BiTE-Secreting CAR-γδT as a Dual Targeting Strategy for the Treatment of Solid Tumors

Abstract

HLA-G is considered as an immune checkpoint protein and a tumor-associated antigen. In the previous work, it is reported that CAR-NK targeting of HLA-G can be used to treat certain solid tumors. However, the frequent co-expression of PD-L1 and HLA-G) and up-regulation of PD-L1 after adoptive immunotherapy may decrease the effectiveness of HLA-G-CAR. Therefore, simultaneous targeting of HLA-G and PD-L1 by multi-specific CAR could represent an appropriate solution. Furthermore, gamma-delta T (γδT) cells exhibit MHC-independent cytotoxicity against tumor cells and possess allogeneic potential. The utilization of nanobodies offers flexibility for CAR engineering and the ability to recognize novel epitopes. In this study, Vδ2 γδT cells are used as effector cells and electroporated with an mRNA-driven, nanobody-based HLA-G-CAR with a secreted PD-L1/CD3ε Bispecific T-cell engager (BiTE) construct (Nb-CAR.BiTE). Both in vivo and in vitro experiments reveal that the Nb-CAR.BiTE-γδT cells could effectively eliminate PD-L1 and/or HLA-G-positive solid tumors. The secreted PD-L1/CD3ε Nb-BiTE can not only redirect Nb-CAR-γδT but also recruit un-transduced bystander T cells against tumor cells expressing PD-L1, thereby enhancing the activity of Nb-CAR-γδT therapy. Furthermore, evidence is provided that Nb-CAR.BiTE redirectes γδT into tumor-implanted tissues and that the secreted Nb-BiTE is restricted to the tumor site without apparent toxicity.

Keywords: antigen heterogenicity; bispecific T-cell engager; chimeric antigen receptor; gamma-delta T; mRNA; nanobody.

Scheme 1

Elevated PD‐L1 in solid tumors increases the risk of immune escape from HLA‐G‐CAR cell therapy. The bicistronic mRNA construct that drives PD‐L1 Nb‐BiTE and HLA‐G Nb‐CAR in γδT cells via electroporation was designed to address this issue. This Nb‐CAR.BiTE‐γδT therapy can overcome HLA‐G and PD‐L1 dilemma and even kill tumor cells with inadequate antigen expression, resulting in potent anti‐tumor activity without apparent toxicity.

Figure 1

PD‐L1 blockade reinforces anti‐HLA‐G Nb‐CAR‐γδT to eliminate PD‐L1‐overexpressing immune escape variants. A) Schematic diagram of anti‐HLA‐G CAR construct. HLA‐G‐targeted Nb#1 and Nb#2 were ligated with CD8α hinge/TM, followed by a 4‐1BB and a CD3ξ intracellular domains, which was driven by EF‐1α promoter using lentiviral vector. B) Generation of immune escape variants following anti‐HLA‐G CAR‐γδT challenge in vivo. Briefly, NSG mice (n = 5) were orthotopically implanted with luciferase‐expressing MDA‐MB‐231 cells. After 7 d, the mice were infused with anti‐HLA‐G Nb‐CAR‐γδT cells (1 × 10⁷) via tail vein injection weekly for 8 weeks. The mice were sacrificed at 63 days after the first challenge of Nb‐CAR‐γδT, and then residual tumors were harvested. Subsequently, the isolated cells were re‐challenged with anti‐HLA‐G Nb‐CAR‐γδT at an E:T ratio of 3:1 for 72 h. C–E) Three immune escape variants generated from individual mice showed consistently upregulated PD‐L1 levels. C) Bioluminescent signals were detected using IVIS after the MDA‐MB‐231 tumor‐bearing mice were treated with or without HLA‐G‐targeted Nb‐CAR‐γδT cells. D) Bioluminescence of the residual clones (231‐R1, ‐R2, ‐R3) was measured using IVIS. Parental MDA‐MB‐231 and MDA‐MB‐231‐luc cells were used as the control (left). Expression levels of ICPs were determined through flow cytometry (right). E) PD‐L1 expression in these clones was also confirmed by immunoblotting. F) Combination with atezolizumab enhanced anti‐HLA‐G Nb‐CAR‐γδT‐induced cytotoxicity against the immune escape variants. Parental MDA‐MB‐231, 231‐R1, ‐R2, and ‐R3 cells were pretreated with or without 10 µg mL⁻¹ atezolizumab for 15 min and followed by coculture with or without parental or anti‐HLA‐G CAR‐γδT cells for 48 h at an E:T ratio of 3:1. Subsequently, we determined specific lysis through a LIVE/DEAD cell‐mediated cytotoxicity assay. G‐I) PD‐L1 blockade reverses the sensitivity of 231‐R3 tumor cells to anti‐HLA‐G CAR‐γδT challenge in vivo. G) NSG mice (n = 5) were orthotopically implanted with 231‐R3 cells. After 10 d, the mice were injected with or without atezolizumab (20 mg kg⁻¹) combined with parental or anti‐HLA‐G CAR‐γδT weekly for 4 weeks through the tail vein. H,I) Bioluminescence of the 231‐R3 tumors was measured (H,I), and J) survival rates were recorded. These results suggested that PD‐L1 is upregulated in persistent tumor cells after HLA‐G CAR‐γδT challenge, and the blockade of PD‐L1 restored their sensitivity to HLA‐G CAR‐γδT cells. Results are representative of at least three independent experiments. Data represent the mean ± SD, n = 4; *p < 0.05; **p < 0.01; ***p < 0.001; and #, not detectable based on paired Student's t‐tests. For tumor growth and survival rate comparisons, the Kaplan–Meier method and log‐rank test were performed.

Figure 2

Expression of PD‐L1 in NSCLC and TNBC cells interferes with their sensitivity to HLA‐G targeted CAR‐γδT cells. A,B) HLA‐G and PD‐L1 were frequently expressed in the same lung (n = 24) and TNBC (n = 30) tumor lesions. HLA‐G and PD‐L1 expressions were determined by 2‐plexed IHC staining using specific antibodies (left). Scale bar = 20 µm of. HLA‐G and PD‐L1 expression levels in tumor lesions and normal adjacent tissues were shown as a tissue cytometry dot‐plot (middle panel) and the quantified data are displayed (right panel). C,D) Expression levels of HLA‐G and PD‐L1 in lung cancer and breast cancer cell lines. C) A549 and H1975; D) MCF‐7, T47D and MDA‐MB‐231 cells were harvested to evaluate the levels of HLA‐G and PD‐L1 by immunoblotting (left panels) and flow cytometry (right panels). E) PD‐L1 blockade re‐enforced cytotoxicity induced by Nb‐CAR‐γδT cells in HLA‐G^high/PD‐L1^high tumor cells. MDA‐MB‐231 and H1975 cells were pretreated with or without 10 µg mL^‐1 atezolizumab for 15 min and then cocultured with parental or Nb‐CAR‐γδT cells (E:T = 3:1) for 48 h. F) PD‐L1 antagonized HLA‐G targeted CAR‐γδT‐induced cytolysis. Mock or PD‐L1‐overexpressed A549 cells were treated with or without 10 µg mL⁻¹ atezolizumab for 15 min and then co‐incubated with parental or Nb‐CAR‐γδT cells (E:T = 3:1) for 48 h. Specific lysis was performed by the LIVE/DEAD cell‐mediated cytotoxicity assay using flow cytometry analysis. These results suggested that elevated PD‐L1 levels on tumor cells may be a risk of immune escape from Nb‐CAR‐γδT therapy in solid tumors. Results are representative of at least three independent experiments. Data represent the mean ± SD, n = 3–4, *p < 0.05; **p < 0.01; and ***p < 0.001 based on paired Student's or Student's t‐tests.

Figure 3

Characterization of the Nb‐CAR.BiTE construct. A) Schematic representation of mRNA and lentiviral HLA‐G Nb‐CAR, with or without combined secretable PD‐L1/CD3ε‐targeted Nb‐BiTE constructs. The bi‐epitopic Nb‐CAR comprises two tandem extracellular HLA‐G Nbs fused with a CD8α hinge/TM, followed by a modified 4‐1BB cytosolic domain (4‐1BB/Tyk) and a modified CD3ξ intracellular domain (CD3ξ/ITAM) linked to a self‐cleavage peptide P2A to separate the secretable bivalent Nb‐BiTE, which includes two PD‐L1 Nbs linked to a CD3ɛ Nb. B) The blocking activity of HLA‐G Nbs. The blockade activity of HLA‐G Nb clones #1 and #2 or HLA‐G mAb 87G to LILRB1/HLA‐G (upper panels) and KIR2DL4/HLA‐G (bottom panel) were determined by competitive ELISA. C) HLA‐G Nb moiety capable to bind to all HLA‐G isoforms. The binding activity of HLA‐G Nb clones and it derived bi‐epitopic Nb moiety to recombinant HLA‐G‐β2M (corresponding to HLA‐G G1 and G5), HLA‐G α1α3 domains (G2 and G6), HLA‐G α1α2 domains (G4) and HLA‐G α1 domain (G5) were examined by ELISA‐based binding assay. D) PD‐L1 blockade Nbs. The activity of PD‐L1 Nb clones #1 and #2 was determined by PD‐L1/PD‐1 blockade bioassay. E) Biofunction and affinity of CD3ε Nb. PBMCs (1 × 10⁶ cells) were treated with/ without 100 ng CD3ε Nb or OKT3. The enriched γδT cells from PBMCs were supplemented with 1000 IL mL^‐1 IL‐2 and then treated with or without 100 ng CD3ε Nb or OKT3. After 6 d, the images were taken (upper panel), Scale bar = 20 µm, and the cell numbers were counted and normalized to the CD3⁺ cell populations (bottom panel). Evaluation of the novel CAR signaling cassette. F) Bi‐epitopic HLA‐G Nb‐CARs were engineered using 4‐1BB‐CD3ξ (BB‐3z) or modified signaling cassettes (IFNAR/BB‐3z or BBTyk‐3z). G) BB‐3z‐, IFNAR/BB‐3z‐, or BBTyk‐3z‐based HLA‐G Nb‐CAR lentiviral particles (MOI = 3) were transduced into γδT cells, and after three days, the expression of Nb‐CAR was determined by flow cytometry. H,I) Parental and BB‐3z‐, IFNAR/BB‐3z‐, or BBTyk‐3z‐Nb‐CAR‐γδT cells were cocultured with MDA‐MB‐231 cells at an E:T ratio of 3:1. H) After 2 h, the phosphorylation status of STAT2 and Syk/ZAP70 in the CAR‐γδT cells was determined; and I) after 24 h, the induced cytotoxicity was determined using a LIVE/DEAD cell‐mediated cytotoxicity assay (upper left panel); CD107a expression in CAR‐γδT cells was detected by flow cytometry using specific antibodies (bottom left panel); and their supernatants were harvested to detect the contents of granzyme B, IFN‐γ, TNF‐α, and IL‐17A by ELISA. These data demonstrated that the elements for engineering the Nb‐CAR.BiTE construct were functional and may be superior than that of the conventional CAR design. Results are representative of at least three independent experiments. Data represent the mean ± SD, n = 3; *p < 0.05; **p < 0.01; and ***p < 0.001 based on paired Student's t‐test.

Figure 4

Characterization of mRNA‐driven Nb‐CAR.BiTE‐γδT cells. A) Nb‐CAR expression was assessed after lentiviral transduction or mRNA electroporation. γδT cells (1 × 10⁶) were electroporated with 2 µg Nb‐CAR/CAR.BiTE IVT mRNA or transduced with lentiviral Nb‐CAR/CAR.BiTE particles (MOI = 3). The Nb‐CAR expression on the indicated days was determined through flow cytometry using a specific antibody against VHH. B,C) Superior cell growth and Nb‐BiTE secretion using the mRNA delivery strategy. B) Cell number and viability were recorded during 7 d after transfection of the mRNA or lentiviral Nb‐CAR.BiTE transgene. C) Contents of PD‐L1‐targeted Nb‐BiTE in the supernatants from mRNA‐ or lentiviral‐driven Nb‐CAR.BiTE‐γδT cell cultures were detected by an ELISA‐based coating with recombinant full‐length PD‐L1. D) Phenotype and purity of mRNA‐driven Nb‐CAR.BiTE‐γδT cells. T effector and central memory phenotype, activating marker, and purity of Nb‐CAR.BiTE mRNA‐electroporated γδT cells were analyzed through flow cytometry using specific antibodies against CD27, CD45RA, CD3, NKG2D, Vδ2, Vγ9, αβTCR, γδTCR, CD4, CD8, CD19, CD14, CD66b, or CD56. Isotype staining was used as a gating control for positive staining. E) Increased secretion of granzyme B and IFN‐γ, but not IL‐17A, from Nb‐CAR.BiTE‐γδT cells after coculture with A549 or MDA‐MB‐231 cells. The contents of granzyme B (upper panel), IFN‐γ (middle panel), and IL‐17A (bottom panel) in culture supernatants were measured using ELISA. These results showed that mRNA‐driven Nb‐CAR.BiTE‐γδT cells would be an effective strategy for the manufacture of Nb‐CAR.BiTE‐γδT cells. Results are representative of at least three independent experiments. Data represent the mean ± SD, n = 3–4, *p < 0.05; **p < 0.01; and ***p < 0.001 based on paired Student's t‐tests.

Figure 5

Evaluation of Nb‐BiTE secreted from Nb‐CAR‐γδT cells. A–D) Functional binding of Nb‐BiTE released from Nb‐CAR‐γδT cells. A) Diagram shows the cell non‐contact coculture system for Nb‐CAR.BiTE‐γδT cells and target cells (left panel). Secreted Nb‐BiTE from Nb‐CAR‐γδT cells was detectable on target cells. We seeded 1 × 10⁵ isolated primary CD3‐positive cells from PBMCs, A549, H1975, or MDA‐MB‐231 cells on the bottom, with or without exposure to 5 × 10⁵ Nb‐CAR.BiTE‐γδT cells on the top, in impenetrable wells for 24 h. Subsequently, the bottom cells were harvested and Nb‐BiTE signals were detected by flow cytometry using specific antibody against VHH (right panel). B) Secreted Nb‐BiTE from Nb‐CAR‐γδT cells promoted CD3‐positive cell proliferation. We cocultured 1 × 10⁶ isolated primary CD3⁺ cells, parental, or Nb‐CAR‐γδT cells with or without 1 × 10⁶ parental, Nb‐CAR, or Nb‐CAR.BiTE‐γδT cells on the top well for 6 days. Subsequently, the CD3⁺ cell numbers were normalized to the non‐coculture group. C) Nb‐BiTE released from Nb‐CAR‐γδT cells strengthened effector cells against PD‐L1‐expressing tumor cells. Schematic illustration of the non‐contact coculture system for Nb‐CAR.BiTE‐γδT cells, which release Nb‐BiTE to engage CD3⁺ effectors and PD‐L1‐overexpressing tumor cells (left panel). The PD‐L1 level on PD‐L1‐overexpressing A549 and PD‐L1 knockdown MDA‐MB‐231 stable clones were examined by flow cytometry analysis (right panel). D) Purity of the isolated CD4‐ and CD8‐positive and CD3‐/CD56⁺ cells from PBMCs were checked by flow cytometry. E) We cocultured 1 × 10⁵ PD‐L1‐overexpressing A549 cells with an individual healthy donor‐derived CD4⁺, CD8⁺, NK, parental γδT, or Nb‐CAR‐γδT cells on the bottom well and with or without 5 × 10⁵ Nb‐CAR.BiTE‐γδT cells on the top well for 72 h. F) PD‐L1 determined the cell‐killing ability of effector cells when exposed to released Nb‐BiTE. PD‐L1‐stable knockdown or scramble‐control MDA‐MB‐231 cells (1 × 10⁵) were cocultured with an individual healthy donor‐derived CD4⁺, CD8⁺, NK, parental γδT, or Nb‐CAR‐γδT cells on the bottom well and with 5 × 10⁵ Nb‐CAR.BiTE‐γδT cells on the top well for 72 h. G,H) PD‐L1‐targeted Nb‐BiTE enhanced anti‐tumor responses of T cells. MDA‐MB‐231 or PD‐L1‐overexpressing A549 cells were cocultured with or without the isolated CD4⁺ or CD8⁺ cells at an E:T ratio of 5:1; or γδT cells at an E:T ratio of 3:1 in the presence/absence of recombinant Nb‐BiTE (1 ng mL^‐1). G) After 24, 48, or 72 h of coculture, we determined the cytotoxicity induced by these effector cells. H) After 48 h of coculture, the supernatants were harvested for detecting the contents of granzyme B and IFNγ by ELISA. Specific lysis was determined using LIVE/DEAD cell‐mediated cytotoxicity assay. These results suggested that the secreted PD‐L1 targeting Nb‐BiTE from Nb‐CAR.BiTE‐γδT could trigger bystander effector cells active against PD‐L1‐expressing tumor cells. Results are representative of at least three independent experiments. Data represent the mean ± SD, n = 4, *,^#,^X p < 0.05; **,^##,^XX p < 0.01; and ***,^###,^XX X p < 0.001 based on paired Student's t‐test. For sub‐Figure G, * represents significant differences between Nb‐BiTE and effector cells (Nb‐BiTE vs effector cells); # represents significant differences between Nb‐BiTE and effector cells+Nb‐BiTE; ^X represents significant differences between effector cells and effector cells+ Nb‐BiTE.

Figure 6

Nb‐CAR.BiTE‐γδT cells are effective to treat tumor cells with HLA‐G and/or PD‐L1 expression in vitro. A,B) Nb‐CAR.BiTE‐γδT cells effectively ablated HLA‐G⁺ and/or PD‐L1‐overexpressed tumor cells. Parental, Nb‐CAR, or Nb‐CAR.BiTE‐γδT cells were cocultured with 1 × 10⁵ MDA‐MB‐231 or H1975 cells A) mock or PD‐L1‐stably overexpressing A549 cells B) at an E:T ratio of 1:1–10:1 for 72 h. Induced cytotoxicity was measured by LIVE/DEAD cell‐mediated cytotoxicity assay C) Nb‐CAR.BiTE‐γδT cells have potent cytotoxicity against the PD‐L1‐overexpressing immune escape variants. We cocultured 1 × 10⁵ parental MDA‐MB‐231, 231‐R1, ‐R2 or ‐R3 cells with Parental, Nb‐CAR, or Nb‐CAR.BiTE‐γδT cells at an E:T ratio of 3:1 for 72 h. Specific lysis was determined using LIVE/DEAD cell‐mediated cytotoxicity assay. D) γδT effector cells retain their natural anti‐tumor effect on tumor cells with low levels of CAR or BiTE antigen expression. HLA‐G and PD‐L1 levels in H1975 and MDA‐MB‐231 cells were detected through flow cytometry following HLA‐G or PD‐L1 stable knockdown (left panel). HLA‐G‐ or PD‐L1‐knockdown H1975 and MDA‐MB‐231 cells (1 × 10⁵) were co‐incubated with parental, Nb‐CAR‐expressing, or Nb‐CAR.BiTE‐expressing γδT cells at an E:T ratio of 1:1, 2:1, 3:1, and 6:1 for 72 h. Subsequently, specific lysis of target cells was determined by a LIVE/DEAD cell‐mediated cytotoxicity assay. E,F) PD‐L1 on tumor cells determines the sensitivity of secreted Nb‐BiTE, and both HLA‐G and PD‐L1 expression affect the cytolysis induced by γδT cells. E) Schematic illustration of non‐contact coculture system for Nb‐CAR.BiTE‐γδT cells, which release Nb‐BiTE to engage γδT cells against H1975 cells. F) H1975 cells were transfected with or without siHLA‐G and/or siPD‐L1 for 48 h, then seeded in the bottom wells (1 × 10⁵ cells per well), and then cocultured with parental or Nb‐CAR‐γδT cells (E:T = 1:1) with or without Nb‐CAR.BiTE‐γδT cells in the top transwell inserts (5 × 10⁵ cells per insert) for 72 h. Subsequently, the specific lysis of target cells was examined using a LIVE/DEAD cell‐mediated cytotoxicity assay. These results suggested that Nb‐CAR.BiTE‐γδT cells are effective to against HLA‐G and/or PD‐L1 positive tumor cells even antigens are low expressed. Results are representative of at least three independent experiments. Data represent the mean ± SD, n = 4, *,^#,^X p < 0.05; **,^##,^XX p < 0.01; ***,^###,^XX X p < 0.001; stars represent significant differences between parental and Nb‐CAR.BiTE‐γδT; pound signs represent significant differences between parental and Nb‐CAR‐γδT; and double daggers represent significant differences between Nb‐CAR‐γδT and Nb‐CAR.BiTE‐γδT based on paired Student's t‐tests.

Figure 7

Anti‐tumor activity of Nb‐CAR.BiTE‐γδT cells in the orthotopic PBMC‐CDX‐NSG mouse models. A) Protocol for evaluating the anti‐tumor activity of mRNA‐driven Nb‐CAR.BiTE‐γδT cells in the PBMC‐CDX‐NSG mice bearing orthotopic HLA‐G^high/PD‐L1^high TNBC or PD‐L1 overexpressing lung tumors. Seven days after inoculation with 1 × 10⁶ tumor cells, the mice received tail vein injection with PBMCs (5 × 10⁶ cells per mouse). Three days later, the parental, Nb‐CAR or Nb‐CAR.BiTE‐γδT cells (1 × 10⁷ cells/mouse) were injected through tail vein once a week for 4 weeks. B‐G) Superior anti‐tumor activity of Nb‐CAR.BiTE‐γδT cells in the PBMC‐CDX‐NSG mouse model. The tumor growth of the orthotopically inoculated B,C) MDA‐MB‐231 (n = 10) and E,F) PD‐L1‐overexpressing A549 cells (n = 5) were monitored weekly through IVIS. D,G) Survival rates were recorded. These results suggested that Nb‐CAR.BiTE‐γδT is effective against HLA‐G/PD‐L1‐double positive solid tumors as well as PD‐L1‐overexpressing solid tumors in vivo. Results are representative of three independent experiments. Data represent the mean ± SD; *p < 0.05; **p < 0.01; and ***p < 0.001 based on the Kaplan–Meier method and log‐rank test.

Figure 8

Biodistribution of Nb‐CAR‐expressing γδT cells in vivo. A‐C) Nb‐CAR expression enhanced γδT cells accumulated in the local breast tumor site. XenoLight DiR dye‐stained parental, Nb‐CAR or Nb‐CAR.BiTE‐γδT cells (1 × 10⁷ cells per mouse) were infused into MDA‐MB‐231 tumor‐bearing NSG mice (n = 5) via tail vein injection. A) Fluorescent signals of each mouse were monitored by IVIS at the indicated time points. B,C) Mice were sacrificed after 24 h, B) organs and tumor tissues were harvested for measuring fluorescent signals by IVIS, and C) the quantitative results were identified and normalized per gram of tissue samples. D) Nb‐BiTE was only detectable in inoculated tumors of Nb‐CAR.BiTE‐γδT‐treated mice bearing MDA‐MB‐231 tumors, but not in other organs. The contents of PD‐L1/CD3ε Nb‐BiTE in the tissue extracts (5 µg) were determined by ELISA‐based method coating with PD‐L1 recombinant protein. The supernatants of Nb‐CAR.BiTE‐γδT cell cultures were used as a positive control. E–G) Nb‐BiTE‐secreting Nb‐CAR‐γδT cells increased the accumulation of infused γδT cells at the PD‐L1‐overexpressing tumor inoculation site. PD‐L1‐overexpressing A549 tumor‐bearing mice were injected with XenoLight DiR dye‐stained parental, Nb‐CAR, or Nb‐CAR.BiTE‐γδT cells (1 × 10⁷ cells per mouse). E) The fluorescent signals were monitored by IVIS at the indicated time points. F,G) 24 h after γδT cell infusion, the mice were sacrificed and fluorescent signals of the organs and tumor tissues were determined by IVIS (F) and quantified (G). H) Presence of Nb‐BiTE was restricted in PD‐L1‐overexpressing tumor inoculation lung tissues but not in other organs in vivo. The tissue extracts were harvested and 5 µg of each sample were analyzed the expression of PD‐L1/CD3ε Nb‐BiTE by ELISA‐based method with the coating of PD‐L1 recombinant protein. The supernatants of Nb‐CAR.BiTE‐γδT cell cultures were used as a positive control. Results are representative of at least three independent experiments. Data represent the mean ± SD, n = 3, *p < 0.05; **p < 0.01; and ***p < 0.001; paired Student's t‐test.