates). b Each data point shows the value measured for one well (n = 3 biological replicates). Standard deviation is indicated by error bars. c Each data point shows the mean of technical duplicates measured for one well (n = 2 biological replicates). The dotted lines indicate luminescence for Jurkat NFAT-cells incubated with tumor samples but without any TCB.

Fig. 8. Stability of Prot-FOLR1-TCB depends on the cleavage site.

a Bioavailability of active Prot-FOLR1-TCB containing different cleavage sites at day 7 after intravenous single-dose injection in non-tumor-bearing NSG mice (n = 3 per group). Active and total Prot-FOLR1-TCB were quantified by ELISA using an anti-PG-LALA antibody (total Prot-FOLR1-TCB) and the anti-idiotypic anti-CD3 antibody (active Prot-FOLR1-TCB). Pharmacokinetic evaluation was conducted by noncompartmental methods. Areas under the serum concentration−time curve were calculated by linear trapezoidal rule. Bioavailabilities F of the active FOLR1-TCB after Prot-FOLR1-TCB administration were calculated by comparing AUC 0−168 h values of FOLR1-TCB following i.v. administration of the respective pro-TCB (AUC from Prot-FOLR1-TCB) and administration of the active TCB (AUC FOLR1-TCB) according to F(%) = (AUC from Prot-TCB/AUC TCB) × 100. Dose corrections were not required, as equimolar doses of Prot-FOLR1-TCB and FOLR1-TCB were used in the respective studies. b Immunohistochemistry of FOLR1 and matriptase in breast PDX tumor at baseline. Some tumors were harvested at start of treatment for the baseline characterization of FOLR1 target antigen and matriptase by immunohistochemistry. Positive staining is observed as a brown precipitate within the sections. The evaluation of FOLR1 and Matriptase expression in the described in vivo model has been conducted twice in two independent samples. Scale bars indicates 200 µm and is included in the image.

Fig. 9. Prot-FOLR1-TCB is efficacious in vivo while there is no hint for tumor-leakage of activated Prot-FOLR1-TCB.

a Tumor growth inhibition curves of breast PDX BC004 model in humanized mice. Humanized mice (n = 9 per group) were weekly i.v. injected with equimolar doses of Prot-FOLR1-TCBs (4 mg/kg) containing different cleavage sites, FOLR1-TCB (3.6 mg/kg) or vehicle. Each dot represents the mean tumor volume ± SEM. Efficacy was evaluated by measuring the reduction of the mean tumor volume at day 62 relative to vehicle control. Statistical analysis was done using one-way ANOVA with Tukey−Kramer correction. No significant efficacy was observed for the Prot-FOLR1-TCB with the noncleavable linker comparing to vehicle control. Significant tumor growth inhibition was induced by the FOLR1-TCB (****p < 0.0001), the Prot-FOLR1-TCB with the matB site (****p < 0.0001) and the Prot-FOLR1-TCB with matC site (*p = 0.034) all compared to vehicle group. The in vivo efficacy study has been conducted once (n = 9 mice per group; n = 8 mice for matB group). b Quantification of human CD3-positive T cells for the different treatment groups of efficacy study using breast PDX in humanized mice. All data points shown in bar chart. Each data point represents the value for one mouse. 95% confidence interval is shown for each group. Two-tailed, unpaired t test was used to calculate statistics. Significantly more huCD3 T cells per mm³ were found for animals treated with FOLR1-TCB (***p = 0.0009) and Prot-FOLR1-TCBs (matC *p = 0.0284 and matB site ***p = 0.0009) than for vehicle group. However also the Prot-FOLR1-TCB with a noncleavable linker had significant more huCD3 T cells per area compared to vehicle (*p = 0.0481). Representative images are shown in Supplementary Fig. 9. c, d Bioavailability of active Prot-FOLR1-TCBs with different linkers in non-tumor-bearing (c) and tumor-bearing humanized mice (d). Bioavailability of active Prot-FOLR1-TCB containing different cleavage sites at day 7 after injection in non-tumor-bearing humanized NSG mice (n = 3 per group) or breast PDX tumor-bearing humanized mice (n = 6 per group). Bioavailabilities F of the active FOLR1-TCB after Prot-FOLR1-TCB administration were calculated (as described in Fig. 8) by comparing AUC 0−168 h values of FOLR1-TCB following i.v. administration of the respective pro-TCB (AUC from Prot-FOLR1-TCB) and administration of the active TCB (AUC FOLR1-TCB) according to F(%) = (AUC from Prot-TCB/AUC TCB) × 100. For tumor-bearing mice AUC were calculated from composite concentration−time data (n = 3/time point). Dose corrections were not required, as equimolar doses of Prot-FOLR1-TCB and FOLR1-TCB were used in the respective studies.