Treatment of Multiple Myeloma Using Chimeric Antigen Receptor T Cells with Dual Specificity

Multiple myeloma is a blood-borne malignancy characterized by clonal neoplastic proliferation of plasma cells in the bone marrow. The survival of patients with multiple myeloma has improved significantly over the past 15 years, owing to an arsenal of new treatment options (1, 2). Nevertheless, multiple myeloma remains an incurable cancer and most patients will continue to experience repeated relapses and eventually succumb to refractory disease (3, 4).

Synthetic chimeric antigen receptors (CAR) represent a major advance in adoptive T-cell therapy. By combining the variable regions of a high‐affinity monoclonal antibody with intracellular signaling components derived from the T-cell receptor (TCR) complex, CARs allow redirection of T-cell cytotoxicity against an antigen of choice, entirely independent of target cell MHC expression (5–8). In light of the exceptional success of CAR therapy in eliminating B-cell malignancies and the regulatory approval of CD19-targeted CARs (9, 10), CAR T cells targeting multiple myeloma have increasingly been the subject of preclinical and clinical research (11–14). Because malignant plasma cells lack CD19 in most patients with multiple myeloma, several myeloma-specific markers have been tested as targets for cell therapy. Anti–B-cell maturation antigen (BCMA) CAR T cells have demonstrated the most promising clinical responses in patients with relapsed/refractory multiple myeloma (15–17). However, relapses associated with downregulation or loss of target antigen have been observed, highlighting the need to identify alternative multiple myeloma–associated targets and develop new CAR T cells (18). Nevertheless, most of the multiple myeloma target antigens are not tumor specific, and are also present on healthy tissues and/or in the hematopoietic system. Therefore, targeting these antigens may result in adverse effects such as “on-target off-tumor” toxicity. In the case of CD19, which is expressed by normal B-lineage cells, these effects can result in manageable B-cell aplasia (19–21). In other instances, the side effects may be intolerable and potentially life-threatening, as is reported for carbonic anhydrase IX or human ERBB2 targeting (22, 23).

To address the unmet need for an effective anti–multiple myeloma CAR-T therapy with reduced toxicity, we adopted a “dual-CAR” T-cell strategy (24–26) to cotarget two multiple myeloma–specific markers. In the dual-CAR strategy, the activation and costimulation signals (signal 1 and signal 2, respectively) are separately provided by two polypeptide chains, each targeting a different tumor-associated antigen through a single-chain variable fragment (scFv) recognition unit. With this split CAR configuration, cotransduced T cells are expected to eliminate tumors expressing both antigens but to have a minimal effect on normal tissues expressing either antigen alone.

In this study, we described the construction and preclinical testing of an anti–multiple myeloma dual CAR. The dual-CAR approach was implemented to require the surface expression of both CD138 (Syndecan-1, SDC-1), and CD38 (ADP-ribosyl cyclase). The former is a transmembrane heparan sulphate proteoglycan, expressed on some B cells, plasma cells and a variety of epithelial cells, and involved in cell adhesion and maturation. The latter is a transmembrane glycoprotein expressed mainly on male tissues (prostate and seminal vesicles), bone marrow, and lymphoid tissues (27–30). CD38 and CD138 are highly coexpressed on multiple myeloma tumor cells in most patients, and represent attractive targets for CAR therapy. We showed that effector cells transfected with the dCAR138-38 exhibited strong reactivity against multiple myeloma both in vitro and in vivo with reduced off-target activity compared with the single target CAR controls.

The following recombinant human proteinswere used: Fc-tagged CD38 (rec-hCD38-Fc; 10818-H02H), CD138 (rec-hCD138; 11429-H08H), Fc-tagged Erb2 (rec-ErbB2-Fc; 10004-H02H; Sino Biological). The following flow cytometry antibodies were used: phycoerythrin (PE)-conjugated anti-human (γ-specific) Fc receptor (Invitrogen; 12–4998–82), allophycocyanin (APC)-conjugated anti-human (γ-specific) Fc receptor (BioLegend; clone M1310G05; 410712), APC-conjugated anti-hCD138 (BioLegend; clone DL-101; 352308), eFluor450-conjugated anti-human CD38 (Invitrogen; clone HB7; 48–0388–42), anti-human CD3-APC (BioLegend; clone UCHT1; 300439), anti-human CD4-Brilliant Violet (BioLegend; clone OKT4; 317434), and anti-human CD8-eFluor450 (Invitrogen; clone RPA-T8; 48–0088–42), 7AAD (BioGems; 61410–00–200). For preparation of activation plates, purified anti-human CD3 (in-house preparation from OKT3 hybridoma kindly provided by Prof. Ronald Levy, Stanford School of Medicine, Stanford, CA) and purified anti-human CD28 (BioLegend; 302934) were used. Other reagents were as follows: RetroNectin (Takara; T100B) was used for preparation of precoated transduction plates, Human IL2 (Novartis-Pharma; 4764111 U57).

293T (human embryonic cells; ATCC CRL-1573), FaDu [pharynx squamous cell carcinoma; kindly provided by Dr. Nidal Muhanna, Tel-Aviv Sourasky Medical Center (TASMC) at 2018], and PG13 (gibbon ape leukemia virus pseudotyping packaging cell line; kindly provided by Ralph Wilson, Rotterdam hospital at 2004) were cultured in DMEM supplemented with 10% fetal calf serum (FCS), 2 mmol/L glutamine and 1 mmol/L sodium pyruvate. Multiple myeloma-derived cell lines: CAG, H929 and U266 (ref. 31; provided by the Hematology Department in TASMC at 2013) as well as human lymphocytes were cultured in RPMI-1640 (Biological Industries) supplemented with 10% FCS, 2 mmol/L glutamine. All media were supplemented with a mixed antibiotic solution containing penicillin (100 U/mL), streptomycin (100 μg/mL), and neomycin (10 μg/mL; Bio-Lab). Human primary cells were purchased from Cell Biologics: alveolar epithelial H-6053, pancreatic epithelial H-6037, cardiac microvascular endothelial H-6024, kidney epithelial H-6034, kidney endothelial H-6014 and from Lonza: Prostate epithelial CC-2555. Each primary cell was thawed and propagated according to the supplier's instructions. We perform periodic Mycoplasma screens by PCR (HyLabs Israel); all cell lines used in this study were confirmed to be Mycoplasma free. The cells were frozen at low passage, and the number of passages was recorded. Cells were maintained at approximately 12 passages. The cells were not authenticated in the past year. Mononuclear cells from human healthy donors were obtained from the Israeli blood bank at Sheba Medical Center, Israel (Helsinki number 0185–14-TLV).

Analysis of cell surface expression of the CD38- and CD138-specific CAR receptors was on the basis of detection of their extracellular scFv moieties. A 2-step staining protocol was used, which included an initial incubation of CAR-transduced T cells with the cognate recombinant proteins: rec-hCD38-Fc and/or rec-hCD138 followed by incubation with fluorescently labeled secondary antibodies: PE-anti–human Fc receptor and/or APC-anti–human CD138. Detection of the Erb2-specific scFv in the context of dual CAR null controls was done by incubation with rec-hErb2-Fc followed by APC or PE-anti–human Fc receptor. Expression of anti–Erb2-CAR was estimated on the basis of green fluorescence protein (GFP) fluorescence [the a-Erb2-CAR gene sequence is followed by the combined sequences of an internal ribosome entry site (IRES) and GFP in the expression vector]. Staining and cell washes were done in 0.1 mL or with 2 mL of FACS Buffer (containing 2% FBS, 0.05% sodium azide, and 2 mmol/L EDTA pH8 in PBS), respectively. Incubation of cells with recombinant proteins or with antibodies was done on ice for 30 to 40 minutes. Staining was monitored with a BD CANTO flow cytometer. Data analysis was carried out using FCS Express software.

The sequences of hCD138- and hCD38-specific scFv's were derived from indatuximab ravtansine (BT062) and daratumumab (DARZALEX; Janssen Biotech), respectively. The N29 monoclonal antibody against human Erb2 was used as a source of the scFv, for construction of the Erb2-directed CAR (32). In all of these scFv's, the variable fragment light (V_L) chain was linked to the variable fragment heavy (V_H) chain through a flexible, 14-amino acid linker (GSTSGSGKSSEGKG; ref. 8). A leader peptide was attached at the 5′ end to direct cell-surface localization. In the single-targeting CARs, the scFv was ligated to a portion of the human CD28 costimulatory molecule (encompassing its cytosolic hinge, transmembrane and inner signalling domains) followed by the activating domain of human FcγR molecule at the 3′ end. The dCAR138-38 was designed as a bicistronic construct and was ordered from Bio Basic Inc. within pUC57 vector. The dCAR138-38 CAR sequence is comprised of the hCD138-specific scFv, located upstream of CD28, and the scFv-targeting hCD38, attached to the FcγR activation signal through the human CD8 (hCD8)-derived cytosolic hinge and transmembrane domain. The CD138(scFv)-CD28 and CD38(scFv)-hCD8-FcγR segments are linked by a T2A self-cleaving peptide sequence from the picornavirus foot-and-mouth disease virus, which allows stoichiometric expression of two separate polypeptides from a single gene (Fig. 1; ref. 33). To generate the CD138-null and CD38-null controls, the CD138 or the CD38 ScFvs were replaced by Erb2-specific scFv in the dCAR138-38 construct. All CAR constructs were sub-cloned into the MSGV-1 retroviral backbone (kindly provided by Dr. Steven A. Rosenberg from the NCI). For cloning, we used Transfer-PCR (TPCR), a restriction enzyme-free cloning platform (34).

293T cells were transfected using Ca₂PO₄ with GAG-POL pCL-Ampho retroviral envelope (35) and the CAR-encoding MSGV-1 vectors. Retroviral supernatant was collected and used to stably transduce the amphotropic PG13-packaging cells. Infected cells were stained according to expressed CAR (see “Flow cytometry analysis” section) and sorted (BD FACSAria III Cell Sorter) to obtain 100% positive CAR-expressing cells. Sorted cells were expanded and frozen in aliquots in liquid nitrogen.

Retroviral transduction of T cells was performed as described previously (36). Briefly, peripheral human blood lymphocytes (PBL) were isolated from the blood of healthy human donors by density gradient centrifugation on Ficoll-Paque (Axis-shield). PBLs were activated in nontissue culture–treated 6-well plates, precoated with anti-human CD3 and anti-human CD28 for 48 hours at 37°C. Activated lymphocytes were harvested and subjected to two consecutive retroviral transductions in RetroNectin precoated, nontissue culture–treated 6-well plates supplemented with human IL-2 (100 IU/mL). After transduction, cells were cultured in the presence of 350 IU/mL IL2 for 24 to 72 hours for in vitro or in vivo assays, respectively. To evaluate transduction, efficiency cells were stained (according to the expressed CAR; see “Flow cytometry analysis” section) and the percentage of positive cells was determined by flow cytometry. Activated but noninfected cells were included as T-cell controls.

Multiple myeloma–specific CAR T or control cells were incubated for 24 hours with target cells at a 2:1 effector:target (E:T) ratio. Cell-free growth medium was collected and analyzed for IFN-γ production by ELISA using a human IFN-γ ELISA kit, according to the manufacturer's instructions (DY285B, R&D systems).

Lentiviral particles were produced in 293T cells by calcium phosphate-mediated transfection involving a four-plasmid expression system (37) In brief, 293T cells were plated into 10 cm² plates at 1.5 × 10⁶ cells/plate for 24 hours. The transfer vector plasmid (pHAGE-CMV-luciferase-W; ref. 38) was cloned in our laboratory, the helper plasmids plp-1 and plp-2 and the envelope plasmid plp-VSVG were mixed with calcium chloride (0.25 mol/L) and HEPES-buffered saline (HBS), and were added to the 293T cells. The virus particles in the medium were harvested 48 hours after transfection, filtered (0.45 μm), and incubated with preseeded CAG cells in the presence of polybrene (8 μg/mL; Sigma). After 3 to 4 hours, viral supernatant was replaced by fresh medium. Limiting dilution of transduced cells was employed to isolate expressing clones. Luciferase expression was analyzed using IVIS 100 Series Imaging, System (Xenogen). The resulting cell line is referred to CAG-Luc.

pCMV_hCD38 and pCMV hCD138 (Sino Biologicals) were used as templates for PCR amplification. Amplified hCD38 and hCD138 sequences were subcloned into the pGEM/4z vector (P2161, Promega) between NcoI and NotI restriction enzymes, using the TPCT platform (34). Vectors were linearized and subjected to in vitro transcription using HiScribe T7 ARCA mRNA kit (NEB, Bio-Lab). For RNA electroporation, FaDu cells were washed twice with OPTI-MEM (Invitrogen) and resuspended in OPTI-MEM at 2 × 10⁷ cells/mL. Subsequently, cells were mixed with 50 or 100 ng hCD38-RNA, or with both hCD38-RNA (500 ng) and hCD138-RNA (10 μg), and electroporated in a 2-mm cuvette (Harvard Apparatus BTX) using an ECM830 Electro Square Wave Porator (Harvard Apparatus BTX) at300V, for 1 msec. Cells and cuvettes were prechilled by putting them on ice for >5 minutes before electroporation. Immediately after electroporation, the cells were seeded in 96-well or 24-well plates for killing assay or flow cytometry analysis, respectively.

The cytotoxicity of T cells was determined by a methylene blue staining‐based assay. dCAR138-38 T cells or noninfected lymphocytes were incubated with target cells in a 96-well plate at E:T ratios of 1:1, 0.5:1 and 0.25:1. After 6 or 16 hours, the plate was washed to remove T cells and dead target cells. Live cells that remained attached to the culture plate were fixed with 4% formaldehyde for 2 hours at room temperature (RT), washed and stained with 0.5% methylene blue (Sigma-Aldrich) diluted in 0.1 mol/L sodium borate for 15 minutes at RT. Cells were then washed to remove excess methylene blue solution. 0.1 mol/L HCL was added before analysis. Absorbance at 620 nm was read on a Multiskan FC ELISA reader (Thermo Fisher Scientific). Percentage of killing was calculated as follows: 100-(absorbance of target cell incubated with CAR T cells/absorbance of target cell incubated without CAR T cells × 100).

All animal experiments were approved by the TASMC ethics committee (see “Ethics” section, following page). Mice used were 8- to 10-week-old NOD/SCID/γ_c^−/− obtained from The Jackson Laboratories. Mice were maintained in a specific pathogen-free facility of the TASMC. Stable luciferase-infected CAG cells (1 × 10⁶) were delivered intravenously 8 days before treatment. Mice were lymph-depleted by full body irradiation (200Rad; flow 60rad/min) using Varian TrueBeam Linear Accelerator Radiation system. One day before lymphocyte infusion. On day 8, engineered CAR-T cells (10 × 10⁶ cells/0.5mL PBS/mouse) were intravenously injected to the multiple myeloma–bearing mice. The general condition of the mice was recorded twice a week. Specifically, we followed the overall appearance of the mice, their activity level, their interaction with the environment and their weight. Mice showing any physical and/or behavioral abnormality or 15% reduction in weight were sacrificed. Anti–multiple myeloma response was evaluated by following survival up to 120 days. In vivo imaging detecting luciferase signal expressed by CAG-Luc cells was performed in some of the experiments. Imaging was done by intraperitoneal injection of luciferin (75 mg/kg), which was monitored with a cooled CCD camera system (IVIS 100 Series Imaging, System, Xenogen). Experiments that included bio-luminescence imaging were carried out at the Weizmann Institute. All procedures were performed under anesthesia with ketamine (100 mg/kg) and xylazine (20 mg/kg; Kepro) diluted in PBS and injected intraperitoneally in accordance with institutional guidelines. Mice were sacrificed when they lost 10% of their weight, or 100 to 120 days from the initiation of the experiment, for histologic analysis.

Paraffin-embedded sections were prepared from all treated mice from in vivo experiments. Indicated tissues were taken for hematoxylin and eosin (H&E) staining, and slides were also stained for human CD138 expression in the pathology department of Sourasky Medical Center.

Animal studies were approved by TASMC under protocol 60-12-17 and carried out in accordance with the recommendations of the Animal Care and Use Committee of TASMC.

Blood donations of normal human lymphocytes were collected according to the principles expressed in the Declaration of Helsinki, and approved by the Institutional Review Board under protocols 0185-14-TLV. There was no need for any patient's consent.

Data were analyzed and graphed using GraphPad Prism V.6. Data are expressed as mean + SEM or SD. A P value of 0.05 was considered statistically significant.

The dCAR138-38 approach was on the basis of coexpression of two receptors on the T-cell surface (Fig. 1). Each receptor was targeted against a distinct tumor associate antigen (TAA). The endodomain of one receptor contained a T-cell–activating domain (incorporating an immunoreceptor tyrosine‐based activation motif, ITAM, motif) and the other included a costimulatory determinant. This split configuration allowed full signal transmission and efficient T-cell stimulation only upon engagement with tumor cells expressing both targeted TAAs. Here, hCD138- and hCD38-specific scFv's were cloned into a T2A-based bicistronic construct, which allows the expression of two separate polypeptides from a single gene (see Fig. 2A and the “Materials and Methods” section). The hCD138- and hCD38-specific scFv's were also incorporated into single-targeting CARs in a second-generation CAR configuration (CD138CAR and CD38CAR). A human-Erb2–specific CAR (constructed previously in our laboratory; ref. 39) was used as a nonrelevant control (Erb2CAR). In addition, the Erb2-specific scFv was used to replace each of the scFv's in the dCAR138-38 construct to generate the CD138-null and CD38-null controls (Fig. 2A). Using these constructs, we sought to mimic a situation of dCAR138-38 engagement with single-presenting cells expressing either CD38 or CD138. All CAR constructs were cloned into the MSGV-1–based retroviral vector and were stably expressed on T cells isolated from healthy human donors. Expression of CARs was confirmed by flow cytometry. The CD138 and CD38 CARs were efficiently expressed in transduced T cells (Fig. 2B). Transduction efficiency was 72%, on average. Transduced cells were 97% to 98% CD3⁺ and comprised of 30% CD4 and 60% CD8⁺ cells on average (Fig. 2C). As shown in Fig. 2D, in vitro expansion of dCAR138-38 T cells and their viability were consistently lower than that of noninfected T cells that were maintained in parallel. Three days following transduction, viability of lymphocytes expressing the dCAR138-38 was 75.32% compared with 94.98% of noninfected cells. Fold expansion of the dual CAR T cells was 1.77, whereas that of noninfected cells was 4.25.

Human T cells were retrovirally transduced with the dCAR138-38, Erb2CAR or left noninfected. The activity of the T cells was evaluated by measuring IFN-γ secretion which reflected their ability to undergo specific stimulation when cocultured with human multiple myeloma cell lines expressing both CD38 and CD138. As shown in Fig. 3A, IFN-γ secretion by the dCAR138-38 transduced T cells was significantly higher (2,800–4,000 pg/mL), compared with Erb2 CAR and noninfected T cells (100–570 pg/mL), demonstrating an effective anti–multiple myeloma response. Direct killing of human multiple myeloma cell line (CAG) by the dCAR138-38 was demonstrated as shown in Fig. 3B, showing 90% killing of target cells in an effector:target (E:T) ratio of 1:1 compared with 20% killing by noninfected cells. These results showed that T cells transduced with the dCAR138-38 had a potent anti–multiple myeloma activity.

To study tumor selectivity of the dCAR138-38, two strategies were applied. First, activity of the dCAR138-38 was compared with single-targeting CARs (CD38CAR and CD138CAR) and to CD138- or CD38-null controls (Fig. 4A and B). Second, activity of the dCAR138-38 toward double-versus single-presenting target cells was assessed (Fig. 4C). As shown in Fig. 4A, we measured IFN-γ secretion by T cells expressing different CARs, when incubated with CAG multiple myeloma cell line. We found that T cells transduced with single-targeting CARs specific to either CD38 or CD138 were effectively stimulated with IFN-γ secretion levels of 0.83- and 1.35-fold relative to dCAR138-38, respectively, confirming the functionality of both scFv's incorporated in the dCAR138-38. T cells expressing the CD38-null control, which can only transmit a costimulatory signal by engaging CD138, were not activated, with a basal IFN-γ secretion level similar to that of noninfected cells (Fig. 4A). This result was expected since activation of T cells was unlikely to occur in the absence of an activating signal. In contrast, T cells expressing the CD138-null control, which can transmit only an activating signal upon CD38 engagement, secreted IFN-γ in response to CAG stimulation, even in the absence of a co-stimulatory signal (Fig. 4A). However, activation was almost doubled when dCAR138-38 was expressed, demonstrating that dual engagement with both CD38 and CD138 antigens and transmission of both activating and costimulatory signals resulted in a stronger T-cell stimulation. Notably, CD38 is a known activation marker and is highly expressed on CAR T cells (27). As such, CD38 can potentially result in autoactivation of CD38-specific CAR-T cells leading to self-lysis (fratricide). Indeed, T cells expressing the full single-targeting CD38CAR secreted a substantial amount of IFN-γ in the absence of target cells (Fig. 4A, white bars). IFN-γ secretion was markedly reduced in T cells expressing the CD138-null control and the dCAR138-38, further demonstrating that transmission of an activating signal was less efficient in the absence of a co-stimulating signal (Fig. 4B). To further evaluate the specificity of the dCAR138-38, we measured its direct killing activity against double-presenting (CD38⁺/CD138⁺) or single presenting (CD38⁺/CD138⁻) target cells (Fig. 4C and D). Using RNA electroporation, CD38 and CD138 were coexpressed in FaDu-target cell lines (which is natively negative for these markers). In addition, CD38 was expressed as a single target at different levels (Fig. 4C). Target cells were incubated with 70% transduced dCAR138-38 T cells at various E:T ratios, and killing activity was measured. The results show that killing of CD38 single-presenting cells by the dCAR138-38 was correlated with CD38 expression level. At a 2:1 E:T ratio, percentage of killing of FaDu cells electroporated with 0.05 μg CD38 RNA was 13% compared with 36% killing when 0.1 μg CD38 RNA was electroporated. Killing of double-presenting target cells by the dCAR138-38 was significantly higher and more efficient compared with single-presenting cells reaching 74% at a 2:1 ratio, similar to myeloma CAG cells presenting both targeted antigens. No killing of FaDu cells electroporated with CD138 was observed, further demonstrating that single expression of CD138 did not trigger dCAR138-38 activation (Fig. 4D).

The potential toxicity of the dCAR138-38 against normal tissues was analyzed using primary human cell lines, derived from healthy tissues. As discussed above, the CD138-speicifc scFv in the dCAR138-38 was fused to the CD28 costimulatory domain, whereas the CD38-targeted scFv was fused to the FcgR activation domain. In this configuration, stimulation of the dCAR138-38 by cells that express only CD138 was not expected, reducing the risk for CD138-related toxicity. Therefore, primary cells were chosen primarily according to Human-Protein-Atlas–derived information regarding CD38 expression on normal tissues (prostate, kidney, pancreas, red blood cells; Table 1). Cells from essential organs (alveolar epithelial cells and cardiac microvascular endothelial cells) were included as well. Target cells were cocultured with the dCAR138-38 as well as with full single CD138CAR and CD38CAR T cells, and IFN-γ secretion was measured. As expected, all CAR T cells were stimulated by the CAG multiple myeloma cell line (secreting 2,700–4,000 pg/mL IFN-γ). Most primary cells—cardiac microvascular endothelial cells, kidney epithelial cells, alveolar epithelial cells, pancreatic epithelial cells, and red blood cells—stimulated the CD38 single-targeting CAR (3,674, 1,446, 1,107, 1,314, and 1,243 pg/mL IFN-γ, respectively), whereas the dCAR138-38 was much less reactive and moderately stimulated only by the pancreatic cells with 747 pg/mL secreted IFN-γ (Fig. 5).

We next tested the efficacy of dCAR138-38 in vivo. We established a multiple myeloma tumor model using NSG mice injected with CAG and treated tumor-bearing mice with different CAR T cells on day 8. Preconditioning treatment, using irradiation of 200RAD, optimized in preliminary studies, was given 1 day before CAR T-cell administration. Treatment efficacy was assessed by following survival rates of treated mice up to 120 days. The distribution of the CAG tumor cells was analyzed by histology following sacrifice. In these experiments 5–10 × 10⁶ T cells transduced with dCAR138-38 or control CARs (CD38/CD138/Erb2 CARs) or noninfected T cells were injected to multiple myeloma–bearing mice. A representative Kaplan—Meier survival curve is shown in Fig. 6A. Figure 6B presents the average median survival of mice in 5 experiments. These results demonstrated that the dCAR138-38 treatment was the most effective against CAG cells in vivo. Median survival of mice treated with the dCAR138-38 was 97.4 days on average, compared with 30 days in mice treated with noninfected T cells (P ≤ 0.005). CD138-null control CAR T cells also provided a beneficial effect on survival but to a lesser extent with a median survival of 70 days. A dose escalating experiment, presented in Fig. 6C, showed that efficacy of dCAR138-38 treatment was dose-dependent. Median survival of mice treated with 2.5 × 10⁶, 5 × 10⁶ or 10 × 10⁶ dCAR138-38 transduced T cells were 68, 73, and 90 days, respectively. These results demonstrated that even the lowest dose (2.5 × 10⁶ T cells/mouse) improved survival of mice compared with no CAR treatment (P < 0.0001). Histologic H&E staining of representative surviving mice (day 95) is presented in Fig. 6D. Three out of five dCAR138-38–treated and surviving mice were tumor free at the end of this experiment. In contrast, only 1 out of 5 mice treated with the CD138-null control CAR survived until day 95, and exhibited tumors in the liver, lung, and bone.

To follow tumor dynamics during CAR T-cell treatment, we transplanted NSG mice with luciferase-expressing CAG cells (CAG-Luc) enabling spatio-temporal monitoring of tumor development. Mice were treated with the various CAR T cells 8 days following tumor injection and were analyzed by bioluminescent imaging every 6 to 8 days. Similar to the survival results, the imaging analysis showed improved survival of dCAR138-38–treated mice (Fig. 6E). In agreement with the observed reduction in tumor mass, at day 62 following tumor injection, 4 out of 5 mice in this group were alive, whereas none of the mice treated with CD138/Erb2 and CD38-null control CARs survived more than 55 days. In the groups of CD138-null control and CD38CAR-treated mice, 2 out of 4/5 mice were viable, most of them had high tumor load. The imaging data demonstrated that the improved survival of dCAR138-38-treated mice correlated with reduced tumor burden. Two mice in this group showed initial signs of tumor development at the relatively delayed time point of approximately 60 days. At this time point, 40% of mice showed complete response with no signs of tumor persistence. These experiments demonstrated that the dual-targeting CAR provided greater efficacy and survival benefit in the treatment of multiple myeloma in a mouse model.

Multiple myeloma is an appealing target-disease for CAR T therapy as it is a blood borne malignancy that responds well to other immune-based interventions, including monoclonal antibodies (daratumumab), immunomodulatory drugs, and stem cell transplantation (40). Clinical responses to BCMA CAR T cells in patients are encouraging, but relapse due to antigen-escape remains a challenge (41). Additional markers have been tested as CAR targets, some of which have reached the stage of clinical testing (11). However, most of these targets are also expressed to some extent on healthy tissues, raising the risk of “on-target off-tumor” toxicity. Molecular manipulations for tuning the affinity of the chimeric receptor have been tested in preclinical settings in an effort to increase CAR selectivity and overcome off tumor toxicity (42). Here, we described the dual targeting approach to develop anti–multiple myeloma CAR with high potency and increased selectivity. In this setting, two receptors are coexpressed in the same T-cell, each is directed against a distinct tumor target. This approach enhances tumor selectivity by targeting a pair of antigens; neither of them is uniquely expressed on malignant cells, but coexpression of both antigens is thought to be tumor specific. Support for this concept was provided by previous reports showing that CARs can be engineered to provide complementary signaling (43, 44). In this study, we developed a dual CAR targeted against two commonly coexpressed multiple myeloma markers, CD38 and CD138. CD138 is an abundant protein expressed on many body tissues (29, 30). Therefore, we incorporated the CD138 recognition unit into the CD28 chain, which is completely inert, as transmission of a costimulatory signal alone is not expected to trigger cytotoxic activity. The activating chain, on the other hand, which in our case targets CD38, resembles a first-generation CAR. Previous preclinical reports show that first-generation CARs mediate antitumor activity in vitro and in animal models. This was shown also in our work via the CD138-null control CAR that mimics a first-generation CAR transmitting an activating signal only. As shown both in vitro and in vivo, the CD138-null control could be stimulated by and exert a cytotoxic activity toward a multiple myeloma cell line. CD38 is expressed by some normal tissues, and therefore the dual CAR may have toxic potential (27, 28). Nevertheless, we envision that in a clinical setting, this toxic potential will be minimal for several reasons. First, according to the two-signal model of T-cell activation, the engagement of a co-stimulatory receptors lowers the T-cell activation threshold decreasing the number of TCRs (or in our case, the activating chain of the dual CAR) that have to be aggregated in the immunologic synapse before a proliferative signal can be transmitted (45). Therefore, dual CAR T cells are not expected to be efficiently activated when engaged with healthy cells expressing low levels of CD38. Indeed, our toxicity experiment showed that human primary cells activated a CD38-targeted second-generation CAR (in which the activating and the costimulatory signal are fused together on the same polypeptide chain), whereas the dual CAR was largely unaffected. Second, it was previously demonstrated that first-generation CARs have poor efficacy in humans due to limited T-cell proliferation and persistence (46, 47). Taken together, dual targeting should promote the selective enhancement of T-cell survival within the tumor niche, because synergistic signals would be preferentially delivered to T cells at that location by malignant cells densely “decorated” by the targeted antigens.

Activation of dual CAR T cells by CD38 as a single antigen may represent an obstacle during the production process in a clinical setting. Before the transduction step, T cells undergo robust activation that upregulates CD38 on their cell surface. This leads to auto-activation and CAR T-cell fratricide, reflected by reduced viability and expansion. Such self-killing may limit large scale expansion, which is required to reach therapeutic doses. Several attempts have been made to prevent fratricide, including CRISPR/Cas9 gene editing to delete the target on the CAR T (48). This approach, however, imposes a risk of generating a therapy-resistant cancer cells due to possible “contamination” of the patient's T-cell harvest with malignant cells. An alternative approach tested uses a blocking antibody to prevent engagement of the CAR with self-antigens during ex vivo expansion (49, 50). Accordingly, using daratumumab, a CD38 blocking mAb, during the production process may be an attractive solution in the case of anti–multiple myeloma dual CAR. Daratumumab is an approved drug for patients with multiple myeloma, and therefore, using it as a part of the manufacturing platform is not expected to affect product safety. Whether daratumumab can reduce T-cell fratricide and enable large-scale expansion with no negative effect on the activity of the dual CAR will be the subject of a later study.

Notably, we found that in addition to its increased selectivity, the dCAR138-38 showed consistently high expression rates and better in vivo performance compared with the CD38 and CD138 CARs. Whereas the inferiority of the CD38CAR might be due to fratricide and was predicted by the in vitro studies, the CD138CAR was less efficient in vivo but performed comparably with the dCAR138-38 in vitro. These findings were presumably attributed to the change in receptor configuration—separating the activating and costimulatory signals onto different receptors. A large body of work has been dedicated to study the role of each CAR component (scFv, its internal linker, hinge, transmembrane domain) and the role of structural and spatial principles in the overall performance of CARs, including their expression, stability, affinity and specificity to target molecule, clustering and interaction with other proteins and ultimately, signal transmission efficiency and intensity (51, 52). Studying the exact mechanism underlying the superiority of the dCAR138-38 is an interesting challenge to be explored in subsequent studies.

In summary, our results showed that the dual CAR concept was very effective against myeloma cells. We demonstrate that dual CAR T cells were both highly effective and safer than any other receptor tested, including two different single-antigen–directed CARs. The split configuration of this CAR is expected to reduce “on-target off-tumor toxicity.” We believe that our work has a high translational potential providing a safe alternative to BCMA-negative or relapsed patients with multiple myeloma.

A. Globerson Levin reports grants from Israel Science Foundation (ISF) and Dotan Hemato-Oncology Center at the Cancer Biology Research Center, Tel Aviv University during the conduct of the study, as well as a patent for PCT/IL2018/051325 pending (in advanced stages of negotiation of research collaboration and an option to license). M. Rawet Slobodkin reports grants from Israel Science Foundation (ISF; grant number 42/906) and Dotan Hemato-Oncology Center, Tel Aviv University during the conduct of the study, as well as a patent for PCT/IL2018/051325 pending (in advanced stages of negotiation of research collaboration and an option to license). T. Waks reports grants from Israel Science Foundation (ISF) and Tel Aviv University during the conduct of the study, as well as patent for PCT/IL2018/051325 pending (and an option to license). G. Horn reports grants from Israel Science Foundation (ISF) and Dotan Hemato-Oncology Center at the Cancer Biology Research Center, Tel Aviv University during the conduct of the study, as well as patent for PCT/IL2018/051325 pending (in advanced stages of negotiation of research collaboration and an option to license). L. Ninio-Many reports a patent for PCT/IL2018/051325 pending (and an option to license). N. Deshet Unger reports grants from Israel Science Foundation (ISF) and Dotan Hemato-Oncology Center at the Cancer Biology Research Center, Tel Aviv University during the conduct of the study, as well as a patent for PCT/IL2018/051325 pending (in advanced stages of negotiation of research collaboration and an option to license). Y. Ohayon reports grants from Israel Science Foundation (ISF) and Dotan Hemato-Oncology Center at the Cancer Biology Research Center, Tel Aviv University during the conduct of the study, as well as a patent for PCT/IL2018/051325 pending (in advanced stages of negotiation of research collaboration and an option to license). S. Suliman reports grants from Israel Science Foundation (ISF) and Dotan Hemato-Oncology Center at the Cancer Biology Research Center, Tel Aviv University during the conduct of the study, as well as a patent for PCT/IL2018/051325 pending (in advanced stages of negotiation of research collaboration and an option to license). Y. Cohen reports grants from Israel Science Foundation (ISF) and Dotan Hemato-Oncology Center at the Cancer Biology Research Center, Tel Aviv University during the conduct of the study, as well as a patent for PCT/IL2018/051325 pending (in advanced stage of negotiation of research collaboration and option to license). B. Tartakovsky reports grants from Israel Science Foundation (ISF; grant number 42/906) and Dotan Center, Tel Aviv University during the conduct of the study, as well as a patent for PCT/IL2018/051325 pending (option to license). E. Naparstek reports grants from Israel Science Foundation (ISF) and Dotan Research Center, Tel Aviv University during the conduct of the study, as well as a patent for PCT/IL2018/051325 pending (in advanced stage of negotiation of research collaboration and an option to license). I. Avivi reports grants from Israel Science Foundation (ISF) and Dotan during the conduct of the study, as well as a patent for PCT/IL2018/051325 pending (in advanced stage of negotiation research collaboration and an option to license). Z. Eshhar reports grants from Israel Science Foundation (ISF) and Tel Aviv University during the conduct of the study, as well as a patent for PCT/IL 2018/051325 pending (and option to license.)

A. Globerson Levin: Conceptualization, supervision, investigation, methodology, writing–original draft, project administration, writing–review and editing. M. Rawet Slobodkin: Conceptualization, formal analysis, investigation, writing–original draft, writing–review and editing. T. Waks: Conceptualization, investigation, writing–review and editing. G. Horn: Conceptualization, investigation, writing–review and editing. L. Ninio-Many: Conceptualization, investigation. N. Deshet Unger: Conceptualization, investigation. Y. Ohayon: Conceptualization, writing–review and editing. S. Suliman: Conceptualization, writing–review and editing. Y. Cohen: Conceptualization. B. Tartakovsky: Conceptualization. E. Naparstek: Conceptualization. I. Avivi: Conceptualization, writing–review and editing. Z. Eshhar: Conceptualization, supervision, writing–review and editing.

This work was supported in part by grant number 42/906 from the Israel Science Foundation (ISF) and by a grant from the Dotan Hemato-Oncology Center at the Cancer Biology Research Center, Tel Aviv University. The authors thank Shelley Schwarzbaum for her remarkable scientific editing.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.