293T cells, Jurkat T cells and K562 cells were obtained from ATCC. Thyroid cancer cell line 8505c cells was obtained from DSMZ (German Collection of Microorganisms and Cell Cultures). 293T cells and 8505c cells were cultured in high glucose DMEM medium with 10% fetal bovine serum (FBS) and 1% penicillin/streptomycin (P/S). Jurkat T cells (clone E6-1) and K562 cells were cultured in IMDM medium with 10% FBS and 1% P/S. All cells were maintained in 5% CO₂ at 37°C.

All plasmids were constructed using ClonExpress II One Step Cloning Kit (Vazyme, Nanjing, China) following manufacturer’s instruction. Retrovirus plasmids encoding TSHR-specific CARs were constructed by inserting the CAR constructs into SFG γ-retroviral vector. For detection of CAR expression, a sequence encoding green fluorescence protein (GFP) was linked in the C-terminus of CAR by a P2A peptide. Plasmids used for cell line transduction were constructed in lentiviral vectors. To prepare the plasmid co-expressing TSHR and luciferase, the sequences encoding TSHR and luciferase reporter were linked by a P2A peptide. The ZAP70 and ERK biosensors were reported by previous studies. The ZAP70 biosensor used the peptide (SREYACISGEL) as ZAP70 substrate (Liu et al., 2021). The ERK biosensor used the peptide (PDVPRTPVDKAKLSFQFPF) as ERK substrate (Komatsu et al., 2011). For co-expression of ZAP70 biosensor or ERK biosensor and CARs in Jurkat T cells, the CAR construct was introduced in the C-terminus of biosensor sequence using a P2A peptide.

Human blood was obtained from healthy donors with written approval. PBMC was isolated using human lymphocyte isolation kit (Dakewe, Shenzhen, China). T cells were further purified using Pan T cell isolation kit (Miltenyi Biotec, Bergisch Gladbach, Germany) and cultured with X-VIVO 15 medium (Lonza) supplemented with 10% FBS, 1% P/S, 5 ng/ml IL-7 (Novoprotein, Shanghai, China) and 5 ng/ml IL-15 (Novoprotein). Immediately after T cell isolation, they were stimulated with CD3/CD28 T cell Activator Dynabeads (Invitrogen, Carlsbad, United States) at a ratio of 1:1. T cells transduction was performed after 48 h. Retrovirus was produced from 293T cell lines. T cells were transduced with retrovirus supernatants in retronectin (Takara, Otsu, Japan)-coated plates. To enhance transduction efficiency, the plates were centrifuged with 3000 rpm for 90 min.

CAR-T cells were co-cultured with 5×10⁴ target cells at different effector/target (E/T) ratio in target cell medium in black-walled 96 well plates. After 18 h, the cells were treated with D-luciferin (GoldBio, St. Louis, United States). Emitted light was detected by a luminescence plate reader (Thermo Varioskan Flash). Lysis percentage was calculated as [1-(luminescence intensity of each sample)/(luminescence intensity of target cells alone)] ×100.

For cytokines and exhaustion markers analysis, CAR-T cells were stimulated with irradiated 8505c-TSHR cells for 24 h. The BD Cytometric Bead Array (CBA) kit were used to quantify the secretion level of IL-2, TNFα and IFN-γ of the supernatant of CAR-T cells following manufacturer’s instruction. For intracellular cytokines staining, the Golgi plug protein transport inhibitor Brefeldin A (eBioscience, San Diego, United States) was added into the cultured medium 4 h before detection. T cells were then fixed and permeabilized with Fixation and Permeabilization kit (BD Biosciences, San Jose, United States) according to manufacturer’s instruction. The following antibodies were used: FITC mouse anti-human TSHR (Santa Cruz Biotechnology, Santa Cruz, United States), Human IL-2 Flex Set (BD Biosciences), Human TNF Flex Set (BD Biosciences), Human IFN-γ Flex Set (BD Biosciences), BV421 rat anti-human IL-2 (BD Biosciences), PE-Cy7 mouse anti-human TNFα (BD Biosciences), PE-Cy7 anti-human CD279 (PD-1) (eBioscience), eFluor 450 anti-human CD223 (LAG-3) (eBioscience). Flow cytometry was performed on a CytoFLEX LX cytometer (Beckman Coulter, Brea, United States). A moflo Astrios EQ cell sorter (Beckman Coulter) was used for cell sorting. Data were analyzed with FlowJo software (FlowJo LLC, Ashland, United States).

Six- to 12-week-old NOD/SCID/IL-2Rγ^null (NSG) mice were obtained from Jihui Shanghai and housed in the Animal Core Facility at Westlake University. All procedures followed the Institutional Animal Care and Use Committee (IACUC) guideline. For the in vivo anti-tumor assay, mice were intravenously injected with 1×10⁶ 8505c-TSHR cells followed by 0.5×10⁶ CAR-T cells the next day. To monitor tumor burden, mice were treated with 200 μL of 15 mg/ml D-luciferin by intraperitoneal injection. Bioluminescence imaging was performed using the Optima small animal imaging system (Biospace Lab, Paris, France).

Time-lapse imaging was performed with a Nikon Eclipse Ti inverted microscope. A Tokai Hit ST Series Stage Top Incubator was used to maintain a 5% CO₂ at 37°C for cells during imaging. The W-VIEW GEMINI imaging splitting optics (Hamamatsu Photonics, Hamamatsu, Japan) with a 438/29 nm excitation filter, a 474/40 nm emission filter, a 535/25 nm emission filter, an iXon Ultra 897 EMCCD camera (Andor Technology, Belfast, United Kingdom) was used to acquire the ECFP and FRET fluorescent signals simultaneously. CAR-Jurkat T cells co-expressing ZAP70 or ERK biosensor were generated by lentivirus transduction. To monitor ZAP70 or ERK signal dynamics upon antigen stimulation, CAR-Jurkat T cells were dropped on the glass-bottom dishes coated with K562-TSHR cells. From that time on, images were acquired at an interval of 30 s for 40 min. The imaging data were analyzed by Fluocell software. The ECFP/FRET ratio of ZAP70 biosensor or the FRET/ECFP ratio of ERK biosensor for each cell was normalized before comparison.

Statistical analysis was performed using GraphPad Prism 8 software. An unpaired, two-tailed student’s t-test were used to determine the statistical differences between two groups. A Kaplan-Meier curve and the log-rank test were used to compare the survival differences between the groups in animal experiments. A p value < 0.05 was considered to be statically significant.

Based on a second-generation CAR construct we investigated the feasibility of CAR-T cells in treating TSHR-positive thyroid cancers. In our design, the ectodomain of CARs was a single chain variable fragment (scFv) that recognize TSHR, and the endodomain contained a CD28 or 4-1BB-derived costimulatory domains and CD3ζ-derived activation domain. To evaluate the expression level of CARs, a green fluorescence protein (GFP) was introduced downstream of CAR through a P2A self-cleavage peptide (Figure 1A). Although not strictly correlated, the expression of GFP could be used to estimate CAR expression level. To find the optimized scFvs, we constructed and compared scFvs from two anti-TSHR monoclonal antibodies (K1-70 and KSAb) that have been previously reported. As the arrangement of variable heavy chain (VH) and variable light chain (VL) fragment may largely affect the antigen recognition property, scFvs with different VH-VL arrangements were constructed for each antibody. Thus, four CD28-based CAR constructs with different scFvs (referred to K70H-28z, K70L-28z, KSAbH-28z and KSAbL-28z) were generated (Figure 1A).

Next, we produced CD28ζ CAR-T cells with different scFvs from primary T cells and compared their cytotoxic effect. Since most thyroid cell lines have lost their TSHR expression (Li et al., 2021), we established a TSHR-positive target cell line by transducing the full-length TSHR gene with a luciferase reporter into the thyroid cancer cell line 8505c (8505c-TSHR) (Figure 1B). Using 8505c-TSHR as the target cell, in vitro killing assay revealed the superior cytotoxic effect of K70H-28z CAR-T cells than its counterparts (Figure 1C). Therefore, the K1-70 derived VH-VL scFv was selected for subsequent studies.

Nowadays, the most widely used CARs in clinic contain either CD28 or 4-1BB as co-stimulatory domains. Both of them have achieved impressive results in clinical studies. However, CAR-T cells bearing CD28 or 4-1BB intracellular domains could have different cytotoxic functions and in vivo persistence (Feucht and Sadelain, 2020). To find out the optimal co-stimulatory domain for thyroid cancer treatment, we compared the in vitro anti-tumor functions of CD28 and 4-1BB-based TSHR CAR-T cells. The CAR structures (K70H-28z & K70H-BBz) have the same scFv but different hinge and transmembrane domains (Figure 1A). Produced by retrovirus transduction, both CAR-T cells showed similar transduction efficiency (Figure 2A). Using either K562-TSHR (Figure 1B) or 8505c-TSHR as target cells, their cytotoxicity was compared by in vitro killing assay. We found comparable cytotoxic effect between K70H-28z and K70H-BBz CAR-T cells (Figure 2B). We then analyzed the cytokines release of CAR-T cells by FACS. Co-culture with 8505c-TSHR cells promoted K70H-28z and K70H-BBz CAR-T cells but not the untransduced T cells to secrete robust IL-2, TNFα and IFNγ (Figure 2C). The K70H-28z CAR-T cells secreted comparable levels of IL-2 and IFNγ, but higher levels of TNFα than K70H-BBz CAR-T cells (Figure 2C). Intracellular cytokines staining further confirmed that antigen stimulation notably induced GFP⁺ CAR-T cells to secrete IL-2 and TNFα but not GFP⁻ T cells (Figure 2D).

After a period of ex vivo culture, which is often required for CAR-T cell expansion before infusion, we observed an increased GFP⁺ ratio of the total populations, representing an enrichment of the CAR-T cells. To evaluate the exhaustion levels, we detected PD-1 and LAG-3 expression of CAR-T cells after antigen stimulation (Figure 2E). Compared to CAR-T cells at day 5, CAR-T cells at day 13 showed increased expression of both exhaustion markers (Figure 2E). Moreover, 60.1% of the K70H-28z CAR-T cells showed double positive of PD-1 and LAG-3, much higher than that of the K70H-BBz group (Figure 2E).

Although previous studies have characterized different signal activities triggered by CD28 or 4-1BB-based CARs (Philipson et al., 2020), their signaling dynamics in the early stage of T cell activation are still elusive. To figure out how co-stimulatory molecules affect CAR-induced signal transductions, we first focused on the membrane-proximal signaling events that were activated at the earliest time after antigen stimulation. ZAP70 has been considered as one of the most important kinases to transduce CAR signals into downstream signaling cascades (Wang et al., 2010; Gudipati et al., 2020). We have applied FRET-based ZAP70 biosensor to monitor the activity changes of ZAP70 during TSHR CAR-T cells activation. The ZAP70 biosensor was designed to include a FRET pair of ECFP and YPet fluorescence proteins, a ZAP70 kinase-specific substrate domain, and a ligand domain. Phosphorylation of the substrate by ZAP70 would cause conformational changes of the biosensor, resulting in reduction of FRET efficiency between ECFP and YPet (Figure 3A) (Liu et al., 2021). Therefore, the relative activity of ZAP70 could be evaluated by detecting the fluorescence ratio of ECFP and FRET (YPet). To assess the ZAP70 activity upon antigen stimulation, the Jurkat T cell line co-expressing ZAP70 biosensor and CARs (CAR-Jurkat T) were generated by lentivirus transduction. Both K70H-28z and K70H-BBz CAR-Jurkat T cells showed similar expression of biosensor (CAR) (Figure 3B). By FRET imaging, we observed remarkable changes of the ZAP70 activity in CAR-Jurkat T cells upon encountering with the K562-TSHR target cells (Figure 3C). The ZAP70 activity increased rapidly within 10 min after antigen stimulation. Thereafter, its signal continued to increase with slower speed (Figure 3D). We further quantified the accumulated signal intensity (area under curve) and maximal signal intensity (maximal ECFP/FRET ratio changes) of K70H-28z and K70H-BBz CARs. Neither the accumulated signal intensity nor the maximal signal intensity of ZAP70 between K70H-28z and K70H-BBz CARs showed significant difference (Figures 3E,F).

To assess whether signals downstream of ZAP70 were affected by different co-stimulatory molecules, we investigated the signal dynamics of ERK during CAR-T cells activation by FRET-based biosensor. The ERK biosensor used a flexible linker to connect the substrate, ligand domain and FRET pairs. It also included a nucleus exporting signal (NES) to increase detection sensitivity of cytosolic signal. The biosensor would increase its FRET efficiency when phosphorylated by ERK (Figure 4A) (Komatsu et al., 2011). The CAR-Jurkat T cells expressing ERK biosensor and K70H-28z or K70H-BBz CAR were generated by lentivirus transduction and showed similar expression of biosensor (CAR) between groups (Figure 4B). By FRET imaging we found that the ERK activity, as visualized by FRET/ECFP fluorescence ratio of CAR-Jurkat T cells increased upon K562-TSHR cells stimulation (Figure 4C). Statistical quantification suggested that the activity of ERK reached to a peak level in about 15 min and then underwent slow decrease (Figure 4D). The accumulated signal intensity of ERK in K70H-28z CAR was slightly higher than that of K70H-BBz CAR but without significant difference (Figure 4E). However, the maximal signal intensity of ERK in K70H-28z CAR was significantly higher than that of K70H-BBz CAR (p < 0.05) (Figure 4F). Collectively, compared with K70H-BBz CAR, K70H-28z CAR triggered similar ZAP70 signaling but significantly enhanced ERK signaling.

Finally, we evaluated the anti-tumor functions of K70H-28z and K70H-BBz CAR-T cells by in vivo studies (Figure 5A). To mimic the distant metastasis of DTCs, of which lung and bone are the major sites, we established a xenograft animal model based on previous studies (Zhang et al., 2014). The 8505c-TSHR cells were intravenously injected into the immunodeficient NSG mice. We found the tumors cells accumulated in the lung at early stage. Bone and other tissue metastasis happened after 14 days and caused animal death in the control group (Figures 5B,C,E). By contrast, mice treated with either K70H-28z or K70H-BBz CAR-T cells successfully controlled tumor growth and survived without obvious side-effect (Figures 5B,C,E). Nevertheless, K70H-28z CAR-T cells were found to eradicate nearly all tumor cells within 7 days, while K70H-BBz CAR-T cells lagged behind (Figures 5C,D). This indicated faster anti-tumor kinetics of K70H-28z CAR-T cells than K70H-BBz CAR-T cells.