Introduction

Chimeric antigen receptor T (CAR-T) cell therapy has demonstrated extraordinary potentials in the treatment of acute lymphocytic leukemia and B-cell lymphoma [1, 2]. However, immunotherapy-associated side effects strongly hampered the development of CAR-T cell therapy, as severe complications are life-threatening for patients [3, 4]. Cytokine release syndrome (CRS; manifested by fever, hypotension, hypoxia, and target-organ damage) and neurotoxicity (characterized by headaches, confusion, seizure, and other neurologic manifestations) are the most representative [3, 4]. Various inflammatory factors were found elevated in serum samples after infusing CAR- T cells; these factors include interleukin-6 (IL-6), interferon-γ (IFN-γ), granulocyte-macrophage colony-stimulating factor (GM-CSF), soluble interleukin-6 receptor (sIL-6R), macrophage inflammatory protein-1 (MCP-1), interleukin-1 (IL-1), etc. [5, 6]. Activated CAR-T cells release various inflammatory cytokines including GM-CSF, which subsequently plays a major role in the stimulation and differentiation of innate monocytes–macrophages lineage [5, 7, 8]. Stimulated monocytes–macrophages massively produce cytokines including IL-6 and IL-1. Inflammatory cascade is thus initiated and leads to CRS and neurotoxicity [6, 8, 9]. Previous studies reported that gene editing of GM-CSF in CAR-T cells could prevent against adverse effects in vitro and in xenograft mouse models without damaging the cytotoxicity of CAR-T cells [8, 10]. However, these studies were conducted using CRISPR/Cas9 or TALENs, which require either a dual transduction of CAR-expressing vector and GM-CSF-gRNA-lentiCRISPRv2 lentiviruses [10] or electroporation of mRNA encoding TRAC TALEN arms and subsequent AAV transfection [8] to prepare GM-CSFKO CAR-T cells. These additional manufacturing procedures might hamper the viability of CAR-T cells. Meanwhile, the knockout efficiency of GM-CSF through double transduction could be unpredictable, and additional examinations would be needed. In the present study, we inserted a short hairpin RNA (shRNA)-expression cassette in the CAR vector to reduce GM-CSF secreted by CAR-T cells. Regular manufacturing procedures would suffice to produce GM-CSFKD CAR-T cells and achieve comparable gene silencing efficiency to gene knockout, thereby promoting the clinical application of CAR-T cell therapy.

Materials and methods

Cells and reagents

K562 (ATCC, #CCL-243) and HEK-293T (ATCC, #ACS4500) cells were obtained from ATCC. Cryopreserved human adult purified PBMCs (MT-BIO, #PB010C) were purchased from MT- BIO. The cryopreserved PBMCs were thawed and cultivated in PRIME-XV T cell CDM (Irvine SCIENTIFIC, #91,154) at 1 × 106/ml supplemented with interleukin-2 (100 IU/ml) to prepare CAR-T cells. Anti-CD3/CD28 Dynabeads (Gibco, #1132D) were added at a bead-to-cell ratio of 1:1 for 24 h to stimulate and expand T cells. The amplified T cells were resuspended at 3 × 105/ml and transfected with lentiviruses (MOI = 3 × 105vg/cell) on RetroNectin (Takara, #T100B) pre-coated plates for 72 h. Transfection efficiency was assessed by detecting GFP expression through flow cytometry and the final transfection efficiencies were approximately at 30–50%.

Generation of GM-CSF knockdown CAR-T cells

Four sets of shRNA sequences (Supplementary Table. S1) for GM-CSF silencing were modified to be constructed on the pLVX-shRNA2 plasmid. The knockdown efficiency in K562 cells was compared through Western blot assay before application to CAR-T cells. The shRNA- expressing fraction was then amplified through polymerase chain reaction and inserted into the anti-human CD19 CAR (CAR19) vector. The CAR19 involved in this research contains an anti-human CD19 single-chain variable fragment (scFv), hinge and transmembrane (TM) regions, and intracellular signaling domains including coactivator 4-1BB as well as CD3 zeta. Lentivirus was produced in 293T cells with Lipofectamine 3000 (Thermo Fisher Scientific, # L3000015) and purified using ultrafiltration kit (Millipore, #UFC910096). Transfection was conducted as described above.

Flow cytometry analysis

For CAR19 expression assay, the transfected 293T cells were incubated with diluted biotinylated human CD19 (Acrobiosystems, #CD9-H8259) and then with BV605 Streptavidin (Biolegend, #405,229) following the manufacturer’s instruction. Flow cytometry was performed on BD LSR Fortessa X-20 Cell Analyzer. As for other cell surface markers, cells were harvested, washed in Dulbecco’s phosphate-buffered saline, and incubated with Human TruStain FcX (Biolegend, #422,302) to block the Fc receptor. Anti-CD3 (Gibco, #555,335, #555,340), anti-CD4 (Gibco, #555,349), anti-CD19 (BD, #562,440), anti-CD14 (Biolegend, #301,804), and anti-CD11b (Biolegend, #101,208) antibodies and Zombie NIR Fixable Viability Kit were added as needed for flow cytometry analyses. All flow cytometric data was analyzed on FlowJo X10.0.7r2.

Enzyme-linked immunosorbent assay

Nalm6 cells were co-cultured with CAR-T cells or mock T cells at 1:1 ratio for 16 h. Cell supernatant was collected for cytokine assays. ELISA kits of human IL6 (Biolegend, #430,504) and GM-CSF (Biolegend, 432,004) were used to detect the concentration of inflammatory cytokines according to manufacturer’s instructions.

Cell killing assay

Nalm6 cells were transduced with lentivirus for co-expression of luciferase-zsGreen (Hanbio Biotechnology) beforehand. Nalm6Luc + cells were later co-incubated with CAR-T cells or mock T cells at effector to target (E: T) ratios of 0.25:1, 0.5:1, and 1:1 for 12 h. D- Luciferin (Thermo Scientific, #8829) was added simultaneously to the cell culture. Residual live cells were measured by detecting bioluminescence on a Thermo Scientific Microplate Reader.

Monocyte cocultivation and endotheliocyte activation assay

Total monocytes were isolated from cryopreserved human PBMCs of the same donor by using the MojoSortTM Human Pan Monocyte Isolation Kit (Biolegend, #480,059) immediately after thawing. CD14 + monocytes were harvested, and about 2 × 105 was placed at the upper chamber of a 24-well transwell plate (pore size:3.0 μm), with equal numbers of CAR-T cells and Nalm6 cells cultured in the lower chamber. After 16 h, supernatant was collected for cytokine measurement.

Multi-analyte flow assay

Beads-based multi-analyte flow cytometry analysis was conducted using LEGENDplexTM Human Inflammation Panel 1 (13-plex, Biolegend, #740,809), which included IL-1β, IFN-α2, IFN-γ, TNF-α, MCP-1, IL-6, IL-8, IL-10, IL-12p70, IL-17 A, IL-18, IL-23, and IL-33, to quantify cytokines and chemokines in the supernatant of cocultured cells. Data were analyzed via Legendplex Version 8.

Statistical analysis

Statistical significance was calculated using Student’s t test on GraphPad Prism. The results were expressed as mean (SD). Two-sided p-values < 0.05 were considered statistically significant.

Results

To obtain GM-CSFKD-shRNA-CAR vector, we first inserted the shRNA sequences for GM-CSF knockdown on the pLVX-shRNA2 plasmid (Supplementary Table. S1). We chose four sets of shRNA and compared their knockdown efficiency in K562 cells by Western blot assays. Compared with the scramble shRNA group, GM-CSFKD-shRNA-2 (shGM2) exhibited moderate GM-CSF silencing capacity, while GM-CSFKD-shRNA-4 (shGM4) was the most effective (Supplementary Fig. S1). The shRNA expression cassette of shGM2/shGM4 was inserted into the anti-human CD19 CAR vector (Fig. 1a). The expression levels of FMC63 and GFP were not hampered in the modified CAR vector (Supplementary Fig. S2).

Cryopreserved human adult purified PBMCs were used to produce activated and amplified T cells. The transfected T cells were divided into five groups: mock T cells (GFP+), CAR19 T cells (CAR19+, GFP+), shGM2-CAR19 T cells, shGM4-CAR19 T cells, and scramble-CAR19 T cells (Supplementary Fig. S3). We used Nalm6 cells (a cell line of CD19+ B cell acute lymphoblastic leukemia) as target cells to coculture with CAR-T cells or mock T cells for 16 h. Cell supernatant was collected, and enzyme-linked immunosorbent assay (ELISA) experiments were conducted to detect the levels of GM-CSF. The average GM-CSF concentration in the regular CAR19 T supernatant was 25,086.21 ng/ml. The GM-CSF concentrations were 2,815.42 ng/ml in the shGM2-CAR19 T group and 284.22 ng/ml in the shGM4-CAR19 T group (Fig. 1b). The average KD ratio in shGM2-CAR19 T cells and shGM4-CAR19 T cells reached 89% and 98.9%, respectively. Co-expressing GM-CSFKD shRNA in CAR-T cells proved to be an efficient method to reduce GM-CSF secretion.

To evaluate the killing efficacy of GM-CSFKD CAR-T cells, we used Nalm6 cells marked with luciferase-zsGreen (Nalm6Luc+ cells) as target. In luciferase-based killing assays, CAR-T cells or mock T cells were added at effector to target (E: T) ratios of 0.25:1, 0.5:1, and 1:1 for 12 h. The killing efficiency was significantly higher in all CAR-T groups than in mock T cells (Fig. 1c). CAR-T cells with different levels of GM-CSF secretion exhibited comparable killing capacity. We indicated that the cytotoxicity of CAR19 T cells was not affected by the reduction of GM-CSF.

To mimic the complicated interplay of immunocytes and inflammatory factors, we used a transwell co-culture system containing mock T/regular/GM-CSFKD CAR-T cells, Nalm6 cells, and monocytes (Fig. 1d). The concentration of GM-CSF in the supernatant of the mock T group was significantly lower than that of CAR19 T cells, regardless of monocyte cocultivation (Fig. 1e). CAR-T cells were proved to be the main source of elevated GM-CSF in CAR-T cell therapy. IL-6 was evidently elevated in the cocultivation of CAR19 T cells, Nalm6 cells, and monocytes (Fig. 1e). GM-CSF KD in CAR19 T cells could alleviate IL-6 approximately to basic secretion (Fig. 1e). We then performed multi-analyte flow assay to assess the profile of inflammatory factors. By comparing secreted cytokines in activated CAR-T cells without participations of CD14+ cells, we noticed that IFN-γ, TNF-α, IL-17 A, MCP-1 and IL-8 were reduced simultaneously with the knockdown of GM-CSF (Fig. 1f). Although a slight raise of IL-6 secretion from GM-CSFKD CAR19 T cells was observed, the majority of IL-6 was secreted upon myeloid cells mediation. The overall production of IL-6 was lower in GM-CSFKD CAR19 T cells. Besides, IFN-γ, TNF-α, IL-17 A, MCP-1, IL-8 and IL-10 were firstly produced by activated CAR-T cells, and elevated when cocultured with monocytes (Fig. 1f). We confirmed that the inflammatory cytokine cascade was initiated at the encounter of CAR-T cells and target cells, while the outburst was mediated by monocytes and macrophages. The supernatant of GM-CSFKD CAR19 T cells cocultured with Nalm6 cells and monocytes exhibited lower levels of IL-6, IL-8, IL-10 MCP-1, IFN-γ, TNF-α, and IL-17 A (Fig. 1f). GM-CSFKD CAR19 T cells have the potential to reduce the extent of CRS/CRES by alleviating the release of pro-inflammatory factors.

Fig. 1
figure 1

Characterization and effects of GM-CSFKD CAR-T cells. (a) Schematic presentation of the GM-CSFKD-shRNA-CAR19 construct used in this study. (b) CAR-T cells (or mock T cells) were cocultured with CD19+ Nalm6 cells for 16 h. Concentration of GM-CSF in the supernatant was analyzed by ELISA. (c) Cytotoxicity of conventional CAR19 T cells and GM-CSFKD CAR19 T cells were compared 12 h after cocultivation with Nalm6Luc+ cells. (d) Diagram of the space division of CAR-T, Nalm6 and CD14+ cells cocultured in transwell (pore size = 0.3 μm). (e) Concentration of GM-CSF (left) and IL-6 (right) in the supernatant collected after coculturing CAR-T/Nalm6/ CD14+ cells for 16 h assessed by ELISA. (f) Cytokine profiles of co-cultured supernatants were measured by multi-analyte flow assay. The concentrations of IFN-α2, IL-12p70, IL-18, IL-23, and IL-33 were below the minimum value and not shown in this figure. (Each experiment was repeated for at least three times, and the results were expressed as mean (SD). * p < 0.05, ** p < 0.01, *** p < 0.001, **** p < 0.0001, ns: not significant)

Full size image

Discussion

Antagonizing CRS biomarkers have been the main strategy in the management of immune-related syndromes. IL-6 receptor blockage with tocilizumab is widely employed in patients with CRS [4, 11]. Our group previously found that the blockage of TNF-α and IL-1β signaling could ameliorate endothelial activation in CAR-T cell therapy [12]. The IL-1 receptor antagonist anakinra managed to protect from CRS and neurotoxicity in vivo [6, 13]. Compared with blocking downstream cytokines, the prevention of immune-related adverse effects is also of great importance. Recent studies set monocytes–macrophages as key mediators and GM-CSF as messenger in CRS and neurotoxicity [6, 8]. Monocyte depletion contributed to prevention from CRS, but it would also impede leukemia clearance in xenograft models [6]. Inhibiting GM-CSF and hence weakening the circuit between CAR-T cells and monocytes have been inspiring. GM-CSF neutralization with lenzilumab and GM-CSF knockout in CAR-T cells did not hamper their anti-tumor effects [8, 10, 14]. Recent study revealed that GM-CSF knockout with CRISPR/Cas9 could ameliorate early activation, contribute to proliferation, and exhibit anti-tumor functions in CAR-T cells [15]. Here, we demonstrated that GM-CSFKD CAR-T cells possess the potential to protect from immune-related adverse effects by reducing the secretion of multiple inflammatory cytokines including not only GM-CSF, but also IL-6, IFN-γ, TNF-α, IL-17 A, MCP-1, IL-8 and IL-10. Among them, moderate decrease of IFN-γ and TNF-α doesn’t affect the overall cytotoxicity of GM-CSFKD CAR-T cells. IFN-γ and TNF-α were found associated with endothelial damage caused by CAR-T cells [16] and held responsible for a lethal cytokine shock in a COVID-19 mouse model [17]. Neuralization of TNF-α with adalimumab could ameliorate endothelial activation without hampering killing capacity in CAR-T cell therapy [12]. GM-CSFKD CAR-T cells possess great potential in developing CRS free CAR-T cells. At the same time, the efficacy and toxicities of the CAR-T cells in vivo are susceptible to the interplay with the innate immune system. We still need to perform further experiments on suitable animal models for verifying the advantages of GM-CSFKD CAR-T cells in the future studies.

In conclusion, we introduced a new method to reduce GM-CSF secretion from CAR-T cells by inserting the GM-CSF shRNA-expression cassette in the CAR vector. This approach can be applied on any newly-developed CAR construct. The manufacture of GM-CSFKD CAR-T cells does not require repeated transfection, and the knockdown efficiency is more predictable and controllable. Hence, the method is practical and feasible for clinical application.