Human induced-T-to-natural killer cells have potent anti-tumour activities

Adoptive cell therapy (ACT) is a particularly promising area of cancer immunotherapy, engineered T and NK cells that express chimeric antigen receptors (CAR) are being explored for treating hematopoietic malignancies but exhibit limited clinical benefits for solid tumour patients, successful cellular immunotherapy of solid tumors demands new strategies. Inactivation of BCL11B were performed by CRISPR/Cas9 in human T cells. Immunophenotypic and transcriptional profiles of sgBCL11B T cells were characterized by cytometer and transcriptomics, respectively. sgBCL11B T cells are further engineered with chimeric antigen receptor. Anti-tumor activity of ITNK or CAR-ITNK cells were evaluated in preclinical and clinical studies. We report that inactivation of BCL11B in human CD8+ and CD4+ T cells induced their reprogramming into induced T-to-natural killer cells (ITNKs). ITNKs contained a diverse TCR repertoire; downregulated T cell-associated genes such as TCF7 and LEF1; and expressed high levels of NK cell lineage-associated genes. ITNKs and chimeric antigen receptor (CAR)-transduced ITNKs selectively lysed a variety of cancer cells in culture and suppressed the growth of solid tumors in xenograft models. In a preliminary clinical study, autologous administration of ITNKs in patients with advanced solid tumors was well tolerated, and tumor stabilization was seen in six out nine patients, with one partial remission. The novel ITNKs thus may be a promising novel cell source for cancer immunotherapy. ClinicalTrials.gov, NCT03882840. Registered 20 March 2019-Retrospectively registered.


Introduction
Adoptive cell therapy (ACT) is a particularly promising area of cancer immunotherapy [1][2][3]. Unlike T cells, NK cells play an important role in immune surveillance by targeting tumor cells that downregulate HLA class I molecules or express stress markers [4]. Engineered T and NK cells that express chimeric antigen receptors (CARs) or cancer-specific T cell receptor (TCR) transgenes with PD-1 ablation have been successfully used to treat hematopoietic malignancies but exhibit limited clinical benefits for solid tumor patients [3,[5][6][7][8][9][10][11][12][13]. New immune cell sources that recognize and eliminate solid tumor cells would therefore be desirable.
Human T cell development also requires BCL11B [35][36][37]. Patients carrying BCL11B mutations exhibit primary immunodeficiency caused by T cell deficiency [38][39][40]. Dysregulation of BCL11B has been implicated in T cell leukemias [41,42]. The inhibition of BCL11B induces apoptosis in T-ALL [43][44][45]. In contrast, knockdown of BCL11B in normal mature human T cells does not affect viability but rather upregulates the expression of ID2 [43]. In addition, the suppression of BCL11B by chimeric antigen receptor (CAR) expression in human lymphoid progenitors represses the expression of T cellassociated genes, including IL7R, GATA3, and NOTCH3, and increases the expression of ID2 and GZMB [46].
The roles of BCL11B in mature human T cells during homeostasis have not yet been fully elucidated. Here, we report that inactivating BCL11B in multiple human T cell subsets reprogrammed them into induced T-to-NK cells (ITNKs). ITNKs retained a functional TCR, upregulated NK cell-associated markers and transcription factors, and contained elongated tubular mitochondria. Mechanistically, BCL11B directly repressed NK-cell associated transcription factors for the maintenance of T cell identity. ITNKs recognized and efficiently lysed cancer cells in culture, in organoids and in murine cancer models. In a clinical investigation, transplanted autologous ITNKs exhibited tumoricidal activities in patients with refractory and advanced solid tumors with no severe adverse effects. The molecular, cellular, preclinical and clinical studies presented here demonstrate that human ITNKs could be further explored as a new cell source for cellbased cancer immunotherapy.

Reagents and antibodies
SR11302 (A8185, APExBIO) was used for the rescue experiment. All antibodies used in the study for fluorescence-activated cell sorting, flow cytometry and cellular stimulation are listed in Table S10. All sgRNAs used in the study are listed in Table S10.

Isolation, transduction, and expansion of primary human T lymphocytes, γδ T cells and NK cells
For all preclinical experiments in this study, PBMCs were isolated from cord blood or from healthy adult donors (median age 33 years, range 29-50 years; two females and three males) using Lymphoprep (Stem Cell Technologies, Vancouver, Canada). T cells were negatively selected from PBMCs using a MACS Pan T Cell Isolation Kit (Miltenyi Biotec, Bergisch Gladbach, Germany) and activated MACS GMP T Cell TransAct (Miltenyi Biotec, Bergisch Gladbach, Germany), in 5 μl at a bead: cell ratio of 1:2 and a density of 2.5 × 10 6 cells/ml for 1 day in T551-H3 (Takara, Japan) medium supplemented with 5% heat-inactivated fetal bovine serum (FBS), 500 U/ml recombinant human IL-2, 10 mM HEPES, 2 mM glutamine and 1% penicillin/streptomycin. γδ T cells were cultured in RPMI 1640 medium supplemented with 10% FBS, antibiotics, IL-2 (100 IU/ml), vitamin C (70 μΜ) and ZOL (50 μM, Τ6739, TargetMol) as reported by Yin et al. [47]. NK cells were cultured in T551-H3 medium with 5% FBS and 500 U/mL IL-2 at an initial density of 1 × 10 6 cells/ ml. Briefly, CB or PBMCs were directly activated with NK Cell Activation/ Expansion Kit (130-094-483, Miltenyi Biotec, Germany), which contains microbeads loaded with NKp46 and CD2 antibodies and were cultured supplemented with IL-2 (500 U/mL). Healthy PBMC donors provided informed consent for the use of their samples for research purposes, and all procedures were approved by the Research Ethics Board of the Guangzhou Institutes of Biomedicine and Health, Chinese Academy of Sciences (GIBH).

Induction and expansion of ITNKs and CAR-ITNKs
For electroporation, on postactivation day 1, T cells from CB were electroporated with sgBCL11B (5 μg plasmid per 1 × 10 7 T cells) or sgBCL11B/PiggyBac (PB)-CAR vectors (3 μg of sgBCL11B plasmid, 2 μg of PB-CAR plasmid, and 1 μg of PBase plasmid per 1 × 10 7 T cells) using an Amaxa Nucleofector 2b (Amaxa ® Human T Cell Nucleofector ® Kit, Lonza, Germany) with electroporation program T-023. Twelve hours after electroporation, T cells were cultured in T cell culture medium containing rh-IL2 (500 U/mL). Cas9 RNPs were prepared immediately before experiments by incubating 20 μM Cas9 with 20 μM sgRNA at a 1:1 ratio in Human T Cell Nucleofector buffer at 37 °C for 10 min to a final concentration of 10 μM [9]. For lentivirus transduction, on postactivation day 1, T cells were transfected with lentivirus at an MOI of 5. Twelve hours after transduction, T cells were electroporated with sgBCL11B (5 μg plasmids per 1 × 10 7 T cells). Subsequently, fresh medium was added every 2 days to maintain cell density within the range of 0.5-1 × 10 6 cells/ml. CD3 + NKp46 + T cells were defined as ITNKs and used in in vitro and in vivo experiments.

Functional assays
For plate-bound antibody stimulation assays, 96-well flat-bottom plates (MaxiSorp, Nunc, Thermo Fisher, USA) were precoated with different antibodies. A total of 1 × 10 5 cells were added to the wells in complete medium. After 1 hour of incubation, GolgiStop (554,724, Biosciences, USA) was added, and the plates were incubated for four additional hours at 37 °C. Afterwards, the cells were washed, fixed, permeabilized and stained with anti-IFN-γ and IL-2 antibodies.

In vitro killing assays
The K562-GL, HeLa-GL, NALM6-GL, SK-OV-3-GL and HepG2-GL target cells were incubated with the indicated killing cells at the indicated ratio in triplicate wells of U-bottomed 96-well plates. Target cell viability was monitored 18 h later by adding 100 μl of the substrate D-luciferin (potassium salt) (Cayman Chemical, Michigan, USA) at 150 μg/ml to each well. Background luminescence was negligible (< 1% of the signal from wells containing only target cells). Spheroids of PDOs were obtained as described in previous studies [48]. Briefly, 5 × 10 5 cells were seeded into wells of a 24-well Kuraray ultralow attachment plate (round-bottom type; Elplasia, Japan) on day 0. On day 2, ITNKs, T cells, NK cells, or culture medium only (as blank control) was cocultured with spheroids in 96-well U-bottomed plates. Cytotoxicity was assayed at 72 h post coculture using a CellTiter-Glo ® 3D Cell Viability Assay (G9683, Promega, USA). The percent cytotoxicity (killing %) values were calculated as (blank signal -experimental signal)/blank signal× 100%.

Xenograft models and in vivo assessment
Animal experiments were performed in the Laboratory Animal Center of Guangzhou Institutes of Biomedicine and Health (GIBH), and all animal procedures were approved by the Animal Welfare Committee of GIBH. All protocols were approved by the relevant Institutional Animal Care and Use Committee (IACUC). NSI mice [49] were maintained in specific pathogen-free (SPF)-grade cages and were provided autoclaved food and water. Direct injection of the indicated tumor cells or leukemia cells in 200 μL of PBS was performed to establish subcutaneous (flank) or intravenous (tail vein) tumors, respectively. At the indicated time for each experiment, 2.5-5 × 10 6 of the indicated killing cells in 200 μL of PBS were adoptively transferred to tumor-bearing mice systemically by tail vein injection. Peripheral blood was obtained by retro-orbital bleeding. The body weight of the mice was measured every 2 or 3 days as indicated. The xenografted mice were then randomized into different groups. The sample size of each group for all mouse experiments was n ≥ 5. Tumors were measured every 3 days with a caliper. The tumor volume was calculated using the following equation: (length×width 2 )/2. In vivo whole-body imaging of luciferase-labeled cells was performed using a cooled CCD camera system (IVIS 100 Series Imaging System, Xenogen, Alameda, CA, USA) [50]. Firefly D-luciferin (potassium salt) was injected at 75 mg/kg. Mice were imaged 5 min after the injection of the substrate. Quantification of total and average emissions was performed using Living Image software. ITNKs and control T cells were cultured under the same conditions (T551-H3 (Takara, Japan) medium supplemented with 5% heat-inactivated FBS, 500 U/ml recombinant human IL-2, 10 mM HEPES, 2 mM glutamine and 1% penicillin/streptomycin), and the expression of CD45RA and CD45RO was characterized before infusion into xenografts. To investigate the function of ITNKs upon adoptive transfer, we first examined the distribution and persistence of ITNKs in vivo. Flow cytometry was used to monitor the proportions of ITNKs in the PB, spleen, bone marrow (BM), liver and lungs on days 7, 14, 21, and 28.

Flow cytometry and cell sorting
Flow cytometric analysis was performed on a FACS-Canto or FACS Fortessa instrument (BD, USA). FACS was performed on the FACSAria II platform (BD, USA). Surface staining for flow cytometry and cell sorting was performed by pelleting cells and resuspending in 50 μl of FACS buffer (2% FBS in PBS) with antibodies for 30 min at 4 °C in the dark. For intracellular staining, cells were fixed and permeabilized with the Foxp3/Transcription Factor Staining Buffer Set (Thermo Fisher, USA), washed, blocked with mouse or rabbit serum and incubated with antibodies for 30 min at 4 °C. Cells were washed once in FACS buffer before resuspension.

Imaging flow cytometry
Surface staining was performed as described above. Cells were washed with 1 × PBS buffer containing 0.5 mM EDTA and 0.2% BSA at pH 7.2 and suspended at a concentration of 1-2 × 10 6 /mL. Positive staining for each antibody-fluorophore combination was determined using FMO controls. Samples were acquired on an Amnis ImageStream X Mark II instrument equipped with 405 nM, 488 nM, 561 nM, and 640 nM lasers utilizing INSPIRE software (Amnis, Seattle, WA). Automatic compensation was performed with a single color, followed by manual adjustment and analysis using IDEAS 6.0 software (Amnis, Seattle, WA).

Cytokine release assays
ITNKs, T cells and NK cells (2 × 10 5 ) were washed twice and mixed with 2 × 10 5 K562 cells in 200 mL of complete medium. Cells were incubated for the indicated time at 37 °C in 5% CO 2 for 18 h. Then, supernatants were collected and stored at − 20 °C for further measurement. The concentrations of cytokines were quantified by a multiplex immunoassay (Milliplex MAP, Millipore, USA) using a kit that detects 21 different cytokines (Cat.# HSTCMAG28SPMX21).

Protein isolation and immunoblotting
Cells were lysed with RIPA buffer (Pierce, Rockford, Illinois, USA), and the protein concentration was quantified using the BCA Protein Assay kit (Pierce, Rockford, Illinois, USA). Samples were loaded onto an 8-12% SDS-PAGE gel, transferred to a PVDF membrane, and sequentially probed with primary antibodies. A species-matched HRP-conjugated secondary antibody was then added, and proteins were detected by autoradiography using an enhanced chemiluminescence kit (ECL Plus, General Electric Healthcare, Little Chalfont, UK).

Histological analysis
Organ or tissue samples were fixed in 10% formalin, embedded in paraffin, sectioned at 4 μm thickness, and stained with hematoxylin and eosin or the indicated antibodies. Images were obtained on a microscope (Leica DMI6000B, Leica Microsystems, Wetzlar, Germany).

CyTOF sample preparation and acquisition
Cells from culture suspensions were stained for viability with 5 mM cisplatin in PBS (Fluidigm, USA) for 5 min on ice and then washed with PBS with 0.5% BSA and 0.02% NaN 3 . Cells were suspended in Fc receptor blocking mix, incubated for 20 min on ice, and subsequently stained with metal-labeled monoclonal antibody (mAb) cocktails against cell surface molecules for 30 min on ice. Antibodies were either purchased preconjugated from Fluidigm or conjugated in-house using mass cytometry antibody conjugation kits (Fluidigm, USA) according to the manufacturer's instructions. Cells were then washed and stained with 200 μl of 1:4000 191/193 Ir DNA intercalator (Fluidigm) diluted in Fix and Perm (Fluidigm) at 4 °C overnight. After treatment with fixation/permeabilization buffer (Thermo Fisher, USA), cells were further incubated with a metallabeled mAb cocktail against intracellular proteins. At the time of acquisition, cells were washed once with PBS with 0.5% BSA and 0.02% NaN 3 , once with ddH 2 O, and then suspended in ddH 2 O containing bead standards (Fluidigm, USA) to approximately 1 × 10 6 cells per ml. Samples were subsequently acquired on a CyTOF instrument (Fluidigm, USA) at an event rate of < 300 events/second [51]. Antibody information is given in Table S10.
Bulk RNA sequencing mRNA extracted from purified T, ITNK and NK cells was prepared according to the TruSeqTM RNA Sample Preparation Guide, and sequencing was performed. Sequenced reads were trimmed to eliminate adaptor sequences and masked to remove low complexity or lowquality sequences. The number of raw reads mapped to genes was calculated by RSEM (rsem-1.2.4), and the sample results were combined and normalized in EDAseq (1.99.1). Gene expression fold changes were calculated using normalized raw reads. The downstream analysis used glbase scripts.

Single-cell RNA sequencing and analysis
To capture single-cell transcriptomic information related to BCL11B inactivation in T cells, we collected sgB-CL11B-edited T cells from days 2-10 (D2, D4, D6, D8, D10) for 10x single-cell RNA-seq. We prepared libraries following the Chromium Single Cell 30 Reagent Kit User Guide. The single-cell libraries were quantified by a high sensitivity Quant-iT dsDNA Assay Kit (Thermo Fisher) on a Qubit 2.0 instrument and then sequenced on an Illumina HiSeq 2500 by Guangzhou Gene Denovo. For data analysis, cellranger-2.1.1 was used to map the 10x singlecell RNA-seq data. The read1 data of pooled cells were split into single-cell data using the barcode sequences contained in the first 16 bp. The next 10 bp were recorded as unique molecular identifiers (UMIs). Read2 75 bp sequences were aligned to the mm10 genome. We used Seurat (v2.3.0) for preprocessing of the data. We excluded cells with fewer than 2500 detected genes. Overall, 11,011 (D2), 6562 (D4), 7464 (D6), 7426 (D8) and 6148 (D10) cells met the quality control criteria and were used for further analysis [52]. Raw scRNA-sequencing data from human blood and splenic NK cells were downloaded from the database listed in Crinier et al. [53].

TCRβ sequencing
cDNA was first generated and amplified by using a human TCRβ profiling kit (635,014, Clontech, Takara). Libraries were sequenced on a HiSeq4000 platform (Illumina, USA). The clean reads were aligned and assembled using MiXCR. The TCRβ clonotypes were exported by parameter '--chains' in the export Clones command of MiXCR [54]. The exported clonotypes were visualized in the form of a chord diagram using VDJtools software (version 1.1.10).

Manufacture of clinical-grade ITNKs
For the clinical trial in this study, ITNKs were derived from autologous T cells that were engineered with CRISPR/Cas9 to knock out BCL11B. ITNKs were manufactured at a Good Manufacturing Practice (GMP) laboratory at Guangdong Zhaotai InVivo Co. Ltd. (GZI). An overview of the manufacturing process is shown in Fig.  S10D. On day 0, autologous T cells were obtained by apheresis. Mononuclear leukocytes were isolated from the apheresis product by Ficoll density gradient centrifugation. T cells were enriched from mononuclear leukocytes with CliniMACS CD4 reagent (200-070-132, Miltenyi Biotec, Germany) and CliniMACS CD8 reagent (200-070-115, Miltenyi Biotec, Germany) and activated by MACS GMP T cell transaction (170-076-156, Miltenyi Biotec, Germany) for 72 h according the manufacture's protocols. On day 4, T cells were electroporated (Amaxa Nucleofector 2b, Amaxa ® Human T Cell Nucleofector ® Kit, Lonza, Germany) with 3 μg sgRNA-BCL11B plasmids for 1 × 10 7 cells. On day 5, sgRNA-BCL11Btransduced T cells were cultured in T551 H3 medium containing 5% CTS immune cell SR (A2596101, Thermo Fisher, USA), 500 U/ml rh-IL2, and gentamycin sulfate (20 μg/ml). On day 7, the cells were transferred to T175 flasks with daily perfusion, and cultures were allowed to continue expansion until the harvest day. On the harvest day, a small sample of cells was collected and screened for pathogenic microorganisms and contaminants (bacteria, fungi, virus, mycoplasma, and endotoxins). Another small sample of cells was collected for evaluation of the killing capacities of ITNKs against K562 cells in vitro. The rest of the cells were harvested, washed, formulated, and cryopreserved in infusible cryomedia. Small-scale cultures were included for mock electroporation controls and potency controls. Absolute cell counts were obtained during large-scale culture using a cell counter (Countstar ® , BioMed, China).

Clinical summaries of subjects
Subject GD001, 47 years old, was first diagnosed with tumor-node-metastasis (TNM) stage II nasopharyngeal carcinoma (NPC) in 2014. Initially, he was treated with local radiotherapy but refused chemotherapy. In 2016, he underwent surgical resection of the left parotid region due to metastasis and refused to undergo chemotherapy again. His right lower lung had a mass with lymph node metastasis in the right hilum and mediastinum, accom- . She experienced slight fever and fatigue upon each cell infusion but did not suffer any CRS or neurotoxicity. Due to disease stability, a lack of adverse effects and the absence of other therapeutic options, the TMC agreed to long-term administration for this subject on a compassionate basis after dosage escalation. The patient received ITNKs once every 30 days at maintenance doses (Table 1). Once per month, the subject received staging evaluation. The subject had not received any other treatments except ITNK cell infusion.
Subject GD003, 42 years old, was first diagnosed with rhabdomyosarcoma in the left armpit in 2017. At first, he received 5 cycles of epirubicin, platinum and ifosfamide adjuvant chemotherapy. He developed liver, pancreas, lung and spine recurrence and was treated with radiofrequency ablation for the liver, high-intensity focused ultrasound (HIFU) for the pancreas, resection including the right lower lung and one vertebral resection and Tomo-Therapy radiation treatment for partial liver and lung metastases. After local therapies, he received apatinib, nivolumab and everolimus but experienced progression prior to T cell harvest. The subject underwent lymphodepleting chemotherapy and received hepatic arterial and ascending aortic injection of ITNKs at dose level 1 on 2 August 2019 (D0). After infusion, this patient experienced fatigue, slight fever, and muscle pain with elevated IL-6 in serum. Although no severe adverse effects were observed in this patient, as the first patient with an IL-6 spike in this clinical investigation, he was administered tocilizumab to prevent potential CRS. In particular, the IL-6 concentration in serum of GD003 (669.3 pg/ml), the patient with the highest IL-6 concentration in this study, was still lower than the highest IL-6 concentrations in anti-HER2 CAR-T recipients (1000 pg/ml) [55], in anti-IL13Rα2 CAR-T recipients (1062.5 pg/ml) [56], and in anti-CEA CAR-T recipients (above 1000 pg/ml) [57]. The subject recovered from CRS quickly and remained eligible for ongoing trial participation. He then received ITNKs at dose level 0 on 16 August 2019 (D14) without infusion-related complications. We repeated the infusion of ITNKs at dose level 1 on 29 August 2019 (D27). At days 30 and 60, his staging evaluations showed progressive disease according to the RECIST 1.1 criteria; however, his serum CA199 was decreased temporarily during the course of ITNK infusion. He was admitted to a local hospital for palliative care.
Subject GD004, 29 years old, was first diagnosed with TNM stage IV melanoma from an unknown site manifested as metastases in the neck and liver in 2017. He was first treated with 4 cycles of albumin-paclitaxel, bevacizumab and pembrolizumab with progression, followed by 4 lines of therapy including various combinations of temozolomide, dacarbazine, cisplatin, vindesine, semustine, ipilimumab, nivolumab and local microwave ablation or radiation, also with progression. He underwent steady-state T cell harvest prior to fludarabine and cyclophosphamide lymphodepleting chemotherapy before ITNK cell infusion. He received an intra-arterial infusion of ITNKs into the hepatic artery and ascending aorta on 16 August 2019 (D0) at dose level 1 without infusionrelated complications. He experienced slight fever but no CRS or neurotoxicity. At day 30, staging evaluation revealed mixed response and stable disease according to the RECIST 1.1 criteria. At day 90, he met the criteria for progression and received surgery on his spine. He remains alive with progression but did not show any new or ongoing trial-related adverse events 14 months post ITNK cell infusion. He received treatment at a local hospital. Subject GD005, 61 years old, was first diagnosed with TNM stage IV gastric carcinoma manifested by metastases in the omentum and ovaries in 2017. She was first treated with 6 cycles of oxaliplatin and tegafur/gimeracil/ oteracil potassium adjuvant chemotherapy. She developed local recurrence and had not received active therapy for 3 months prior to T cell harvest and lymphodepleting chemotherapy and ITNK cell infusion. As ITNKs for this subject could not be manufactured at dose level 1, the ITNKs produced were given at dose level 0 intravenously on 28 August 2019 (D0), with the agreement of the TMC. She experienced no CRS or neurotoxicity. At day 28, her staging evaluation showed stable disease according to the RECIST 1.1 criteria. At day 60, she met the criteria for progression and started to receive other chemotherapy. The subject was treated at a local hospital.
Subject GD006, 28 years old, was first diagnosed with TNM stage III appendiceal mucinous cystadenocarcinoma identified during surgery in 2016. She was first treated with 1 cycle of oxaliplatin and 5-FU, followed by a second surgery and 3 rounds of intraperitoneal heat perfusion of oxaliplatin, irinotecan and raltitrexed. She progressed with local recurrence followed by surgery and 3 rounds of intraperitoneal heat perfusion of oxaliplatin, irinotecan and raltitrexed, followed by 4 cycles of Xeloda (capecitabine) with progression. She underwent steady-state T cell harvest. Before ITNK cell infusion, she received cyclophosphamide as lymphodepleting chemotherapy, which caused acute nausea and vomiting. Therefore, fludarabine administration was suspended for this patient. The subject underwent intra-arterial infusion of ITNKs into both uterine arteries at dose level 1 on 9 June 2019 (D0) without infusion-related complications. She then received intravenous infusion of ITNKs at dose level 2 on 11 August 2019 (D63) without infusion-related complications. She did not experience any CRS or neurotoxicity. At days 28, 60 and 90, staging evaluations revealed stable disease according to the RECIST 1.1 criteria. However, at day 120, she had progressive disease and was begun on PD-1 antibody and bevacizumab with progression. She is now on PD-1 and CTLA-4 antibodies and bevacizumab and has shown an initial response; she also underwent a third debulking surgery in the abdomen. She remains alive without new or ongoing study-related adverse events post ITNK cell infusion.
Subject GD007, 63 years old, was first diagnosed with TNM stage IV nonsmall cell lung adenocarcinoma in his right upper lung in 2017. He was first treated with erlotinib but experienced progression and was then treated with osimertinib. He developed EGFR-TKI resistance and was treated with 4 cycles of chemotherapy, including pemetrexed, carboplatin and apatinib, with continued progression. He had not received active therapy for 3 months prior to T cell harvest for ITNK cellular immunotherapy. The subject underwent lymphodepleting chemotherapy and received intravenous infusion of ITNKs at dose level 1 on 24 October 2019 (D0) without any infusion-related complications. He experienced slight fever and fatigue without CRS or neurotoxicity. At day 28, staging evaluation showed mixed response and stable disease according to the RECIST 1.1 criteria; in January 2020, disease progression was noted and was followed by other unknown therapies. His disease is stable, and he had no new or ongoing study-related adverse events 90 days post ITNK infusion.
Subject GD008, 59 years old, was first diagnosed with TNM stage IV rectal adenocarcinoma manifested by multiple metastases in the liver and abdominal lymph nodes in 2019. He was first treated with 9 cycles of mFOLFOX6 plus bevacizumab and local radiotherapy upon the fifth cycle of chemotherapy. He progressed and received transarterial chemoembolization twice for liver metastases. After lymphodepleting chemotherapy, the subject received an intratumoral injection of ITNKs into three liver lesions at dose level 1 on 4 March 2020 (D0). Because this subject was the first intratumoral recipient of ITNKs, dose escalation was not performed. He received an intravenous infusion of ITNKs at dose level 1 on 9 March 2020 (D5). He experienced slight fever and fatigue upon the first 2 days of cell infusion but no CRS or neurotoxicity. At day 45, staging evaluation revealed stable disease according to the RECIST 1.1 criteria. However, the second harvest of PBMCs failed, and he had to leave the trial to seek other therapies.
Subject GD009, 68 years old, was diagnosed with TNM stage IIIB melanoma from the upper palate mucosa manifested by invasion of the palate bone in 2017. He was first treated with 4 cycles of dacarbazine, cisplatin and bevacizumab with progression followed by 2 lines of therapy, including oxaliplatin, epirubicin and ipilimumab, with progression. The patient underwent T cell harvest prior to fludarabine and cyclophosphamide lymphodepleting chemotherapy before ITNK cell infusion. He received a venous infusion of ITNKs at dose level 1 on 27 August 2019 (D0) without infusion-related complications. He then received intravenous infusion of ITNKs at dose level 2 on 1 October 2019 (D35) without infusion-related complications. He experienced slight fever and no CRS or neurotoxicity. At day 45, staging evaluation revealed progressive disease according to the RECIST 1.1 criteria. He was begun on other chemotherapy with albumin-paclitaxel and carboplatin and radiotherapy outside of this trial.

Statistic
Statistical significance was determined using Student's t test (two groups) or ANOVA with Tukey's multiple comparison test (three or more groups). Survival was plotted using a Kaplan-Meier survival curve, and statistical significance was determined by the log-rank (Mantel-Cox) test. All statistical analyses were performed using Prism software, version 7.0 (GraphPad, Inc., San Diego, CA, USA). Statistical significance was indicated at *P ≤ 0.05, ** P ≤ 0.01, and ***P ≤ 0.001.

Reprogramming human cord and peripheral blood T cells into ITNKs
To inactivate BCL11B, we transfected human cord blood (CB) T cells with a plasmid expressing Cas9-EGFP and sgRNAs targeting either exon 2 or 3, alone or in combination (Figs. 1A and S1A). Cells transfected with both sgRNA and Cas9 cultured under T cell conditions showed limited cell death (Fig. S1B) and still expressed GFP 5 days after electroporation (Fig. S1C), and up to 47.2% of the cells had become NKp46 positive at Day 14 (Fig. S1D). DNA sequencing confirmed gene editing at the BCL11B locus, including frameshifts in exons 2 and 3 (Fig. S1E), and western blot analysis revealed a loss of the BCL11B protein in NKp46 + CD3 + cells (Fig. 1B). In the experiments presented below, we used two sgRNAs targeting Fig. 1 Reprogramming of primary human T cells into ITNKs by inactivating BCL11B. A sgRNA targeting exon 2 and exon 3 of the BCL11B locus. sgRNA, Cas9 and EGFP elements were integrated into a single vector. B Western blot analysis of BCL11B (120 kDa) levels in three representative samples of CB-derived T cells that were transduced with sgCtrl or NKp46 + CD3 + cells (purity: 92.41 ± 2.60%) that were sorted from sgBCL11B-engineered T cells. C Representative flow cytometric detection of CD3, CD56, NKp30 and NKp46 in T cells: T cells transduced with sgCtrl, T cells transduced with sgBCL11B and normal NK cells (CD3 − CD56 + ). Data are representative of five independent experiments. D Graph summarizing the percentages of CD56 + , NKp30 + , and NKp46 + cells in CD3 + T cells that received sgBCL11B or sgCtrl at 14 days post electroporation. The mean values of five independent healthy donors are shown. P < 0.001 for CD56 + , NKp30 + and NKp46 + T cells in sgBCL11B-electroporated T cells compared to sgCtrl-electroporated T cells. ***P ≤ 0.001, two-way ANOVA with Sidak's multiple comparisons test. E TCR diversity in sgCtrl T and NKp46 + CD3 + cells purified from sgBCL11B-edited T cells from the same donor based on variable chain sequencing data for the TCRβ locus. The 20 variable chain sequences at the TCRβ locus were analyzed exons 2 and 3 of BCL11B (sgBCL11B) to increase the efficiency of BCL11B gene editing. In addition to increasing expression of NKp46, sgBCL11B-transduced T cells also upregulated the NK cell markers CD56 and NKp30 while retaining expression of the T cell marker CD3 (Fig. 1C, D). Conversely, T cells expressing an unrelated sgRNA (sgCtrl) lacked NKp46 expression (Fig. 1C, D). NKp46 + CD3 + cells were viable, as verified by imaging flow cytometric analysis (Fig. S1F). The T cell origin of NKp46 + CD3 + cells was confirmed by analyzing their TCRβ repertoire and comparing it to that of polyclonal T cells from the same donor (Fig. 1E). Compared to sgCtrltransduced T cells, sgBCL11B-transduced T cells did not show an effect on growth through 21 days post electroporation (Fig. S1G). In addition to their derivation from CB T cells, ITNKs could be efficiently generated by ablating BCL11B in peripheral blood T cells (Fig. S2A-B). Since cell populations derived from either CB or peripheral blood mononuclear cells (PBMCs) contain various types of lymphocytes, including both CD4 + T cells and CD8 + T cells, we investigated the cellular heterogeneity of ITNKs. A subset of ITNKs from CB or PBMCs were also either CD8 + or CD4 + , with CD8-expressing cells being more frequent (Fig. S2C, D). CD4 + ITNKs strongly expressed CD56 and NKp30 (Fig. S2C, D). BCL11B deletion did not affect the survival or expansion of CD8 + and CD4 + T cells during 3 weeks of culture (Fig. S2E). Furthermore, upon BCL11B loss, memory (CD45RA − /CD45RO + ) and effector (CD45RA + /CD45RO − ) CD4 + and CD8 + T cells were reprogrammed into NKp30 + CD3 + CD4 + and NKp46 + CD3 + CD8 + ITNKs, respectively (Fig. S3A). Of interest, the percentages of ITNKs in CD8 + and CD4 + effector T cell cultures were higher than those in CD8 + and CD4 + memory T cell cultures, respectively (Fig.  S3A), suggesting that effector T cells reprogram into ITNKs more efficiently than do memory T cells after BCL11B ablation. In addition, we detected γδTCR + ΙTNKs with a γδTCR + NKp30 + phenotype (Fig. S3B) and mucosal-associated invariant T (MAIT)-derived ITNKs, which expressed both TCRvα7.2, a MAIT cell marker [58], and NKp30 (Fig. S3C), following BCL11B inactivation. We therefore surmise that acute loss of BCL11B reprograms major human T cell subtypes, including CD8 + T, CD4 + T, γδ T, and MAIT cells, into ITNKs.

ITNKs acquire transcriptional profiles of NK cells
We further performed time-resolved scRNA-seq analysis of the same samples evaluated by CyTOF and identified three CD8 + clusters and CD4 + clusters (Fig. 3A, B). Analysis of the expression profiles of T cell-associated surface markers (CD3, CD8A, and CD4) and transcription factors (TCF7 and LEF1) , and the NK cell-associated genes (NCAM1, NCR3, ID2, IL2RB, and NFIL3), [64,65] revealed that CD8 + ITNKs were identified mainly in Cluster 5, whereas CD4 + ITNKs were enriched in Cluster 2 (Fig. 3B, C). At the individual gene level, violin plot analysis further confirmed the upregulation of NK cellassociated genes and the concomitant downregulation of T cell-related genes in Clusters 5 and 2 compared to other clusters of CD8 T cells and CD4 T cells, respectively (Fig. 3D), which were consistent with the CyTOF analysis (Fig. S4A). Other highly upregulated genes in ITNKs included cell proliferation-related genes (CCNB1, CCNB2, and CDKN2D) [66] and AP-1 family members (JUN, JUNB, and JUND), which regulate T cell activation and proliferation [67] (Fig. 3D).
To further investigate the ITNK gene expression signature, we conducted bulk RNA-seq analysis of T cells (CD3 + ), ITNKs (CD3 + CD4 + NKp30 + and CD3 + CD8 + NKp30 + ) and NK cells (CD3 − CD56 + ) derived from PBMCs. Principal component analysis (PCA) and unsupervised hierarchical clustering analysis revealed that ITNKs exhibited global transcriptomic features of both T cells and NK cells (Fig. 3E, F). Although ITNKs still expressed genes associated with TCR signaling, in general, they downregulated TCF7 and LEF1 ( Fig. 3F and Tables S1, S2, S3), which was consistent with our scRNA-seq analysis (Fig. 3C, D). In addition to NCRs, genes encoding key NK cell-associated transcription factors, including ID2, ZBTB16, NFIL3, and ZNF35 (the human homolog of Zfp105), [68], exhibited increased expression in ITNKs compared to T cells (Fig. 3F). KEGG analysis of the upregulated differentially expressed genes (DEGs) showed that NK cell-associated pathways (e.g., natural killer cell-mediated cytotoxicity and the Fc epsilon RI signaling pathway) were enriched in ITNKs compared to T cells (Tables S1, S2, S3), which is in line with previous findings showing that NK cell-associated genes are upregulated in Bcl11b-deficient murine T cells [16]. Thus, these analyses show that ITNKs acquire expression of NK cell-associated genes at the transcriptional level.
We next evaluated the antitumor effects of ITNKs in xenograft models. ITNKs and T cells shared similar expression profiles for CD45RA and CD45RO ( Fig. S5F and Table S4). K562 cells expressing luciferase for bioluminescence imaging (BLI) were inoculated into immunocompromised mice (NSI: NOD/ SCID/IL2RG −/− ) [74], followed by infusion of ITNKs, NK cells, or T cells (Fig. S5G). ITNK or NK cell infusion caused a significant decrease in the tumor burden measured by BLI 28 days after K562 cell inoculation (Fig. 4E, F) and prolonged survival (Fig. 4G). Notably, ITNK transplantation did not cause any detectable symptoms of graft-versus-host disease (GVHD) up to 3 months after infusion, as the GVHD scores [75] of the ITNK and NK groups remained at zero (Fig. S5H). Similar results showing ITNK-mediated suppression of tumor cell xenografts were obtained for solid tumor cells (human primary HCC tumor cells in Fig. 4H and HeLa cells in Fig. S5I, J). Human ITNKs were therefore found to be effective in inhibiting both blood-derived and solid tumors in the xenograft models tested.
Engineered ITNKs were also evaluated in vivo by injecting K562-CD19 or HepG2 cells into immunocompromised NSI mice, followed by administration of CAR-ITNKs, ITNKs, CAR-T cells, or T cells (Fig. S6F). CAR19-ITNKs (Fig. S6G, H) and CARGPC3-ITNKs (Fig. 5D) demonstrated substantially more potent antitumor activities than CAR-T cells in the xenograft models. Therefore, engineering ITNKs with CAR molecules targeting specific tumor antigens further enhances the cytotoxicity of ITNKs, as CAR signaling and NCR signaling have been shown to synergistically achieve target cell recognition and cytotoxicity [80].

ITNKs as a potential cell source for treating refractory and advanced solid tumors
The potent antitumor capacity of human ITNKs encouraged us to explore their potential in preclinical and clinical applications. We first examined the distribution of human ITNKs in samples of peripheral blood, spleen, bone marrow, liver, and lung of immunocompromised NSI mice taken on days 1, 7, 14, 21 and 180 post ITNK infusion (Fig. S7A). ITNKs persisted for 2 to 3 weeks in the recipients but were undetectable 180 days post infusion (Fig. S7B). The presence of ITNKs did not appear to cause GVHD, as mentioned above (Fig. S5H). We next carefully evaluated possible off-target genetic changes induced by the BCL11B-CRISPR strategy by wholegenome sequencing of purified CD3 + NKp46 + ITNKs derived from two individual donors (Fig. S7C). A computational tool predicted 669 potential off-target sites with 3 mismatches and 1 bulge size [81,82]. After excluding candidate sites with no insertions or deletions (indels) near the predicted off-target sites in ITNKs, we analyzed 16 candidate sites, including the two on-target loci (Table  S5). Whole-genome sequencing results showed no indels in the other 14 candidates off-target sites (Table S5).
Encouraged by these results, we designed a clinical investigation to assess the safety and efficacy of infusing autologous ITNKs in patients diagnosed with advanced and metastatic cancer who did not respond to multiple prior therapies (Tables 1 and S6). Autologous ITNKs were manufactured from the PBMCs of nine patients. The manufacturing process and clinical protocol (NCT03 882840) for the ITNKs and the CONSORT diagram are depicted in Figs. S7D, S8, and S9. The median age of the patients was 49.6 years, and six of the patients were men (Table 1). Tumor biopsies from patients GD001-GD006 were available and evaluated for the expression of HLA-I molecules that suppress NK cell activation [83] or for NCR ligands, which stimulate NK cytotoxicity against tumor cells [84] (Fig. S10A). Autologous ITNKs were manufactured from the PBMC fraction of the blood from the nine patients. All products were subjected to detailed release criteria testing (Table S7). We also sequenced the ITNKs from the 9 patients and applied Cas-OFFinder to predict potential off-target sites [85] (Fig. S10B). The median frequency of mutations, all of which were intron variants, at the 20 off-target sites tested was 0.1% (range, 0-1%). In contrast, the median frequency of mutations at the BCL11B locus, including intron variants, frameshifts, and in-frame deletions, was 9.8% (range, 7.0-12.0%) (Fig. S10C). ITNKs were cultured for up to 21 days with a 3-to 76-fold expansion while they showed no longterm growth prior to infusion (Table S8). ITNKs used for this trial, defined as CD3 + NKp30 + cells, comprised 17.2-58.1% of all cells (Fig. S11A and Table S8) and were capable of efficiently lysing K562 cells in vitro (Table 1 and Fig. S11B).
The eight patients received intravenous or intra-arterial infusion of 0.31-102.04 × 10 6 ITNKs per kg of body weight after cyclophosphamide (Cy) and/or fludarabine (Flu) lymphodepletion [86] (Table 1). In all cases, patients readily recovered from cell infusion-related side effects such as slight fever, fatigue, or muscle pain and were released after at least 1 week of inpatient observation (Table 1). Serum levels of 40 cytokines, including IL-6, a common indicator of immunotherapy response and cytokine release syndrome (CRS) [87], were measured in patients before and at different times after ITNK infusion ( Fig. S11C and Table S9). Elevated levels of serum IL-6 were detected in seven of nine ITNK-treated patients (Fig. 6A). Of interest, the patients with the highest concentration of IL-6 (GD003) suffered only slight fever, fatigue and muscle pain for 2 days (Table 1 and Fig. 6A). Notably, several cytokines, including PDGF-BB, MIP-1β, GM-CSF, M-CSF and TNF-α, were significantly elevated in GD002 patients from day 3 post ITNK infusion onwards ( Fig. 6B and S11C and Table S9), consistent with in vitro assays showing GM-CSF secretion (Fig. 4B).
No patients experienced CRS, neurotoxicity, or other severe adverse effects attributed to the cell infusion (Table 1, also described in the Materials and Methods section). They were subjected to clinical outcome evaluation according to the RECIST 1.1 standards at 1 month post infusion. Computed tomography (CT) scan imaging diagnostic analysis showed that tumors in patients GD004, GD005, GD006, GD007, and GD008 were clinically stable post ITNK treatment ( Fig. 6C and S11D). Notably, patient GD002, who had undergone relapse with sacrum metastases after previous surgery, chemotherapy, and radiotherapy, achieved partial remission (PR) as revealed by both MRI and PET-CT scan imaging after 16 months. This subject had been intravenously infused 14 times with ITNKs but without any other medical treatment (Fig. 6C). In patient GD003, the tumor marker CA199 decreased in serum 1 month after receiving three doses of ITNKs by hepatic artery and ascending aortic injection (Fig. S11E). However, his tumors, as well as patient GD009, continued to progress (Fig. 6C and S11D). The overall characteristics of the patients and their clinical outcomes are summarized in Table 1. Retrospective analyses showed that stable responses are more likely to be achieved in patients whose ITNKs lysed K562 cells more efficiently and whose tumors exhibited high levels of NCR ligands and HLA-I expression (Table 1, Fig. 6D). In particular, the cytotoxicity of ITNKs in subject GD002 was the most potent and this patient achieved PR (Table 1). Taken together, the results of our preliminary clinical investigation indicated that ITNK infusions TNF-α were significantly elevated in patients (GD002, GD004-GD007) after ITNK infusion on day 3 based on Benjamini-Hochberg-adjusted p value. C Representative CT scans and PET-CT (MRI) of patients before (baseline) and after ITNK infusion at indicated time points post initial ITNK infusion. Tumors in patients GD004, GD005, GD006, GD007 and GD008 were stabilized (all tumors increased less than 20% in size), and GD002 achieved partial remission (PR), as revealed by both CT and PET-CT scan imaging diagnostic analysis at 16 months (34.6% decrease in tumor size), whereas tumors in patients GD001, GD003 and GD009 progressed following ITNK treatment (all tumors increased more than 30% in size). Red circles indicate the monitored tumor sites. D Percentages of ITNK cytotoxicity for patients with progressing diseases and patients with stable diseases. The assay was performed against K562 cells at an E: T ratio equal to 1:1 measured 24 h after coculture. Data are shown as the mean ± SD; P = 0.0082 (stable vs. progressing); unpaired t test, **P ≤ 0.01 following Cy/Flu lymphodepletion are well tolerated and provided clinical benefit in 6/9 patients and one patient achieved partial remission.

Discussion
Here, we demonstrate that acute loss of BCL11B reprograms mature human T cells into ITNKs, which express a diverse TCR repertoire and both T-and NK cell-associated genes and exhibit functional properties of both cell types (Table S11). Down-regulation of BCL11B also occurs naturally in the NKG2C + CD8 + T cells with an NK-like genetic signature and phenotype in HCMVseropositive individuals [88]. The alterations in gene expression caused by BCL11B deletion in human T cells resemble, in various aspects, how Bcl11b ablation affects murine T cells [16][17][18], suggesting evolutionarily conservation. Innate lymphoid cells (ILCs) are a recently identified family of lymphoid effector cells comprising both "cytotoxic" ILCs (NK cells) and "helper" ILCs. The orchestration of ILC and CD4 + T cell subset differentiation is remarkably similar [89,90]. Thus, similar to Th1 cells, ILC1 cells express TBX21 and produce IFN-γ [91], while ILC2 development requires Bcl11b [92]. ILC3 cells in turn express RORγt [93], which is absent from CD4 + ITNKs. However, future studies are required to compare the transcriptomes of CD4 + ITNKs with the different types of ILC cells.
Human ITNKs effectively eliminate leukemia and solid tumor-derived cells in culture. ITNKs expressed both TCRs and NCRs and efficiently killed leukemia cells in vitro through these TCRs and NCRs, such as NKp30 and NKp46 (Figs. 4A-D and S5D). Consistent with these findings, murine ITNKs use their TCR to recognize and lyse lymphoma cells expressing MHC-I molecules in vitro [16]. Human ITNKs were also found to target tumor cells in organoids and after transplantation into xenograft models. Of note, although ITNKs did not lyse K562 or primary HCC organoids as efficiently as NK cells did in vitro (Figs. 4C and S5E), they were superior in suppressing the growth of K562 tumors and multiple types of solid tumors in xenograft models (Figs. 4E-H, S5J). The differences between in vitro and in vivo antitumor effects may be due to the greater proliferation of ITNKs than NK cells in vivo. The mechanisms underlying ITNK activation and elimination of tumors in an in vivo microenvironment demand further investigation. ITNKs could be derived from various T cell subsets, including CD8, CD4, and γδT cells, and found to be functionally heterogeneous. CD8 + ITNKs expressed NKp46 and lysed K562 cells efficiently, while CD4 + ITNKs did not. However, CD4 + ITNKs secreted Th1 cytokines, including TNF-α (Fig. S5A, B). Further preclinical studies are required to identify the ITNK subsets with the most potent antitumor effects.
Our preliminary clinical results with infusions of ITNKs against solid tumors showed that they are safe and elicited clinical benefits in 6/9 patients. No clonal T/NK malignancies were observed in the limited number of phase I trial recipients treated to date, but long-term follow-up studies will be needed for all gene-modified cellular therapies to monitor possible late events. Optimization of CRISPR/Cas9 delivery into T cells, such as through the use of ribonucleoprotein (RNP) [9], may not only prevent plasmid-induced toxicity but also increase ITNK output by improving the deletion efficiency of BCL11B in T cells. With an improved ITNK manufacturing protocol, we can avoid undesired variability in dose levels and further escalate ITNK dosages, given that ITNKs were well tolerated in this study. In addition, due to the high lentiviral transduction efficiency of ITNKs, we can generate CAR-ITNKs to improve the specificity of ITNK recognition against tumors and introduce suicide genes such as inducible Caspase 9 (iCasp9) into ITNKs to avoid potential T cell leukemogenesis.
Human ITNKs express functional T cell receptors and NK cell receptors, and effectively kill both blood and solid cancer cells in culture, organoids, and mouse xenograft models. Remarkably, ITNKs resulted in tumor stabilization in six out of nine patients with advanced solid tumors in a preliminary clinical trial. Critically, no severe adverse effects have been observed in all these patients. ITNKs thus offers several major potential advantages over the immune cells currently being test clinically, e.g. CAR-T, CAR-NK: Robust proliferation, potent killing capacity, broad killing spectrum, no obvious severe adverse effects, and allowing further genetic engineering for example CAR-ITNKs to enhance specific killing. We think ITNK cells could serve as a more feasible and powerful choice compared to T cells or NK cells for clinicians to apply to general oncologic practice for multiple types of hematological or solid tumors, especially for those cancers with low scores of HLA-I or/and high scores of NCR ligands.

Conclusions
Taken together, these studies show that ITNKs may provide a new cellular source for adoptive cell therapy.