SW480 tumor cells (EGFR⁺) were obtained from American Type Culture Collection (ATCC, CCL-228) and maintained in Dulbecco’s Modified Eagle Medium (DMEM, 1965-039, Gibco) supplemented with 10% (v/v) fetal calf serum (FCS, 04-007-1A, Biological Industries), 0.05 mM β‐mercaptoethanol (β‐ME, 200-646-6, Merck), and 100 IU/ml sodium penicillin, 100 μg/ml streptomycin sulphate, and 2.0 mM L‐glutamine (PSG, 10378-016, Life technologies). 22Rv1 tumor cells (PSMA⁺) were obtained from the European Collection of Authenticated Cell Cultures (ECACC, 5092802) and MM.1s tumor cells (kind gift from R. Groen, Amsterdam UMC, Vrije Universiteit, Amsterdam, the Netherlands) were transfected with CD1d as described previously (40) (referred to as MM.1s.CD1d). The 22Rv1 and MM.1s tumor cell lines were maintained in Roswell Park Memorial Institute 1640 (RPMI-1640, 22400089, Gibco) medium supplemented with 10% FCS, β-ME, and PSG. Tumor cell lines were tested regularly for Mycoplasma using PCR.

Healthy donor- and cancer patient peripheral blood mononuclear cells (PBMC) were isolated from whole blood by density gradient centrifugation using Lymphoprep™ (AXI-1114547, Fresenius). Healthy donor blood samples were obtained under written informed consent from Sanquin (Amsterdam, the Netherlands). Blood samples from cancer patients were obtained after approval by the institutional review board (medical ethical committee Amsterdam UMC, location VUmc) and written informed consent was obtained from all the participants from the Amsterdam UMC (location VUmc, Amsterdam, the Netherlands). Expanded Vγ9Vδ2 T-cells were generated from healthy donor PBMC-derived Vγ9Vδ2 T-cells as described before (33). In short, Vδ2⁺ T cells were isolated from healthy donor PBMC using magnetic bead sorting using FITC-conjugated Vδ2 antibody ( Supplementary Table S1 ) in combination with anti-mouse IgG microbeads (130-048-401, Miltenyi Biotec). Purified Vγ9Vδ2 T-cells were stimulated weekly with irradiated feeder mix consisting of healthy donor PBMC (1 x 10⁶ cells/ml), JY cells (1 x 10⁵ cells/mL, 94022533, ECACC), IL-7 (10 U/mL, 207-IL-025, R&D Systems), IL-15 (10 ng/mL, 34-8159-85, eBioscience), and PHA (50 ng/mL, R30852801, Thermo Fisher Scientific). Purity of Vγ9Vδ2 T-cells used in experiments was ≥95%. All functional experiments were performed in RPMI medium supplemented with 10% FCS, β-ME, and PSG. Cell cultures were kept at 37˚C in a humidified atmosphere containing 5% CO₂.

Human Vδ2-TCR specific llama-derived VHHs were generated and screened as previously described (41). In short, two Lama glamas were immunized with human Vγ9Vδ2 T-cells and Vδ2-TCR-specific VHHs were selected using phage display followed by confirmation of specific binding to Vδ2⁺ T cells by flow cytometry. Bivalent constructs were generated by linking two monovalent Vδ2-TCR-specific VHHs with low (VHH-5C7, Kd ~350nM), intermediate (VHH-5D3, Kd ~21nM) or high (VHH-6H4, Kd ~0.4nM) affinity in various combinations using a Gly4Ser amino acid based linker [(G4S)₂; referred to as 10 amino acid (AA) linker], resulting in the following bivalent constructs: 6H4-(G4S)₂-5C7, 5C7-(G4S)₂-6H4, 6H4-(G4S)₂-6H4, 5D3-(G4S)₂-6H4, 6H4-(G4S)₂-5D3, 5D3-(G4S)₂-5D3, 5D3-(G4S)₂-5C7, 5C7-(G4S)₂-5D3, 5C7-(G4S)₂-5C7. The bivalent construct 6H4-5C7 was generated with multiple linker lengths; G4S, (G4S)₂ or (G4S)₄ (referred to as 5, 10 or 20 AA linker). Purified protein was produced by ImmunoPrecise Antibodies (IPA, Utrecht, The Netherlands) using DNA transfected HEK293E cells and rmp Protein A affinity chromatography followed by preparative size exclusion. Proteins used were >95% pure and monomeric.

Three different tumor associated antigen (TAA) specific VHHs [i.e. directed against EGFR (VHH-7D12 (42)], PSMA [VHH-JVZ-007 (43)] or CD1d [VHH-1D12 (44)] were linked with a G4S linker to the N-terminus of 6H4-5C7 (termed Vδ2^hi-lo): EGFR-G4S-6H4-(G4S)₂-5C7, PSMA-G4S-6H4-(G4S)₂-5C7 and CD1d-G4S-6H4-(G4S)₂-5C7 (referred to as EGFR-Vδ2^hi-lo bsVHH, PSMA-Vδ2^hi-lo bsVHH and CD1d-Vδ2^hi-lo bsVHH). Next, Fc domain and anti-albumin VHH containing molecules were created to allow for in vivo plasma half-life extension. For the Fc domain, knobs-into-holes technology was used for heterodimerization (HC1, knob mutation T366W and HC2, hole mutations T366S, L368A, Y407V), LFLE mutations were included to silence the Fc domain except for the neonatal Fc receptor (FcRn) (L234F and L235E silencing, EU numbering), a Cys220 deletion was incorporated to avoid an unpaired cysteine and a modified hinge was used (HC1: 5C7-G4S-6H4-hinge-Fc paired with HC2: EGFR-hinge-Fc, PSMA-hinge-Fc or CD1d-hinge-Fc; referred to as EGFR-Vδ2^hi-lo bsVHH-Fc, PSMA-Vδ2^hi-lo bsVHH-Fc and CD1d-Vδ2^hi-lo bsVHH-Fc. Expression of these tumor targeting VHHs was performed via pcDNA3.1(+) (100µg, Genscript) transfection in Expi293F cells (A14527, Gibco) using the ExpiFectamine 293 Transfection Kit (A14524, Gibco). Proteins were purified using rmp Protein A Sepharose Fast Flow resin (17-5138-03, GE healthcare) (EGFR-Vδ2^hi-lo bsVHH, PSMA-Vδ2^hi-lo bsVHH and CD1d-Vδ2^hi-lo bsVHH) or CaptureSelect C-tagXL Affinity Matrix (2943072050, Thermo Scientific) (EGFR-Vδ2^hi-lo bsVHH-Fc, PSMA-Vδ2^hi-lo bsVHH-Fc or CD1d-Vδ2^hi-lo bsVHH-Fc). Quality control was done using size exclusion ultra performance liquid chromatography. For fusion to an anti-albumin (Alb) VHH (45), the following construct was generated: EGFR-(G4S)₂-Alb-(G4S)₂-6H4-G4S-5C7 (referred to as EGFR-Vδ2^hi-lo bsVHH-albumin; illustrated in Supplementary Figure S4A ). Purified protein for the construct was produced by IPA as described above. Proteins used were >95% pure and >94% heterodimer shown by native mass spectrometry analysis. Lastly, bispecific engagers containing a monovalent Vδ2-TCR-specific VHH linked with a TAA-specific VHH, i.e. EGFR-G4S-5C8, PSMA-G4S-5C8 and CD1d-G4S-5C8 (referred to as EGFR-Vδ2 bsVHH, PSMA-Vδ2 bsVHH and CD1d-Vδ2 bsVHH), were generated as described before (31, 32) and are illustrated in Supplementary Figure S3A .

Cells were resuspended in PBS (1073508600, Fresenius Kabi) supplemented with 0.5% bovine serum albumin (M090001/03, Fisher Scientific) and 20 μg/ml NaN3 (247-852-1, Merck) and incubated with fluorochrome-labeled monoclonal antibodies and viability dyes ( Supplementary Table S1 ) for 30 min at 4°C. Unbound fluorochrome-labeled antibodies were washed away. Cells were analyzed using the FACS LSRFortessa XL-20 (BD Biosciences) and data analysis was performed using FlowJo v10.8.1 (BD Biosciences).

Binding of the various bivalent and bispecific VHH constructs to Vδ2, EGFR, PSMA or CD1d was evaluated using flow cytometry or ELISA, as indicated. For flow cytometry analysis, Vγ9Vδ2 T-cells, SW480, 22Rv1 or MM.1s.CD1d tumor cells were incubated with a concentration range of the constructs for 45 min at 4°C, followed by extensive washing (5x) and incubation for 30 min at 4°C with FITC-labeled rabbit-anti-llama polyclonal Ab ( Supplementary Table S1 ). For the ELISA, wells of a clear bottom F96-well Maxisorp plate (439454, Nunc) were coated overnight with human Gamma9Delta2 TCR (0.5 µg/mL, produced by IPA). Wells were blocked with 2% BSA (A2153, Sigma) for 60 min, and test samples were incubated for 120 min at room temperature. The rabbit anti-camelid VHH cocktail (60-min incubation, 19L002038, GenScript) and swine anti-rabbit immunoglobulins HRP (60-min incubation, 41289300, Dako) were used for detection and 3,3’,5,5’-Tetramethylbenzidine (TMB, Life Technologies, #SB02) was used as a substrate for color development. The absorbance (450 nm) was measured using a plate reader (Molecular Devices, iD5) and the data analysis was done using SoftMax Pro software v 7.1.

Vγ9Vδ2 T-cell percentages in healthy donor and cancer patient-derived PBMC (patient characteristics provided in Supplementary Table S2 ) were assessed using flow cytometry using fluorescently-labelled antibodies against CD3, Vγ9, Vδ2 ( Supplementary Table S1 ). The capacity of the various bivalent and bispecific VHH constructs to induce Vγ9Vδ2 T-cell expansion was assessed by incubating healthy donor or cancer patient-derived PBMC (patient characteristics provided in Supplementary Table S3 ) with indicated concentrations of the bivalent or bispecific constructs, 10 µM pamidronate disodium salt hydrate (referred to as pamidronate, P2371-10MG, Sigma) (positive control) or medium (negative control). Twenty-four hrs later, 100 IU/ml IL-2 (Proleukin, Clinigen) was added to all conditions (including medium, termed IL-2 control). Fold expansion and enrichment of Vγ9Vδ2 T-cells (compared to baseline) were assessed using flow cytometry at the end of the 8-day culture period using fluorescently-labelled antibodies against CD3, Vγ9, Vδ2 and 123counting eBeads™ (01-1234-42, Thermofisher). Depending on the experiment, 7-AAD (A9400-1MG, Sigma), fixable viability dyes eFluor™ 506 or eFluor™ 780 ( Supplementary Table S1 ) were used to identify viable cells. The phenotype of expanded Vγ9Vδ2 T-cells was determined using fluorescently-labelled antibodies against CD27, CD45RA, CD25, CD69, HLA-DR, DNAM-1, NKG2D, NKG2A, PD-1, CTLA-4 and TIGIT (see Supplementary Table S1 for clones and fluorophores).

Vγ9Vδ2 T-cell degranulation and subsequent tumor cell lysis were assessed by flow cytometry. Expanded healthy donor-derived Vγ9Vδ2 T-cells were incubated with SW480, 22Rv1 or MM.1s.CD1d tumor cells (1:1 E:T ratio) for 24hrs with a concentration range of the TAA-Vδ2^hi-lo bsVHH with or without Fc domain or anti-albumin-VHH, or medium control. Degranulation and activation of Vγ9Vδ2 T-cells was assessed using fluorescently-labelled antibodies against CD3, Vγ9, Vδ2, CD25 and CD107a ( Supplementary Table S1 ) and tumor cell lysis was determined using 7-AAD and 123counting eBeads^TM. To assess degranulation and tumor cell lysis of 8-day expanded Vγ9Vδ2 T-cells in healthy donor derived PBMC, γδ T-cells were isolated from the PBMC cultures using the untouched human TCR γ/δ⁺ T-cell isolation kit (130-092-892, Miltenyi Biotec) and used as described above in the presence or absence of the respective TAA-Vδ2^hi-lo bsVHH or TAA-Vδ2^hi-lo bsVHH-Fc engagers. Supernatants were collected and stored at -20°C until further analysis for IL-2, TNF and IFN-γ secretion using the human Th1/Th2/Th17 cytometric bead array (CBA) kit (560484, BD). CBA data were analyzed with FCAP Array software v3.0 (BD).

Sub-lethally irradiated (1.75 Gy on day 0) NOD.Cg-Prkdcscid Il2rgtm1Sug/JicTac (NOG)-hIL-15 mice (8.5 weeks old, #13683-F Taconic Biosciences) were kept under pathogen-free conditions (Laboratory Animal Center of the Netherlands Cancer Institute, Amsterdam, The Netherlands) and used for evaluation of in vivo expansion of Vγ9Vδ2 T-cells in PBMC. This in vivo experiment was approved by the Animal Welfare Committee of the Netherlands Cancer Institute and was performed in accordance with national guidelines. Mice were randomized into 2 groups (n=4/group) and intravenously (i.v.) inoculated with 10x10⁶ healthy human donor-derived PBMC (n=1 healthy donor with 0.89% Vγ9Vδ2 T-cells of total CD3⁺ T-cells) on day 0. PBS or the CD1d-Vδ2^hi-lo bsVHH (0.5 mg/kg) were intraperitoneally (i.p.) administered on days 0 and 4. Blood (day 1 and 8) and lungs, spleens and livers (day 8) were collected and prepared for flow cytometry analysis. Erythrocytes were lysed using lysis buffer (0.16 M NH₄Cl; 76050139.1000, Boom and 0.011 M KHCO₃; 104854, Merck). Lungs were dissociated while stirring in a flask for 45 min at 37°C using Iscove’s Modified Dulbecco’s Medium (IMDM) supplemented with 5% FCS, PSG, 0.1% DNAse I (10104159001, Roche) and 0.14% collagenase A (10103586001, Roche). Spleens and livers were dissociated using a 100 µM cell strainer (352360, Falcon) and a syringe plunger, and dissociated cells were washed with PBS. Blood and tissue-derived single cell suspensions were used for flow cytometry analysis using a combination of 7-AAD, fluorescently-labelled antibodies against CD45 (mouse), CD45 (human), CD3, Vγ9, Vδ2, CD27, CD45RA and 123counting eBeads™ for Vγ9Vδ2 T-cell analysis ( Supplementary Table S1 ).

Analysis were performed using Prism v.9.1.0 (GraphPad Software). Data were analyzed using unpaired or paired t test, one-way analysis of variance (ANOVA) with Tukey’s comparisons test, or two-way ANOVA with Tukey’s multiple comparisons test, as appropriate. Binding curves and dose–response curves with EC₅₀ values were calculated using nonlinear regression analysis. To calculate the correlation between Vγ9Vδ2 T-cell frequency and donor age, Spearman’s rank correlation analysis was used. P=< 0.05 was considered significant and showed with asterisks: P=< 0.05: *, P=< 0.01: **, P=< 0.001: ***, P=< 0.0001: ****.

Given that the Vγ9Vδ2 T-cell frequency in peripheral blood has been reported to decline with age and can be reduced due to underlying malignancies (9–14), we assessed the frequency of Vγ9Vδ2 T-cells in peripheral blood of healthy donors [n=121; 68 females, 53 males, mean age (range): 41 (18-77)] and cancer patients with various solid tumors (n=91; see Supplementary Table S2 for patient characteristics). As shown in Supplementary Figure S1A , the Vγ9Vδ2 T-cell frequency varied substantially between individuals and was significantly higher in healthy donor compared to cancer patient PBMC (3.5 ± 0.3% versus 2.2 ± 0.4% of total T cells; mean ± SEM, P = 0.01). In healthy female and male donors, the Vγ9Vδ2 T-cell frequency negatively correlated with donor age ( Supplementary Figure S1B ). This negative correlation between Vγ9Vδ2 T-cell frequency and age was not (statistically significantly) observed in cancer patients, though the population of adult patients with solid tumors evaluated in our analysis was generally skewed towards older age (66 ± 0.9 yr versus 40 ± 1.5 yr in healthy donors; mean ± SEM; Supplementary Figure S1C ). Of interest, the frequency of Vγ9Vδ2 T-cells was comparable in cancer patients and healthy donors with age ≥ 39 yr (Vγ9Vδ2 T-cell frequency 2.0 ± 1.9% versus 2.3 ± 3.3% of total T cells in healthy donors ≥ 39 yr and cancer patients ≥ 39 yr respectively; mean ± SEM; Supplementary Figure S1D ), suggesting that the Vγ9Vδ2 T-cell frequencies in peripheral blood of cancer patients are predominantly influenced by age, rather than by the underlying malignancy.

As TCR cross-linking, which is not achieved with monovalent TCR binding, initiates signaling typically resulting in activation, differentiation, cytokine production and proliferation, we evaluated whether linkage of two Vδ2-TCR specific VHHs could promote Vγ9Vδ2 T-cell expansion. Three Vδ2-TCR specific VHHs were selected based on affinity (i.e. VHH-5C7, Kd ~350 nM; VHH-5D3, Kd ~21 nM; and VHH-6H4, Kd ~0.4 nM) (ref (41) and not shown), and differentially combined and linked using a 10 AA linker into nine different bivalent VHHs. Binding of the various bivalent Vδ2-VHHs to Vγ9Vδ2 T-cells was assessed using flow cytometry or ELISA and showed that the binding EC₅₀ was comparable for compounds that contained a high affinity VHH linked to either a low or intermediate affinity VHH or two intermediate-binding VHHs (EC₅₀ = 0.05-0.43 nM) ( Figure 1A ). For the high and low affinity VHH combination, N-terminal positioning of the high affinity VHH-6H4 resulted in a lower EC₅₀ than when this VHH was positioned C-terminally. Compounds consisting of an intermediate and a low affinity VHH bound with higher EC₅₀ values of 17-22 nM. The binding EC₅₀ of the linked low affinity VHHs could not be reliably assessed using flow cytometry and was therefore determined using ELISA, resulting in a value of 211 nM ( Figure 1A right panel). To assess whether the bivalent Vδ2-VHHs could support Vγ9Vδ2 T-cell expansion, healthy donor derived PBMC were cultured for 8 days in the presence or absence of 1 nM (optimal concentration across the compounds as determined by titration, data not shown) of the bivalent Vδ2-VHHs or 10 µM pamidronate. After 24 hrs, 100 IU/ml of recombinant human (rh) IL-2 was added to the cultures (including the medium control, termed IL-2 control). Figure 1B illustrates the design of the experiment and Figure 1C shows a representative gating strategy for the assessment of the Vγ9Vδ2 T-cell frequency. As shown in Figure 1D , maintained or increased Vγ9Vδ2 T-cell numbers were only observed in the conditions exposed to bivalent VHHs containing either 2 linked low affinity VHHs or the low affinity VHH linked to the high affinity VHH. When enrichment of Vγ9Vδ2 T-cells was assessed, a similar pattern was observed. Enrichment and expansion were most consistent and pronounced when PBMC were cultured with the bivalent high-low affinity 6H4-5C7 (i.e. N-terminally positioned 6H4) VHH (enrichment 35.3 ± 49.4%; median ± IQR; P = 0.03, and expansion 5.6 ± 12.9 fold; median ± IQR; P = 0.07, paired t tests). Most of the other bivalent Vδ2-VHHs did not support enrichment and expansion and in some cases even resulted in a reduction of Vγ9Vδ2 T-cell numbers, possibly due to fratricide. As expected, enrichment and expansion were also observed when PBMC were cultured in the presence of pamidronate. In the absence of IL-2, neither pamidronate nor bivalent Vδ2-VHHs triggered Vγ9Vδ2 T-cell expansion ( Supplementary Figure S2 ). To assess the impact of different linker lengths, bivalent VHHs with the high affinity Vδ2-TCR VHH-6H4 and the low affinity Vδ2-TCR VHH-5C7 were generated with either a G4S, (G4S)₂ or (G4S)₄ linker (referred to as 5, 10 or 20 AA linker). As shown in Figures 1E, F , these bivalent (6H4-5C7) Vδ2-VHHs showed a similar binding profile (i.e. similar EC₅₀ and maximum binding) and similar enrichment and expansion of Vγ9Vδ2 T-cells in the 8 day PBMC cultures. Enrichment and expansion of Vγ9Vδ2 T-cells was already observed using 0.1 nM of the bivalent (6H4-5C7) Vδ2-VHHs and not further enhanced with higher concentrations. The bivalent Vδ2-VHH with the high affinity VHH-6H4 (N-terminal) a 10 AA linker and the low affinity VHH-5C7 was selected for further experiments and termed bivalent Vδ2^hi-lo VHH.

To explore whether a bispecific engager that includes the bivalent Vδ2^hi-lo VHH as well as a VHH specific for a tumor associated antigen (TAA) could combine TAA directed Vγ9Vδ2 T-cell effector functions and Vγ9Vδ2 T-cell expansion, six bispecific engagers were generated (see Figure 2A for design). The engagers included the bivalent Vδ2^hi-lo VHH either directly linked to a VHH directed against EGFR [clone 7D12 (42)], PSMA [clone JVZ-007 (43)] or CD1d [clone 1D12 (44)] at the N-terminus or (indirectly) linked to these VHHs using Fc domains that heterodimerize using knobs-into-holes technology (see Materials and Methods). The latter formats would allow for half-life extension of the engagers when applied in vivo.

All six constructs bound with similar EC₅₀ values of 0.2-0.5 nM to Vγ9Vδ2 T-cells ( Figure 2B ) and supported significant Vγ9Vδ2 T-cell enrichment (median 20.6-58.6% of total T cells, versus 2.9% in IL-2 control) and expansion (median 12.2-32.4 fold expansion, versus 1.4 fold in IL-2 control) when added at a concentration of 1 nM (TAA-Vδ2^hi-lo bsVHH) or 100 nM (TAA-Vδ2^hi-lo bsVHH-Fc) to healthy donor derived PBMC for 8 days (optimal concentrations determined by titration, not shown), illustrating that the ability to induce Vγ9Vδ2 T-cell expansion was retained in these molecular formats ( Figure 2C ). Enrichment and expansion of Vγ9Vδ2 T-cells was not observed when a monovalent Vδ2-VHH was linked to the EGFR or PSMA-specific VHHs, underscoring the importance of including the bivalent Vδ2^hi-lo VHH when expansion is desired ( Supplementary Figures S3A, B ). Of note, enrichment and expansion of Vγ9Vδ2 T-cells was observed when PBMC were cultured with CD1d specific engagers containing either the monovalent Vδ2-VHH or the bivalent Vδ2^hi-lo VHH, which was likely related to CD1d expressing cells in PBMC (e.g. monocytes, B cells, circulating dendritic cells) that allowed Vδ2-TCR crosslinking to support Vγ9Vδ2 T-cell expansion (32).

After binding of these TAA-Vδ2^hi-lo bsVHH and TAA-Vδ2^hi-lo bsVHH-Fc molecules to the TAA expressing tumor cells was confirmed ( Figure 2D ), the antitumor effector functions of these molecules were evaluated in co-cultures of Vγ9Vδ2 T-cells and tumor cells expressing the respective TAAs (i.e. SW480 (EGFR⁺), 22Rv1 (PSMA⁺) or MM.1s.CD1d (CD1d⁺)). A representative gating strategy is shown in Supplementary Figure S5A . As shown in Figures 2E, F , all constructs were able to induce Vγ9Vδ2 T-cell degranulation and subsequent tumor cell lysis with EC_50s in the low-picomolar range. Addition of the Fc domain resulted in a 3-9 fold reduced potency for tumor lysis, possibly due to steric hindrance or an increased intermembrane distance between the Vγ9Vδ2 T-cell and TAA⁺ tumor target cell which may interfere with efficient immune synapse formation (46).

Incorporation of an anti-albumin binding domain can also provide a means to extend in vivo half-life (45). To explore this, we added an anti-albumin VHH to the EGFR-Vδ2^hi-lo bsVHH (see Supplementary Figure S4A for design) and confirmed binding to both Vγ9Vδ2 T-cells and EGFR⁺ tumor cells ( Supplementary Figure S4B ). The EGFR-Vδ2^hi-lo bsVHH-albumin retained the ability to support Vγ9Vδ2 T-cell enrichment and expansion in healthy donor PBMC ( Supplementary Figure S4C ) and also triggered Vγ9Vδ2 T-cell degranulation and lysis of SW480 tumor cells ( Supplementary Figure S4D ). The potency (i.e. EC₅₀ for degranulation and tumor lysis) of the EGFR-Vδ2^hi-lo bsVHH-albumin construct was ~10 fold lower compared to the EGFR-Vδ2^hi-lo bsVHH-Fc perhaps as a result of impaired immune synapse formation when albumin was bound to the engager in this specific orientation.

The functional properties of the expanded Vγ9Vδ2 T-cells were next explored. For this purpose, expanded Vγ9Vδ2 T-cells were enriched from the PBMC cultures using negative MACS isolation and co-cultured with either EGFR⁺ SW480 colorectal cancer cells, PSMA⁺ 22Rv1 prostate cancer cells, or CD1d⁺ MM.1s multiple myeloma cells for an additional 24 hr; the TAA-Vδ2^hi-lo bsVHH/bsVHH-Fc engagers were re-added to the respective conditions. Figure 3A illustrates the experimental design and Supplementary Figures S5B, C show the Vγ9Vδ2 T-cell frequency post expansion and post enrichment and a representative dot-plot thereof. Compared to Vγ9Vδ2 T-cells cultured with IL-2 only (IL-2 control), Vγ9Vδ2 T-cells expanded using pamidronate, the Vδ2^hi-lo bivalent VHH, or the Vδ2^hi-lo bsVHH/bsVHH-Fc typically expressed higher levels of the activation marker CD25 upon co-culture with tumor cells ( Figure 3B ). In the presence of the TAA-Vδ2^hi-lo bsVHH and bsVHH-Fc molecules, the expanded Vγ9Vδ2 T-cells expressed significantly higher levels of the degranulation marker CD107a and, in line, triggered more robust tumor lysis ( Figures 3C, D ; plots from a representative donor are shown in Supplementary Figures S5D, E ). Supernatants obtained from these 24 hr co-cultures demonstrated that the TAA-Vδ2^hi-lo bsVHH molecules promoted the production of higher levels of various proinflammatory cytokines, i.e. IL-2, TNF and IFN-γ ( Figure 3E ). Overall, the increase in proinflammatory mediators and cytolytic effects were most pronounced when expanded Vγ9Vδ2 T-cells were cultured with TAA-Vδ2^hi-lo bsVHH. Taken together, Vγ9Vδ2 T-cells expanded from PBMC with the TAA-Vδ2^hi-lo bsVHH and bsVHH-Fc exhibit an activated effector phenotype and mediate tumor lysis when exposed to TAA expressing tumor cells.

We next evaluated whether the TAA-Vδ2^hi-lo bsVHH and bsVHH-Fc molecules could also support expansion of Vγ9Vδ2 T-cells from cancer patient PBMC. For this purpose, PBMC were obtained from patients with gastric cancer, esophageal cancer or melanoma. The majority of these patients had advanced-stage disease and was treatment naïve (see Supplementary Table S3 for patient characteristics). As shown in Figure 4A , pamidronate, the bivalent Vδ2^hi-lo VHH as well as the TAA-Vδ2^hi-lo bsVHH and TAA-Vδ2^hi-lo bsVHH-Fc constructs supported the enrichment and expansion of Vγ9Vδ2 T-cells in 8 day PBMC cultures. Enrichment (55.9 ± 33.2%; median ± IQR) and expansion (69.8 ± 79.5 fold; median ± IQR) was most pronounced with the CD1d-Vδ2^hi-lo bsVHH and less striking when the Fc domain was connected to this bsVHH. With the EGFR and PSMA specific Vδ2^hi-lo engagers, expansion was similar in the presence and absence of the Fc domain.

The phenotype of Vγ9Vδ2 T-cells expanded from healthy donor and cancer patient PBMC using pamidronate, the Vδ2^hi-lo bivalent VHH, the various TAA-Vδ2^hi-lo bsVHH and TAA-Vδ2^hi-lo bsVHH-Fc molecules and the IL-2 control was characterized (representative gating strategy shown in Supplementary Figure S6 ). At baseline, most Vγ9Vδ2 T-cells in healthy donor PBMC had a central memory (CD27⁺CD45RA^-, 47.1 ± 4.5%; mean ± SEM) phenotype with naïve (CD27⁺CD45RA⁺, 23.6 ± 4.2%; mean ± SEM), effector memory (CD27^- CD45RA^-, 17.1 ± 4.6%; mean ± SEM) and terminally differentiated (CD27^-CD45RA⁺, 12.1 ± 1.9%; mean ± SEM) Vγ9Vδ2 T-cell populations representing smaller fractions ( Figure 4B ). After an 8-day culture of PBMC with 100 IU/ml IL-2, pamidronate or the various TAA-Vδ2^hi-lo bsVHH and bsVHH-Fc engagers, expanded Vγ9Vδ2 T-cells exhibited a shift towards an effector memory phenotype with a concomitant reduction in the frequency of naïve and terminally differentiated subsets while the fraction of central memory cells overall remained similar in size. Of interest, Vγ9Vδ2 T-cells expanded with the TAA-Vδ2^hi-lo bsVHH-Fc molecules retained larger proportions of Vγ9Vδ2 T-cells with a terminally differentiated phenotype (EGFR-Vδ2^hi-lo bsVHH; P = 0.0007, PSMA-Vδ2^hi-lo bsVHH; P = 0.02, CD1d-Vδ2^hi-lo bsVHH; P = 0.01, unpaired t tests) and naïve phenotype (EGFR-Vδ2^hi-lo bsVHH; P = 0.008, PSMA-Vδ2^hi-lo bsVHH; P = 0.0002, CD1d-Vδ2^hi-lo bsVHH; P = 0.002, unpaired t tests). When compared to Vγ9Vδ2 T-cells in healthy donor PBMC, the proportion of central memory Vγ9Vδ2 T-cells was increased in cancer patient PBMC at baseline (72.2 ± 6.8% versus 47.1 ± 4.4% in healthy donors; mean ± SEM, P = 0.001; unpaired t test). This was accompanied by a lower proportion of Vγ9Vδ2 T-cells with a naïve phenotype (9.7 ± 2.6% versus 23.6 ± 4.2% in healthy donors; mean ± SEM, P = 0.02; unpaired t test), which may suggest their priming and involvement in the antitumor immune response ( Figure 4C ). The Vγ9Vδ2 T-cells expanded from cancer patient PBMC using the TAA-Vδ2^hi-lo bsVHH and bsVHH-Fc were in general skewed to an effector memory phenotype. However, while the proportion of naïve and terminally differentiated effector memory cells was strongly reduced upon expansion in Vγ9Vδ2 T-cells expanded from healthy donor PBMC, these were maintained in Vγ9Vδ2 T-cells expanded from cancer patient PBMC ( Figure 4C ).

Healthy donor Vγ9Vδ2 T-cells expanded by the TAA-Vδ2^hi-lo bsVHH and bsVHH-Fc constructs as well as pamidronate were highly activated as shown by strong increases in CD25 and HLA-DR expression ( Figure 4D ). During culture, expression of DNAM-1, an activating receptor expressed by most Vγ9Vδ2 T-cells, remained highly expressed and was even expressed by a larger proportion in several cases. Of interest, NKG2D, an activating receptor also expressed by most Vγ9Vδ2 T-cells, remained highly expressed when Vγ9Vδ2 T-cells were expanded with pamidronate or the TAA-Vδ2^hi-lo bsVHH, but was expressed by a lower proportion of Vγ9Vδ2 T-cells when these were expanded with the TAA-Vδ2^hi-lo bsVHH-Fc engagers, though this did not reach statistical significance. The upregulation of activation markers on expanded Vγ9Vδ2 T-cells was accompanied by an increase in the proportion of Vγ9Vδ2 T-cells expressing co-inhibitory receptors including NKG2A, CTLA-4, and TIGIT. Of note, the proportion of Vγ9Vδ2 T-cells expressing the inhibitory immune checkpoint receptor PD-1 was not altered. Overall, the increased expression of multiple inhibitory receptors on expanded Vγ9Vδ2 T-cells likely reflects their increased activation state (47, 48). Indeed, NKG2A was previously reported to identify a population of Vγ9Vδ2 T-cells with greater cytotoxic potential (11). In the conditions where cancer patient PBMC were cultured with pamidronate or the Vδ2^hi-lo bsVHH containing constructs, Vγ9Vδ2 T-cells expressed increased levels of the activation markers CD25 and HLA-DR ( Figure 4E ). At baseline, expression of DNAM-1 and NKG2D was more variable on cancer patient Vγ9Vδ2 T-cells and expression did not significantly change upon expansion with the TAA-Vδ2^hi-lo bsVHH and bsVHH-Fc molecules. Of interest, while PD-1 was typically expressed by a low proportion of Vγ9Vδ2 T-cells in healthy donor PBMC at baseline, more variable expression was noted on Vγ9Vδ2 T-cells in cancer patient PBMC with a further increase in the proportion of PD-1 expressing Vγ9Vδ2 T-cells noted in several patients after expansion with either pamidronate or the Vδ2^hi-lo bsVHH constructs. NKG2A expression by Vγ9Vδ2 T-cells was variable at both baseline and upon expansion. CTLA-4 and TIGIT were expressed by an approximately similar proportion of Vγ9Vδ2 T-cells in cancer patient and healthy donor PBMC and were in both cases upregulated upon expansion, though the proportion of Vγ9Vδ2 T-cells expressing CTLA-4 appeared to be more consistently increased upon activation and expansion in cancer patient PBMC.

While all TAA-Vδ2^hi-lo bsVHHs supported expansion of Vγ9Vδ2 T-cells, results were consistently robust with the CD1d-Vδ2^hi-lo bsVHH and this engager was therefore selected to assess whether Vγ9Vδ2 T-cell expansion could also be induced in vivo. As mice lack the phosphoantigen responsive γδ T-cell population that humans have, an immunodeficient NOG-hIL-15 mouse model was selected as this allowed for inoculation of human PBMC and could simultaneously provide a relatively low level of IL-15 cytokine support. Mice (n=4/group) were i.v. inoculated with human PBMC (n=1 healthy donor with 0.89% Vγ9Vδ2 T-cell of total CD3⁺ T-cells) on day 0 followed by i.p. administration of either PBS or CD1d-Vδ2^hi-lo bsVHH on days 0 and 4 ( Figure 5A ). On day 8, animals were sacrificed and the proportion of Vγ9Vδ2 T-cells was assessed in peripheral blood, spleen, liver and lungs (gating strategy illustrated in Supplementary Figure S7 ). As shown in Figures 5B, C , the proportion of Vγ9Vδ2 T-cells was significantly higher in mice treated with the CD1d-Vδ2^hi-lo bsVHH compared to the PBS control. Administration of the CD1d-Vδ2^hi-lo bsVHH resulted in an impressive 11525 (± 7692; median ± IQR, P = 0.004) fold expansion of the Vγ9Vδ2 T-cells in peripheral blood. CD1d-Vδ2^hi-lo bsVHH expanded Vγ9Vδ2 T-cells isolated from spleen, liver and lungs all expressed a central or effector memory phenotype, whereas 4.8% (± 1.0; mean ± SEM) of the Vγ9Vδ2 T-cells in peripheral blood retained a naïve phenotype ( Figure 5D ). In summary, these results indicate that Vδ2^hi-lo bsVHH based engagers can also trigger expansion of Vγ9Vδ2 T-cells in a NOG-hIL-15 mouse in vivo model.