C57BL/6 mice (female, 8-11 weeks old) were acquired from Envigo. RAG^-/- mice and C57BL/6 mice on different congenic backgrounds (CD45.2^+/+, CD45.1^+/+, CD45.1^+/-, CD90.1^+/+, CD90.1^+/-) were derived from in-house breeding under specific pathogen-free conditions at our mouse facility at the Technical University Munich, Institute for Medical Microbiology, Immunology and Hygiene. The performed animal experiments were approved by the district government of Upper Bavaria (Department 5—Environment, Health and Consumer Protection ROB- 55.2-2532.Vet_02-17-138).

Murine splenocytes were cultured in RMPI (Life Technologies, Cat #31870074) with 10% fetal bovine serum (FCS) (Sigma, Cat #F7524), 0.025% L-glutamine (Sigma, Cat #G8540), 0.1% HEPES (Roth, Cat #HN77.4), 0.001% gentamycin (LifeTechnologies, Cat #15750037), 0.002% penicillin/streptomycin (LifeTechnologies, Cat #10378016) and 15 ng/ml recombinant human (rh) IL-15 (Peprotech, Cat # 200-15). The Platinum-E packaging cell line (PlatE) was cultured in DMEM (Life Technologies, Cat #10938025) supplemented with 10% FCS, 0.025% L-glutamine, 0.1% HEPES, 0.001% gentamycin, 0.002% streptomycin. All cells were grown in a humidified incubator at 37°C and 5% CO2.

scFvs were generated by fusing the variable regions of the heavy (VH) and light (VL) chains with a short (G4S)₃ linker. For CAR constructs, the signal peptide of CD8α was followed by the scFv, a hinge domain (a triple repetitive sequence of Strep-tag II (STII) for anti-CD19 CAR and a CD8a spacer domain for the anti-CAR CAR) and parts of the IgG4-Fc molecule. This extracellular domain was followed by a transmembrane region originated from the CD28 chain and the intracellular signaling domains CD3-zeta. A truncated version of the EGFR (EGFRt) was added at the 5’ and separated from the CAR sequence by a self-splicing viral P2A element. As control, a construct with the same elements as the anti-CD19 CAR but without the scFV (defined as ‘mock’) was generated. DNA templates were designed in silico and synthesized by GeneArt (Thermo Fisher Scientific) or Twist Bioscience into pMP72 vectors. scFv sequence of anti-CAR CAR (clone: 5G2) were kindly provided by the lab of Prof. Riddell.

For the production of retroviral particles, PlatE cells were transiently transfected with a pMP72 expression vector by calcium phosphate precipitation. For this, 15 µl of a 3.31 M MgCl₂ was mixed with 18 µg vector DNA and filled up to 150 µl with ddH₂O. This solution was slowly added under vortexing to 150 µl transfection buffer (274 mM NaCl, 9.9 mM KCl, 3.5 mM Na₂HPO₄ and 41.9 mM HEPES), and the final transfection mix was added to PlatE cells for 6 h at 37°C, followed by a complete medium exchange. The virus-containing supernatant was harvested 48 and 72 h after transfection by filtration (0,45 µm sterile filter) and stored at 4°C for up to one week. Mouse splenocytes were obtained from CD45/CD90 congenic mice by mashing the spleen into single-cell suspensions. After red cell lysis, splenocytes were stimulated at a concentration of 10⁷ cells/mL with purified anti-mouse CD3 (BD Biosciences, clone 145-2C11, 1:1000, Cat #553058) and anti-mouse CD28 (BD Biosciences, clone 37.51, 1:3000, BD Cat #553295) antibodies in presence of 25 U/ml recombinant human interleukin-2 (Peprotech, Cat #200-02). For anti-CAR CAR transduction, splenocytes were previously depleted of B cells by magnetic purification (Miltenyi Biotec, Cat #130-048-701) using an anti-mouse CD19-FITC antibody (BD Biosciences, clone 1D3, 1:100, Cat #553785) according to the manufacturer’s protocol. After 24 h, stimulated splenocytes were collected and transduced via spinoculation. For this, tissue-culture untreated 24-well plates were coated overnight with retronectin (TaKaRa, Cat #T100B), purified anti-mouse CD3 (1:1000) and anti-mouse CD28 (1:3000) antibodies. The virus-containing supernatant was centrifuged onto the plates at 3000 g at 32°C for 2 h. After removal of the supernatant, cells were transferred and centrifuged at 800 g at 32°C for 1.5 h. Cells were then rested for two days before administration to the mice.

For in vitro assays, cells were harvested from the culture and washed twice with FACS buffer (1x PBS, 0.5% (w/v) bovine serum albumin (BSA), pH 7.45) before staining. For blood analysis, blood was sampled from the tail vein and stored in heparin at room temperature. Bone marrow was acquired from the hind legs by removal of muscles and rinsing with complete RPMI medium whereas lymph nodes and spleens were homogenized into single-cell suspension by mashing. Red cell lysis was performed for single-cell suspensions from all organs and blood with RBC lysis buffer (90% 0.17M NH4Cl in Tris-HCl). After lysis, cells were first stained with anti-CD16/32 antibody (Biolegend, clone 93, 1:400, Cat #101301) for 20 min at 4°C (only cells from lymphoid tissues), followed by surface marker staining for 20 min at 4°C in the dark. In this study we used the following antibodies for surface staining: CD3-APC (Biolegend, clone 145-2C11, 1:100, Cat #100312), CD3-FITC (Life Technologies, clone 17A2, 1:100, Cat #11003282), CD8-PacO (Life Technologies, clone 5H10, 1:50, Cat #MCD0830), CD19-ECD (BD Bioscience, clone 1D3, 1:300, Cat #562291), CD45.1-PacBlue (Life Technologies, clone A20, 1:100, Cat #48045382), CD45.1-APC (Life Technologies, clone A20, 1:100, Cat #17045382), CD90.1-PacBlue (Life Technologies, clone HIS51, 1:500, Cat #48-0900-82), CD90.2-APC (Life Technologies, clone 53-2.1, 1:500, Cat #17090283), EGFR-PE (Biolegend, clone AY13, 1:2000, Cat #352904), PD-1-PE-Cy7 (Life Technologies, clone J43, 1:100, Cat #25-9985-80), TIM3-BV421 (Biolegend, clone RMT3-23, 1:100, Cat #119723), CD69-BV510 (Biolegend, clone H1-2F3, 1:100, Cat #104531). Live/dead discrimination was performed either by addition of propidium iodide (LifeTechnologies, 1:100, Cat #P1304MP) 3 min before the end of the staining or with ethidium-monoazide-bromide (LifeTechnologies, 1:1000, Cat #E1374) for 15 min at 4°C under bright light. Absolute quantification was performed by using 123count eBeads (Life Technologies, Cat #01-1234-42) according to the manufacturer’s instructions. Specific information is further provided for each experiment. Samples were acquired on a CytoFLEX S flow cytometer (Beckman Coulter). All flow cytometry data were analyzed with FlowJo v10.

5x10⁴ effector anti-CAR CAR-T cells (EGFRt⁺) were co-cultured with target anti-CD19 CAR-T cells (EGFRt⁺) at different effector-to-target ratios in 96-well plates. After 1 h co-incubation, 1x GolgiPlug was added (BD PharMingen, Cat #555029) and cells were incubated for additional 4 h at 37°C. 25 ng/ml Phorbol-12-myristat-13-acetate (Sigma, Cat #P1585) and 1 μg/ml ionomycin (Sigma, Cat #I9657) were used as positive controls, while untransduced cells were used as negative controls. After co-culture, staining for live/dead discrimination was performed with ethidium-monoazide-bromide as described before, followed by surface marker antibody staining for congenic markers (CD45.1, CD45.2) and EGFRt for 20 min at 4°C. Cells were permeabilized using Cytofix/Cytoperm (BD Biosciences, reference #554714) and stained intracellularly for TNF-α-PE-Cy7 (BD Biosciences, clone MP6-XT22, 1:100, Cat #557644), INF-γ-APC (LifeTech, clone XGM1.2, 1:400, Cat #17-7311-82) and IL-2 (LifeTech, clone JES6-5H4, 1:100, Cat #12-7021-82). As bulk CAR-T cell products were used, the number of transferred cells always refers to the amount of congenic marker⁺EGFRt⁺ cells. The total amount of transferred cells was defined by flow cytometry according to the expression of the EGFRt marker.

4x10⁴ effector anti-CAR CAR-T cells (EGFRt⁺) were co-cultured with target anti-CD19 CAR-T cells (EGFRt⁺) at different effector to target ratios in 96-well U-bottom plate at a concentration of 7.5x10⁵ cells/mL for 48 h. At the beginning of the incubation (0 h), and after 24 and 48 h, 200 µL of the cell suspension was harvested for surface marker antibody staining of congenic markers (CD90.1, CD45.1), CD3, CD8, CD19 and EGFRt. Staining for live/dead discrimination was performed with propidium iodide. As bulk CAR-T cell products were used, the number of transferred cells always refers to the amount of congenic marker⁺ EGFRt⁺ cells. The total amount of transferred cells was defined by flow cytometry according to the expression of the EGFRt marker.

Transduced mouse splenocytes were rested for 5 days or activated with anti-CD3 (1:1000) and anti-CD28 (1:3000) antibodies for 24 h prior analyses. Cell lysates were generated by addition of 200 µl 1X SDS sample buffer (62.5 mM Tris pH 6.8, 2% w/v SDS, 10% Glycerol, 0.01% Bromophenol blue, 50 mM DTT) to 2x10⁶ cells, followed by sonication for 10 min and incubation at 95°C for 5 min. The solution was centrifuged for 5 min at 13.000 g, and 15 µl of the supernatant were run on an 8% SDS-PAGE gel (stacking gel: TRIS 0.5 M pH 6.5, 0.4% SDS, 40% Acrylamide, 10% APS TEMED; separating gel: Tris 1.5 M pH 8.8, 0.4% SDS, 40% Acrylamide, 10% APS TEMED) for 1 h at 150 V. The proteins were semi-dry blotted onto a nitrocellulose membrane at 10 V for 30 min. The efficiency of the transfer was evaluated via Ponceau staining (0.5% Ponceau, 1% acetic acid). The membrane was blocked in Tris-buffered saline containing 0.1% Tween (TBS-T buffer) with 5% nonfat dry milk (blocking buffer) for 1 h at room temperature. Afterwards the membrane was stained with anti-pCD3ζ antibody (Life Technologies, clone EM54, 1:3000, Cat #PIMA528538) in TBS-T buffer containing 5% BSA at 4°C. As loading control, GAPDH was stained with anti-GAPDH antibody (Thermo Fisher Scientific, clone 1D4, 1:3000, Cat #MA1-16757). After 12 h, the membrane was washed in TBS-T buffer and incubated with a horseradish peroxidase-conjugated anti-mouse secondary antibody (Cell Signaling, 1:1000, Cat #7076S) in blocking buffer for 1 h at room temperature. After washing, the membrane was incubated with 200 µl enhanced chemiluminescence solution (Biorad, Cat #1705061) for 5 min at room temperature and analyzed with the INTAS chemoluminescence detection system.

For the co-injection model, Rag^-/- or C57BL/6 mice were sublethally irradiated (5 Gy) the day before the simultaneous adoptive transfer of 1x10⁶ anti-CD19 CAR-T cells and 2x10⁶ anti-CAR CAR-T cells by intravenous injection (i.v.). Transferred cells and B cells were followed in the blood at different time points and in bone marrow, spleen and lymph nodes at the endpoint.

For the therapeutic model, 2x10⁶ anti-CD19 CAR-T cells were transferred i.v. into sublethally irradiated mice (5 Gy). After 27 days, mice were either or not irradiated with a 2 Gy dose and treated the day after i.v. with 4x10⁶ or 8x10⁶ anti-CAR CAR-T cells, or 4x10⁶ mock T cells (CAR construct without the scFv). Moreover, one group received the monoclonal antibody Cetuximab (1mg per mouse) (Bristol-Myers Squibb) by intraperitoneal injection. Transferred cells and B cells were followed in the blood at different time points and in bone marrow, spleen and lymph nodes at the endpoint.

As bulk CAR-T cell products were used, the number of transferred cells always refers to the number of congenic marker⁺EGFRt⁺ cells. The total amount of transferred cells was defined by flow cytometry according to the expression of the EGFRt marker. The specific congenic marker combinations are indicated for each experiment.

Data were visualized using GraphPad Prism (V9.1). All statistical tests were performed using GraphPad Prism (V9.1). Specific details regarding statistical analyses can be found in the corresponding figure legends. Levels of significance were defined as the following: ns = p > 0.05, *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001, ****p ≤ 0.0001. Data are presented as mean ± standard deviation or mean + standard deviation, as indicated in figure legends.

In this study, we used a fully syngeneic second-generation anti-CD19 CAR as the primary ‘therapeutic’ CAR. We further included in the hinge region a sequence containing a three-time in tandem repetition of an 8 amino acid peptide, named Strep-tag II (27). Previous investigations have demonstrated that the inclusion of this triple Strep-tag II in the CAR construct does not compromise the functionality of the corresponding engineered T cells (28). To target anti-CD19 CAR-T cells, we generated a second CAR construct where the scFv could specifically recognize the Strep-tag II. This CAR, referred to as ‘anti-CAR’ CAR from now on, was additionally equipped with the same CD3zeta/CD28 intracellular domains and truncated human EGFR (23, 24) as the anti-CD19 CAR ( Figure 1A ). This engineering approach theoretically enables the depletion of fully functional anti-CD19 CAR-T cells by the second anti-CAR CAR-T cells ( Figure 1B ). For appropriate discrimination, the two CARs were always re-expressed in splenocytes from mice harboring distinct CD45/CD90 congenic marker combinations. Furthermore, B cells were depleted from spleen-derived cell suspensions before viral transduction to avoid stimulation of anti-CD19 CAR-T cells prior any in vitro or in vivo assay. Both receptors could be successfully expressed in primary murine splenocytes through retroviral gene delivery and showed comparable surface expression levels ( Figure 1C ).

To demonstrate the in vitro functionality of the newly designed anti-CAR CAR-T cells, we first examined intracellular cytokine release upon target-specific stimulation during co-culture with Strep-tag II-expressing anti-CD19 CAR-T cells. The anti-CAR CAR-T cells showed reliable cytokine production (IFN-γ, TNF-α and IL-2) in a dose-dependent manner in presence of the target cells. Moreover, there was little to no unspecific signaling when co-cultured with untransduced control cells ( Figures 1D, E ). For evaluating cytotoxic capacity, we again co-cultured anti-CD19 CAR-T cells with anti-CAR CAR-T cells at different effector-to-target (E:T) ratios and monitored the absolute number of each T cell population over 48 hours. By using bulk populations, we could also assess the specificity of killing by tracking the survival of unedited (EGFRt^-) cells. Importantly, while the untransduced control cells without antigen expression, as well as the effector cells, proliferated over time, we observed a significant decrease of the target anti-CD19 CAR-T cells ( Figures 1F, G ). Altogether, these data demonstrate the specific target recognition of our anti-CAR CAR-T cells, which ultimately translates into effective cytotoxic T-cell functions.

Finally, we investigated whether anti-CAR CAR-T cells showed any tonic signaling activity. For this purpose, we firstly analyzed the baseline phosphorylation of CD3zeta in rested splenocytes, using freshly restimulated cells as control. While we observed elevated levels of pCD3z in restimulated anti-CAR CAR-T cells, we did not detect any level exceeding those found in control cells in rested splenocytes ( Supplementary Figure 1A ). To confirm this phenotype, we also assessed the persistence and activation state of adoptively transferred anti-CAR CAR-T cells under homeostatic conditions ( Supplementary Figure 1B ). We observed overall similar numbers of circulating anti-CAR CAR-T cells compared to untransduced mock control cells ( Supplementary Figure 1C ), and, as expected, no influence on the B-cell compartment except for the drop in the first week due to the pre-conditioning regimen prior T-cell transfer ( Supplementary Figure 1D ). Furthermore, we did not detect neither consistent upregulation of activation markers (CD69 and PD-1) nor exhausted cells (PD-1⁺TIM3⁺) ( Supplementary Figures 1E, F ). These observations were confirmed also in anti-CAR CAR-T cells infiltrating secondary lymphoid organs, except from the detection of a small fraction of CD69⁺ cells in the bone marrow ( Supplementary Figure 1G ). Altogether, we could not find robust indications of tonic signaling.

Our in vitro data indicated a certain degree of functionality for the designed anti-CAR receptor. However, such evidence could not definitively predict the in vivo functionality, as this relies on additional factors such as the overall antigen load, target cell accessibility and the receptor moiety itself, among others (29, 30).

To assess the in vivo functionality of our anti-CAR construct with minimal confounders, we first conducted a short-term in vivo killing assay in the imunocompromised RAG1^-/- mouse model, which lacks mature T cells. As before, we generated anti-CD19 and anti-CAR CAR-T cells from different congenically labelled splenocytes, which were simultaneously injected in pre-conditioned mice. We monitored the adoptively transferred cells in the periphery through blood analysis for 6 days and in tissues on day 10 ( Figure 2A ). While anti-CD19 CAR-T cells were reliably detected in the blood when transferred alone, we could not detect circulating anti-CD19 CAR-T cells when both T-cell products were co-transferred. This observation was consistent in secondary lymphoid tissues ( Figures 2B, C ; Supplementary Figure 2 ), suggesting infiltration properties that are expected from a cell-based mechanism.

As the next step, we setup up a similar experiment using an immunocompetent wild type C57BL/6 mouse model as a recipient. We also included an additional control group treated solely with anti-CAR CAR-T cells ( Figure 3A ; Supplementary Figure 3A ). In this model, the administration of anti-CD19 CAR-T cells induces long-lasting B-cell aplasia (31), similar to what reported in patients receiving anti-CD19 CAR therapy (5, 32). Thus, this syngeneic model allowed us to monitor the efficacy of anti-CAR CAR-T cells in terms of depleting the anti-CD19 CAR-T cells as well as reconstituting the B-cell compartment eliminated by the primary therapeutic CAR-T cells (24). Moreover, this model presented challenges more relevant to potential applications in humans, such as the competition with other immune cell populations and, most importantly, a stimulus for the primary anti-CD19 CAR, which was still expected to functionally recognize B cells.

Looking at the dynamics of anti-CD19 CAR-T cells in the blood ( Supplementary Figure 3B ), we observed a well-known pattern of expansion and contraction within the first two weeks in mice treated solely with the primary CAR-T cells ( Figures 3B, C ; Supplementary Figure 3C ). This phenomenon is commonly observed in CD19⁺ preclinical tumor models as well as in patients receiving anti-CD19 CAR-T cells, and, in our model, was triggered by the endogenous B cells. In sharp contrast, as anticipated from our previous data, this kinetic did not occur in presence of the anti-CAR CAR-T cells. Moreover, we did not detect anti-CD19 CAR-T cells throughout the observation period, further indicating the persistent functionality of the treatment over time ( Figures 3B, C ; Supplementary Figure 3C ). While we observed peripheral expansion and contraction of anti-CAR CAR-T cells in mice receiving both CAR-T cell types, this occurred to a much lesser extent. The high circulating levels of the anti-CAR CAR-T cells when transferred alone, instead, were a sign of T cells with no specificity in the used experimental model ( Figures 3B–D ; Supplementary Figure 3C ). In secondary lymphoid tissues, we found sustained depletion of anti-CD19 CAR-T cells, despite anti-CAR CAR-T cells were barely detectable ( Figure 3E ).

We finally investigated the endogenous B-cell compartment, which represents the off-tumor target of our primary CAR-T cells. In the anti-CAR-treated control group, B cells recovered within seven days post irradiation, while mice receiving anti-CD19 CAR-T cells experienced long-term aplasia, confirming that the addition of the Strep-tag in the hinge region did not affect functionality. Importantly, in mice receiving the combination of the two CAR-engineered T cells, B cells reconstituted early at lower frequency and eventually reached levels comparable to our control group approximately six weeks after infusion ( Figures 3F, G ).

In summary, our findings show that anti-CAR CAR-T cells can effectively deplete primary CAR-T cells in an immunocompetent host, leading to the reversal of the unwanted long-term B-cell aplasia induced by anti-CD19 CAR-T cells as a side effect.

Our previous data proved the functionality and reliability of the anti-CAR approach in a model of simultaneous target engagement. Next, we sought to investigate whether it could effectively reverse a long-term side effect caused by a primary therapeutic CAR-T cell product. To address this question, we induced stable B-cell aplasia in wild-type immunocompetent C57BL/6 mice by administering anti-CD19 CAR-T cells, and then transferred anti-CAR CAR-T cells four weeks later ( Figure 4A ). Achieving successful engraftment of transferred T cells usually requires lymphodepletion, a step commonly used in clinical practice to enhance the effectiveness of CAR-based therapy in patients (33). Nevertheless, consenting patients for a second round of pre-conditioning, in particular in cases of disease remission, can be challenging. Therefore, we compared the efficiency of anti-CAR CAR-T cell transfer with and without pre-conditioning, using different cell doses. Additionally, we included a CAR construct without the scFv (mock) as a negative control, and the monoclonal anti-EGFR Cetuximab as a positive control, known for depleting adoptively transferred EGFRt-expressing CAR-T cells and promoting B-cell reconstitution (24) ( Figure 4A ; Supplementary Figure 4A ).

All mice showed B-cell aplasia until 21 days from anti-CD19 CAR-T cell transfer, which was the last time point of monitoring prior to the second T-cell transfer. Mice receiving anti-CAR CAR-T cells exhibited reliable B-cell reconstitution as early as two weeks post-administration, reaching levels comparable to Cetuximab-treated mice ( Figures 4B, C ). Noteworthy, there were no significant differences among the four anti-CAR-treated groups, except in mice receiving the lower T-cell dose without pre-conditioning. However, despite a slower reconstitution kinetic, B-cell frequencies in this group eventually reached comparable levels with the other groups at later time points. Importantly, mice receiving the control mock CAR-T cells maintained stable B-cell aplasia throughout the observation period also in combination with the second pre-conditioning, except for one mouse ( Figures 4B, C ). This observation suggested that anti-CD19 CAR-T cells reliably persisted and remained functional after irradiation. Moreover, it also attributed the B-cell reconstitution to the specific depletion of anti-CD19 CAR-T cells by the secondary safeguard CAR-T cells. This was further supported by the clear presence of anti-CD19 CAR-T cells in mice receiving the mock CAR-T cells at both early (day 21) and late (day 70) time points, regardless of irradiation, while anti-CAR-treated mice showed no detectable anti-CD19 CAR-T cells ( Figure 4D ; Supplementary Figure 4B ).

Similarly to what observed in the peripheral circulation, tissue analysis revealed high B-cell levels and undetectable anti-CD19 CAR-T cells also in the spleen, lymph nodes and bone marrow of anti-CAR CAR-T cell treated mice. Intriguingly, the group receiving low-dose anti-CAR without pre-conditioning showed more variable B-cell levels and persisting infiltrating anti-CD19 CAR-T cells ( Figures 4E, F ), despite no detectable B cells in the blood.

We finally examined the persistence of the anti-CAR CAR-T cells and observed a clear relationship between the cell dose and pre-conditioning. As expected, irradiation facilitated engraftment of the transferred cells regardless of cell doses. Indeed, circulating anti-CAR CAR-T cells remained detectable only in irradiated mice ( Figures 5A, B ), with the highest frequencies observed in mice receiving 8x10⁶ cells plus irradiation in both blood and tissues ( Figure 5 ). Intriguingly, the higher cell dose compensated for the absence of pre-conditioning, at least during the initial weeks in the circulation, as indicated by comparable levels between the groups 8x10⁶ cells vs 4x10⁶ cells plus irradiation ( Figure 5B ). Although the amount of anti-CAR CAR-T cells declined to undeletable levels in the blood of mice receiving this high cell dose at later time points, tissue infiltration remained consistent and comparable to the other groups ( Figure 5C ). Taking into account the similar kinetics of B-cell reconstitution and complete anti-CD19 CAR-T cell depletion, we could conclude that this condition (8x10⁶ cells, no irradiation) would be still efficient.

In summary, our findings demonstrate that administration of anti-CAR CAR-T cells can reverse established B-cell aplasia in a model simulating a possible therapeutic application, potentially eliminating the need for pre-conditioning before T-cell transfer.