A T-cell engaging bispecific antibody with a tumor-selective bivalent folate receptor alpha binding arm for the treatment of ovarian cancer

Immunizations, next-generation sequencing, clonotype analysis, and cloning

Methods used for immunization, NGS, and cloning were previously described.²³ Briefly, UniRats were immunized with recombinant FRα protein (Antibody Solutions, CA, USA). After 8 weeks, draining lymph nodes were harvested, and total RNA was isolated. Ig heavy chains were sequenced using next-generation sequencing (NGS) (Illumina MiSeq). The data were analyzed as previously described, and the top clonotypes were recombinantly expressed in HEK293 cells.²⁴

Protein expression and purification

Antibodies were expressed in ExpiCHO cells (Thermo Fisher Scientific). Clarified supernatants were harvested and affinity purified using CaptureSelect CH1-XL resin (Thermo Fisher Scientific) and cation exchange (Mono S). Antibodies were analyzed by size exclusion chromatography-ultra-high performance liquid chromatography (SEC-UPLC) and sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE).

Cell lines

Tumor cell lines OVCAR-3, SKOV-3, HT-29, and HeLa were obtained from ATCC. LNCaP and MCF-7 were obtained from Millipore Sigma. IGROV-1 was obtained from NCI. Choroid plexus, retinal pigment, pulmonary alveolar, bronchial, and renal cortical epithelial cells were obtained from ScienCell Research Laboratories. All cell lines were maintained in a 5–8% CO₂ buffered incubator at 37°C and maintained as per manufacturer’s recommendations.

Thermal stress and stability characterization

Antibodies were formulated in 20 mM citrate and 0.1 M NaCl at pH 6.2. High-molecular-weight species were measured by SEC-UPLC (ThermoFisher UltiMate 3000 HPLC) before (T₀) and after temperature stress (T₃₀) for 1 month at 37°C.

Quantification of FRα expression

Estimation of antibodies bound per cell (ABC) was described previously.²⁴ Briefly, cell lines were stained with a saturating concentration of a PE-labeled FRα antibody (BioLegend, clone LK26) and tested via flow cytometry using BD Quantibrite beads. Antigen density (or FRα molecules per cell) was calculated by multiplying ABC values by 2.

Scatchard analysis

Methods for Scatchard analysis were described previously.²⁴ Briefly, the cell surface affinity of bivalent TNB-928B and monovalent FRαxCD3 TCEs was measured using IGROV-1. Antibodies were labeled with Alexa Fluor-488 (Thermo Fisher), and Quantum Alexa Fluor-488 MESF beads (Bangs Laboratories) were used to generate the standard curve. Affinity (K_D) was calculated as previously described.²⁴

Activation and proliferation of T-cells

To measure the activation of CD4⁺ and CD8⁺ T-cells, T-cells were incubated with IGROV-1 or SKOV-3 (E:T of 5:1) in the presence of an indicated antibody for 48–72 hours. Cells were harvested and incubated with LIVE/DEAD™ Fixable Near IR dye (Thermo Fisher Scientific) and Human TruStain FcX (Biolegend). The cells were washed with 1XPBS + 1% BSA + 0.1% NaN₃ and stained with antibodies for CD45 (Biolegend, clone 2D1), CD4 (Biolegend, clone OKT4), CD8 (Biolegend, clone RPA-T8) and CD69 (Biolegend, clone FN50). Cells were washed twice and analyzed by flow cytometry. T-cell proliferation assays were similar but with an incubation period of 3 d, and cells were stained via intracellular staining for Ki67 (Biolegend, clone Ki-67). CD4⁺ Tregs were stained with antibodies for CD25 (Biolegend, clone BC96) and FOXP3 (Biolegend, clone 206D) using intracellular staining with the FOXP3 fix/perm kit (Biolegend).

In vitro cytotoxicity and cytokine assays

Target cells were plated at a density of 0.1 × 10⁶ to 0.2 × 10⁶ in RPMI-1640 + 10% FBS + 1% P/S, and frozen PBMCs (STEMCELL Technologies) were thawed, washed, and rested overnight. After 24 hours, pan T-cells were isolated (Miltenyi Biotec). Pan T-cells were added to tumor cells at the E:T ratios indicated along with diluted antibody. After 48 hours, an aliquot of supernatant was removed, frozen, and subsequently analyzed for IL-2 and IFNγ release (MSD U-plex platform). Cytotoxicity was measured using Cell Proliferation Reagent WST-1 (Sigma) or CytoTox 96 LDH substrate (Promega). Percentage of cell lysis was calculated as previously described.²⁵

Ex vivo cytotoxicity, cytokine, and cytotoxic granule measurement

Fresh, de-identified, surgically removed ovarian tumor tissue (obtained under an MTA from Dr Chapman) was cut into small pieces, washed with PBS, and digested with 2.5 U/mL dispase II (STEMCELL Technologies) for 30 minutes at 37°C with manual agitation every 5 minutes. Dissociated cells were centrifuged and resuspended in DMEM with 10% FBS and 1% P/S at 0.4 × 10⁶ cells/mL. Cells were incubated with the indicated antibodies for 48–72 hours. Cytotoxicity was measured using CytoTox 96 LDH substrate (Promega). Cytokines were measured as described above. Perforin and granzyme B were measured by ELISA (Cell Sciences and Thermo Fisher Scientific, respectively). T-cells were quantified, and the effector to target cell ratio in these samples was determined by measuring the subsets of live T-cells, positive for CD4 and CD8, vs. the target cells staining positive for FRα (CD4⁻/CD8⁻).

Analysis of FRα protein expression in tumor tissue

Representative blocks of formalin-fixed, paraffin-embedded tumor were available for all nine patients that were tested ex-vivo. Briefly, 4-µm tissue sections were deparaffinized, rehydrated, and pretreated in appropriate buffer for antigen retrieval by using a Biocare Decloaking Chamber. Tissue slides were then incubated at room temperature with a primary antibody, anti-FRα (Abcam, EPR20277, 4 µg/ml) followed by an appropriate secondary antibody. FRα staining intensity was scored according to the following grading scheme: 0 = no stain; 1+ = weak (minimal) stain; 2+ = mild stain; and 3+ = intense stain.

In vivo mouse efficacy study

NOD-Prkdc^em26Cd52I2rg^em26,d22 (NCG) mice (7–9 weeks, GemPharmatech Co., Ltd.) were subcutaneously injected with 5 × 10⁶ IGROV-1 or SKOV-3 cells with 50% Matrigel and intraperitoneally injected with 10 × 10⁶ resting human PBMCs (AllCells) on day 0. On day 3, mice were randomized into groups (n = 3) by body weight and injected intravenously (i.v.) with test article. Test article was injected every 3 d at respective doses for a total of 8–10 treatments. Twice per week, the length and width of each tumor was measured using calipers and the volume was expressed in mm³ using the formula: V = (L × W × W)/2, where V is tumor volume, L is tumor length, and W is tumor width. In a similar follow-on study, NCG mice were injected with activated human PBMCs (AllCells). Prior to injection, PBMCs were activated for 3 d with 1 µg/mL plate-coated anti-CD3 (eBioscience, clone OKT3) and 1 µg/mL soluble anti-CD28 (eBioscience, clone CD28.2). At the end of the study, the tumors were harvested, processed into formalin-fixed paraffin-embedded sections, and stained with anti-human CD45 (Cell Signaling Technology, 13917, 1:800) and anti-human FRα (Abcam, EPR20277, 1:1000). The stained sections were scanned with NanoZoomer-2.0 HT Image System (Hamamatsu) at 40x magnification.

Mouse pharmacokinetic (PK) evaluation

The PK of TNB-928B was evaluated in female BALB/c mice using the study design described previously.²⁴ Mice were injected via tail vein with 1 mg/kg or 10 mg/kg TNB-928B (n = 3/group * 6 groups). Serum samples were collected for 14 d post-dose at timepoints ranging from 30 minutes to 14 d. Serum TNB-928B concentrations were measured using human IgG4 AlphaLISA (Perkin Elmer). PK parameters were estimated using Phoenix WinNonlin version 7.0 software (Pharsight Corp).

Statistics

Statistical analysis was performed using a two-tailed unpaired t-test using GraphPad Prism 9. P values <.05 were considered statistically significant.

FRα expression on normal cells and ovarian tumor cells

FRα is expressed on normal healthy tissues including lung and kidney, and targeting FRα with a TCE could result in on-target, off-tumor toxicity. We sought to quantify cell surface expression of FRα on various ovarian and other solid tumor cell lines as well as normal primary cells by flow cytometry. The FRα antigen density on the ovarian tumor cell lines OVCAR-3, SKOV-3, and IGROV-1 ranged from 44 × 10³ to 1,871 × 10³ (Table 1), recapitulating the range observed in clinical samples of relapsed ovarian cancer patients.²⁶ In contrast, normal primary alveolar and bronchial epithelial lung cells express 0.1 × 10³ and 0.8 × 10³ FRα/cell respectively, and renal cortical epithelial cells express 6.7 × 10³ FRα/cell (Table 1). FRα expression has also been reported on choroid plexus and retinal epithelial cells;^13,14 however, we only measured FRα antigen densities of 0.7 × 10³ and 1.3 × 10³, respectively. Since the high end of expression on normal tissues reported to express FRα is in the range of 7 × 10³, the cell lines HT-29 and MCF-7 with FRα antigen densities of 8 × 10³ and 5 × 10³, respectively, were selected as cell lines representing the FRα expression threshold at or below which no TCE-dependent cytotoxicity was desirable.

Generation and functional characterization of a FRαxCD3 TCE

Fully human heavy-chain only antibodies (UniAbs) targeting FRα were generated by immunization of UniRats followed by a next-generation sequencing (NGS)-based antibody repertoire discovery approach.^23,27 FRα-specific UniAbs were identified by screening recombinant antibody candidates for cell surface binding to FRα⁺ tumor cell lines and binding to FRα protein. Antibodies against CD3 were generated in OmniFlic animals, and the previously identified low-affinity anti-CD3_F2B arm that mediates tumor cell killing with minimal cytokine release was included in the creation of FRαxCD3 TCEs.^23,28

We sought to improve the selectivity of a FRαxCD3 TCE to minimize the risk of on-target, off-tumor activity and increase the therapeutic window of a FRα targeting TCE. To achieve tumor selectivity, we utilized a reduced-affinity bivalent format to specifically target FRα overexpressing ovarian tumor cells. This 2 + 1 bispecific antibody format has previously been applied to TCEs for selective targeting of solid tumors overexpressing CEA and HER2.^29,30 FRαxCD3 TCEs were generated by pairing the lead anti-FRα VH (variable heavy region of the antibody) in either monovalent or bivalent formats with OKT3, a high-affinity CD3 binding arm (for a positive control, PC) or with CD3_F2B, a low-affinity CD3 binding arm (lead bivalent antibody, TNB-928B) using knobs-in-holes technology as previously described (Figure 1a).²⁸ TNB-928B showed robust expression (3.8 g/L) and low aggregation propensity after thermal stress (Supplemental Figure S1 and Supplemental Table S1). The cell surface affinity of the bivalent anti-FRα arm of TNB-928B was 3.7 ± 1.6 nM, and the monovalent FRαxCD3 was 123 ± 4 nM to IGROV-1 cells as determined by Scatchard analysis (Figure 1b and Supplemental Figure S2).

To determine the effect of FRα valency on the in vitro functional activity of FRαxCD3 TCEs, cytotoxicity assays were performed. The ability of TNB-928B to mediate the lysis of IGROV-1 (high FRα antigen density) was compared to a corresponding bispecific antibody containing the same VH, which is monovalent for FRα. Both TNB-928B and the bivalent PC antibodies exhibited an avidity effect observed in lower EC₅₀ values (increased potency) compared to the respective monovalent antibodies (Figure 1c and Supplemental Table S2). Importantly, TNB-928B exhibited similar maximum tumor cell lysis compared to the PC at saturating doses with ~75% lysis. Taken together, these data demonstrate that the bivalent format of TNB-928B exhibited a strong avidity effect for killing high FRα expressing tumor cells and TNB-928B achieves maximum activity comparable to the PC.

T-cell exclusion from solid tumors could present an obstacle for T-cell mediated therapies. To evaluate if TNB-928B can mediate tumor cell lysis at low effector to target cell (E:T) ratios, cytotoxicity assays were conducted with several E:T ratios ranging from 5:1 to 1:5. Substantial lysis of SKOV-3 and OVCAR-3 tumor cells was observed at E:T ratios of 1:5 and 1:3 with 44% and 61% maximum lysis, respectively (Supplemental Figure S3A). IGROV-1 was more resistant to lysis at low E:T ratios with minimal tumor cell lysis at 1:1 and below. These results demonstrated that TNB-928B is capable of tumor cell lysis at very low E:T ratios that reflect the predicted ratios in solid tumors.

Like other GPI-anchored proteins, FRα is cleaved from the cell surface. Soluble FRα protein (sFRα) is elevated in the serum of ovarian cancer patients, and in both early and advanced ovarian cancer patients, high sFRα is associated with shorter PFS.^31,32 Moreover, expression of FRα on tumor cells is strongly correlated with sFRα levels. To determine the effect of sFRα on TNB-928B mediated tumor cell lysis, a cytotoxicity assay was performed in the presence of exogenously added soluble recombinant FRα. A minimal decrease in potency of TNB-928B mediated SKOV-3 lysis was detected in the presence of sFRα up to 1,000 ng/mL, which is approximately 100-fold higher than sFRα measured in the serum of EOC patients,^31,33 but maximum cell lysis was unchanged (Supplemental Figure S3B). The measured levels of sFRα after 48 hours were comparable to the amount of exogenous sFRα added, suggesting sFRα was stable for the duration of the experiment (Supplemental Figure S3C). These data suggest that the cytotoxic activity of TNB-928B is not affected by sFRα levels comparable to physiological levels detected in the serum of ovarian cancer patients.

TNB-928B activates resting T-cells and induces T-cell proliferation

To further characterize the mechanism of action of TNB-928B, we measured the upregulation of CD69 in subsets of CD3⁺ T-cells as a marker of T-cell activation following co-culture of resting T-cells and IGROV-1 cells. TNB-928B induced dose-dependent activation of both CD4⁺ and CD8⁺ T-cells to a similar extent as the PC (Figure 2a). Approximately 73% of CD4⁺ and 69% of CD8⁺ T-cells were CD69-positive following co-culture of IGROV-1 cells and pan T-cells at saturating concentrations of TNB-928B. The EC₅₀ values for TNB-928B mediated CD4⁺ and CD8⁺ T-cell activation were 16.1 pM and 8.3 pM, respectively, compared to approximately 1 pM for the PC in both T-cell subtypes. Negligible CD69 upregulation was observed with treatment with a negative control (NC) with the same anti-CD3_F2B arm as TNB-928B but with an irrelevant tumor targeting arm. Similar T-cell activation results were observed with SKOV-3 cells (Supplemental Figure S4). To measure the relative activation of Tregs by TNB-928B, we measured the percentage of Tregs in the T-cell population by gating for CD4⁺CD25⁺Foxp3⁺ cells. Importantly, TNB-928B induced Tregs approximately two-fold less than the PC at antibody concentrations saturating for T-cell activation compared to PC treated cells (Figure 2c), suggesting TNB-928B preferentially activates effector T-cells compared to Tregs.

To evaluate the ability of TNB-928B to induce T-cell proliferation, we measured the upregulation of Ki67 as a marker of proliferation following co-culture with IGROV-1 tumor cells. TNB-928B induced dose-dependent increases in Ki67 with approximately 43% of CD4⁺ and 71% of CD8⁺ T-cells (Figure 2b). The EC₅₀ values for TNB-928B mediated CD4⁺ and CD8⁺ proliferation were 48.9 pM and 17.8 pM, respectively. No Ki67 proliferation was detected with the NC, demonstrating target specificity of TNB-928B. Taken together, these data are consistent with the mechanism of action of TCEs.

TNB-928B exhibits selective cytotoxicity of high FRα-expressing tumor cells

To evaluate the ability of TNB-928B to selectively lyse high FRα expressing cell lines while sparing low FRα expressing cell lines, we examined the tumor cell lysis activity of TNB-928B compared to a monovalent FRαxCD3_F2B (or a PC that contains CD3_OKT3) containing the same anti-FRα VH on SKOV-3, HeLa, HT-29, and MCF-7 cells. SKOV-3, and HeLa represent high FRα density cell lines, whereas HT-29 and MCF-7 are low FRα density cell lines with expression levels similar to normal cells (Table 1). While the monovalent FRαxCD3_F2B molecule was efficacious in lysing SKOV-3 (75% maximum lysis), this molecule also reached high levels of maximum lysis of HT-29 (60% maximum lysis, Figure 3 a and b), which likely would result in unacceptable in vivo on-target, off-tumor toxicity associated with targeting FRα with a monovalent TCE. In contrast, TNB-928B exhibited significantly reduced lysis of HT-29 and MCF-7 cells as compared to the bivalent PC (10% vs 74% maximum HT-29 lysis, respectively, Figure 3b and Supplemental Figure S5B). Nevertheless, TNB-928B exhibited similar maximum cytotoxicity of SKOV-3 cells as the PC with ~70% maximum lysis (Figure 3a). TNB-928B also mediated the lysis of IGROV-1, HeLa, and OVCAR-3 cell lines, demonstrating activity against tumor cell lines with high FRα antigen densities (Table 1 and Supplemental Figures S3A and S5A). Taken together, these results demonstrate that the bivalent format of TNB-928B has similar efficacy as that of a strong TCE on high-density FRα tumor cell lines but minimal activity on low FRα density cell lines, indicating that TNB-928B is selective for FRα overexpressing tumor cells due to its avidity-mediated activity.

TNB-928B decouples tumor cell cytotoxicity from cytokine release

CAR-Ts and TCEs are frequently associated with CRS. To minimize the risk of CRS, TNB-928B contains the previously described and clinically validated anti-CD3_F2B (Figure 1a).^24,25,28,34 To evaluate TNB-928B-mediated cytokine release from several PBMC donors, aliquots of cell culture supernatant were collected from cytotoxicity assays and analyzed for IL-2 and IFNγ cytokine release. At concentrations of TNB-928B that mediate robust lysis of IGROV-1, SKOV-3, and OVCAR-3, the levels of IL-2 and IFNγ were lower than those induced by the OKT3 containing PC, irrespective of the donor (Figure 4). Furthermore, TNB-928B displayed no cytotoxicity nor IL-2 or IFNγ release when tested against the FRα-negative LNCaP cell line (Figure 4).

Activity of TNB-928B on ex vivo patient ovarian tumor samples

We evaluated TNB-928B mediated cytotoxicity in patient-derived samples using fresh surgically removed ovarian tumor biopsies. Ex vivo ovarian tumor samples were incubated with TNB-928B, PC, or NC for 48–72 hours. TNB-928B mediated substantial ex vivo tumor cell lysis comparable to the PC in dissociated ovarian tumor samples from six of nine different patients (Figure 5a and Supplemental Table S3). No exogenous T-cells were added to these samples, and of the five samples that were immunophenotyped, four had E:T ratios <1, indicating that the few endogenous T-cells present in the tumor were sufficient for mediating tumor cell cytotoxicity. We measured FRα antigen density on seven of the nine patient samples (IHC expression identified in eight of nine patient samples; Supplemental Table S3) and found a trend between samples displaying <10% (non-responders) and >10% (responders) tumor lysis (Figure 5b). Two of the non-responders had FRα antigen densities <1 × 10³ (Supplemental Table S3) and therefore were not expected to show activity based on our in vitro results. In a representative dissociated ovarian tumor sample, the EC₅₀ of TNB-928B was 45.2 pM, consistent with our in vitro results (Figure 5c). Also, consistent with our in vitro cytotoxicity results, TNB-928B induced reduced levels of cytokine release as measured by IFNγ and IL-2 compared to the PC (Figure 5 d and e). TNB-928B mediated tumor cell lysis was accompanied by the release of cytotoxic granules, with levels of perforin and granzyme B in the supernatant comparable to the PC (Supplemental Figure S6A and B). Importantly, when dissociated ovarian tumor samples were incubated with patient-matched PBMCs at an E:T ratio of 1:1, TNB-928B induced approximately 2.5-fold less activation of Treg cells compared to the PC (figure 5f). These data suggest TNB-928B mediates robust ovarian tumor cell killing of primary patient tumor samples that express high levels of FRα, decouples cytotoxicity from cytokine release, and preferentially activates effector T-cells over Treg cells.

In vivo antitumor efficacy and pharmacokinetics (PK) of TNB-928B in murine models

In vivo efficacy of TNB-928B was assessed in immunocompromised NCG mice implanted with IGROV-1 subcutaneously (s.c.) and resting PBMCs intraperitoneally (i.p.) in a prevention of tumor engraftment model. Mice were treated with TNB-928B administered intravenously (i.v.) every 3 d. TNB-928B displayed dose-dependent tumor growth inhibition (TGI) of 90.5%, 94.6%, or 96.4% with doses 1, 10, or 100 μg/animal, respectively (Figure 6a). In a separate study with activated PBMCs, both TNB-928B and the PC resulted in substantial anti-tumor activity of IGROV-1 and SKOV-3 tumors and human immune cell infiltration into IGROV-1 tumors as measured by IHC staining for human CD45 in tumors harvested at the end of the study (Figure 6b and Supplemental Figure S7A and B). Moreover, both TNB-928B and PC resulted in the elimination of nearly all FRα-positive tumor cells by day 30 (Figure 6b). Taken together, these results suggest TNB-928B exhibits strong in vivo efficacy in multiple murine xenograft models.

The PK of TNB-928B was evaluated in BALB/c mice following a single tail vein injection at 1 and 10 mg/kg. TNB-928B clearance ranged from 7.9 to 10.3 mL/d/kg. The half-life ranged from 3.9 to 8 d for TNB-928B (Figure 6c). Since neither arm of TNB-928B cross-reacts with FRα or CD3 in rodent species, the observed linear PK was expected and is consistent with nonspecific clearance mechanisms. These results suggest TNB-928B is stable in vivo with favorable PK and with a half-life similar to conventional antibodies.