Preventing trogocytosis by cathepsin B inhibition augments CAR T cell function

Chimeric antigen receptor (CAR) T cell therapy has shown remarkable efficacy in cancer treatment. Still, most patients receiving CAR T cells relapse within 5 years of treatment. CAR-mediated trogocytosis (CMT) is a potential tumor escape mechanism in which cell surface proteins transfer from tumor cells to CAR T cells. CMT results in the emergence of antigen-negative tumor cells, which can evade future CAR detection, and antigen-positive CAR T cells, which is hypothesized to lead to CAR T cell fratricide and dysfunction. Using a system to selectively degrade trogocytosed antigen in CAR T cells, we show that the presence of trogocytosed antigen in CAR T cells directly causes CAR T cell fratricide and exhaustion. By performing a small molecule screening using a custom high throughput CMT-screening assay, we identified the cysteine protease cathepsin B (CTSB) as a key driver of CMT. We show that overexpression of cystatin A (CSTA), an endogenous human inhibitor of CTSB, reduces trogocytosis resulting in prolonged antitumor activity and increased CAR T cell expansion/persistence. Overall, we show that targeting CMT is an effective approach to enhance CAR T cell function, which may improve their clinical efficacy. .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted June 14, 2024. ; https://doi.org/10.1101/2024.06.11.598379doi: bioRxiv preprint

Chimeric antigen receptors (CAR) are genetically engineered proteins combining antigen binding domains with immune cell activation domains 1–3. T cells and natural killer (NK) cells engineered to express CARs recognize tumor-associated antigens with high specificity and are clinically effective treatments for patients with hematologic malignancies and autoimmune diseases 1,4–9. It has previously been shown that CAR T cells and CAR NK cells rapidly transfer the targeted antigen from tumor cells to their own cell surface in a process called trogocytosis 10–12. CAR-mediated trogocytosis (CMT) is associated with increased expression of apoptotic and exhaustion markers as well as reduced expansion10–12. In addition, it has been shown that T cells engineered to express the tumor antigen CD19 are efficiently killed by other CD19 CAR T cells (“fratricide”) 10. Mechanistically linking CMT to exhaustion, fratricide, and reduced CAR T cell expansion has, so far, been elusive due to the lack of specific inhibitors of trogocytosis. Analogously, no approaches exist to inhibit CMT therapeutically to enhance CAR T cell function. In this study, we develop an analytical framework to explore the mechanistic basis and functional consequences of trogocytosis in CAR T cells. Using this approach, we demonstrate that CAR T cell dysfunction is directly caused by CMT and we identify a genetically encoded, fully-human approach for the persistent suppression of CMT to increase expansion and antitumor activity of CAR T cells. .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted June 14, 2024. ; https://doi.org/10.1101/2024.06.11.598379doi: bioRxiv preprint

CAR-mediated trogocytosis directly causes CAR T cell dysfunction CAR-mediated trogocytosis (CMT) is the extraction of the targeted tumor antigen from the tumor cell surface and its incorporation into the CAR T cell plasma membrane 10–13 (Fig. 1A). Ex vivo, CMT occurs across cancer types 10,11 and target antigens, including CD1910,11, CD22 10, mesothelin 10, and BCMA 13. We found robust transfer of CD19 to CD19-targeting CAR T cells (clone: FMC63) as well as CD19 loss on tumor cells ( Fig. 1B/C), which can be observed just minutes after CAR T cells contact target cells ( Fig. 1D, Suppl Fig. S1A). In the clinical setting, CMT had only been shown to occur in B cell lymphoma patients treated with CAR NK cells 12. By analyzing peripheral blood mononuclear cells (PBMCs) from patients who had recently received CD19 CAR T cells, we observed that CMT also occurs in patients treated with CAR T cells ( Fig. 1E/F). The presence of trogocytosed antigen on CAR T cells is correlated with increased exhaustion and reduced viability 10–12 but it remains unknown if CMT is the cause of these effects. We, therefore, developed an approach for the targeted degradation of trogocytosed CD19 fused to GFP (CD19-GFP) in CAR T cells ( Fig. 1G). Expression of a trogocytosed antigen degrader (TAD), the Nslmb E3-targeting domain 14,15 fused to a GFP-binding protein (TADGFP), results in a significant reduction of total and surface CD19 in 293T cells engineered to express CD19-GFP ( Suppl Fig. S1B/C). Expression of TAD GFP in CAR T cells did not alter their expansion during manufacturing or their short-term antitumor activity (Suppl Fig. S1D/E). When cocultured for 2 hours with K562 cells expressing CD19-GFP, unmodified FMC63 CAR T cells showed increased levels of CD19 ( Fig. 1H) and GFP ( Fig. 1I) indicative of .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted June 14, 2024. ; https://doi.org/10.1101/2024.06.11.598379doi: bioRxiv preprint CMT. When overexpressing TADGFP in FMC63 CAR T cells, the amounts of CD19 and GFP were significantly reduced in the CAR T cells after coculture (Fig. 1H/I). In addition, TADGFP expression resulted in significantly increased CAR T cell numbers ( Fig. 1M ) independent of CAR T cell proliferation ( Suppl. Fig. S1F ) indicating reduced levels of fratricide. Targeted degradation of trogocytosed antigen was further validated using a fully human TAD (TAD CD19) construct directly targeting CD19. To recognize CD19, TADCD19 uses the Lyn kinase SH2 domain ( Fig. 1J), which has been shown to bind the intracellular domain of CD19 with high affinity. Expression of TAD CD19 in FMC63 CAR T cells resulted in reductions in CMT comparable to TADGFP (Fig. 1K/L). It has previously been shown that the FMC63 CAR can bind to CD19 in cis16. Following acquisition of CD19 by CMT, CD19 + CAR T cells may therefore experience persistent CAR signaling, which in turn may result in T cell exhaustion (Suppl. Fig. 1G ). We found that FMC63 CAR T cells exposed to CD19-GFP + K562 cells showed increased expression of exhaustion markers TIM-3 and LAG-3 compared to negative control T cells expressing a CAR without a binding domain ( ∆ scFv, Fig. 1N ). Expression of TADGFP prevented TIM-3/LAG-3 upregulation demonstrating that increased exhaustion marker expression is the direct consequence of antigen transfer to CAR T cells ( Fig. 1N). CMT-induced fratricide and exhaustion could substantially limit long-term CAR T cell antitumor activity. In a serial coculture stress test of CAR T cell efficacy ( Fig. 1O), we found that TAD GFP substantially increased the ability of CAR T cells to control B cell acute lymphoblastic leukemia (B-ALL, Fig. 1P). .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted June 14, 2024. ; https://doi.org/10.1101/2024.06.11.598379doi: bioRxiv preprint These data demonstrate that CMT directly causes CAR T cell dysfunction and fratricide, limiting expansion and antitumor activity. Extracellular cathepsin B is a key driver of CAR-mediated trogocytosis Little is known regarding the molecular and cellular drivers of CMT, and to date there are no high-throughput assays to systematically probe potential modulators of CMT as it occurs in real-time. To address this, we developed a luciferase complementation assay for the real-time detection of CMT (CompLuc, Fig. 2A ). In this assay, luciferase complementation occurs following the transfer of CD19 fused to the C-terminal fragment of NanoLuc 17 (cLuc) to CAR T cells, which express a complementary N-terminal NanoLuc (nLuc) fragment with high affinity for cLuc ( Suppl. Fig. S2A-C). The CompLuc assay allows highly sensitive detection of small numbers of nLuc +cLuc+ CAR T cells in cocultures (Suppl. Fig. S2D). We demonstrate that CMT, as quantified by CompLuc, is antigen-dependent and correlates with effector-target ratio ( Fig. 2B/C ) and flow cytometry endpoint analysis ( Fig. 2B ). To further validate the CompLuc assay, we determined the effect of an established modulator of CMT, the actin polymerization inhibitor cytochalasin D 10,18. Treatment of CAR T cells with cytochalasin D significantly reduced CMT in a dose-dependent manner as measured by the CompLuc assay (Fig. 2D). Using CompLuc, we next investigated the possibility of targeting CMT without altering CAR T cell-mediated tumor cell killing. Actin polymerization inhibitors, including cytochalasin D, significantly limit T cell cytotoxicity ( Fig. 2E) but we hypothesized that CMT could be mechanistically distinct from cytotoxic function and could be targeted .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted June 14, 2024. ; https://doi.org/10.1101/2024.06.11.598379doi: bioRxiv preprint selectively. Because it has previously been shown that CMT requires direct cell-to-cell contact10, we explored the effect of inhibiting various processes at the T cell immune synapse on CMT 10,18–21 ( Fig. 2F ). We found that inhibition of cathepsin B (CTSB) in FMC63 CAR T cells reduced CMT in a dose-dependent manner as measured by CompLuc (Fig. 2G) without substantially altering CAR T cell cytotoxicity (Fig. 2H). Using a membrane-impermeable inhibitor of CTSB, we found that inhibition of extracellular CTSB is sufficient to prevent CMT (Fig. 2I). CTSB is a ubiquitously expressed cysteine protease, primarily localized to lysosomal and endosomal compartments under physiological conditions, where it is primarily involved in protein degradation and turnover 22,23. However, CTSB is also found in the cytosol, retains its catalytic activity at neutral pH 24, and has been shown to line exocytic granules of cytotoxic T lymphocytes 25. In addition to proteolytic degradation within the endosome and lysosome, CTSB has been implicated in the degradation and remodeling of components of the extracellular matrix, such as collagen and fibronectin, thereby promoting the invasion and metastasis of tumor cells 24,26,27. Similarly, CTSB may contribute to the extraction of antigen-rich membrane fragments prior to their transfer to CAR T cells. To determine whether CTSB localizes to the immune synapse upon antigen contact, we generated FMC63 CAR T cells expressing CTSB fused to mCherry. When exposed to immobilized CD19 (Fig. 2J), we found that CTSB in FMC63 CAR T cells, but not ∆ scFv CAR T cells, rapidly localized to the immune synapse following CAR T cell antigen recognition ( Fig. 2K-M, Suppl. Fig. 1A ), correlating with increased trogocytosis as measured by CompLuc and flow cytometry (Fig. 2N). .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted June 14, 2024. ; https://doi.org/10.1101/2024.06.11.598379doi: bioRxiv preprint Taken together, quantification of CMT by CompLuc enabled the identification of extracellular CTSB produced by CAR T cells as a key driver of CMT that is dispensable for CAR-mediated killing. These data indicate that CTSB is a potential therapeutic target to prevent trogocytosis and thereby improve CAR T cell efficacy. Expression of human cystatins increases CAR T cell expansion by preventing CTSB- mediated CMT. To date there are no clinically approved small molecule inhibitors of CTSB, and its systemic inhibition may result in substantial toxicities due to the ubiquitous expression of CTSB 28. Cystatin A (CSTA), a cysteine protease inhibitor in the cystatin superfamily29, is an endogenous human inhibitor of CTSB 30. Similar to Ca-074-Me, CSTA prevents access to the active-site cleft of CTSB through insertion of a hydrophobic wedge 31 (Fig. 3A). We hypothesize that overexpression of CSTA in CD19 CAR T cells may selectively and efficiently reduce CMT by inhibiting CTSB (Fig. 3B). Stable overexpression of CSTA in FMC63 CAR T cells (CAR CSTA) resulted in significantly increased amounts of CSTA (Fig. 3C) and reduced CTSB activity (Fig. 3D). CARCSTA cells showed equivalent expansion ( Fig. 3E), phenotype ( Fig. 3F), and short- term antitumor activity ( Fig. 3G) compared to conventional FMC63 CAR T cells. We observed significantly reduced CMT in these cells as measured by CompLuc (Fig. 3H/I). CMT was similarly reduced when overexpressing the closely related superfamily member cystatin B 32 or a truncated variant of CSTA (CSTA 1-57) containing the hydrophobic wedge ( Suppl. Fig. S3A-D) . CSTA overexpression reduced both antigen transfer to CAR T cells ( Fig. 3K, Suppl. Fig. S3F ) and antigen loss on tumor cells ( Fig. .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted June 14, 2024. ; https://doi.org/10.1101/2024.06.11.598379doi: bioRxiv preprint 3J, Suppl. Fig. S3E ), suggesting inhibition of CMT at the antigen extraction step. This CSTA-mediated inhibition of CMT resulted in significantly increased CAR T cell expansion when exposed to tumor cells (Fig. 3L, Suppl. Fig. S3G ). Taken together, these data demonstrate that inhibition of CTSB through overexpression of human CSTA significantly reduces CMT and improves CAR T cell expansion in vitro. Cystatin overexpression improves long-term CAR T cell expansion and efficacy We have shown that CMT drives CAR T cell exhaustion and fratricide resulting in reduced CAR T cell numbers and potentially affecting the cells' long-term antitumor activity. To explore this, we assessed CAR CSTA T cell persistence and antitumor activity in an extended serial coculture model (Fig. 4A ). We observed significantly higher numbers of CARCSTA T cells over time (Fig. 4B) despite no differences in activation (Fig. 4C) or proliferation (Fig. D ), indicating that CSTA overexpression results in improved expansion, likely driven by reduced fratricide. In addition, we observed that CAR CSTA T cells exhibit superior long-term in vitro tumor control ( Fig. 4E). Similarly, we observed significantly increased in vivo CARCSTA T cell expansion (Fig. 4F/G). Cystatin A CAR T cells show reduced trogocytosis in a BCMA CAR T cell model CMT has previously been shown to occur across cancer types and target antigens 10,13. In addition to CD19, the FDA has approved two CAR T cell products targeting B cell maturation antigen (BCMA) for the treatment of multiple myeloma 4,5,33,34 (Fig. 5A). We found that, like FMC63 CAR T cells, BCMA CAR T cells are subject to CMT as BCMA is transferred to CAR T cells ( Fig. 5B). In addition, we observed BCMA loss on multiple .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted June 14, 2024. ; https://doi.org/10.1101/2024.06.11.598379doi: bioRxiv preprint myeloma cells when cocultured with both FDA-approved CAR T cell products ( Fig. 5C). We next assessed whether CSTA overexpression is able to prevent BCMA-specific CMT. As observed for CD19 CAR T cells, CSTA expression in BCMA CAR T cells did not alter CAR T cell expansion during manufacturing ( Fig. 5D) or their phenotype ( Fig. 5E). BCMA CAR CSTA T cells showed slightly improved short-term antitumor activity against multiple myeloma cell lines ( Fig. 5F) and increased expansion at the end of the coculture (Fig. 5G) compared to conventional BCMA CAR T cells. Following exposure to K562 cells overexpressing a BCMA-GFP fusion protein, BCMA CAR CSTA T cells exhibited significantly reduced CMT as measured by CompLuc ( Fig. 5H/I ) and flow cytometry ( Fig. 5J/K, Suppl. Fig. 4A/B ), resulting in increased CAR T cell expansion (Fig. 5L). These data indicate that CSTA overexpression is an effective approach to reduce CMT across target antigens and to increase CAR T cell expansion. .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted June 14, 2024. ; https://doi.org/10.1101/2024.06.11.598379doi: bioRxiv preprint

CAR T cells have revolutionized cancer immunotherapy, but most patients receiving CAR T cell therapy relapse within 5 years of treatment 35–37. Several mechanisms of CAR-mediated relapse have been described, including CAR T cell-mediated trogocytosis (CMT). CMT affects CAR T cell function in three key ways: 1) by rendering tumor cells temporarily antigen negative, 2) by causing CAR T cells to be killed by other CAR T cells, and 3) by inducing exhaustion, potentially through CAR binding tumor antigen in cis, leading to persistent CAR signaling 10–12. Previous studies have demonstrated the presence of tumor antigen on CAR T cells following tumor cell cocultures and that the presence of trogocytosed antigen was correlated with increased apoptotic markers and exhaustion 10,11. However, it had remained unknown if the two were caused by CMT. In this study, using targeted degradation of trogocytosed antigen-GFP fusion proteins, we now causally link CMT to CAR T cell fratricide and exhaustion. We also demonstrate that the trogocytosed antigen degrader (TAD) approach can be adapted to degrade antigens by using endogenous protein domains binding to the target antigen’s intracellular domain. Further work will be needed to explore target specificity/orthogonality of such endogenous binding domains to ensure CAR T cell functionality. Additionally, using small molecule- and protein-based inhibitors, we show that the cysteine protease cathepsin B (CTSB) produced by CAR T cells is essential for CMT. When expressed in tumor cells, extracellular CTSB has been shown to contribute to the metastatic process by degrading and remodeling components of the extracellular matrix 24,26,27 . Analogously, our data show that CTSB produced by CAR T cells rapidly .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted June 14, 2024. ; https://doi.org/10.1101/2024.06.11.598379doi: bioRxiv preprint localizes to the immune synapse and that CMT, specifically antigen extraction from tumor cells, is conferred by extracellular CTSB. To develop an approach to specifically inhibit CMT in CAR T cells, we explored the overexpression of cystatins belonging to the stefin family, established inhibitors of cathepsins 29,31. We found that both cystatin A and cystatin B potently inhibited CTSB in CAR T cells. While the small molecule inhibitors used throughout this study have been shown to be specific for CTSB 38,39, the more robust inhibition of CMT by cystatins could also indicate a role of other cathepsins, such as cathepsin L and H in CMT 29,31. Future work will be needed to further delineate the precise molecular mechanism by which extracellular CTSB and potentially other cathepsin proteins are able to confer antigen extraction from tumor cells. Our study focused on CAR T cell approaches targeting the two clinically validated CAR T cell antigens CD19 and BCMA approved for the treatment of several hematologic malignancies. However, we propose that future work should include an in-depth exploration of CMT in solid malignancies. In this setting, exhaustion-driven T cell dysfunction is generally more pronounced 40,41 and CMT may play a key role. In addition, a denser tumor microenvironment and constrained T cell mobility may favor more extensive intratumoral fratricide. In conclusion, we show that CMT directly causes CAR-mediated fratricide and exhaustion, and that CMT can be mitigated by inhibition of CTSB using overexpression of human endogenous cystatins. Inhibition of CMT in turn results in prolonged antitumor activity and increased CAR T cell persistence. .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted June 14, 2024. ; https://doi.org/10.1101/2024.06.11.598379doi: bioRxiv preprint

Study approval All recombinant DNA and biosafety work was approved by the institutional biosafety committees at University of Maryland, Baltimore (protocol IBC-6040). Animal experiments were approved by the institutional animal care and use committee at the University of Utah (protocol 18-1104). Cell lines and primary human cells Raji, NALM6, Daudi, DB, Toledo, MM.1S, RPMI8226, U266B1, K562, and Phoenix- AMPHO cells were purchased from the American Type Culture Collection (ATCC) and cultured according to ATCC instructions. Lenti-X 293T cells were purchased from Takara and cultured according to the manufacturer’s instructions. Cell lines were authenticated by their respective supplier. Healthy donor buffy coats were obtained from the New York Blood Center. PBMCs from healthy donors were isolated from buffy coats by density gradient using FicollPaque (GE) as previously described 34,42. Primary human T cells were cultured in AIM-V medium (Invitrogen 12055-083) supplemented with 5% Human serum (Sigma H4522-100ML), 1% Pen/Strep (Thermofisher 15140-122) (T cell media), and 40 IU IL-2 (R&D Systems #202-IL-10). All cells were cultured at 37 °C, 5% CO2. Vector constructs All vectors generated for this study were produced by Twist Biosciences. FMC63-nLuc contains the CD19-specific scFv fragment FMC63 43. All CAR constructs contain the CD8α hinge/transmembrane domain, 4-1BB costimulatory domain, and CD3z domain. .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted June 14, 2024. ; https://doi.org/10.1101/2024.06.11.598379doi: bioRxiv preprint For some constructs, the full-length sequence of human cystatin A (UniProt, P01040) was synthesized and cloned downstream of the respective CAR and nLuc fragment separated by a P2A sequence (Twist Bioscience). DNA was isolated using Endofree Plasmid Maxi Kit (Qiagen 12362) following the manufacturer’s protocol. Plasmid concentration was measured using a NanoDrop One instrument (Thermo). All DNA constructs were stored at -20°C. Clinical data Whole blood was drawn from patients receiving CAR T cell therapy 7-28 days after CAR T cell injection. Peripheral blood mononuclear cells (PBMCs) were isolated by density gradient using Ficoll Paque as previously described 34,42. PBMCs were stained with anti- hCD3, anti-hCD4, anti-hCD8, anti-CAR, and anti-hCD19, anti-hCD27, anti-hCD137, and 7-AAD and analyzed by flow cytometry. Samples were collected under Institutional Review Board (IRB)-approved protocol 2043GCCC (IRB H0091736, PI: D. Atanackovic). Gammaretrovirus Production Gammaretrovirus was produced using Phoenix-AMPHO cells (ATCC, catalog no. CRL- 3213). Phoenix-AMPHO cells were transiently transfected with 16 µg of plasmid DNA using Opti-MEM Reduced Serum Medium (Thermofisher, catalog no. 31985070) and lipofectamine 2000 (Invitrogen 11668-019) according to manufacturer’s instructions. During transfection, cells were cultured in Dulbecco’s Modified Eagle Medium (DMEM, Thermofisher 11995073) supplemented with 10% FBS. Virus-containing supernatant was filtered using Steriflip™ Sterile Disposable Vacuum Filter Units (Millipore Sigma .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted June 14, 2024. ; https://doi.org/10.1101/2024.06.11.598379doi: bioRxiv preprint SEIM003M00). Virus was concentrated using Retro-X Concentrator (Takara, 631456). The following day, concentrated virus was centrifuged at 1500 x g for 45 minutes at 4°C. Supernatant was removed and concentrated virus was resuspended in 1.5 mL complete T cell media. Transgenic T cell production and expansion CAR T cells were generated as previously described 11,34,42. Buffy coats from healthy donors were obtained from the New York Blood Center and peripheral blood mononuclear cells (PBMCs) were isolated using Ficoll-Paque and cryopreserved until use. PBMCs were thawed and cultured overnight in complete T cell media. PBMCs were stimulated for 2 days with anti-CD3/anti-CD28 T cell activation beads (Thermo, catalog no. 11131D) in the presence of interleukin-2 (IL-2; 40 IU/ml; R&D Systems, catalog no. 202-IL-010) in complete T cell media and incubated at 37°C, 5% CO 2. Bead- stimulated cells were transferred to RetroNectin-coated (Takara) virus–containing plates and incubated overnight. Transduction was repeated the next day before counting and diluting cells to 0.4 × 10 6 cells/ml. After the second transduction, cells were grown for an additional 7 days before removing beads using a DynaMag-15 magnet (Thermo Fisher Scientific). IL-2 was replenished every 2 days to 40 IU/ml. Cells were frozen in 90% fetal bovine serum/10% dimethyl sulfoxide and stored in liquid nitrogen. CAR T cell transduction efficiency and phenotype were determined by flow cytometry. CAR T cells were washed with FACS buffer and stained with antibodies targeting hCD3, HA, hCD4, hCD8, hCD95, hCD62L, and hCD45RA. Imaging Substrate Preparation .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted June 14, 2024. ; https://doi.org/10.1101/2024.06.11.598379doi: bioRxiv preprint Eight-well chambers (Cellvis, catalog no. C8-1.5H-N) were used for all experiments. For CAR T cell activation on CD19-coated surfaces, 8-well chambers were coated with 0.01% poly-L-lysine (PLL) solution diluted in distilled water for 10 minutes at room temperature. PLL was aspirated from each well, and the chambers were allowed to air- dry for 1 hour at 37°C. PLL-coated dishes were then incubated overnight at 4°C with a 10 μ g/mL solution of NeutrAvidin (Thermo Scientific, catalog no. 31000) in 1x Dulbecco’s phosphate-buffered saline (DPBS). After overnight incubation, coated wells were washed with 1x DPBS at room tem perature and incubated at 37°C with a 10 μ g/mL solution of biotinylated human CD19 protein (ACRO Biosystems catalog no. CD9-H82E9) in DPBS fo r 2 hours at 37°C. Prior to the experiment, coated wells were washed three times with RPMI 1640 phenol red-free imaging medium. For coculture experiments, 8-well chambers wells were incubated with 10 μ g/mL fibronectin (MilliporeSigma, catalog no. 34-163-11MG) in DPBS at room temperature for 1 hour prior to seeding with 293T cells stably expressing CD19-GFP. 293T cells were seeded at a concentration of 5 x 10 5 cells per well in complete growth media, followed by an overnight incubation at 37°C, 5% CO2. Prior to imaging, wells were washed three times with warm complete imaging media, consisting of a 1:1 ratio of RPMI 1640 supplemented with 5% fetal bovine serum (FBS) and DMEM supplemented with 10% FBS. The washing process was performed thrice to ensure removal of any residual media while leaving a known volume in the wells. Confocal Microscopy Confocal microscopy was conducted using an inverted microscope (Nikon Ti-E PFS, Nikon Inc.) equipped with a 100× Silicone objective lens. Imaging was performed with a .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted June 14, 2024. ; https://doi.org/10.1101/2024.06.11.598379doi: bioRxiv preprint Prime BSI camera (Photometrics). Image acquisition protocols were managed using Nikon Elements software, and images were cropped in FIJI for further analysis. All live cell imaging was done with imaging chambers placed in a stage-top Okolab Incubator (Okolab S. R. L.) pre-equilibrated to 37°C with 5% CO 2. For live cell imaging on glass, activated CAR T cells suspended in RPMI 1640 medium supplemented with 5% FBS were deposited onto CD19-coated surfaces. Imaging was started between 3 to 6 minutes after CAR T cells expressing CTSB-mCherry were added to a biotin-CD19- coated coverslip. For each well, timelapse images of a 3D volume (planes with z spacing of 0.3 μ m to span the cell from the top to the bottom) were acquired every 3 minutes for 60 minutes. For coculture experiments, imaging was started between 3 to 6 minutes after CAR T cells expressing CTSB-mCherry were dropped onto a layer of 293T cells expressing CD19-GFP seeded on a coverslip, at a concentration of 7 x 10 4 cells per drop. Brightfield imaging was used to identify cells attached to the apical surface of HEK293 cells. The synaptic plane between a CAR-T cell and a HEK293 cell was identified and designated as the home plane for acquisition of Z-stack time-lapse movies. Two-channel images using 488 nm and 561 nm lasers (for GFP and mCherry imaging respectively) were acquired every 3 minutes for 1 hour and 15 minutes utilizing a Z-spacing of 0.3 or 0.6 μ m, with brightfield images taken at the home plane. Image analysis was carried out in MATLAB (Mathworks, Inc.) using custom scripts. The plane of the synapse was determined using the actin channel. After background subtraction, axial intensity gradients are estimated for every voxel of sufficient intensity within the ROI. Below the cell, these gradients are typically positive due to the Airy .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted June 14, 2024. ; https://doi.org/10.1101/2024.06.11.598379doi: bioRxiv preprint pattern of the PSF. The synapse is taken as the first plane for which the gradients are no longer consistently positive. Estimating the average distance of CTSB to the synapse The center of fluorescence (COF) of CTSB is defined similarly to the center of mass. The voxel positions are weighted by CTSB intensity after background subtraction and thresholding to obtain the COF. The average CTSB distance is then calculated from the axial (z) distance of the Cathepsin B COF to the plane of the synapse. Characterizing CTSB axial dispersion The axial dispersion is an estimate of the average distance of CTSB molecules to the COF. The axial (z) distance of each voxel to the COF is determined, and the axial dispersion is defined as the average of these distances weighted by CTSB intensity after background subtraction and thresholding. Voxels with sufficiently low signal do not contribute to the calculation due to the thresholding procedure. CTSB clustering at the synapse The pair auto-correlation function g(r) of CTSB is computed at the synapse using the actin channel as a mask 44. The actin channel is segmented by smoothing with a Gaussian filter, generating an initial mask by applying k-means clustering (k = 2) after a log transformation, and then applying morphological operations to connect and smooth the initial mask. The clustering coefficient, gave, is computed by averaging over all radial bins with 0.25 m or 0.5 m. CompLuc-based trogocytosis assay .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted June 14, 2024. ; https://doi.org/10.1101/2024.06.11.598379doi: bioRxiv preprint Transduced, cryopreserved nLuc+ CAR T cells were thawed and cultured in complete T cell media supplemented with 40 IU IL-2 for 48 hours prior to use. nLuc-expressing CAR T cells were cocultured with CD19-cLuc-expressing K562 tumor cells at the specified effector-target ratios. CAR T cells and tumor cells were resuspended in Opti-MEM reduced serum media. Live Cell Substrate (Promega, catalog no. N2011) was prepared according to manufacturer’s instructions. K562 tumor cells and prepared Live Cell Substrate were added to wells of a black 96-well plate and luminescence was measured to assess baseline luminescence. CAR T cells were added to appropriate wells and luminescence was measured every minute for three hours at 37˚C. Luminescence was measured using a Spark multi-mode plate reader (Tecan). Flow cytometry-based trogocytosis assay A flow-cytometry based trogocytosis assay was used to confirm results observed in CompLuc, to assess CD19 levels on CAR T cells and tumor cells, and to quantify CAR T cells. 2 x 10 5 target cells were seeded in wells of a 96-well round bottom plate. Various ratios of CAR T cells were cocultured with target cells for 1 hour at 37˚C, 5% CO 2. Following coculture, cells were resuspended by gentle pipetting and transferred to wells of a 96-well V bottom plate for washing and staining. Cells were stained with Zombie violet fixable viability dye, and antibodies for hCD3, HA, and hCD19. Accucheck counting beads (Life Technologies) were added to the cells. Samples were acquired on an LSR II flow cytometer (BD) or an Aurora full-spectrum flow cytometer (Cytek). Luciferase-based cytotoxicity assay To determine in vitro CAR T cell cytotoxicity, cell lines (Raji, NALM6, Daudi, Toledo) were transduced with pHIV-Luc-ZsGreen lentivirus and sorted on a FACS Aria flow .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted June 14, 2024. ; https://doi.org/10.1101/2024.06.11.598379doi: bioRxiv preprint cytometer (BD) for GFP expression. 3 x 10 4 target cells were seeded in wells of a 96- well round bottom plate. CAR T cells were cocultured with target cells at the indicated effector-target ratios and incubated for 16 hours at 37˚C, 5% CO2. Following incubation, 80 μ L of supernatant was harvested from each well. Cells were suspended by gentle pipetting, and 100 μ L was transferred to a 96-well black flat bottom plate. D-Luciferin (Gold Biotechnology, catalog no. LUCNA-2G) at 150 μ g/ml was added to the cells and incubated for 5 minutes at 37°C. Luminescence was determined on a Spark multimode plate reader (Tecan). Serial coculture repeat stimulation assay To determine the long-term in vitro control and exhaustion of CAR T cells, luciferase- expressing tumor cells were plated at 3 x 10 4 cells/well. CAR T cells were cocultured at a defined effector-target ratio and incubated for 24 hours at 37˚C, 5% CO 2. Following incubation, cells were transferred to a 96-well black flat bottom plate. D-luciferin was added to cells and luminescence was determined on a Spark multimode plate reader (Tecan). Next, remaining CAR cells were pooled together and normalized based on expansion. CAR T cells were redistributed to wells and fresh tumor cells were added to each well. Plates were incubated for 48 hours at 37˚C, 5% CO 2. Luminescence measurements and normalization were repeated until a loss of cytotoxicity was observed. Cystatin A enzyme-linked immunosorbent assay Cystatin A concentration was assessed using a Human Cystatin A ELISA kit (Invitrogen, catalog no. EH140RB). Total cell lysates were extracted from CAR T cells using radioimmunoprecipitation assay (RIPA) buffer (Thermo) containing protease inhibitor .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted June 14, 2024. ; https://doi.org/10.1101/2024.06.11.598379doi: bioRxiv preprint cocktail (Roche). Total protein concentration was determined using Pierce BCA assay (Thermo). Cystatin A levels were determined by enzyme-linked immunosorbent assay (ELISA) according to manufacturer’s instructions (Invitrogen) and calculated using standard curve interpolation. Absorbance was measured at the recommended wavelength on a Spark multimode plate reader (Tecan). Cathepsin B activity assay Cathepsin B activity was assessed using the InnoZyme Cathepsin B Activity Assay Kit (MilliPore Sigma, catal og no. CBA001). Total cell lysates were extracted from CAR T cells using the provided cell lysis buffer according to manufacturer’s instructions. Total protein concentration was determined using Pierce BCA assay (Thermo Fisher Scientific). Cathepsin B activity was determined fluorometrically according to manufacturer’s instructions (Calbiochem). Fluorescence was measured on a Spark multimode plate reader (Tecan). Treatment of CAR T cells with inhibitors FMC63 CAR T cells were treated with small-molecule inhibitors targeting Actin (inhibitor: Cytochalasin D, Sigma Aldrich catalog no. C8273), Dynamin (inhibitor: DynaSore, Sigma Aldrich catalog no. D7693), Cathepsin B (inhibitor: Ca-074-Me, SelleckChem catalog no. S7420), Clathrin (inhibitor: PitStop, Abcam catalog no. ab120687), or LFA-1 (inhibitor: BI-1950, Boehringer Ingelheim) at the indicated concentrations for one hour. Following treatment, cells were washed with T cell media and centrifuged at 400 x g for 5 minutes. To determine the effect of the non-membrane permeable CTSB inhibitor CA -074 (Millipore Sigma, catal og no. 205530), the inhibitor was added directly to the co-culture at the indicated concentrations. As a control, CAR T .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted June 14, 2024. ; https://doi.org/10.1101/2024.06.11.598379doi: bioRxiv preprint cells treated with DMSO. CAR T cells were cocultured with CD19-expressing tumor cells for the indicated intervals. Western blot Total cell lysates were extracted from CAR T cells using radioimmunoprecipitation assay (RIPA) buffer (Thermo Fisher Scientific) containing protease inhibitor (Roche). Total protein concentration was determined using Pierce BCA assay (Thermo Fisher Scientific). Samples were separated by SDS–polyacrylamide gel electrophoresis, and separated proteins were transferred to nitrocellulose membranes using an iBlot2 transfer system (Thermo Fisher Scientific). Membranes were blocked with 5% nonfat milk–tris-buffered saline and incubated with primary antibodies against CD19 or β-actin. Membranes were washed and developed using species-specific secondary anti- IgG/horseradish peroxidase antibodies (R&D Systems) and Western Lightning Plus- ECL solution (PerkinElmer). Bands were visualized and quantified on an iBright 1500 imaging system (Thermo Fisher Scientific). In vivo cystatin A model of CAR T cell expansion 6–8-week-old male NRG mice (Jackson Laboratory) were irradiated with a sublethal dose of 200 cGy (Day -1). The next day (Day 0), mice were injected with 4 x 10 5 NALM6 tumor cells via tail vein injection. On Day 4, mice were injected with 1 x 106 FMC63 CAR T cells ± Cystatin A overexpression or CAR T cells lacking a binding domain (∆ scFv) via tail vein injection. Animals were weighed twice weekly and monitored for signs of distress in accordance with institutional regulations. On day 28, animals were euthanized and tissues collected for analysis by flow cytometry. Statistical Analysis .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted June 14, 2024. ; https://doi.org/10.1101/2024.06.11.598379doi: bioRxiv preprint The respective statistical tests are stated in the figure legends. Generally, statistical significance between two groups was determined by two-sided Student’s t-test or Mann- Whitney U test. Statistical significance between groups of three or more was determined by one- or two-way analysis of variance (ANOVA). Significance of differences between cell numbers, mean fluorescence intensity, and area under curve, was calculated by two-sided Student’s t test. Significance of differences in murine tumor control was determined by two-way ANOVA. All statistical tests were performed using Prism 10 (GraphPad). Results were considered significant when p < 0.05. .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted June 14, 2024. ; https://doi.org/10.1101/2024.06.11.598379doi: bioRxiv preprint

The authors thank the Flow Cytometry Shared Service of the University of Maryland Marlene and Stewart Greenebaum Comprehensive Cancer Center for help with sorting cells and panel design. The authors also thank the Huntsman Cancer Institute at the University of Utah for the use of the Preclinical Research Resource (PRR). This work was supported by funds through the Maryland Department of Health's Cigarette Restitution Fund Program (CH-649-CRF), the National Cancer Institute Cancer Center Support Grant (P30CA134274), the National Institutes of Health to AU (R35 GM145313), and by an NIAID-funded predoctoral fellowship to JMB (T32 AI095190). The plasmid pHIV-Luc-ZsGreen was a gift from Dr. B. Welm (Addgene #39196). The plasmid SFG.CNb30_opt.IRES.eGFP was a gift from Dr. M. Pule (Addgene #22493). AUTHOR CONTRIBUTIONS KAD and TL conceived the project, planned and performed experiments, analyzed data, and wrote the manuscript. KAD performed the majority of the described in vitro experiments, generated all transgenic T cells, developed the CompLuc assay, identified and validated the role of CTSB in CMT, and developed the CSTA-based approach of CMT inhibition in CAR T cells. KN and AP performed confocal microscopy and fixed coculture staining and imaging. JMB planned experiments, analyzed data, and edited the manuscript. EG performed CAR T cell phenotyping and analyzed data. FF wrote analysis scripts for CTSB dispersion. WS, ER, and DL planned and performed in vivo models and analyzed data. XF developed full-spectrum flow cytometry panel and performed flow cytometry analysis. AW planned experiments and analyzed data. DA .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted June 14, 2024. ; https://doi.org/10.1101/2024.06.11.598379doi: bioRxiv preprint supervised CAR T cell phenotyping. AU planned and supervised microscopy experiments and analyzed data. TL supervised all work related to this project. All authors reviewed and approved the manuscript. .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted June 14, 2024. ; https://doi.org/10.1101/2024.06.11.598379doi: bioRxiv preprint

1. Sadelain, M. CD19 CAR T Cells. Cell 171, 1471 (2017). 2. Maude, S. L. et al. Tisagenlecleucel in Children and Young Adults with B-Cell Lymphoblastic Leukemia. N. Engl. J. Med. 378, 439–448 (2018). 3. Sadelain, M., Rivière, I. & Riddell, S. Therapeutic T cell engineering. Nature 545, 423–431 (2017). 4. Raje, N. et al. Anti-BCMA CAR T-Cell Therapy bb2121 in Relapsed or Refractory Multiple Myeloma. N. Engl. J. Med. 380, 1726–1737 (2019). 5. D’Agostino, M. & Raje, N. Anti-BCMA CAR T-cell therapy in multiple myeloma: can we do better? Leukemia 34, 21–34 (2020). 6. Romanski, A. et al. CD19/i4 CAR engineered NK /i4 92 cells are sufficient to overcome NK cell resistance in B/i4 cell malignancies. J. Cell. Mol. Med. 20, 1287–1294 (2016). 7. Boissel, L. et al. Retargeting NK-92 cells by means of CD19- and CD20-specific chimeric antigen receptors compares favorably with antibody-dependent cellular cytotoxicity. OncoImmunology 2, e26527 (2013). 8. Mackensen, A. et al. Anti-CD19 CAR T cell therapy for refractory systemic lupus erythematosus. Nat. Med. 28, 2124–2132 (2022). 9. Mueller, F. et al. CD19-Targeted CAR-T Cells in Refractory Systemic Autoimmune Diseases: A Monocentric Experience from the First Fifteen Patients. Blood 142, 220 (2023). 10. Hamieh, M. et al. CAR T cell trogocytosis and cooperative killing regulate tumour antigen escape. Nature 568, 112–116 (2019). .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted June 14, 2024. ; https://doi.org/10.1101/2024.06.11.598379doi: bioRxiv preprint 11. Olson, M. L. et al. Low-affinity CAR T cells exhibit reduced trogocytosis, preventing rapid antigen loss, and increasing CAR T cell expansion. Leukemia 36, 1943–1946 (2022). 12. Li, Y. et al. KIR-based inhibitory CARs overcome CAR-NK cell trogocytosis- mediated fratricide and tumor escape. Nat Med 28, 2133–2144 (2022). 13. Camviel, N. et al. Both APRIL and antibody-fragment-based CAR T cells for myeloma induce BCMA downmodulation by trogocytosis and internalization. J. Immunother. Cancer 10, e005091 (2022). 14. Marei, H. et al. Antibody targeting of E3 ubiquitin ligases for receptor degradation. Nature 610, 182–189 (2022). 15. Békés, M., Langley, D. R. & Crews, C. M. PROTAC targeted protein degraders: the past is prologue. Nat. Rev. Drug Discov. 21, 181–200 (2022). 16. Ruella, M. et al. Induction of resistance to chimeric antigen receptor T cell therapy by transduction of a single leukemic B cell. Nat. Med. 24, 1499–1503 (2018). 17. Dixon, A. S. et al. NanoLuc Complementation Reporter Optimized for Accurate Measurement of Protein Interactions in Cells. ACS Chem. Biol. 11, 400–408 (2016). 18. Miyake, K. et al. Trogocytosis of peptide-MHC class II complexes from dendritic cells confers antigen-presenting ability on basophils. Proc Natl Acad Sci U S A 114, 1111–1116 (2017). 19. Gilmartin, A. A., Ralston, K. S. & Petri, W. A. Inhibition of Amebic Cysteine Proteases Blocks Amebic Trogocytosis but Not Phagocytosis. J. Infect. Dis. 221, 1734–1739 (2020). .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted June 14, 2024. ; https://doi.org/10.1101/2024.06.11.598379doi: bioRxiv preprint 20. Reed, J., Reichelt, M. & Wetzel, S. A. Lymphocytes and Trogocytosis-Mediated Signaling. Cells 10, 1478 (2021). 21. Hudrisier, D., Aucher, A., Puaux, A.-L., Bordier, C. & Joly, E. Capture of Target Cell Membrane Components via Trogocytosis Is Triggered by a Selected Set of Surface Molecules on T or B Cells. J. Immunol. 178, 3637–3647 (2007). 22. Cavallo-Medved, D., Moin, K. & Sloane, B. Cathepsin B. AFCS Nat Mol Pages 2011, A000508 (2011). 23. Mort, J. S. & Buttle, D. J. Cathepsin B. Int. J. Biochem. Cell Biol. 29, 715–720 (1997). 24. Vidak, E., Javoršek, U., Vizovišek, M. & Turk, B. Cysteine Cathepsins and Their Extracellular Roles: Shaping the Microenvironment. Cells 8, 264 (2019). 25. Balaji, K. N., Schaschke, N., Machleidt, W., Catalfamo, M. & Henkart, P. A. Surface Cathepsin B Protects Cytotoxic Lymphocytes from Self-destruction after Degranulation. Journal of Experimental Medicine 196, 493–503 (2002). 26. Olson, O. C. & Joyce, J. A. Cysteine cathepsin proteases: regulators of cancer progression and therapeutic response. Nat Rev Cancer 15, 712–729 (2015). 27. Gabrijelcic, D. et al. Cathepsins B, H and L in human breast carcinoma. Eur J Clin Chem Clin Biochem 30, 69–74 (1992). 28. Uhlén, M. et al. Tissue-based map of the human proteome. Science 347, 1260419 (2015). 29. Estrada, S. et al. The Role of Gly-4 of Human Cystatin A (Stefin A) in the Binding of Target Proteinases. Characterization by Kinetic and Equilibrium Methods of the .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted June 14, 2024. ; https://doi.org/10.1101/2024.06.11.598379doi: bioRxiv preprint Interactions of Cystatin A Gly-4 Mutants with Papain, Cathepsin B, and Cathepsin L. Biochemistry 37, 7551–7560 (1998). 30. Pavlova, A., Krupa, J. C., Mort, J. S., Abrahamson, M. & Björk, I. Cystatin inhibition of cathepsin B requires dislocation of the proteinase occluding loop. Demonstration by release of loop anchoring through mutation of His110. FEBS Lett. 487, 156–160 (2000). 31. Pavlova, A. & Björk, I. Grafting of Features of Cystatins C or B into the N-Terminal Region or Second Binding Loop of Cystatin A (Stefin A) Substantially Enhances Inhibition of Cysteine Proteinases. Biochemistry 42, 11326–11333 (2003). 32. Jenko, S. et al. Different propensity to form amyloid fibrils by two homologous proteins—Human stefins A and B: Searching for an explanation. Proteins: Struct., Funct., Bioinform. 55, 417–425 (2004). 33. Cohen, A. D. et al. B cell maturation antigen-specific CAR T cells are clinically active in multiple myeloma. J. Clin. Investig. 129, 2210–2221 (2019). 34. Radhakrishnan, S. V. et al. CD229 CAR T cells eliminate multiple myeloma and tumor propagating cells without fratricide. Nat. Commun. 11, 798 (2020). 35. Grover, P. et al. Chimeric antigen receptor T-cell therapy in adults with B-cell acute lymphoblastic leukemia. Blood Adv 6, 1608–1618 (2022). 36. Xu, X. et al. Mechanisms of Relapse After CD19 CAR T-Cell Therapy for Acute Lymphoblastic Leukemia and Its Prevention and Treatment Strategies. Front. Immunol. 10, 2664 (2019). .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted June 14, 2024. ; https://doi.org/10.1101/2024.06.11.598379doi: bioRxiv preprint 37. Gu, T., Zhu, M., Huang, H. & Hu, Y. Relapse after CAR-T cell therapy in B-cell malignancies: challenges and future approaches. J. Zhejiang Univ.-Sci. B 23, 793– 811 (2022). 38. Towatari, T. et al. Novel epoxysuccinyl peptides A selective inhibitor of cathepsin B, in vivo. FEBS Lett. 280, 311–315 (1991). 39. Murata, M. et al. Novel epoxysuccinyl peptides Selective inhibitors of cathepsin B, in vitro. FEBS Lett. 280, 307–310 (1991). 40. Newick, K., O’Brien, S., Moon, E. & Albelda, S. M. CAR T Cell Therapy for Solid Tumors. Annu. Rev. Med. 68, 139–152 (2016). 41. Sterner, R. C. & Sterner, R. M. CAR-T cell therapy: current limitations and potential strategies. Blood Cancer J 11, 69 (2021). 42. Mause, E. R. V. et al. Systematic single amino acid affinity tuning of CD229 CAR T cells retains efficacy against multiple myeloma and eliminates on-target off-tumor toxicity. Sci. Transl. Med. 15, eadd7900 (2023). 43. Lee, D. W. et al. T cells expressing CD19 chimeric antigen receptors for acute lymphoblastic leukaemia in children and young adults: a phase 1 dose-escalation trial. Lancet 385, 517–528 (2015). 44. Veatch, S. L. et al. Correlation Functions Quantify Super-Resolution Images and Estimate Apparent Clustering Due to Over-Counting. PLoS ONE 7, e31457 (2012). .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted June 14, 2024. ; https://doi.org/10.1101/2024.06.11.598379doi: bioRxiv preprint FIGURE LEGENDS Figure 1: CA R-m ediated trog oc ytosis dire ctly cau ses CA R T c e ll dysfun ction. ( A) Sch ema of CAR-med iated trog o cyt osis (C MT) . (B) CD19 transfer to CAR T c ells lacking an antig en bi nding d om ain (∆scF v) o r C D19 CA R T c ells (FM C63) afte r a 1- hou r c o cultu re wit h CD19-p ositi v e cell lin es at a 0.5:1 effect or-tar get ratio u sing flow cyt ometr y. (C) C D19 loss on tumo r cells after a 1-ho ur c oc ultu re wi th ∆scFv o r FM C63 CAR T c ells at a 0.5:1 effecto r-ta rget rati o usi ng flow c ytom etr y. (D) Obser vation of CMT in b rig htfield and c onfo cal i mages of ∆scF v or FM C63 CA R T cells c ultur ed w ith CD19-GFP-exp re ssing 293T c ells. CD19 is s ho wn i n g re en. Scal e b ar r epr es ents 5μm. (E-F) Amount of CD19 on CAR T cells is olated from patients 7-28 days after r ec ei ving CD19 CAR T cell th erap y us ing fu ll-spectr um flow cyt omet ry . Data indicate indi vidual patients. Statistical signif ican ce was determin ed by Wilc oxo n matched-pa irs sign ed- rank test. (G) Sc hema of trog oc yt osed antigen deg rade r (TAD). (H ) CD19 and (I) GFP expr ess ion on FM C63 CAR T c e lls expr essing TA D GFP or T cells expr essing a CA R lacking a binding d omain (∆s c Fv) follo wing a 1-h ou r co cultu r e with CD19-GFP- expr ess ing NAL M6 c ells as d et ermin ed b y flow c ytom etr y. (J) Sc hema of a full y human-de ri ved TAD s ystem (T AD CD19 ). VHL = v on Hipp el-Lin dau prote in; Ub = ubiquit in. (K) GFP and (L) CD19 express ion on F MC63 CAR T cells exp ress ing TAD CD19 or ∆s cF v CAR T cells a fter a 1-h ou r co cultu re with CD 19-GFP-expre ssing NALM6 c ells at a 0.5:1 effe cto r-target rati o. (M) Quantif icati on of CAR T c ells after a 24-hou r c oc ultur e with CD19-G FP-expr essi ng NALM6 cells us i ng flow c ytomet ry . Values a re n ormaliz ed to w ells c ontaining onl y CAR T cells. (N) E xpre ssi on of LAG-3 and TIM-3 on CAR T c ells exp re ssing TAD GFP after a 24-h ou r c oc ultur e with CD19- .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted June 14, 2024. ; https://doi.org/10.1101/2024.06.11.598379doi: bioRxiv preprint GFP-expr ess ing NAL M6 cells using f ull-spe ctr um flow c yto metry . ( H-N) Data indicate s mean ± S .D. fr om th r ee t ec hni cal r epli cates. Statisti c al signifi canc e was determ ined b y tw o-tailed Stud e nt’s t-test . (O) Sc he m a o f i n vi tr o se rial c oc ultur e expe rim ent to dete rmin e long-t er m CAR T c ell e xpansi on and anti tumor acti vity . (P) Sur vi val of C D19-GFP-exp res sing NALM6 c ells e xpr essing fi refl y l ucif eras e (Fl uc) at a 0.5:1 effe cto r-target rati o us ing a luminesc en ce-bas ed c ytoto x icit y assay. Data indicate s mean ± S .D . fr om thre e te ch nical repli cates . T u mor s ur vi val was normaliz ed to unt reated t umo r cells. Statisti cal signif ican ce was dete rmin ed by multiple t- tests and c or re cted f or m ultiple c ompar isons us ing t he B onf er ron i-Dun n

. Figure 2: Extracellular cathepsi n B is a key driver of CAR-med iated trogo cy tosis. (A) S ch ema of luc ife rase c om plementation as say ( CompL uc) . (B) CM T du ring co cultu re of F MC63 CAR T cell s expr essing nLu c with K562 c ells expr essing full length or tru ncated (tr19) CD19 f used to cL uc us ing CompLu c o r f low cyt ometr y aft er 1 hour at the indi cated eff ect or-t arget ratios . C MT usi ng CompLu c was measu red fo r a total 2.5 hou rs at 1-minut e in tervals . Data r epr es ents best f it of th re e te chni cal repli cates. (C) Ar ea-unde r- cu rv e quantificati on of l umine sc enc e i n Fig. 2B . (D) Peak lumines cen c e of FM C63 CAR T cells co cultu r ed wit h K562 c ells exp ress ing CD19- cLu c at a 1:1 effector-targ et ratio as determined b y CompL uc . CA R T cells we re pr e- treated wit h c yto chalasin D f or 1 hou r at the ind icated con c entrati ons. (E) S ur vi val of Raji-Flu c c ells after 16-h ou r c oc u lture w ith F MC63 CAR T cells p re-tr eated w ith th e indicated c onc entrati ons of c yto chalasin D . C MT valu es ar e no r malized to a DM SO .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted June 14, 2024. ; https://doi.org/10.1101/2024.06.11.598379doi: bioRxiv preprint cont rol. Data repr es ent mean ± S.D. of th re e tec hni cal repl icat es. (F) Sc hema of potential modulato rs of CM T. (G ) Peak luminesc en ce as det ermi ned by C ompLu c of FMC63 CAR T c ells treated with inhibit ors of th e indi cated p rote i ns co cult ur ed with K562 cells exp ress ing CD19- cLu c at a 1:1 effector-targ et rati o. Val ues ar e no rmaliz ed to a DM SO cont rol. Data rep res ent mean ± S .D. of th re e te chn ical repl icate s. (H ) Sur vi val of Raj i-Flu c c ells after 16-hou r c oc ultur e wit h FM C63 CAR T c ells treated with in hibit ors of t he i ndicat ed prote ins at a singl e c on cent rati on. T umo r su r vi val was meas ur ed u sing a luc ife ras e-based c ytoto xi city assay . Data rep re sent mean ± S.D . of thr ee te chn ical r eplicate s. (I) Peak luminesc en ce as dete r mined by C ompLu c of FM C63 CAR T c ells co cultu r ed wit h K562 c ells exp r essing CD19-cL uc at a 1:1 effe cto r-target rati o. FM C63 CAR T cells wer e pre-t reated with the indi cated con cent rations of Ca-074-Me (membrane pe rmeable) o r Ca-074 (membrane imperm eable). Value s are no rmalized t o a D MSO c ontr ol. Data rep res ent mean ± S. D. of th re e te chn ical repli cates . (J) Sch ema of mic ro sc opy set u p used to as sess catheps in B (CTS B) locali zation to the immune s ynapse . Bioti nylated CD19 was immobiliz ed on plate-bo und N e utrAv idin. F MC63 CAR T c ells expr essing C TSB- mChe rr y w er e add ed to slid es a nd imaged u sing conf ocal mi c ros cop y. (K) Sid e (x- z) vie w of the dist ributi on of CT SB-mC he rr y in ∆scF v or FM C6 3 CAR T cells on immobiliz ed CD19 at indi cated ti me points using spinning disk c o nfocal imaging . (L) Axial dispers ion of CTSB in fi xe d ∆scFv and FMC63 CAR T c el ls 5 minutes after plating. (M) Av erag e distan ce of CTSB to immun e s ynaps e i n fi xed ∆scF v and FM C63 CAR T cells 5 minut es and 15 minutes afte r plating. (M- N) Sta tistical signif ican ce was determi ned by tw o-tailed S tudent’s t-te st. C MT dur ing c oc ultur e of ∆scF v or .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted June 14, 2024. ; https://doi.org/10.1101/2024.06.11.598379doi: bioRxiv preprint FMC63 CAR T cells expr ess i ng nLuc w ith K562 cells exp ress ing C D19-cLu c measur ed b y C ompLu c and fl ow cyt ometr y. C MT using CompL uc was measu red fo r a total 2.5 hou rs at 1-min ute int e rvals. Fl ow cyt ometr y was pe rf ormed immediat ely following completi on of Comp Luc. Data repr es ents best fit of thr ee te chn ical repli cates. Figure 3: Expre ssion of h uma n cys tatins in creases CA R T cell expansion by reducin g CTSB-mediat ed C MT. (A) Cry stal stru ctu res of bo vi ne cat heps in B in comple x wit h Ca-074 (left, PDB : 1QDQ) and h uman cath epsi n B in compl ex w ith cystatin A ( rig ht, PD B: 3K9 M). T he a ctiv e sit e c yst eine of cathe psin B i s sh own in pink; Ca-074 and cystatin A ar e sh own in t eal. (B) S ch ema of const ru ct us ed fo r prod ucti on of CAR CSTA T cells. (C ) Levels of C STA in CAR or CAR CSTA T cells m easu red by ELISA. Data r epr ese nt mean ± S. D. of two te ch nical repli cate s. (D) Cat heps in B activit y in CAR o r CAR CSTA T cell s measur ed by flu or esc en ce . Data repr esent mean ± S.D . of tw o te chn ical r eplicat es. (E ) Expansion of CAR or CAR CSTA CAR T c ells du ring manufactur ing as det ermin ed b y c ell co unting. (F ) CAR T cells wer e analyz ed v ia flow c ytom etr y afte r pr odu cti on f or (I) CAR exp re ssi on, (II ) T c ell s ubsets , and (III-IV ) T cell ph enot ype . (G) S ur vi val of Raji-Flu c, NAL M6-Flu c, Daud i-F luc, or T oled o-Flu c cells after 16-ho ur c oc ultu re with FMC63 or CAR CSTA T cells at the indicated effe cto r- target rati os. Tumo r s ur vi val w as measu red using a l uc ife rase -based c ytot oxi city assay. Data r epr es ent mean ± S. D. of t hr ee te ch nical repli cates. ( H) CMT and (I) ar ea unde r c ur ve q uantifi cation as d e termin ed by CompL uc using K562 cells exp re ssing CD19-cL uc after 3 h ou rs. Data r e pres ent mean ± S. D. of t hr ee te c hnical repl icate s. (J) .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted June 14, 2024. ; https://doi.org/10.1101/2024.06.11.598379doi: bioRxiv preprint CD19 loss on tumo r cells, (K) CD 19 transfer to CAR T cells, and (L) total CAR T cells following a 30-minut e c oc ultur e of F MC63 o r CAR CSTA T c el ls and K562 cells expr ess ing C D19-cL uc at a 0.5 :1 eff ect or-targ et ratio. CAR T c ell numb ers ar e normaliz ed to wells c ontaining only CAR T cells us ing co u nting beads. Data rep res ent m ean ± S. D. of thr ee te chni cal repli cates . Figure 4: Cy statin o vere xpressi on improves long- term CA R T cell expansion and effica cy. (A) Sc hema of in v itr o exp er iment to d ete rmin e lon g-term CAR T c ell expansi on and antitum or acti vit y. (B) Quantifi catio n of CAR T cells d uri ng se rial co cultu re u sing flow c ytom etry . CAR T cells we re no rmaliz ed usi ng counting b eads. Data repr es ent mean ± S. D. of thre e te chn ical r epli cates. (C) Mean flu o res cen c e intensit y of CD69 in CAR T c ells after 24-ho ur c oc ultur e. Data rep r esent m ean ± S.D . of thre e tec hni cal repli cates. (D) Mean fluor es cen c e intensit y of Ki67 on Day 5 of ser ial co cultu re using flow c yto metry . (E) Su rv ival of C D19-GFP-expr ess ing NALM6 cells exp res sing l uc ife rase at a 0. 5:1 eff ect or-targ et ratio using a lu mines cen ce-bas ed cyt otoxi cit y assay . Data rep r ese nt mean ± S. D. of thr ee t ec hni cal repli cates . (F) Sch ema of in v ivo e xpe rime nt to measur e CAR T cell e xpansi on and persi sten ce . (G) Quantificati on of CAR T cells in murin e sple en, p er iph eral blo od, and bone mar ro w. Statistical s ignifi canc e was det er mined b y tw o-tailed Stud ent’s t-test. Figure 5: CSTA overe xpression reduc es CMT and increas es e xpansion in BCM A CAR T c ells. (A) Sc hema of va rio us CAR BCMA T cells. (B) BCMA transfer to CAR T c ells and BC MA loss on BC MA-GFP- expr ess ing K562 c ells. (C) B C M A loss o n M M .1S, .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted June 14, 2024. ; https://doi.org/10.1101/2024.06.11.598379doi: bioRxiv preprint RPMI8226, and U266B1 c ells wh en c o cultu red with CAR BCMA , Ab ecma, and Car v ykti CAR T c ells. (D ) Expansi on of BC MA CA R o r CAR CSTA C AR T cells d uring manufactur ing as det erm ined b y cell c ount ing. (E) B C MA CAR T c ells we re analy zed via flow c ytomet ry afte r prod ucti on for (I) CAR exp re ssi on, (II) T c ell subsets, and (III- IV) T cell ph enot ype . (F) Su rvi val of M M.1S- , RP MI8226-, and U26 6B1-Fluc c ells after 16-hou r c oc ultur e wit h BC MA C AR or CAR CSTA T cells at th e indi cated eff ect or-targ et ratios . Tumo r s ur vi val was measur ed us ing a lucif eras e-based cyt otoxi cit y assay . Data rep re sent m ean ± S .D . of thr ee te chn ical repli cates . (G) BC MA CAR T c ell expansi on after 16-ho ur c oc ultu re with th e indi cated tumor c ell lines at an 8: 1 effe cto r-target ratio. ( H) CMT and (I) area und er cu rv e quantif icat ion as det erm ined by C ompLu c usi ng K562 cells expr essi ng BC MA-GFP- cLu c after 3 h ou rs. Data rep res ent mean ± S. D. of th re e tec hni cal repli cates. (J ) BC MA los s on U266B1 tumor cells afte r 2- ho ur c oc ultur e w ith B CMA CAR o r CA R CSTA T c el ls. (K) BCMA-GFP transfer to CAR T cells, and (L) total CAR T cells followi ng a 30- minute c oc ultu re of BC MA CAR o r CAR CSTA T c ells and K562 cells expr ess ing CD19- cL uc at a 1:1 effe cto r- target ratio . CAR T c ell numbe rs are no rmaliz ed to w ells c ontainin g only CA R T c ells using c ounting beads . (H-L) Data repr esent mean ± S. D. of th re e tec hni cal replicat es. (G, I-L) Statistical s ignifi canc e w as dete rmin ed b y tw o-tailed Stud ent’s t-t est. .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted June 14, 2024. ; https://doi.org/10.1101/2024.06.11.598379doi: bioRxiv preprint .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted June 14, 2024. ; https://doi.org/10.1101/2024.06.11.598379doi: bioRxiv preprint .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted June 14, 2024. ; https://doi.org/10.1101/2024.06.11.598379doi: bioRxiv preprint .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted June 14, 2024. ; https://doi.org/10.1101/2024.06.11.598379doi: bioRxiv preprint .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted June 14, 2024. ; https://doi.org/10.1101/2024.06.11.598379doi: bioRxiv preprint .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted June 14, 2024. ; https://doi.org/10.1101/2024.06.11.598379doi: bioRxiv preprint .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted June 14, 2024. ; https://doi.org/10.1101/2024.06.11.598379doi: bioRxiv preprint