In vitro effects of combinatorial treatments of Tb535H with immunomodulatory mAbs on the stimulation of hPBMCs

We firstly investigated on novel combinations of multiple immunomodulatory mAbs, recently isolated in our laboratory, specific for PD-L1, PD-1 or LAG-3 and able to strong activate T cells [24,25,26,27,28], with the parental bi-specific tribody (Tb535H) targeting the 5T4-TAA on tumor cells and CD3 on T cells, which already demonstrated anti-tumor efficacy in preclinical studies [23]. These novel combinations were tested on PD-L1-positive MDA-MB 231 and BT-549 breast cancer cells expressing 5T4, co-cultured for 48 h with lymphocytes (Effector:Target cells ratio 5:1), by using the Tb535H tribody (1 nM) plus each of the immunomodulatory mAbs (50 nM) in comparison with each single parental compound (Supplementary Fig. 1A and B). Cell surface expression of 5T4 and PD-L1 on these cells were confirmed by flow cytometric analyses (data not shown). Interestingly, we found that the addition of the anti-PD-L1, -PD-1 or LAG-3 mAb to the tribody significantly increased the tumor cells lysis accordingly to the higher secretion of IFNγ cytokine, detected after treatments (Supplementary Fig. 1). In parallel, the same combinations tested on PD-L1, 5T4-negative MCF-7 cancer cells showed no significant increase of these effects, thus confirming the specificity of the combinatorial treatments (Supplementary Fig. 1C).

Construction and purification of novel 53X tribodies

The promising results of the combined treatments prompted us to generate five novel tri-specific tribodies, as shown in Fig. 1A and B and listed in Table 1. The tribody is a multi-specific antibody construct made up of a Fab domain as a scaffold, in which Fd fragment of H-chain (VH + CH1) and L-chain (VL + CL) can naturally heterodimerize in vivo, and additional functions, such as scFvs, can be incorporated on the scaffold. Parental Tb535H is a bi-specific tribody, endowed with a bivalent binding for 5T4 through its Fab and one of scFv domains, and a monovalent CD3 binding through another scFv [23]. The CD3 binding scFv used both in the parental Tb535H and novel 53X tribodies was generated from OKT3 mAb, and was chosen as not able to induce the non-specific activation of T cells and cytokines secretion in the absence of tumor cells at concentrations up to 1,000 ng/mL, as previously described in the patent documents of parental Tb535H (PCT/EP2015/080795, WO2016/097408 A1). The novel 53X tribodies generated here, have binding specificity to 3 different molecules, 5T4, CD3 and an IC, such as PD-L1, PD-1, or LAG3, in a single molecule. Indeed, the 5T4 binding scFv arm of Tb535H [23] was replaced by a binding moiety derived from PD-1_1, PD-L1_1, 10_12, LAG3_1 mAbs [24, 25]. The mAb 10_12 is an affinity-matured variant of PD-L1_1 [24], whereas Palivizumab [31] was used as a negative control. The novel tribodies were named as 53D [(Fab)5T4 x (scFv) CD3 x (scFv) PD-1(PD-1_1)], 53L1 [(Fab)5T4 x (scFv) CD3 x (scFv) PD-L1(PD-L1_1)], 53L10 [(Fab)5T4 x (scFv) CD3 x (scFv) PD-L1(10_12)], 53G [(Fab)5T4 x (scFv) CD3 x (scFv) LAG3(LAG3_1)], 53P [(Fab)5T4 x (scFv) CD3 x (scFv) Palivizumab], respectively. The latter one was constructed as an isotype control for the other 53X tribodies. All the tribodies were expressed and purified as described in Methods, and their purity and stability was checked by SDS-PAGE and size-exclusion chromatography (data not shown).

Fig. 1

Schematical representation of 53X tribodies. A Scheme of constructs for 53X tribodies. SP, human interleukin (IL-2) signal peptide; VH_5T4, VL_5T4, amino acid sequences coding for the immunoglobulin heavy and light chain variable regions from Tb535H with specificity for human 5T4; CH1, CL, amino acid sequences coding for the human immunoglobulin heavy chain constant region 1 and the kappa light chain constant region, respectively; VH_CD3, VL_CD3, amino acid sequences coding for the variable heavy and light chain regions from humanized OKT3 building a scFv with specificity for human CD3; L1, L2, aminoacid sequences with flexible linker, GPGGGSPG, and GGGGSGGGGSGGGGS [(GGGS)3], respectively. VH_C, VL_C, amino acid sequences coding for the variable heavy and light chain from PD-1_1, PD-L1_1, 10_12, LAG3_1, scFvs with specificities for human PD-1, PD-L1 or LAG-3, respectively. The scFv from Palivizumab was used in an additional tribody for isotype control; 6xHis, amino acid sequences coding for a hexahistidine tag. B Scheme of assembled tribody proteins for Tb535H [(5T4)2 × CD3], 53D [5T4 x CD3 x PD-1], 53L1 [5T4 x CD3 x PD-L1], 53L10 [5T4 x CD3 x PD-L1], 53G [5T4 x CD3 x LAG-3] and 53P [5T4 x CD3 x isotype control], respectively

Table 1 List of tribodies generated in the studyBinding affinity of 53X tribodies for recombinant 5T4 and CD3 proteins by Enzyme-Linked Immuno Sorbent Assays (ELISA)

To verify whether the novel generated tri-specific 53X tribodies 53D, 53L1, 53L10, 53G and 53P retain the ability of parental Tb535H to bind to 5T4 antigen, binding affinity of purified tribodies to 5T4 protein was examined by ELISA (Fig. 2). Recombinant human 5T4 protein was immobilized onto 96-well ELISA plates. The binding assays of parental Tb535H and purified novel tribodies were performed by testing them at increasing concentrations (0 – 500 nM) according to the procedure described in Materials and Methods. As shown in Fig. 2A, all the constructed tribodies showed concentration dependent binding to 5T4 with similar or even better EC50 (nM) values than that of the parental Tb535H (Fig. 2B). Even though the parental Tb535H has two 5T4 binding arms derived from a Fab and a scFv, differently from the novel constructed 53X tribodies that have a monovalent 5T4 binding arm derived from a single Fab, the results indicated that the novel tribodies show a comparable or better binding affinity (Kd), thus demonstrating that a single anti-5T4 arm (Fab) is sufficient for retaining similar 5T4 binding of the parental Tb535H in all the novel 53X tribodies.

Fig. 2

Binding affinity of parental Tb535H and novel 53X tribodies to recombinant 5T4 protein by ELISA. A Binding curves by ELISA assays of tribodies (0–500 nM) to immobilized recombinant human 5T4 protein. Binding values were reported as the mean of determinations obtained in three independent experiments. B Table reporting the binding affinity or EC50 (nM) values for 5T4 of each indicated construct. Standard Deviations were ≤ 3—10%

To verify whether the constructed novel 53X tribodies 53D, 53L1, 53L10, 53G and 53P, retain the ability of the parental Tb535H to bind to CD3 antigen, binding affinity of purified tribodies to recombinant CD3 protein was examined by ELISA (Fig. 3). Recombinant human CD3δ/ε heterodimeric protein was immobilized onto 96-well ELISA plates. The binding assays of Tb535H and purified novel tribodies tested at increasing concentrations (0 – 500 nM) were carried out according to the procedure described in Materials and Methods. As shown in Fig. 3A, all constructed tribodies showed concentration dependent CD3 binding with similar EC50 (nM) values (Fig. 3B). The results suggest that the novel tribodies fully retain the ability of parental Tb535H to bind to human CD3 protein.

Fig. 3

Binding affinity of parental Tb535H and novel 53X tribodies to recombinant CD3 protein by ELISA. A Binding curves by ELISA assays of tribodies (0 – 500 nM) to immobilized recombinant human CD3ε/δ heterodimer. Binding values were reported as the mean of determinations obtained in three independent experiments. B Table reporting the EC50 (nM) values for CD3 binding of each indicated tribody. Standard Deviations were ≤ 3—10%

Binding by ELISA of novel 53X tribodies to recombinant IC proteins and activated hPBMCs

To verify whether the novel tribodies including a binding arm to PD-L1 or PD-1 have a satisfactory binding affinity for their target recombinant proteins, we performed ELISA assays on immobilized recombinant PD-1 or PD-L1 proteins (Fig. 4). Tribodies 53L1 and 53L10, which have a binding arm specific for PD-L1, showed binding to recombinant PD-L1 protein in a concentration dependent manner with EC50 values of 79.85 nM and 29.87 nM, respectively (Fig. 4A). The tribody 53D, which contains a binding arm to PD-1, also showed a concentration-dependent binding to recombinant PD-1 protein (Fig. 4B).

Fig. 4

Binding affinity of indicated tribodies for recombinant PD-L1 or PD-1 proteins by ELISA. A Binding curves obtained by ELISA assays of 53L1 and 53L10 tribodies (0 – 500 nM) on immobilized recombinant human PD-L1 proteins. B Binding curves obtained by ELISA of 53D tribody (0–500 nM) tested on immobilized recombinant human PD-1 protein. Binding values were reported as the mean of at least three determinations obtained in three independent experiments. Standard Deviations were ≤ 3—10%

The Kd values obtained on PD-1 and PD-L1 for the tribodies are comparable to those of their parental mAbs even though the maximum absorbance value was found to be lower.

These results were confirmed when the novel 53X tribodies were tested by cell ELISA assays (Supplementary Fig. 2) for their binding to the target proteins on SEB (Staphylococcal enterotoxin B) activated lymphocytes expressing CD3 and the target ICs in their native conformation.

All the novel tribodies 53D, 53L1, 53L10, 53G were found able to bind to activated hPBMCs with a higher affinity than the parental Tb535H lacking the IC binding moiety (Supplementary Fig. 2).

To further confirm the binding specificity of the tribody 53G for LAG-3, we investigated also its binding on HuT 78 cell line, which is derived from CD4+ human T cell lymphoma and expresses high levels of LAG-3 receptor, but low or absent levels of PD-1 and PD-L1. We tested it at increasing concentrations, in parallel with the LAG-3_1 parental mAb, by cell ELISA and we found that the tribody 53G bound to the cells with a comparable specificity (Supplementary Fig. 3).

Antagonistic effects of the 53X tribodies tested by competitive ELISA assays

To verify whether the novel human tribodies 53D, 53L1 and 53L10, derived from PD-1_1, PD-L1_1 and 10_12 antibody binding domains, respectively, retain the ability of the parental mAbs to interfere in PD-1/PD-L1 interaction, we performed competitive ELISA assays by measuring the binding of biotinylated PD-L1 ligand to immobilized PD-1 receptor in the absence or in the presence of a molar excess (5:1 M/M) of the tribodies 53D, 53L1, 53L10 or their parental mAbs PD-L1_1,PD-1_1 and 10_12, used in parallel assays as positive controls. An unrelated human IgG was used as negative control. As reported in Fig. 5A and B, the binding of biotinylated PD-L1 ligand was significantly reduced in the presence of the tribodies with respect to the binding signal of the biotinylated PD-L1 used alone, thus suggesting that the novel tribodies retain the ability of the parental mAbs to interfere in the interactions of PD-L1 with PD-1.

Fig. 5

Antagonistic effects of the 53X tribodies tested by Competitive ELISA assays. The ELISA assays were performed by measuring the binding of each biotinylated chimeric PD-L1 or MHC II protein (striped bars) to immobilized PD-1 (A-B) or LAG-3 (C) proteins, respectively, in the absence or in the presence of the unlabelled competitive anti-PD-L1 (53L1 and 53L10) (A), anti-PD-1 (53D) (B) or anti-LAG-3 (53G) (C) tribodies (black bars) or their respective parental mAbs (grey bars) used at saturating concentrations (5:1 M/M for anti-PD-L1 or anti-PD-1, and 3:1 M/M for anti-LAG-3 excess ratio). Unrelated IgGs were used as negative controls (empty bars). Binding values were reported as the mean of at least three determinations obtained in three independent experiments. Error bars depicted means ± SD. P-values for the indicated compounds are: ** P < 0.01; * P < 0.05 by student’s t-test (two variables)

In parallel assays, we analyzed the ability of the tribody 53G, to interfere in LAG-3/MHCII (HLA-DRA) interaction, by performing similar competitive ELISA assays. To this aim, LAG-3-His-GST recombinant protein was coated on plates at the concentration of 50 nM, the tribody 53G or its parental mAb LAG-3_1 were added at saturating concentration of 2 μM and the binding of the Biotinylated-MHCII protein (700 nM) was measured. As shown in Fig. 5C, the binding of biotinylated MHCII was strongly reduced in the presence of the tribody 53G or the parental LAG-3_1 mAb, used as control, with respect to the binding signal of the biotinylated MHC II used alone.

T cell activation bioassays in the presence of 5T4-expressing target cells

Activation of T cell receptor (TCR)/CD3 on effector cells by novel tribodies and parental Tb535H was examined on 5T4-expressing CHO-K1 (CHO-5T4) cells, used as target cells. The assays were performed by using T cell activation bioassays (NFAT). Figure 6A shows the schematic representation of the assay. When cell surface 5T4 on target cells and CD3 on the effector cells are cross-linked with a bi-specific T cell engager, TCR/CD3 transduces intracellular signals resulting in NFAT-response element (NFAT-RE)-mediated luminescence. Genetically engineered Jurkat T cell line which expresses a luciferase reporter driven by a NFAT-RE (“TCR/CD3 effector cells”) were co-cultured with CHO-5T4 cells (“target cells”) in the presence of increasing concentrations of parental Tb535H or novel 53X tribodies, as indicated (Fig. 6B). Parental Tb535H induced luciferase activity in a concentration dependent manner and EC50 was found to be 0.11 nM. The novel tribodies also induced a concentration-dependent luciferase activity with EC50 values in the range of 0.03 nM – 0.16 nM (Fig. 6C). The 5T4-mediated CD3 activation by novel tribodies (with the exception of 53D) was slightly better than that of the parental Tb535H; the EC50 values of novel tribodies were within almost three-fold lower than that of parental Tb535H.

Fig. 6

T cell activation bioassays in the presence of 5T4-expressing target cells. A Schematic representation of TCR/CD3 activation in the presence of 5T4-expressing cells by T cell activation bioassay (NFAT). B Genetically engineered Jurkat T cells which expresses a luciferase reporter driven by a NFAT-response element (NFAT-RE) (“TCR/CD3 effector cells”) were incubated with increasing concentrations of the indicated tribodies in the presence of CHO-K1-5T4 cells. After 4 h incubation at 37℃, Bio-Glo™ reagent was added and luminescence was measured by a luminometer. Luminecence values were reported as the mean of determinations obtained in three independent experiments. C Table reporting the EC50 (nM) values for 5T4-mediated TCR/CD3 activation of each indicated construct

In vitro cytotoxic effects of the novel human tribodies on human tumor cells co-cultured with human lymphocytes

To verify whether the novel tribodies targeting PD-L1, PD-1 and LAG-3, efficiently induce the activation of lymphocytes against cancer cells, we investigated their effects on co-cultures of tumor cells and lymphocytes. To this aim, MDA-MB-231 breast cancer cells or lung A-549 cancer cells expressing high levels of PD-L1 and 5T4 were co-cultured with hPBMCs (Effector: Target ratio of 5:1) in the absence or presence of increasing concentrations of the tribodies 53D, 53G, 53L1 and 53L10 for 48 h at 37°C. In parallel assays, we tested their parental mAbs PD-L1_1, 10_12, PD-1_1 and LAG-3_1 used in combination with the parental tribody Tb535H at the same concentrations and a control tribody (53P), derived from Tb535H but lacking immunomodulatory antibody fragments. After the incubation, cells lysis was measured by the lactate dehydrogenase (LDH) released in the cell supernatants by using LDH assays. The results show that all the novel tribodies, with the exception of 53P, used as a negative control, induced tumor cells lysis with higher efficacy than the parental mAbs or their combinations with Tb535H (see Figs. 7 and 8). Specifically, all the tribodies at least doubled the levels of LDH release of their corresponding parental mAbs and parental Tb535H tribody used in combination, with the tribodies 53L1 and 53L10 reaching about 80–90% of lysis at a concentration of 1 nM.

Fig. 7

Cytotoxic effects of the novel tribodies or the combinations of Tb535H with the parental mAbs on MDA-MB-231 cells co-cultured with hPBMCs. Breast MDA-MB-231 tumor cells were co-cultured with hPBMCs (Effector:Target cells ratio 5:1) and treated for 48 h with (A) 53L1 and 53L10 (striped bars), (B) 53D (striped bars), (C) 53G (striped bars). Tb535H (light grey bars), parental immunomodulatory mAbs (dark grey bars) or their combinations (black bars) were also tested at the indicated concentrations. Co-cultures untreated or treated with the unrelated Palivizumab-derived tribody 53P (C, empty bars) were used as negative controls. Cell lysis was measured by detecting LDH release, as described in Methods. Error bars depict means ± SD. P-values for the indicated compounds are: *** P ≤ 0,001; ** P < 0.01; * P < 0.05 by student’s t-test (two variables)

Fig. 8

Cytotoxic effects of the novel tribodies or the combinations of Tb535H with the parental mAbs on A-549 cells co-cultured with hPBMCs. Lung A-549 tumor cells were co-cultured with hPBMCs (Effector:Target cells ratio 5:1) and treated for 48 h with (A) 53L1 and 53L10 (striped bars), (B) 53D (striped bars) or (C) 53G (striped bars). Tb535H (light grey bars), parental immunomodulatory mAbs (dark grey bars) or their combinations (black bars) were also tested in parallel assays at the indicated concentrations. Co-cultures untreated or treated with the unrelated tribody 53P (C, empty bars) were used as negative controls. Cell lysis was measured by detecting LDH release, as described in Methods. Error bars depict means ± SD. P-values for the indicated compounds are: *** P ≤ 0,001; ** P < 0.01; * P < 0.05 by student’s t-test (two variables)

In order to clarify whether the enhanced tumor cell lysis mediated by the novel tribodies is correlated to the increased activation of lymphocytes, we measured the levels of IFNγ, as a marker of hPBMCs functionality [32], released in the supernatants of these co-cultures, by comparing the effects of the novel tribodies with those of the parental mAbs when used in combination with parental tribody Tb535H. As shown, IFNγ levels were much higher in the supernatants of MDA-MB-231 (Fig. 9) and A-549 (Fig. 10) co-cultures treated with the novel tribodies with respect to those observed in the treatments with their parental tribody Tb535H, their parental mAbs PD-L1_1, 10_12, PD-1_1 and LAG-3_1, or their combinations.

Fig. 9

Effects of novel tribodies or the combinations of Tb535H with the parental mAbs on the secretion of IFNγ by co-cultures of MDA-MB-231 with hPBMCs. The secretion of IFNγ was measured by ELISA on supernatants of MDA-MB-231 cells co-cultured with hPBMCs and treated for 48 h with (A) 53L1 and 53L10 (striped bars), (B) 53D (striped bars), (C) 53G (striped bars). Tb535H (light grey bars), parental immunomodulatory mAbs (dark grey bars) or their combinations (black bars) were also tested at the indicated concentrations for comparison. Co-cultures untreated or treated with the unrelated tribody 53P (C, empty bars) were used as negative controls. Error bars depict means ± SD. P-values for the indicated compounds are: *** P ≤ 0,001; ** P < 0.01; by student’s t-test (two variables)

Fig. 10

Effects of novel tribodies or the combinations of Tb535H with the parental mAbs on the secretion of IFNγ by co-cultures of A-549 cells with hPBMCs. The secretion of IFNγ was measured by ELISA on supernatants of A-549 cells co-cultured with hPBMCs and treated for 48 h with (A) 53L1 and 53L10 (striped bars), (B) 53D (striped bars), (C) 53G (striped bars). Tb535H (light grey bars), parental immunomodulatory mAbs (dark grey bars) or their combinations (black bars) were also tested at the indicated concentrations for comparison. Co-cultures untreated or treated with the unrelated tribody 53P (C, empty bars) were used as negative controls. Error bars depict means ± SD. P-values for the indicated compounds are: *** P ≤ 0,001; ** P < 0.01; * P < 0.05 by student’s t-test (two variables)

To exclude the possibility that the 53X tribodies exert nonspecific side effects on circulating naive lymphocytes in the absence of tumor cells, the novel tribodies were incubated with unstimulated lymphocytes for 48–66 h. The levels of IL-2 and IFNγ cytokines released from unstimulated lymphocytes treated with tribodies were found to be very low (Supplementary Fig. 4), thus proving that the effects of the tribodies are exerted only on activated lymphocytes expressing the ICs in the presence of tumor cells.

Thus, these results clearly confirm the validity of our strategy to insert an immunomodulatory moiety in the parental Tb535H to increase its anti-tumor efficacy with no significant enhancement of off-target side effects.

In vivo anti-tumor efficacy of novel 53X tribodies

In vivo anti-tumor efficacy of the novel tribodies was evaluated on A-549 (human lung cancer cells) xenograft mouse models. Human PBMCs were activated by Dynabeads Human T activator CD3/CD28 for 4 days before the transplantation (activated PBMCs) to amplify the T cell population responsible for cytotoxicity of Tb535H against tumor cells. A-549 lung cancer cells were subcutaneously transplanted in the right flank with equal number of activated human PBMCs in immunodeficient NOD-SCID mice (day 0). Intravenous administrations of the indicated tribodies were performed every other day from day 0 to day 10 (total 6 treatments) (Fig. 11). In the parental Tb535H treatment group, significant tumor growth inhibition was observed at the dose of 20 µg/mouse compared with vehicle treatment (P < 0.05), although the lower dosage (2 µg/mouse) did not show significant tumor growth inhibition (Fig. 11A). On the other hand, the anti-tumor activity of 53L1 (20 µg/mouse) was significantly potentiated compared with that of Tb535H used at the same dosage (Fig. 11B). Furthermore, in 53L10 treatment group, significant tumor growth inhibition of A-549 tumor by treatment with both 2 µg/mouse and 20 µg/mouse dosage was observed compared with vehicle treatment (Fig. 11C). The tumor growth inhibition (TGI; %) in 53L10 (2 µg/mouse) treatment group on day 50 reached 41.5% compared with that of vehicle treatment group (P < 0.05) and 100% TGI by 53L10 at higher dosage treatment group (20 µg/mouse) was observed (P < 0.05). Complete tumor regression in all mice even at the final day (day 50) of the study was observed in 53L10 (20 µg/mouse) treatment group.

Fig. 11

In vivo Anti-tumor efficacy of parental Tb535H, 53L1 and 53L10 novel tribodies. Tumor growth curves of A-549 subcutaneous tumors treated in the presence of human PBMCs with (A) parental Tb535H, (B) 53L1, and (C) 53L10 at two different doses (〇: 2 μg/mouse treatment, △: 20 μg/mouse treatment). ●: vehicle (PBS) treatment. Five × 106 cells of A-549 cells were subcutaneously transplanted into right flank of NOD-scid mice together with equal number of human activated PBMC (hPBMC). hPBMC were stimulated with Dynabeads Human T-Activator CD3/CD28 (Veritas) for 4 days before the transplantation. Intravenous treatment of indicated samples at the indicated dosage were performed at day 0, 2, 4, 6, 8, 10 (total 6 treatment) after the cell transplantation. Tumor volumes were expressed as mean ± standard deviation (SD). *P < 0.05 by Dunnett test (vs vehicle treatment group)

Tumor regression by treatment with 53L10 was reproduced also in a different set of the study (Supplementary Fig. 5). The growth of A-549 subcutaneous tumors was again significantly inhibited by 53L10 administration in a dose dependent manner and the complete tumor regression in all mice even at the final day (day 42) was observed in 53L10 (20 µg/mouse) treatment group.

Anti-tumor activities of 53D and 53G, respectively, was also evaluated in the same models in comparison with the control 53P tribody (Fig. 12). Although the patterns of tumor growth inhibition by 53D was similar to that of the control 53P, tumor growth inhibition obtained with 53G treatment (both at 2 µg/ and 20 µg/mouse dose) was significantly stronger compared with vehicle treatment. Significant tumor regression during the study was observed in 53G (20 µg/mouse) treatment group.

Fig. 12

In vivo anti-tumor efficacy of 53D and 53G compared to 53P. Tumor growth curves of A-549 subcutaneous tumors treated in the presence of human PBMCs with (A) 53P (negative control), (B) 53D, and (C) 53G at two different doses (〇: 2 μg/mouse treatment, △: 20 μg/mouse treatment). ●: vehicle (PBS) treatment. Tumor volume was expressed as mean ± standard deviation (SD). *P < 0.05 by Dunnett test (vs vehicle treatment group)