Human cell lines SEM, Jurkat, and K562 were purchased from DSMZ. Cells were cultured in RPMI 1640 containing 10% FBS, at 37°C with 5% CO₂. Peripheral blood mononuclear cells were obtained from healthy donors. CD3⁺ T cells were isolated using EasySep™ Human T Cell Enrichment Kit (STEMCELL Technologies, 19051) and then cultured in CTS™ AIM V™ SFM medium (Thermo Fisher Scientific, 0870112DK) containing 5% human serum AB (GEMINI, 100-512), 1% GlutaMAX (Thermo Fisher Scientific, 35050061), and 100 U/mL human IL-2 (PeproTech, 200-02). The CD3⁺ T cells were activated by Dynabeads™ Human T-Activator CD3/CD28 (Thermo Fisher Scientific, 11131D).

The surface CAR was detected by using APC or Brilliant Violet 421™ anti-FLAG Tag Antibody (Biolegend, 637308/637322). The total CAR was staining intracellular FLAG expression by APC or Brilliant Violet 421™ anti-FLAG Tag Antibody after using BD Cytofix/Cytoperm Buffer System (BD, 554714). The CD19 protein was detected by APC anti-human CD19 Antibody (BD, 555415). Data produced by the flow cytometer (BD LSRFortessa™ X-20) were analyzed using the FlowJo software.

Lentivirus plasmids pHR-FLAG-Anti-CD19 single-chain variable fragment (FMC63)-CD8 hinge-CD8α or CD1a transmembrane domain-CD28 and 4-1BB costimulatory domain-CD3ζ-copGFP (pHR-FMC63-CD8α/1a-CAR) was generated by inserting CAR sequence [generated according to the patent previously reported (20)] into the pHR plasmids (Addgene). The pCDH-CD19-mRuby2 plasmid was generated by inserting CD19 and mRuby2 sequences into the pCDH lentiviral vector (System Biosciences). The pCDH-Luciferase-mRuby2 plasmid was generated by inserting Luciferase and mRuby2 sequences into the pCDH lentiviral vector. For CAR-T cells, human CD3⁺ T cells activated for 72 h were collected and resuspended with concentrated lentiviruses (pHR-FMC63-CD8α/1a-CAR, MOI=15). 0.5 million cells were plated on a 24-well plate and spun at 1200 g for 1 h at 37°C. The virus supernatant was aspirated and replaced with fresh medium after 24 h infection. The CAR-T cells were checked 3 days after infection by flow cytometry. The endotoxin levels in CAR-T cells were within permissible ranges (Yeasen, 36723ES). And there was no mycoplasma contamination (Yeasen, 40601ES). For CAR-Jurkat cells, after infection, cells were sorted according to surface CAR levels to get different surface CAR expression cells. For SEM-mRuby2-Luciferase and CD19-K562 cells, after infection (pCDH-Luciferase-mRuby2 and pCDH-CD19-mRuby2, respectively), positive cells were sorted according to mRuby2.

The target cells (SEM or CD19-K562) were stained with CellTrace™ Violet/Yellow/Far Red (Thermo Fisher Scientific, C34571/C34573/C34572) as wanted. Then CAR-T/CAR-Jurkat cells were co-cultured with target cells at different E/T ratios for the indicated time, and the apoptosis of target cells was detected using the Annexin V-APC or Annexin V-Pacific Blue Apoptosis Detection Kit (BioLegend, 640930/640926). Data produced by the flow cytometer were analyzed using the FlowJo software.

For antibody-based assay (11), CAR-T or CAR-Jurkat cells were stained with Brilliant Violet 421™ anti-FLAG Tag Antibody at 4°C for 30 min and washed with PBS. Then these cells were incubated at 37°C for 0, 1, 2, or 4 h in the medium. Next, these cells were stained with Alexa Fluor^® 647 Goat anti-rat IgG (minimal x-reactivity) Antibody (BioLegend, 405416) at 4°C for 30 min. The decrease in the percentage of cells staining positive for the secondary antibody was quantified by flow cytometry as an indication of the internalization rate. For Brefeldin A-based assay (12), CAR-T cells were treated with 10 μM Brefeldin A (Topscience, T6062) at 37°C for 0, 1, 2, or 4 h, and then the surface CAR was detected by the anti-FLAG antibody. CAR-Jurkat cells were co-cultured with SEM at a 1:1 ratio for 0, 1, 2, or 4 h with the presence of 10 μM Brefeldin A, and the surface CAR was detected by anti-FLAG antibody.

For antibody-based assay (11), CAR-T or CAR-Jurkat cells were stained with Brilliant Violet 421™ anti-FLAG Tag Antibody at 37°C for 30 min and washed with PBS. Then the cells were stained with Alexa Fluor^® 647 Goat anti-rat IgG (minimal x-reactivity) Antibody at 4°C for 30 min (baseline) or at 37°C for 30 min or 60 min. The increase in the percentage of cells staining positive for the secondary antibody was quantified by flow cytometry as an indication of recycling.

CAR-T cells alone or co-cultured with SEM at a 1:1 ratio was treated with 50 µg/mL cycloheximide (MedChemExpress, HY-12320) at 37°C for 0, 1, 2, or 4 h. The total CAR was staining intracellular FLAG expression by anti-FLAG Tag Antibody after using BD Cytofix/Cytoperm Buffer System. The decrease in the percentage of total CAR was quantified by flow cytometry as an indication of degradation rate.

CAR-T or CAR-Jurkat cells were co-cultured with SEM cells at a 1:1 ratio for 4 h, and the cells were then stained with mouse anti-FLAG antibody (Sigma, F1804) in combination with one of the following rabbit antibodies: anti-human Rab5 (CST, 3547P), anti-human Rab11 (Proteintech, 15903-1-AP), or anti-human LAMP1 (Abgent, AP1823a-ev). Then Goat anti-Mouse IgG (H+L) Secondary Antibody, DyLight™ 650 (Thermo Fisher Scientific, SA5-10174) was used to stain mouse anti-FLAG antibody, and Goat anti-Rabbit IgG (H+L) Highly Cross-Adsorbed Secondary Antibody, Alexa Fluor™ 568 (Thermo Fisher Scientific, A11036) was used to stain rabbit anti-human Rab5, Rab11, or LAMP1. Positive CAR-T cells were identified by staining with the anti-FLAG antibody.

CAR-T cells were sorted after co-culture with SEM cells at 1:1 for 24 h. The mRNA-Seq library was constructed and then sequenced using the MGISEQ-2000RS (BGI Technology Service). Differentially expressed genes were selected based on |log2FoldChange| > 1 and Q-value < 0.05.

CAR-T cells were co-cultured with target cells for the indicated time. Then CD69 was detected by APC anti-human CD69 antibody (BioLegend, 310910) and CD25 was detected by PE anti-human CD25 antibody (BioLegend, 302606) as an indication of CAR-T activation. For exhaustion detection, CAR-T cells were co-cultured with target cells (SEM or CD19-K562) at the indicated E/T ratio for 3 days, and exhaustion markers on CAR-T cells were detected by APC anti-human CD279 (PD-1) Antibody (BioLegend, 367405), PE anti-human CD223 (LAG-3) Antibody (BioLegend, 369305), and PerCP/Cyanine5.5 anti-human CD366 (TIM-3) Antibody (BioLegend, 345015).

CAR-T cells were co-cultured with SEM cells at a 1:1 ratio for 24 h, and the supernatant was collected for detection of cytokines by LEGENDplex™ Human CD8/NK Panel (BioLegend, 741065).

CAR-T cells alone or co-cultured with SEM at a 1:1 ratio for the indicated time, and 200 nM TMRE (MedChemExpress, HY-D0985A) was used to detect CAR-T cells mitochondrial activity, 200 nM MitoTracker Deep Red FM (Beyotime Biotechnology, C1032) was used to detect CAR-T cells mitochondrial mass, and 5 μM MitoSOX Red (MedChemExpress, HY-D1055) was used to detect CAR-T cells mitochondrial ROS.

CAR-T cells were co-cultured with SEM cells at 1:1 for 3 days, and CD62L or CD45RA on CAR-T cells were detected by APC anti-human CD62L (BioLegend, 304809) or PE anti-human CD45RA (BioLegend, 304108) antibodies.

NOD/SCID mice were purchased from Vital River Laboratory. For survival assay, 1 × 10⁶ SEM-mRuby2-Luciferase tumor cells were injected into mice (5 weeks, female) intravenously. Mice were allocated randomly into different experimental groups (3 groups [T, 19CAR-T(8α), 19CAR-T(1a)], n=4) and then intravenously administered 1 × 10⁶ CAR-T cells on the third day after injection of tumor cells. Tumor growth was monitored by the quantitative imaging system. For CAR-T cell detection in vivo, 1 × 10⁶ SEM-mRuby2-Luciferase tumor cells were injected into mice (5 weeks, female) intravenously. Mice were allocated randomly into different experimental groups (3 groups [T, 19CAR-T(8α), 19CAR-T(1a)], n=3) and then intravenously administered 1.5 × 10⁶ CAR-T cells on the 17th day after injection of tumor cells. Then mice were sacrificed on the third day after injection of CAR-T cells. The total cells in bone marrow were collected. First, the dead cells were stained by Fixable Viability Dye eFluo™ 780 (Thermo Fisher Scientific, 65-0865-14). Then the cells were treated by TruStain Fc™ PLUS (anti-mouse CD16/32) Antibody (BioLegend, 156604) on ice for 10 min. Finally, PerCP/Cyanine5.5 anti-human CD45 (BioLegend, 982312), Brilliant Violet 605™ anti-human TIM-3 (BioLegend, 345017), Brilliant Violet 785™ anti-human LAG-3 (BioLegend, 369321), and APC anti-human PD-1 antibodies were stained at 4°C for 30 min. Data produced by the flow cytometer (BD FACSymphony™ A3) were analyzed using the FlowJo software.

The two-tailed Student t-test was used to analyze the differences between the control and experimental groups. The log-rank Kaplan-Meier survival test was used to compare the survival distributions of the different treatment groups in an SEM-mRuby2-Luciferase xenograft mouse model. The Pearson correlation coefficient was calculated by ImageJ. The statistical significance level was calculated by GraphPad Prism indicated as * for P < 0.05, ** for P < 0.01, *** for P < 0.001, or **** for P < 0.0001, and the ns indicate no significant difference. Error bars reflect ± SD of three independent experiments.

Here, we constructed a novel CAR incorporating the TMD from the CD1a molecule, differing from traditional CD8α transmembrane CAR (13, 21, 22). The other domains of the CAR were FLAG, CD19 single-chain variable fragment (FMC63), CD8 hinge, CD28 and 4-1BB costimulatory domain, CD3ζ signal domain, and copGFP ( Figure 1A ). We named these two different CAR-T 19CAR-T(8α) and 19CAR-T(1a), respectively. Both CAR-T cells were generated by infecting CD3⁺ T cells with the same multiplicity of infection (MOI = 15) ( Figure 1B ). Here, the CAR levels in 19CAR-T(1a) were nearly the same as those in 19CAR-T(8α) based on GFP signals, although 19CAR-T(1a) showed slightly lower levels ( Figure 1C ). However, we measured surface CAR on CAR-T cells using anti-FLAG antibody by flow cytometry and discovered that the CAR molecules on 19CAR-T(8α) cell surface were almost 3-fold than 19CAR-T(1a), showing a difference from GFP signal ( Figures 1D, E ). We hypothesized that this difference was due to the accumulation of CAR molecules in the cytoplasm of 19CAR-T(1a) cells rather than on the cell surface. To test this, intracellular flow cytometry was used to detect total CAR in CAR-T cells. The result showed that the total CAR in 19CAR-T(1a) was relatively high and consistent with the GFP signal ( Figures 1F, G ). These data indicated that 19CAR-T(1a) had low surface CAR levels, with a significant amount of CAR retained intracellularly. Since surface CAR expression has been shown to influence CAR-T cell cytotoxicity (8–10), we performed both short- and long-time co-culture of CAR-T cells with SEM tumor cells. Surprisingly, there was no significant difference in SEM cell apoptosis between these two CAR-T cells ( Figures 1H, I ). This suggests that despite the lower surface CAR expression, 19CAR-T(1a) cells possess sufficient cytotoxicity to kill tumor cells, comparable to 19CAR-T(8α) cells.

According to the study that the CD1a TMD controls the endocytosis of CD1a protein (19), the exploration of dynamic change of CAR molecules on 19CAR-T(1a) is needed. First, CAR-T cells were co-cultured with SEM cells, and the surface CAR was detected. We observed that 19CAR(1a) decreased less than 19CAR(8α) ( Figure 2A ). Given the different expressions of surface CAR on these two CAR-T cells, the results may be influenced by baseline surface CAR levels. Therefore, we constructed 19CAR-Jurkat(8α) cell lines expressing high, middle, and low levels of CAR (named 19CAR-J(8α)^H, 19CAR-J(8α)^M, and 19CAR-J(8α)^L), with 19CAR-J(8α)^M expressing surface CAR levels similar to 19CAR-J(1a) ( Supplementary Figures S1A, B ). Then the cytotoxicity of 19CAR-J(8α)^M and 19CAR-J(1a) was confirmed by detecting the apoptosis of SEM cells under different E/T ratios. The results reflected similar cytotoxicity between these two CAR-J cells, though 19CAR-J(1a) could induce slightly more tumor cell apoptosis under low E/T after 72 h of co-culture ( Supplementary Figure S1C ). Next, we employed these CAR-J cells to co-culture with SEM cells at 1:2 for 24 h and measured the surface CAR changes. We found that CAR levels decreased across all CAR-J cells, and 19CAR-J(8α)^H and 19CAR-J(8α)^M decreased more CAR molecule than 19CAR-J(8α)^L, reflecting the surface CAR levels influence the statistical analysis of the decrease of CAR ( Supplementary Figures S2A, B ). Importantly, 19CAR-J(1a) exhibited a more significant decrease in surface CAR than 19CAR-J(8α)^M ( Supplementary Figures S2A, B ), and the result was confirmed by co-culture of SEM and 19CAR-J(8α)^M and 19CAR-J(1a) at different E/T ratios ( Supplementary Figures S2C–E ). There was a difference in the results between the CAR-J and CAR-T cells due to the different surface CAR expressions. Thus, to directly assess CAR internalization, we employed a previously reported assay (11). The results showed that 19CAR(1a) had a faster internalization rate than 19CAR(8α), both in CAR-T and CAR-J cells ( Figure 2B ; Supplementary Figure S2F ). To further validate the conclusion, brefeldin A (BFA), a compound inducing fusion of early endosomes with the trans-Golgi network, was used to study the internalization of CAR (12). The results suggested that 19CAR(1a) had rapid internalization rate than 19CAR(8α) under BFA treatment ( Figure 2C ). What’s more, the same result could be seen in co-cultured CAR-J and SEM cells under BFA treatment ( Supplementary Figure S2G ). Besides the internalization of CAR, we also investigated the recycling rate of CAR molecules according to the method reported previously (11). The results indicated that 19CAR-T(1a) had a higher rate of CAR recycling compared to 19CAR-T(8α), and the same results could be seen on the CAR-J cells ( Figure 2D ; Supplementary Figure S2H ). Collectively, these data suggest that the CD1a transmembrane CAR shows a rapid internalization and recycling rate.

Interactions between cells, such as trogocytosis, have been shown to affect CAR-T cell function (23). Hence, we measured CAR and CD19 proteins on effector or target cells after co-culture. We found that SEM acquired a small amount of CAR molecules after co-culture with CAR-T cells, with 19CAR-T(8α) transferring more CAR proteins to SEM than 19CAR-T(1a) ( Figure 2E ; Supplementary Figure S3A ). To confirm these results, we repeated the experiment using CAR-J cells. The results showed that 19CAR-J(8α)^H transferred more CAR proteins to SEM than 19CAR-J(8α)^M, and there was no significant difference between 19CAR-J(8α)^M and 19CAR-J(1a), suggesting that surface CAR levels, rather than the TMD, influenced the CAR transfer ( Supplementary Figure S3B ). Although only a small amount of CAR was transferred, we explored whether this part of CAR will affect CAR-J cell cytotoxicity. First, we sorted SEM cells after co-culture with 19CAR-J(8α)^H cells at 1:1 for 24 h and tracked the persistence of transferred CAR on the SEM surface over time. We found that the transferred CAR could not sustain on the SEM surface and almost disappeared after 72 h ( Supplementary Figure S3C ). Furthermore, when the sorted SEM cells were co-cultured again with 19CAR-J(8α)^H cells, there was no significant difference in apoptosis between SEM, SEM^J, and SEM^{19CAR-J(8α)H} cells, suggesting that the small amount of transferred CAR on SEM cells had no impact on its apoptosis induced by CAR-J cells ( Supplementary Figures S3D, E ). Similarly, we tracked the levels of CD19 proteins on CAR-T cells over time and found that 19CAR-T(8α) cells acquired slightly more CD19 proteins than 19CAR-T(1a) ( Figure 2F ). However, only a small amount of CD19 was transferred compared to CD19 on SEM cells ( Supplementary Figure S3F ). Co-culture experiments with 19CAR-J(8α)^M and 19CAR-J(1a) cells showed that CD19 levels on SEM cells decreased significantly, consistent with the report that anti-CD19 CAR induced CD19 internalization (24), and there was no obvious difference between 19CAR-J(8α)^M and 19CAR-J(1a) ( Supplementary Figure S3G ). These data reveal that only a small amount of CD19 or CAR is transferred to effector or target cells, and the majority of CAR molecules undergo internalization.

Next, due to the 19CAR(1a) showing a rapid internalization rate, we explored whether the degradation of CAR would be influenced by TMD. We treated resting and activated CAR-T cells with cycloheximide to inhibit protein synthesis and observed that 19CAR(8α) underwent more degradation than 19CAR(1a) in both conditions ( Figures 2G, H ), suggesting the TMD influence the CAR degradation, likely due to the difference in the localization of CAR molecules post-internalization. CD1a molecule has been demonstrated to co-localize with early and recycling endosomes after internalization (25, 26). Here, we co-cultured CAR-T cells with SEM and explored the co-localization of early endosome marker Rab5 and recycling endosome marker Rab11 (27, 28) with CAR by immunofluorescence microscopy analysis. The results showed that more internalized 19CAR(1a) co-localized with Rab5 and Rab11 ( Figures 3A, B ), explaining the result that 19CAR-T(1a) had a rapid recycling rate. In contrast, more internalized 19CAR(8α) co-localized with LAMP1, a lysosome marker (26), indicating the fact that more CAR molecules were degraded in 19CAR-T(8α) compared to 19CAR-T(1a) ( Figures 3C ). In addition, these findings were further corroborated by similar experiments conducted in CAR-J cells, which yielded consistent results ( Supplementary Figure S4 ). In summary, these data suggest that internalized CD1a transmembrane CAR co-localizes more with early and recycling endosomes, which explains the rapid recycling rate observed in 19CAR-T(1a) cells.

Next, given the different surface CAR expression levels, we further investigated whether the CAR-T cell activation could be altered by CD1a transmembrane CAR. The CAR-T cells were sorted after co-culture with SEM cells at 1:1 for 24 h, followed by RNA sequencing (RNA-seq). The heatmap of differentially expressed genes (DEGs) showed that after co-culture, CAR-T cells were all activated and most of the DEGs were upregulated ( Supplementary Figure S5A ). Furthermore, the Venn diagram revealed that 19CAR-T(1a) had distinct transcriptomes from 19CAR-T(8α) after activation, among which 1920 DEGs were unique in 19CAR-T(1a) ( Supplementary Figure S5B ). In addition, a direct comparison of DEGs between 19CAR-T(1a) and 19CAR-T(8α) showed that most DEGs were downregulated in 19CAR-T(1a) ( Figure 4A ). Then, we used a volcano plot to show the signally upregulated or downregulated DEGs in 19CAR-T(1a) compared to 19CAR-T(8α) after activation. Notably, the cytokine genes IL2, IL3, IL4, and IL13 were significantly downregulated in 19CAR-T(1a) ( Figure 4B ). Pathway analysis of downregulated DEGs in 19CAR-T(1a) revealed different enriched pathways, such as positive regulation of cytokine production, positive regulation of leukocyte activation, cytokine-cytokine receptor interaction and so on ( Figure 4C ; Supplementary Figure S5C ), suggesting the lower activation levels in 19CAR-T(1a). This was further supported by a reduction in CD69 and CD25, T cell activation markers (29), in 19CAR-T(1a) after 24 h of co-culture with SEM cells at a 1:1 ratio ( Figure 4D ). In addition, the 12 cytokines analysis also showed lower production of IL-4, IL-10, and IFNγ in 19CAR-T(1a) ( Figure 4E ), confirming its reduced activation levels. Previous study reveals that the CAR-T cell activation levels are associated with exhaustion (8). Thus, we co-cultured CAR-T cells and SEM cells at 1:1 for 72 h and then the exhaustion markers PD-1, LAG-3, and TIM-3 were detected (30). We found that these exhaustion markers on 19CAR-T(1a) were all lower than 19CAR-T(8α) ( Figures 4F–H ), suggesting the reduced exhaustion in 19CAR-T(1a).

Next, to ensure that these results were independent of the target cells, we constructed another target cell, CD19-K562, which possessed exogenously introduced CD19 proteins ( Supplementary Figure S6A ). First, we co-cultured CAR-T cells and CD19-K562 cells at different E/T ratios, and surface CAR on CAR-T cells was detected. The results were the same as the CAR-T cells co-cultured with SEM cells ( Supplementary Figure S6B ). Most importantly, 19CAR-T(1a) still showed lower CD69 level than 19CAR-T(8α) after co-culture with CD19-K562 cells ( Supplementary Figure S6C ), suggesting the lower activation levels in 19CAR-T(1a). Furthermore, the exhaustion markers PD-1, LAG-3, and TIM-3 on 19CAR-T(1a) were all lower than 19CAR-T(8α) after co-culture with CD19-K562 cells at different E/T ratios (1:2, 1:4 or 1:8) for 72 h, which was the same as the results on SEM cells ( Supplementary Figure S6D ). In summary, these data reflect that CD1a transmembrane CAR-T cells exhibit reduced activation and exhaustion levels, and these results are due to the CAR TMD rather than target cells. The high production of cytokines may result in CRS (3), and lower cytokine production in 19CAR-T(1a) could potentially alleviate CRS.

Due to the different activation levels in 19CAR-T(8α) and 19CAR-T(1a) cells, we further explored whether the mitochondrial activity was influenced according to the research which suggested that mitochondrial function was associated with CAR-T cell function (31–33). Therefore, we measured the mitochondrial activity of these two CAR-T cells by TMRE and mitochondrial mass by using Mitotracker. We found that there was no significant difference in mitochondrial activity or mass between these two CAR-T cells at resting status ( Supplementary Figures S7A, B ). After short-time activation (4 h), both mitochondrial activity and mass increased, but no differences were observed between 19CAR-T(8α) and 19CAR-T(1a) ( Supplementary Figures S7C, D ). However, following long-time activation (24 h), both mitochondrial activity and mass were higher in 19CAR-T(8α) compared to 19CAR-T(1a) ( Figures 5A, B ), suggesting the higher activation level in 19CAR-T(8α), according to the study that T cell activation could increase mitochondrial mass (34). Here, we used the per-sample ratio of the TMRE/Mitotracker mean fluorescent index (MFI) as an indicator of activity per mitochondrion (35). The results reflected that 19CAR-T(1a) had higher mitochondrial activity per mitochondrion after both short- and long-time activation, although no difference was observed at resting status, suggesting the increased activity per mitochondrion was due to the CAR-T cell activation ( Figure 5C ; Supplementary Figures S7E, F ). In addition, we also measured the mitochondrial ROS by flow cytometry using MitoSOX. Similar to TMRE, no significant differences in ROS levels between these two CAR-T cells were observed at resting status or after short-time activation ( Supplementary Figures S7G, H ). After long-time activation, these two CAR-T cells all had increased mitochondrial ROS levels than T cells, with 19CAR-T(8α) displaying higher ROS levels, also suggesting the high activation level in 19CAR-T(8α) ( Figure 5D ). Similarly, the ROS level per mitochondrion was indicated by using MitoSOX/Mitotrcker MFI ratio. Interestingly, 19CAR-T(1a) showed higher ROS per mitochondrion after long-time activation ( Figure 5E ; Supplementary Figures S7I–J ). Previous studies have shown that increased mitochondrial activity correlated with a higher population of memory CAR-T cells (36). Thus, we detected memory CAR-T cells markers CD62L and CD45RA (37) after co-culture CAR-T cells with SEM at 1:1 for 72 h and found that 19CAR-T(1a) cells differentiated more into central memory CAR-T cells (Tcm, CD62L⁺CD45RA^-) ( Figure 5F ). At the same time, 19CAR-T(1a) cells had fewer population of effector T cells (Teff, CD62L^-CD45RA⁺) compared to 19CAR-T(8α) after 72 h of activation ( Figure 5G ). Collectively, our data indicate that CD1a transmembrane CAR-T cells have increased mitochondrial activity per mitochondrion upon activation and differentiate more into memory T cells, which may contribute to better clinical outcomes, as previously reported (35).

Here, given the lower activation level of 19CAR-T(1a) cells compared to 19CAR-T(8α) cells ex vivo, it was important to assess the anti-tumor efficacy of 19CAR-T(1a) cells in vivo. By injecting SEM-mRuby2-Luciferase tumor cells into NOD/SCID mice intravenously, we constructed an SEM xenograft model. After tumor cells injection for 3 days, T cells, 19CAR-T(1a) cells, and 19CAR-T(8α) cells were injected, respectively. Tumor growth was monitored by the quantitative imaging system ( Figure 6A ). We found that the tumor progression could be delayed by 19CAR-T(1a) treatment ( Figures 6B, C ). Survival analysis revealed that both 19CAR-T(1a) and 19CAR-T(8α) cells prolonged mouse survival, with 19CAR-T(1a) treatment leading to significantly better outcomes ( Figure 6D ). In addition, mice were sacrificed on the third day after injection of CAR-T cells ( Figure 6E ). Interestingly, the exhaustion markers PD-1 and LAG-3 were all lower on 19CAR-T(1a) cells, consistent with in vitro results ( Figure 6F ). Taken together, these data suggest that 19CAR-T(1a) cells have a superior anti-tumor ability in vivo, and CD1a TMD could reduce CAR-T cell exhaustion.