Treatment with Humanized Selective CD19CAR-T Cells Shows Efficacy in Highly Treated B-ALL Patients Who Have Relapsed after Receiving Murine-Based CD19CAR-T Therapies

Clinical trials using CD19 chimeric antigen receptor (CAR)-T cells have shown impressive results in treatment of B-cell malignancies such as acute lymphoblastic leukemia (ALL; refs. 1–3), chronic lymphocytic leukemia (CLL) (4), and non-Hodgkin lymphoma (5, 6). The scFv sequences used in these published clinical trials were based on murine antibody sequences (FMC63- or SJ25C1-mAbs; refs. 7–9). Recent studies have confirmed that host immune responses can recognize the epitopes of the murine scFv and rendered subsequent infusions ineffective (10–14).

Another problem with current technology is that generation of CAR-T cells before infusion normally involves nonselective expansion of all CD3⁺ T cells. Patient-derived T cells showed a large variation in the proportion of CAR-transduced T cells (5%–50%) in the final products (15, 16). The variation in CAR transduction rates, and the nonselective expansion of both CAR-positive and -negative T cells prior to infusion, added uncertainty to quality control and interpretation of clinical results, sometimes even leading to failure of CAR-T manufacture. To address this issue, a selective domain can be incorporated in the linker sequence between the heavy chain and light chain. LaSS-B, a stress-dependent shuttling autoantigen containing 408 amino acids, is a natural protein that exists in human cell nucleus (17, 18). A short peptide derived from LaSS-B, an epitope termed E-tag, can be incorporated into the linker area of two particular CARs recognizing prostate stem cell antigen (PSCA) or CD33, without affecting their functionality (19).

In this study, we developed a humanized CD19 CAR and incorporated the 10-amino acid E-tag as a selective domain between the heavy and light chains of scFv. Through activation with selective domain-specific mAbs, SmAb, CAR-transduced T cells were selectively expanded that led to a higher purity and a greater proportion of T central memory (Tcm) cells in the final product. Humanized selective (hs) CD19-specific CAR T cells were used to treat 5 B-ALL patients who had previously received murine scFv CD19 CAR-T therapies.

CD19 hsCAR comprised a Kozak consensus ribosomebinding sequence, a human CD8 signal peptide (SP), the humanized sequences encoding the VH and VL regions referring to US7635472B2 with the order of VH and VL reversed, and a 20-residue linker peptide containing a 10-amino acid selective domain derived from human nuclear protein, LaSS-B. The humanization of the VH and VL sequences was confirmed by using a web-based program at www.abysis.org. The CAR also included a hinge domain, a transmembrane region of the human CD8a molecule, and an intracellular signaling domain containing 41BB used as a costimulation domain, and CD3ζ domains used as activation regions. CD19 hsCAR gene was cloned into pLenti6.4 lentiviral vector backbone (Thermo Fisher Scientific) for production of the third-generation lentivirus.

CD19 hsCAR lentivirus was generated and stored at −80°C. Concentrated viral titers were determined by a limiting dilution method on HT1080 cell line using flow cytometry, as described previously (20).

The binding affinity of CD19 hsCAR was assayed by using microscale thermophoresis (MST) following the protocol as described previously (21). Briefly, purified recombinant extracellular domains of human CD19, CD19 hsCAR and murine-based CAR (derived from FMC63 clone, named as CD19 mCAR) were obtained by expression in HEK293T, and then purified by going through Ni-NTA columns. The extracellular domain of CD19 was labeled by NT-647 using Mo-L001 MonolithNT^MT protein labeling kit RED-NHS (NanoTemper) following the manufacturer's instructions, and was added into the reaction system as ligand; then purified extracellular domains of CD19 hsCAR and CD19 mCAR were incubated with target for binding affinity test at room temperature. Affinity value was measured by Monolith NT.115 (NanoTemper). The results were analyzed by MO Affinity analysis software (version 5).

The cell lines used in this study were obtained from the following sources. HEK293T was a kind gift from Palmer lab at Stanford University (Stanford, CA). Raji, a human Burkitt lymphoma cell line expressing CD19, and HT1080 cell line were purchased from TongPai Biotechnology Co., Ltd (Shanghai, China). Raji cells constitutively expressing luciferase (Raji-Luc) was generated by Beijing Vitalstar Biotechnology Co., Ltd. The cell lines were tested to be Mycoplasma-free by using Rapid Mycoplasma Test Kit (Cellapybio). HEK293T was cultured in DMEM. Raji, Raji-Luc, and HT1080 were maintained in RPMI1640. DMEM and RPMI1640 media were supplemented with 2 mmol/L l-glutamine, 10 mmol/L HEPES, and 10% heat-inactivated FBS. After thawing, the cell lines were used within 5 passages for the experiments in this study.

Peripheral blood mononuclear cells (PBMC) were isolated using Ficoll–Paque gradient centrifugation. CD3⁺ T cells were purified and activated by CD3/CD28 magnetic beads (Gibco) at the ratio of 1:2. Purified T cells were then seeded into T75 culture flasks with a density of 1.0 × 10⁶/mL for activation in X-VIVO 15 medium (Lonza) containing 200 IU/mL IL2 (XinLuoEr). The transduction was implemented by adding CD19 hsCAR- or CD19 mCAR-lentivirus with MOI of 25 within 24 to 48 hours. Transduction efficiency was evaluated by flow cytometric assay at day 5. For SmAb-induced restimulation, SmAb mAb (Sino Biological Inc.) was diluted by PBS and precoated in the culture flask at a concentration of 20 μg/mL overnight at 4°C. Expanded CD19 hsCAR-transduced T cells were transferred into SmAb-coated flask for restimulation on day 6 or day 7, according to the cell proliferation status. Final product was harvested from day 9 to day 12, depending on the cell proliferation status and yield, and was subjected to quality control for subsequent experiments or clinical administration (Supplementary Fig. S1A). For proliferation assay and population analysis, T cells were cultured for up to 15 days in vitro. PBMCs isolated from three healthy donors were used in preclinical tests for CAR-T manufacturing and functional evaluation.

Quality control for in-process product and final product was conducted according to the previous description (16). In general, sterility was tested by a 3-day cultivation prior to final product harvest. Mycoplasma test was performed by using Rapid Mycoplasma Test Kit (Cellapybio) following the manufacturer's instructions. Transduction efficiency and constitution were examined by using flow cytometry.

The following anti-human antibodies were purchased from BioLegend (BioLegend): anti-human CD3-FITC, anti-human CD19-PE, anti-human CD8-PerCP, anti-human CD4-APC, anti-human CD56-PerCP, anti-human CD14-APC, anti-human CD45RO-PE, anti-human CD197-APC, anti-human CD27-FITC, anti-human CD45RA-FITC, anti-human PD-1-APC, streptavidin-PE. Biotin-labeled protein-L was obtained from ACROBiosystems (ACROBiosystems) for CAR-positive T-cell measurement, and detection of biotinylated protein-L was implemented by adding streptavidin-PE (22). Flow cytometry was performed by using FACSCalibur (BD), and data were analyzed by using FlowJo (Treestar).

The cytotoxic activity on Raji was conducted by standard lactate dehydrogenase (LDH) release method. Target Raji cells (1 × 10⁴) were incubated with T cells transduced by CD19 hsCAR or CD19 mCAR at various effector/target (E/T) ratios (from 1:1 to 25:1) in 96-well microplates. After 12-hour cultivation at 37°C, the supernatant was harvested and cytotoxicity measured by using CytoTox 96 Non-Radioactive Cytotoxicity Assay Kit (Promega), following the manufacturer's instructions.

Cytokines were measured by ELISA or electrochemiluminescence (MSD) assay (Meso Scale Discovery). In vitro cell-mediated cytotoxic cytokines, including IL2, IFNγ, and TNFα, were tested by ELISA. Briefly, samples of supernatants were collected and measured using ELISA Kits (Neobioscience). Cytokines in serum samples from patients and animals were measured by Cytokine Panel 1 (human) Kit of MSD MULTI-SPOT Assay system (Meso Scale Diagnostics). ELISA data were acquired by using VARIOSKAN FLASH (Thermo Fisher Scientific), and MSD assay data acquired by using QuickPlex SQ120 system (Meso Scale Diagnostics).

Six- to 8-week-old female NOD/SCID IL2Rγc^−/− NSG mice were obtained from the Beijing Vitalstar Biotechnology Co., Ltd. Mice were injected intravenously with 1.0 × 10⁶ Raji-luc tumor cells through the tail vein, and received injections of 1.0 × 10⁶ CAR-modified or control T cells intravenously 3 days after tumor inoculation. Tumor progression and distribution were evaluated by serial bioluminescence imaging at indicated time points. Mice were given intraperitoneal injections of luciferin substrate (Promega) resuspended in PBS (15 mg/kg), and were anesthetized with chlorine ammonia ketone (100 mg/kg). Imaging was performed by using an IVIS Lumina K Serial III System (Caliper) from 12 to 14 minutes after luciferin injection in small binning mode at an acquisition time of 1 s to 1 min to obtain unsaturated images. Luciferase activity was analyzed by using Living Image Software (Caliper) and the photon flux analyzed within regions of interest that encompassed the entire body of each individual mouse. The animal studies have been approved by the local institutional review board.

ELISA assay was used to determine anti-serum response targeting CD19 CAR containing murine-based scFv or humanized scFv. Serum was collected from patients who previously received murine-based CD19 CAR-T infusions. Recombinant extracellular domain of CD19 mCAR and CD19 hsCAR (1 mg/mL in PBS) were diluted to 4 μg/mL using 0.1 mmol/L PBS, and were coated to the bottom of 96-well ELISA plates. Test wells were blocked for 30 minutes at 37°C using 1% BSA. Samples (100 μL) were added into wells and incubated for 1 hour at 37°C. HRP-labeled goat-anti-human antibodies specific for IgA, IgG, or IgM (Beijing ZhuangMeng Biotechnology Co., Ltd) were added into wells after 5 washes. TMB substrate solution was used for color development, and measurement performed by using microplate reader at 450 nm. Positive results were set at an OD450 value ≥0.2.

Phase I clinical trials (ChiCTR1800014761 and ChiCTR1800017439) were conducted to evaluate the safety and efficacy of CD19 hsCAR-T treatment. The following major inclusion criteria were used: (i) age < 75 years; (ii) patients with CD19⁺ refractory/relapse B-cell lymphoma, acute B lymphocytic leukemia, and chronic B lymphocytic leukemia, including patients who had previously been administered murine-based CD19 CAR-T cells; and (iii) patients with MRD-positive refractory/relapse acute B lymphoblastic leukemia, B-cell lymphoma, and chronic B lymphoblastic leukemia. All clinical investigations were conducted according to the ethical guidelines of Declaration of Helsinki. The clinical studies were performed after approval by the Institutional Review Board. The investigators have obtained written consent from all patients enrolled and treated in this trial.

Prior to CAR-T cell infusion, patients were treated with a fludarabine/cyclophosphamide (FC)–based lymphodepletion chemotherapy scheme consisting of intravenous administration of fludarabine (30 mg/m² for 3 days) and cyclophosphamide (250 mg/m² for 3 days) to remove endogenous lymphocytes. On day 0, the patients received a single dose of CD19 hsCAR-T infusion at a dosage of 0.3 to 3 × 10⁶ cells/kg body weight. The hsCAR-T dosage was determined by tumor burdens, previous murine CAR-T dosage used, and side effects, particularly cytokines release syndrome (CRS) grades following previous murine CAR-T treatment(s). Patients 1 and 2 were treated with autologous CAR-T cells, and patients 3, 4, and 5 with CAR-T cells produced from hematopoietic stem cell transplantation (HSCT) donors' PBMCs. Safety and efficacy were assayed according to the criteria of the National Comprehensive Cancer Network (NCCN) guideline, Acute Lymphoblastic Leukemia (Version 1. 2018).

Blood samples were obtained from patients before and at intervals after CD19 hsCAR-T infusions, and flow cytometry was performed to analyze the persistence of CAR-T cells. Circulating CD19 hsCAR-T cell numbers per 100 μL were calculated on the basis of the absolute CD3⁺ T lymphocyte counts. In addition, CAR DNA copies were assessed by qPCR (23) using primers specific to the scFv region of CD19 hsCAR. The primer sequences were listed below:

• Forward primer, 5′-GTCTGAACCGGCTGGGGATG-3′;

• Reverse primer, 5′-TGGATGCAGCCCGCCATTAG-3′.

Data were presented as mean ± SEM. Statistical analyses were conducted using Prism Software (GraphPad Software). For comparison between two groups, Student t test was conducted as a two-sided paired test with a confidence interval of 95%. For comparisons of three groups and above, the analysis was performed by using ANOVA. Results with a P value less than 0.05 were considered significant.

The design of humanized selective CAR was described in Materials and Methods and illustrated in Fig. 1A.

An improvement to the traditional CD19 CAR was the humanized sequence of the scFv region (Fig. 1B and C). Murine CD19 scFv sequence may be recognized by the host immune system and rendered the second murine CD19 CAR-T cell infusion ineffective (10, 14, 24). The humanization of the scFv region was in agreement with the prediction of the online software at www.abysis.org (Fig. 1B and C). An important determinant of the quality of CAR is the specificity to antigen. Using purified human CD19 protein as target antigen, we tested the affinity of murine CD19 CAR and CD19 hsCAR. The kinetics of binding between murine and humanized selective CARs to the ligand (hCD19) was plotted by using MST assay (Monolith NT.115, NanoTemper; Fig. 1D), and the KD values for CD19 mCAR and hsCAR were 509.4 ± 89.8 and 83.4 ± 12.2 nmol/L, respectively (Fig. 1E), suggesting that hsCAR possessed an affinity that was about 6-fold greater than that of mCAR.

Also included in the design of CD19 hsCAR was the selective epitope E-tag (17). Selective domain-specific mAb, SmAb, was coated to the bottom of plates during the ex vivo expansion stage. Compared with CD19 mCAR-transduced PBMCs with or without SmAb stimulation and CD19 hsCAR-transduced PBMCs without exposure to SmAb, CD19 hsCAR-transduced PBMCs exposed to SmAb showed a significantly greater proliferative capacity. On day 14, hsCAR-infected PBMCs exposed to SmAb expanded 81.6 ± 3.38 times, while hsCAR-infected PBMCs without exposure to SmAb expanded 61.7 ± 2.58 times, mCAR-infected PBMCs without SmAb stimulation expanded 58.2 ± 2.55 times and 53.26 ± 1.79 times with SmAb treatment (Fig. 1F).

On day 14, among the final product, more than 95% of the total cells were CD3⁺ T cells; a small percentage of the total cells (<5%) were CD3⁻/CD56⁺ natural killer (NK) cells, and few CD3⁻/CD19⁺ B cells or CD3⁻/CD14⁺ cells were detected by flow cytometry (Fig. 1G). Among the CAR-positive cells, the proportions of each population were presented in Supplementary Fig. S1B.

Evaluation of CAR-positive cells among the final products revealed that, on day 14, 64.73% ± 4.4% of total cells were positive for CAR in the “CD19hsCAR + SmAb” group, significantly higher than those of the other three groups (Fig. 1H–J). The results confirmed that the selective domain can be used to specifically expand the CAR-transduced cells.

CAR-T cells of different groups (CD19 mCAR, CD19 mCAR with SmAb, CD19 hsCAR w/o SmAb, CD19 hsCAR with SmAb) were incubated with Raji cell line at various E/T ratios. At the four ratios tested (1:1, 6.25:1, 12.5:1, and 25:1), CD19 hsCAR with SmAb all showed superior cytotoxicity versus the other three groups (Fig. 2A). Interestingly, at lower E/T ratios (6.25:1), CD19 hsCAR w/o SmAb also showed a greater cytotoxicity compared with CD19 mCAR-T cells with and without SmAb restimulation. Accordingly, CD19 hsCAR with SmAb, when incubated with Raji cells at E/T ratio 25:1, led to increased cytokine release of IL2, IFNγ, and TNFα in the medium, as detected by ELISA (Fig. 2B).

It is interesting to observe that prior activation with SmAb resulted in an improved cytotoxic function of T cells in vitro. We next examined whether such activation led to changes in proportions of T memory cell subpopulations. Compared with CD19 mCAR, CD19 mCAR with SmAb and CD19 hsCAR w/o SmAb groups, exposure of CD19 hsCAR group to SmAb led to a greater proportion of central memory T (Tcm) cell subpopulation and a smaller proportion of effector memory T-cell subpopulation (Fig. 2C–E). The proportions of each subpopulation among the CAR-positive cells were presented in Supplementary Fig. S1C and S1D.

To investigate which downstream signaling pathways had contributed to the enlarged Tcm subpopulation, we examined various pathways associated with T-cell differentiation, and found that ERK1/2, STAT3, P38, and AKT pathways were all activated by SmAb activation (Fig. 2F–K). The results indicated that SmAb restimulation could potentially induce T-cell differentiation via regulation on different signaling pathways.

Raji cells that expressed luciferase were intravenously infused into the immunodeficient (NOD/SCID IL2Rγc^−/− NSG) mice to establish the leukemia mouse models (n = 5/group). Normal healthy human PBMCs were used to generate various groups of CD19 CAR-T cells, which included mCAR, hsCAR, and a control hsCAR Stopper CAR (CD19 hsCAR lacking intracellular costimulation domain). PBS or each group of the CAR-T cells were intravenously infused into the leukemia mouse models 3 days after Raji cell injection (1 × 10⁶/100 μL Raji cells/mouse). T cells engineered by lentiviral transduction of EGFP were used as a mock group. CAR-positive T cells (1 × 10⁶/100 μL) were infused into each mouse in different CAR-T treatment groups; 1 × 10⁶/100 μL EGFP-positive T cells were infused for mock group, and equal volume of PBS for PBS group. Tumor load and distribution were monitored prior to treatment and 7, 14, 21, 35, 49, and 70 days after treatment (Fig. 3A; Supplementary Fig. S2). Mice treated with hsCAR-T cells activated with SmAb showed significantly longer survival than those with hsCAR-T cells w/o SmAb, and than those with mCAR-T cells (Fig. 3B–D). Analysis of mouse sera from 20 to 50 days posttreatment revealed a higher concentration of proinflammatory cytokines IFNγ, IL2, TNFα, IL6, and a lower concentration of IL15 (Fig. 3E–I).

Five refractory/relapsed B-ALL patients were included in the trial who had received previous murine CAR-T treatments and then relapsed with CD19⁺ B lymphoblastic cells (Supplementary Tables S1 and S2). Patient 1 was a 9-year old male as in 2018, who was diagnosed of B-ALL with a fusion genotype of E2A-HLF in September 2017 (treatment history listed in Supplementary Table S2). In May 2018, patient 1 received hsCAR-T therapy following lymphodepletion. Grade 1 CRS was observed after hsCAR-T treatment. On day 14, patient 1 achieved CMR (MDR-) and remained in CMR (MDR-) on day 28. Two months later, patient 1 received allo-HSCT and remained in CMR as of March 2019, when the manuscript was prepared. This case proved the concept that humanized CAR-T cells may effectively treat patients who relapsed after murine CAR-T therapy.

Patient 2 was a 14-year-old male patient as in 2018, and, in June 2017, he was diagnosed of B-ALL with a fusion genotype E2A-HLF and a complex chromosome karyotype (Supplementary Table S2). This patient received mCAR-T treatment twice but did not respond to the second murine CAR-T treatment (NR, Supplementary Table S2), which did not reduce the tumor load in bone marrow at any time point examined posttreatment and the disease remained progressive (lymphoblastic cells in bone marrow: day 0 before infusion of the second mCAR-T, 0.5% by morphology, 0.53% by flow cytometry, E2A-HLF fusion gene 0.94%; day 15, 1.5% by morphology, 0.6% by flow cytometry, E2A-HLF fusion gene 13.3%; day 36, 30% by morphology, 46.81% by flow cytometry, E2A-HLF fusion gene 119.58%). In May, 2018, patient 2 received CD19 hsCAR-T treatment. Considering the high tumor load, the dosage of hsCAR-T cells was lowered to 0.3 × 10⁶/kg body weight, to avoid possible side effect of severe CRS. On day 0 prior to hsCAR-T infusion (after lymphodepleting conditioning), the tumor loads in the bone marrow was 36% according to morphology evaluation, and 34.86% by flow cytometry. On day 15 posttreatment, tumor loads in bone marrow evaluated by morphology and flow cytometry were reduced to 10.5% and 14.98%, respectively; B-cell percentage in peripheral blood was also brought down to almost zero on day 15 (Supplementary Fig. S3B); but on day 30, the tumor loads in bone marrow bounced back to 82% (morphology) and 71.84% (flow cytometry), respectively. The results suggested that hsCAR-T cells might have been effective but exhausted after day 15 possibly due to the low dosage infused and the relative high tumor load. Indeed, the proportion of hsCAR-T cells in peripheral blood was brought down sharply after day 15 and they almost disappeared on day 30 (Fig. 5C). Unfortunately, when we were trying to infuse hsCAR-T cells for another time with a higher dosage, patient 2 quitted the trial by some unknown reasons.

Patient 3 was a 17-year-old male patient as in 2018, who was diagnosed of B-ALL with a fusion genotype BCR-ABL1 in the end of 2014. He had an extramedullary disease with the central nervous system (CNS) involvement, and received targeted therapy (Imatinib followed by dasatinib) and chemotherapy. After achieving CMR, the patient underwent haploidentical HSCT from his mother as donor in May 2015 (Supplementary Table S2). In June, 2017, patient 3 relapsed with extramedullary disease with CNS involvement with a tumor load of 8.96% in the cerebral spinal fluid (CSF). In August, 2017, patient 3 received CD19 mCAR-T treatment, and achieved CMR after experiencing grade 3 CRS and neurotoxicity. In March, 2018, patient 3 received another CD19 mCAR-T as a preventive measure (at this time, the patient had not relapsed). In late April, 29 days after the second time mCAR-T treatment, the patient relapsed with 66.13% lymphoblastic cells of the recipient origin in CSF and 0.02% in BM (evaluated by morphology). In June, patient 3 received CD19 hsCAR-T treatment following intrathecal chemotherapy and lymphodepleting conditioning. On day 5, the patient experienced grade 1 CRS. On day 15, CMR (MRD-) was achieved in bone marrow, peripheral blood, and the CNS (Supplementary Tables S1 and S2). On day 63, the patient relapsed with extramedullary disease with CNS involvement but still remained in CR in bone marrow and peripheral blood. Considering that patient 3 was clear of cancer cells in bone marrow and peripheral blood, he was given another haploidentical HSCT after intrathecal chemotherapy. As of March, 2019, patient 3 remained in CR in bone marrow, peripheral blood, and CNS.

Patient 4 was a 14-year-old female as in 2018 and was diagnosed of B-ALL in 2016 (Supplementary Table S2). She received CD19 mCAR-T treatment bridging to HLA-identical HSCT from her sister as donor. In August 2018, she relapsed and received CD19 hsCAR-T treatment in September 2018 (Supplementary Tables S1 and S2); the patient remained in CMR as of March 2019, when the manuscript was prepared.

Patient 5 was a 21-year-old female as in 2018 and was diagnosed of B-ALL in 2016. Similar to patient 4, patient 5 received mCAR-T treatment bridging to haploidentical HSCT from her father as donor. In September 2018, patient 5 relapsed and received CD19 hsCAR-T treatment in October 2018. This patient remained in CMR as of March 2019 (Supplementary Tables S1 and S2).

After infusion of hsCAR-T cells, various levels of expansion of hsCAR-T cells were detected in the peripheral blood of all 5 patients within 3 weeks postinfusion (Fig. 4A–D). Accordingly, the number and proportion of B cells in peripheral blood remained close to zero on day 30 in patient 1 and patient 3 who achieved CMR. In patient 2, who did not achieve CR, B cells started to be detected after day 20 (Supplementary Fig. S3A–S3C). In patient 4, B cells started to be detected after day 30 (Supplementary Fig. S3D); in patient 5, the proportion of B cells remained close to zero two months after hsCAR-T treatment (Supplementary Fig. S3E). The cytokine concentrations in the peripheral blood were measured and presented in Fig. 4E–L. A surge of IL2, IL6, sCD25, and TNFα levels was detected in the sera of patients 1 and 3, but not in that of patient 2. A surge of IL6 was also detected in patients 4 and 5.

We also compared the expansion of CAR-T cells and cytokine levels after hsCAR-T versus mCAR-T treatments (Fig. 5; Supplementary Fig. S4). Patient 1 received 0.3 × 10⁶ mCAR-T/kg body weight and 1 × 10⁶ hsCAR-T/kg body weight; prior to the mCAR-T and hsCAR-T treatments, the tumor loads in bone marrow were 5.04% and 4.00%, respectively. After treatments, the proportion of CAR-T cells in peripheral blood reached the highest value of 0.5% on day 3 following mCAR-T versus 61% on day 11 following hsCAR-T treatment (Fig. 5A). The number of CAR-T cells in 100 μL peripheral blood reached the highest value of 43 on day 11 following mCAR-T versus 15,000 on day 15 following hsCAR-T treatment (Fig. 5B). Patient 2 received mCAR-T infusions twice and hsCAR-T once, and all the three were at the same dosage of 0.3 × 10⁶/kg body weight. The tumor loads in bone marrow before each of the three infusions were 0.43%, 0.50%, and 46.8%, respectively. The proportion of CAR-T cells in peripheral blood reached the highest value of 6.36% on day 11 following the first time mCAR-T infusion and remained close to zero following the second time mCAR-T infusion (Fig. 5C). In contrast, following subsequent hsCAR-T treatment, the proportion went up to 79.64% on day 15. The number of CAR-T cells in 100 μL peripheral blood reached the highest value of 693 on day 11 following the first time mCAR-T infusion versus 26,800 on day 15 following hsCAR-T infusion (Fig. 5D). Very few mCAR-T cells were detected following the second time mCAR-T infusion. As to patient 3, the peak of CAR-T expansion was detected earlier (day 7) following the second time mCAR-T infusion versus following the first time mCAR-T infusion (day 11; Fig. 5E and F). Compared with the two mCAR-T treatments, the peak values of hsCAR-T cells were higher (17.96% vs. 4.45% and 3.85%, Fig. 5E; 6,690/100 μL vs. 2,280/100 μL and 3,070/100 μL, Fig. 5F). Patient 4 and 5 both received mCAR-T treatments before HSCT and hsCAR-T after. The percentages of CAR-T cells in peripheral blood and the numbers of CAR-T cells in 100 μL peripheral blood in patient 4 were comparable following mCAR-T versus hsCAR-T treatments (Fig. 5G and H). Patient 5 showed a higher degree of CAR-T expansion following hsCAR-T treatment compared with mCAR-T treatment (Fig. 5I and J).

The cytokine levels in the sera of the 5 patients following each treatment were presented in Supplementary Fig. S4A–S4T. With regard to patient 1, after mCAR-T treatment and relapse, hsCAR-T treatment still elicited proinflammatory cytokine response, such as IL6 and IFNγ (Supplementary Fig. S4A and S4F); and the level of anti-inflammatory cytokine, IL10, was lower (Supplementary Fig. S4K). As to patient 2, the second time mCAR-T treatment induced a surge in both IL6 and IL10. hsCAR-T treatment caused a higher expression of IFNγ, compared with mCAR-T results. No data of sCD25 after the second time mCAR-T treatment were available for patient 2 (Supplementary Fig. S4B, S4G, S4L, and S4Q). As to patient 3, after mCAR-T infusion for the second time, although the CAR-T cells seemed to have expanded to some extent (Fig. 5E and F), proinflammatory cytokine response was dampened compared with that after the first time mCAR-T treatment (Supplementary Fig. S4C and S4H). However, hsCAR-T treatment following the two mCAR-T treatments elicited proinflammatory cytokine response (Supplementary Fig. S4C and S4H) and reduced the level of anti-inflammatory cytokine, IL10 (Supplementary Fig. S4M). In patient 4 and 5, levels of proinflammatory cytokines IL6 and IFNγ were both lower following hsCAR-T versus mCAR-T treatments (Supplementary Fig. S4D, S4I, S4E, and S4J), possibly due to the suboptimal reconstruction of the hematopoietic system following HSCT (Supplementary Table S3), which might have subsequently affected the cytokine responses.

One of the reasons that might underlie the diminished efficacy following the second murine CAR-T treatment in patients 2 and 3 was that the patients may have developed antibodies that recognized the murine scFv sequences. To test this hypothesis, we analyzed the sera isolated from the 5 patients (Fig. 6A and B) and two healthy control people (Fig. 6C). The serum was tested for the reaction to the extracellular domains of murine and humanized CD19 CARs before and after hsCAR-T treatment. Patients 1-3 who received murine CD19 CAR-T cell treatment showed a positive IgA reaction to murine CAR; after treatment with hsCAR-T, no immunoglobulins reactive to humanized CAR was detected in the sera of the three patients (Fig. 6A). Similarly, sera from patients 4 and 5 did not show immunoglobulins reactive to humanized CAR (Fig. 6B, no blood samples available before HSCT). The data of reactivity to murine and humanized CD19 CARs from each individual patient were presented in Supplementary Table S4.

To further confirm the inhibitory effects, T cells derived from patients 1–3 were infected with murine CAR or hsCAR, and cocultured with Raji cell line at different E/T ratios with or without patient serum, to test the cytotoxic functions. In one experimental group, protein G was added together with the serum to block the effects of immunoglobulins (Fig. 6D–F). The cytotoxic functions of mCAR-T cells were inhibited by the patient serum, and incubation with protein G could rescue the inhibitory effects. In contrast, the cytotoxicity of hsCAR-T cells was not inhibited by patient serum (Fig. 6D–F).

One issue associated with the current CAR-T production process is that both CAR-transduced and nontransduced T cells are expanded. In an attempt to enhance the purity and yield, we developed the selective CD19 CAR. After activation with coated SmAb, CAR-transduced T cells in the final product were increased from 32.2% to 64.7% (Fig. 1H); the percentage was almost doubled. The yield of total CD3⁺ T cells on day 14 was also significantly greater, achieving around 80-fold expansion with SmAb stimulation versus around 60-fold expansion without (Fig. 1F). Theoretically, hsCAR-transduced T cells can be further purified by magnetic sorting using beads-conjugated SmAb, an approach that can be used to address whether CAR-negative T cells affect efficacy. Another merit of the selective CAR was that stimulation with SmAb during expansion stage can reshape the composition of T-cell subpopulations, leading to a higher proportion of Tcm in the product. According to the results, this impact was associated with activation of several signaling pathways, including ERK1/2, AKT, P38, and STAT3, which have been shown to be involved in T-cell differentiation process (25–28). Previous studies have shown that persistence of GD2 CAR-T cells is concordant with the percentage of Tcm in the final product (29). In our study, leukemia mouse models treated with SmAb-stimulated hsCAR-T cells displayed a longer survival compared with the other groups (Fig. 3C and D), suggesting an improved antitumor efficacy associated with SmAb-stimulated hsCAR-T cells. The CD4/CD8 ratios and other subpopulation composition in the final products for the 5 patients were also measured and presented in Supplementary Table S5. It seemed there existed a large variation in the CD4/CD8 ratios between the final products of each individual. The kinetics of CD8 Tcm cells after infusion into the 5 patients were presented in Supplementary Fig. S5.

The other major improved feature of the current CAR design was the humanized scFv sequence. This hsCAR showed a 6-fold higher affinity to CD19, compared with the murine counterpart. Other groups also reported the design of humanized CD19 CARs (30–32), but the affinity was comparable to the murine counterpart or not measured. At a higher E/T ratio which may lead to a saturated cytotoxic effect, hsCAR-T cells with a higher affinity and mCAR-T cells with a lower affinity showed comparable anti-tumor cytotoxicity. At a lower E/T ratio, hsCAR-T cells exhibited a better cytotoxicity (Fig. 2A). The comparable efficacies as observed in the animal experiments in mCAR-T group versus hsCAR-T group without SmAb stimulation were probably due to the saturated dosage of CAR-T cells in vivo (Fig. 3D). Also, the microenvironments in which humanized CAR-T cells kill human Raji cells were different in mice versus in humans. The mouse milieu may not preferably favor humanized CAR over murine CAR to implement their functions.

Humanized CAR-T cells have a potential to break the barrier caused by immune recognition of murine scFv following repetitive infusions of murine CAR-T cells. The two FDA-approved products, Kymriah and Yescarta, were based on murine scFv sequences (33, 34). Many of those patients will relapse eventually, even if CAR-T therapy is bridged to allo-HSCT. With an increasing number of patients undergoing such treatments, there is a pressing need to find a new interventional strategy to treat the patients who have relapsed with CD19⁺ leukemia cells after receiving murine CAR-T therapies. In this study, we enrolled 5 highly treated, refractory/relapsed, B-ALL patients who had exhausted all available options and who might represent different complex conditions. Patient 2 and 3 both received mCAR-T treatments twice. After the second time, mCAR-T cells failed to expand in patient 2 (Fig. 5C and D); mCAR-T cells in patient 3 expanded to some extent after the second time (Fig. 5E and F) but without triggering proinflammatory cytokine response (Supplementary Fig. S4C and S4H) or influencing B cell percentages in the PB (Supplementary Fig. S3C). Analysis of the sera from the three patients revealed that there were IgA that were reactive to the extracellular domain of CD19 mCAR (Fig. 6A). We further confirmed that there was an inhibitory component in the patient serum that could reduce the cytotoxic effect of mCAR-T cells (Fig. 6D–F), and the inhibitory effect was reversed by pulling down immunoglobulins through adding protein G (Fig. 6D–F). It was unclear whether patient 4 and 5 had developed antibodies to murine CAR before HSCT due to the unavailability of the blood samples prior to transplantation. Even though they had, the HSCT process would have eliminated the B-cell clones that produced those antibodies in the recipient, and reconstructed the hematopoietic system with the cells of donor genotype. The cases of patients 4 and 5 indicated that hsCAR-T may also replace mCAR-T as an earlier line of treatment option. The comparison of efficacies of hsCAR-T versus mCAR-T would require a separate study with a larger patient cohort. In all of the 5 patients, no antibodies reactive to hsCAR were detected in the sera, suggesting that hsCAR-T might be repetitively administered to the patients and remain efficacious.

The 5 patients represented various complex conditions. Patient 2 and 5 exhibited complex chromosomes; patient 3 had Philadelphia chromosome and CNS involvement. All of the above factors are associated with poor prognosis with conventional treatments. Patient 1 and 2 did not undergo HSCT; patient 3 went through HSCT before murine CAR-T treatments; patient 4 and 5 received HSCT after murine CAR-T therapies. For patients 3, 4, and 5, we used the healthy HSC donors' PBMCs for production of CAR-T cells, and no symptoms of graft-versus-host disease (GvHD) were observed. The results suggest that CD19 hsCAR-T may be employed in a wide spectrum of situations involved in B-cell malignancies.

Another noteworthy result was that the 5 patients all showed very mild side effects after hsCAR-T treatment (Supplementary Tables S6 and S7). Only a temporary fever occurred which quickly resolved. The CRS was only at grade 1 for all the 5 patients. One possible reason for the mild side effects might be due to the higher affinity of the scFv region of hsCAR. Compared with other reported CD19 humanized CARs (30, 31), hsCAR still possessed higher affinity to CD19. The higher affinity may be due to a different humanized scFv sequence used in our study, the reversed order of heavy chain and light chain, and/or incorporation of the selective short peptide, which might have contributed to a change in the three-dimensional configuration of the antigen-recognizing domain of CAR. Further studies are needed to elucidate the underlying mechanisms.

In summary, the humanized selective CD19 CAR is superior to its murine counterpart in terms of antigen-binding affinity, product yield, purity, and in vitro antitumor efficacy. The results from the highly treated patients confirmed that hsCAR-T therapy is effective to treat those who have relapsed from murine CAR-T therapies.

No potential conflicts of interest were disclosed.

Conception and design: P. Lu, Z. Chen

Development of methodology: Y. Zhao, J. Long, P. Lu, Z. Chen

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): Y. Zhao, Z. Liu, X. Wang, H. Wu, J. Zhang, J. Yang, F. Zhang, J. Long, P. Lu, Z. Chen

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): Y. Zhao, Z. Liu, X. Wang, H. Wu, J. Zhang, J. Yang, F. Zhang, L. Liu, J. Long, P. Lu, Z. Chen

Writing, review, and/or revision of the manuscript: Y. Zhao, Z. Liu, X. Wang, H. Wu, J. Zhang, J. Yang, F. Zhang, L. Liu, J. Long, P. Lu, Z. Chen

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): Y. Zhao, Z. Liu, X. Wang, H. Wu, J. Zhang, J. Yang, F. Zhang, J. Long, P. Lu, Z. Chen

Study supervision: J. Long, P. Lu, Z. Chen

The work was supported by the following grants. Stem Cell and Translation National Key Project (2016YFA0101403 to Z. Chen), National Natural Science Foundation of China (81661130160, 81422014, 81561138004 to Z. Chen), Beijing Municipal Natural Science Foundation (5142005 to Z. Chen), Beijing Talents Foundation (2017000021223TD03 to Z. Chen), Support Project of High-level Teachers in Beijing Municipal Universities in the Period of 13th Five–year Plan (CIT & TCD20180333 to Z. Chen), Beijing Medical System High Level Talent Award (2015-3-063 to Z. Chen), Beijing Municipal Health Commission Fund (PXM 2018_026283_000002 to Z. Chen), Beijing Talents Fund (2018A03 to Z. Chen), Beijing Municipal Administration of Hospitals Clinical Medicine Development of Special Funding Support (ZYLX201706 to Z. Chen), the Royal Society-Newton Advanced Fellowship (NA150482 to Z. Chen), and a special fund on CAR-T from Xuanwu Hospital (to Z. Chen).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.