key: cord-0029546-iryxp5td
authors: Liu, Xin; Xu, Yixiang; Xiong, Wei; Yin, Bingnan; Huang, Yuqian; Chu, Junjun; Xing, Changsheng; Qian, Chen; Du, Yang; Duan, Tianhao; Wang, Helen Y; Zhang, Ningyan; Yu, John S.; An, Zhiqiang; Wang, Rongfu
title: Development of a TCR-like antibody and chimeric antigen receptor against NY-ESO-1/HLA-A2 for cancer immunotherapy
date: 2022-03-25
journal: J Immunother Cancer
DOI: 10.1136/jitc-2021-004035
sha: 7122c3ee22545e6a32d55f40131a799034966a67
doc_id: 29546
cord_uid: iryxp5td

BACKGROUND: The current therapeutic antibodies and chimeric antigen receptor (CAR) T cells are capable of recognizing surface antigens, but not of intracellular proteins, thus limiting the target coverage for drug development. To mimic the feature of T-cell receptor (TCR) that recognizes the complex of major histocompatibility class I and peptide on the cell surface derived from the processed intracellular antigen, we used NY-ESO-1, a cancer-testis antigen, to develop a TCR-like fully human IgG1 antibody and its derivative, CAR-T cells, for cancer immunotherapy. METHODS: Human single-chain variable antibody fragment (scFv) phage library (~10(∧11)) was screened against HLA-A2/NY-ESO-1 (peptide 157–165) complex to obtain target-specific antibodies. The specificity and affinity of those antibodies were characterized by flow cytometry, ELISA, biolayer interferometry, and confocal imaging. The biological functions of CAR-T cells were evaluated against target tumor cells in vitro. In vivo antitumor activity was investigated in a triple-negative breast cancer (TNBC) model and primary melanoma tumor model in immunocompromised mice. RESULTS: Monoclonal antibody 2D2 identified from phage-displayed library specifically bound to NY-ESO-1(157-165) in the context of human leukocyte antigen HLA-A*02:01 but not to non-A2 or NY-ESO-1 negative cells. The second-generation CAR-T cells engineered from 2D2 specifically recognized and eliminated A2+/NY-ESO-1+tumor cells in vitro, inhibited tumor growth, and prolonged the overall survival of mice in TNBC and primary melanoma tumor model in vivo. CONCLUSIONS: This study showed the specificity of the antibody identified from human scFv phage library and demonstrated the potential antitumor activity by TCR-like CAR-T cells both in vitro and in vivo, warranting further preclinical and clinical evaluation of the TCR-like antibody in patients. The generation of TCR-like antibody and its CAR-T cells provides the state-of-the-art platform and proof-of-concept validation to broaden the scope of target antigen recognition and sheds light on the development of novel therapeutics for cancer immunotherapy.

ABSTRACT Background The current therapeutic antibodies and chimeric antigen receptor (CAR) T cells are capable of recognizing surface antigens, but not of intracellular proteins, thus limiting the target coverage for drug development. To mimic the feature of T-cell receptor (TCR) that recognizes the complex of major histocompatibility class I and peptide on the cell surface derived from the processed intracellular antigen, we used NY-ESO-1, a cancer-testis antigen, to develop a TCR-like fully human IgG1 antibody and its derivative, CAR-T cells, for cancer immunotherapy. Methods Human single-chain variable antibody fragment (scFv) phage library (~10 ∧11 ) was screened against HLA-A2/ NY-ESO-1 (peptide 157-165) complex to obtain targetspecific antibodies. The specificity and affinity of those antibodies were characterized by flow cytometry, ELISA, biolayer interferometry, and confocal imaging. The biological functions of CAR-T cells were evaluated against target tumor cells in vitro. In vivo antitumor activity was investigated in a triple-negative breast cancer (TNBC) model and primary melanoma tumor model in immunocompromised mice. Results Monoclonal antibody 2D2 identified from phagedisplayed library specifically bound to NY-ESO-1 157-165 in the context of human leukocyte antigen HLA-A*02:01 but not to non-A2 or NY-ESO-1 negative cells. The secondgeneration CAR-T cells engineered from 2D2 specifically recognized and eliminated A2+/NY-ESO-1+tumor cells in vitro, inhibited tumor growth, and prolonged the overall survival of mice in TNBC and primary melanoma tumor model in vivo. Conclusions This study showed the specificity of the antibody identified from human scFv phage library and demonstrated the potential antitumor activity by TCR-like CAR-T cells both in vitro and in vivo, warranting further preclinical and clinical evaluation of the TCR-like antibody in patients. The generation of TCR-like antibody and its CAR-T cells provides the state-of-the-art platform and proof-of-concept validation to broaden the scope of target antigen recognition and sheds light on the development of novel therapeutics for cancer immunotherapy.

Immunotherapies of therapeutic monoclonal antibodies are emerging as a major development to cancer treatment. 1 Similarly, chimeric antigen receptor (CAR) engineered T cells show impressive clinical benefits in leukemia and lymphoma. 2 3 However, due to a paucity of ideal and targetable tumorassociated antigens expressed on tumor surface, the application of CAR-T therapy has been severely limited in patients with solid cancers. [4] [5] [6] Despite the limited surface antigens, there are a large number of intracellular antigens that could serve as cancerspecific immune targets, which include cancer-testis antigens and mutation-derived neoantigens. 7 8 Cancer-testis antigens are considered promising target antigens for cancer immunotherapy. 8 9 New York Esophageal Squamous Cell Carcinoma-1 (NY-ESO-1) is not Open access expressed in normal tissues with exception of the testis but is highly expressed in many types of cancer, including melanoma, sarcoma, and multiple myeloma. [10] [11] [12] [13] Others and our studies show that NY-ESO-1 is one of the most immunogenic antigens recognized by T cells derived from patients with cancer. [14] [15] [16] [17] [18] [19] [20] [21] Recent studies show that immunotherapy using HLA-A2-restricted NY-ESO-1specific T cell receptor (A2-ESO-1 TCR) engineered T cells led to a 55% clinical response rate in metastatic synovial sarcoma and melanoma and 80% in myeloma without toxicity. [22] [23] [24] Furthermore, TCR-like antibody approaches have been developed to show that, like TCRs, antibodies could also recognize peptide/major histocompatibility complex (MHC) complexes on the cancer cell surface, thus opening the new opportunity to target intracellular cancer antigens that are processed and presented onto the cell surface by MHC molecules. Antigen-specific antibodies could be screened from synthetic human single-chain variable antibody fragment (scFv), Fab phage library, or hybridomas generated from immunized animals such as mice. 25 26 Those TCR-like monoclonal antibodies (mAbs) displayed highly specific binding affinity and killing ability in vitro and in vivo. 27 Moreover, mAbs can be readily further modified and optimized, and derivatives from TCR-like antibodies, such as antibodydrug conjugates, bispecific T cells engagers, and CAR-T cells have yielded impressive results. [28] [29] [30] [31] [32] [33] Here, we report the development of a TCR-like antibody mAb 2D2 that recognizes HLA-A2+/NY-ESO-1 157-165 peptide complexes and tumor cells with high specificity. CAR construct derived from scFv of 2D2 demonstrates its ability to eliminate target tumor cells. This is the first successful report of CAR-T cells targeting NY-ESO-1, with both in vitro and in vivo antitumor activity and specificity. Collectively, our data highlights a novel CAR-T technology and a new strategic application to immunotherapy. For the TNBC model, female mice were first challenged with 2 million MDA-MB-231-ESO tumor cells at fat pad. On day 4, mice were intravenously treated with 10 million control T cells or 2D2-CAR T cells suspended in 200 µL PBS, respectively, followed by three doses of 50,000 IU recombinant human IL-2 intraperitoneally. Mice were closely monitored, and tumor volume was measured via vernier caliper at indicated days.

For the melanoma model, 2 million Mel 1558 tumor cells were subcutaneously implanted at right flank of male NSG mice. A total of 2.5 million 2D2-CAR T cells were injected on day 4 followed by 50, 000 IU rhIL-2 intraperitoneally.

For safety assessment, NSG-A2 mice (n=6) were treated with the same dose used to achieve effective therapy (10 million T cells, intravenous injection). The in vivo toxicity was determined based on the body weight loss and histopathological microscopic evaluation of the major tissues. On day 7, three mice were euthanized by carbon dioxide (CO 2 ) inhalation, and tissue samples were collected. The remaining three mice were euthanized on day 14 for key tissues collection. Bloods were collected in mini lavender top tubes (Greiner, 450573VET) for complete blood count (CBC), while sera were collected in golden top clot activator serum separation tubes (Greiner, 450571VET) for superchem analysis (Antech Diagnostics).

HEK293T, MDA-MB-231-ESO1, and PC3-A2-ESO1 were cultured in DMEM supplied with 10% inactivated FBS and 100 unit/mL of penicillin and 100 µg/mL of streptomycin. T2 cells, Mel 586, Mel 624 and Mel 1558 were cultured with RPMI-1640 containing 10% inactivated FBS and 100 unit/mL of penicillin and 100 µg/mL of streptomycin. All cells were routinely tested as mycoplasma negative.

Panning of phage-displayed scFv antibody library Human scFv phage-display antibody library (~10 11 clones) was constructed in lab. 34 For peptide-MHC complex panning (synthesized by the NIH tetramer facility), an in-solution method was adopted to avoid the disadvantage of the immobilization method that may cause conformational changes of the pMHC complex. Briefly, the scFv phage was blocked with an equal amount of 10% milk for 1 hour at room temperature. One hundred microliters of streptavidin-conjugated Dynabeads M-280 (Thermo Fisher, 11 206D) were blocked with 5% milk at room temperature for 1 hour. The first round of negative selection against control biotinylated-pMHC (pp65 peptide/ HLA*0201, 200 nM) was mixed with blocked phage and beads. After incubation at room temperature for 1 hour, Open access beads were pulled down through a magnetic rack, and the deselected phages were transferred to a new tube and incubated with the biotinylated NY-ESO-1/HLA*0201 complex for 1 hour. Mixed with 100 µL streptavidinconjugated dynabeads and incubated for one more hour. Beads were pulled down and washed quickly with PBST (0.05%) twice, followed by a three-time wash with a 5 min incubation, and then quickly washed with PBS twice. The remaining phages on beads were eluted by adding 200 µL fresh 100 mM TEA with 20 min incubation. The supernatant was transferred to a 50 mL tube and neutralized with 0.5 volume 1 M Tris-HCI (pH 7.5). Next, 10 mL fresh prepared TG1 cells (OD600: 0.5-0.8) were added to the neutralized phage and incubated at 37°C for 1 hour at 250 rpm. The infected TG1 cells were inoculated on a 2xYTA plate with antibiotics and incubated at 30°C overnight. The input and output phage were tittered, and the enrichment ratio was calculated. The panning process was repeated three rounds, with decreased biotinylated NY-ESO-1/HLA*0201 complex concentration (200 nM-50 nM-10nM).

After three rounds of phage panning, the output phageinfected TG1 clones were picked by colony picker (Molecular Devices) into 96-well plates and cultured at 775 rpm at 37°C with 80% humidity overnight. Phage ELISA was exactly performed as follows: first, transfer 3 µL overnight cultured bacteria to a new 96-well plate with 120 µL 2xYTAG medium; second, incubate 1.5-2 hours at 37°C at 775 rpm with 80% humidity until OD600 reaches 0.4-0.8; then, add 20 µL/well diluted M13K07 helper phage to obtain a ratio of 10 helper phages: 1 bacterial cell and incubate 1 hour; after that, pellet the bacteria and resuspend with 150 µL/well 2xYTAK medium containing 0.5 mM IPTG and culture at 30°C, 775 rpm with 80% humidity overnight; finally, centrifuge the plates at 2500 g for 20 min and use the supernatants for the ELISA.

ELISA was performed as follows: 96-well high binding plates were coated with 100 µL/well streptavidin (2 µg/ mL) overnight at 4°C. The next day, plates were blocked with 5% bovine serum albumin (BSA) and then incubated for 1 hour at room temperature with 100 µL/ well NY-ESO-1/HLA-A2 complex (2 µg/mL). Then 100 µL/well supernatants were added and incubated for 1 hour followed by washes three times with phosphatebuffered saline with Tween 20 (PBST) and two times with phosphate-buffered saline (PBS). HRP conjugated anti-M13 IgG (Santa Cruz Biotechnology) was used to detect the remaining phage. TMB was used as the substrate, and OD450 was measured.

Antibody production and purification Phagemids of positive clones were extracted for DNA sequencing. After sequencing analysis with IMGT, CDR diversity was summarized, and unique combinations of heavy chain and light chain nucleotide sequences were obtained. Specific infusion primers were designed to PCR the heavy chain and light chain from the phagemid DNA. The purified PCR product was infused into human IgG heavy chain and light chain expression vector. The clones were picked for Sanger sequencing to confirm the correct infusion of antibody sequence. Heavy and light chain plasmids were cotransfected in Expi-293 cells, and antibodies were purified by protein A resin.

ELISA of purified NY-ESO-1 antibodies ELISA was performed similarly as mentioned in phage ELISA, but used purified antibodies in replacement of phage and used HRP anti-human-IgG as a secondary antibody for detection.

Codon optimized 2D2 scFv was synthesized from Integrated DNA Technologies, Inc (Skokie, Illinois, USA), with a (G 4 S) 3 linker between the heavy chain and light chain. A GM-CSFR leader sequence was added in front of the viable heavy chain. The fragment was cloned into a pMSGV1 retroviral vector containing CD8α hinge and transmembrane domain, followed by 4-1BB costimulatory domain and CD3ζ signaling domain. All constructs were sequence confirmed.

Blood from healthy donors was obtained from Gulf Coast Regional Blood Center. Fresh PBMCs were isolated with Ficoll reagent following the manufacturer's instructions. Buffy layer was collected and washed twice with PBS. T cell medium suspended PBMCs then were seeded into an antihuman CD3 antibody (OKT3) coated plate for activation. Furthermore, retrovirus was packaged in HEK 293 T cells with envelope plasmid RD114 and packaging plasmid Gag-pol. Virus was harvested at 48-hour and 72-hour post-transfection and filtered with a 0.45 µm filter. Activated PBMCs were transduced twice with retrovirus in the presence of RetroNectin as per the manufacturer's guide. T cells were cultured for 3-7 days before use. On the following day, 50 µL supernatants were added into human IFN-γ (1:1000, Thermo Fisher, M700A) precoated and 1% BSA blocked plate for a 1-hour incubation at room temperature with gentle shake. The plate was then washed twice and incubated with a biotin-conjugated IFN-γ antibody (1:1000, Thermo Fisher, M700B) for 1 hour, followed by another two washes and avidin-HRP (1:5000) incubation for 30 min in the dark. After washing, 100 µL TMB was added for reaction, which was stopped Open access by 50 µL 2.5 N sulfuric acid after 15 min. The absorbance was read on a spectrophotometer (Bio-Tek) at 450 nm.

Flow cytometric analysis was performed by using BD LSR II. T cells were stained with protein L and biotinylated HLA-A2/NY-ESO-1 monomer for CD19 CAR-T and 2D2 CAR-T cells, respectively, followed by phycoerythrin (PE)-conjugated streptavidin, while using biotinylated HLA-A2/CMV monomer as control. Tumor cells were stained with mAb 2D2 for 30 min and followed by goatanti-human IgG-Alexa fluor 594. Data were analyzed with FlowJo V.10.0.7 software (TreeStar).

Cells were seeded in a four-chamber glass bottom dish. HEK 293 T cells were pulsed with 20 µg/mL NY-ESO-1 or control peptide. Cells were fixed by 4% PFA for 15 min at room temperature. After washing, cells were blocked with 5% normal goat serum for 1 hour and incubated with 10 µg/mL mAb 2D2 and APC conjugated antihuman HLA-A2 antibody at 4°C overnight. The next day, cells were washed three times with PBS and incubated with 5 µg/mL Alexa Fluor 488 or 594 conjugated goat-antihuman (H+L) secondary antibody for 1 hour in the dark. Nucleus was stained with DAPI for 5 min. Imaging was acquired with a Nikon A1 confocal microscopy.

Formalin-fixed samples were embedded by paraffin. Blocks were cut at 5 µm and mounted onto positively charged glass slides. After deparaffinization and hydration using graded concentrations of ethanol to deionized water, sections were stained with mouse-anti-human CD3 (Invitrogen, MA5-12577, 1:20) and mouse-antihuman CD8α (Abcam, ab187279, 1:200), respectively. DAB (3,3′diaminobenzidine) was used as a chromogen, followed by nuclear counterstaining. Then slides were washed and dehydrated and covered by coverslips with a permount mounting medium (Fisher, SP15-100). Imaging was captured by Olympus microscopy equipped with a DP74 camera.

LDH cytotoxicity release assay Transduced T cells were cocultured with tumor cells with a series of effector to target (E:T) ratios in 96-well plates for 6 hours and 24 hours, respectively. Supernatant containing LDH was transferred into a new enzymatic plate for cytotoxicity assay as per the manufacturer's instruction (Promega, G1780). The absorbance was read on a spectrophotometer (Bio-Tek) at 490 nm. Percentage of specific lysis was calculated as the formula: % cytotoxicity = (experimental -effector spontaneous -target spontaneous) / (target maximum -target spontaneous) * 100. All values were substrated by medium control first.

Toxicity studies HLA-A2 transgenic mice were treated with the same dose used to achieve effective therapy (10 million T cells, intravenously), and the in vivo toxicity was determined by potential clinical signs such as body weight loss and histopathological microscopic evaluation of the major tissues at days 7 and 14.

Statistical analysis Data were presented as mean±SD or SEM if indicated. Unpaired Student's t-test was used to study the differences between groups; one-way analysis of variance (ANOVA) was used for comparing more than two groups while two-way ANOVA was used for comparing more than two groups at different time points. Kaplan-Meier analysis and log-rank tests were used in the survival comparisons; statistical analysis was performed with GraphPad Prism V.8 (GraphPad Software, Inc). P values of <0.05 were considered to be significant.

Screening and selection of specific scFv for HLA-A2/NY-ESO-1 complex and engineering of full-length human mAb Human scFv was selected by a negative biopanning of a phage-displayed scFv antibody library on biotinylated HLA-A2/CMV pp65 495-503 (NLVPMVATV) control peptide monomer followed by a positive biopanning on HLA-A2/ NY-ESO-1 157-165 (SLLMWITQC) monomer. Hence, phages bound to HLA-A2 and an irrelevant peptide or biotin would have been eliminated in this step (figure 1A). After 3~4 rounds of panning, 40 clones were found to have positive binding to HLA-A2/NY-ESO-1 157-165 via phage ELISA (figure 1B, online supplemental figure S1A). Those clones were sequenced, and diversity of genotype was analyzed according to the IMGT repertoire (www. imgt.org) (online supplemental figure S1B). Ten clones with unique antibody sequences were further cloned into expression vectors. Among them, eight fully human IgG1 antibodies were expressed and purified. Two purified mAb E5 and mAb 2D2 had high specificity in recognition of 10 µg/mL HLA-A2/NY-ESO-1 157-165 complex but not HLA-A2/CMV pp65 495-503 complex (online supplemental figure S1C). However, mAb E5 showed no recognition of HLA-A2/NY-ESO-1 157-165 complex at a low concentration of 1 µg/mL, while mAb 2D2 maintained the specificity in recognizing target antigen but not control antigen ( figure 1C ). An antibody titration experiment against HLA-A2/NY-ESO-1 157-165 revealed that EC 50 of E5 was 66 nM, while mAb 2D2 had an EC 50 of 0.608 nM ( figure 1D, online supplemental figure S1D ). This may explain the difference in recognition of HLA-A2/ NY-ESO-1 157-165 between mAb E5 and mAb 2D2 mAb at the low concentration. Thus, mAb 2D2 was selected for further studies.

To further characterize the specificity of mAb 2D2, we first measured the affinity of mAb 2D2 to HLA-A2/ NY-ESO-1 157-165 complex by the biolayer interferometry on an Octet instrument. A moderate high affinity figure 2B) . Similarly, analysis of mean fluorescence intensity also indicated the specificity of mAb 2D2 binding to HLA-A2/NY-ESO-1 157-165 complex, with a saturated concentration above 16 µg/mL (figure 2C). We also used immunofluorescence to confirm our result by confocal microscopy imaging. HEK 293 T cells were pulsed with NY-ESO-1 157-165 and CMV pp65 495-503 peptide, respectively. Cells were costained with anti-HLA-A2 and mAb 2D2, as well as DAPI for the nucleus. mAb 2D2 colocalized with HLA-A2 in NY-ESO-1 157-165 peptidepulsed HEK 293 T cells but not in CMVpp65 495-503 control peptide-pulsed cells (figure 2D), indicating the antigenspecific binding of mAb 2D2 on the cell surface, which is a critical prerequisite for targeted therapy such as CAR-T.

The 2D2-CAR was engineered using variable heavy and light chain sequences from the 2D2 scFv connected by a (Gly4Ser)3 flexible linker. The 2D2 scFv with a GM-CSFR leader sequence was ligated upstream to the CD8α hinge and transmembrane and 4-1BB costimulatory domain and CD3ζ signaling domain. The resulting 2D2-BBZ gene was cloned into a pMSGV1 gamma-retroviral vector and confirmed by sequencing ( figure 3A) . Following retroviral transduction, T cells expressed high levels of 2D2-CAR, as assessed by flow cytometric analysis using a biotinylated HLA-A2/NY-ESO-1 157-165 monomer followed by streptavidin-PE (figure 3B). The transduction efficiency of the CAR T cells was between 30% and 90% for all experiments. To confirm that the 2D2-CAR was not artificially binding to 

HLA-A2/NY-ESO-1 157-165 monomer due to high avidity for the antigen, cytomegalovirus (CMV) pp65 495-503 -HLA-A2 monomer was used to stain CAR T cells. No binding of the HLA-A2/CMVpp65 495-503 monomer was observed on 2D2-CAR T cells (online supplemental figure S2A ).

The specific recognition of 2D2-CAR T cells was assessed by IFN-γ ELISA against a range of HLA-A2 + and/or NY-ESO-1 + cancer cell lines and HLA-A2 + cells pulsed with NY-ESO-1 157-165 peptide. 2D2-CAR T cells specifically recognized TAP-deficient T2 cells (HLA-A2+) pulsed with NY-ESO-1 157-165 peptide, but not T2 cells loaded with CMV pp65 495-503 peptide (figure 3C). To further determine the specificity of 2D2-CAR-T cells against NY-ESO-1 petide, alanine scaning within NY-ESO-1 peptide was conducted. Nine peptides containing an alanine substitution at each position, together with wildtype NY-ESO-1 peptide and control peptide derived from a different cancer testis antigen, were pulsed onto T2 cells at a concentration of 2 µg/mL. 2D2-CAR-T cells were cocultured with T2 cells pulsed with different peptides to determine specific 

recognition. We found that alanine substitution of the residue positions 4 (Met) and 5 (Trp) completely abolished T cell activity. Furthermore, substitution of residue positions 2 (Leu) and 3 (Leu) also markedly reduced T cell activity, while substitution of residue position 1 (Ser) reduced T cell activity but statistically not significant 4D) , and the weight of tumor was significantly reduced about 50% compared with control mice (p=0.0018) ( figure 4E ). In addition, mice treated with 2D2-CAR T cells greatly enhanced overall survival with a median survival of 41 days (Ctrl vs 2D2-CAR T: p=0.0027) (figure 4F). Through further investigation in the mice spleen, we detected more CD3 +T cells in 2D2-CAR-T cell treated group than the control group (online supplemental figure S4A ). The CD3 positive T cells were counted, and the percentage in the spleen indicated 10-fold higher in the 2D2-CAR-T group compared with the control group (online supplemental figure S4B ). Furthermore, significantly T cell infiltration (CD3 and CD8) into tumor cells was detected in the tumor section from 2D2-CAR T cell treated mice (online supplemental figure S4C,D) . These data suggest that 2D2-CAR T cells can effectively infiltrate into the tumor, kill tumor cells, impair tumor growth and prolong mice overall survival.

To further demonstrate the antitumor response in a more clinically relevant application, we chose a tumor cell line Mel1558 that endogenously expresses HLA-A2/NY-ESO-1. Two million tumor cells were subcutaneously implanted at right flank of NSG mice. total of 2.5 million 2D2-CAR T cells were injected on day 4 followed by 50,000 IU rhIL-2 intraperitoneally. Cytokine was detected on day 6 after T cell injection from mice serum (figure 5A). We noticed a significant difference in IFN-γ release between control and treated groups, even though the amount is very low (figure 5A). 2D2-CAR T-cell treated mice had a significant reduction in tumor growth starting from day 35 (figure 5B). The tumors harvested from 2D2-CAR T cell treated mice were about 50% smaller compared with control groups (figure 5C). Further analysis in the tumor section by IHC revealed that a significant increment of T cells infiltrated in the tumor in the 2D2-CAR T treated group than the control group (figure 5D). Consistently, flow cytometry also proved the existence of T cells in the tumor section with more than fivefold higher in the 2D2-CAR-T group (figure 5E). All these data suggest that 2D2-CAR T cells can inhibit the growth of tumor endogenously expressing A2/NY-ESO-1 in vivo.

To evaluate the potential side effects of 2D2-CAR T cells, human HLA-A2 transgenic NSG (NSG-A2) mice (n=6) were treated with 1×10 7 of 2D2-CAR T cells in 200 µL PBS or an equivalent volume of PBS as control. Body weight and health condition were monitored daily. Mice were euthanized to harvest key tissue organs on day 7 and day 14 for pathological analysis (figure 6A). We found that 2D2-CAR-T cell treatment did not cause any marked decrease in mouse body weight and displayed no significant difference in growth compared with control group Open access (figure 6B). Major organs including brain, heart, liver, spleen, lung, and kidney were freshly collected from euthanized mice on days 7 and 14 after 2D2-CAR-T cell treatment. There was no siginificant difference in organ weight between 2D2-CAR-T cell-treated group and control group (online supplemental table S1). Consistently, we did not observe any potential tissue damage, such as organ discoloration, apparent side/shape change, or liquid in the abdominal/thoracic cavity. H&E staining and microscopic examination did not show obvious abnormality Mice were sacrificed when the tumor reaches 2 cm in diameter, which was considered as an endpoint. Data in figure part B were present as mean (SD). Difference between the control group and treatment group was analyzed by unpaired Students' t-test. Data in figure parts C and E were presented as mean (SE). Difference was analyzed by two-way ANOVA followed by multiple comparisons with Bonferroni adjustment. Data in figure part F was analyzed by Kaplan-Meier survival analysis with a log-rank test. **P<0.01; ***p<0.001. ANOVA, analysis of variance; CAR, chimeric antigen receptor.

Open access or pathology in the brain, heart, liver, spleen, lung, or kidney from 2D2-CAR-T cell injected mice compared with the control group ( figure 6C ). Furthermore, we performed CBC and biochemistry analysis (Superchem) of mice treated with 2D2-CAR-T cells and found a slight but not significantly increase in absolute blood cell numbers such as white cell count, red blood cells, neutrophils, monocytes, eosinophils, and basophils, in 2D2-CAR-T cell treated mice compared with PBS treated mice. The relative number was slightly reduced in neutrophils and significantly increased in monocytes. No appreciable changes in platelet, hemoglobin, hematocrit, mean corpuscular volume, mean corpuscular hemoglobin, and mean corpuscular hemoglobin concentration were observed (online supplemental table S2). In the serum chemistry study, we did not observe any significant difference between 2D2-CAR-T cells treated mice and PBS treated control mice. Liver function was evaluated by the serum level of total protein, albumin, globulin, albumin/globulin ratio, aspartate transaminase, alanine transaminase, Alkaline phosphatase, gamma-glutamyl transferase, and total bilirubin. Key electrolytes such as phosphorus, chloride, calcium, magnesium, sodium, potassium, and sodium/potassium ratio were balanced in both groups. Concentrations of glucose, cholesterol, and triglyceride were in the normal range. Kidney function as measured by blood urea nitrogen (BUN), creatinine and BUN/creatinine ratio, pancreatic function as measured by amylase and precision PSL, and muscle function measured by creatine phosphokinase, were all normal in 2D2-CAR-T treated mice compared with PBS treated mice (online supplemental figure S5 ). These results 

suggest that 2D2-CAR-T cell treatment does not induce any potential toxicity or tissue damage in NSG-A2 mice.

In this study, we identified mAb 2D2 as a novel TCR-like human full-length IgG1 monoclonal antibody against an intracellular cancer-testis antigen NY-ESO-1 presented by HLA-A2 molecules. The mAb 2D2 could specifically recognize HLA-A2/NY-ESO-1 157-165 monomer, NY-ESO-1 157-165 pulsed T2 or HEK 293 T cells, and tumor cells expressing HLA-A2 and NY-ESO-1, but not HLA-A2/CMV pp65 495-503 . We further show that genetically engineered 2D2-CAR T cells could specifically recognize HLA-A2/ NY-ESO-1 complex and lyse tumor cells both in vitro and in vivo. Mice treated with 2D2-CAR T cells significantly impaired the growth of A2 and NY-ESO-1 double positive breast cancer cells and enhanced overall survival. Notably, no adverse effect was observed in the key organs of the normal mice treated with 2D2-CAR T cells along, demonstrating the safety of the 2D2-CAR T cells.

A previous study used a semisynthetic Fab repertoire to identify antibodies that could bind to HLA-A2/NY-E-SO 157-165 . 35 Although they removed streptavidin binders, there was no negative selection round against the HLA-A2 complex with non-relevant peptides. The affinity of antibody (3M4E5) isolated from the screening was 60 nM, which was further modified to increase the affinity via the comparison between the high-resolution structure of 1G4 TCR and 3M4E5 bound to HLA-A2/NY-ESO-1 157-165 . Using computer model analysis, a new library with point Figure 6 Safety assessment of 2D2-CAR T cells in vivo. (A) A schematic diagram of the animal experiment. NSG-A2 mice were infused with 1×10 7 2D2 CAR T cells or PBS on day 0. Body weight was measured as indicated. On day 7 and day 14, mice were sacrificed for histopathology and hematology assessment. (B) Body weight change after T cells injection. (C) H&E staining for the key organs, heart, spleen, lung, kidney, brain, and liver with 10× magnification. CAR, chimeric antigen receptor; ns, not significant.

Open access mutations in the residues not contacting with the peptide was generated. 36 This super-high affinity antibody (T1) exceeded the affinity of TCR by 1000-fold. Specific lysis of HLA-A2 positive T2 cells pulsed with NY-ESO-1 peptide was observed. However, no killing activity of primary tumor cells was demonstrated. In a subsequent study using a CAR construct containing an scFv from T1 antibody, it was found that T1 scFv-CAR-T cells failed to show the antigen specificity probably due to the high affinity of T1 antibody to HLA-A2/NY-ESO-1 157-165 complex as well as HLA-A2 molecules. 37 Similar to TCR, [38] [39] [40] [41] these studies suggest that increasing TCR-like antibody affinity may increase the likelihood of cross-reactivity against HLA molecules. Therefore, mAb 2D2 has an affinity of 5.74 nM, a moderate high affinity for a monoclonal antibody, which somehow maintains the specificity, promoting the success of CAR-T therapy. Interestingly, the immunoglobin type of the mAb 2D2 VL chain was the same as T1, while it differed in the heavy chain variable region. However, 2D2-CAR T cells did not display 'off-target' issues, as demonstrated by our in vivo experiments.

Even though antibodies usually have higher affinity than TCRs, CAR-T cells derived from TCR-like antibodies may need a higher density of pMHC complex presented at the cell surface to achieve a similar level of activation compared with TCR-T cells. 42 αβ-TCR-T cells with relatively low-affinity maintain the specificity and cytotoxic activity, while the high-affinity CAR-T cells may reduce specificity and killing ability. 43 One possible reason is the difference of conformation between in vitro refolded complex and bona fide pMHC complex on the cell surface for TCR-like antibody recognition. To compare HLA-A2 NY-ESO-1 specific (A2-ESO TCR)-T cells with 2D2-CAR-T cells, we cloned 1G4 with two amino acid substitutions 44 into retroviral expression vector. The same numbers of A2-ESO TCR T-and 2D2 -CAR-T cells were cocultured with CMV pp65 495-503 peptide or NY-ESO-1 157-165 peptide, respectively. We found that A2-ESO TCR-T cells released higher (10-fold) IFN-γ than 2D2-CAR-T cells in response to A2/NY-ESO-1 complex (online supplemental figure  S6A ). Importantly, the killing activity of 2D2-CAR-T cells was similar to that of A2-ESO TCR-T cells (online supplemental figure S6B ). A previous study shows that the target antigen density required for inducing T cell cytokine production is much higher than that required for stimulating CAR-mediated lysis. 45 The lowest density for CD20-specific CAR-T cells to lyse target cells is ~200 molecules per cell, but cytokine production required a higher density of CD20 (~5000 molecules per cell). 45 Generally, TCR-T cells require 10 pMHC complexes per cell on the target surface to be activated, while CAR-T cells require more than 100 copies per cell. 46 Importantly, the lower cytokine production by 2D2 CAR-T cells may reduce the risk of potential toxicity associated with the cytokine storm, while maintaining similar killing ability of tumor cells. To further test this possibility, we performed in vivo experiment to demonstrate the therapeutic efficacy of A2-ESO TCR-T cells and 2D2-CAR-T cells. We found that both A2-ESO TCR-T cells and 2D2-CAR-T cells showed strong antitumor response (online supplemental figure  S6C ). Finally, A2-ESO TCR α and β chains may mispair with endogenous TCR α and β chains, causing potential off-target cross reactivity, while 2D2-CAR-T cells could not mispair with endogenous TCR α and β chains. Thus, the lower amounts of cytokine release and no potential mispairing of 2D2 CAR-T cells may have advantages over A2-ESO TCR-T cells. However, further studies are needed to investigate potential clinical application.

Currently, CAR-T cell technology does not work well in the solid cancers. 2D2-CAR T cells could not completely eradicate tumor cells in mouse models. We showed that the in vitro cytolytic activity of 2D2-CAR-T cells was correlated with the level of HLA-A2/NY-ESO-1 complex on the cell surface, suggesting that the low cytolytic activity of 2D2-CAR-T cells toward MDA-MB-231-ESO and Mel1558 in vivo is due to a low density of HLA-A2/ NY-ESO-1 complex on the cell surface. Furthermore, T cell trafficking and persistence could also influence the anti-tumor activity of CAR-T or TCR-T cells in vivo. We noticed that T cells would die within a week after injection into the NSG mice in the absence of antigen stimulation or IL-2 (data not shown). Murine CAR-T cells may exert superior antitumor activity in immunocompetent mice. 47 Combination therapy together with checkpoint blockade or negative regulator inhibition may also augment antitumor activity in vivo. 48 T cell trafficking to tumor sites is critically important for CAR-T cell therapy in the solid cancers. For example, CXCR3 and CCR5 are often highly expressed in tumor-infiltrating lymphocytes from melanoma, colorectal and breast cancers, implying the importance in regulating T cell trafficking. 49 Thus, further studies are needed to promote 2D2-CAR-T cell trafficking and persistence in the solid cancer and overcome immune suppression in the tumor microenvironment, thus developing more effective CAR-T or TCR-like CAR-T cell therapy.

In summary, our results presented here provide a proof of concept that TCR-like antibody derived CAR-T cells could successfully inhibit tumor cell growth and enhance mouse overall survival. Unlike current CAR technology that only recognizes surface antigens, TCR-like CAR may broaden the scope of target antigens, including intracellular antigens. The specificity and cytotoxicity of our TCRlike CAR would significantly impact our ability to develop new immunotherapy against various types of cancer. 6 Neurosurgical Oncology in the Department of Neurosurgery, Cedars-Sinai Medical Center, Los Angeles, California, USA

Monoclonal antibodies for cancer immunotherapy

Chimeric antigen receptormodified T cells for acute lymphoid leukemia

Chimeric antigen receptormodified T cells in chronic lymphoid leukemia

Current progress in CAR-T cell therapy for solid tumors

CAR-T cell therapy: current limitations and potential strategies

CAR T cells for solid tumors: new strategies for finding, infiltrating, and surviving in the tumor microenvironment

T-cell receptor mimic (TCRm) antibody therapeutics against intracellular proteins

Immune targets and neoantigens for cancer immunotherapy and precision medicine

CAR-Engineered NK cells targeting wild-type EGFR and EGFRvIII enhance killing of glioblastoma and patient-derived glioblastoma stem cells

NY-ESO-1 and LAGE-1 cancer-testis antigens are potential targets for immunotherapy in epithelial ovarian cancer

Immunohistochemical analysis of NY-ESO-1 antigen expression in normal and malignant human tissues

Cloning genes encoding MHC class II-restricted antigens: mutated CDC27 as a tumor antigen

A testicular antigen aberrantly expressed in human cancers detected by autologous antibody screening

A breast and melanomashared tumor antigen: T cell responses to antigenic peptides translated from different open reading frames

Generation of NY-ESO-1-specific CD4+ and CD8+ T cells by a single peptide with dual MHC class I and class II specificities: a new strategy for vaccine design

Identification of CD4+ T cell epitopes from NY-ESO-1 presented by HLA-DR molecules

CD4(+) T cell recognition of MHC class II-restricted epitopes from NY-ESO-1 presented by a prevalent HLA DP4 allele: association with NY-ESO-1 antibody production

Human tumor antigens for cancer vaccine development

Simultaneous humoral and cellular immune response against cancer-testis antigen NY-ESO-1: definition of human histocompatibility leukocyte antigen (HLA)-A2-binding peptide epitopes

A survey of the humoral immune response of cancer patients to a panel of human tumor antigens

NY-ESO-1 may be a potential target for lung cancer immunotherapy

Tumor regression in patients with metastatic synovial cell sarcoma and melanoma using genetically engineered lymphocytes reactive with NY-ESO-1

A pilot trial using lymphocytes genetically engineered with an NY-ESO-1-reactive T-cell receptor: long-term follow-up and correlates with response

NY-ESO-1-specific TCR-engineered T cells mediate sustained antigen-specific antitumor effects in myeloma

Targeting the intracellular WT1 oncogene product with a therapeutic human antibody

Activity of 8F4, a T-cell receptorlike anti-PR1/HLA-A2 antibody, against primary human AML in vivo

T-cell-receptor-like antibodies -generation, function and applications

Therapeutic bispecific T-cell engager antibody targeting the intracellular oncoprotein WT1

A novel TCR-like CAR with specificity for PR1/HLA-A2 effectively targets myeloid leukemia in vitro when expressed in human adult peripheral blood and cord blood T cells

Optimized T-cell receptormimic chimeric antigen receptor T cells directed toward the intracellular Wilms tumor 1 antigen

TCR-like antibody drug conjugates mediate killing of tumor cells with low peptide/HLA targets

Elimination of melanoma by sortase A-generated TCR-like antibody-drug conjugates (TL-ADCs) targeting intracellular melanoma antigen MART-1

CAR-T cells targeting a nucleophosmin neoepitope exhibit potent specific activity in mouse models of acute myeloid leukaemia

Molecular determinants and mechanism for antibody cocktail preventing SARS-CoV-2 escape

Dissecting cytotoxic T cell responses towards the NY-ESO-1 protein by peptide/MHC-specific antibody fragments

Rational development of high-affinity T-cell receptor-like antibodies

An MHC-restricted antibodybased chimeric antigen receptor requires TCR-like affinity to maintain antigen specificity

TCRs with high affinity for foreign pMHC show self-reactivity

Class II-restricted T cell receptor engineered in vitro for higher affinity retains peptide specificity and function

High-affinity TCRs generated by phage display provide CD4+ T cells with the ability to recognize and kill tumor cell lines

Directed evolution of human T-cell receptors with picomolar affinities by phage display

Construction and molecular characterization of a T-cell receptor-like antibody and CAR-T cells specific for minor histocompatibility antigen HA-1H

Functional comparison of engineered T cells carrying a native TCR versus TCR-like antibody-based chimeric antigen receptors indicates affinity/avidity thresholds

Single and dual amino acid substitutions in TCR CDRs can enhance antigen-specific T cell functions

Target antigen density governs the efficacy of anti-CD20-CD28-CD3 ζ chimeric antigen receptor-modified effector CD8+ T cells

TCR-Like CAR-T cells targeting MHC-Bound minor histocompatibility antigens

Il-7 and CCL19 expression in CAR-T cells improves immune cell infiltration and CAR-T cell survival in the tumor

A guide to cancer immunotherapy: from T cell basic science to clinical practice

Non-redundant requirement for CXCR3 signalling during tumoricidal T-cell trafficking across tumour vascular checkpoints

Acknowledgements We would like to thank Ms Hui Deng for her technical support. We would like to thank NIH tetramer facility for providing monomers.Contributors RW served as guarantor and supervised the entire project. XL, ZA and RW designed the experiments. YX conducted phage library screening. YX and WX performed protein purification and affinity measurement. XL contributed to most experiments and data analysis. XL, ZA and RW wrote and revised the manuscript, with input from other authors.Funding This work was supported in part by a Welch Foundation grant AU-0042-20030616 and Cancer Prevention and Research Institute of Texas (CPRIT) Grants RP150551 and RP190561 (ZA) as well as grants from the NCI, NIH (R01CA101795, R01CA246547, and U54CA210181), and the Department of Defense (DoD) CDMRP BCRP (BC151081) and LCRP (LC200368) to RW (Wang).

Competing interests None declared.Patient consent for publication Not applicable.Ethics approval Not applicable.Provenance and peer review Not commissioned; externally peer reviewed.Data availability statement Data are available on reasonable request.Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.Open access This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See http://creativecommons.org/licenses/by-nc/4.0/.

Wei Xiong http://orcid.org/0000-0002-8087-5047 Rongfu Wang http://orcid.org/0000-0002-8834-2763