Tonic Signaling and Its Effects on Lymphopoiesis of CAR-Armed Hematopoietic Stem and Progenitor Cells

T cells expressing chimeric Ag receptors (CARs) represent a promising tool for cancer therapy. In particular, treatment of hematological malignancies via CAR T cells shows impressive antitumor responses, which is underlined by the Food and Drug Administration approval of two CD19-specific CAR constructs in 2017 (1, 2). CARs are artificial fusion molecules that consist of an extracellular Ag-binding domain, mostly a single-chain fragment variable (scFv), a transmembrane region, and cytoplasmic T cell signaling domains. In contrast to first generation CARs, second or third generation CARs contain costimulatory signaling domains (e.g., CD28, CD137), which improve cytotoxicity, expansion, and in vivo persistence (3–6). Together with high cytolytic activity, long-term survival of CAR T cells is particularly essential for positive outcomes (7, 8). Besides immunogenicity and the choice of costimulatory domains, several groups demonstrated that basal/tonic signaling is also of central importance for CAR T cell persistence (9–11). Based on the available scientific data, tonic signaling can be defined as a constitutive or chronic activation of T cells in the absence of a ligand (12). Physiologically, low-level continuous tonic signaling via interactions between the endogenous TCR and self-peptide–loaded MHC molecules constitutes an important mechanism to regulate T cell homeostasis (13). In contrast, tonic signaling mediated by T cell–engrafted CAR constructs appears to be more complex. According to recent studies, varying degrees of Ag-independent tonic activation are frequently associated with second-generation CAR T cells (e.g., anti-GD2, anti-CD22, anti–HER-2, anti–c-Met) (10, 11). Tonic signaling can be dependent or independent of the relevant Ag and has major effects on CAR-based therapies. On the one hand, CAR tonic signaling may promote T cell expansion by providing stimulating signals (14). On the other hand, CAR tonic signaling can trigger terminal effector T cell differentiation, exhaustion and/or enhanced activation-induced cell death, and, therefore, limits in vivo persistence as well as antitumor potential (10, 11).

In contrast to finally differentiated T cells, hematopoietic stem cells (HSCs) are highly regenerative and possess a life-long potential to replenish all hematopoietic subpopulations. Thus, CAR-transduced HSCs should enable a rapid and long-lasting supply of Ag-specific effector cells, which can be not only T cells but also NK cells and myeloid cells (15–17). Therefore, such a multilineage immunotherapy in combination with the sustained replenishment of anergic or exhausted effector CAR T cells could even be superior over ex vivo–expanded CAR T cells (17). Furthermore, integration of CAR molecules into HSCs could be combined with classical HSC transplantation, which is used for treatment of hematological malignancies, CNS tumors, or sarcoma (15). Indeed, De Oliveira et al. (16) proved that infusion of human HSCs carrying a second-generation anti-CD19 CAR into immunodeficient NSG mice leads to functional CAR T cells, B cells, NK cells, and myeloid cells. These results were further confirmed by Larson and colleagues (15). However, Dolnikov’s group observed a suppression of CAR B cell development (18). Moreover, expression of an HIV-specific CD3ζ CAR construct in murine (mu) HSCs and its ligand-driven stimulation during hematopoiesis led to the arrest of T cell development (19, 20). To learn whether these controversial data could be related to tonic signaling of the respective CAR constructs, we decided to investigate the effect of tonic signaling on hematopoiesis in more detail.

For this purpose, we looked for two CAR constructs: one showing a strong Ag-independent tonic signaling and the other one lacking the signaling domains. Both CAR constructs were then transduced into mu bone marrow (BM) cells. CAR⁺ HSCs successfully were engrafted in lethally irradiated mice and differentiated into hematopoietic progenitors. Although mice transplanted with HSCs expressing CARs lacking signaling domains showed a normal hematopoiesis, mice transplanted with HSCs expressing the signaling CAR construct lacked both mature B and T cells.

The second-generation CAR construct, CAR-28/ζ, consists of an extracellular humanized scFv derived from the E5B9-specific mAb anti-La (clone E5B9) (21). The following hinge, transmembrane, and costimulatory signaling domain originate from the mu CD28 domain. Additionally, the construct carries at the C terminus the cytoplasmic signaling domains of the mu CD3ζ molecule. In contrast to this, the CAR-Stop construct lacks all signaling domains. To detect the extracellular CAR expression, the peptide epitope E7B6 of the human La/SS-B protein (22, 23) was inserted between the extracellular scFv and the hinge region. Cloning of the respective molecules was based on a cloning vector encoding the scFv anti-La E5B9 and the vector pEX_E7B6-mu 28/ζ or pEX_E7B6-mu 28 Stop synthesized and purchased from the company Eurofins Genomics (Ebersberg, Germany). The CAR-28/ζ as well as the CAR-Stop open reading frames were subsequently cloned into the lentiviral vector backbone p6NST60 (24). Thereby, the respective constructs were fused to the independently translated enhanced GFP (EGFP) (24). EGFP served as a marker protein to identify successfully transduced cells (22, 24, 25).

The anti-muCD19 target module (TM) consists of a CD19-specific rat-derived scFv (clone 1D3) and a peptide epitope of the human La/SS-B protein (E5B9) (21, 26). The cloning strategy of the transduction vector was similar to an already described procedure (27). For that purpose, the respective gene NotI–anti-muCD19 (V_LV_H)-XhoI was synthesized by the company Eurofins Genomics. To permanently produce the anti-muCD19 TM, Chinese hamster ovary cells were transduced with the generated lentiviral vector p6NST50_anti-muCD19 as previously described (28, 29). Subsequently, the permanently expressed anti-muCD19 TM was purified and analyzed by SDS-PAGE as published before (21, 30, 31).

The human embryonic kidney cell line HEK293T (ATCC CRL-11268) was cultured in complete DMEM (28). For virus particle production, cells were kept in StemPro-34 SFM (1×) medium (Thermo Fisher Scientific, Schwerte, Germany). The TM-producing cell line CHO (ATCC CCL-61) was grown in complete RPMI 1640 medium (Biochrom, Berlin, Germany) (28). The C57BL/6J-derived mu pre-B cell line RW7938, which was kindly provided by the group of Prof. Roers (Institute of Immunology, Medical Faculty Carl Gustav Carus, Technical University Dresden) (32), was kept in complete RPMI 1640 supplemented with 50 μM 2-ME. All cells were cultured at 37°C in a humidified atmosphere (5% CO₂).

All animal experiments were performed with C57BL/6J mice (Janvier, France), which were kept under specific pathogen-free conditions at the Experimental Center of Technical University Dresden according to the institutional guidelines of German Regulations for Animal Welfare. The study was approved by the Landesdirektion Sachsen (24-9168.11-1/2014-10).

BM cells were isolated by crushing long bones using mortar and pestle. Cells were flushed with RPMI 1640 medium containing 2% FCS (Biochrom) and 2 mM EDTA (VWR International, Darmstadt, Germany) and filtered through a 40-μm mesh. After centrifugation (360 × g, 4°C, 8 min) and erythrocyte lysis using ACK Lysing Buffer (Thermo Fisher Scientific), filtration of the BM cells was repeated.

Peripheral blood was drawn by retrobulbar puncture using glass microcapillaries and mixed with PBS/heparin (500 U/ml; Biochrom). To lyse erythrocytes, cells were incubated twice with ACK Lysing Buffer for 5 min.

Thymocyte suspensions were obtained by rubbing the respective organ through a 40-μm cell strainer. After erythrocyte lysis, cells from the thymus were washed with RPMI 1640/2% FCS/2 mM EDTA and filtered through a 40-μm mesh.

BM cells (5 × 10⁶ cells), peripheral blood cells (PBCs) (2 × 10⁶ cells), or thymocytes (2 × 10⁶ cells) were stained at 4°C with various Ab mixtures to identify different subpopulations. Staining procedure was done as described before (28, 33). mAbs directed against CD3 (145-2C11 or 17A2), CD4 (GK1.5), CD8 (53-6.7), CD11b (M1/70), CD19 (eBio1D3), CD25 (PC61.5), CD44 (IM7), CD45R (B220, RA3-6B2), CD49b (DX5), CD93 (AA4.1), CD117 (c-Kit, 2B8), CD127 (IL-7Rα, A7R34), CD135 (Flk2, A2F10), IgM (11/41), Ly-6A/E (Sca-1, D7), Ly-6G (Gr1, RB6-8C5), NK1.1 (PK136), and TER-119 (TER-119) were purchased from eBioscience (Thermo Fisher Scientific) and were either fluorochrome conjugated or biotin labeled. Depending on the subpopulation of interest, different lineage (Lin) Ab mixtures were used to distinguish Lin⁻ from Lin⁺ cells. The Lin mixture used for the isolation of Lin⁻Sca-1⁺cKit⁺ (LSK) cells or Lin⁻cKit⁺ (LK) cells contained mAbs recognizing CD3, CD11b, B220, Gr1, and TER-119. To stain common lymphoid progenitors (CLPs), a mixture of mAbs against CD3, CD4, CD8, CD11b, CD19, B220, Gr1, NK1.1, and TER-119 was generated. In comparison, the Lin Ab mixture for analysis of thymocytes was lacking mAbs against CD3, CD4, and CD8 but included mAbs directed against F4/80 (BM8) and CD11c (N418). In the case of biotin-labeled molecules, cells were detected using Streptavidin-V500 acquired from BD Biosciences/BD Pharmingen (Heidelberg, Germany) or Streptavidin-PE/Cy7 (BioLegend, London, U.K.). Detection of CAR surface expression was performed via the CAR-included peptide epitope E7B6. Therefore, T cells were incubated with the anti-La 7B6 mAb and subsequently stained with a fluorochrome-labeled goat anti-mouse IgG detection Ab (Beckman Coulter, Krefeld, Germany). For tonic signaling analysis, mAbs against human CD69 (REA824), PD-1 (PD1.3.1.3), and LAG-3 (REA351) were purchased from Miltenyi Biotec (Bergisch Gladbach, Germany). LIVE/DEAD cell discrimination was performed using propidium iodide ([PI] Thermo Fisher Scientific) or DAPI (Thermo Fisher Scientific) (24, 26, 34). Stained cells were analyzed on a MACSQuant Analyzer (Miltenyi Biotec) or an LSR II (BD Biosciences/BD Pharmingen), and data were evaluated by using MACSQuantify (Miltenyi Biotec) or FlowJo software (BD Biosciences/BD Pharmingen).

Primary (1°) human T cells were isolated from PBMCs from buffy coats as previously described (29). After using a human Pan T cell isolation kit (Miltenyi Biotec), T cells were kept in complete RPMI medium supplemented with 50 U/ml IL-2 (ImmunoTools, Friesoythe, Germany). Subsequently, T cells were transduced with the lentiviral vector p6NST60_CAR-28/ζ or with p6NST60_CAR-Stop. Generation of lentiviral particles and transduction of T cells was performed as recently published (24, 35). During modification, lentiviral supernatant was added at a multiplicity of infection (MOI) of 1–3. After transduction, genetically modified cells expressing the marker gene EGFP were purified using a FACSAria III (BD Biosciences/BD Pharmingen). During T cell transduction procedure and following cultivation, T cells were cultured in complete RPMI medium supplemented with 200 U/ml IL-2 (Proleukin S; Novartis Pharmaceuticals, Horsham, U.K.), 5 ng/ml IL-7, and 5 ng/ml IL-15 (ImmunoTools). One day before experiments, expansion medium was substituted for complete RPMI medium without any recombinant cytokines.

Prior to gene expression analysis, resting EGFP⁺PI⁻ CAR-transduced T cells or PI⁻ control T cells (500 cells per tube) were sorted in triplicates into eight-well PCR strips containing 5 μl of elution buffer (QIAGEN, Hilden, Germany) with 0.3% IGEPAL and 0.1% BSA using a FACSAria Fusion (BD Biosciences/BD Pharmingen) and quickly frozen on dry ice. Gene expression profile of cells was obtained as described previously (36) with some modifications. Briefly, after the thawing of cells by 10-min incubation on ice, cDNA was synthesized directly from cells applying qScript cDNA SuperMix (Quanta BioSciences, Gaithersburg, MD) by incubating 10 min at 25°C, 60 min at 42°C, and 5 min at 85°C. Total cDNA was preamplified for 18 cycles (1 × 95°C for 5 min, 18 × (95°C for 45 s, 60°C for 1 min, and 72°C for 1.5 min, and 1 × 72°C for 7 min) using the TATAA PreAmp GrandMaster Mix (TATAA Biocenter, Göteborg, Sweden) in a total reaction volume of 35 μl in the presence of primer pairs (25 nM for each primer) for the analyzed genes (listed in Supplemental Table I). Subsequently, preamplified cDNA (10 μl) was treated with 1.2 U exonuclease I, and gene expression was measured by real-time PCR on the BioMark HD System (Fluidigm, CA) using the 96.96 Dynamic Array IFC, the GE 96 × 96 Fast PCR⁺ Melt protocol, the SsoFast EvaGreen Supermix with Low-ROX (Bio-Rad Laboratories, CA), and 5 μM primers for each assay. Raw data were analyzed using the Fluidigm Real-Time PCR analysis software. The n-fold expression change between CAR-28/ζ versus CAR-Stop T cells was calculated according to the crossing point ΔΔC_P method.

To investigate the T cell activation level, triplets of 5 × 10³ T cells were seeded in the presence or absence of mu B cells (RW7938) in 96-well plates. Cocultivation was performed in the presence or absence of the anti-CD19 TM. After 48 h, cells were harvested, stained for CD3 and CD25 surface expression, and analyzed using the MACSQuant Analyzer.

To analyze the killing potential of CAR⁺ T cells, they were cocultivated with tumor cells in the presence or absence of a tumor-specific TM. Detection of specific target cell lysis was determined by flow cytometry–based viability assays using the MACSQuant Analyzer as published previously (24, 33).

To isolate LSK and LK cells from whole BM cells, Lin⁻ cells were pre-enriched by magnetic bead separation on an autoMACS Pro Separator using a mouse Lineage Cell Depletion Kit (both from Miltenyi Biotec). Following this, cells were stained at 4°C for 30 min with a mixture of Lin mAbs and Abs against cKit as well as Sca-I. Afterwards, pre-enriched cells were sorted on a FACSAria III to obtain LSK or LK cells.

Directly after isolation, LSK cells were transduced with the lentiviral vector p6NST60_CAR-28/ζ and LK cells either with p6NST60_CAR-Stop or p6NST60_CAR-28/ζ. Production of lentiviral particles by HEK293T cells was performed as described previously (24). For that purpose, HEK293T cells were kept in StemPro-34 SFM (1×) medium. For transduction, hematopoietic cells were cultivated in wells coated with recombinant human fibronectin fragment RetroNectin (Takara Bio, Shiga, Japan) at densities of 1 × 10⁵ cells/ml. StemPro-34 SFM (1×) medium supplemented with 100 μg/ml penicillin/streptomycin (Biochrom), 2 mM N-acetyl-l-alanyl-l-glutamine (Biochrom), 4 μg/ml polybrene (Merck, Darmstadt, Germany), 50 ng/ml recombinant mu stem cell factor, 50 ng/ml recombinant mu thrombopoietin, and 50 ng/ml recombinant mu Flt3-L (all purchased from PeproTech, Rocky Hill, NJ) served as transduction medium. Transduction was performed by adding the respective lentiviral virus supernatant twice at an MOI of 3–6 for the next 24 h. After genetic modification, transduced cells were rested in transduction medium lacking polybrene for an additional 12 h before transfer into recipient mice.

First, experimental mice received a lethal total body radiation of 9 Gy (YXLON Maxi Shot Source, Hamburg, Germany). One day after conditioning, animals were i.v. injected with 1 × 10⁵ transduced LSK cells mixed with 2 × 10⁶ Sca-1^- “rescue”-BM cells. Alternatively, mice were transplanted with 2 × 10⁶ transduced LK cells. For the following 3 wk, 1.17 g/l neomycin (Sigma-Aldrich, Steinheim, Germany) was added to the drinking water of γ-irradiated animals. To generate secondary (2°) recipient mice, serial transplantation was performed using EGFP-sorted BM cells (>95% purity) from 1° recipients 14 wk posttransplantation. After isolation of EGFP⁺ BM cells, 1 × 10⁶ genetically modified cells were i.v. injected into lethally irradiated 2° recipient mice. To obtain a sufficient number of mice carrying CAR-transduced BM cells, tertiary (3°) recipients were generated based on unsorted whole BM samples of 2° recipient mice 18 wk posttransplantation.

Over the past years, we have described and established a universal modular Ab-based CAR platform technology (UniCAR) (see also Fig. 1A) (22, 27, 30, 35, 37–41). As schematically summarized in Fig. 1A, UniCAR T cells do not directly bind to the surface of tumor cells. Instead, UniCARs recognize a peptide epitope derived from the nuclear protein La/SS-B, which is not accessible on the surface of intact living cells (42, 43). Consequently, UniCAR T cells alone are inert (see also Fig. 1A) as was confirmed previously both in vitro and in vivo in mouse models (e.g., 39). To redirect UniCAR T cells to target cells, a second component, a TM, is required, which mediates the cross-linkage between UniCAR T cells and the cell surface molecule on the tumor cell. To fulfill this function, a TM represents a bispecific fusion molecule consisting of a binding moiety against a tumor-associated Ag (TAA) and the epitope recognized by the UniCAR.

During development of the UniCAR platform, a series of CAR constructs varying with respect to their extracellular peptide binding domain, spacer region, transmembrane domain (TMD), and also of intracellular signaling domains were established and analyzed (M.P. Bachmann, unpublished observations). Among them was the anti-La CAR CAR-28/ζ presented in this article. This construct came into our focus as in contrast to T cells expressing UniCAR T cells genetically manipulated to express the anti-La CAR-28/ζ construct, which showed a strong preactivation and ligand-independent tonic signaling (see also below, Figs. 2B–D, 3B). The structure of the CAR-28/ζ construct is schematically summarized in Fig. 1B. It differs from UniCARs as it contains the TMD of the mu CD28 molecule and a combined mu CD28/CD3ζ signaling domain. For detailed analysis of the effects of the CAR-28/ζ on T cell activation and function, we created the corresponding control construct (CAR-Stop) lacking any signaling moiety. To detect the expression of the respective CAR on the cell surface, both constructs contain the peptide epitope E7B6 in their extracellular domain as described previously for UniCARs (see also 2Materials and Methods). For differentiation of genetically modified cells from wild type (wt) cells, both CAR constructs allow a cotranslational expression of EGFP.

To confirm these features, in a first step, human T cells were lentivirally transduced, and the CAR expression was measured by flow cytometry. To discriminate between genetically modified and wt cells, coexpression of the extracellular peptide epitope E7B6 and EGFP marker expression was compared (Fig. 2A, presort). The results confirm that expression of the respective CAR construct and the EGFP marker protein tightly correlates. Thus, detection of the reporter protein is sufficient for further identification of CAR⁺ cells. Furthermore, transduction rates of five independent T cell donors were comparable between both CAR constructs (CAR-Stop: 53 ± 18%; CAR-28/ζ: 50 ± 17%). Before experimental usage, modified T cells were sorted to >90% purity to ensure similar starting conditions (Fig. 2A, postsort).

Having confirmed the successful expression of both CAR constructs, we next aimed to verify that engraftment with the CAR-28/ζ induces a strong tonic signal in the absence of target Ag. For that purpose, transcriptional expression of 55 T cell–related genes was analyzed in resting nontransduced, CAR-Stop, or CAR-28/ζ–modified human T cells (Fig. 2B, 2C). Obtained data clearly reveal that all of the investigated genes are expressed at much higher levels in cells engineered with a signaling CAR construct than in cells lacking a signal-transmitting endodomain. In accordance with studies describing “continuous” CARs (10, 11), CAR-28/ζ T cells were characterized by a high cytokine production (IFN-γ, GM-CSF, and IL-4), elevated levels of granzyme B and perforin, and an enhanced expression of exhaustion-associated transcription factors (EOMES, Helios, T-bet). Other investigators report an upregulation of Fas (CD95) and FasL associated with sustained ligand-independent T cell activation (44), which also applied to T cells transduced with the CAR-28/ζ construct. Further evidence for tonic signaling was gathered by detection of the exhaustion markers PD-1 and LAG-3 as well as the activation marker CD69 at the protein level. As shown in Fig. 2D and confirmed by CTLA4 gene array analysis (Fig. 2C), CAR-28/ζ-expressing T cells display an exhausted phenotype, which is considered one of the hallmarks of CAR tonic signaling (11, 12).

Subsequently, we wanted to verify functionality and antitumor activity of the CAR-28/ζ construct. First, the activation status of transduced T cells was analyzed in the presence or absence of CD19⁺ target cells (mu B cell line RW7938) and in the presence or absence of a cross-linking anti-muCD19 TM. The CD25 protein expression data confirm a strong basal activation level of CAR 28/ζ T cells already in the absence of target cells and without a cross-linking TM (Fig. 3A). Noteworthy to mention, there is still an additional upregulation of CD25 expression in the presence of the TM. In contrast to CAR-28/ζ–expressing T cells, neither CAR-Stop T cells nor control T cells display enhanced CD25 levels upon incubation with target cells and the TM. As the upregulation of CD25 in T cells expressing CAR-28/ζ does not differ in the absence or presence of target cells, we concluded that the activation is not simply a xeno reaction to the target cell. Next, we analyzed the killing capability of T cells expressing the CAR-28/ζ construct. The lysis experiment was performed over 48 h in complete RPMI 1640 in the absence of recombinant cytokines. T cells expressing the CAR-Stop construct and T cells not expressing CARs served as negative controls. As shown in Fig. 3B, efficient eradication of CD19⁺ RW7938 target cells was exclusively induced by the TM-mediated cross-linkage between CAR-28/ζ⁺ T cells and the tumor cells. In contrast, no significant tumor cell lysis was observed in the absence of the CD19-directed TM or by using CAR-Stop T cells.

Taken together, these results confirm that CAR-28/ζ–armed T cells exhibit a high preactivation level; hence, the CAR itself causes tonic signaling in the absence of target ligand. Nevertheless, as tumor cell lysis is only occurring in dependence on a cross-linking TM, CAR T cell functionality remains highly specific.

After having established both CAR constructs in human T cells, CAR expression was assessed in two mu BM populations. For that purpose, BM was isolated and either LK cells or the LSK subpopulation were purified (Fig. 4A), which are enriched for HSCs and multipotent progenitors (45). Following isolation of the respective population, cells were genetically modified with the CAR-Stop or the CAR-28/ζ vector. As confirmed by flow cytometry, both LK and LSK cells efficiently expressed the EGFP reporter gene and, consequently, the corresponding CAR construct (Fig. 4B). Using an MOI of 3, LK cells showed similar transduction efficiencies for the CAR-Stop (24.2%) and CAR-28/ζ construct (22.8%). In addition to LK cells, we analyzed whether the LSK subpopulation could be directly transduced. Because of the low percentage of LSK cells in the BM compartment and, accordingly, the limited number of cells isolated, LSK cells were only modified with the CAR-28/ζ vector using an MOI of 6. The results in Fig. 4B confirm that, similarly to LK cells, LSK cells also express the novel signaling CAR construct (67.1%). The higher transduction efficiency obtained for these cells is most likely because of the increased MOI used.

Following transduction, modified BM cells were transplanted into lethally irradiated C57BL/6J mice (Fig. 5A). The 1° recipient mice received either LK cells armed with the CAR-Stop construct or LSK cells expressing the CAR-28/ζ molecule. Three months postinjection, BM was harvested to assess in vivo engraftment. CAR⁺ BM cells were identified using the EGFP marker. Despite varying transduction efficiencies (CAR-Stop⁺ LK cells: 24.2%; CAR-28/ζ⁺ LSK cells: 67.1%), there was no significant difference in engraftment. Independent of the respective CAR construct, roughly 32–36% of the BM cells were identified as EGFP⁺/CAR⁺ (Fig. 5B). To further increase EGFP⁺/CAR⁺ donor chimerism, isolated BM cells from 1° recipient mice were sorted for EGFP expression (purity >95%) and transplanted into lethally irradiated 2° recipient mice (Fig. 5A). After 4 mo, these 2° recipient mice showed an efficient engraftment in the BM compartment with a percentage of over 95% CAR-armed cells as based on EGFP expression (Fig. 5B). Additionally, 3° recipient mice were generated by serial transplantation of modified BM cells. Four months posttransplantation, BM chimerism was above 90% for all mice (n = 3; CAR-Stop⁺ BM cells: 95 ± 1%; CAR-28/ζ⁺ BM cells: 92 ± 1%), which confirms successful serial long-term repopulation of CAR-modified cells (Fig. 5B). Owing to this efficient engraftment, 2° and 3° recipients were suitable for subsequent phenotypic characterizations.

Following the establishment of 2° recipient mice with high EGFP⁺/CAR⁺ donor chimerism, the impact of both constructs on hematopoiesis was analyzed. For that purpose, PBCs of four representative mice were isolated, and EGFP⁺/CAR⁺ subpopulations were characterized by flow cytometry. Cells were stained against the cell surface markers CD3 (T cells), B220 (B cells), CD11b (myeloid cells), Gr1 (mainly granulocytes and monocytes), CD49b (mainly NK cells), and NK1.1 (NK cells and subset of NK-T cells). The obtained results underline that modified BM cells successfully engrafted and differentiated, depending on the integrated CAR construct, into both myeloid and lymphoid Lin (Fig. 6). Mice with CAR-Stop–modified BM cells developed CAR⁺ T cells (n = 4; 14 ± 3%), B cells (n = 4; 57 ± 3%), and myeloid cells (n = 4; 33 ± 3%), whereas BM cells transduced with the CAR-28/ζ construct developed neither T cells (n = 4; 2 ± 1%) nor B cells (n = 4; 1 ± 1%) as shown for one representative recipient in Fig. 6A. Thus, transplanted BM cells carrying the signaling CAR lack all CAR⁺ cell types of the adaptive immune system. Additional analysis on NK cells showed that CAR-Stop– as well as CAR-28/ζ-transduced BM cells differentiated into NK cells (Fig. 6B). However, the percentage of CAR-28/ζ⁺ NK cells (13.7%) within the PBCs was higher than of CAR-Stop⁺ NK cells (0.9%). This relative increase of NK cells was caused by the absence of CAR-28/ζ⁺ T and B cells.

In general, the phenotypic analysis reveals that only CAR-Stop⁺ BM cells undergo normal hematopoiesis, whereas expression of the signaling construct perturbs B as well as T cell lymphopoiesis.

After discovering the absence of mature CAR-28/ζ⁺ B and T cells, we investigated the reason underlying these results. For that purpose, development of the respective progenitor populations was analyzed in detail. To reduce the impact of recipient-dependent variations, the number of experimental animals was increased, and at this time, the 3° recipient generation was established. Four months posttransplantation, hematopoietic cells from BM, PB, and thymus were isolated and subsequently analyzed by flow cytometry. First, we compared the expression of stem cell markers between BM cells of a nontransplanted wt mouse and three CAR-Stop⁺ or three CAR-28/ζ⁺ transplanted animals. As shown for one representative recipient in Fig. 7, CAR-Stop as well as CAR-28/ζ mice contained similar percentages of EGFP⁺/CAR⁺ HSC-enriched LSK cells (n = 3; CAR-Stop⁺ LSK: 1.1 ± 0.1%; CAR-28/ζ⁺ LSK: 0.9 ± 0.3%) and CLPs (n = 3; CAR-Stop⁺ CLPs: 0.6 ± 0.2%; CAR-28/ζ⁺ CLPs: 0.5 ± 0.2%). Moreover, data demonstrate that the percentage of these populations is slightly reduced in comparison with a wt mouse (n = 1; LSK: 2.9%; CLPs: 1.9%). Decrease of LSK cells and CLPs is probably due to serial BM transplantation. Nevertheless, CAR⁺ hematopoietic progenitors exist and, thus, production of T and B cells is theoretically possible.

In principle, CLPs have the potential to differentiate into T cells, B cells, and NK cells (Fig. 8A). Still, CAR-28/ζ⁺ T cells are missing, whereas CAR-Stop⁺ T cells are present. Therefore, T cell differentiation in the thymus was characterized in detail (Fig. 8). After removing the respective thymi, the size of the organs was compared between a control mouse and mice with modified BM cells (Fig. 8B). Results indicated that animals with CAR-28/ζ cells not only lack T cells but additionally possessed smaller thymi. Furthermore, thymocytes were isolated, and the expression of surface Ags was analyzed to distinguish basic T cell differentiation stages. Briefly, double-negative (DN) T cells that neither express CD4 nor CD8 progress through four DN stages (DN1–DN4) and develop into double-positive (DP) T cells expressing CD4 and CD8, which become single-positive (SP) cells during positive selection (Fig. 8A). As can be seen in Fig. 8C, 8D, CAR-Stop⁺ thymocytes differentiate into DN (n = 3; 5 ± 1%), DP (n = 3; 86 ± 2%), and finally SP T cells (n = 3; CD4⁺: 8 ± 1%, CD8⁺: 2 ± 0.1%) in a similar way as wt cells. In contrast to this, thymocytes expressing the CAR-28/ζ construct underwent incomplete T cell differentiation. Investigating CAR-28/ζ⁺ thymocytes reveals that the main population were DN cells (n = 3; 93 ± 3%). Moreover, nearly all of them belonged to the DN1 (n = 3; 57 ± 5%) and DN2 (n = 3; 27 ± 15%) stages. Thus, the number of pre-TCR–expressing DN3 and DN4 T cells is dramatically reduced compared with CAR-Stop mice (n = 3; CAR-Stop⁺: 66 ± 6%; CAR-28/ζ⁺: 16 ± 11%). Because of the loss of DN3 and DN4 cells, only marginal numbers of DP (n = 3; 1 ± 1%) and SP T cells (n = 3; CD4⁺: 3 ± 2%; CD8⁺: 3 ± 1%) have developed (Fig. 8D). This arrest in T cell differentiation is consistent with the reduced thymi size as lymphocytes influence thymic development (46).

Besides loss of mature T cells, our data revealed that expression of the CAR-28/ζ construct additionally interfered with B cell development. To investigate this, BM cells were isolated and analyzed for major B cell stages. Theoretically, CLPs can give rise to pro-B cells, which differentiate into pre-B cells carrying a pre-BCR. After productive rearrangement of Ig H and L chain genes, pre-B cells become immature IgM⁺ B cells that exit the BM for peripheral organs (Fig. 9A). Analyses of B cell development in CAR-Stop⁺, CAR-28/ζ⁺, and wt BM cells revealed that pro- (n = 3; CAR-Stop⁺: 16 ± 2%; CAR-28/ζ⁺: 6 ± 2%) and pre-B cells (n = 3; CAR-Stop⁺: 84 ± 2%; CAR-28/ζ⁺: 94 ± 2%) were generated irrespective of CAR expression in these cell populations (Fig. 9B). Nevertheless, CAR-28/ζ⁺ pre-B cells did not give rise to immature IgM⁺ B cells (n = 3; 1 ± 0.3%), whereas wt mice (n = 1; 8%) as well as CAR-Stop mice (n = 3; 8 ± 1%) possessed IgM⁺ B cells (Fig. 9B, 9C). Thus, as soon as a functional BCR is expressed on the surface, CAR-28/ζ⁺ B cells are eliminated.

Together, these phenotypic studies demonstrate that expression of the CAR-28/ζ construct has a negative effect on hematopoiesis and interferes with T cell as well as B cell development, whereas NK cell production and differentiation of the myeloid compartment are unaffected.

Despite improvement of traditional cancer treatments, including surgery, chemotherapy and radiation, novel therapeutic approaches are urgently needed. The clinical success of CAR T cells underlines their impressive potential to specifically eliminate tumor cells. However, proliferation capacity of differentiated T cells is limited, and exhaustion may occur over time. Moreover, the immunosuppressive microenvironment in solid tumors may convert, entering CAR T cells into anergic T cells. One possibility to overcome these limitations could be the transfer of HSCs genetically modified to express CARs. So far, infusion of autologous or allogeneic HSCs is part of standard medical procedures to induce an immune response against hematological malignancies (47). Hence, additional engraftment of HSCs with tumor-specific CARs could provide a persistent source of multilineage effector cells to establish an immunological memory and, thus, to further increase graft-versus-cancer activity (15, 16, 19, 48, 49). Besides CAR-expressing T cells, efficient antitumor effects were also demonstrated for genetically modified NK and myeloid cells (16, 48, 50). Thus, Ag-specific innate immune cells constitute a potent tool for cancer treatment because of their intense circulation and vascular diapedesis properties (17). Furthermore, HSC differentiation into CAR⁺ NK and myeloid cells is faster than T cell formation as they do not depend on thymopoiesis (16, 51). Hence, cells of the innate immune system represent the initial effector cells, whereas CAR-armed T cells further augment tumor eradication.

However, controversial data with respect to hematopoiesis from CAR⁺ HSCs are reported in the literature (15, 16, 18–20). Therefore, we investigated the effects of CAR-mediated tonic signaling frequently observed in second-generation CAR T cells (10, 11, 44). In addition to ligand-driven tonic signaling, T cell stimulation can be triggered by Ag-independent aggregation of CARs, which may promote T cell exhaustion and, thus, limit therapeutic effects (11, 52). The data presented in this article are in line with these studies. One consequence of our finding for the development of a novel clinically applicable CAR construct is that a simple replacement of the scFv may not be sufficient. The qualitative and quantitative effect on tonic signaling might also be taken into account. To assess mainly the impact of tonic signaling on HSCs, two La-specific CAR constructs were selected for our studies: 1) a second-generation CAR containing a mu CD28/CD3ζ signaling domain (CAR-28/ζ) and 2) a control construct lacking the cytoplasmic signal transduction moiety (CAR-Stop). Considering the aim of this project, the nuclear autoantigen La/SS-B appeared as an ideal target to study Ag-independent tonic signaling because it is not accessible on the cell surface under physiological conditions (42, 43). Therefore, CAR T cells are inactive in vitro and in vivo in experimental mice, until they are cross-linked to cancer cells via tumor-specific TMs (e.g., 22, 27, 39). According to transcriptional profiling of resting gene-modified T cells and flow cytometry–based analysis of the activation markers CD69 and CD25 as well as exhaustion-associated markers PD-1 and LAG-3, the CAR-28/ζ construct presented in this article led to a basal tonic signaling in transduced T cells, whereas the CAR-Stop lacks this signal. Of note, a tonic signaling could be beneficial in the context of switchable CARs like the UniCAR system, which are per se inert, by promoting their engraftment. It is therefore of interest that CAR-28/ζ CARs are fully functional and kill target cells in a TM dependent and specific manner despite the activated state. Thus, this pair of CAR constructs appeared to us as a suitable model to study the effects of tonic signaling on the repopulation potential of modified CAR-expressing HSCs.

To evaluate in vivo effects, we used immunocompetent mice because 1) their hematopoietic system is highly analogous to humans (53), and 2) it is critical to test immunotherapies in model systems with a fully functional immune system. First, BM cells of C57BL/6J mice could be efficiently modified with the CAR-Stop as well as the CAR-28/ζ construct yielding similar transduction rates. These CAR-armed BM cells were successfully transplanted into lethally irradiated syngeneic mice. Importantly, none of the CAR constructs impaired long-term BM engraftment, which is consistent with previous descriptions (15, 16, 19, 54). Moreover, the establishment of 1°, 2°, or 3° recipients with CAR⁺ BM cells confirms long-term and serial repopulation activity despite tonic signaling.

To analyze the differentiation of modified HSCs, peripheral blood was checked for CAR⁺ T cells, B cells, NK cells, and myeloid cells. The control CAR lacking any intracellular signaling domain was detected in all tested hematopoietic cell populations. Thus, as expected, the expression of the CAR-Stop molecule, which cannot induce any reaction, including a tonic signaling, has no apparent effect on hematopoiesis. In contrast, mice engrafted with CAR-28/ζ–transduced BM cells developed no mature CAR⁺ T cells or B cells. These findings differ from previous studies using CD19-specific CAR constructs. Using these CARs, not only modified NK and myeloid cells were present but also anti-CD19 CAR⁺ T cells (15, 16). However, the incorporated CD19 scFv (clone FMC63) was already tested by Long and colleagues (11) and mediated no CAR clustering and, therefore, no tonic signaling. Thus, to the best of our knowledge, we are the first ones demonstrating that Ag-independent tonic signaling of CARs may interfere with hematopoiesis.

To define the time point at which expression of the signaling CAR blocks T cell and B cell development, we examined the presence of the respective progenitor populations. Irrespective of the CAR construct, modified LSK cells (containing HSCs) as well as oligopotent CLPs were successfully produced. Nevertheless, development of CAR-28/ζ T cells from transplanted HSCs is arrested after DN2 stage. In normal mice, T cells of the DN3 stage are characterized by the formation of a pre-TCR, which associates with the CD3 complex, including the CD3ζ chain harboring three ITAMs. Stimulation via these ITAMs is essential for transition of DN3 to DN4 stage (55, 56). As the CAR-28/ζ construct contains the ITAMs of the CD3ζ molecule, we hypothesize that CAR-mediated tonic signaling causes an aberrant ITAM-based signal, leading to a blockage in T cell development. This assumption is further supported by the results of Lin and Roberts (20) who showed that because of Ag-driven stimulation, mature CAR⁺ T cells were missing (19). In addition to the absence of CAR-28/ζ T cells, we unexpectedly observed that tonic signaling also impairs B cell development. Similar to T cells, the development of mature B cells depends on appropriate signaling via the pre-BCR, which is associated with the ITAM-bearing Igα and Igβ heterodimer (57). Thus, it seems likely that, similarly to T cells, the aberrant ITAM-mediated signal affects B cell differentiation, leading to the unexpected complete loss of B cells if such a signal occurs in the BM during differentiation.

In summary, the presented data indicate that tonic signaling of CAR molecules may have adverse effects on HSCs regarding the development of cells of the adaptive immune system. It seems that tonic signaling leads to a loss of adaptive immune cells as soon as recombined immune receptors become functional, and the developing cells are checked for autoreactivity. Consequently, tonic signaling becomes detrimental during T and B cell development by mimicking autorecognition via the newly recombined immune receptors. Therefore, tonic signaling of CAR constructs to be introduced into HSCs need to be assessed for such limiting side effects. Alternatively, the expression of CARs from promoters, which are active following the critical stages of T cell development, may overcome such constraints in the future.

We thank Simon Loff, Annett Lindner, Doreen Löbel, and Katja Müdder for technical assistance.

M.P.B., G.E., and A. Ehninger hold patent applications related to the UniCAR platform and the anti-La Abs and epitopes used. The other authors have no financial conflicts of interest.