Simultaneous secretion of a PD-L1 x 4-1BB bispecific antibody improves antileukemic efficacy of STAb-T cells secreting a CD19-specific T-cell engager

ABSTRACT

Adoptive therapy with CAR-T cells and systemic administration of bispecific T-cell engagers (TCE) have achieved unprecedented success in the treatment of relapsed/refractory (R/R) B-cell malignancies. However, high relapse rates remain a major challenge. STAb (Secretion of T cell-redirecting bispecific Antibodies)-T-cell immunotherapy represents a promising alternative by enabling both polyclonal T-cell recruitment and sustained bispecific antibody release. Here, we describe an evolution of STAb-T19 therapy, which has demonstrated superior outcomes to those of CAR-T-19 cells in preclinical models of B-ALL, on the basis of the simultaneous secretion of two bispecific antibodies: a CD19 × CD3 TCE and a PD-L1 × 4-1BB bsAb. The combined approach aims to increase the antitumor efficacy of STAb-T19 cells by blocking the PD−1/PD-L1 axis with conditional 4-1BB costimulation to ensure the long-term persistence of STAb-T cells. Preclinical data show that this combination improves cytotoxic activity and prolongs antileukemic efficacy compared with low-dose single STAb-T therapy. Our findings suggest that the integration of PD-L1 × 4-1BB bsAbs into STAb-T19 therapy may maximize efficacy, thereby opening a promising avenue to address resistance and relapse in B-ALL. Moreover, this approach may pave the way for the development of next-generation cell-based therapies for hematologic and solid malignancies.

Introduction

In recent years, immunotherapy strategies based on CD19 targeting, such as chimeric antigen receptor (CAR)-T19 therapy or systemic administration of the anti-CD19/CD3 bispecific antibody (bsAb) blinatumomab, have revolutionized the treatment of B-cell leukemias and lymphomas.^1,2 However, disease recurrence is still observed in 30−60% of patients receiving CAR-T-19 therapy,³ and up to 80% of patients relapse after blinatumomab treatment.^4,5 To address these challenges, an emerging strategy combines adoptive cell therapy with bsAbs in an approach known as STAb (Secretion of T cell redirecting bispecific Antibodies)-T-cell immunotherapy. STAb-T-cell therapy is based on the modification of T-cells to in vivo secrete bsAbs that target a tumor-associated antigen (TAA) and CD3, also known as T-cell engagers (TCEs).^6,7 In contrast to systemically administered bsAbs, STAb-T cells provide long-term and sustained release of TCEs, compensating for the rapid blood clearance of small-sized bsAbs.⁸ Importantly, in contrast to membrane-anchored CARs, soluble secreted bsAb enables the polyclonal recruitment of both STAb-T cells and unmodified bystander T cells to the tumor, increasing the antitumor response.^9,10 We have recently demonstrated that STAb-T19 cells, which secrete a CD19 × CD3 bsAb, outperform CAR-T19 therapy in relevant in vivo models of B-ALL.¹¹ However, STAb-T-cell therapy may face certain challenges. Although bsAbs have demonstrated the ability to induce potent T cell cytotoxicity without the need for additional costimulatory signals^8,9,11 and the in vivo expansion of memory STAb-T cells has been reported,¹¹ concerns might arise regarding the long-term persistence of STAb-T cells in the absence of costimulation. In this context, first-generation CAR-T cells devoid of costimulatory domains have shown limited or no clinical efficacy. The inclusion of CD28 or 4-1BB (CD137) costimulatory domains in second-generation CARs is necessary to increase CAR-T-cell persistence and achieve strong therapeutic responses in patients.¹² In addition, two agonistic antibodies targeting 4-1BB, urelumab and utomilumab, have been tested in clinical trials to promote antitumor T cell responses.¹³ However, poor outcomes have been obtained due to toxicity or a lack of efficacy.¹⁴ More recently, multivalent antibodies have been designed to combine the 4-1BB-stimulating component with a tumor-targeting domain to reduce systemic toxicity.^13,15–17

On the other hand, the expression of inhibitory checkpoint molecules on tumor cells reduces the antitumor efficacy of T lymphocytes. To overcome this limitation, the administration of antibodies that block the PD−1/PD-L1 axis has been widely evaluated and has achieved remarkable success in several cancers.¹⁸ PD-L1 has been shown to be upregulated in blasts from relapsed B-ALL patients and from those refractory to blinatumomab.¹⁹ Moreover, PD-L1 blockade enhanced blinatumomab-induced antileukemic T cell responses in vitro.¹⁹ Different clinical trials evaluating anti-PD−1 mAbs in B-ALL (NCT05310591, NCT04546399, and NCT02879695) are underway, some of which involve combination therapy with blinatumomab. In addition, certain aggressive B-cell lymphomas express PD-L1.²⁰ Thus, anti-PD−1 antibodies are being evaluated in different clinical trials for the treatment of B-cell non-Hodgkin lymphoma in combination with CAR-T cells, kinase inhibitors or immunomodulators (NCT04539444, NCT02950220, and NCT03015896), and pembrolizumab has been approved for the treatment of large B-cell lymphoma.¹⁸

More recently, the preclinical synergy of CD137 agonists with PD-(L)1 blockade has prompted the development of different PD-L1 × 4-1BB bispecific antibodies,^21–23 which have demonstrated improved outcomes compared with those of anti−4-1BB antibodies in terms of hepatotoxicity.¹³ The IgG1-based Fc-silenced PD-L1 x 4-1BB bsAb acasunlimab (Gen1046) showed, in combination with pembrolizumab, a manageable safety profile, deep responses and durable disease control in phase I and II clinical trials (NCT05117242 and ongoing) in PD-L1⁺ NSCLC patients.^24,25 A similar PD-L1 x 4-1BB bsAb, MCLA−145, given alone or in combination with pembrolizumab, has shown good tolerability and a manageable safety profile with encouraging clinical activity in an ongoing phase I clinical trial with patients with advanced/metastatic solid tumors.²⁶ Finally, CAR-T cells engineered to secrete a fusion protein comprising the soluble trimeric 4-1BB ligand linked to an anti-PD−1 scFv have shown enhanced antitumor efficacy in vivo compared with conventional CAR-T cells in both hematologic and solid malignancies.²⁷

Here, we propose a STAb-T-cell therapy based on the simultaneous secretion of two different bispecific antibodies: CD19 × CD3 (19-bsAb) and PD-L1 × 4-1BB (P4-bsAb). Our study suggests that the secretion of both bsAbs improves the antitumor efficacy of STAb-T19 cells and extends their cytotoxic activity over time. Therefore, the combination of PD-L1 x 4-1BB with TAA x CD3 bsAbs could improve the outcomes of STAb-T19 and other STAb-T therapies.

Materials and methods

Cell lines and culture conditions

HEK293 (CRL−1573), HEK293T (CRL−3216), Jurkat Clon E6−1 (TIB−152), CHO-K1 (CCL−61), Nalm6 (CRL3273), and K562 (CCL−243) cells were purchased from the American Type Culture Collection (Rockville, MD, USA). Nalm6 and K562 cells expressing the firefly luciferase (Luc) gene (Nalm6^Luc and K562^Luc) were generated in house by transduction with the lentiviral vector pRRL‐Luc‐IRES‐EGFP.²⁸ Nalm6^PD-L1 and Nalm6^PD-L1/Luc were produced in house by transduction of Nalm6 or Nalm6^Luc, respectively, with a pCCL lentiviral vector encoding PD-L1 cDNA synthesized by GeneArt AG (Thermo Fisher Scientific, Regensburg, Germany). Jurkat T cells expressing human PD−1 and NFAT-induced luciferase (Jurkat^{NFAT-RE-luc/PD-1}), referred to as Jurkat^PD-1, and CHO-K1 cells expressing human PD-L1 (PD-L1 aAPC/CHO-K1), referred to as CHO^APC/PD-L1, were obtained from Promega (Cat. No. J1250, Promega, Madison, WI, USA). Jurkat T cells stably expressing human 4-1BB and NFAT-induced luciferase (Jurkat^{NFAT-RE-luc/-4-1BB}), referred to as Jurkat^4-1BB, were obtained from Promega (Cat. No. JA2352, Promega). Jurkat T cells expressing GFP-tagged 4-1BB (Jurkat^4-1BB/GFP) were generated by lentiviral transduction using commercial lentiviral particles (Cat. No. RC200664L4V, Origene, Rockville, MD, USA) as previously described.²⁹ Jurkat ^{4-1BB/PD-L1-KO} cells were generated by CRISPR-Cas9 gene editing. Recombinant Cas9 protein and single-guide RNAs (sgRNAs) were obtained from Integrated DNA Technologies (IDT). sgRNAs were designed using multiple software tools, including the CRISPR Design Tool (crispr.mit.edu) and Benchling (benchling.com), which prioritize high on-target scores and minimal off-target probabilities. Three sgRNAs targeting the CD274 (PD-L1) coding sequence were employed: sgCD274_1: GGTTCCCAAGGACCTATATG; sgCD274_2: ACCGTTCAGCAAATGCCAGT; and sgCD274_3: AATAGACAATTAGTGCAGCC. Recombinant Cas9 protein (2.5  µg) was preincubated with each sgRNA (2 µg) for 10−20 minutes at room temperature to form ribonucleoprotein (RNP) complexes. Cell transfections were performed using the Neon Transfection System (Cat. No. N10096; Thermo Fisher Scientific). Briefly, 2 × 10⁵ Jurkat^4-1BB/GFP cells per reaction were electroporated in 10 µL of Buffer R (Cat. No. BR5, Thermo Fisher Scientific). Electroporation was carried out using 3 pulses of 10 ms duration at 1325 V. Postelectroporation, the cells were seeded into 24-well plates containing prewarmed complete medium and allowed to recover. Following culture expansion, single cells were sorted into 96-well plates using fluorescence-activated cell sorting (FACS). Clonal populations were established and maintained for subsequent screening and validation of PD-L1 knockout efficiency.

HEK293- and CHO-K1-derived cell lines were cultured in Dulbecco’s modified Eagle’s medium (DMEM) (Cat. No. 10313021, Life Technologies, Carlsbad, CA, USA) supplemented with 10% (vol/vol) heat-inactivated fetal bovine serum (FBS) (Cat. No. F7524, Merck Life Science, Darmstadt, Germany), 2 mmol/L L-glutamine (Cat. No. 25030081, Life Technologies) and antibiotics (100 units/mL penicillin, 100 µg/mL streptomycin; Cat. No. P4333, Sigma-Aldrich, St. Louis, Missouri, USA), referred to as complete DMEM. Jurkat-, Nalm6-, and K562-derived cell lines were cultured in RPMI−1640 (Cat. No. 12-702Q; Lonza Bioscience, Basel, Switzerland) supplemented with heat-inactivated 10% FBS, 2 mmol/L L-glutamine and antibiotics, referred to as complete RPMI medium (RCM). All the cell lines were grown at 37°C in 5% CO₂ and were routinely screened for mycoplasma contamination by PCR using the Mycoplasma Gel Detection Kit (Cat. No. 90.021−4542, Biotools, Madrid, Spain).

Vectors

The SAP3.28 scFv¹⁷ from the pMA-RQ−4-1BB-scFv plasmid, synthesized by GeneArt AG, was subcloned as SalI/BglII into the plasmid pCR3.1αPD-L1-αCD3 (unpublished) to replace the anti-CD3 scFv. The resulting expression vector, pCR3.1αPD-L1-α4-1BB, encodes the oncostatin M signal peptide, an N-terminal FLAG/Strep tag, atezolizumab scFv (VH-VL), a five-residue linker (G₄S), SAP3.28 scFv (VH-VL), and C-terminal myc and polyHis tags. The lentiviral vector pCCL-PD-L1−4-1BB (encoding Oncostatin M signal peptide, FLAG/Strep tags, atezolizumab scFv, G₄S, SAP3.28 scFv, and myc/polyHis tags) was generated by cloning the fragment BamHI/XbaI from pCR3.1αPD-L1-α4-1BB into pCCL-EF1α-LiTE.³⁰ This fragment was also subcloned as MuI/BmgBI into pCCL-CD19-CD3-T2A-tdTomato¹¹ to obtain pCCL-PD-L1−4-1BB-T2A-tdTomato. pCCL-CD19-CD3 (pCCL-EF1a-BiTE19) and pCCL-CD19-OKT3-T2A-GFP (pCCL-EF1a-BiTE19-T2A-GFP) have been previously described.^11,31

Expression and purification of PD-L1 × 4-1BB antibody

HEK−293 cells (4 × 10⁵ cells/well) were transfected with the pCR3.1αPD-L1-α4-1BB vector using the Lipofectamine 3000 kit (Cat. No. L3000008, Fisher Scientific, Waltham, MA, USA) following the manufacturer’s instructions and cultured at 37°C. To generate a stable cell line, transfected cells were selected in DMEM complete medium supplemented with 500 μg/mL of G418 (Cat. No. 10131035, Invitrogen Life Technologies, Carlsbad, CA, USA).

For recombinant antibody purification, conditioned media collected from the stable cell line was centrifuged, filtered through 0.22 μm (Cat. No. 431097, Corning, New York, US) and purified using a Strep-Tactin® purification system (Cat. No. 2−1235−001 IBA Life Sciences, Göttingen, Germany) in an ÄKTAprime plus system (Cat. No. 1100131313, GE Healthcare Life Science, Chicago, IL, USA). The resulting fractions were then dialyzed at 4°C with PBS (pH 7.4) supplemented with 150 mM NaCl and concentrated using Vivaspin 20 centrifugal concentrator with a MWCO of 10 kDa (Cat. No. VS2001 Sartorius, Gotinga, Germany). All the fractions were analyzed by sodium dodecyl sulfate (SDS)–polyacrylamide gel electrophoresis (PAGE) under reducing conditions.

Mass spectrometry

Mass spectrometry analysis was performed as previously described.^15,32 Briefly, a 2 μL aliquot of purified recombinant antibody was desalted using ZipTip® C4 microcolumns (Cat. No. ZTC04S008; Merck Life Science) and eluted with 0.5 μL of a sinapinic acid (SA) matrix (10 g/L in a [70:30] acetonitrile:0.1% trifluoroacetic acid solution) onto a groundsteel-massive 384 target plate (Cat. No. 8280784; Bruker Daltonics, Billerica, MA, USA). Analysis was performed using an Autoflex III MALDI-TOF/TOF spectrometer (Bruker Daltonics) in linear mode with the following parameters: a mass range of 5,000–40,000 Th, linear positive mode, ion source 1 at 20 kV, ion source 2 at 18.5 kV, lens voltage at 9 kV, pulsed ion extraction set to 120 ns, and high gating ion suppression up to 1,000 m/z. External mass calibration was conducted using a protein 1 standard calibration mixture (Cat. No. 8206355, Bruker Daltonics). Data acquisition, peak processing, and spectral analysis were carried out using FlexControl 3.0 and FlexAnalysis 3.0 software (Bruker Daltonics).

Size exclusion chromatography-multiangle light scattering (SEC-MALS)

The SEC-MALS study was conducted as previously described.^15,32,33 In summary, the experiments were performed on a Superdex 200 Increase 10/300 GL column (Cat. No. 28990944, GE Healthcare Life Science) attached in-line to a DAWN EOS light scattering photometer and an Optilab rEX differential refractometer (Wyatt Technology, Santa Barbara, CA, USA). The chromatography was run at room temperature, and the two detectors were thermostatted at 23 °C. The column was equilibrated with PBS (pH 8.0 + 150  mM NaCl) and 0.1 μm filtered, and the system was calibrated with BSA at 2 g/L in the same buffer. The PD-L1 × 4-1BB protein sample was centrifuged for 5 minutes at 14000 rpm, and 100 μL of the solution at 0.1 g/L was injected into the column at a flow rate of 0.5 mL/minute. The column had an exclusion volume of 8.6 mL, and no absorbance (no aggregated protein) was observed in the chromatogram at this volume. Data acquisition and analysis were conducted using ASTRA software (Wyatt Technology). The reported molar mass corresponds to the center of the chromatography peak. Based on multiple measurements of BSA samples at 1 g/L under similar conditions, we calculate the experimental error in molar mass to be approximately 5%. Due to the low concentration of the PD-L1 × 4-1BB sample used, the error in this case may be higher.

Circular dichroism

A circular dichroism (CD) assay was performed as previously described.^15,32 In summary, the measurements were conducted with a Jasco J−810 spectropolarimeter (JASCO, Tokyo, Japan). The spectrum was obtained on a sample at 0.08 g/L in PBS (pH 7.4) and 150 mM NaCl using a 0.2 cm path length quartz cuvette at 25 °C. Thermal denaturation curves were recorded on the same protein sample and cuvette by increasing the temperature from 5 to 95 °C at a rate of 1°C/minute and measuring the change in ellipticity at 210 nm every °C with a 32 s response and 4 nm bandwidth. The CD data were processed with the program Origin (OriginLab, MA, USA). We estimate that the uncertainty in the molar ellipticity is approximately 5% and that the uncertainty in the midpoint denaturation temperature is 0.5 °C.

Serum stability

The serum stability experiment was conducted as previously described.³³ In summary, purified PD-L1 × 4-1BB was incubated in 60% (v/v) human serum (Cat. No. H4522, Sigma-Aldrich) at 37 °C for 96 hours. The binding activity of the sample at time point 0 was set as 100% to calculate the corresponding decay in PD-L1 and 4-1BB binding by ELISA, as described below. The samples were analyzed with a Multiskan FC Photometer and GraphPad Prism software. The results correspond to one experiment performed in duplicate.

PD−1/PD-L1 blockade bioassay

The PD−1/PD-L1 blockade bioassay was performed as previously described³³ and following the manufacturer’s instructions (Cat. No. J1259, Promega). Briefly, 2.5 × 10⁴ CHO^APC/PD-L1 cells/well were seeded in 96-well white plates in complete DMEM and incubated overnight at 37°C. The medium was subsequently removed, and different amounts of HER2 IgG, PD-L1 IgG, and PD-L1 × 4-1BB proteins (final concentrations of 400, 66.7, 6.67, 0.667, 0.0667, and 0.00667 nM) were added to 40 μl of RPMI 1% FBS/well. Then, 1.25 × 10⁵ Jurkat^PD-1 cells/well were added to 40 μl of RPMI 1% FBS/well and incubated for 6 hours at 37°C. Then, 80 μL of BioGlo Reagent (Cat. No. G7941, Promega) was added, and bioluminescence, as an indicator of activation, was measured in a Tecan Infinite F200 fluorescence microplate reader (Tecan Life Sciences, Männedorf, Switzerland).

Antigen-dependent jurkat 4-1BB activation assay

The 4-1BB bioassay was performed as previously described²⁹ and following the manufacturer’s instructions (Cat. No. J2332, Promega). Briefly, 96-well white plates were precoated with BSA or 5 μg/mL hPD-L1 (Cat. No. 156-B7−100, R&Amp D Systems, Minneapolis, MN, USA) overnight at 4°C. Then, the medium was removed, and different amounts of 4-1BB IgG or PD-L1 × 4-1BB (final concentrations of 400, 66.7, 6.67, 0.667, 0.0667, and 0.00667 nM) were added in 30 μL of RPMI 1% FBS. Then, Jurkat^4-1BB cells (1.25 × 10⁵ cells/well) were added in RPMI 1% FBS (40 μL/well) and incubated for 6 h at 37°C. Finally, 80 μL/well of BioGlo Reagent was added, and bioluminescence was measured with a Tecan Infinite F200 Fluorescence Microplate Reader (Tecan Life Sciences).

Lentivirus production and titration

Lentiviral particles encoding PD-L1 × 4-1BB, PD-L1 × 4-1BB-dTo, CD19 × CD3 and CD19 × CD3-GFP were generated by transfection of HEK293T cells with the corresponding transfer vectors together with the packaging plasmids pMDLg/pRRE and pRSVrev, and the envelope plasmid pMD2. G VSV (all from Plasmid Factory, Bielefeld, Germany) using linear polyethyleneimine (Cat. No. 23966−1, Polysciences, Warrington, PA, USA). After 48 hours, the supernatants containing the viral particles were collected and ultracentrifuged at 26,000 rpm for 2 hours at 4°C. The pellets were subsequently resuspended in RPMI 1640 and stored at −80°C until use.

The functional titers (TU/mL) of PD-L1 × 4-1BB- and CD19 × CD3-encoding lentiviruses were determined by flow cytometry analysis after limiting dilution in Jurkat cells and performing intracellular poly-His tag detection, as described below. Titers of lentivirus encoding PD-L1 × 4-1BB-dTo and CD19 × CD3-GFP were determined by flow cytometry by analyzing tdTo or GFP expression, respectively.

T-cell transduction and culture conditions

Peripheral blood mononuclear cells (PBMCs) were isolated from the peripheral blood of healthy volunteer donors by density-gradient centrifugation using Lymphoprep (Cat. No. AXS−1114544; Axis-Shield, Oslo, Norway). All donors provided written informed consent in accordance with the Declaration of Helsinki. PBMCs were activated with plate-coated anti-CD3 (1 μg/mL; clone OKT3, Cat. No. 566685 BD Biosciences, Franklin Lakes, NY, USA) and anti-CD28 (1 μg/mL; clone CD28.2, Cat. No. 555725, BD Biosciences) mAbs for 2 days. T cells were subsequently transduced with lentiviral particles encoding CD19 × CD3, PD-L1 × 4-1BB or a mixture of both at the indicated multiplicity of infection (MOI) in the presence of interleukin−7 (IL−7) (10 ng/mL; Cat. No. 30−095−367 Miltenyi Biotec, Bergisch Gladbach, Germany) or IL−15 (10 ng/mL; Cat. No. 30−095−764 Miltenyi Biotec) to generate STAb-T19, STAb-TP4 or STAb-T19-P4 cells, respectively. T cells were expanded in RCM supplemented with IL−7 and IL−15 for 5−7 days before the experiments were performed. Nontransduced T cells (NT-T) were expanded under the same culture conditions used as negative controls.

Enzyme-Linked Immunosorbent Assay (ELISA)

Human CD19-Fc (5 μg/mL) (Cat. No. 9269-CD−050, R&D Systems), human PD-L1-Fc (2.5 μg/mL) (Cat. No. 156-B7−100, R&D Systems), or human 4-1BB-Fc (2.5 μg/mL) (Cat. No. 838-4B−100, R&D Systems) chimeras were immobilized on Maxisorp 96-well plates (Cat. No. M9410-1CS, NUNC, Roskilde, Denmark) overnight at 4°C. After washing and blocking, purified protein solution (1 μg/mL) or conditioned medium from transduced T cell cultures was added and incubated for 1 hour at room temperature. The wells were subsequently washed 3 times with PBS containing 0.05% Tween 20 (Cat. No. P1379, Sigma-Aldrich) and 3 times with PBS. For CD19 × CD3 bsAb detection, an anti-His mAb (Cat. No. 34660, Qiagen, Venlo, Netherlands) was used (1 μg/mL), and after washing, horseradish peroxidase (HRP)-conjugated goat anti-mouse (GAM) IgG, Fc specific (Cat. No. 115−035−166, Jackson ImmunoResearch, Baltimore, PA, USA) (400 ng/mL), was added. For PD-L1 × 4-1BB bsAb detection, HRP-conjugated anti-FLAG (M2 clone, Cat. No. A8592, Sigma-Aldrich) was used (1 μg/mL). Finally, the plate was developed with 3,3,5,5-tetramethylbenzidine (TMB) (Cat. No. T0440, Sigma-Aldrich), and the reaction was stopped with 1 N H₂SO₄. The absorbance was read at 450–620 nm using Multiskan FC photometer (Thermo Fisher Scientific).

Western blotting

Protein samples and conditioned media from transduced T cell cultures were separated under reducing conditions on 10−20% Tris-glycine gels (Cat. No. XP10202BOX, Life Technologies), transferred onto nitrocellulose membranes (Cat. No. #IB23002, Thermo Fisher Scientific), and probed with anti-FLAG IgG₁ (clone M2, Cat. No. #3165, Sigma-Aldrich) (200 ng/mL) or anti-G₄S linker IgG (clone E702V, Cat. No. 71645, Cell Signaling Technology, Danvers, MA, USA) (110 ng/mL), followed by incubation with HRP-conjugated GAM IgG, Fc specific (1:10,000 dilution) (Cat. No. #A2554, Sigma-Aldrich) or HRP-conjugated goat anti-rabbit (GAR) IgG (Cat. No. 31460, Invitrogen;40 ng/mL). Visualization of protein bands was performed with Pierce ECL Western Blotting substrate (Cat. No. 32132; Thermo Fisher Scientific), using ChemiDoc MP Imaging System and Image Lab software (both from Bio-Rad, Hercules, CA, USA).

Cytotoxicity assays

Cytotoxicity was assessed as previously described.³⁰ Briefly, 5 × 10⁴ luciferase-expressing CD19-positive (Nalm6^PD-L1/Luc) or CD19-negative (K562^Luc) tumor target cells were cocultured with STAb-TP4, STAb-T19 or mixtures of both cell types, as indicated, maintaining a constant 1:1 effector-to-target (E:T) ratio. When necessary, NT-T cells were added to obtain 5 × 10⁴ total T cells. After 48 hours, 20 μg/mL D-luciferin (Cat. No. E1602, Promega) was added before bioluminescence quantification using a Victor® luminometer (PerkinElmer, Waltham, MA, USA). The percent-specific cytotoxicity was calculated using the following formula: 100– [(bioluminescence of each sample/mean bioluminescence of NT-target cells) × 100]. A lysis control was established by adding 5% Triton X−100 to target cells.

Multiplex bead-based immunoassays

Cytokine levels in culture supernatants were measured in multiplex bead-based immunoassays with a ProcartaPlex Cytokine 10-Plex Human Panel (Cat. No. LHC0001M, Thermo Fisher), following the manufacturer's  instructions, and was developed using LABScan3D multiplex flow analyzer (One lambda, Canoga Park, LA, USA).

Flow cytometry

The antibodies used for flow cytometry analysis are detailed in Table S1. DAPI (Cat. No. D9542−324 10MG, Sigma-Aldrich) or 7-aminoactinomycin D (7-AAD) (Cat. No. 559925, BD Biosciences) were used as viability markers. Cell surface-bound CD19 × CD3 or PD-L1 × 4-1BB bsAbs were detected with an APC-conjugated anti-His mAb. Intracellular bsAbs were detected using the Inside Stain Kit (Cat. No. 130−090−477; Miltenyi Biotec), following the manufacturer’s instructions, and an anti-His APC mAb. Alternatively, CD19 × CD3 and PD-L1 × 4-1BB expression were estimated based on GFP and dTo expression, respectively. Cell acquisition was performed with a BD FACSCAnto II flow cytometer using BD FACSDiva software (both from BD Biosciences) or with a Dx-Flex flow cytometer (Beckman Coulter, Brea, CA, USA). Analysis was performed using FlowJo V10 software (Tree Star, Ashland, OR, USA).

In vivo B-ALL xenograft model

Animal procedures adhered to European Union Directive 86/609/EEC and Recommendation 2007/526/EC, which are enforced in Spanish law under RD 1201/2005. All animal experiments were approved by the respective Ethics Committee of Animal Experimentation of the Instituto de Investigación Hospital 12 de Octubre; they were performed in accordance with the guidelines stated in the International Guiding Principles for Biomedical Research Involving Animals, established by the Council for International Organizations of Medical Sciences. The experimental study protocols were additionally approved by the local government (PROEX 166/19).

Nine-week-old female NSG mice (NOD. Cg-Prkdc^scid-IL2rg^tm1Wjl/SzJ; Charles River Laboratories, Wilmington, MA, USA) were infused intravenously (i.v.) with 1 × 10⁶ Nalm6^PD-L1/Luc cells, and after 2 days received i.v 4 × 10⁶ T cells (STAb-T19 group: 2 × 10⁶ STAb-T19 cells + 2 × 10⁶ NT-T cells; STAb-TP4 group: 2 × 10⁶ STAb-TP4 cells + 2 × 10⁶ NT-T cells; STAb-T19-P4 group: 4 × 10⁶ STAb-T19-P4 cells). Tumor growth was evaluated weekly by bioluminescence imaging at the indicated time points. Briefly, 125  mg/kg of D-Luciferin (Cat. No. E1605, Promega) was administered intraperitoneally in 200 μL of sterile PBS. Animals were imaged 7 minutes after D-Luciferin injection using the Bruker In Vivo Xtreme II System (Bruker Corporation). The photon flux emitted by the luciferase-expressing cells was measured as the average radiance (photons/sec/mm²). Imaging analysis was performed using the Bruker Molecular Imaging Software (Bruker Corporation). The mice were euthanized on day 21. Tumor burden and T-cell persistence were analyzed by flow cytometry in peripheral blood (PB), which was obtained by tail vein puncture, at day 14, and in PB, spleen and bone marrow (BM) samples after euthanasia. Body weight was monitored over time. Animals showing endpoint clinical disease or signs of graft-versus-host disease were euthanized.

Statistical analysis

All plots and statistical tests were performed using GraphPad Prism 8.0 (GraphPad Software, La Jolla, California, USA). In general, the in vitro assays were performed in triplicate, and the values are presented as the means ± SD. Significant differences (P values) were determined using a two-tailed, unpaired Student’s t test, which assumes a normal distribution, or one-way analysis of variance (ANOVA), which was adjusted by Dunnett’s test for multiple comparisons, as indicated. Two-way ANOVA was used to analyze experiments that evaluated the interaction of two variables following multiple comparison testing using either Dunnett’s test or Tukey’s test, as appropriate. Significance is defined as follows: *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001.

Results

Generation and characterization of a PD-L1 × 4-1BB bispecific antibody

We developed a novel STAb-T immunotherapy based on the simultaneous secretion of CD19 × CD3 and PD-L1 × 4-1BB tandem scFv (ta-scFv) bsAbs by engineered T cells. The CD19 × CD3 bsAb, consisting of CD19 and CD3 scFvs fused by a glycine-serine-based linker (G₄S), has been extensively characterized by our group^11,34 (Figure 1A,B). The PD-L1 × 4-1BB bsAb is a ta-scFv made by joining a PD-L1 scFv to a 4-1BB scFv through a G₄S module. N-terminal FLAG/Strep-II and C-terminal myc/6xHis tags were incorporated for purification and detection (Figure 1C,D). In addition, constructs incorporating the reporter genes GFP and tdTo to CD19 × CD3 or PD-L1 × 4-1BB, respectively, were generated to facilitate the detection of transduced T cells (Figure 1A–D).

Figure 1.

Characterization of the novel PD-L1 × 4-1BB bsAb. (A, B) Schematic diagrams showing the genetic (A) and domain (B) structures of the anti-CD19 × anti-CD3 bsAb, which bears a signal peptide from the human K light chain (grey box), the anti-CD19 scFv (A3B1) gene (blue boxes), the anti-CD3 (OKT3) scFv gene (grey boxes) and the His tag (yellow box). (C, D) Schematic diagrams showing the genetic (C) and domain (D) structures of the anti-PD-L1 × anti−4-1BB bsAb, which bears the oncostatin M signal peptide (grey box), the anti-PD-L1 scFv (atezolizumab) gene (red boxes) and the anti−4-1BB (SAP3.28) scFv gene (brown boxes). N-terminal FLAG-Strep and C-terminal Myc-His tags (yellow boxes) were appended for purification and detection purposes. (E–I) HEK293 cells were stably transfected with a plasmid encoding the PD-L1 × 4-1BB bsAb. The recombinant antibody was purified from the conditioned medium by affinity chromatography using Strep-Tactin® columns and subjected to characterization assays. (E) Western blot detection of the PD-L1 × 4-1BB bsAb purified from the supernatant of stably transfected HEK293T cells. Conditioned media from nontransfected HEK293T cells (NTf) were used as a negative control. One representative of three experiments is shown. (F) Detection of purified PD-L1 × 4-1BB bsAb by ELISA against a plastic-immobilized human PD-L1-Fc chimera (hPD-L1-Fc), a human 4-1BB-Fc chimera (h4-1BB-Fc) or BSA. PBS was used as a negative control. The data are presented as the mean ± SD (n = 3). (G) The functional binding of purified PD-L1 × 4-1BB (5 μg/mL) to human PD-L1 and 4-1BB expressed on the cell surface of PD-L1⁺4-1BB^- Nalm6^PD-L1 and PD-L1^-4-1BB⁺ Jurkat^{4-1BB/PD-L1-KO} cells, respectively, was demonstrated by flow cytometry using an APC-conjugated anti-His mAb. PD-L1 and 4-1BB IgGs (5 μg/mL) served as positive controls and were detected with a PE-conjugated antihuman Fc antibody. (H) For PD−1/PD-L1 blockade bioassays, Jurkat^PD-1 cells were cocultured with CHO^APC/PD-L1 in the presence of 10-fold increasing concentrations of PD-L1 × 4-1BB bsAb. After 6 hours at 37 °C, luminescence was determined. The Y-axis represents the reporter gene fold induction relative to the values obtained from Jurkat^PD-1 cells alone. HER2 IgG was used as a negative control, and PD-L1 IgG was used as a positive control. The results are expressed as mean ± SD (n = 3). Significance was calculated by an unpaired Student’s t test. (I) Antigen-dependent Jurkat 4-1BB activation assay with 10-fold increasing concentrations of PD-L1 × 4-1BB protein against plastic-immobilized human PD-L1-Fc (hPD-L1) or BSA. 4-1BB IgG was used as a negative control. The data are presented as the fold induction relative to the values obtained from unstimulated Jurkat^4-1BB cells. The results are expressed as mean ± SD (n = 3). Significance was determined by an unpaired Student’s t test. Significance is defined as follows: *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001.

The PD-L1 × 4-1BB antibody was purified from the conditioned medium of stably transfected HEK293 cells by affinity chromatography. A single band with the expected molecular weight (57 kDa) was identified in reducing SDS‒PAGE (Figure S1A) and Western blotting (Figure 1E). Mass spectrometry revealed a molar mass of 57 kDa, indicating that the signal sequence was absent in the secreted protein (data not shown). SEC-MALS analysis confirmed that the protein eluted as a major symmetric peak (Figure S1B), with a molar mass close to the theoretical value calculated for a monomer (57 kDa). A minor peak eluted at a smaller volume, but the scattering signal was too low to determine its molar mass (Figure S1B). The circular dichroism (CD) spectrum indicates a mixed α-helix and β-sheet secondary structure (Figure S1C), and a folded stable tertiary structure is shown by cooperative thermal denaturation (with a midpoint of ~ 59 °C; Figure S1D). Denaturation is irreversible, as protein precipitation occurs inside the cuvette. The PD-L1 x 4-1BB molecule showed high stability in human serum at 37°C, retaining binding activities of almost 100% and 60% for 4-1BB and PD-L1, respectively, after 96 hours (Figure S1E, F). The binding of the purified protein to plastic-immobilized PD-L1 and 4-1BB was demonstrated by ELISA (Figure 1F). In addition, the binding of the PD-L1 × 4-1BB bsAb to its cognate antigens was assessed in a cellular context, using Nalm6^PD-L1 (4-1BB^-PD-L1⁺) and Jurkat^{4-1BB/PD-L1-KO} (4-1BB⁺PD-L1^-) as antigen-positive cells, and Nalm6 (4-1BB^-PD-L1^-) as a negative control (Figures 1G and S2). Monospecific PD-L1 and 4-1BB IgG antibodies were included in the analysis as positive controls for antibody binding (Figure 1G).

To demonstrate the ability of PD-L1 × 4-1BB to block the PD−1/PD-L1 interaction, Jurkat^PD-1 T cells expressing human PD−1 and a luciferase reporter driven by an NFAT response element were cocultured with CHO^APC/PD-L1 cells, with the ability to activate Jurkat^PD-1 cells only when the PD−1/PD-L1 interaction was blocked. The binding of the PD-L1 × 4-1BB bsAb to PD-L1 blocked the PD−1/PD-L1 interaction in a dose-dependent manner, as shown by the significant increase in luciferase activity (Figure 1H). The PD-L1 IgG mAb, which was used as a positive control, significantly increased luciferase activity as well, whereas no induction was observed in the presence of the HER2-targeting IgG mAb (Figure 1H).

The ability of the PD-L1 × 4-1BB bsAb to activate 4-1BB was subsequently determined using Jurkat^4-1BB cells, which are characterized by constitutive h4-1BB expression and by luciferase expression dependent on 4-1BB crosslinking. PD-L1 x 4-1BB-mediated 4-1BB signaling due to crosslinking with plate-bound PD-L1 significantly induced luciferase activity, whereas no induction was observed in the absence of PD-L1 or when 4-1BB IgG was used (Figure 1I).

STAb-T cells efficiently secrete CD19 x CD3 and PD-L1 x 4-1BB bsAbs, which synergize to induce antitumor specific cytotoxicity

Primary T cells were transduced with lentiviral vectors encoding 19– or P4-bsAb at an MOI of 10 to generate STAb-T19 or STAb-TP4 cells. Both bsAbs were secreted with expected molecular weights of 55 and 57  kDa, respectively (Figure 2A,B), and specifically recognized plastic-immobilized human CD19 (Figure 2C) or human PD-L1 and 4-1BB (Figure 2D), respectively. Analysis of intracellular expression revealed comparable transduction efficiencies (Figure 2E), ranging from 10−30% (Figure S3A), whereas decoration by the secreted bsAbs bound to the cell surface was notably higher in STAb-TP4 cells (Figures 2E and S3B). This could be due to the strong binding of the P4-bsAb to PD-L1 expressed on activated T cells, as suggested by the sharp decrease in PD-L1 detection in STAb-TP4 cells compared with that in NT-T and STAb-T19 cells (Figure S3C). To assess whether PD-L1 x 4-1BB could potentiate the cytotoxic ability of 19-bsAb, 5 × 10⁴ luciferase (Luc)-expressing CD19⁺PD-L1⁺ Nalm6^PD-L1/Luc or CD19^-PD-L1^- K562^Luc target cells were co-cultured for 48 hours with decreasing numbers of STAb-T19 cells (2.5 × 10⁴, 2.5 × 10³, 2.5 × 10² and 2.5 × 10¹), in the presence or in the absence of a constant number of STAb-TP4 (2.5 × 10⁴). NT-T cells were added as needed to achieve a total T cell:target cell ratio of 1:1. The combination of both STAb-T populations resulted in more potent cytotoxic activity against CD19⁺PD-L1⁺ Nalm6^PD-L1/Luc cells than did STAb-T19 alone (Figure 2F). STAb-TP4 cells, which exert a cytotoxic effect comparable to that of NT-T cells (Figure S4A), barely affected target cell viability in the absence of STAb-T19 (Figure 2F). No cytotoxicity was observed when CD19^-PD-L1^- cells (K562^Luc) were used as targets (Figure 2F). Interestingly, when cocultures with decreasing E:T ratios were extended over 9 days, low doses of STAb-T19 cells (1:4 and 1:8 T cell:target ratios) did not prevent the escape of Nalm6^PD-L1 leukemic cells, whereas the addition of STAb-TP4 cells, which alone had effects comparable to those of NT-T cells under similar conditions (Figure S4B), controlled tumor growth over time at these ratios (Figures 2G, S5). Notably, at high E:T ratios and despite their limited cytotoxic capacity (Figure 2F), both NT-T and STAb-TP4 T cells proliferated robustly and maintained a numerical advantage in culture even in the absence of specific antigen recognition, likely due to strong prior activation during in vitro transduction and expansion, as well as to donor-dependent alloreactivity toward Nalm6 cells (Figures 2G and S4B).

Figure 2.

STAb-T19 and STAb-TP4 cells efficiently secrete CD19 × CD3 and PD-L1 × 4-1BB, respectively, and synergize to induce potent specific cytotoxicity. (A, B) Western blot detection of secreted CD19 × CD3 (A) or PD-L1 × 4-1BB (B) in the conditioned media from lentivirally transduced human primary T cells (STAb-T19 or STAb-TP4, respectively). Conditioned media from nontransduced T cells (NT-T) was used as a negative control. (C, D) Detection of soluble functional CD19 x CD3 (C) or PD-L1 x 4-1BB (D) in the conditioned media from STAb-T19 or STAb-TP4 by ELISA against plastic-immobilized human CD19-Fc chimera (hCD19-Fc), human PD-L1-Fc chimera (hPD-L1-Fc), human 4-1BB-Fc chimera (h4-1BB-Fc) or BSA. Conditioned media from NT-T cells was used as a negative control. The data are presented as mean ± SD (n = 3). (E) Representative analysis of intracellular and cell surface-bound (decoration) CD19 × CD3 and PD-L1 × 4-1BB in NT-T, STAb-T19 and STAb-TP4 cells by flow cytometry. One representative experiment out of four independent experiments is shown. The numbers represent the percentage of cells that stained positive for the His tag. (F) Cell cytotoxicity assay. Decreasing numbers of STAb-T19 cells were cocultured with 5 × 10⁴ CD19⁺PD-L1⁺ (Nalm6^PD-L1/Luc) or CD19^-PD-L1^-(K562^Luc) luciferase-expressing target cells in the presence or absence of a constant number of STAb-TP4 cells. NT-T cells were added as indicated to maintain a constant 1:1 effector:target ratio. Cocultures of target cells with STAb-TP4 cells were used as controls. The percentage of cytotoxicity was calculated after 48 h and normalized to the target cell death observed in cocultures with NT-T cells. The data are shown as mean ± SD (n = 3). Significance was determined by two-way ANOVA with Sidak’s multiple comparisons test. Significance is defined as follows: *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001. (G) Nalm6^PD-L1 cells were cocultured with 1:1 mixtures of STAb-TP4 + NT-T, STAb-T19 + NT-T or STAb-T19 + STAb-TP4 cells at the indicated E:T ratios, and the relative percentages of T cells (CD2⁺CD10^-) and leukemic cells (CD2^-CD10⁺) were analyzed by flow cytometry after 0, 2, 5, and 9 days.

PD-L1 × 4-1BB promotes persistence of STAb-T19 cells and prevents leukemia escape in vitro

Next, we delved into the ability of STAb-TP4 cells to prolong the antitumor activity of STAb-T19 cells over time. For this purpose, we performed long-term (>20 days) cocultures of STAb-T19 and STAb-TP4 cells. To quantify the percentage of bsAb-secreting cells, T lymphocytes were transduced with reporter gene constructs (Figure 1A,C). The transduction efficiency was determined based on the percentage of GFP- or tdTo-reporter protein expression by STAb-T19 or STAb-TP4 cells, respectively (Figure 3A). No differences were observed in the CD4:CD8 ratio between STAb-T19⁺ and STAb-TP4⁺ cells (Figure 3B), with comparable proportions of naïve, central memory, and effector memory subsets among NT-T, STAb-T19 and STAb-TP4 cells (Figure 3C). However, a trend toward a higher CD4:CD8 ratio was observed among bsAb-secreting cells (STAb-T19⁺ and STAb-TP4⁺) compared to non-secreting cells within the transduced populations (Figure 3B). Cocultures of tumor cells with T cells at different ratios revealed that only the combination of STAb-T19 and STAb-TP4 cells fully controlled tumor growth beyond day 24 at the lowest E:T ratio (Figures 3D, S6). Moreover, when co-cultures were rechallenged with tumor cells on day 16 to mimic relapse, only the combination fully eliminated tumor cells at a 1:2 ratio (Figure 3D).

Figure 3.

PD-L1 × 4-1BB potentiates and prolongs the antitumor effect of CD19 × CD3. Long-term cocultures of Nalm6^PD-L1 target cells with reporter gene-expressing STAb-T cells were generated. (A) Percentage of reporter protein expression in primary STAb-T19 or STAb-TP4 cells (GFP or tdTo, respectively). (B, C) Percentages of CD4⁺ and CD8⁺ T cells (B) and percentages of naïve (TN), central memory (TCM), effector memory (TEM) and effector memory RA⁺ (TEMRA) T cells (C) among nontransduced (NT-T), STAb-T19-transduced and STAb-TP4-transduced T cells. The transduced populations were further classified into bsAb-secreting (+) or nonsecreting (−) cells on the basis of reporter gene expression. The data are shown as mean ± SD (n = 3). (D) Nalm6^PD-L1 target cells were cocultured with STAb-TP4 + NT-T or STAb-T19 + NT-T or STAb-T19 + STAb-TP4 cells at the indicated E:T ratios, and the relative percentages of CD2⁺CD10^- and CD2^-CD10⁺ cells were analyzed by flow cytometry at days 4, 12, 20, and 28. On day 16, the cocultures were split into two wells, and 5,000 target cells were added to one of them; the relative percentages of T cells and tumor cells were analyzed at days 20 and 28. One of two similar experiments is shown.

To elucidate the mechanisms underlying the enhanced persistence and efficacy of the STAb-T19 and STAb-TP4 cell combination, we analyzed the expression of exhaustion markers at different time points, along with the secretion of key cytokines relevant to T-cell function and those involved in, or predictive of, cytokine release syndrome (CRS). Overall, the percentages of T cells expressing LAG−3 and TIM−3 were high across all groups, including NT-T cells, likely due to the activation stimulus used during expansion and transduction (Figure S7A, B). LAG−3 expression remained relatively stable over time, whereas TIM−3 expression decreased, particularly at 1:1 E:T ratios. As shown, NT-T and STAb-TP4 cells exhibited lower and comparable expression levels, while STAb-T19 cells and the combination showed higher percentages and more similar between them (Figure S7A, B). PD−1 expression decreased over time, with no significant differences observed among the groups (Figure S7C). Thus, the observed differences in persistence and efficacy between STAb-T19 cells and the STAb-T19 + STAb-TP4 combination cannot be attributed to the differential expression of exhaustion markers. With respect to cytokine secretion, STAb-T19 cells produced significantly higher levels of IL-1β, IL−2, and IL−4 (at E:T ratios of 1:1 and 1:8), as well as IL−10 (at an E:T ratio of 1:8), than did the STAb-T19 + STAb-TP4 combination (Figure S8). IL−6 did not significantly differ between these two groups, whereas IFN-γ, GM-CSF, and TNF-α tended to be secreted at higher levels in the combination cocultures. The significant differences between STAb-T19 or the combination and NT-T or STAb-TP4 alone are summarized in Table S2.

From a clinical perspective, manufacturing a single STAb-T cell product instead of two would be more cost-effective and would streamline the process. Therefore, we compared the antileukemic effect of STAb-T19 cells with that of T cells simultaneously transduced with lentiviral vectors encoding 19-bsAb and P4-bsAb to generate STAb-T19-P4 cells. To avoid differences due to the dose-dependent cytopathic effect of viral infection, we equalized the total MOIs in all the transductions. Thus, T cells were transduced with 19-bsAb (MOI of 15), P4-bsAb (MOI of 15) or both 19-bsAb (MOI of 8) and P4-bsAb (MOI of 7) lentiviral particles. The transduction efficiencies of the different T cell populations generated are shown in Figure 4A. No differences were observed in the CD4:CD8 ratio (Figure 4B) or in the proportions of naïve, central memory and effector memory subsets (Figure 4C) between the distinct STAb-T populations. Interestingly, a trend towards an increased CD4:CD8 ratio among bsAb-secreting cells (STAb-T19⁺ and STAb-TP4⁺) compared with non-secreting cells was again observed in the transduced populations. Singly transduced T cell populations were diluted with nontransduced T cells to achieve reporter gene expression levels similar to those of doubly transduced cells, and long-term cocultures with leukemic Nalm6^PD-L1 cells were established. STAb-T19-P4 cells showed a more potent response, both in primary cultures and after rechallenge with tumor cells, confirming that the cotransduced population enhances the efficiency of STAb-T19 cells (Figure 4D).

Figure 4.

Simultaneous secretion of CD19 × CD3 and PD-L1 × 4-1BB by STAb-T19-P4 cells improves the antitumor efficacy of STAb-T19 cells in vitro. (A) Percentage of reporter protein expression in primary STAb-T19, STAb-TP4, and STAb-T19-P4 cells (GFP or tdTo). (B, C) Percentages of CD4⁺ and CD8⁺ T cells (B) and percentages of naïve (TN), central memory (TCM), effector memory (TEM) and effector memory RA⁺ (TEMRA) T cells (C) among nontransduced (NT-T), STAb-T19-, STAb-TP4-, or STAb-T19-P4-transduced T cells. The transduced populations were further classified into bsAb-secreting (+) or nonsecreting (−) cells on the basis of reporter gene expression. Data are shown as mean ± SD (n = 3). (D) Nalm6^PD-L1 target cells were cocultured with STAb-TP4, STAb-T19, or STAb-T19-P4 cells at the indicated E:T ratios, and the relative percentages of CD2⁺CD10^- and CD2^-CD10⁺ cells were analyzed by flow cytometry at days 4, 12, 20, and 24. On day 16, the cocultures were split into two wells, and 5,000 target cells were added to one of them; the relative percentages of T cells and tumor cells were analyzed on days 20 and 24.

Secretion of PD-L1 × 4-1BB bsAb enhances the antitumor efficacy of STAb-T19 in vivo

To evaluate the antitumor effects of the combination strategy in vivo, immunodeficient NSG mice were injected i.v. with 1 × 10⁶ Nalm6^PD-L1/Luc leukemic cells. Two days later, mice were divided into three groups and treated with 4 × 10⁶ T cells: 2 × 10⁶ STAb-T19 + 2 × 10⁶ NT-T cells (STAb-T19 group), 2 × 10⁶ STAb-TP4 + 2 × 10⁶ NT-T cells (STAb-TP4 group) or 4 × 10⁶ STAb-T19-P4 co-transduced T cells (STAb-T19-P4 group) (Figure 5A). As shown by bioluminescence (BLI) imaging, leukemia progression was significantly reduced in mice treated with STAb-T19-P4-cotransduced cells compared with that in mice receiving STAb-T19 or STAb-TP4 cells (Figure 5B,C). No decrease in body weight was observed during the treatment in any of the groups (Figure 5D).

Figure 5.

Simultaneous secretion of 19- and P4-bsAbs by STAb-T19-P4 cells improves the antitumor efficacy of STAb-T19 in vivo. Xenograft murine model of B-ALL. NSG mice received 1 × 10⁶ Nalm6^PD-L1/Luc cells intravenously followed 2 days after by intravenous infusion of T cells (4 × 10⁶): STAb-TP4 (2 × 10⁶ STAb-TP4 + 2 × 10⁶ NT-T), STAb-T19 (2 × 10⁶ STAb-T19 + 2 × 10⁶ NT-T), or STAb-T19-P4 cells (4 × 10⁶ STAb-T19-P4). (A) Timeline of the experimental design. (B) Total radiance quantification at the indicated time points. Statistical significance was calculated by two-way ANOVA test corrected a Tukey’s multiple comparison test. Significant differences between STAb-TP4 (red asterisks) or STAb-T19 (blue asterisks) and STAb-T19-P4 are indicated. (C) Bioluminescence images showing disease progression. (D) Changes in body weight over time. (E–H) Percentages of B-ALL cells (CD19⁺) and T cells (CD3⁺) in the peripheral blood (PB) at day 14 (E) and at the time of euthanasia in the PB (F), in bone marrow (BM) (G) and spleen (H), as assessed by flow cytometry. Statistical significance was calculated one-way ANOVA test corrected with a Tukey’s multiple comparison test. Significance is defined as follows: *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001.

Flow cytometry analysis of PB on day 14 revealed significantly enhanced leukemia progression (CD19⁺ cells) in the STAb-TP4-treated mice compared to the STAb-T19- and STAb-T19-P4 groups (Figure 5E). In contrast, T cell expansion was notably higher in mice treated with co-transduced T cells (Figure 5E). Endpoint analysis of PB, BM, and spleen confirmed the BLI results. STAb-T19-P4-treated mice presented a lower percentage of blasts in all organs, with significant differences observed in the BM and spleen (Figure 5F, G, H). With respect to T cell expansion, significant differences in CD3⁺ cell percentages were observed only in the spleen, where T cells expanded more in STAb-T19-P4-treated mice than in those receiving STAb-T19 cells (Figure 5H).

Discussion

The infusion of engineered STAb-T lymphocytes secreting bispecific antibodies that redirect T cells against a TAA⁸ is emerging as a promising strategy for cancer immunotherapy. We and others have demonstrated the superior antitumor effects of STAb-T cells over CAR-T cells in relevant preclinical models^9–11,35 and the expansion of memory STAb-T cells in vivo.⁹ However, CD3-engaging bsAbs, also known as T cell engagers (TCEs), do not provide costimulatory signaling to T cells, and although TCEs recognizing different TAAs have shown potent T cell cytotoxicity in the absence of costimulation,^8–11 concerns may arise regarding the long-term persistence of STAb-T cells in patients. Only the imminent translation of STAb-T therapies to the clinic will shed light on this issue. Nonetheless, we believe that providing 4-1BB costimulatory signals to STAb-T cells would be beneficial. On the other hand, the clinical efficacy of STAb-T cells might be compromised by the expression of inhibitory molecules, such as PD-L1, on tumor cells and other cells within the tumor microenvironment. The potential of both 4-1BB agonistic antibodies and PD−1/PD-L1 blocking antibodies in cancer therapy has been explored, with anti-PD−1 or PD-L1 antibodies achieving remarkable success in certain cancers.^36–39 Furthermore, PD-L1 × 4-1BB bsAbs have shown promising results in clinical trials.^24–26,40

Here, we have developed an evolution of STAb-T therapy by arming STAb-T cells, which secrete a TAA-targeting TCE, with an additional PD-L1 × 4-1BB bsAb. For this purpose, we designed and characterized a novel anti-PD-L1 x anti−4-1BB molecule (P4-bsAb) in the ta-scFv format and combined its in situ secretion by genetically engineered T cells with the simultaneous secretion of a preclinically validated anti-CD19 x anti-CD3 bsAb.¹¹ Purified PD-L1 × 4-1BB effectively induced PD-L1-dependent 4-1BB activation and blockade of the PD−1/PD-L1 interaction. Moreover, functional P4-bsAb was efficiently secreted in situ by lentivirally transduced human primary T cells (STAb-TP4). The ability of 4-1BB activation to enhance tumor-specific T cell-mediated cytotoxicity has been widely described,^15,41,42 and accordingly, P4-bsAb enhanced STAb-T19 cytotoxicity against CD19⁺PD-L1⁺ target cells. In addition, the incorporation of the 4-1BB costimulatory domain into CAR constructs has been shown to enhance T-cell persistence compared with first-generation CARs, which lack a costimulatory domain, and CD28-costimulated second-generation CARs.⁴³ Consistent with this, the P4-bsAb demonstrated the ability to prolong T-cell persistence and the antitumor efficacy of the 19-bsAb at low E:T ratios.

Interestingly, the P4-bsAb did not exhibit antitumor activity in the absence of the 19-bsAb. In clinical trials, PD-L1 × 4-1BB bsAbs have shown evidence of antitumor activity as monotherapies.^25,40 However, their efficacy relies on the reactivation of pre-existing tumor-specific T lymphocytes, whose function is compromised by the PD−1/PD-L1 interaction, insufficient costimulation or both. Our combination approach goes beyond merely blocking PD-L1 and providing costimulation for effective T-cell activation; it can also induce T-lymphocyte cytotoxicity in tumors devoid of tumor-specific T-cells. In addition, the cosecretion of P4-bsAb potentiates and prolongs the antitumor efficacy of STAb-T cells, enabling sustained leukemia growth control at E:T ratios where STAb-T cells alone are inefficient in the long term.

The observed differences in antitumor efficacy and persistence did not appear to be associated with the differential expression of exhaustion markers, which in vitro seemed to be driven by the nonspecific activation and expansion process, with only modest increases in TIM−3 and LAG−3 upon antigen-specific activation. In terms of cytokine profiles, with a focus on those significantly elevated in cocultures with STAb-T19 alone, previous studies have shown that IL-1β production enhances the proliferation, effector function, and memory differentiation of CD8⁺ T cells across various experimental models,⁴⁴ whereas IL−2, while essential for T-cell proliferation, may promote a mature phenotype linked to reduced persistence.⁴⁵ IL−4 drives Th2 differentiation, which has been associated with reduced antitumor activity,⁴⁶ cytotoxic function,⁴⁷ memory T cell formation and long-term persistence.⁴⁸ Finally, although generally antiinflammatory, IL−10 has been shown to enhance CAR-T-cell proliferation, effector function, and stem cell–like memory formation, improving persistence.⁴⁹ Thus, disentangling the overall balance of cytokine effects remains challenging and is further complicated by their pleiotropic and context-dependent nature. Importantly, in vivo effects are likely to be more complex, as cytokines can act on a broader range of immune, tumor, and stromal cell populations. Notably, preclinical models using immunodeficient mice do not fully reproduce the tumor microenvironment. Syngeneic models with murine T cells in immunocompetent mice, where bsAb constructs are designed to recognize murine antigens, could offer additional opportunities for studying solid tumors, with highly immunosuppressive environments, and combination therapies. These models have been successfully employed in preclinical CAR-T studies for both hematological and solid tumors⁵⁰ and may provide valuable insights for future investigations with STAb-T cells. On the other hand, agonistic anti–4-1BB antibodies have been shown to modulate the metabolic reprogramming of T cells,⁵¹ enhancing mitochondrial respiration, which in turn increases the persistence of adoptively transferred T cells and promotes improved tumor control in preclinical models.⁵² In fact, metabolic reprogramming has been suggested to be the dominant effect of 4-1BB therapy.⁵³ Analyzing the metabolic changes induced by the PD-L1x4-1BB molecule to determine its contribution to the increased persistence and efficacy of combination therapy over STAb-T19 alone will be an interesting objective for future studies.

When designing combination approaches, enhancing efficacy while maintaining tolerable toxicity remains a key concern. Liver toxicity is the main safety concern associated with 4-1BB agonistic antibodies. It may result from 4-1BB engagement with liver-resident Kupffer cells, promoting local inflammation and T-cell infiltration, which amplifies the inflammatory cascade and ultimately results in severe hepatotoxicity.⁵⁴ Notably, FcγRIIB-mediated crosslinking, rather than mere epitope binding, has been proposed as the primary mechanism driving urelumab-induced hepatitis.¹⁴ Thus, the absence of an Fc domain eliminates the toxicity associated with Fc receptor crosslinking, as reported in preclinical^22,55,56 and clinical studies.²¹ Moreover, PD-L1–dependent activation of 4-1BB is expected to reduce off-tumor toxicity by localizing 4-1BB signaling to PD-L1–expressing sites, primarily within the tumor microenvironment. In terms of the potential toxicities elicited by STAb-T19 cells, the adverse effects are expected to be similar to those observed with the anti-CD3/CD19 bispecific antibody blinatumomab and with CD19-directed CAR-T-cell therapies, primarily CRS and immune effector cell-associated neurotoxicity syndrome (ICANS). The upcoming first-in-human trials will provide critical insights into the actual toxicity profile. Nevertheless, the expected toxicities are anticipated to be clinically manageable. A further relevant concern is whether combination therapy could exacerbate toxicity in addition to enhancing efficacy. Importantly, we did not observe any signs of toxicity in the preclinical animal model. Moreover, our group recently reported that systemic administration of mRNA encoding an EGFR-targeting TCE and a PD-L1 × 4-1BB costimulatory antibody induced tumor regression without associated toxicity.²² Although that approach results in transient antibody secretion, whereas our strategy promotes sustained secretion over time, these findings offer a reassuring indication of the potential safety profile of combining TCEs with 4-1BB costimulation.

Another advantage of our strategy is that tumor homing relies not only on passive diffusion but also on active T-cell trafficking. Additionally, the small size of the ta-scFv format may result in improved tissue penetration compared with that of clinically tested PD-L1 x 4-1BB bispecific antibodies.^25,40 While its smaller half-life is reduced, this limitation is offset by the sustained in vivo secretion by STAb-T cells.

One remarkable observation in our study was the elevated level of decoration in STAb-TP4 cells compared with that in STAb-T19 cells, despite similar transduction efficiencies. The most plausible explanation is that activated T cells express high levels of PD-L1, providing a high density of anchor points through both 4-1BB and PD-L1. This issue deserves special attention because the expression of PD-L1 on activated T cells and the potential effects of its activation have not been sufficiently addressed. Specifically, the impact of these bsAbs on T cell functionality through the cross-linking of 4-1BB and PD-L1, both of which are expressed on activated T lymphocytes, remains underexplored. PD-L1 delivers intrinsic prosurvival signals to cancer cells.^57,58 However, conflicting results have been reported regarding the effects of PD-L1-delivered signals on T cells. While some studies maintain that PD-L1 engagement on T cells can mitigate T-cell activation⁵⁹ or lead to T-cell apoptosis,⁶⁰ other studies suggest that the PD-L1 interaction might be essential for activated T-cell survival.⁶¹ Finally, a phased effect has also been proposed, in which ligation of PD-L1 on T cells initially costimulates T-cell growth and cytokine secretion but then switches to programmed cell death.⁶² In any case, the overall effect on our system appears to be positive, although further experiments specifically analyzing the impact of crosslinking of PD-L1 expressed on T cells are warranted.

Finally, the in vivo secretion of P4-bsAb delayed leukemia progression in a xenograft murine model of B-ALL compared with the effects of low doses of STAb-T19 alone. In previous studies,¹¹ we demonstrated that different doses of 19-bsAb-secreting cells could completely abrogate leukemia growth. However, in this study, the lowest dose of STAb-T19⁺ cells tested was not sufficient to completely eliminate leukemic cells. Notably, the cosecretion of P4-bsAb exhibited a clear synergistic effect with 19-bsAb, enhancing its ability to inhibit leukemia progression.

Considering the potential clinical translation of our strategy, an important factor is that, as with CAR-T-cell therapies, variations in STAb-T-cell engraftment, expansion, and persistence in vivo may critically impact both efficacy and toxicity. Poor engraftment and expansion could lead to treatment failure, while robust proliferation may enhance tumor clearance but increase the risk of severe toxicities, including CRS and ICANS.⁶³ Persistence is crucial for long-term disease control but may lead to chronic B-cell aplasia or prolonged cytopenia.⁶³ CAR-T-cell therapies incorporating a 4-1BB costimulatory domain, rather than CD28, exhibit slower expansion, delayed CRS onset and prolonged persistence.⁶³ The incorporation of 4-1BB stimulation into STAb-T approaches may similarly enhance persistence. Lymphodepletion prior to T-cell administration, likely based on fludarabine/clophosphamide, is also a key factor for T-cell engraftment and persistence. While it represents a risk factor for CRS, lymphodepletion is critical for CAR-T-cell efficacy since it facilitates engraftment and reprograms the microenvironment.⁶⁴ Optimizing this conditioning to balance efficacy with the risks of immunosuppression and opportunistic infections will be key for clinical translation.⁶⁴

In summary, we demonstrated for the first time the potential of a PD-L1 × 4-1BB bsAb, which is secreted in situ by engineered T cells, to enhance the antitumor activity of a simultaneously secreted TAA-targeting TCE. Further studies are warranted to evaluate this strategy in both hematological and, particularly, solid tumor models, where PD-L1 x 4-1BB may be especially valuable in overcoming the highly immunosuppressive tumor microenvironment, paving the way for its clinical translation as a novel cancer immunotherapy.

Supplementary Material

Supplementary material

Figure S1. Functional characterization of the purified PD-L1 × 4-1BB bsAb.(A) SDS‒PAGE analysis of purified PD-L1 × 4-1BB (three fractions) under reducing conditions. BSA curve (200, 100, 50, 25, and 12.5 μg/mL) were used as standard curve. The migration of molecular mass markers is indicated (kDa). (B) SEC-MALS analysis of the PD-L1 × 4-1BB protein. The black line corresponds to the UV absorbance at 280 nm (left axis), and the red line corresponds to the measured molar mass (right axis). (C) PD-L1 x 4-1BB protein far-UV CD spectra. (D) PD-L1 × 4-1BB bsAb thermal denaturation profile. (E, F) Serum stability of purified PD-L1 × 4-1BB was analyzed by determining its relative binding to immobilized h4-1BB (E) or hPD-L1 (F) after sample incubation in human serum at 37 °C for different time periods. The data are shown as mean ± SD (n=2).Figure S2. Characterization of target cell lines.PD-L1 and 4-1BB cell surface expression in Nalm6, Nalm6^PD-L1, and Jurkat^{4-1BB/PD-L1-KO} cells was analyzed by flow cytometry. The percentages of PD-L1⁺ and 4-1BB⁺ cells are indicated.Figure S3. Characterization of transduced primary T cells. (A, B) Intracellular expression (A) and cell-surface decoration (B) by CD19 × CD3 or PD-L1 × 4-1BB in NT-T, STAb-T19 or STAb-TP4 cells are shown; the results are presented as mean ± SD of at least four different transductions. (C) Detection, by flow cytometry using fluorochrome-conjugated mAbs, of PD-L1 and 4-1BB on the NT-T, STAb-T19, and STAb-TP4 cell surface.Figure S4. Comparative cytotoxic effects of NT-T and STAb-TP4 cells cocultured with CD19⁺PD-L1⁺ Nalm6^PD-L1 target cells. (A) Cytotoxic effect of NT-T or STAb-TP4 cells on Nalm6^PD-L1/Luc cells at a 1:1 effector:target ratio, normalized to target cell death in cultures with Nalm^6PD-L1/Luc cells alone. The data are shown as mean ± SD (n=9). (B) Nalm6^PD-L1 cells were cocultured with NT-T or STAb-TP4 cells at the indicated E:T ratios, and the relative percentages of T cells (CD2⁺) and target cells (CD10⁺) were measured by flow cytometry after 0, 4, 12, and 28 days.Figure S5. Leukemia escape from immune pressure. Nalm6^PD-L1 cells were cocultured with 1:1 mixtures of STAb-TP4+NT-T, STAb-T19+NT-T, or STAb-T19+STAb-TP4 at the indicated E:T ratios. The percentages of the different cell populations present in the cocultures, on the basis of CD2 and CD10 expression, were assessed by flow cytometry after 0, 2, 5, and 9 days.Figure S6. Replication of the leukemia escape assay.Nalm6^PD-L1 cells were cocultured with 1:1 mixtures of STAb-TP4+NT-T, STAb-T19+NT-T or STAb-T19+STAb-TP4 at the indicated E:T ratios, and the expression of CD2 and CD10 was analyzed by flow cytometry after 4, 12, 20, and 24 days. One of two similar experiments is shown.Figure S7. Exhaustion marker analysis in T cell–tumor cell cocultures.Nalm6^PD-L1 cells were cocultured with 1:1 mixtures of STAb-TP4+NT-T, STAb-T19+NT-T or STAb-T19+STAb-TP4 at the indicated E:T ratios (1:1 and 1:8), and the expression of LAG-3 (A), TIM-3 (B) and PD-1 (C) was analyzed by flow cytometry after 0, 3, 6, and 9 days. The data are shown as means ± SEM (n=3). Significance was determined by two-way ANOVA with Tukey’s multiple comparisons test. Significance is defined as follows: *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001.Figure S8. Multiplex cytokine profiling in T cell–tumor cell cocultures.Nalm6^PD-L1 cells were cocultured with 1:1 mixtures of STAb-TP4+NT-T, STAb-T19+NT-T, or STAb-T19+STAb-TP4 at the indicated E:T ratios (1:1 and 1:8), and the supernatants were collected at days 3 and 9. Cytokine secretion was measured in multiplex bead-based immunoassays. The data are shown as mean ± SD (n=3). Significance was determined by two-way ANOVA with Tukey’s multiple comparisons test. Significance is defined as follows: *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001 (see Table S2).Table S1. Antibodies used for flow cytometry analysis.Table S2. Significant results from multiplex bead-based immunoassays were analyzed by two-way ANOVA followed by Tukey’s multiple comparisons test.

Funding Statement

BB is funded by Fundación Científica de la Asociación Española Contra el Cáncer (INNOV211832BLAN), the Comunidad de Madrid (IND2022/BMD−23732) and the Instituto de Salud Carlos III (ISCIII) (PI20/01030, PI23/01256). LA-V is funded by the Spanish Ministry of Science and Innovation MCIN/AEI/10.13039/501100011033 (PID2023-148429OB-I00, PID2020-117323RB-I00, PDC2021−121711-I00, CPP2022−009762, CPP2022−009765, CPP2023−010827), ISCIII (DTS20/00089, PMPTA22/00167), the Asociación Española contra el Cáncer (PROYE19084ALVA and PRYGN234844ALVA), the CRIS Cancer Foundation (FCRIS−2021−0090 and FCRIS−2023−0070), the Fundación ‘‘La Caixa’’ (HR21−00761 project IL7R_LungCan), the Comunidad de Madrid (P2022/BMD−7225 Next Generation CART MAD), and the Fundación FERO (BBASELGAFERO2024.01). M.G.-R. is supported by an industrial PhD fellowship from the Comunidad de Madrid (IND2022/BMD−23732). FJB is funded by PID2023-152271NB-I00 from MCIN/AEI/10.13039/501100011033. We thank the Proteomics service at CIC bioGUNE and the Molecular Interactions service at CIB-CSIC for assisting with the mass spectrometry and SEC-MALS experiments. This study has adhered to ARRIVE guidelines ('la Caixa' Foundation) (FCRIS−2021−0090,FCRIS−2023−0070) (IND2022/BMD-23732) (Fundación Científica Asociación Española Contra el Cáncer) (Ministerio de Ciencia e Innovación) (PDC2021-121711-I00, CPP2022-009762)

Supplemental material

Supplemental data for this article can be accessed at https://doi.org/10.1080/2162402X.2025.2570224.

Disclosure of potential conflicts of interest

BB and LA-V are cofounders of STAb Therapeutics, a spin-off company from the Instituto de Investigación Sanitaria Hospital 12 de Octubre (imas12). BB and LA-V are inventors on the patent application “Anti-CD19/anti-CD3 bispecific antibody, T cells secreting the same, method of preparation and use thereof” (EP21708942). LD-A and LA-V are inventors on the patent application “T cells expressing anti-BCMA/anti-CD3 antibodies and uses thereof” (EP23383410.0). LA-V is a cofounder of Leadartis, a biotech company focused on an unrelated interest.

Author contributions

Conceptualization: MG-R, LA-V, BB; Data curation: MG-R, LR-P; Formal analysis: MG-R, LR-P, FJB, MC, BB; Funding acquisition: BB; Investigation: MG-R, LR-P, IZ, CD-A, LD-A, OH, RN, RT-R, SR-P, FJB, BB; Methodology: MG-R, BB; Project administration: BB; Resources: BB; Software: MG-R, FJB, BB; Supervision: MC, LA-V, BB; Validation: LA-V, BB; Visualization: MG-R, BB; Writing- original draft: MG-R, BB; Writing- review and editing: all authors.

Data availability statement

The data that support the findings of this study are available from the corresponding author, BB, upon reasonable request.