Vγ2 x PD-L1, a Bispecific Antibody Targeting Both the Vγ2 TCR and PD-L1, Improves the Anti-Tumor Response of Vγ2Vδ2 T Cell

The potent cytotoxic property of Vγ2Vδ2 T cells makes them attractive for adoptive T cell transfer therapy. The transfusing of the expanded Vγ2Vδ2 T cells into cancer patients shows well-tolerated, but the clinical response rates are required to be improved, implying that there is still an unmet efficacy with low toxicity for this novel anti-tumor therapy. In this study, we test the anti-tumor efficacy of a Y-body-based bispecific antibody (bsAb) Vγ2 x PD-L1 that preferentially redirects Vγ2Vδ2 T cells to combat PD-L1 positive tumor cells. With nanomolar affinity levels to Vγ2Vδ2 T cells and PD-L1+ tumor cells, Vγ2 x PD-L1 bridges a Vγ2Vδ2 T cell with a SKOV3 tumor cell to form a cell-to-cell conjugation. In a PD-L1-dependent manner, the bsAb elicits effective activation (CD25+CD69+), IFNγ releasing, degranulation (CD107a+), and cytokine production (IFNγ+ and TNFα+) of expanded Vγ2Vδ2 T cells. The activations of the Vγ2Vδ2 T cells eliminate PD-L1-expressing human cancer cell lines, including H1975, SKOV3, A375, H1299, and H2228 cells, but not PD-L1 negative cells including HEK-293 (293) cells and healthy PBMCs. Finally, we show that combining Vγ2 x PD-L1 with adoptively transferring Vγ2Vδ2 T cells inhibits the growth of existing tumor xenografts and increases the number of Vγ2Vδ2 T cells into the tumor bed. Vγ2 x PD-L1 represents a promising reagent for increasing the efficacy of adoptively transferred Vγ2Vδ2 T cells in the treatment of PD-L1 positive malignant tumors.


INTRODUCTION
Vg2Vd2 T cells, a unique fast-acting subset of innate gd T cells found exclusively in primates (1), have been widely employed for adoptive cell immunotherapy in clinical studies for treating malignancies in past years (2). These cells have NK and cytotoxic T cell features, as well as potential and intrinsic rapid anti-tumor effector capabilities (3,4), and appear to be a more promising candidate for allogeneic T cell therapy than ab T cell-based CAR-T cells by participating in immune surveillance and killing a broad spectrum of cancer cells through a major histocompatibility complex (MHC)-independent activation mechanism (5). Recently, the adoptive transfer of Vg2Vd2 T cells to cancer patients has recently been shown to extend the survivals of latestage liver cancers (23.1 vs 8.1 months) and lung cancers (19.1 vs 9.1 months) (6), and well tolerated as well (7). Yet, this therapy provided moderate clinical benefits with stable disease being the mostly outcome for patients who respond to this therapy (7).One of the reasons for this suboptimal effectiveness is the hostile tumor microenvironment that negatively regulates the anti-tumor functional characteristics of Vg2Vd2 T cells by the engagement between the programmed death-ligand 1 (PD-L1) expressed on the tumor cells and PD-1 expressed on the Vg2Vd2 T cells (8). Several groups proposed a combination approach of the Vg2Vd2 T cell-based adoptive immunotherapy with a PD-1 checkpoint blockade for the immunity against leukemia (9), follicular lymphoma (10), and prostate cancer (11). Likely, the anti-PD-L1 mAb enhances the cytotoxicity of Vg2Vd2 T cells against PD-L1 high cancer cells by adding ADCC activity (12).
After the success of targeting PD-1/PD-L1 axes, extensive efforts were directed to explore bsAb-based strategies to increase the anti-tumor activity of the adoptively transferred Vg2Vd2 T cells (13,14). As a result, two representative series of Vg2Vd2 T cell-targeting bsAbs were constructed, one targeting to Vg2-TCR and the other targeting to Vd2-TCR. BsAb [(Her-2) 2 × Vg2] increased the cytotoxicity of Vg2Vd2 T cell against Her2overexpressing pancreatic, ovarian and breast cancer cells showed by in vitro assay and in a PDAC grafted mouse model (15,16). Similarly, Vg2 x CD123 was created to treat acute myeloid leukemia (17). Lately, Vd2 x EGFR elicits Vg2Vd2 T cell-mediated killing of colon cancer cell line SW480 both in vitro and in vivo (18), Vd2 x CD1d for chronic lymphocytic leukemia (19), and Vd2 x CD40 for b-cell malignancies (20). Moreover, these Vg2Vd2 T cell-specific targeting strategies were thought to overcome T cell over-activation induced by current CD3targeting bsAbs, which could lead to cytokine storm syndrome, a severe side effect due to Treg stimulation. For example, the FDA-approved CD3 x CD19 bsAb, blinatumomab, could increase the numbers of Treg cells, which were correlated with non-responsiveness to blinatumomab in ALL patients (21) and further led to abnormal macrophage activation-dependent cytokine storm syndrome (22). Taken together, T cell engagers designed to activate Vg2Vd2 T cells exclusively might represent a feasible approach balanced between efficacy and safety.
Here, we describe the preclinical evaluation of Vg2 x PD-L1. Our findings reveal that Vg2 x PD-L1 activates selectively the fresh and expanded Vg2Vd2 T cells to kill tumor cells in vitro, enhances the migration of the transfused Vg2Vd2 T cells into tumor sites, and inhibits the growth of the existing tumors in nude mice. These data suggest that Vg2 x PD-L1 plus adoptively transferred Vg2Vd2 T cells is potential to treat PD-L1 positive solid malignancies.

Generation of the Recombinant Antibodies
The bsAbs, including Vg2 x PD-L1 and Vg2 x Null, were generated similarly to Y111 described previously by Yang et al. (23). Briefly, the expression plasmids for Vg2 x PD-L1 and Vg2 x Null were synthesized and verified by sequencing in AuGCT Biotech (Wuhan, China). Then these expression vectors were transfected into cGMP banked CHO-S cells (Invitrogen, Carlsbad, USA) using the Fecto PRO Reagent (Ployplus, New York, USA) according to the manufacturer's protocol, respectively. After a week, the cell culture supernatant was collected and serially purified by Sepharose Fast Flow protein A affinity chromatography column (GE, Milwaukee, USA), Fab Affinity KBP Agarose High Flow Resin (ACROBio systems, Newark, USA), and SP cation exchanged chromatography column (GE, Milwaukee, USA). Finally, the purified proteins were analyzed by SDS-PAGE and size-exclusion chromatograms. The Vg2 x Null served as the control molecule for Vg2 x PD-L1, with both molecules sharing the same backbone and Vg2-targeting scFv part. Similarly, its two parental monoclonal antibodies (Vg2 mAb (Clone 7A5) and PD-L1 mAb (23)) were produced.

Tumor Cell Lines Culture
Tumor cell lines, including NCI-H1975 (human adenocarcinoma epithelial cell line, CRL-5908), SKOV3 (human ovarian adenocarcinoma cell line, HTB-77), A375 (human malignant melanoma cell line, CRL-1619), NCI-H1299 (human NSCLC metastatic cell line, CRL-5803), NCI-H2228 (human NSCLC adenocarcinoma cell line, CRL-5935), and nonmalignant kidney cell line HEK-293 were purchased from ATCC (Manassas, USA) and used as target cells. These cell lines were first transduced with firefly luciferase gene-containing pseudo-typed lentiviral particles purchased from GeneCopoeia (Shanghai, China), and the stable luciferase-expression cells were then selected under pressure of puromycin (Gibco, New York, USA). CHO-PD-L1 was generated from the parental CHO-K1 cell line (CCL-61, ATCC) through overexpressing human PD-L1. Tumor cells were cultured in RPMI 1640 (Biosharp, Hefei, China), DMEM or F-12K medium (purchased from Hyclone, New York, USA) supplemented with 10% FBS (Excell, Clearwater, USA) and penicillin/streptomycin (Gibco, New York, USA) and maintained in a humidified incubator with 5% CO 2 at 37°C. All cell lines in use were routinely tested for Mycoplasma infection using a commercial PCR kit (Vazyme, Nanjing, China), and new cultures were established monthly from frozen stocks as described previously (24).

Expansion of Vg2Vd2 T Cells
The sampling protocols for human blood and in vitro experimental procedures were evaluated and approved by the institutional review boards for human subjects' research and institutional biosafety committees at Hubei Province Food and Drug Safety Evaluation Center (Wuhan, China). All subjects are volunteer adults who signed on the informed consent.
Frozen or fresh human peripheral blood mononuclear cells (PBMCs) were obtained from LeiDeBio (Guangzhou, China) or Milestone (Shanghai, China). The ex vivo expansion protocol was described previously (23,25). Briefly, PBMCs were cultured in RPMI 1640 medium (Gibco, New York, USA) supplemented with 10% FBS (Excell, Clearwater, USA), at 2×10 6 cells/mL with the stimulation of 2.5 mM Zoledronic Acid (Sigma Aldrich, Darmstadt, Germany) and 1000 IU/mL IL2 (Sihuan Pharma, Beijing, China) for 10-14 days. The expanded Vg2Vd2 T cells were negatively enriched from the cultures by a TCR g/d + T Cell Isolation Kit (Miltenyi Biotech, Teterow, Germany). The purity and quality of the isolated cells were assessed by surface staining Vg2/Vd2 and CD86/CD69/HLA-DR as described previously ( (23)). Then, the purified Vg2Vd2 T cells were maintained in RPMI 1640 medium supplemented with 10% FBS overnight for rest before use. In this study, effector Vg2Vd2 T cells were expanded and purified from a total of eight healthy individuals for in vitro functional analysis and two healthy donors for in vivo anti-tumor evaluations.

Binding Ability of Antibodies to Cells
A flow cytometry-based method was used to determine the affinities of Vg2 x PD-L1 of its anti-Vg2 arm to Vg2Vd2 T cells and its anti-PD-L1 arm to PD-L1 positive tumor cells. The sorted Vg2Vd2 T cells or tumor cells were incubated with serially diluted antibodies (Vg2 x PD-L1, Vg2 x Null, Vg2 mAb, and PD-L1 mAb) for one hour at 4°C. After wash, the cells were stained for 30 minutes at room temperature with APC or PEconjugated mouse-anti-human IgG Fc antibody (HP6017, Biolegend, San Diego, USA) diluted in 1:100. The cells were then resuspended in 200 mL FACS buffer (PBS with 2% FBS) and analyzed by a BD FACSelesta flow cytometer. For tumor cells, the cell-bound antibodies were quantified by the median fluorescence intensity (MFI) values, and the MFI were plotted against antibody concentrations to obtain the EC 50 . For Vg2Vd2 T cells, APC positive populations were used to determine the specific binding%.
The formation of an in-tans bridge between T cells and tumors cells was accessed by a flow cytometry method. Briefly, SKOV3 cells were stained with 50 nM CFSE, and the PBMC cultures (treated by Zol+IL2 for 10-14 days) were labeled by PKH26 according to the manufacturer's protocol. Then, the CFSE-stained SKOV3 cells were co-cultured with PKH26labelled PBMC cultures at a ratio of 1:1 with 1 ug/mL of Vg2 x PD-L1 or Vg2 x Null at an incubator for 0.5 hours. After washing, the cells were recorded on the FACSelesta (BD, San Jose, USA). The percentages of the CFSE + PKH26 + doublepositive cells among the total cells have represented the ratios of cells engaged in cell-to-cell association.

PD-L1 Blockade Reporter Assay
The assay was carried out following the manufacturer's instructions (Promega, Cat#J1250). Briefly, PD-L1 aAPC/ CHO-K1 cells were seeded at 4 × 10 4 cells/well at 100 mL in white 96-well plates followed by a cultured overnight in an incubator at 37°C with 5% CO2. The next day, the supernatant was discarded and the PD-L1 aAPC/CHO-K1 cells were incubated with serially diluted antibodies and PD-1 effector cells (5 × 10 4 /well) for 6 h. Then the relative luminescence units (RLU) of each well were determined using a Bio-Luc kit from Vazyme (Nanjing, China).

PD-L1 Expression Scores Determination
Tumor cell lines were incubated with 40 mg/mL Vg2 X PD-L1 (Target) or Vg2 X Null (Null) for 1 hour at 4°C, then stained with APC-conjugated mouse-anti-human IgG Fc antibody (HP6017, Biolegend, San Diego, USA) for 30 minutes at room temperature. The APC positive populations and MFI of the APC channel were determined by flow cytometry. The expression scores were defined by [log 10 (Target APC positive populations -Null APC positive populations ) + log 10 (Target APC MFI/ Null APC MFI )]/2.

Evaluate T Cell Activation by Surface Staining and Intracellular Cytokine Staining
Flow cytometry was performed to evaluate T-cell activation as described in the other reports (26,27). Expanded Vg2Vd2 T cells were enriched from PBMCs cultures (Zol+IL2 for 10-14 days), and cultured overnight. In parallel, 0.2 million H1975 or SKOV3 cells were plated in a 24-well-plate overnight. For activation assay, 0.2 million expanded and negatively enriched Vg2Vd2 T cells were added into either the tumor cell wells or empty wells with 1 mg/mL of Vg2 X PD-L1 or Vg2 X Null for 24 hours. Then

Antibodies Mediated Cytotoxicity In Vitro
Two in vitro methods including luciferase-activity based assays and CFSE-PI staining-based assay were developed to access the killing ability of Vg2Vd2 T cells mediated by antibodies.
Luciferase-activity based assays: 2 x 10 4 firefly luciferaseexpressing tumor cells (Target: T) were co-incubated with expanded Vg2Vd2 T cells (Effector: E) at an E:T ratio of 0.5:1 (or other indicated E: T ratios), or fresh enriched gd T cells (Effector) at an E:T ratio of 5:1, in the presence of a serial of diluted antibodies for 12 hours in a white 96-well-flat bottom plate. A Bio-Luc kit from Vazyme (Nanjing, China) was used to measure luciferase activity. Then the "Specific lysis" was calculated as follows: % Specific lysis = [1 -(RLU Ab-treated wells )/(RLU Targetonly wells )] × 100.
CFSE-PI staining-based assay: Unrelated healthy PBMCs were stained with CFSE according to the manufacturer's protocol. Then these cells were co-cultured with Vg2Vd2 T cells at a 1:1 E: T ratio in the presence of various doses of indicated antibodies for 12 hours. Then 1mg/mL of PI (Sigma) was added to the wells. The percentages of CFSE + PI + cells among the total of target cells (CFSE + ) were defined as "Specific Cytotoxicity%" values.

Measuring Vg2Vd2 T Cell Releasing IFNg
The supernatant was collected from T cell and tumor cell coculture wells and stored at -80°C until measurement. Human IFNg were quantified with the ELISA kits from Proteintech (KE00063, Wuhan, China).

Mouse Tumor Model
Female nude mice were obtained from the VITALSTAR (Beijing, China) at age of 6-8 weeks and were used in this study under a protocol approved by the Animal Care and Use Committee from Hubei Province Food and Drug Safety Evaluation Center (#202110191).
Firstly, 5 million SKOV3 cells were subcutaneously inoculated into the right dorsal flank of nude mice on Day 0. After one week, tumor volumes had reached around 200 mm 3 , these mice were randomly divided into three groups receiving PBS, 2 million purified Vg2Vd2 T cells i.v. through lateral tail vein plus 8 mg/kg Vg2 X Null i.p. or 8 mg/kg Vg2 X PD-L1 i.p. on Days 7,11,14, and 18 (Q2W, two weeks, four times). After treatment, tumor volumes and mice body weights were measured three times a week. The tumor volume was calculated using the formula: Tumor Volume (mm 3 ) = (a x b 2 )/2, where a is the longitudinal length and b is the transverse width. On day 34, these mice were sacrificed and tumor xenografts were excised for tumor weighting and IHC staining.

IHC Analysis
The tumor tissues were cut into small pieces embedded in 4% paraformaldehyde for fixation. Then these tumor pieces were sectioned and examined by IHC staining using a rabbit-antihuman CD3 antibody (Clone SP7). Tissue sections were then counter-stained with hematoxylin. Positive cells were counted in five randomly selected microscopic fields (magnification 20X) and supplied for further quantification analysis.

Statistical Analysis
Statistical analyses were performed with GraphPad Prism 6.0 (La Jolla, USA). Before performing nonlinear regression analysis for in vitro assays (cell binding and killing), the antibody concentrations (on the x-axis) were transformed in a log scale.
Then, the "log (agonist) vs. response-Variable slope (four parameters)" method was applied to calculate EC 50 . P values were assessed by one-way or two-way ANOVA, followed by Dunnett test or Tukey multiple comparisons as appropriate. P values <0.05 were considered to be significant. P values were reported in Supplementary Table 1 RESULTS Design, Generation, and Characterization of Vg2 x PD-L1 We initially designed and constructed four recombinant antibodies, i.e. Vg2 x PD-L1, Vg2 x Null, PD-L1 mAb and Vg2 mAb to test their activities. The structural properties of these generated antibodies were summarized in Figure 1A. Firstly, the molecular weights of these recombinant proteins were confirmed through SDS-PAGE under both reducing and non-reducing conditions (Supplementary Figure 1A). Then, the SEC results indicated that the purities of the prepared antibodies were more than 95% (Supplementary Figure 1B). Next, we used three PD-L1 expression cell lines (CHO-PD-L1, SKOV3, and H1975) to compare antibody binding ability to the cells between Vg2 x PD-L1 and PD-L1 mAb. The mean EC 50 values for Vg2 x PD-L1 binding to CHO-PD-L1, SKOV3, and H1975 were 1.444 nM, 0.594 nM, and 1.687 nM, respectively ( Figure 1B, Figure 2). Both Vg2 x PD-L1 bsAb and PD-L1 mAb had a similar affinity to the cellular surface PD-L1 ( Figure 1B), due to these two antibodies having the same variable regions for PD-L1 binding (23). Furthermore, we determined the PD-L1 expression scores for a series of target tumor cells using Vg2 x PD-L1 bsAb, which confirmed that Vg2 x PD-L1 exhibited potent affinity toward tumor cells with variable PD-L1 expression levels (Supplementary Figure 2). In addition, the binding affinity to the expanded Vg2Vd2 T cells of Vg2 x PD-L1 was about 60-folds weaker than that of the parental Vg2 mAb, as the mean EC 50 values for Vg2 x PD-L1 and Vg2 mAb were 12.39 nM and 0.21 nM, respectively ( Figures 1C, 2). Moreover, Vg2 x PD-L1 retained the blocking ability as PD-L1 mAb, which was demonstrated in the PD1/PD-L1 cell-based reporter assay ( Figure 1D). In summary, Vg2 x PD-L1 bound with nanomolar affinity to the sorted and expanded Vg2Vd2 T cells and PD-L1 expressing tumor cells.

Vg2 x PD-L1 Efficiently Bridges Vg2Vd2 T Cells to PD-L1 Positive Tumor Cells
Subsequently, we checked whether the Vg2 x PD-L1 prompted the formation of the biphasic cell-to-cell conjugates between Vg2Vd2 T cells and PD-L1 expressing tumor cells. For this purpose, Vg2Vd2 T cells stained with CFSE were co-cultured with PKH26-labelled SKOV3 cells for 30 minutes at 37°C with Vg2 x PD-L1 or Vg2 x Null, then the percentages of doublepositive cells among total cells were measured to represent the bridging ability. In the presence of Vg2 x Null at 1 mg/mL, the double-positive cell population (Q2) was 2.21%, while this population was increased up to 20.1% by Vg2 x PD-L1 (Supplementary Figure 3). In contrast, Vg2 x PD-L1 failed to prompt the co-binding of Vg2Vd2 T cells and HEK-293 cells. (Supplementary Figure 3) Vg2 x PD-L1 Selectively Activates Vg2Vd2 T Cells Exposed to PD-L1 Expressing Tumor Cell Lines Next, we investigated whether the activation of Vg2Vd2 T cells mediated by Vg2 x PD-L1 was dependent on the presence of PD-L1 + tumor cells. Vg2Vd2 T cells were co-cultured with H1975 and SKOV3 cells, the two cell lines that expressed high levels of PD-L1 (Supplementary Figure 2A). Vg2Vd2 T cells secreted little amount of IFNg and did not exhibit activation phenotype (measured by CD25 + CD69 + ) in response to the bsAbs treatment alone ( Figures 3A, B, Figure 4). Of note, in the presence of H1975 and SKOV3 cells, Vg2 x PD-L1, but not Vg2 x Null, triggered significantly the release of IFNg and active phenotype of Vg2Vd2 T cells (Figures 3A, B, 4). Accordingly, Vg2 x PD-L1 further enhanced significantly both the IFNg and TNFa productions and degranulation levels of Vg2Vd2 T cells only in the presence of PD-L1 positive SKOV3 and H1975 cells ( Figures 3C-F, 4). Moreover, these Vg2Vd2 T cells activated jointly by Vg2 x PD-L1 and PD-L1 tumor cells displayed multifunctional effector phenotypes, which co-expressed IFNg, TNFa, and CD107a ( Figures 3D, F, Figure 4). In contrast, Vg2 x Null did not exert agonistic effects on Vg2Vd2 T cells even when co-cultured with PD-L1 expressing target cells in the above conditions ( Figures 3C-F, 4). Together, these data demonstrated that Vg2 x PD-L1 revoked robust effector functions of Vg2Vd2 T cells, including activation, degranulation, and cytokines secretion, in dependent on the engagement of target tumor cells. Vg2Vd2 T cells were co-cultured with SKOV3, H2228, and H1299 cell lines in E:T ratios ranging from 5:1 to 0.3125:1 for 12 hours. We selected SKOV3, H2228 for this test as these two cell lines expressed PD-L1 at high or low levels as determined using Vg2 x PD-L1 staining (Supplementary Figure 2B). Vg2Vd2 T cells alone showed E: T ratio-dependent cytotoxicity for SKOV3 and H2228 ( Figures 5A, B). The addition of Vg2 x PD-L1, but not Vg2 x Null, significantly enhanced tumor cell death even at the lowest E: T ratio (0.3125:1) for the both cell lines (Figures 5A, B). Furthermore, the larger amount of IFNg was only detected in the Vg2 x PD-L1 treated cultures, demonstrating that Vg2 x PD-L1 elicited PD-L1specific IFNg production from Vg2Vd2 T cells (Figures 5C, D). We then evaluated whether Vg2 x PD-L1 could enhance cytotoxicity towards tumor cells that were resistant and refractory to Vg2Vd2 T cells' killing. Indeed, Vg2Vd2 T cell alone lysed less than 20% of H1299 cells even at a 5:1 ratio ( Figure 5E). However, Vg2 x PD-L1 strongly increased the lysis of H1299 with the increased IFNg production by Vg2Vd2 T cells ( Figures 5E, F). Importantly, Vg2 x PD-L1 induced efficient tumor cell lysis, and IFNg secretion was observed at an E: T ratio as low as 0.3125:1 for these three cell lines ( Figure 5).

Vg2 x PD-L1 Potency in Killing PD-L1 Positive Tumor Cell Lines Is Mediated by Both Fresh and Expanded Vg2Vd2 T Cell
To confirm whether Vg2 x PD-L1 could redirect Vg2Vd2 T cells to kill a broad spectrum of tumor cells, we took 5 different human solid tumor cell lines expressing PD-L1 for the test. For these PD-L1 expressing tumor cells, Vg2Vd2 T cells alone did not exert an appreciable killing effect, nor did the PD-L1 mAb ( Figure 6A). However, a dose-dependent effective killing mediated by Vg2Vd2 T cells was observed with the addition of Vg2 x PD-L1 irrespective of tumor cells' origin, but not for Vg2 x Null ( Figure 6A). As expected, Vg2Vd2 T cells exhibited a dosedependent IFNg secretion treated with Vg2 x PD-L1, compared with no such effect with control Abs ( Figure 6B). We further observed that the Vg2 x PD-L1-induced Vg2Vd2 T cells' cytotoxicity (killing EC 50 ) towards tumor cells was correlated significantly with these tumor cells' PD-L1 expression scores, while the release IFNg EC 50 showed a negative trend with the PD-L1 expression scores ( Figure 6C). Moreover, the viability of PD-L1 neg HEK-293 cells remained unaffected in all tested concentrations in the presence of Vg2 x PD-L1 ( Figure 6D). In addition, allogeneic PBMCs were used as target cells to check if the killing activity of Vg2 x PD-L1 was specific to tumor cells. The Vg2Vd2 T cell-mediated killing percentages of allogeneic PBMCs were low even in the presence of Vg2 x PD-L1, indicating the Vg2 x PD-L1 activated Vg2Vd2 T cells' killing activity was indeed restricted to tumor cells ( Figure 6E). Moreover, fresh Vg2Vd2 T cells enriched from healthy donors also exerted concentration-dependent killing of SKOV3 cells mediated by Vg2 x PD-L1, but not by Vg2 x Null or PD-L1 mAb ( Figure 6F).
Taken together, these results demonstrated that Vg2 x PD-L1 could redirect Vg2Vd2 T cells to kill PD-L1+ tumor cell lines with IFNg secretion, but to leave PD-L1 negative tumor cells and healthy cells un-attacked.

Vg2 x PD-L1 Enhances the Efficacy of Adoptively Transferred Vg2Vd2 T Cells In Vivo
We further studied the effect of Vg2 x PD-L1 on the outgrowth of established PD-L1 expressing tumors. SKOV3 cells were injected into nude mice, and the tumor cells were allowed to grow out and engraft for one week before the mice received twice-weekly i.v. injections with human Vg2Vd2 T cells, followed by twice-weekly i.p. injections with either 8 mg/kg Vg2 x PD-L1 or Vg2 x Null, or PBS. The mice were sacrificed at the time of severe disease symptoms ( Figure 7A). The Vg2Vd2 T cells alone, or Vg2Vd2 T cells plus Vg2 x Null did not control the tumor growth ( Figures 7B, C). In contrast, the combo treatment with Vg2 x PD-L1 and Vg2Vd2 T cells significantly delayed the tumor growth, with lower tumor weights at the end of the study (Figures 7B-D) than those of the control groups. After 16 days of treatment, Vg2Vd2 T cell counts were significantly higher in the Vg2 x PD-L1+Vg2Vd2 T cells group, compared with the Vg2 x Null+ Vg2Vd2 T cells group or Vg2Vd2 T cells group (Figures 7E, F).

DISCUSSION
The clinical investigations of PD-1/PD-L1 inhibitors have resulted in a paradigm shift in the treatment of advanced cancer patients, as well as longer overall survival time (30). However, due to the limited efficacy (only 20 to 30% objected response) and resistance to PD-1/PD-L1, there is still an unmet medical need for exploring novel agents to improve PD-L1 targeting therapeutic effectiveness (31). The inadequate infiltration of T lymphocytes into the cold tumor is one of the reasons for this therapeutic resistance (32). Several clinical studies showed that transferred Vg2Vd2 T cells migrated into the tumor bed, leading to encouraging clinical responses and tumor reduction in treated patients (33). Here, the bispecific antibody and Vg2Vd2 T cells transfer combination approach provided a potential strategy to circumvent the PD-L1 blockade therapy limitations. The approach for targeting potent cytotoxicity Vg2Vd2 T cells by constructing Vg2 x PD-L1 on the Y-body platform, based-on which two novel candidate medications are currently on clinical trials, noted as M701 (NCT04501744) and M802 (NCT04501770) (34). Vg2 x PD-L1 preserved high affinity to PD-L1 as well as the PD1/PD-L1 blocking activity. However, consistent with other reports, the PD-1/PD-L1 blocking activity did not contribute to the killing ability of Vg2Vd2 T cells (12), possibly because the PD-L1 mAb used in our study contained silent Fc without ADCC capability. Vg2 x PD-L1 had a slower affinity for the Vg2 TCR than Vg2 mAb, which was desired for clinical use to prevent cytokine release storm (35). Additionally, Vg2 TCR-targeting Y-body platform allowed for the simple replacement of the PD-L1 Fab to create a sequence of Vg2 x TAAs, which enabled Vg2Vd2 T cells to target a broader spectrum of tumor types and helping a larger population of cancer patients.
In vitro, Vg2 x PD-L1-activated Vg2Vd2 T cells were able to selectively kill tumor cells selectively without killing PD-L1 negative non-malignant cells or normal cells. In fact, the activation, degranulation, and subsequent tumor cell killing mediated by Vg2 x PD-L1 were all dependent on simultaneous binding to Vg2Vd2 T cell and PD-L1 expressing tumor cells, demonstrating the safety of our strategy in comparison to PD-L1 chimeric antigen receptor NK cells (36). In line with these in vitro observations, Vg2 x PD-L1 was found to improve Vg2Vd2 T cell mediated tumor growth inhibition in vivo. Mechanically, Vg2 x PD-L1 generated a greater Vg2Vd2 T cell infiltration.
Meanwhile, there are several limitations in this study. First, because Vg2Vd2 T cells are species specific, we employed an immunodeficiency mouse model to investigate the efficacy of Vg2 x PD-L1 plus Vg2Vd2 T cells, without examining whether this combination therapy could change or reshape the suppressive tumor microenvironment, or the in vivo toxicity of combo usage. Second, this combo treatment was not fully curative because tumor volumes did not reach to near zero by the end of treatment. As a small amount of Vg2Vd2 T cells and a fixed bsAb dose were used in the current treatment protocol, we intended to improve the present therapy approach involving a modest number of Vg2Vd2 T cells and bsAb dosage. Third, we were unable to determine the TCR sequence of tumor bed infiltrating Vg2Vd2 T cells, which would provide valuable information for further TCR-T design.
In conclusion, we developed a novel and potential therapeutic T cell engager bispecific antibody Vg2 x PD-L1, which caused Vg2Vd2 T cells to destroy PD-L1 expressing tumor cells efficiently and selectively. Vg2 x PD-L1 offers promising therapy options for solid tumors, including ovarian cancer (28,37,38), melanoma (38,39), and non-small cell lung cancer (NSCLC) (38). The infiltrating Vg2Vd2 T cells in tumor acted as protective anti-tumor effector population and were linked with positive outcomes. As PD-L1 is a clinically well-established tumor target, its widespread expression pattern suggested that our combination approach might be beneficial for the PD-L1 positive cancer patients who had refractory or relapsed for PD-L1 inhibitor treatment.

DATA AVAILABILITY STATEMENT
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

ETHICS STATEMENT
The studies involving human participants were reviewed and approved by the institutional review boards for human subjects' research and institutional biosafety committees at Hubei Province Food and Drug Safety Evaluation Center (Wuhan, China