2022-Anti-CD19 chimeric antigen receptor T cells secreting anti-PD-L1 single-chain variable fragment attenuate PD-L1 mediated T cell inhibition

International Immunopharmacology 113 (2022) 109442 Contents lists available at ScienceDirect International Immunopharmacology journal homepage: www.elsevier.com/locate/intimp Anti-CD19 chimeric antigen receptor T cells secreting anti-PD-L1 single-chain variable fragment attenuate PD-L1 mediated T cell inhibition Pornpimon Yuti a,b,1, Yupanun Wutti-in a,b,c,1, Nunghathai Sawasdee a,b, Katesara Kongkhla a,b, Nattaporn Phanthaphol a,b,c, Kornkan Choomee a,b, Thaweesak Chieochansin a,b, Aussara Panya d, Mutita Junking a,b, Pa-thai Yenchitsomanus a,b,*, Jatuporn Sujjitjoon a,b,* a Siriraj Center of Research Excellence for Cancer Immunotherapy (SiCORE-CIT), Faculty of Medicine Siriraj Hospital, Mahidol University, Bangkok, Thailand b Division of Molecular Medicine, Research Department, Faculty of Medicine Siriraj Hospital, Mahidol University, Bangkok, Thailand c Graduate Program in Immunology, Department of Immunology, Faculty of Medicine Siriraj Hospital, Mahidol University, Bangkok, Thailand d Department of Biology, Faculty of Science, Chiang Mai University, Chiang Mai, Thailand ARTICLE INFO ABSTRACT Keywords: Adoptive T cell therapy using second-generation anti-CD19 chimeric antigen receptor T cells (anti-CD19-CAR2- Cancer immunotherapy T) induced complete remission in many heavily pretreated patients with B cell acute lymphoblastic leukemia (B- Chimeric antigen receptor ALL) or diffuse large B cell lymphoma (DLBCL). However, poor clinical efficacy was observed in treating CAR aggressive B cell lymphomas (BCL). The limited T cell function was reported by programmed cell death protein 1 CD19 ligand (PD-L1) expressed on BCL cells bound to the PD-1 receptor on T cells. To overcome this problem, we Anti-PD-L1 scFv generated anti-CD19-CAR4-T cells secreting anti-PD-L1 single-chain variable fragment (scFv), namely anti-CD19- Immune checkpoint blockade CAR5-T cells, and evaluated their functions in vitro. Both anti-CD19-CAR-T cells contain an anti-CD19 scFv derived from a monoclonal antibody, FMC63, linked to CD28/4-1BB/CD27/CD3ζ. The secreting anti-PD-L1 scFv is derived from atezolizumab. Our results showed that secreted anti-PD-L1 scFv could bind to PD-L1 and block the binding of anti-PD-L1 monoclonal antibodies on PD-L1high tumor cells. Anti-CD19-CAR4-T and anti-CD19- CAR5-T cells efficiently killed CD19+ target tumor cells in two-dimensional (2D) and three-dimensional (3D) co-culture systems. However, anti-CD19-CAR5-T cells demonstrated superior proliferative ability. Interestingly, at a low effector (E) to target (T) ratio of 0.5:1, anti-CD19-CAR5-T cells showed higher cytotoxicity against CD19+/PD-L1high cells compared to that of anti-CD19-CAR4-T cells. The cytotoxicity of anti-CD19-CAR4-T cells against CD19+/PD-L1high could be restored by adding anti-PD-L1 scFv. Our findings demonstrate the combina­ tion antitumor efficiency of anti-CD19-CAR4-T cells and anti-PD-L1 scFv against CD19+/PD-L1high tumors. As such, anti-CD19-CAR5-T cells should be further investigated in vivo antitumor efficiency and clinical trials as a treatment for aggressive B cell lymphoma. 1. Introduction standard treatment [1,2]. Thus, second-line drugs and several treatment regimens have been developed to improve the effectiveness of treat­ Lymphomas are the most common hematopoietic malignancies, ments for BCL [3]. Immunotherapies that employ monoclonal anti­ approximately 95 % of types are B cell in origin, and type behavior bodies and adoptive T cell transfer to target tumor antigens, immune varies from slow growing to highly aggressive [1]. Chemotherapy is a checkpoint molecules, and T cell-suppressing cytokines are emerging standard treatment that yields a good therapeutic response in certain promising treatments for leukemias and solid cancers [4,5]. types of lymphomas. However, approximately 30–40 % of aggressive B cell lymphomas (BCL), such as Burkitt lymphoma (BL) and diffuse large Adoptive T cell therapy using chimeric antigen receptor (CAR)-T B cell lymphoma (DLBCL), fail to respond to treatment or relapse after cells was recently approved by the United States Food and Drug Federation (US FDA) for treating B cell malignancies [6–9] and multiple * Corresponding authors at: Division of Molecular Medicine, Research Department, Faculty of Medicine Siriraj Hospital, Mahidol University, 2 Wanglang Road, Bangkoknoi, Bangkok 10700, Thailand. E-mail addresses: [email protected] (P.-t. Yenchitsomanus), [email protected] (J. Sujjitjoon). 1 These two authors contributed equally to this work. https://doi.org/10.1016/j.intimp.2022.109442 Received 18 August 2022; Received in revised form 30 October 2022; Accepted 6 November 2022 Available online 23 November 2022 1567-5769/© 2022 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by- nc-nd/4.0/).

P. Yuti et al. International Immunopharmacology 113 (2022) 109442 myeloma [10,11]. The basic structure of a CAR molecule comprises an the two common ICBs widely used in clinics. Atezolizumab, which is a extracellular domain for binding a tumor antigen, a hinge and trans­ monoclonal antibody that blocks PD-L1, has been approved as a first-line membrane domain, and an intracellular costimulatory domain that is therapy in children with non-Hodgkin lymphoma and Hodgkin lym­ linked to the T cell receptor (TCR) activation domain - CD3ζ [12]. CAR-T phoma [33]. Combination therapy using anti-CD19-CAR-T cells and cells targeting CD19 (anti-CD19-CAR-T cells) showed promising clinical atezolizumab is being evaluated in an ongoing clinical trial, but the outcomes for treatment of relapsed and refractory acute lymphoblastic preliminary results suggest the safety and efficacy in DLBCL (ClinicalT leukemia (r/r ALL) [6]. The four US FDA-approved anti-CD19-CAR-T rials.gov identifier: NCT02926833). This preliminary finding suggests cell products, including tisagenlecleucel [6], axicabtagene ciloleucel that combination therapy using CAR-T cells and ICB may improve the [7], brexucabtagene autoleucel [8], and lisocabtagene maraleucel [9], treatment outcome in patients with aggressive BCL. Although several are second-generation CAR-T cells (CAR2-T). These CAR-T cells were studies reported satisfactory clinical outcomes of ICB monotherapy [32], generated using an anti-CD19 single-chain variable fragment (scFv) that this treatment strategy requires several cycles of ICB administration, was derived from the mouse monoclonal antibody mFMC63 as an which may cause immune-related adverse events, and the treatment is extracellular domain linked to a single intracellular costimulatory highly expensive. domain of CD28 or 4-1BB, and TCR activation domain – CD3ζ (CD28/ CD3ζ or 4-1BB/CD3ζ) [6–9]. The incorporation of different intracellular CAR-T cells engineered to secrete ICB, such as anti-PD-1 or anti-PD- costimulatory domains, either CD28 or 4-1BB, in the CAR molecule in­ L1 antibody, were developed to improve the antitumor efficiency of fluences the properties of CAR-T cells. CD28-based CAR-T cells exhibited CAR-T cells in both in vitro and in vivo studies [34–36]. For example, rapid expansion with less durability, whereas the 41BB-based CAR-T anti-PD-L1 IgG secreted by engineered CAR-T cells targeting carbonic cells demonstrated slower cell expansion with more sustained persis­ anhydrase (CAIX) [34] and mesothelin [36] resulted in enhanced anti­ tence [13]. Recently, CD27 costimulatory signals in CAR2-T cells tumor efficiency, and could control tumor growth in mice. Li, et al. re­ effectuated both efficient killing in solid tumor models and prolonged T ported that anti-CD19-CAR-T cells secreting anti-PD-1 scFv were cell survival [14,15]. functionally more efficient than the parental CD28-based CAR-T cells for targeting B cell malignancies [37]. In addition, there were evidences Despite the promising treatment outcomes of anti-CD19-CAR2-T cell from previous in vitro [38,39] and in vivo [40] studies that the use of PD- therapy for ALL that were demonstrated [6], poor treatment response L1 blockade was more effective than the use of PD-1 blockade. However, was reported in aggressive large B cell lymphomas (BCL) [16,17]. the generation of anti-CD19-CAR4-T cells to secrete anti-PD-L1 scFv Improvement in anti-CD19-CAR-T cell treatment for BCL is, therefore, against tumor cells with CD19+ and high PD-L1 expression (CD19+/PD- urgently needed. There was evidence from a clinical study of non- L1high) has never been reported. Hodgkin’s lymphoma (NHL) that third-generation anti-CD19-CAR-T cells incorporating CD28 and 4-1BB costimulatory domains resulted in To increase the antitumor efficiency of CAR-T cells targeting CD19+/ better CAR-T cell expansion than that of second-generation anti-CD19- PD-L1high tumor cells, we engineered anti-CD19-CAR4-T cells to secrete CAR-T cells [18]. Interestingly, a recent phase I clinical trial demon­ anti-PD-L1 scFv (hereafter referred to as anti-CD19-CAR5-T cells) and strated a durable response and few adverse events in NHL patients who we evaluated their cytotoxic functions against tumor cells. We hypoth­ received fourth-generation anti-CD19-CAR-T cells containing CD28/ esized that anti-CD19-CAR5-T cells will be able to target the CD19 an­ CD27 costimulatory domains and inducible caspase 9 [19]. To further tigen on tumor cells and, at the same time, secrete anti-PD-L1 scFv to improve CAR-T, our group created fourth-generation CAR-T cells block PD-1/PD-L1 interaction on anti-CD19-CAR5-T cells and target (CAR4-T cells) that contain three costimulatory domains (CD28, 4-1BB, tumor cells. The cytotoxic functions, cytokine secretion levels, and and CD27). These CAR4-T cells exhibited high cytotoxic activity against proliferation of anti-CD19-CAR5-T cells were evaluated and compared different tumor models, including breast cancer [20], chol­ to those of anti-CD19-CAR4-T cells. angiocarcinoma [21–23], and recently B cell malignancies [24]. More­ over, we found and reported that anti-CD19-CAR4-T cells carrying 2. Materials and methods mFMC63 scFv exerted superior antitumor efficiency against CD19+ Raji cells compared to the antitumor efficiencies of anti-CD19-CAR2-T cells 2.1. Ethical approval and anti-CD19-CAR3-T cells [24]. The project proposal and the protocol for the collection of venous The programmed cell death 1 (PD-1; CD279) and PD-ligand 1 (PD- blood samples from healthy volunteers after obtaining written informed L1; B7-H1 and CD274) axis plays a crucial role in the suppression and consent to do so were both approved by Siriraj Institutional Review exhaustion of T cells [25]. PD-1/PD-L1 interaction usually increases Board (SIRB) of the Faculty of Medicine Siriraj Hospital, Mahidol Uni­ after CAR-T cells encounter the target antigen on tumor cells and in the versity, Bangkok, Thailand (COA number: Si 762/2016). tumor microenvironment. Increased expression of PD-1 on T cells, and increased expression of PD-L1 on tumor cells are commonly reported to 2.2. Cell lines and cell culture conditions impair the potency of CAR-T cells [26,27]. Thus, PD-1/PD-L1 interac­ tion is a significant barrier that limits the efficacy of CAR-T cell therapy. Raji, which is a human Burkitt lymphoma cell line (American Type Moreover, it was reported that 15 % of tumor-infiltrating lymphocytes Culture Collection [ATCC] cat# CCL-86, RRID: CVCL_0511; Manassas, (TILs) in lymphoma tissues of patients with DLBCL exhibited PD-1 VA, USA), was used as CD19+/PD-L1low cell in this study. Raji cells were expression, which was found to be associated with poor survival out­ also engineered to express the human PD-L1 protein by using the coding comes [28]. In patients with lymphomas, PD-1 expression on CD4+ CAR- sequences of the CD274 gene (NM_014143), namely Raji-PD-L1, which T cells was elevated up to threefold from the first infusion [29]. Retro­ was also used as CD19+/PD-L1high cell. Jurkat T cells, which are spective analyses found PD-L1 to be expressed in approximately 30 % of immortalized human T lymphocytes (ATCC cat# TIB-152, RRID: DLBCL malignant cells [30,31]. Importantly, a favorable treatment CVCL_0065), were transduced with lentiviruses carrying cDNA se­ outcome was observed when patients with high PD-L1 expression on quences encoding anti-CD19-CAR4 or anti-CD19-CAR4 secreting anti- tumor cells were treated with CAR-T cells together with PD-1/PD-L1 PD-L1 ScFv (anti-CD19-CAR5) for use in anti-PD-L1 scFv binding inhi­ blockade [32]. bition assays. These cell lines were cultured in Roswell Park Memorial Institute (RPMI)-1640 Medium (Gibco; Invitrogen Corporation, Carls­ Immune checkpoint blockade (ICB) has been approved for the bad, CA, USA) supplemented with 10 % heat-inactivated fetal bovine treatment of several cancers, such as non-small cell lung cancer, mela­ serum (FBS) (Gibco; Invitrogen) and penicillin (100 U/ml)/strepto­ noma, hepatocellular carcinoma, and aggressive BCL. ICB was designed mycin (0.1 mg/ml). Human cervical carcinoma (HeLa) cells (ATCC cat# to disrupt PD-1/PD-L1 interaction, thereby improving the cytotoxic CRL-7923, RRID: CVCL_0030) and HeLa-CD19+ cells were created as function of T cells. At present, anti-PD-L1 and anti-PD-1 antibodies are 2

P. Yuti et al. International Immunopharmacology 113 (2022) 109442 described previously [24]. Lenti-X™ 293T cell line (Takara Bio, Shiga, supplemented with cytokines every 2–3 days. Japan) was used for lentivirus packaging. Human cholangiocarcinoma cell line, KKU-213A (JCRB1557; Japanese Collection of Research Bio­ 2.6. Flow cytometry resources [JCRB] Cell Bank, Osaka, Japan), expressing almost 100 % PD-L1 on the cell surface was used as native cancer cells expressing PD- To examine the expression of the CD19 and PD-L1 proteins on cancer L1 for anti-PD-L1 scFv functional testing. These cell lines were main­ cell surface, anti-CD19-APC (Clone LT19, ImmunoTools) or anti-PD-L1- tained in Dulbecco’s Modified Eagle’s Medium (DMEM) (Gibco; Invi­ PE (Clone 29E. 2A3; BioLegend, San Diego, CA, USA) was incubated trogen), 10 % FBS, and penicillin (100 U/ml)/streptomycin (0.1 mg/ml) with the cell lines at an antibody dilution of 1:50 in 2 % FBS in at 37 ◦C in a 5 % CO2 atmosphere. phosphate-buffered saline (PBS). To examine the transduction effi­ ciencies of anti-CD19-CAR4-T cells and anti-CD19-CAR5-T cells, CAR 2.3. Plasmid construction surface expression was detected using anti-c-Myc-FITC (Clone ab1394; Abcam, Cambridge, UK). T cell immunophenotypes were examined A self-inactivated lentiviral transfer plasmid, pCDH.EF1α.SIN.WPRE, using anti-CD3-FITC (Clone UCHT-1), anti-CD4-APC (Clone MEM-241), harboring anti-CD19-CAR4, was constructed as previously described anti-CD8-APC (Clone UCHT-4), anti-CD16-APC (Clone 3G8), anti-45RO- [24]. In brief, a murine-derived anti-CD19 scFv, mFMC63, was cloned FITC (Clone UCHL1), and CD62L-PE (Clone HI62L) – all of which were in-frame with the sequence of the human c-Myc peptide tag connected purchased from ImmunoTools. Anti-CD56-PE (Clone AB_2563925), anti- with the sequence of the CAR4 containing the human CD8 hinge PD-1-PE (Clone EH12.2H7), anti-LAG3-PE (Clone C9B7W), and anti- domain, the CD28 transmembrane domain, three intracellular cos­ TIM3-PE (Clone F38-2E2) were purchased from BioLegend. Flow timulatory domains (CD28/4-1BB/CD27), and the CD3ζ activation cytometry was conducted using a BD AccuriTM C6 Plus Flow Cytometer domain. The anti-PD-L1 scFv comprised a T2A peptide, a signal peptide (BD Biosciences, Franklin Lakes, NJ, USA). The acquired data were (SP), a variable heavy chain (VH), a (GGGGS)x3 linker, a variable light analyzed using FlowJo 10.0 software (FlowJo LLC, Ashland, OR, USA). chain (VL) derived from an anti-PD-L1 monoclonal antibody (atezoli­ zumab), and a hemagglutinin (HA) epitope tag. The cDNA sequence 2.7. Binding and binding inhibition assays encoding anti-PD-L1 scFv was synthesized by Integrated DNA Technol­ ogies (Coralville, IA, USA). The anti-PD-L1 scFv cDNA sequence was To examine the binding of secreted anti-PD-L1 scFv to the PD-L1 inserted downstream of the sequence of anti-CD19-CAR4 to create anti- protein, and its ability to inhibit binding of the anti-PD-L1 antibody to CD19-CAR4 secreting anti-PD-L1 scFv (anti-CD19-CAR5). They were the PD-L1 protein, KKU-213A cells that highly express the PD-L1 protein linked via a T2A peptide for cleavage of anti-PD-L1 scFv to be secreted on their cell surface were used as target cells. The KKU-213A cells were from the cells. The cloned cDNA sequence was under the control of a incubated with culture medium of Jurkat T cells transduced with len­ human elongation factor-1α (EF-1α) promotor. The plasmid construct tiviruses expressing anti-CD19-CAR5 for 60 min at 4 ◦C. The KKU-213A was transformed into Stbl3 Escherichia coli and extracted using a Mid­ cells were then washed twice with 2 % FBS/PBS. For the direct binding iprep Kit (Qiagen, Hilden, Germany). assay, anti-HA (Clone 2–2.2.14; Invitrogen) was added to the KKU213A cells incubated with anti-PD-L1 scFv, and the cells were left to incubate 2.4. Lentivirus production for 60 min at 4 ◦C. The Alexa-488 conjugated goat anti-mouse antibody was then added to facilitate fluorescence detection via flow cytometry. Lenti-X™ 293T cells were used for production of lentiviral particles For the binding inhibition assay, KKU213A cells were initially incubated by co-transfection of three lentiviral plasmid vectors; psPAX2, pMD2.G, with anti-PD-L1 scFv for 60 min at 4 ◦C, and then subsequently incu­ and pCDH-EF1α-anti-CD19-CAR4 or pCDH-EF1α-anti-CD19-CAR5 bated with the anti-PD-L1-PE antibody (clone 29E.2A3; Biolegend) for transfer vector into the cells using the calcium phosphate precipitation 30 min at 4 ◦C. After incubation, the KKU213A cells were washed and method. At 48- and 72-hours after transfection, culture supernatants resuspended in fluorescence-activated cell sorting (FACS) buffer before containing lentiviral particles were collected and filtered through a 0.45 analysis using a BD Accuri™ C6 Plus Flow Cytometer (BD Biosciences). µm membrane to remove cell debris. Virus particles were concentrated The acquired data were analyzed using FlowJo 10.0 software (FlowJo by high-speed centrifugation at 20,000g for 150 min at 4 ◦C. The virus LLC). The data were reported as relative mean fluorescence intensity particles were stored at − 80 ◦C until use in subsequent experiments. The (MFI) that was normalized to the MFI of KKU-213A cells incubated with virus titer was determined by transduction into Jurkat T cells or quan­ mock culture medium and stained with isotype antibody. tified using a qPCR Lentiviral Titration Kit (Applied Biological Materials [ABM], Richmond, BC, Canada) according to the manufacturer’s 2.8. Immunoblot analysis instructions. Full-length anti-CD19-CAR4 and anti-CD19-CAR5 proteins were 2.5. Preparation of T cells and lentiviral transduction examined by immunoblot analysis. Lenti-X 293T cells were transfected with the lentiviral transfer plasmid, pCDH-EF1α-anti-CD19-CAR4 or Peripheral blood mononuclear cells (PBMCs) were isolated from pCDH-EF1α-anti-CD19-CAR5, by using Lipofectamine® 2000 Reagent heparinized venous blood samples obtained from healthy volunteers (Life Technologies Corporation, Carlsbad, CA, USA). After transfection using Lymphocyte Separation Medium (Corning, Inc., Corning, NY, and culturing for 48 h, the culture supernatant was collected, and the USA). Non-adherent lymphocytes were separated and activated for 72 h transfected cells were briefly lysed with 1 % NP-40 lysis buffer. To using 5 µg/ml phytohaemagglutinin (PHA)-L (Sigma-Aldrich Corpora­ examine anti-PD-L1 scFv protein, the culture medium (CM) of anti- tion, Saint Louis, MO, USA) in TexMACS Medium (Miltenyi Biotec, CD19-CAR5-transduced-Jurkat T cells was collected and analyzed by Bergisch Gladbach, Germany) supplemented with 5 % human AB serum immunoblot method. Transduced Jurkat T cells were separately seeded (Sigma-Aldrich Corporation, Saint Louis, MO, USA) and 20 ng/ml of into individual well of a 96-well plate (2x105 cells per well/ 200 μl recombinant human interleukin (rhIL)-2, 10 ng/ml rhIL-7, and 40 ng/ml culture medium). Then, the cells and culture supernatant from each well rhIL-15 (all Immunotools, Friesoythe, Germany). To create anti-CD19- were collected at different culture times as indicated. An equal volume CAR4-T cells and anti-CD19-CAR5-T cells, PHA-activated T cells were (25 μl) of culture supernatant was taken for electrophoretic separation transduced with lentiviruses using 10 μg/mL of protamine sulfate and immunoblot analysis. Whole-cell lysates and culture supernatants (Sigma-Aldrich) and spinoculated at 1,200g for 90 min at 32 ◦C. After were loaded to separate on 12 % sodium dodecyl sulfate-polyacrylamide transduction and culturing for 48 h, the virus-containing medium was gel electrophoresis (SDS-PAGE) and blotted onto a nitrocellulose replaced with TM5 medium (5 % human AB serum in TexMACS) membrane. The membrane was blocked with 5 % skim milk in Tris- 3

P. Yuti et al. International Immunopharmacology 113 (2022) 109442 buffered saline and 0.1 % Tween-20 for 30 min. To detect the full-length chloromethylfluorescein diacetate) Dye (Thermo Fisher Scientific). The anti-CD19-CAR4 protein, the membrane was incubated with mouse anti- mock-T cells, anti-CD19-CAR4-T cells, and anti-CD19-CAR5-T cells were CD3ζ antibody (clone sc-166435; Santa Cruz Biotechnology Dallas, TX, co-cultured with CMFDA-labeled target cells at E:T ratios of 0.5:1, 1:1, USA). To examine the HA-tagged anti-PD-L1 scFv protein, the mem­ and 5:1 for 24, 48, and 72 h. After co-culturing, counting beads brane was incubated with mouse anti-HA antibody (Clone 2-2.2.14; (123count™ eBeads Counting Beads; Thermo Fisher Scientific) were Invitrogen). Anti-GAPDH antibody (clone sc-32233; Santa Cruz mixed into each sample for analysis by flow cytometry. The absolute Biotechnology) was used to detect GAPDH, the protein loading control. numbers of effector cells and target cells were calculated according to The membrane was then incubated with goat anti-mouse horseradish the manufacturer’s instructions. Cytotoxicity was then calculated using peroxidase (HRP)-conjugated antibody (Invitrogen), and chemilumi­ the following formula: [1-(target cells in each condition/target cells nescent signals were generated via incubation with SuperSignal West alone at the indicated time)] × 100. These data were also calculated as Pico Substrate (Thermo Fisher Scientific, Waltham, MA, USA). fold-changes to determine T cell proliferation using the formula: [effector cells in each condition/effector cells alone at the starting time]. 2.9. Cytotoxicity assay by crystal violet staining 2.12. T cells proliferation assay To conduct the cytotoxicity assay, HeLa or HeLa-CD19+ cells (2x104 cells/well), which were used as target (T) cells, were cultured in a 96- To examine the effect of secreted anti-PD-L1 scFv on T cell prolif­ well plate for 16 h to allow them to firmly adhere to the plate. eration, T cells were labeled with 5 μM carboxyfluorescein succinimidyl Untransduced T (mock-T) cells, anti-CD19-CAR4-T cells, or anti-CD19- ester (CFSE) (Invitrogen) and co-cultured with target cells at E:T ratio of CAR5-T cells were used as effector (E) cells to co-culture with the 1:1 in the absence or presence of culture medium (CM) containing target (T) cells at E:T ratios of 1:1, 5:1, and 10:1 for 18 h. After removal secreted anti-PD-L1 scFv. After incubation for 72 h, CFSE dilution was of the culture medium and the effector cells, the target cells were washed measured by flow cytometry and T cell proliferation was calculated. twice with PBS, and adherent (living) cells were stained with crystal violet dye (Merck Millipore, Burlington, MA, USA) and incubated for 30 2.13. Degranulation assay min. After washing to remove the excess dye, the crystal violet-stained cells were dissolved in 1 % sodium dodecyl sulfate (SDS) for at least The mock-T cells, anti-CD19-CAR4-T cells, or anti-CD19-CAR5-T 30 min. Once all of the stained cells were entirely dissolved, the optical cells (5x104 cells) were co-cultured with Raji or Raji-PD-L1 cells at the density (OD) at 590 nm was measured using a Synergy™ Microplate E:T ratio of 5:1 in the culture medium containing anti-CD107a-APC Reader, and analyzed using Gen5™ software (both BioTek Instruments, antibody (560664, H4A3, BD Biosciences) to examine degranulation Winooski, VT, USA). The percentage of cytotoxicity was calculated using marker, CD107a. Four hours after incubation, Brefeldin A (00–4506-51; the following formula: [1- (experimental OD/(spontaneous OD)] × 100. Invitrogen) and Monensin (00–4505-51; Invitrogen) were added in order to block golgi exportation and intracellular protein transportation. 2.10. Three-dimensional (3D) tumor spheroid formation and cytotoxicity After another 20 h of incubation, the cultured cells were collected, assay immunostained by using FITC-conjugated anti-CD8a antibody (11–0088-42; Invitrogen). Viable cells were gated according to the To generate three-dimensional (3D) spheroids, 2x103 cells of HeLa typical forward/side scatter characteristics and CD107a expression on cells or HeLa-CD19+ cells were firstly labeled with 1 µM of CellTracker™ live CD8a+ cells was analyzed by flow cytometry. Green 5-chloromethylfluorescein diacetate (CMFDA) Dye (Thermo Fisher Scientific). Spheroid formation was initiated by adding 3 % 2.14. Cytokine production matrigel matrix (Corning) into the cells, which were seeded in a 96-well U-shaped plate (Spheroid ULA/CS; PerkinElmer, Waltham, MA, USA), The mock-T cells, anti-CD19-CAR4-T cells, or anti-CD19-CAR5-T followed by centrifugation at 1,200g for 10 min at 4 ◦C. Formation of the cells were co-cultured with Raji cells or Raji-PD-L1 cells at an E:T tumor spheroid was observed at 48 h after centrifugation. To conduct ratio of 5:1 for 24 h in a serum-free medium. The culture supernatants the cytotoxicity assay, the mock-T cells, anti-CD19-CAR4-T cells, or anti- were then harvested, centrifuged to remove cell debris, and stored at CD19-CAR5-T cells were resuspended in culture medium containing 1 -70 ◦C. Cytokine levels in culture supernatants (25 μl) were measured by µg/ml propidium iodide (PI) dye, and then the cells were gently added using Cytokine Bead Array (CBA) of the LEGENDplexTM Human CD8/NK into the spheroid culture at an E:T ratio of 5:1. The green tumor cells and cell panel (#741065, BioLegend), following the manufacturer’s in­ PI-stained dead cells were monitored, and their images were captured at structions. This panel allows simultaneous quantification of 13 human 24, 48, and 72 h using a confocal microscope (Nikon Instruments Inc., cytokines and proteins including: IL-2, IL-4, IL-6, IL-10, IL-17A, IFN-γ, Melville, NY, USA). The mean fluorescence intensity (MFI) of PI-stained TNF-α, soluble Fas, soluble FasL, granzyme A, granzyme B, perforin, and cells was analyzed using NIS-Element’s software (Nikon Instruments). granulysin. Cytotoxicity was calculated using the following formula: [(experimental MFI-spontaneous MFI)/(maximum MFI-spontaneous MFI)] × 100. The The samples were analyzed on CytoFLEX flow cytometer (Becton experimental MFI was defined as the MFI of spheroids co-cultured with Dickinson (BD) Biosciences, New Jersey, USA). Additionally, IFN-γ and mock-T cells, anti-CD19-CAR4-T cells, or anti-CD19-CAR5-T cells. The TNF-α levels were also measured by enzyme-linked immunosorbent spontaneous MFI was defined as the MFI of the spheroids alone (without assay (ELISA) using a Quantikine® ELISA Kit (R&D Systems, Inc., Min­ T cells). Maximum MFI was defined as the MFI of spheroids alone neapolis, MN, USA), according to the manufacturer’s instructions. (without T cells) that were treated with 0.05 % Triton-X 100 to deter­ mine total tumor cell death. 2.15. Statistical analysis 2.11. Cytotoxic effect and T cell proliferation assays as analyzed by flow GraphPad Prism 7 software (GraphPad Software, Inc., San Diego, CA, cytometry USA) was used for statistical analysis and graphical figure presentation. The results are summarized as mean ± standard error of the mean (SEM) To examine the cytotoxic effects of anti-CD19-CAR4-T cells and anti- from at least three independent experiments. Student’s t-test and two- CD19-CAR5-T cells against CD19+ tumors, Raji and Raji-PD-L1 that way analysis of variance (ANOVA) was used for comparisons between endogenously expressed CD19 were used as target cells. The target cells and among groups. The p-value < 0.05 was considered statistically were labeled with 1 µM of CellTracker™ CMFDA (5- significant for all tests. 4

P. Yuti et al. International Immunopharmacology 113 (2022) 109442 3. Results scFv derived from atezolizumab, which is a humanized monoclonal antibody specific to PD-L1. The schematic representations of the lenti­ 3.1. Expression of anti-CD19-CAR4 and anti-CD19-CAR5 in Lenti-X viral constructs containing anti-CD19-CAR4 and anti-CD19-CAR5 se­ 293T and Jurkat T cell lines quences are shown in Fig. 1A. These two lentiviral constructs are composed of cDNA sequences encoding a signal peptide, anti-CD19 scFv Our group previously reported that anti-CD19-CAR4-T cells exhibi­ (namely mFMC63) c-Myc tag, CD8 hinge, CD28 transmembrane, and ted superior cytotoxic functions over those of anti-CD19-CAR2 T cells intracellular costimulatory domains of CD28, 4-1BB (or CD137), CD27, and anti-CD19-CAR3 T cells [24]. We, therefore, selected anti-CD19- joined to intracellular activation domain – CD3ζ, which was designated CAR4-T cells to generate anti-CD19-CAR4-T cells secreting anti-PD-L1 as anti-CD19-CAR4. The original version of the anti-CD19-CAR4 scFv, namely anti-CD19-CAR5-T cells via incorporation of anti-PD-L1 construct (Fig. 1A: the upper construct) was modified by Fig. 1. Construction, expression, and characterization of the anti-CD19-CAR4 and anti-CD19-CAR5 proteins. (A) Upper: Schematic representation of the anti-CD19- CAR4 construct in the lentiviral transfer vector containing the serial sequences of 5′ LTR, EF-1α promoter, signal peptide, mFMC63 scFv, human c-Myc peptide tag, human CD8 hinge domain, CD28 transmembrane domain, three intracellular costimulatory domains (CD28/4-1BB/CD27), and CD3ζ, Lower: Schematic repre­ sentation of the anti-CD19-CAR5 construct in the lentiviral transfer vector containing the same sequence of anti-CD19-CAR4 construct and the sequence of anti-PD-L1 scFv containing the signal peptide inserted downstream of the sequences of the T2A cleavage peptide, and joined with the sequence of HA-tag. (B) Upper: Immunoblot analysis of the full-length anti-CD19-CAR4 protein (first row) and the anti-PD-L1 scFv protein (second row) in cell lysates of transfected Lenti-X 293 T cells compared to the untransfected cells in which the glyceraldehyde 3-phosphate dehydrogenase (GAPDH) protein was used as loading control (third row). Lower: anti-PD-L1 scFv detected in culture medium (CM). (C) Histogram plots demonstrate CAR surface expression of anti-CD19-CAR4 and anti-CD19-CAR5 in transfected Lenti-X 293 T cells. (D) Bar graph summarizes CAR surface expression. (E) Immunoblot analysis of HA-tagged anti-PD-L1 scFv secreted from Jurkat T cells transduced with lentiviruses containing anti-CD19-CAR5 construct, which increased in a time-dependent manner. (F) Binding activity of the secreted anti-PD-L1 scFv to PD-L1 expressed on KKU213A cells, and (G) inhibitory activity of secreted anti-PD-L1 scFv against the binding of monoclonal anti-PD-L1 antibody to PD-L1 expressed on KKU-213A cells. The results are summarized as mean ± standard error of the mean (SEM) from at least three independent experiments. Statistical differences were determined by Student’s t-test, and asterisks indicate the following p-values: *p < 0.05, **p < 0.01, and ***p < 0.001. (Abbreviations: CAR, chimeric antigen receptor; CD, cluster of differentiation; HA-tag, human influenza hemagglutinin; Ctrl, control; MFI, mean fluorescence intensity; h, hours). 5

P. Yuti et al. International Immunopharmacology 113 (2022) 109442 incorporating a cDNA sequence encoding anti-PD-L1 scFv in cis at the 3́- These results indicate that anti-PD-L1 scFv secreted from the Jurkat T end of anti-CD19-CAR4 joined with a HA tag sequence, and the modified cells transduced with anti-CD19-CAR5 construct exhibited binding ac­ version was designated as the anti-CD19-CAR5 construct (Fig. 1A: the tivity and blocking activity against the binding of monoclonal anti-PD- lower construct). Both constructs were controlled by a human elon­ L1 antibody to PD-L1 expressed on KKU-213A cells. gation factor 1α (EF-1α) promoter to obtain high levels of expression in primary human T cells. 3.2. Generation and characterization of anti-CD19-CAR4 and anti- CD19-CAR5 in primary human T cells To examine anti-CD19-CAR4 protein expression, these two lentiviral constructs were transfected into Lenti-X 293 T cells. The full-length anti- To generate and characterize the anti-CD19-CAR4 and anti-CD19- CD19-CAR4 proteins (72 kDa) that were produced by these two con­ CAR5 in primary human T cells, PHA-L activated human T cells were structs were detected by immunoblot method using an anti-CD3ζ anti­ transduced with lentiviral particles harboring anti-CD19-CAR4 or anti- body (Fig. 1B). The anti-PD-L1 scFv protein containing HA-tag (30 kDa) CD19-CAR5 sequence. Transduction efficiencies were examined by that was produced from the anti-CD19-CAR5 lentiviral construct was CAR expression on T cells by flow cytometric analysis, and the repre­ detected by immunoblot analysis in both whole-cell lysate (upper panel) sentative histograms are shown in Fig. 2A. The summarized results from and in condition medium (CM; lower panel) (Fig. 1B), which indicates 4 independent experiments showed the CAR surface expression of anti- that it was secreted from the transfected Lenti-X 293 T cells. CD19-CAR4-T cells and anti-CD19-CAR5-T cells were 47.5 ± 2.4 % and 50.0 ± 3.9 %, respectively, compared with the background staining of Cell surface expression of anti-CD19-CAR4 and anti-CD19-CAR5 in 4.22 ± 0.45 % on mock-T cells (Fig. 2B). These results indicated that the transfected Lenti-X 293 T cells was also detected by flow cytometric human anti-CD19-CAR proteins were expressed on T cells, and that the analysis using an FITC-conjugated anti-c-Myc antibody, which recog­ anti-CD19-CAR4-T cells and anti-CD19-CAR5-T cells were successfully nized c-Myc-tag locating between the mFMC63 scFv and the hinge re­ generated. gion. The representative histograms of the results from flow cytometric analysis are shown in Fig. 1C. The mean of CAR protein expression on The CAR-T cell phenotypes, including immune cell subpopulation, the surface of Lenti-X 293 T cells from the anti-CD19-CAR4 and anti- exhaustion profile, and memory T cell subtypes, were then investigated. CD19-CAR5 lentiviral constructs was 50.0 ± 6.4 % (p = 0.0018) and T cells in PBMCs, PHA-L activated T cells, mock-T, anti-CD19-CAR4-T 53.4 ± 7.8 % (p = 0.0028), respectively, compared to untransfected cells cells, and anti-CD19-CAR5-T cells were stained to determine the ex­ (2.29 ± 0.84 %) (Fig. 1D). These results indicate successful production pressions of cell surface markers by flow cytometry. After PHA-L acti­ of full-length anti-CD19-CAR4, anti-CD19-CAR5, and anti-PD-L1 scFv vation, the proportions of CD3+/CD4+ and CD3+/CD8+ T cells were secreted from the transfected cells. similar. However, the proportions of CD3+/CD8+ and CD3+/CD4+ T cells were changed from the ratio of 1:1 to the ratio of 2:1 in all trans­ To further examine the basic functions of the secreted anti-PD-L1 duced T cell products (Fig. 2C). Other populations of immune cells, scFv, anti-PD-L1 scFv protein secretion and stability in culture me­ including NK cells, NKT cells, and B cells, were also determined; how­ dium, its direct binding ability to the PD-L1 antigen expressed on the ever, these immune cells were represented as small proportions (Sup­ tumor cells, and its ability to block the binding of the monoclonal anti- plementary Fig. 2A-C). PD-L1 antibody were investigated. Jurkat T cells, which is an immor­ talized human T cell line, were transduced with lentiviruses carrying the Expression of exhaustion markers [including PD-1, T cell immuno­ anti-CD19-CAR5 construct to produce the anti-PD-L1 scFv protein. The globulin domain and mucin domain (TIM)-3, and lymphocyte-activation culture media were collected at 0, 18, 24, 48, and 72 h. The secretion gene (LAG) 3] of PHA-L activated T cells were significantly increased and stability of anti-PD-L1 scFv were determined by immunoblot anal­ compared to their respective expressions in PBMCs (Fig. 2D). However, ysis. The results showed that the anti-PD-L1 scFv protein was detected at these exhaustion markers showed no significant differences among increasing levels at each of the 18-, 24-, 48-, and 72-hour timepoints mock-T cells, anti-CD19-CAR4-T cells, and anti-CD19-CAR5-T cells. (Fig. 1E). LAG3 expression in mock-T cells, anti-CD19-CAR4-T cells, and anti- CD19-CAR5-T cells was significantly lower than in PHA-L activated T The binding ability of anti-PD-L1 scFv to the PD-L1 protein was cells (Fig. 2D). The expressions of CD45RO and CD62L were used to examined using KKU-213A cells, which is a human cholangiocarcinoma analyze memory T cell subtypes in each transduction condition cell line that endogenously expresses a high level of PD-L1 protein on the (Fig. 2E), and the gating strategies are shown in Supplementary cell surface (Supplementary Fig. 1). The Jurkat T cells were transduced Fig. 2D. Central memory T cell (Tcm) was a major T cell subtype of with lentiviruses containing anti-CD19-CAR5 construct, and the culture mock-T cells, anti-CD19-CAR4-T cells, and anti-CD19-CAR5-T cells, media were collected as the source of anti-PD-L1 scFv for testing in KKU- followed by naïve T cell subset, effector memory T cell (Tem) subset, and 213A cells. After incubation, the binding of anti-PD-L1 scFv containing effector T cell (Teff) subtype. Teff was the least expressed subset. Tem HA-tag to the PD-L1 protein on the cell surface of KKU-213A cells was showed no significant difference in all conditions. These results suggest detected by flow cytometric analysis. The results showed the binding of no significant differences in memory T cell phenotypes or the effector anti-PD-L1 scFv present in the condition media collected at the different subpopulations of mock-T cells, anit-CD19-CAR4-T cells, or anti-CD19- time points (0, 18, 24, 48, and 72 h) was significantly increased from CAR5-T cells. 1.41 ± 0.06 to 2.85 ± 0.057, 3.25 ± 0.65, 4.17 ± 0.61, and 4.37 ± 0.53, respectively, relative to the control condition medium from the mock 3.3. Antitumor efficiency and specificity of anti-CD19-CAR4-T cells and cell culture (1.39 ± 0.06) (Fig. 1F). anti-CD19-CAR5-T cells against HeLa-CD19+ cells in co-culture systems The anti-PD-L1 scFv secreted in the condition media collected from HeLa cells, which naturally lack CD19, were engineered to over­ the transduced Jurkat T cells at different time points (0, 18, 24, 48, and express CD19, namely HeLa-CD19+ cells to evaluate specific cytotox­ 72 h) was also tested for inhibitory activity against the binding of icity in 2D and 3D spheroid co-culture systems (Supplementary Fig. 3A- monoclonal anti-PD-L1 antibody (anti-PD-L1-PE) to the PD-L1 protein B). Parental HeLa cells were used as negative control cells, and HeLa- on the cell surface of KKU-213A cells. The control conditions for these CD19+ cells were used as target cells overexpressing CD19. These two experiments consisted of untreated KKU-213A cells (1.00), and KKU- cells had low PD-L1 expression (Supplementary Fig. 3A-B). Untrans­ 213A cells treated with culture medium of mock-transduced Jurkat T duced T cells (mock-T cells; T cells lacking CAR), anti-CD19-CAR4-T cells (0.97 ± 0.01). The secreted anti-PD-L1 scFv collected at 18, 24, 48, cells, or anti-CD19-CAR5-T cells were co-cultured with HeLa or HeLa- and 72 h reduced the binding ability of the monoclonal anti-PD-L1 CD19+ cells at E:T ratios of 0:1, 1:1, 5:1, and 10:1 for 18 h. After antibody from the levels in untreated controls to 0.85 ± 0.03, 0.78 ± washing, the killed tumor cells were aspirated and the remaining living 0.04, 0.62 ± 0.02, and 0.62 ± 0.02, respectively (Fig. 1G). The results showed that the secreted anti-PD-L1 scFv could significantly inhibit the binding of monoclonal anti-PD-L1 antibody to PD-L1 on KKU-213A cells. 6

P. Yuti et al. International Immunopharmacology 113 (2022) 109442 Fig. 2. Generation of anti-CD19-CAR4-T cells and anti-CD19-CAR5-T cells in primary human T cells. (A) Histogram plots demonstrate CAR surface expression in primary human T cells detected using an anti-c-Myc antibody. (B) Summary of CAR surface expression on primary human T cells from 4 independent experiments. (C) Proportions of helper CD3+/CD4+ T cells and cytotoxic CD3+/CD8+ T cells of unactivated peripheral blood mononuclear cells (PBMCs), PHA-L activated T cells, mock-T cells, anti-CD19-CAR4-T cells, and anti-CD19-CAR5-T cells. (D) Expressions of exhaustion markers, including PD-1, TIM3, and LAG3, on different CAR-T cell products. (E) Memory T cell subtypes of different CAR-T cell products. The results are summarized as mean ± standard error of the mean (SEM) from at least three independent experiments using blood samples from different healthy donors. Statistical differences were determined by Student’s t-test, and asterisks indicate the following p-values: *p < 0.05, **p < 0.01, and ****p < 0.0001. (Abbreviation: CAR, chimeric antigen receptor; CD, cluster of differentiation; Naïve T, naïve T cells; Tcm, central memory T cells; Tem, effector memory T cells; Teff, effector T cells). cells were stained by crystal violet dye. The results showed that anti- parental HeLa cells and HeLa-CD19+ cells. CD19-CAR4-T cells and anti-CD19-CAR5-T cells could efficiently kill To mimic the cytotoxicity activities of anti-CD19-CAR4-T cells and HeLa-CD19+ cells, but that they inefficiently killed HeLa cells as demonstrated by less reduction of blue-stained cells left in the wells anti-CD19-CAR5-T cells against solid tumor of B cell lymphoma, 3D (Fig. 3A and Supplementary Fig. 4). The crystal violet-stained cells spheroids of parental HeLa cells and HeLa-CD19+ cells were generated were dissolved and the percentages of cytotoxicity were analyzed by mixing CMFDA-labeled HeLa cells or HeLa-CD19+ cells with matrigel (Fig. 3B). The results showed that while anti-CD19-CAR4-T cells and and culturing them for two days. These 3D spheroids were then co- anti-CD19-CAR5-T cells exhibited low killing activity against the cultured with the effector anti-CD19-CAR4-T cells or anti-CD19-CAR5- parental HeLa cells, they exhibited specific killing activity against HeLa- T cells in the presence of propidium iodide (PI). The death of green- CD19+ cells at E:T ratios of 5:1 (51.9 ± 13.5 % and 52.8 ± 9.0 %, fluorescence tumor spheroids was examined by the presence of red- respectively) and 10:1 (70.9 ± 10.6 % and 76.6 ± 2.2 %, respectively). fluorescence PI staining that was detected by confocal microscope at Very low cytotoxic activity of mock-T cells was observed in both 24, 48, and 72 h. The captured representative images, which were clearly demonstrated at 48 and 72 h, are illustrated in Fig. 3C-D. The 7

P. Yuti et al. International Immunopharmacology 113 (2022) 109442 Fig. 3. Cytotoxic activities and specificity of anti-CD19-CAR4-T cells and anti-CD19-CAR5-T cells against parental HeLa cells and HeLa-CD19+ cells in two- dimensional (2D) and three-dimensional (3D) co-culture systems. (A) Representative image of a crystal violet assay showing the cytotoxic activities of mock-T cells, anti-CD19-CAR4-T cells, and anti-CD19-CAR5-T cells against parental HeLa cells (left panel) and HeLa-CD19+ cells (right panel) at effector to target (E:T) ratios of 0:1, 1:1, 5:1, and 10:1 in a 2D co-culture system. (B) Percentages of cytotoxic activity of mock-T cells, anti-CD19-CAR4-T cells, and anti-CD19-CAR5-T cells against parental HeLa cells (left panel) and HeLa-CD19+ cells (right panel) at E:T ratios of 1:1, 5:1, and 10:1. (C) Representative image of the cytotoxic activities of mock-T cells, anti-CD19-CAR4-T cells, and anti-CD19-CAR5-T cells against parental HeLa cells at 48 and 72 h (h) in a 3D spheroid system. (D) Representative image of the cytotoxic activities of mock-T cells, anti-CD19-CAR4-T cells, and anti-CD19-CAR5-T cells against HeLa-CD19+ cells at 48 and 72 h in a 3D spheroid system. (E) Percentages of cytotoxic activity of mock-T cells, anti-CD19-CAR4-T cells, and anti-CD19-CAR5-T cells against parental HeLa cells at 24, 48, and 72 h. (F) Percentages of cytotoxic activity of mock-T cells, anti-CD19-CAR4-T cells, and anti-CD19-CAR5-T cells against HeLa-CD19+ cells at 24, 48, and 72 h. The scale bar in the inset images is 200 µm. The data shown in the histograms reflect the mean ± standard error of the mean (SEM) from at least three independent experiments. Statistical differences between the results of different cell types and mock-T cell results were determined by two-way analysis of variance (ANOVA), and asterisks indicate the following p-values: *p < 0.05, **p < 0.01, and ***p < 0.001. (Abbreviations: CAR, chimeric antigen receptor; CD, cluster of differentiation; CMFDA, 5-chlorome­ thylfluorescein diacetate; h, hour; PI, propidium iodide). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.) results showed that while the green-fluorescence 3D spheroids of HeLa (Fig. 3D). In addition, the intensities of green-fluorescence HeLa-CD19+ cells were not killed by anti-CD19-CAR4-T cells or anti-CD19-CAR5-T tumor cells were dramatically decreased, indicating the death of tumor cells (Fig. 3C), the 3D spheroids of HeLa-CD19+ cells were killed by cells. The percentages of cytotoxicity were calculated from the mean both anti-CD19-CAR4-T cells and anti-CD19-CAR5-T cells, as illustrated fluorescence intensities of PI-stained spheroids in each condition. The by the presence of red-fluorescence stained areas in the spheroids results showed low cytotoxicity of both anti-CD19-CAR4-T cells and 8

P. Yuti et al. International Immunopharmacology 113 (2022) 109442 anti-CD19-CAR5-T cells co-cultured with HeLa cell spheroids at 24, 48, showed, among other cytokines and proteins, the levels of three cyto­ and 72 h (Fig. 3E). In contrast, the cytotoxicity of anti-CD19-CAR4-T kines, including IL-2, IFN-γ, and TNF-α, in the culture media of anti- cells and anti-CD19-CAR5-T cells were significantly increased in a CD19-CAR4-T cells and anti-CD19-CAR5-T cells were significantly time-dependent manner when they were co-cultured with HeLa-CD19+ higher than those of mock-T cells, when they exposed to Raji cells or cell spheroids for 48 h (48.4 ± 12.9 % and 37.2 ± 10.3 %, respectively) Raji-PD-L1 cells (Fig. 4C-D). IL-2 levels in the culture media of anti- and 72 h (66.8 ± 9.5 % and 63.2 ± 5.6 %, respectively), when compared CD19-CAR4-T cells and anti-CD19-CAR5-T cells that were exposed to with the cytotoxicity of mock-T cells at both timepoints (6.2 ± 4.0 % at Raji cells were 12,692 ± 621 pg/ml and 10,432 ± 1764 pg/ml (p = 48 h, and 21.4 ± 9.1 % at 72 h) (Fig. 3F). After co-culturing for 48 and 0.2937). IFN-γ levels were 9,263 ± 1121 pg/ml and 6,529 ± 1172 pg/ 72 h, the cytotoxic activities of anti-CD19-CAR4-T and anti-CD19-CAR5- ml (p = 0.0800). TNF-α levels were 765 ± 112 pg/ml and 578 ± 83 pg/ T cells were significantly increased, compared to that of the mock-T ml (p = 0.2528) (Fig. 4C). When these cells were co-cultured with Raji- cells. However, at both timepoints, the cytotoxicities of anti-CD19- PD-L1 cells, the levels of IL-2, IFN-γ, and TNF-α in the culture media of CAR4-T cells and anti-CD19-CAR5-T cells were not significantly anti-CD19-CAR4-T cells and anti-CD19-CAR5-T cells were significantly different (p = 0.6466 and p = 0.9560, respectively). These results higher those of the mock-T cells (Fig. 4D). IL-2 levels in culture media of indicate that both anti-CD19-CAR4-T cells and anti-CD19-CAR5-T cells anti-CD19-CAR4-T cells and anti-CD19-CAR5-T cells were 10,984 ± demonstrated specific and efficient killing cytotoxic activities against 1315 pg/ml and 9,401 ± 2191 pg/ml (p = 0.5691). IFN-γ levels were HeLa-CD19+ cell spheroids, which mimicked the condition of a solid 6,779 ± 1443 pg/ml and 6,429 ± 841 pg/ml (p = 0.8442). TNF-α levels tumor. were 558 ± 92 pg/ml and 527 ± 47 pg/ml (p = 0.7796). The levels of these three cytokines in the culture media of anti-CD19-CAR4-T cells 3.4. Cytokine secretion of anti-CD19-CAR4-T cells and anti-CD19-CAR5- and anti-CD19-CAR5-T cells, exposed to either Raji cells or Raji-PD-L1 T cells exposed to target cells expressing CD19 with low or high PD-L1 cells, were not significantly different. These data were agreeable with expression the results that were examined by ELISA, which showed that the levels of IFN-γ and TNF-α in the culture media of anti-CD19-CAR4-T cells and When CAR-T cells exposed to specific antigen on the target cancer anti-CD19-CAR5-T cells were significantly higher than those of the cells, the CAR T cells were activated, leading to degranulation to secrete mock-T cells, when they exposed to either Raji cells or Raji-PD-L1 cells, perforin and granzymes, and secretion of cytokines, which then trig­ but the levels of these two cytokines in the culture media of anti-CD19- gered apoptosis of the target cancer cells. To evaluate the activation of CAR4-T cells and anti-CD19-CAR5-T cells, when they exposed to either anti-CD19-CAR-T cells exposed to target cancer cells expressing CD19 Raji cells (p = 0.3016 and p = 0.2643) or Raji-PD-L1 cells (p = 0.4905 and PD-L1, Raji (Burkitt human B lymphoma) cells naturally expressing and p = 0.5433), were not significantly different (Supplementary high level of CD19 (99.7 %) but low level of PD-L1 (22.9 %), and Raji- Fig. 5C-D). PD-L1 cells expressing high CD19 (95.0 %) and high PD-L1 (86.7 %) (Supplementary Fig. 3C-D) were co-cultured with the anti-CD19-CAR- Interestingly, the levels of IL-6, a proinflammatory cytokine that T cells. The relative mean fluorescence intensities (MFI) of PD-L1 ex­ causes cytokine-mediated toxicity or cytokine release syndrome (CRS) in pressions in HeLa, HeLa-CD19+, and Raji cells were not significantly CAR-T cell therapy, secreted from anti-CD19-CAR4 T cells, were higher different (1.66 ± 0.14, 1.98 ± 0.40, 2.14 ± 0.08, respectively) (Sup­ than those of anti-CD19-CAR5-T cells, when they exposed to either Raji plementary Fig. 3E). In contrast, the relative MFI of PD-L1 expression in cells or Raji-PD-L1 cells (Fig. 4E). Our data showed that the level of IL-6 Raji-PD-L1 cells were as high as 9.10 ± 1.31 folds, compared to other in the culture media of mock-T cells was 1.68 ± 0.56 pg/ml while its target cell lines (Supplementary Fig. 3E). level in the culture media of anti-CD19-CAR4-T cells was 8.33 ± 2.10 pg/ml; however, its level in the culture media of anti-CD19-CAR5-T cells The early step of anti-CD19-CAR-T cell activation was examined by was 3.74 ± 1.11 pg/ml, which was not significantly different from that measuring the degranulation marker, CD107a, on cytotoxic (CD8+) T of the mock-T cells (p = 0.5765). The level of IL-6 in the culture media of cells after the exposure to Raji or Raji-PD-L1 target cells. The results anti-CD19-CAR5-T cells was lower than that of anti-CD19-CAR4-T cells showed that the percentages of CD8+ CD107a+ T cells of anti-CD19- although they were not significantly different (p = 0.0961). Neverthe­ CAR4-T cells and anti-CD19-CAR5-T cells were slightly higher but not less, the level of IL-6 in culture media of anti-CD19-CAR4-T cells when significantly different, compared to that of the mock-T cells (Supple­ exposed to Raji-PD-L1 was 9.66 ± 2.38 pg/ml, which was significantly mentary Fig. 5A-B). The levels of perforin, granzyme A, granzyme B, higher than those of the mock-T cells (2.93 ± 0.41 pg/ml, p = 0.0147) granulysin, and cytokines (altogether 13 cytokines and proteins) were and anti-CD19-CAR5-T cells (3.92 ± 0.43 pg/ml, p = 0.0356), while the simultaneously analyzed by the Cytokine Bead Array (CBA) of the levels of IL-6 in culture media of the two latter were not significantly LEGENDplexTM Human CD8/NK cell panel. The levels of perforin in the different (p = 0.8756). culture media of mock-T cells, anti-CD19-CAR4-T cells, and anti-CD19- CAR5-T cells were not significantly different, when they exposed to Raji 3.5. Effects of secreted anti-PD-L1 scFv on T cell proliferation and cells (Fig. 4A). However, the levels of granzyme A, granzyme B, and cytotoxicity of anti-CD19-CAR4-T cells and anti-CD19-CAR5-T cells granulysin in the culture media of anti-CD19-CAR4-T cells after the co- against CD19+/PD-L1high tumor cells culturing with Raji cells were significantly increased, compared to that of mock-T cells (p = 0.0281, p < 0.0001, and p = 0.0227, respectively). The effects of secreted anti-PD-L1 scFv on T cell proliferation, and In contrast, only the level of granzyme B, but not granzyme A and the cytotoxicity of anti-CD19-CAR4-T cells and anti-CD19-CAR5-T cells granulysin, in the culture media of anti-CD19-CAR5-T cells was signif­ were evaluated by flow cytometric methods. Anti-CD19-CAR4-T cells or icantly increased, when the cells exposed to Raji cells (Fig. 4A). The anti-CD19-CAR5-T cells were co-cultured with CMFDA-labeled Raji cells levels of perforin, granzyme A, grazyme B, and granulysin in the culture or CMFDA-labeled Raji-PD-L1 cells at an E:T ratio of 1:1 for 24, 48, and media of anti-CD19-CAR4-T cells and anti-CD19-CAR5-T cells, when 72 h (Supplementary Fig. 6). After counting beads were added, pro­ they exposed to Raji-PD-L1, were signigicantly increased, compared to liferation of effector CAR T cells was analyzed by flow cytometry. The those of mock-T cells (Fig. 4B). These indicates that anti-PD-L1 scFv absolute numbers of effector CAR-T cells and CMFDA-labeled target cells secreted by anti-CD19-CAR5-T cells did not affect secretion of perforin, were determined by their fluorescence intensities. The results showed granzyme A/B, and granulysin, when the cells exposed to cancer cells relative increases in cell proliferation in fold-changes in a time- expressing PD-L1. dependent manner (Fig. 5A-B). After co-culturing with CMFDA- labeled Raji cells for 72 h, the proliferation of anti-CD19-CAR4-T cells The cytokines and proteins that were secreted by CAR-T cells, which was 4.64 ± 0.56, which was not significantly different from that of play roles in CAR-T cell survival and cytolytic activity, were also mock-T cells (3.66 ± 0.6, p = 0.26). However, the proliferation of anti- measured by the Cytokine Bead Array (CBA) method. The results 9

P. Yuti et al. International Immunopharmacology 113 (2022) 109442 Fig. 4. Measurements of granulytic proteins/enzymes and cytokines secreted from anti-CD19-CAR4-T cells and anti-CD19-CAR5-T cells after exposure to Raji and Raji-PD-L1 cell lines. Granulytic proteins/enzymes, including perforin, granzyme A, granzyme B, and granulysin, were measured in the culture media of mock-T cells, anti-CD19-CAR4-T cells, and anti-CD19-CAR5-T cells, exposed to (A) Raji cells or (B) Raji-PD-L1 cells. The levels of anti-tumor cytokines, including interleukin-2 (IL- 2), interferon gamma (IFN-γ), and tumor necrosis factor alpha (TNF-α), secreted in the culture media of mock-T cells, anti-CD19-CAR4-T cells, and anti-CD19-CAR5-T cells after exposure to (C) Raji cells or (D) Raji-PD-L1 cells were similarly measured. (E) Proinflammatory cytokine (IL-6) was also measured in the culture media of the three cells exposed to either of the two target cancer cells. All experiments were conducted by setting the effector to target ratio of 5:1, which were co-cultured for 24 h, and the culture media were analysed by the Cytokine Bead Array (CBA) method using the LEGENDplexTM Human CD8/NK cell panel. The results are sum­ marized as mean ± standard error of the mean (SEM) from the datasets of at least three independent experiments. Statistical differences were determined by Student’s t-test or two-way analysis of variance (ANOVA), and asterisks indicate the significant p-values as follows: *p < 0.05 **p < 0.01, ***p < 0.001, and ****p < 0.0001. (Abbreviation: CAR, chimeric antigen receptor; CD, cluster of differentiation). 10

P. Yuti et al. International Immunopharmacology 113 (2022) 109442 Fig. 5. Anti-PD-L1 scFv secreted from anti-CD19-CAR5-T cells promoted T cell proliferation and increased the cytotoxicity of anti-CD19-CAR4-T cells and anti-CD19- CAR5-T cells. (A) Time-dependent proliferation of anti-CD19-CAR4-T cells and anti-CD19-CAR5-T cells co-cultured with Raji cells or (B) Raji-PD-L1 cells at an effector to target (E:T) ratio of 1:1. (C) Dose-dependent cytotoxicity of anti-CD19-CAR4-T cells and anti-CD19-CAR5-T cells against Raji cells or (D) Raji-PD-L1 cells at E:T ratios of 0.5:1, 1:1, and 5:1 after co-culturing for 24 h (h). The results are summarized as mean ± standard error of the mean (SEM) from at least 3 independent experiments. Statistical differences were determined by two-way analysis of variance (ANOVA), and asterisks indicate the signigicant p-values as follows: *p < 0.05, **p < 0.01. (Abbreviations: CAR, chimeric antigen receptor; CD, cluster of differentiation; ns, not significant). CD19-CAR5-T cells was 5.39 ± 0.64, which was significantly higher cells at E:T ratios of 1:1 and 5:1 (p = 0.9985, p = 0.9997, respec­ than that of mock-T cells (p = 0.02). Interestingly, after co-culturing tively). Interestingly, in the co-culture with the Raji-PD-L1 cells at the E: with CMFDA-labeled Raji-PD-L1 cells for 72 h, only the proliferation T ratio of 0.5:1, the cytotoxicity of anti-CD19-CAR5-T cells was signifi­ of anti-CD19-CAR5-T cells (5.77 ± 0.66) was significantly higher than cantly higher than that of anti-CD19-CAR4-T cells (49.0 ± 8.4 % vs 30.5 that of mock-T cells (3.45 ± 0.55, p = 0.0038), and also significantly ± 2.5 %, p = 0.0497) (Fig. 5D). These data indicate that, at a low E:T higher than that of anti-CD19-CAR4-T cells (4.01 ± 0.56, p = 0.027). ratio (0.5:1), anti-PD-L1 scFv secreted from anti-CD19-CAR5-T cells These results indicate that the secreted anti-PD-L1 scFv that would could alleviate the PD-L1 mediated CAR-T cell inhibition as observed interfere with PD-1/PD-L1 interaction could promote proliferation of when anti-CD19-CAR4-T cells were examined. anti-CD19-CAR5-T cells. 3.6. Effects of secreted anti-PD-L1 scFv on maintaining cytotoxicity of The cytotoxicity of anti-CD19-CAR4-T cells and anti-CD19-CAR5-T anti-CD19-CAR5-T cells cells co-cultured with CMFDA-labeled Raji cells or CMFDA-labeled Raji-PD-L1 cells at E:T ratios of 0.5:1, 1:1, and 5:1 for 24 h were also Since the cytotoxic effect of anti-CD19-CAR5-T cells against Raji-PD- evaluated by flow cytometric analysis (Fig. 5C-D). The cytotoxicities of L1 cells was higher than that of anti-CD19-CAR4-T cells at the low E:T anti-CD19-CAR4-T cells against the Raji cells were 38.2 ± 6.2 %, 69.6 ± ratio (0.5:1), when they were co-cultured for 24 h. We wondered 3.5 %, and 82.2 ± 5.3 %, respectively. Similarly, the cytotoxicities of whether this effect would also be observed at the E:T ratio of 1:1 when anti-CD19-CAR5-T cells against the Raji cells were 49.6 ± 3.6 %, 63.9 ± they were co-cultured for 48 h. The results of this experiment showed 6.5 %, and 76.9 ± 5.5 %, respectively (Fig. 5C). The cytotoxicities of that the cytotoxicity of mock-T cells against Raji cells was 27.8 ± 4.0 % anti-CD19-CAR4-T cells against the Raji-PD-L1 cells were 30.5 ± 2.5 %, while those of anti-CD19-CAR4-T cells and anti-CD19-CAR5-T cells were 49.4 ± 3.5 %, and 75.7 ± 4.3 %, respectively. The cytotoxicities of anti- 48.7 ± 3.1 % (p = 0.0149) and 54.5 ± 8.0 % (p = 0.0419), respectively, CD19-CAR5-T cells against the Raji-PD-L1 cells were 49.0 ± 8.4 %, 53.4 which both were significantly higher than that of the mock-T cells. ± 3.3 %, and 78.9 ± 4.7 %, respectively (Fig. 5D). However, the cyto­ However, the cytotoxicity of mock-T cells against Raji-PD-L1 cells was toxicities of anti-CD19-CAR4-T cells and anti-CD19-CAR5-T cells after 29.3 ± 6.4 % whereas that of anti-CD19-CAR4-T cells was 38.5 ± 4.7 %, co-culturing with Raji cells were similar, even after increasing the which was not significantly different (p = 0.3170). Interestingly, the effector to target ratios (p = 0.6412, p = 0.9914, p = 0.9948, respec­ cytotoxicity of anti-CD19-CAR5 T cells against Raji-PD-L1 cells was tively). These data indicate that secreted anti-PD-L1 scFv from anti- significantly higher (65.3 ± 5.3 %, p = 0.0200) (Fig. 6A), suggesting at CD19-CAR5-T cells did not interfere the cytotoxicities against tumor low E:T ratios the cytotoxic effects of anti-CD19-CAR5-T cells were cells expressing CD19+. Similar results were observed after these anti- sustained and observed at the longer period of co-culturing time. CD19-CAR-T cells exposed to the Raji-PD-L1 cells. The cytotoxicities of anti-CD19-CAR4-T cells were similar to that of anti-CD19-CAR5-T The expressions of PD-1 on effector cells (Fig. 6B) and PD-L1 on 11

P. Yuti et al. International Immunopharmacology 113 (2022) 109442 Fig. 6. Effect of secreted anti-PD-L1 scFv on enhancing cytotoxicity of anti-CD19-CAR5-T cells. (A) Cytotoxicities of anti-CD19-CAR4-T cells and anti-CD19-CAR5-T cells against Raji cells or Raji-PD-L1 cells at E:T ratio of 1:1 after co-culture for 48 h. (B) Percentage of PD-1 expression in the population of CD3+ effector T cells, and (C) Mean fluorescence intensity (MFI) of PD-L1 expression on target cancer cells after co-culture at E:T ratio of 1:1 for 48 h. (D) Cytotoxicity of anti-CD19-CAR4-T cells and anti-CD19-CAR5-T cells against Raji cells or Raji-PD-L1 cells at an E:T ratio of 0.5:1 after co-culture for 48 h. (E) Effect of secreted anti-PD-L1 scFv in condition medium (CM) on the cytotoxicity of anti-CD19-CAR4-T cells compared to anti-CD19-CAR4-T cells (without CM), and anti-CD19-CAR5-T cells against Raji- PD-L1 cells at E:T ratio of 0.5:1 after co-culture for 48 h. (F) Proliferations of anti-CD19-CAR4-T cells in the conditions with or without CM at E:T ratio of 1:1 after co- culture for 72 h. The results are summarized as mean ± standard error of the mean (SEM) from at least 3 independent experiments. Statistical differences were determined by Student’s t-test or two-way analysis of variance (ANOVA), and asterisks indicate the significant p-values as follows: *p < 0.05, **p < 0.01. (Ab­ breviations: CAR, chimeric antigen receptor; CD, cluster of differentiation). target cancer cells (Fig. 6C) were examined. PD-1 expression levels of However, the MFI levels of PD-L1 expression on Raji-PD-L1 cells were the mock-T cells, anti-CD19-CAR4-T cells, and anti-CD19-CAR5-T cells remarkedly increased after co-culture with mock-T cells (34,399 ± were similar both in the conditions without target cells and co-culturing 2,926) and anti-CD19-CAR4-T cells (30,535 ± 616), but it was lower with Raji cells (Fig. 6B). Once anti-CD19-CAR4-T cells were exposed to after co-culture with anti-CD19-CAR5-T cells (27,682 ± 2,408), which Raji-PD-L1 cells, the PD-1 level was significantly increased (32.8 ± 3.8 was significantly lower than that of the mock-T cells (p = 0.0485). The %) when compared to that of the mock-T cells (24.0 ± 4.1 %, p = upregulations of PD-1 on anti-CD19-CAR4-T cells and PD-L1 on Raji-PD- 0.0311). In contrast, the PD-1 level of anti-CD19-CAR5-T cells was 23.3 L1 cells when they were co-cultured would possibly reduce the cytotoxic ± 1.6 %, similar to that of the mock-T cells, which was significantly effect of anti-CD19-CAR4-T cells but this effect did not occur when anti- lower (p = 0.0216), compared to that of anti-CD19-CAR4-T cells. CD19-CAR5-T cells secreting anti-PD-L1 scFv were co-cultured with Raji-PD-L1 cells. The anti-PD-L1 scFv from anti-CD19-CAR5-T cells The expressions of PD-L1 on Raji and Raji-PD-L1 cells after exposures could affect the expressions of both PD-1 on the effector cells and PD-L1 to mock-T cells, anti-CD19-CAR4-T cells or anti-CD19-CAR5-T cells were on the target cancer cells. also examined. The mean fluorescent intensity (MFI) levels of PD-L1 expression on Raji cells did not alter after co-culture with the mock-T To focus on the levels of cytotoxicity at the lower E:T ratio in the cells, anti-CD19-CAR4-T cells, and anti-CD19-CAR5-T cells (Fig. 6C). extended co-culturing time, the cytotoxicities of anti-CD19-CAR4-T cells 12

P. Yuti et al. International Immunopharmacology 113 (2022) 109442 and anti-CD19-CAR5-T cells on Raji cells or Raji-PD-L1 cells were major problem that causing reduced CAR-T cell efficacy [25,27]. Pre­ investigated at the E:T ratio of 0.5:1 for 48 h (Fig. 6D and Supple­ vious studies revealed that a high PD-L1 level suppressed tumor- mentary Figure 7). The results showed that in the co-culture with Raji associated T cell activities, such as T cell activation, T cell prolifera­ cells, the cytotoxicity of anti-CD19-CAR4-T cells (43.8 ± 5.1 %) was tion, and IFN-γ secretion [31,47,48]. Several studies, thus, attempted to similar to that of anti-CD19-CAR5-T cells (46.4 ± 4.9 %), whereas co- block the PD-1/PD-L1 interaction to improve T cell functions culturing with the Raji-PD-L1 cells, the cytotoxicity of anti-CD19- [47,49,50]. Preclinical studies showed the antitumor efficiency of CAR- CAR5-T cells (51.6 ± 2.8 %) was significantly higher than that of both T cells was increased when they were combined with immune check­ mock-T cells (24.8 ± 4.2 %, p = 0.001) and anti-CD19-CAR4-T cells point blockades (ICBs), such as CAR-T cells secreting full-length anti-PD- (23.8 ± 5.2 %, p = 0.036). Therefore, the cytotoxicity of anti-CD19- L1 antibody [34,36] or anti-PD-1 antibody. Previous BCL studies CAR5-T cells was markedly increased over that of anti-CD19-CAR4-T demonstrated that anti-PD-1 scFv secreted by anti-CD19-CAR-T cells cells when co-cultured with the Raji-PD-L1 cells at this low E:T ratio. [35] or anti-CD20-CAR-T cells [51] or dominant-negative PD-1 armored This data indicates that the cytotoxic activity of anti-CD19-CAR4-T cells anti-CD19 CAR T cells [49] could enhance CAR-T cell antitumor func­ was suppressed by PD-1/PD-L1 interaction. tions and improve the survival of CAR-T cells and tumor-specific T cells in vivo. However, a disruption of PD-1/PD-L1 signaling by engineering Next, we investigated whether the cytotoxic activity of anti-CD19- anti-CD19-CAR-T co-expressed with PD-1-specific VHH domain of anti- CAR5-T cells at this low E:T ratio was improved by the effect of anti- PD-1 nanobody affected CAR-T cell survival and also diminished CAR-T PD-L1 scFv. Culture medium (CM) containing secreted anti-PD-L1 scFv function, which it was suggested that PD-1 downregulation should not was collected from the culture of anti-CD19-CAR5-T cells at 48 h be considered as the way to improve the quality of therapeutic CAR-T (Supplementary Figure 8). This condition medium was added into the cells [50]. Several evidences supporting that therapeutic anti-PD-L1 co-culture of anti-CD19-CAR4-T cells and Raji-PD-L1 cells to compare antibodies were more effective than that of anti-PD-1 antibodies, the cytotoxicity with that of anti-CD19-CAR5-T cells co-cultured with which were demonstrated by in vitro study, in vivo mouse model, and in Raji-PD-L1 cells (Fig. 6E). The results showed that the cytotoxicity of asymptomatic multiple myeloma (AMM) patients [38–40]. An in vitro anti-CD19-CAR4-T cells with the addition of condition medium con­ functional assay using an efficient T cell reporter platform revealed that taining secreted anti-PD-L1 scFv against the Raji-PD-L1 cells was PD-L1 antibodies were superior to PD-1 antibodies in reverting PD-1 significantly higher (45.7 ± 2.7 %) than that of anti-CD19-CAR4-T cells signaling [38]. Moreover, humanized mice received PD-L1 blockade, without the addition of condition medium (27.4 ± 5.4 %, p = 0.040). but not PD-1 blockade, could restore CAR T cell activities by directly This higher cytotoxicity is similar to that of anti-CD19-CAR5-T cells inhibiting M2 macrophages-induced immunosuppression [40]. The ef­ secreting anti-PD-L1 scFv (53.2 ± 3.4 %, p = 0.161). Furthermore, the fect of anti-PD-L1, atezolizumab, antibody to enhance DC maturation direct effect of secreted anti-PD-L1 scFv on T cell proliferation was also leading to greater antigen-specific T cell expansion was observed in investigated (Fig. 6F and Supplementary Figure 9). The results showed patients with AMM [39]. These evidences suggest that tumor-specific that CM containing anti-PD-L1 scFv did not affect anti-CD19-CAR4-T CAR-T cells secreting anti-PD-L1 scFv are likely to enhance therapeu­ cells when they were co-cultured with Raji cells (p = 0.2127). Howev­ tic efficacy. er, the CM containing anti-PD-L1 scFv could increase proliferation of anti-CD19-CAR4-T cells when they were co-cultured with Raji-PD-L1 To overcome PD-L1-mediated T cell inhibition, we generated anti- cells (p = 0.0198). These data indicate that the anti-PD-L1 scFv CD19-CAR5-T cells that could locally secrete anti-PD-L1 scFv to block secreted from anti-CD19-CAR5-T cells could block PD-L1 on Raji-PD-L1 PD-1/PD-L1 interaction and enhance CAR-T cell cytotoxic functions. cells, which restore the cytotoxic activity of anti-CD19-CAR4-T cells and Our group previously developed CAR4-T cells and demonstrated their promote cell proliferation. higher efficiency compared to those of CAR2-T cells and CAR3-T cells against hematopoietic malignancy [24] and solid tumors [20–23]. In the 4. Discussion present study, we designed the lentiviral construct to drive the expres­ sion of anti-CD19-CAR4 secreting anti-PD-L1 scFv, which is referred to The current standard chemotherapy treatment for aggressive B cell as anti-CD19-CAR5 (Fig. 1A). The designed anti-CD19-CAR5 construct lymphomas (BCL) is not satisfactory because almost half of patients do could express CAR and anti-PD-L1 scFv proteins (Fig. 1B-1D), which not respond to treatment or they relapse after treatment [2]. Adoptive T were cleaved at the inserted T2A peptide by an intracellular protease cell therapy using second-generation anti-CD19-CAR-T cell products has enzyme. The anti-PD-L1 scFv protein was able to secrete into the culture been approved by the US FDA for refractory aggressive BCL [6,7] and medium (Fig. 1B and 1E) and exert bioactivities since it could bind to mantle cell lymphoma [8]. Clinical trials using the anti-CD19-CAR2-T the PD-L1 protein expressed on tumor cells, and also inhibit the binding cell treatment reported an overall response rate (ORR) of 52 %-80 %, of the anti-PD-L1 monoclonal antibody to the PD-L1 protein (Fig. 1E-G). and a complete response rate (CRR) of 40 %-55 % in patients with diffuse large B cell lymphoma (DLBCL) [41]. A clinical trial using third- Furthermore, anti-CD19-CAR5-T cells were successfully generated generation anti-CD19-CAR-T cells demonstrated better expansion and from primary human lymphocytes that showed a similar level of CAR persistence of these CAR-T cells than those of second-generation CAR in protein expression as that of anti-CD19-CAR4-T cells (Fig. 2A-B). The patients with relapse or refractory non-Hodgkin lymphomas (r/r-NHL) major proportion of the generated anti-CD19-CAR5-T cells were cyto­ [18]. A recent phase I clinical trial in relapse or refractory non-Hodgkin toxic T cells (CD3+/CD8+) (Fig. 2C). The exhaustion profiles of both lymphomas (r/r-NHL) using anti-CD19-CAR-T cells harboring CD28/ anti-CD19-CAR4-T cells and anti-CD19-CAR5-T cells during CAR-T CD27ζ and inducible caspase 9 (iCas9) reported an ORR of 67 %, and a production were similar to that of PHA-L activated T cells (except for partial response rate (PRR) of 24 % [19]. The results of these clinical CD3+/LAG3+), and mock-T cells (Fig. 2D), which suggests that anti- studies suggest that the problems of treatment resistance and disease CD19-CAR5-T cell production did not significantly alter the pheno­ recurrence still exist. Therefore, improvement to enhance the efficiency types of T cells. Additionally, central memory T cell subtype (Tcm) was a of adoptive T cell therapy using CAR-T therapy is urgently needed. major population in both anti-CD19-CAR4-T cells and anti-CD19-CAR5- T cells (Fig. 2E). This may be explained by the fact that IL-2, IL-7, and IL- PD-L1 is abundantly expressed in human B cell malignancies, 15 cytokines were added into the culture media of CAR-T cells, which including B cell lymphoma [42], non-Hodgkin lymphomas (NHL) [31], promotes less differentiation of phenotypes [52]. These results indicate mantle cell lymphoma, acute monocyte leukemia [43]. It is also highly that our production protocol encouraged the functionality of both anti- expressed on macrophages, some activated T and B cells, and myeloid- CD19-CAR4-T cells and anti-CD19-CAR5-T cells by maintaining cyto­ derived suppressor cells (MDSCs) [44,45]. The upregulation of PD-L1 toxic T cells (CD3+/CD8+) and Tcm phenotypes, which may help to could be mediated by the IFN-γ cytokine [46], resulting in impaired T promote CAR-T persistence [53] and cytotoxic function. cell antitumor functions. Increased PD-L1 expression on tumor cells is a The mFMC63 was reported to have high affinity for binding to the 13

P. Yuti et al. International Immunopharmacology 113 (2022) 109442 CD19 molecule [54]. The specific binding ability of mFMC63 scFv was the cytotoxicity assay to achieve high levels of tumor cell lysis because, still maintained after it was fused to the construct CAR molecule. The at a low E:T ratio, CAR-T cell exhaustion would occur, resulting in specific killing abilities of anti-CD19-CAR4-T cells and anti-CD19-CAR5- reduced tumor cell lysis. The results of the present study demonstrated T cells against HeLa-CD19+ cells were evaluated, and were found to be that anti-CD19-CAR5-T cells and anti-CD19-CAR4-T cells exerted similar (Fig. 3A-B). Since lymphoma is a type of solid tumor, physical similar cytotoxicity against Raji-PD-L1 cells when they were co-cultured barrier and the tumor microenvironment limit the ability of CAR-T cells at high E:T ratios (Fig. 5D). The similar cytotoxicities of these CAR-T to access the tumor site [55]. To mimic some physical barrier charac­ cells may be due to that both effector cells have reached their teristics of a solid tumor, we generated 3D spheroid cultures of parental threshold of killing capacities. In addition, we also observed the high HeLa cells and HeLa-CD19+ cells for testing with anti-CD19-CAR4-T levels of non-specific killing of mock-T cells against Raji cells and Raji- cells and anti-CD19-CAR5-T cells in co-culture systems that presented PD-L1 cells at high E:T ratio of 5:1. These non-specific killing might cell-cell and cell-extracellular matrix interaction [21]. Our results abolish the difference of specific killing of anti-CD19-CAR5-T cells and demonstrated that both anti-CD19-CAR4 T cells and anti-CD19-CAR5-T anti-CD19-CAR4-T cells. Moreover, we found that, in this co-culture cells could infiltrate and specifically kill HeLa-CD19+ spheroids in a condition, both PD-1 in anti-CD19-CAR4-T cells and PD-L1 in Raji-PD- time-dependent manner (Fig. 3C-F). The cytotoxicities of anti-CD19- L1 cells were upregulated (Fig. 6A-C). These findings were similar to CAR4 T cells and anti-CD19-CAR5-T cells in HeLa-CD19+ spheroid tu­ our previous report that anti-GD2-CAR-T cells upregulated PD-1 mors were unexpectedly similar. One explanation is that HeLa-CD19+ expression after exposure to target cancer cells resulting in the sup­ spheroid expressed a low level of PD-L1 protein on their cell surface pression of CAR-T cell functions [27]. To prevent CAR-T cell exhaustion, (Supplementary Fig. 3B). Accordingly, the effect of anti-PD-L1 scFv anti-PD-1 antibody could be added in combination with the CAR-T cell secreted from anti-CD19-CAR5-T cells to block PD-L1 protein and treatment [57]. Similar to PD-1/PD-L1 axis blockade, our data indicate enhance their cytotoxicity against HeLa-CD19+ spheroid could not be that a low number of anti-CD19-CAR5-T cells that could secrete anti-PD- demonstrated. Thus, increase of co-culturing time or use of HeLa-CD19+ L1 scFv were able to kill the target Raji-PD-L1 cells (Fig. 6D). Thus, stably expressing high level of PD-L1 for generation of 3D spheroid tu­ when the PD-1/PD-L1 interaction was blocked, anti-CD19-CAR5-T cells mors may be more suitable for evaluation of this cytotoxic function. We, could efficiently kill the target cancer cells expressing a high level of PD- therefore, decided to change to engineer Raji B-cell lymphoma stably L1. expressing PD-L1 protein for using in the next experiments to examine cytokine release, T cell proliferation, and cytotoxic activity. Moreover, we also demonstrated the cytotoxicities of anti-CD19- CAR4-T cells and anti-CD19-CAR5-T cells against Raji cells without When anti-CD19-CAR-T cells recognized CD19 antigen expressed on PD-L1 expression were similar, but the cytotoxicity of anti-CD19-CAR4- Raji cells, the CAR-T cells would be activated and release granulytic T cells against Raji cells expressing high PD-L1 was clearly reduced protein/enzymes and cytokines, which in turn triggered apoptosis of the (Fig. 6D). However, the reduced cytotoxicity of anti-CD19-CAR4-T cells cancer cells. The levels of granzyme A, granzyme B, and granulysin against Raji-PD-L1 cells could be restored by adding condition medium secretion from anti-CD19-CAR4-T cells and anti-CD19-CAR5-T cells (CM) containing anti-PD-L1 scFv secreted from anti-CD19-CAR5-T cells were significantly increased after these cells were exposed to the CD19+ (Fig. 6E). In addition, in the presence of CM, the proliferation of anti- Raji cells (Fig. 4A-B). Perforin secretion was not significantly increased CD19-CAR4-T cells was also increased (Fig. 6F). when the two anti-CD19-CAR-T cells were exposed to parental Raji cells but it was similarly increased when both anti-CD19-CAR-T cells were The potential benefits of anti-CD19-CAR5-T cells include effective exposed to Raji-PD-L1 cells. The levels of IL-2, IFN-γ and TNF-α cytokine cytotoxicity against cancer cells expressing both CD19 and PD-L1, in secretion from anti-CD19-CAR4-T cells and anti-CD19-CAR5-T cells which the cytotoxicity eventuates at low effector to target ratio. The were significantly and similarly increased, compared to those of the unique characteristics that anti-CD19-CAR5-T cells generate low IL-6 mock-T cells (Fig. 4C-D). These findings indicate that the anti-CD19- and the requirement of using a low number of anti-CD19-CAR5-T cells CAR5-T cells did not highly increase cytokine secretion, which is will be likely to cause less cytokine-mediated toxicity. Comparing to the different from the result previously reported by Lal, et al. [36] that anti- use of combined CAR-T cells and immune checkpoint blockade (ICB), mesothelin-CAR-T cells secreting anti-PD-L1-scFv-Fc antibody had which requires several cycles of costly ICB infusions [32] that may cause dramatically increased the production of IL-2, IFN-γ, and TNF-α cyto­ immune-related adverse effects, the use of anti-CD19-CAR5-T cells may kines when these cells exposed to the target cancer cells at a high E:T minimize the adverse effect and also reduce cost of treatment. However, ratio of 10:1. IL-6 is a proinflammatory cytokine that plays roles in it is possible that CAR-T cells that worked well in preclinical in vitro immune response and inflammation. In addition, high serum levels of IL- studies might fail to provide substantial killing of target cancer cells in 6 in CAR T cell treatment are associated with cytokine release syndrome vivo. Accordingly, the safety and efficacy of using anti-CD19-CAR5-T (CRS) [56]. Interestingly, it is important to note that the levels of IL-6 cells for treatment of aggressive BCL should be further evaluated by in secretion from anti-CD19-CAR5-T cells, exposed to both Raji cells and vivo studies and clinical trials. Thus, lacking of in vivo antitumor efficacy Raji-PD-L1 cells, were significantly lower than those of anti-CD19- data of anti-CD19-CAR5-T cells is a major limitation of the present CAR4-T cells (Fig. 4E). Thus, anti-PD-L1 scFv secreted from anti- study. Nevertheless, this study provides crucial proof-of-principal data CD19-CAR5-T cells could somehow reduce IL-6 secretion. anti-CD19- supporting that simultaneously secreted anti-PD-L1 scFv from anti- CAR5-T cells exhibit high cytotoxic activity against target tumor cells CD19-CAR5-T cells is a feasible strategy to improve CAR-T antitumor expressing high PD-L1 may be beneficial for reducing the side effect of efficiency prominently against PD-L1-positive tumors. cytokine release syndrome (CRS) when these anti-CD19-CAR5-T cells are used in patients. Taken together, the results of this study yield substantial evidence demonstrating that: (i) anti-CD19-CAR5-T cells exerted more effective Anti-CD19-CAR5-T cells demonstrated greater proliferation cytotoxicity than that of anti-CD19-CAR4-T cells; (ii) the secreted anti- compared to that of anti-CD19-CAR4-T cells when they were exposed to PD-L1 scFv promoted self-proliferation of anti-CD19-CAR5-T cells; (iii) both Raji cells and Raji-PD-L1 cells (Fig. 5A-B). This finding was also the secreted anti-PD-L1 scFv could restore the cytotoxic effect of anti- observed in a study of CAR-T cells secreting anti-PD-L1 scFv [34]. Thus, CD19-CAR4-T cells that was inhibited by PD-L1 expression on the the blockade of PD-1/PD-L1 interaction could promote CAR-T cell target cancer cells; (iv) anti-CD19-CAR5-T cells released comparatively function and survival because it could prevent CAR-T cell exhaustion lower level of proinflammatory IL-6 cytokine that may cause less [34] and increase CAR-T cell expansion [37]. cytokine-mediated toxicity; and, (iv) a comparatively lower number of anti-CD19-CAR5-T cells used could induce cytotoxicity against target Interestingly, at the low effector-to-target (E:T) ratio of 0.5:1, anti- cancer cells expressing PD-L1. This study demonstrated that PD-L1 CD19-CAR5-T cells were more efficient than anti-CD19-CAR4-T cells blockade mediated by anti-PD-L1 scFv secreted from anti-CD19-CAR5- in killing Raji-PD-L1 cells. Normally, high E:T ratios were employed in T cells can improve CAR-T antitumor efficiency prominently against 14

P. Yuti et al. International Immunopharmacology 113 (2022) 109442 PD-L1-positive tumors. However, the anti-CD19-CAR5-T cells may also donors that participated in this study; Dr. Sansanee Noisakran from the be effective against PD-L1-negative tumors mediated through the inhi­ National Center of Genetic Engineering and Biotechnology (BIOTECH) bition of M2 macrophages. Further studies are needed to compare the for providing the Raji cell line; and, Assoc. Prof. Dr. Naravat Poungvarin efficacy and safety of anti-CD19-CAR4-T cells and anti-CD19-CAR5-T of the Department of Clinical Pathology, Faculty of Medicine Siriraj cells by in vitro co-culture with macrophages or by using humanized Hospital, Mahidol University for providing the lentiviral packaging mice model, after which anti-CD19-CAR5-T cells should be evaluated in plasmids. animal model and human clinical trials. Appendix A. Supplementary data 5. Authors’ contributions Supplementary data to this article can be found online at https://doi. PYu, YW, and JS designed and conducted the experiments, analyzed org/10.1016/j.intimp.2022.109442. the data, interpreted the results, prepared the figures, and drafted the manuscript. NP performed the three-dimensional spheroid killing assay. References NS, KK, KC, and TC generated the plasmid constructs, prepared the plasmids, and produced the lentiviruses. AP and MJ provided materials [1] R. Kuppers, Mechanisms of B-cell lymphoma pathogenesis, Nat. Rev. Cancer 5 (4) and reagents. JS and PYe conceptualized and managed the study, su­ (2005) 251–262. pervised the experiments, and edited the manuscript. All authors have read and approved the final version of the manuscript submitted for [2] A. Galaznik, C. Reich, G. Klebanov, Y. Khoma, E. Allakhverdiiev, G. Hather, et al., journal publication. Predicting Outcomes in Patients With Diffuse Large B-Cell Lymphoma Treated With Standard of Care, Cancer Inform. 18 (2019), 1176935119835538. 6. Funding disclosure [3] E. Van Den Neste, N. Schmitz, N. Mounier, D. Gill, D. Linch, M. Trneny, et al., This study was financially supported by grants from the National Outcome of patients with relapsed diffuse large B-cell lymphoma who fail second- Science and Technology Development Agency (NSTDA) (grant number line salvage regimens in the International CORAL study, Bone Marrow Transplant. P1650727); Mahidol University (grant numbers R016110006 and 51 (1) (2016) 51–57. R016210017); the Program Management Unit for National Competi­ tiveness Enhancement (PMU-C) under the Office of National Higher [4] U. Greenbaum, K.M. Mahadeo, P. Kebriaei, E.J. Shpall, N.Y. Saini, Chimeric Education Science Research and Innovation Policy Council (NXPO) Antigen Receptor T-Cells in B-Acute Lymphoblastic Leukemia: State of the Art and (grant number C10F630063); Siriraj Research Fund of the Faculty of Future Directions, Front. Oncol. 10 (2020) 1594. Medicine Siriraj Hospital, (grant number R016034008); the Interna­ tional Research Network (IRN) and the Thailand Research Fund (TRF) [5] J. Zhang, L.J. Medeiros, K.H. Young, Cancer Immunotherapy in Diffuse Large B- (grant number IRN58W0001). YW was supported by an International Cell Lymphoma, Front. Oncol. 8 (2018) 351. Research Network (IRN)-Ph.D. Scholarship (IRN5801PHDW01). PYu was supported by the Faculty of Medicine Siriraj Hospital, Mahidol [6] S.L. Maude, N. Frey, P.A. Shaw, R. Aplenc, D.M. Barrett, N.J. Bunin, et al., Chimeric University. antigen receptor T cells for sustained remissions in leukemia, N. Engl. J. Med. 371 (16) (2014) 1507–1517. 7. Ethics statement [7] N. Bouchkouj, Y.L. Kasamon, R.A. de Claro, B. George, X. Lin, S. Lee, et al., FDA The use of human blood samples from healthy donor volunteers for Approval Summary: Axicabtagene Ciloleucel for Relapsed or Refractory Large B- experimental studies was approved by the Siriraj Institutional Review cell Lymphoma, Clin. Cancer Res. 25 (6) (2019) 1702–1708. Board (SIRB) of the Faculty of Medicine Siriraj Hospital, Mahidol Uni­ versity (COA number: Si 762/2016). [8] M. Wang, J. Munoz, A. Goy, F.L. Locke, C.A. Jacobson, B.T. Hill, et al., KTE-X19 CAR T-Cell Therapy in Relapsed or Refractory Mantle-Cell Lymphoma, N. Engl. J. 8. Informed consent statement Med. 382 (14) (2020) 1331–1342. Written informed consent was obtained from all subjects involved in [9] J.S. Abramson, M.L. Palomba, L.I. Gordon, M.A. Lunning, M. Wang, J. Arnason, et the study. al., Lisocabtagene maraleucel for patients with relapsed or refractory large B-cell lymphomas (TRANSCEND NHL 001): a multicentre seamless design study, Lancet 9. Data availability statement 396 (10254) (2020) 839–852. Study data will be made available upon request to the corresponding [10] N.C. Munshi, L.D. Anderson Jr., N. Shah, D. Madduri, J. Berdeja, S. Lonial, et al., authors. Idecabtagene Vicleucel in Relapsed and Refractory Multiple Myeloma, N. Engl. J. Med. 384 (8) (2021) 705–716. Declaration of Competing Interest [11] W.H. Zhao, J. Liu, B.Y. Wang, Y.X. Chen, X.M. Cao, Y. Yang, et al., A phase 1, open- The authors declare that they have no known competing financial label study of LCAR-B38M, a chimeric antigen receptor T cell therapy directed interests or personal relationships that could have appeared to influence against B cell maturation antigen, in patients with relapsed or refractory multiple the work reported in this paper. myeloma, J Hematol Oncol. 11 (1) (2018) 141. Data availability [12] J. Jayaraman, M.P. Mellody, A.J. Hou, R.P. Desai, A.W. Fung, A.H.T. Pham, et al., CAR-T design: Elements and their synergistic function, EBioMedicine. 58 (2020), Data will be made available on request. 102931. Acknowledgements [13] O.U. Kawalekar, R.S. O’Connor, J.A. Fraietta, L. Guo, S.E. McGettigan, A. D. Posey Jr, et al., Distinct Signaling of Coreceptors Regulates Specific Metabolism The authors gratefully acknowledge the healthy volunteer blood Pathways and Impacts Memory Development in CAR T Cells, Immunity 44 (2) (2016) 380–390. [14] H. Chen, F. Wei, M. Yin, Q. Zhao, Z. Liu, B. Yu, et al., CD27 enhances the killing effect of CAR T cells targeting trophoblast cell surface antigen 2 in the treatment of solid tumors, Cancer Immunol. Immunother. 70 (7) (2021) 2059–2071. [15] D.G. Song, Q. Ye, M. Poussin, G.M. Harms, M. Figini, D.J. Powell Jr., CD27 costimulation augments the survival and antitumor activity of redirected human T cells in vivo, Blood 119 (3) (2012) 696–706. [16] K. Cappell, R.M. Sherry, J.C. Yang, S.L. Goff, D. Vanasse, L. McIntyre, et al., Long- term follow-up of anti-CD19 CAR T-cell therapy for B-cell lymphoma and chronic lymphocytic leukemia, J. Clin. Oncol. 38 (15_suppl) (2020) 3012. [17] Y. Ping, F. Li, S. Nan, D. Zhang, X. Shi, J. Shan, et al., Augmenting the Effectiveness of CAR-T Cells by Enhanced Self-Delivery of PD-1-Neutralizing scFv, Front. Cell Dev. Biol. 8 (2020) 803. [18] C.A. Ramos, R. Rouce, C.S. Robertson, A. Reyna, N. Narala, G. Vyas, et al., In Vivo Fate and Activity of Second- versus Third-Generation CD19-Specific CAR-T Cells in B Cell Non-Hodgkin’s Lymphomas, Mol. Ther. 26 (12) (2018) 2727–2737. [19] X. Zhou, S. Tu, C. Wang, R. Huang, L. Deng, C. Song, et al., Phase I Trial of Fourth- Generation Anti-CD19 Chimeric Antigen Receptor T Cells Against Relapsed or Refractory B Cell Non-Hodgkin Lymphomas, Front. Immunol. 11 (2020), 564099. [20] P. Luangwattananun, M. Junking, J. Sujjitjoon, Y. Wutti-In, N. Poungvarin, C. Thuwajit, et al., Fourth-generation chimeric antigen receptor T cells targeting folate receptor alpha antigen expressed on breast cancer cells for adoptive T cell therapy, Breast Cancer Res. Treat. 186 (1) (2021) 25–36. [21] N. Phanthaphol, C. Somboonpatarakun, K. Suwanchiwasiri, T. Chieochansin, J. Sujjitjoon, S. Wongkham, et al., Chimeric Antigen Receptor T Cells Targeting Integrin alphavbeta6 Expressed on Cholangiocarcinoma Cells, Front. Oncol. 11 (2021), 657868. 15

P. Yuti et al. International Immunopharmacology 113 (2022) 109442 [22] K. Supimon, T. Sangsuwannukul, J. Sujjitjoon, N. Phanthaphol, T. Chieochansin, [39] N. Bar, F. Costa, R. Das, A. Duffy, M. Samur, S. McCachren, et al., Differential N. Poungvarin, et al., Anti-mucin 1 chimeric antigen receptor T cells for adoptive T effects of PD-L1 versus PD-1 blockade on myeloid inflammation in human cancer. cell therapy of cholangiocarcinoma, Sci. Rep. 11 (1) (2021) 6276. JCI, Insight. 5 (12) (2020). [23] T. Sangsuwannukul, K. Supimon, J. Sujjitjoon, N. Phanthaphol, T. Chieochansin, [40] Y. Yamaguchi, J. Gibson, K. Ou, L.S. Lopez, R.H. Ng, N. Leggett, et al., PD-L1 N. Poungvarin, et al., Anti-tumour effect of the fourth-generation chimeric antigen blockade restores CAR T cell activity through IFN-gamma-regulation of CD163+ receptor T cells targeting CD133 against cholangiocarcinoma cells, Int. M2 macrophages, J. ImmunoTher. Cancer 10 (6) (2022). Immunopharmacol. 89 (Pt B) (2020), 107069. [41] S.S. Neelapu, F.L. Locke, N.L. Bartlett, L.J. Lekakis, D.B. Miklos, C.A. Jacobson, et [24] Y. Wutti-in, J. Sujjitjoon, N. Sawasdee, A. Panya, K. Kongkla, P. Yuti, et al., al., Axicabtagene Ciloleucel CAR T-Cell Therapy in Refractory Large B-Cell Development of a Novel Anti-CD19 CAR Containing a Fully Human scFv and Three Lymphoma, N. Engl. J. Med. 377 (26) (2017) 2531–2544. Costimulatory Domains. Frontiers, Oncology. (2022) 11. [42] J. Yang, G. Hu, Significance of PD-L1 in the diagnosis and treatment of B-cell [25] S. Rafiq, C.S. Hackett, R.J. Brentjens, Engineering strategies to overcome the malignant lymphoma, Oncol Lett. 17 (3) (2019) 3382–3386. current roadblocks in CAR T cell therapy, Nat. Rev. Clin. Oncol. 17 (3) (2020) 147–167. [43] X. Chen, S. Liu, L. Wang, W. Zhang, Y. Ji, X. Ma, Clinical significance of B7–H1 (PD-L1) expression in human acute leukemia, Cancer Biol. Ther. 7 (5) (2008) [26] T. Gargett, W. Yu, G. Dotti, E.S. Yvon, S.N. Christo, J.D. Hayball, et al., GD2- 622–627. specific CAR T Cells Undergo Potent Activation and Deletion Following Antigen Encounter but can be Protected From Activation-induced Cell Death by PD-1 [44] W. Xiao, J.D. Klement, C. Lu, M.L. Ibrahim, K. Liu, IFNAR1 Controls Autocrine Blockade, Mol. Ther. 24 (6) (2016) 1135–1149. Type I IFN Regulation of PD-L1 Expression in Myeloid-Derived Suppressor Cells, J. Immunol. 201 (1) (2018) 264–277. [27] J. Sujjitjoon, E. Sayour, S.T. Tsao, M. Uiprasertkul, K. Sanpakit, J. Buaboonnam, et al., GD2-specific chimeric antigen receptor-modified T cells targeting [45] M. Ballbach, A. Dannert, A. Singh, D.M. Siegmund, R. Handgretinger, L. Piali, et retinoblastoma - assessing tumor and T cell interaction, Transl. Oncol. 14 (2) al., Expression of checkpoint molecules on myeloid-derived suppressor cells, (2021), 100971. Immunol. Lett. 192 (2017) 1–6. [28] M. Cohen, A.G. Vistarop, F. Huaman, M. Narbaitz, F. Metrebian, E. De Matteo, et [46] K. Abiko, N. Matsumura, J. Hamanishi, N. Horikawa, R. Murakami, K. Yamaguchi, al., Cytotoxic response against Epstein Barr virus coexists with diffuse large B-cell et al., IFN-gamma from lymphocytes induces PD-L1 expression and promotes lymphoma tolerogenic microenvironment: clinical features and survival impact, progression of ovarian cancer, Br. J. Cancer 112 (9) (2015) 1501–1509. Sci. Rep. 7 (1) (2017) 10813. [47] Y. Iwai, M. Ishida, Y. Tanaka, T. Okazaki, T. Honjo, N. Minato, Involvement of PD- [29] J.N. Kochenderfer, M.E. Dudley, R.O. Carpenter, S.H. Kassim, J.J. Rose, W. L1 on tumor cells in the escape from host immune system and tumor G. Telford, et al., Donor-derived CD19-targeted T cells cause regression of immunotherapy by PD-L1 blockade, Proc. Natl. Acad. Sci. USA 99 (19) (2002) malignancy persisting after allogeneic hematopoietic stem cell transplantation, 12293–12297. Blood 122 (25) (2013) 4129–4139. [48] V.R. Juneja, K.A. McGuire, R.T. Manguso, M.W. LaFleur, N. Collins, W.N. Haining, [30] J. Kiyasu, H. Miyoshi, A. Hirata, F. Arakawa, A. Ichikawa, D. Niino, et al., et al., PD-L1 on tumor cells is sufficient for immune evasion in immunogenic Expression of programmed cell death ligand 1 is associated with poor overall tumors and inhibits CD8 T cell cytotoxicity, J. Exp. Med. 214 (4) (2017) 895–904. survival in patients with diffuse large B-cell lymphoma, Blood 126 (19) (2015) 2193–2201. [49] X. Liu, Y. Zhang, K. Li, Y. Liu, J. Xu, J. Ma, et al., A novel dominant-negative PD-1 armored anti-CD19 CAR T cell is safe and effective against refractory/relapsed B [31] D.J. Andorsky, R.E. Yamada, J. Said, G.S. Pinkus, D.J. Betting, J.M. Timmerman, cell lymphoma, Transl. Oncol. 14 (7) (2021), 101085. Programmed death ligand 1 is expressed by non-hodgkin lymphomas and inhibits the activity of tumor-associated T cells, Clin. Cancer Res. 17 (13) (2011) [50] R.S. Kalinin, V.M. Ukrainskaya, S.P. Chumakov, A.M. Moysenovich, V. 4232–4244. M. Tereshchuk, D.V. Volkov, et al., Engineered Removal of PD-1 From the Surface of CD19 CAR-T Cells Results in Increased Activation and Diminished Survival, [32] D. Modi, B. Potugari, J. Uberti, Immunotherapy for Diffuse Large B-Cell Front. Mol. Biosci. 8 (2021), 745286. Lymphoma: Current Landscape and Future Directions, Cancers (Basel) 13 (22) (2021). [51] M. Nakajima, Y. Sakoda, K. Adachi, H. Nagano, K. Tamada, Improved survival of chimeric antigen receptor-engineered T (CAR-T) and tumor-specific T cells caused [33] B. Geoerger, C.M. Zwaan, L.V. Marshall, J. Michon, F. Bourdeaut, M. Casanova, et by anti-programmed cell death protein 1 single-chain variable fragment-producing al., Atezolizumab for children and young adults with previously treated solid CAR-T cells, Cancer Sci. 110 (10) (2019) 3079–3088. tumours, non-Hodgkin lymphoma, and Hodgkin lymphoma (iMATRIX): a multicentre phase 1–2 study, Lancet Oncol. 21 (1) (2020) 134–144. [52] Y. Xu, M. Zhang, C.A. Ramos, A. Durett, E. Liu, O. Dakhova, et al., Closely related T-memory stem cells correlate with in vivo expansion of CAR.CD19-T cells and are [34] E.R. Suarez, K. Chang de, J. Sun, J. Sui, G.J. Freeman, S. Signoretti, et al., Chimeric preserved by IL-7 and IL-15, Blood 123 (24) (2014) 3750–3759. antigen receptor T cells secreting anti-PD-L1 antibodies more effectively regress renal cell carcinoma in a humanized mouse model, Oncotarget. 7 (23) (2016) [53] L. Gattinoni, E. Lugli, Y. Ji, Z. Pos, C.M. Paulos, M.F. Quigley, et al., A human 34341–34355. memory T cell subset with stem cell-like properties, Nat. Med. 17 (10) (2011) 1290–1297. [35] S. Rafiq, O.O. Yeku, H.J. Jackson, T.J. Purdon, D.G. van Leeuwen, D.J. Drakes, et al., Targeted delivery of a PD-1-blocking scFv by CAR-T cells enhances anti-tumor [54] S. Ghorashian, A.M. Kramer, S. Onuoha, G. Wright, J. Bartram, R. Richardson, et efficacy in vivo, Nat. Biotechnol. 36 (9) (2018) 847–856. al., Enhanced CAR T cell expansion and prolonged persistence in pediatric patients with ALL treated with a low-affinity CD19 CAR, Nat. Med. 25 (9) (2019) [36] S.W.G. Lal, Z. Chen, X. Xiang, C. Liang, X.F. Huang, Co-expression of PD-L1 1408–1414. Antibodies Enhances the Anti-Tumor Efficacy of Chimeric Antigen Receptor T Cells, J. Cancer Res. Immuno-oncology. 5 (2019) 1–9. [55] D.W. Scott, R.D. Gascoyne, The tumour microenvironment in B cell lymphomas, Nat. Rev. Cancer 14 (8) (2014) 517–534. [37] S. Li, N. Siriwon, X. Zhang, S. Yang, T. Jin, F. He, et al., Enhanced Cancer Immunotherapy by Chimeric Antigen Receptor-Modified T Cells Engineered to [56] M. Du, P. Hari, Y. Hu, H. Mei, Biomarkers in individualized management of Secrete Checkpoint Inhibitors, Clin. Cancer Res. 23 (22) (2017) 6982–6992. chimeric antigen receptor T cell therapy, Biomark Res. 8 (2020) 13. [38] L.A. De Sousa, C. Battin, S. Jutz, J. Leitner, C. Hafner, J. Tobias, et al., Therapeutic [57] L. Cherkassky, A. Morello, J. Villena-Vargas, Y. Feng, D.S. Dimitrov, D.R. Jones, et PD-L1 antibodies are more effective than PD-1 antibodies in blocking PD-1/PD-L1 al., Human CAR T cells with cell-intrinsic PD-1 checkpoint blockade resist tumor- signaling, Sci. Rep. 9 (1) (2019) 11472. mediated inhibition, J. Clin. Invest. 126 (8) (2016) 3130–3144. 16