PD-L1 targeting bispecific T cell engager kills PD-L1 tumor cells while blocking PD-1 and changing the tumor immune environment

The sequences of PD-L1-specific VHH clones 104D2, 104F5, and 104E12 were sourced from US Patent No. US-20160280786-A1. To generate PD-L1xVδ2 bsTCEs, these VHH clones were fused to the Vδ2-TCR specific VHH clone 5C816 using a Gly4Ser linker (VHH-G4S-VHH). Constructs were generated in both C-terminal and N-terminal orientations, resulting in the following bsTCEs: 104D2×5C8, 5C8×104D2, 104F5×5C8, 5C8×104F5, 104E12×5C8, and 5C8×104E12. Purified proteins were produced by ImmunoPrecise Antibodies using DNA-transfected HEK293E cells, rmp-Protein-A affinity chromatography and preparative size-exclusion chromatography.

HEK293T wild type (WT) cells were transfected to express PD-L1 using 50 µg pLX304 (HsCD00440445, DNASU) and 100 µg polyethylenimine (23 966–1, Polysciences). PD-L1 expression was checked after 24 hours using flow cytometry and PD-L1⁺ HEK293T cells were enriched using magnetic cell separation with antigen-presenting cell (APC)-labeled PD-L1 (clone MIH1, 17-5983-42, eBioscience) combined with anti-mouse IgG microbeads (130-048-402, Miltenyi Biotec). Melanoma cell lines BRO,17 WM918 and MEL-AKR19 were described previously and their identities were confirmed by short tandem repeat analysis, using the human cell line authentication service provided by Eurofins based on 21 independent PCR-single-locus-technology, following the ISO 17025 standard guidelines (https://www.eurofinsgenomics.eu/en/genotyping-gene-expression/applied-genomics-services/cell-line-authentication/↗). A375 was obtained from ATCC (CRL-1619). MEL-23 and MEL-25 were in-house generated from metastatic tumor samples collected from advanced-stage patients. These patients were enrolled under informed written consent in an institutional review board (IRB)-approved clinical study of autologous whole-cell vaccination at the VU University medical center between 1987 and 1998.20 The collected tumor tissue was minced with a scalpel and dissociated three times for 45 min with 0.02% DNase I (Boehringer) and 0.14% collagenase (Boehringer) in Hanks’ Balanced Salt Solution (Whittaker Bioproducts). The obtained cell suspensions were processed and cryopreserved within 24 hours of surgical removal.21 After thawing single-cell suspensions were cultured until confluency after which non-adherent cells were removed.22 Melanoma identity was confirmed by flow cytometry-based staining for melanoma-associated chondroitin sulfate proteoglycan.22 Above-mentioned tumor cell lines were maintained in Dulbecco’s Modified Eagle’s Medium (41 965–039, Gibco) supplemented with 10% (v/v) fetal calf serum (04-007-1A, Biological Industries), 0.05 mM β-mercaptoethanol (200-646-6, Merck), 100 IU/mL sodium penicillin, 100 µg/mL streptomycin sulfate and 2.0 mM L-glutamine (PSG, 10 378–016, Life Technologies). Vγ9Vδ2-T cells were isolated from healthy donor peripheral blood mononuclear cells (PBMCs), expanded as described before.6 In short, Vγ9Vδ2-T cells were isolated from PBMC using magnetic bead sorting with fluorescein isothiocyanate (FITC)-labeled Vδ2 mAbs (online supplemental table S1) in combination with anti-mouse IgG microbeads (Miltenyi). Purified Vγ9Vδ2-T cells were stimulated weekly with irradiated feeder mix consisting of healthy donor PBMC (1×106 cells/mL), JY cells (1×105 cells/mL, 94022533, ECACC), interleukin (IL)-7 (10 U/mL, R&D Systems), IL-15 (10 ng/mL, eBioscience), and PHA (50 ng/mL, Thermo Fisher Scientific) and used in experiments if purity of Vγ9Vδ2-T cells was >95% of total cells. Dead feeder cells and debris were removed from Vγ9Vδ2-T cells by density gradient centrifugation using Lymphoprep (AXI-1114547, Fresenius) before using them in functional experiments. Cell lines were kept at 37°C in a humidified atmosphere containing 5% CO₂.

Cells were resuspended in phosphate-buffered saline (PBS) (1073508600, Fresenius Kabi) supplemented with 0.5% bovine serum albumin (M090001/03, Fisher Scientific) and 20 mg/mL NaN3 (247-852-1, Merck) and incubated with fluorochrome-labeled antibodies (Abs,) for 30 min at 4°C. Unbound fluorochrome-labeled Abs were washed away. The LSR Fortessa XL-20 (BD) was used for data acquisition and flow cytometry data were analyzed using Kaluza Analysis V.1.3 (Beckman Coulter) or FlowJo V.10.6.1 and 10.7.2 (Becton Dickinson). Cytometric bead array (CBA) data were analyzed with FCAP Array software V.3.0 (BD) online supplemental table S1

Binding of PD-L1xVδ2 bsTCEs to PD-L1 and Vγ9Vδ2-T cells was determined by incubating HEK293T WT (control), HEK293T PD-L1⁺ or expanded healthy donor-derived Vγ9Vδ2-T cells with a concentration range of the PD-L1xVδ2 bsTCE for 45 min at 4°C. Cells were extensively washed (5×) to remove unbound PD-L1xVδ2 bsTCE and binding was detected using FITC-labeled rabbit-anti-llama polyclonal antibody (BET A160-100F, Bioke) which was incubated for 30 min at 4°C. Unbound detection antibody was washed away and binding was analyzed using flow cytometry.

Two assays were used to determine if the PD-L1xVδ2 bsTCE can block PD-1/PD-L1 interactions. Recombinant biotin-labeled PD-1 (10 µg/mL, 71 109–1, BPS Biosciences) was pre-incubated with HEK293T PD-L1⁺ cells for 45 min at 4°C. Cells were washed and a concentration range of the PD-L1xVδ2 bsTCE was added and incubated for 45 min at 4°C. Cells were extensively washed (5×) to remove unbound PD-L1xVδ2 bsTCE and binding of the PD-1 protein was detected using APC-labeled streptavidin conjugate which was incubated for 30 min at 4°C. Unbound streptavidin conjugate was washed away and binding was analyzed using flow cytometry.

Second, the commercially available PD-1/PD-L1 blockade bioassay (J1250, Promega) was used. To do so, CHO-K1 cells (stably expressing PD-L1 and a cell surface protein designed to activate cognate TCRs in an antigen-independent manner) were thawed, and incubated for 24 hours at 37°C and 5% CO₂. Jurkat T cells (stably expressing human PD-1 and NFAT-induced luciferase) were thawed and co-cultured with CHO-K1 cells in the presence or absence of a concentration range of PD-L1xVδ2 bsTCE, nivolumab (Bristol Myers Squibb) or durvalumab (AstraZeneca BV) for 6 hour. Bio-Glo reagent was added and luminescence was measured using a luminescence plate reader (Glomax, Promega). The following calculation was used: relative light units (RLU) antibody – background / RLU no antibody control – background=fold induction.

HEK293T WT (control), HEK293T PD-L1⁺, BRO, WM9, MEL-AKR, A375, MEL-23 or MEL-25 were co-cultured for 24 hours with expanded healthy donor-derived Vγ9Vδ2-T cells (1:1 effector to target (E:T) ratio) in the presence or absence of 1 or 100 nM PD-L1xVδ2 bsTCE. PD-L1xVδ2 bsTCE-mediated Vγ9Vδ2-T cell activation was determined by assessing expression of CD107a (anti-CD107a mAb added at the start of culture), CD25 and 4-1BB. Tumor cytotoxicity was determined using a viability dye in combination with 123count eBeads counting beads (01-1234-42, Thermo Fisher Scientific).

For functional experiments, patient-derived cryopreserved dissociated primary renal cell carcinoma (RCC) and metastatic melanoma single-cell suspensions (dissociation process described before21) were used. RCC and melanoma single-cell suspensions were thawed and cultured for 2 or 5 days with or without expanded healthy donor-derived Vγ9Vδ2-T cells and 10 µg/mL durvalumab, 100 nM PD-L1xVδ2 bsTCE or medium control. Expression of activation (CD25, 4-1BB) and degranulation (CD107a) markers and expression of PD-1 and TIM-3 on Vγ9Vδ2-T cells, CD4⁺ T cells and CD8⁺ T cells was determined after 2 days using fluorescently labeled antibodies and flow cytometry. Supernatants collected from these 2-day cultures were used to analyze IL-2, IL-4, IL-6, IL-10, IFNγ, TNF and IL-17A secretion through the human Th1/Th2/Th17 CBA kit (560484, BD). Phenotypic analysis of CD45⁺HLA-DR⁺CD11c⁺ myeloid cells was done after 5 days using fluorescently labeled antibodies against CD80, CD83, CD86, CD14, CD163, BDCA-3 and TIM-3 throughflow cytometry. Tumor cytotoxicity was determined after 2 and 5 days using 7AAD (A9400-1MG, Sigma) and 123counting eBeads (01-1234-42, Thermo Fisher) followed by flow cytometry analysis.

In order to determine PD-L1xVδ2 bsTCE-mediated Vγ9Vδ2-T cell activation, tumor cell cytotoxicity and infiltration in a three-dimensional in vitro setting, BRO tumor cells were used to generate tumor spheroids. BRO tumor cells (5,000) were cultured overnight in nuclon sphere-treated 96-well plates (174925, Thermo Fisher Scientific) to generate spheroids. Expanded healthy donor-derived Vγ9Vδ2-T cells (6,000) were co-cultured with 1 BRO spheroid per condition for 24 hours in the presence or absence of 100 nM PD-L1xVδ2 bsTCE. Spheroid-infiltrated Vγ9Vδ2-T cells (termed IN fraction) were separated from non-infiltrated Vγ9Vδ2-T cells (termed OUT fraction) by washing twice with PBS and carefully removing the supernatant from the spheroids after they had sunk to the bottom. Fluorescently labeled antibodies against 4-1BB, CD25 and PD-1 were used for phenotypic analysis of Vγ9Vδ2-T cells. Cytotoxicity of BRO spheroids was determined by trypsinizing them to retrieve a single-cell suspension (also comprising the IN Vγ9Vδ2-T cell fraction). Cells were prepared and stained for flow cytometry analysis.

Vγ9Vδ2-T cell infiltration into BRO spheroids was visualized using confocal microscopy. To do so, Vγ9Vδ2-T cells were stained with a SPY555-DNA dye (1000×, SC201, Spirochrome) for 24 hours at 37°C while BRO tumor cells were stained with a SPY650-DNA dye (1000×, SC501, Spirochrome) during spheroid formation. Cells were washed, transferred to a high bioinert 8-well imaging µ-slide (80800, Ibidi) and co-cultured (1 BRO spheroid and 6,000 Vγ9Vδ2-T cells) in the presence or absence of 100 nM PD-L1xVδ2 bsTCE in the environmental chamber at 37°C and 5% CO₂ within the Nikon AXR LC for 8 hours. Each well was imaged every 20 min with 52 Z-stacks and a 10x objective. Images were analyzed with Imaris (Oxford Instruments), which was used to identify spheroids by using the “surface” tool, and Vγ9Vδ2-T cells by using the “spots” tool. The program calculated the shortest distance of each Vγ9Vδ2-T cell (spot) to the spheroid (surface). The sum of the negative values (Vγ9Vδ2-T cells within spheroid) was calculated per condition per time point to determine the amount of spheroid-infiltrating Vγ9Vδ2-T cells over time.

GraphPad Prism V.9.1.0 (GraphPad Software) was used for statistical analyses. Data were analyzed using paired t test or one-way analysis of variance with Dunnett multiple comparisons test. p<0.05 was considered significant and indicated with asterisks: *p<0.05, **p<0.01, ***p<0.001 and ****p<0.0001. EC50 values were calculated using non-linear regression analysis with GraphPad Software.

Six PD-L1xVδ2 bsTCEs were generated by fusing PD-L1-specific VHH clones 104D2, 104F5, and 104E12 to the high-affinity Vδ2-TCR-specific VHH clone 5C8, using a Gly4Ser linker, in both N-terminal and C-terminal orientations. While PD-L1xVδ2 bsTCEs containing PD-L1 VHH clones 104D2 and 104F5 bound to PD-L1⁺ HEK293T tumor cells, no binding was observed with the PD-L1xVδ2 bsTCEs incorporating clone 104E12 (figure 1A,B; figure 1A shows PD-L1 expression on PD-L1 transfected HEK239T cells compared with WT cells). The ability of the 104D2-VHH and 104F5-VHH containing PD-L1xVδ2 bsTCEs to inhibit PD-1/PD-L1 interactions was assessed by studying recombinant biotin-labeled PD-1 displacement from PD-L1⁺ HEK293T tumor cells. As shown in figure 1C, increasing concentrations of both PD-L1xVδ2 bsTCEs resulted in a dose-dependent reduction of PD-1 binding. Displacement of PD-1 was most pronounced with C-terminal positioning of the PD-L1 VHHs (EC₅₀ values: 5C8×104D2 0.62 nM and 5C8×104F5 0.52 nM vs 104D2×5C8 11.05 nM and 104F5×5C8 7.87 nM) with no notable difference between clones 104D2 and 104F5. Based on its higher binding intensity to PD-L1⁺ HEK293T tumor cells, the 5C8×104D2 based bsTCE was selected for further experiments and from here on referred to as PD-L1xVδ2 bsTCE. This PD-L1xVδ2 bsTCE bound with high affinity (EC₅₀ for binding 0.70 nM) to PD-L1⁺ HEK293T cells and showed no binding to non-transfected HEK293T cells (figure 1D). The EC₅₀ for binding to Vγ9Vδ2-T cells was 0.75 nM which is in line with our earlier observations for the Vδ2-specific VHH9 (figure 1E). To evaluate whether the PD-L1xVδ2 bsTCE could also functionally block the interaction between PD-1 and PD-L1, a cell-based bioassay was used in which release of the PD-1/PD-L1 axis was tested in co-cultures of PD-1⁺ Jurkat cells and PD-L1⁺ CHO cells. As shown in figure 1F, the PD-L1xVδ2 bsTCE resulted in a dose-dependent abrogation of the PD-L1 mediated inhibition of PD-1⁺ Jurkat cells, as reflected by an increase in luciferase activity. A similar effect was observed with the positive control antibodies in these assays (ie, the PD-L1 blocking mAb durvalumab and the PD-1 blocking mAb nivolumab).

Collectively, these data show that the generated PD-L1xVδ2 bsTCE mediates specific and high-affinity binding to PD-L1-expressing cells as well as Vγ9Vδ2-T cells and has the ability to block the interaction between PD-1 and PD-L1.

We next explored the ability of the PD-L1xVδ2 bsTCE to induce Vγ9Vδ2-T cell activation and tumor cell lysis. In 24 hours co-cultures of Vγ9Vδ2-T cells and HEK293T WT or HEK293T PD-L1⁺ tumor cells, Vγ9Vδ2-T cell degranulation and tumor cell lysis were only observed in co-cultures with PD-L1⁺ tumor cells (figure 2A,B). No reactivity was observed using a control bsTCE (104E12×5C8). As these HEK293T PD-L1⁺ tumor cells expressed high levels of PD-L1, we next explored the activity of the PD-L1xVδ2 bsTCE in co-cultures of Vγ9Vδ2-T cells and the patient-derived melanoma tumor cell lines BRO, WM9, AKR and A375 that express lower to substantially lower levels of PD-L1 (figure 2C). As shown in figure 2D,E, the PD-L1xVδ2 bsTCE triggered Vγ9Vδ2-T cell activation, as assessed by upregulation of the activation markers CD25 and 4-1BB and the degranulation marker CD107a, and subsequent tumor cell lysis of all of these tumor cell lines, despite PD-L1 expression being quite low in WM9 and AKR cell lines. It is known that IFNγ, which is secreted at high levels by activated Vγ9Vδ2-T cells,9 10 can upregulate PD-L1 expression.23 IFNγ induced PD-L1 upregulation on the melanoma tumor cell lines was confirmed, and may therefore have contributed to the observed potency of the PD-L1xVδ2 bsTCE against PD-L1^low melanoma tumor cells (figure 2F). The activity of the PD-L1xVδ2 bsTCE was also tested using low-passage melanoma cell lines (two to three passages) that were isolated in-house from lymph node metastases of two patients. Both of these patient-derived cell lines expressed PD-L1, which could be further increased by IFNγ (figure 2G,H). When Vγ9Vδ2-T cells were co-cultured with these low-passage melanoma cell lines, the PD-L1xVδ2 bsTCE triggered robust Vγ9Vδ2-T cell degranulation and tumor cell lysis (figure 2I).

In conclusion, the generated PD-L1xVδ2 bsTCE efficiently cross-links Vγ9Vδ2-T cells and PD-L1 expressing tumor cell lines (both established and low-passage patient-derived) to trigger Vγ9Vδ2-T cell activation and lysis of PD-L1⁺ tumor cells.

Next we assessed whether the PD-L1xVδ2 bsTCE could also induce Vγ9Vδ2-T cell activation and tumor lysis in co-cultures of Vγ9Vδ2-T cells with patient-derived dissociated RCC and metastatic melanoma tumor samples, comprising the entire TME. The majority of tumor cells present in these samples was found to express PD-L1 (figure 3A). Vγ9Vδ2-T cells were added to tumor samples in a 1:10 E:T ratio as infiltrating Vγ9Vδ2-T cell frequencies were relatively low and variable between tumor suspensions (figure 3B). After a 2-day co-culture, Vγ9Vδ2-T cells expressed significantly higher levels of the activation markers CD25 and 4-1BB, the degranulation marker CD107a, and the immune checkpoint receptors PD-1 and TIM-3 (figure 3C). Upregulation of these markers was generally less pronounced when Vγ9Vδ2-T cells were exposed to melanoma as compared with RCC samples. Levels of these receptors on Vγ9Vδ2-T cells did not change in co-cultures containing the anti-PD-L1 mAb durvalumab rather than the PD-L1xVδ2 bsTCE (figure 3D). Consistent with the observed activation of Vγ9Vδ2-T cells in response to the PD-L1xVδ2 bsTCE, supernatants from the 2-day co-cultures showed significantly increased levels of IFNγ, TNF, and IL-2. There was a trend for higher levels of IL-4 in the PD-L1xVδ2 bsTCE treated conditions, while levels of IL-6 and IL-10 did not notably change. In none of the conditions IL-17A could be detected. Lysis of the patient RCC and melanoma cells was assessed after 2 and 5 days of co-culture. Vγ9Vδ2-T cells alone exhibited variable tumor cytotoxicity against both RCC and melanoma cells, but this was significantly enhanced in the presence of the PD-L1xVδ2 bsTCE. The PD-L1xVδ2 bsTCE-induced tumor cell lysis increased between day 2 and day 5 (figure 3E). Durvalumab did not induce any Vγ9Vδ2-T cell activation nor enhance tumor cytotoxicity in these co-cultures.

Overall, these data showed that despite the presence of the (suppressive) TME, the PD-L1xVδ2 bsTCE could still trigger a robust Vγ9Vδ2-T cell-mediated antitumor response in tumor samples obtained from patients with RCC and metastatic melanoma.

PD-L1xVδ2 bsTCE-mediated Vγ9Vδ2-T cell engagement leads to coactivation of conventional CD4⁺ and CD8⁺ T cells and predominance of mature dendritic cells in the tumor microenvironment.

As it is known that interactions between PD-1 and PD-L1, expressed by tumor cells and tumor-infiltrating myeloid cells, drive effector T-cell dysfunction in the TME,24 we next explored whether the PD-L1xVδ2 bsTCE could overcome such effects in tumor samples of patients with RCC and melanoma. For this purpose, dissociated RCC and melanoma samples were co-cultured for 48 hours with expanded Vγ9Vδ2-T cells at an E:T ratio of 1:10 in the presence or absence of either durvalumab or the PD-L1xVδ2 bsTCE. In the presence of the PD-L1xVδ2 bsTCE there was a statistically significant increase in the expression of the activation marker 4-1BB and the degranulation marker CD107a on both RCC infiltrating conventional CD4⁺ and CD8⁺ T cells (figure 4). A similar pattern was observed in melanoma tumor samples, though here it did not always reach statistical significance. With the exception of upregulation of CD25 on melanoma-infiltrating CD8⁺ T cells and TIM-3 on RCC-infiltrating CD8⁺ T cells, no other statistically significant changes were noted for CD25, PD-1 and TIM-3. While exposure of the tumor samples to durvalumab in some cases resulted in minor (not statistically significant) differences in marker expression, the modulatory effects triggered by the presence of the PD-L1xVδ2 bsTCE were clearly more pronounced and suggestive of a direct contribution of the activated Vγ9Vδ2-T cells over inhibition of the PD-1/PD-L1 checkpoint alone.

We next assessed whether the PD-L1xVδ2 bsTCE could affect the frequency and phenotype of myeloid cells. For this purpose, melanoma patient-derived single-cell tumor suspensions were cultured with expanded Vγ9Vδ2-T cells (E:T ratio 1:10) in the presence or absence of durvalumab or the PD-L1xVδ2 bsTCE for 5 days. At baseline, PD-L1 was found to be expressed by a median of 17.5% (range 7.7–35.9%, figure 5A) of the myeloid cells (defined as CD45⁺HLA-DR⁺CD11c^high) and gated to exclude lymphocytes25. As these cultures were performed from cryopreserved samples, no surviving granulocytes were present in the cultures and we have previously shown that all thus gated myeloid cells are from monocytic origins based on CD88 and CD89 expression.25 on co-culture with Vγ9Vδ2-T cells, the PD-L1xVδ2 bsTCE triggered substantial lysis of myeloid cells (figure 5B). The observed lysis typically exceeded the percentage of myeloid cells expressing PD-L1 at baseline, which may be due to upregulation of PD-L1 during the co-culture (ie, by IFNγ produced by activated Vγ9Vδ2-T cells).23 Indeed, myeloid cell PD-L1 was typically increased at day 5, except in the conditions exposed to durvalumab (where reliable PD-L1 assessment was complicated by interference with the mAb used for PD-L1 detection) and the PD-L1xVδ2 bsTCE, likely because of lysis of the PD-L1 expressing myeloid cells (figure 5C). Of interest, in the cultures exposed to the PD-L1xVδ2 bsTCE, a clear phenotypic shift could be noted in the remaining myeloid cells. While CD80 expression remained unchanged, expression of CD83 and CD86 was significantly increased on these cells, indicative of a phenotype consistent with mature dendritic cells (figure 5D). These cells also exhibited significantly reduced expression of CD14, CD163 and BDCA-3, along with a trend towards lower TIM-3 levels, suggesting a phenotype associated with attenuated suppressive and enhanced T-cell stimulatory functionality (figure 5E).

Overall, these data demonstrate that the PD-L1xVδ2 bsTCE can trigger Vγ9Vδ2-T cells to mediate a shift in the TME characterized by an increase in 4-1BB and CD107a expressing CD4⁺ and CD8⁺ T cells, elimination of PD-L1 expressing tumor and myeloid cells and an associated increase in rates of mature dendritic cells expressing higher levels of co-stimulatory molecules and lower levels of suppression related receptors.

Tumor spheroids were generated and used as a model to determine PD-L1xVδ2 bsTCE-mediated infiltration and cytolytic effects of Vγ9Vδ2-T cells in a three-dimensional tumor model (illustrated in figure 6A). Spheroids were generated by culturing melanoma-derived BRO tumor cells overnight in non-adhesive culture plates. During this overnight incubation, BRO tumor cells retained PD-L1 expression (figure 6B). Spheroids were subsequently co-cultured with expanded Vγ9Vδ2-T cells for an additional 8 hours, during which confocal microscopy in combination with Z-stacks (stack of images captured at different focal depths along the Z-axis) was used to image and track Vγ9Vδ2-T cells, or 24 hours after which the supernatant, containing non-infiltrated Vγ9Vδ2-T cells (termed OUT), was harvested and these OUT Vγ9Vδ2-T cells were compared with infiltrated Vγ9Vδ2-T cells retrieved from the spheroids after trypsinization (termed IN). The PD-L1xVδ2 bsTCE promoted rapid infiltration of Vγ9Vδ2-T cells into BRO spheroids within the first 2–3 hours, after which it stabilized (figure 6C). In the absence of the PD-L1xVδ2 bsTCE, some Vγ9Vδ2-T cell infiltration was observed in the first hour, but decreased thereafter. As depicted in figure 6D, in the absence of the PD-L1xVδ2 bsTCE, Vγ9Vδ2-T cells (green) predominantly surrounded the BRO spheroid (purple) without showing substantial infiltration. In the presence of the PD-L1xVδ2 bsTCE, clear infiltration of the Vγ9Vδ2-T cells into the spheroid was noted. As expected, Vγ9Vδ2-T cells cultured in the presence of the PD-L1xVδ2 bsTCE triggered variable, but statistically significant, lysis of BRO spheroid tumor cells during the 24 hours co-culture (figure 6E). Lysis was highest in the donors that exhibited higher baseline tumor reactivity, possibly related to differences in BTN2A1/3A1, NKG2D and DNAM-1 interactions between Vγ9Vδ2-T cell donors. Co-cultures that were performed in the presence of the PD-L1xVδ2 bsTCE resulted in significant upregulation of the activation markers 4-1BB and CD25 on both spheroid-infiltrated and non-infiltrated Vγ9Vδ2-T cells (figure 6F). In contrast, PD-1 expression levels on Vγ9Vδ2-T cells that had infiltrated the spheroids were actually significantly lower than in the medium control while they were similar on the Vγ9Vδ2-T cells that had not infiltrated the spheroids. Interestingly, in the presence of the PD-L1xVδ2 bsTCE, infiltrated Vγ9Vδ2-T cells expressed significantly higher levels of 4-1BB (p=0.017) and a trend towards lower levels of PD-1 (p=0.067) compared with non-infiltrated Vγ9Vδ2-T cells that were also exposed to the PD-L1xVδ2 bsTCE. No differences were observed for CD25 levels between infiltrated and non-infiltrated Vγ9Vδ2-T cells (p=0.781) exposed to the PD-L1xVδ2 bsTCE.

Collectively, these findings demonstrate that the PD-L1xVδ2 bsTCE enhances Vγ9Vδ2-T cell infiltration into tumor spheroids, leading to their lysis. Within these spheroids, Vγ9Vδ2-T cells exhibit elevated expression of activation markers 4-1BB and CD25, alongside low expression of the checkpoint inhibitor PD-1, suggesting reduced sensitivity to inhibitory signaling.

Data are available upon reasonable request.