Phoenix-Ampho packaging cell line (RRID: CVCL_H716) was obtained from ATCC (Manassas, VA, USA). The ACC-MESO-4 malignant mesothelioma cell line (RRID: CVCL_5114) was obtained from the RIKEN Cell Bank (Ibaraki, Japan). The generation of the JurkatΔαβCD8a cell line, a subline of Jurkat (RRID: CVCL_0065) in which the TCRA and TCRB gene loci have been deleted and which carries a CD8A transgene, has been described elsewhere (8). JurkatΔαβCD8a cells were transfected with NFAT-RE (pGL4.30) or NFκB-RE (pGL4.32) luciferase reporter plasmid (Promega, Madison, WI, USA) using a NEPA21 electroporator (NEPAGENE, Chiba, Japan) and selected using hygromycin (Thermo Fisher Scientific, Waltham, MA, USA). Stable transfectants, designated JurkatΔαβCD8a-NFAT-RE-luc and JurkatΔαβCD8a-NFκB-RE-luc, respectively, were confirmed by reporter responsiveness to stimulation with 16 nM PMA (Sigma-Aldrich, St. Louis, MO, USA) and 14 μM Ionomycin (Adipogen Life Sciences, Fuellinsdorf, Switzerland) and used for CAR introduction. Male NSG (NOD. Cg-Prkdc^scid Il2rg^tm1Wjl /SzJ) mice were purchased from Jackson Laboratories Japan (Kanagawa, Japan) and housed in the animal facility of Kanagawa Cancer Center Research Institute. Mice were purchased at 5 weeks of age and used for experiments before 8 weeks of age. FITC anti-CD3e (BioLegend Cat# 300405, RRID: AB_314059), PE/Cy7 anti-CD4 (BioLegend Cat# 317413, RRID: AB_571958), PerCP-Cy5.5 anti-CD4 (BioLegend Cat# 317427, RRID: AB_1186124), APC anti-CD8a (BioLegend Cat# 301014, RRID: AB_314132), PE-Cy7 anti-CD8a (BioLegend Cat# 301012, RRID: AB_314130) Pacific Blue anti-PD-1 (BioLegend Cat# 329915, RRID: AB_1877194), APC-Cy7 anti-Tim-3 (BioLegend Cat# 345025, RRID: AB_2565716), APC anti-LAG-3 (BioLegend Cat# 369211, RRID: AB_2728372), AlexaFluor 488 anti-CCR7 (BioLegend Cat# 353206, RRID: AB_10916389), APC CD62L (BioLegend Cat# 385105, RRID: AB_3097359), APC-Cy7 anti-CD45RA (BioLegend Cat# 304128, RRID: AB_10708880) and Pacific Blue anti-CD25 (BioLegend Cat# 356129, RRID: AB_2563589) monoclonal antibodies (mAbs) were obtained from BioLegend (San Diego, CA, USA). AlexaFluor594- or PE-labeled anti-G4S linker (Cell Signaling Technology; Cat# 38907, RRID: AB_3626304, Cat# 39614, RRID: AB_3626305) and anti-Whitlow/218 linker (Cell Signaling Technology; Cat# 62405, RRID: AB_3626306, Cat# 61465, RRID: AB_3626309) mAbs were purchased from Cell Signaling Technology (Danvers, MA, USA).

Retroviral vectors containing CAR constructs for SKM-28z and SKM-BBz CARs, which have the (G4S)4 linker in scFv, have been described previously (4). The linker part of these SKM-(G4S)4-CAR vectors was modified to the Whitlow 218 linker sequence (GSTSGSGKPGSGEGSTKG) using Gibson assembly (New England Biolabs, Ipswich, MA). CD19-CARs were generated by replacing the SKM9-2 scFv sequence with the scFv sequence derived from anti-CD19 mAb FMC63 (7) linked by the Whitlow/218 linker sequence.

Human peripheral blood mononuclear cells (PBMCs) were collected by density gradient centrifugation (Lymphoprep; Alere Technologies AS, Oslo, Norway) from the peripheral blood of three independent healthy volunteers who provided written informed consent. The study was conducted in accordance with the tenets of the Declaration of Helsinki, and the protocol was approved by the Institutional Review Board of Kanagawa Cancer Center (study number: 28-KEN-29). PBMCs (1×10⁶ cells/mL) were stimulated using plate coated with 5ug/ml anti-CD3 mAb (clone UCHT-1; BD Biosciences, San Jose, CA, USA) and 10 ng/mL human IL-2 (Peprotech, Rocky Hill, NJ, USA) for 2 days in AIM-V medium (Thermo Fisher Scientific) supplemented with 10% human AB serum (MP Biomedicals, Irvine, CA, USA). A retroviral vector containing CAR cDNA was transfected into Phoenix-Ampho packaging cells using X-fect reagent (Takara Bio, Shiga, Japan), according to the manufacturer’s instructions, to obtain the virus-containing supernatant. The virus-containing supernatant was added to an untreated 24-well plate coated with retronectin (Takara Bio) and centrifuged at 2000 ×g for 2 hours at 32°C to facilitate the binding of virus particles to retronectin. Activated PBMCs (2.5×10⁵ cells/well) were then added to the virus-coated plates for infection. CAR-T cells expressing SKM-28z CAR with the (G4S)₄ linker, SKM-BBz CAR with the (G4S)₄ linker, SKM-28z CAR with the Whitlow/218 linker, and SKM-BBz CAR with the Whilow/218 linker were designated SKM-G4S-28z, SKM-G4S-BBz, SKM-W218-28z, and SKM-W218-BBz, respectively ( Figure 1A ). Twenty-four hours later, secondary infection was performed, and cells were cultured with medium containing 10ng/ml of human IL-2 and passaged every 3 to 4 days at 5x10⁵ cells/ml.

CAR antigen specificity was assessed by incubation with the SKM-Epmin3 peptide, a His-tagged epitope of HEG1 recognized by the SKM9-2 mAb (9), followed by staining with a PE-labeled anti-His tag Ab (BioLegend Cat# 362603, RRID: AB_2563634). CAR expression on PBMCs and the surface phenotypes of CAR+ cells were determined by multicolor flow cytometry using PE anti-G4S or anti-Whitlow/218 linker mAb, PerCP-Cy5.5 anti-CD4 mAb, FITC or PE-Cy7 anti-CD8a mAb, APC-Cy7 anti-Tim-3 mAb, APC anti-LAG3 mAb, Pacific Blue anti-PD-1 mAb, AlexaFluor 488 anti-CCR7 mAb, APC anti-CD62L mAb, APC-Cy7 anti-CD45RA mAb and Pacific Blue anti-CD25 mAb. For the in vitro cytotoxicity assay, CAR-positive cells were sorted using a FACS Aria II cell sorter (BD Biosciences) after staining with PE anti-linker mAbs. For CAR expression in Jurkat reporter cell lines, the reporter cells (1 × 10⁵ cells) were single-infected with a retronectin-bound CAR-encoding retrovirus in a 24-well plate.

To image CAR distribution on the cell surface, CAR-T cells were stained with Alexa Fluor 594-labeled rabbit mAbs against the corresponding linkers in scFv (i.e. anti-G4S for G4S-CARs and anti-Whitlow/218 linker for W218-CARs). Cells were fixed with 4% formaldehyde and embedded in VECTASHIELD Antifade mounting medium (Vector Laboratories, Newark, CA, USA) in a glass-bottom dish (MatTek Life Sciences, Ashland, MA, USA). Confocal images were captured using a Zeiss LSM 710 (Carl Zeiss AG, Oberkochen, Germany) equipped with a 63x/1.4 NA PlanApo objective in 1.0 μm sections from cell top to bottom. For quantitative analysis, regions of interest (ROI) were defined on the cell surface area of the equatorial section image, and mean fluorescence intensities (MFI) and standard deviations (SDs) were obtained using Fiji software (RRID: SCR_002285) (10). To quantify the relative variability of CAR expression, the coefficient of variation (CV=SD/MFI) was calculated for 12 cells of each CAR (11).

ACC-MESO-4 cells (5 × 10³ cells) were mixed with sorted CAR-T cells (5 × 10³ to 2.5 × 10⁴ cells; at E:T ratios from 1:1 to 5:1) in a 96-well V-bottom culture plate and cultured for 18 h. The amount of LDH in the culture supernatant was measured using the cytotoxic LDH assay kit-WST (Dojindo, Kumamoto, Japan). Specific LDH release was calculated by comparison with the maximum LDH release after lysis with 0.5% Triton X-100. The in vitro cytotoxicity kinetics were measured using an xCELLigence real-time cell analyzer (Agilent, Santa Clara, CA, USA). ACC-MESO-4 cells (2 × 10⁴ cells) were seeded onto impedance plates and cultured overnight. After 22 h, the CAR-T cells (2 × 10⁴ cells/well) were added to each well. The impedance of each culture well was measured continuously for 28 h. The cytokine levels in the culture supernatant of the LDH release assay were evaluated using a bead-based multiplex assay. In this assay, cytokines were measured in 50 μL aliquots of 4-fold diluted supernatant of E:T=5:1 cultures using the Bio-Plex 200 system (Bio-Rad Laboratories, Hercules, CA). The Bio-Plex Pro Human Cytokine Th1/Th2 Assay (Bio-Rad Laboratories) was used to measure the following 9 cytokines: IL-2, IL-4, IL-5, IL-10, IL-12 (p70), IL-13, GM-CSF, IFNγ, and TNF-α. IL-2 measurement was not evaluated because the culture media contained 10 ng/mL recombinant human IL-2, making it difficult to measure the IL-2 secreted by CAR-T cells. In some experiments, remaining cells after killing assay at E:T = 5:1 condition were recovered and stained for FACS analysis.

To quantify NFAT or NFκB reporter activity, CAR-expressing JurkatΔαβCD8a-NFAT-RE-luc cells or JurkatΔαβCD8a-NFκB-RE-luc cells (2×10⁵ cells) were seeded in 96-well white plates. BrightGlo luciferase assay substrate (Promega) was added to the cultures, and luminescence was measured using an Arvo plate reader (PerkinElmer, Waltham, MA, USA). For antigen stimulation, ACC-MESO-4 cells (1×10⁴ cells) were seeded in 96-well white plates one day before the assay. CAR-T cells (1×10⁴ cells) were added and co-cultured for 5 h before luciferase activity measurement.

CAR-infected PBMCs were stained with an anti-mouse IgG polyclonal antibody, and CAR-positive cells were sorted on day 9 after infection using a FACSAria II cell sorter. After two days of culture, total RNA was purified from the sorted cells using an RNeasy Mini RNA Extraction Kit (Qiagen, Germantown, MD, USA). Three biological replicates were used for this preparation. The cDNA library was generated using the SMART-Seq v4 Ultra Low Input RNA kit (TaKaRa Bio), sequenced using a NovaSeq 6000 sequencer (Illumina, San Diego, CA), and mapped to GRCh38 [GENCODE v42 (RRID: SCR_014966)] using KI Biobank - STAR (2.7.10b) (RRID: SCR_005923). Principal component analysis (PCA) was performed using the DESeq2 software (RRID: SCR_015687) (12). Based on the transcriptome data, differentially expressed genes (DEGs) were identified using DESeq2 with the criteria of q<0.05, p<0.05, and |log2FC|>1.

In an ectopic mouse model, 25 male NSG mice were subcutaneously injected with ACC-MESO-4 cells (1x10⁶ cells), 7 days later (day 0), mice were randomly divided into 4 groups and each group received intravenous injection of G4S-28z or W218-28z CAR-T cells (1x10⁶ cells) or left untreated (control) (n=6~7/group). Tumor sizes were measured using calipers, and tumor volumes were calculated using the following formula: (long axis) x (short axis)² x 1/2. Exact Wilcoxon-Mann-Whitney test was used to statistically evaluate tumor volumes at the end of observation (day 70). P values < 0.05 were considered statistically significant. During the experiments, no mice reached the prespecified exclusion criteria of tumor volume greater than 1500mm³ or 10% weight loss.

PBMC transduced with SKM-G4S-28z or SKM-G4S-BBz CAR were positively stained with an antibody against the G4S linker. Conversely, cells transduced with SKM-W218-28z and SKM-W218-BBz CAR were positively stained with an antibody against the Whitlow/218 linker, indicating the successful construction of the retroviral vectors ( Figure 1B ). We observed that the staining intensities of W218 CARs were weaker than those of G4S CARs. This may be because of the different affinities of each mAb for the corresponding epitopes. To directly compare the expression levels of each CAR type and confirm the antigen specificities of the modified CARs, CAR-T cells were incubated with a (His)6-tagged epitope peptide for SKM9-2 mAb. As shown in Figure 1C , CAR-T cells with different linkers showed no significant differences in epitope peptide staining, indicating that the modified CARs retained their original antigen specificity and had comparable expression levels. During CAR-T cell production, all CAR-T cells proliferated equally to CAR (–) cells, while W218 CAR-T cells tended to show lower viability compared to G4S CAR-T cells. ( Supplementary Figure S1 ).

In our previous study, SKM-G4S-CARs, independent of intracellular co-stimulatory domains, showed a tendency to aggregate on the cell surface compared to CD19-CAR with the Whitlow/218 linker (4). To determine whether the introduction of the Whitlow/218 linker into SKM-CARs reduced their tendency to aggregate on the cell surface, the cell surface distribution of each CAR was observed using confocal microscopy ( Figure 2A ). Modification of the linker sequence did not affect the CV values of cell surface CAR staining. However, all SKM-CARs were significantly more aggregated than CD19-CAR with the Whitlow/218 linker ( Figure 2B ).

In our previous study, SKM-G4S-28z CAR-T cells expressed higher levels of T cell exhaustion markers, Tim-3, LAG3, and PD-1, than CD19-CAR-T cells due to spontaneous signaling independent of antigen ligation (tonic signal). SKM-G4S-BBz CAR-T cells also generated a tonic signal, but with a different quality caused by the 4-1BB co-stimulatory domain and did not upregulate exhaustion markers. To investigate the influence of linker sequences on tonic signal-induced surface phenotypes, the expression of Tim-3, LAG3, and PD-1 on SKM-W218 CAR-T cells was compared with that on SKM-G4S CAR-T cells. Consistent with the above observation that there was no difference in the cell surface aggregation of G4S and W218 CARs, SKM-W218-28z CAR-T cells showed similar upregulation of these T cell exhaustion markers as SKM-G4S-28z CAR-T cells ( Figure 3 ). Expression of exhaustion markers were almost the same between CD8⁺ and CD4⁺ CAR-T cells except for LAG3, which is expressed much higher in the CD8⁺ CAR-T cells ( Supplementary Figure S2 ). The expression of exhaustion markers were up-regulated at day 11 compared to day 4 from retrovirus infection, suggesting these are induced by persistent tonic signaling from SKM-CARs ( Figure S3 ). To further address changes in the phenotypes of CAR-T cells with different CAR constructs, T cells differentiation markers CCR7, CD62L and CD45RA, as well as activation marker CD25 were stained. All CAR-T cells showed decrease in the CD62L and CD45RA compared to CAR (–) control, suggesting differentiation from naïve stage. Interestingly the two SKM-BBz CAR-T cells showed remarkable upregulation of CCR7 compared to the two SKM-28z CAR-T cells, suggesting differentiational arrest before effector memory/terminal exhausted stage ( Supplementary Figure S4 ). While these phenotypes were consistent with our previous report, again, there is no noticeable difference between G4S CAR-T cells and W218 CAR-T cells. Expression of CD25 were upregulated in all SKM-CAR-T cells compared to CAR (–) control, suggesting presence of the tonic signal.

To investigate the downstream signaling pathways induced by tonic signals in each CAR construct, CAR molecules were expressed on TCR-deficient Jurkat cells transduced with the NFAT reporter gene, called JurkatΔαβCD8a-NFAT-RE-luc cells. Consistent with our previous results, strong NFAT activation was observed in the JurkatΔαβCD8a-NFAT-RE-luc cells expressing SKM-28z CARs, but almost no NFAT activity was detected in those expressing SKM-BBz CARs ( Figure 4A ). Notably, cells expressing either SKM-G4S-28z or SKM-W218-28z showed NFAT activation at the same level, suggesting that there was no change in the magnitude of tonic signals in SKM-28z CARs with different linkers ( Figure 4A ). Similarly, the level of NFAT reporter activity after 5 h of antigen stimulation was not different between the SKM-G4S-28z and SKM-W218-28z CARs ( Figure 4B ).

To investigate another downstream signaling pathway, NFκB, in each CAR construct, CARs were expressed on TCR-deficient Jurkat cells with the NFκB reporter gene, i.e., JurkatΔαβCD8a-NFκB-RE-luc cells. NFκB reporter activity was significantly upregulated in the SKM-CAR-expressing cells without stimulation ( Figure 4C ) and was increased after antigen stimulation: SKM-BBz CARs showed stronger activities than SKM-28z CARs after antigen stimulation ( Figure 4D ). Notably, the levels of NFκB signals between SKM-BBz CARs with the different linkers as well as between SKM-28z CARs with the different linkers were not significantly different ( Figure 4C, D ).

To compare the in vitro killing activity of CAR-T cells expressing each CAR construct, CAR-T cells were cocultured with ACC-MESO4 cells expressing the SKM9-2 antigen. Consistent with our previous results obtained using the G4S linker (4), SKM-G4S-28z CAR-T cells tended to have a lower ability to kill ACC-MESO4 cells than SKM-G4S-BBz CAR-T cells. In contrast, in the case of the Whitlow/218 linker, both SKM-W218-28z and SKM-W218-BBz CAR-T cells showed good killing abilities, comparable to those of SKM-G4S-BBz CAR-T cells ( Figure 5A ). Because the LDH assay can only detect endpoint differences, the killing kinetics were measured using a real-time cell analyzer that detects plate impedance. Consistent with our previous report, SKM-G4S-28z CAR-T cells (red) exhibited slower kinetics than SKM-G4S-BBz CAR-T cells (orange). Notably, SKM-W218-28z CAR-T cells (green) exhibited similar kinetics to SKM-G4S-BBz (orange) and SKM-W218-BBz CAR-T cells (purple). This result demonstrated that replacing the G4S linker with the W218 linker improved the killing activity of SKM-28z CAR-T cells, whereas there were no differences in CAR aggregation, cell surface phenotypes, or tonic signal intensity between them. Following the differences in in vitro killing ability despite the lack of differences in surface antigens, the phenotype of CAR-T cells after killing target cells was examined. As expected, the surface expression of exhaustion markers (PD-1, LAG3 and Tim-3) was upregulated after target cell killing. However, there was no difference between G4S CAR T cells and W218 CAR T cells in the degree of exhaustion marker upregulation ( Supplementary Figure S5 ).

To further investigate the functional differences between SKM-CAR-T cells with the G4S and Whitlow/218 linkers, cytokine secretion from activated CAR-T cells was measured ( Figure 4C ). Culture supernatants from CAR-T cells co-cultured with ACC-MESO4 cells were applied to a bead-based multiplex assay to quantify T cell cytokines: IL-4, IL-5, IL-10, IL-12 (p70), IL-13, GM-CSF, IFNγ, and TNF-α. There were considerable differences in the secretion of each cytokine between SKM-G4S-28z and SKM-G4S-BBz CAR-T cells; SKM-G4S-28z CAR-T cells secreted more IL-4 and IL-10 than SKM-G4S-BBz CAR-T cells, while SKM-G4S-BBz CAR-T cells secreted more IL-5, IL-12, IL-13, GM-CSF, IFNγ, and TNF-α. A comparison between SKM-W218-28z and SKM-W218-BBz CAR-T cells revealed the same trend. Notably, when comparing SKM-G4S-28z and SKM-W218-28z CAR-T cells, SKM-W218-28z CAR-T cells secreted significantly higher levels of all cytokines tested than SKM-G4S-28z CAR-T cells did.

To investigate the molecular changes caused by switching the linker sequence from G4S to Whitlow/218, the gene expression of each CAR-T cell in the absence of antigen stimulation was compared using RNA sequencing. CAR-T cells expressing SKM-G4S-28z CAR or SKM-G4S-BBz CAR and T cells without CAR expression exhibited different gene expression patterns in the PCA plot ( Figure 6 ). Interestingly, SKM-W218-28z and SKM-W218-BBz CAR-T cells showed gene expression patterns similar to those of their respective counterparts with the G4S linker. When DEGs were identified using the criteria |log2FC>1, p<0.05 and q < 0.05, 16 DEGs were observed between SKM-W218-28z and SKM-G4S-28z CAR-T cells ( Table 1 ), whereas there were only four DEGs between SKM-W218-BBz and SKM-G4S-BBz CAR-T cells ( Table 2 ). These results suggest that no significant differences were observed in most genes owing to the substitution of linker sequences at steady state without stimulation.

We investigated the ability of SKM-G4S-28z and SKM-W218-28z CAR-T to control tumor growth in vivo using a human tumor cell line xenograft model. The ectopic xenograft model was established by subcutaneous injection of the malignant mesothelioma cell line ACC-MESO-4 into immunodeficient NSG mice. Unexpectedly, despite the improved in vitro killing activity ( Figure 5A, B ), intravenous administration of SKM-W218-28z CAR-T did not delay tumor growth compared to SKM-G4S-28z CAR-T treated or untreated mice, as shown in Figure 7 . At the endpoint of the mouse experiment, we isolated cells from tumors or spleens and examined CAR expression in the isolated cells by flow cytometry using mAbs against the G4S or W218 linker. However, no CAR-T cells were detected in the tumors or spleens of SKM-W218-28z or SKM-G4S-28z CAR-T treated mice (data not shown).