Targeting CD20⁺ Aggressive B-cell Non–Hodgkin Lymphoma by Anti-CD20 CAR mRNA-Modified Expanded Natural Killer Cells In Vitro and in NSG Mice

B-cell non-Hodgkin lymphoma (B-NHL), including Burkitt lymphoma, makes up approximately 60% of all malignant NHL that occurs in children and adolescents (1). The outcome for children and adolescents with de novo mature B-NHL has improved significantly, as we previously demonstrated that short but intensive chemotherapy is associated with a 90% 5-year event-free survival (EFS; refs. 2–4). We demonstrated in an international multi-cooperative group study, which comprised the Children's Oncology Group (COG), The Société Française d'Oncologie Pédiatrique (SFOP), and the United Kingdom Children's Cancer Study Group (UKCCSG), that children and adolescents with newly diagnosed mature B-NHL [French–American–British (FAB)/Lymphome malins de Burkitt (LMB) 96], have a 90% 5-year overall survival (OS) following treatment with a short, intensive course of chemotherapy (4). Unfortunately, for children and adolescents who relapse or progress with de novo mature B-NHL, the prognosis is dismal due to chemoradiotherapy resistance (4, 5). Similarly, the prognosis in adults with recurrent/refractory Burkitt lymphoma is dismal (6). Therefore, development of alternative cellular targeted therapeutic strategies is required to improve outcomes in children, adolescents, and adults with recurrent/refractory de novo mature B-NHL.

CD20 is a glycosylated phosphoprotein expressed on the surface of B cells on all developmental stages, except pro-B cells or plasma cells (7). It is also expressed in >98% of childhood, adolescent, and adult mature B-cell NHLs and therefore is an attractive cancer therapeutic target (6, 8, 9). Rituximab, a chimeric anti-CD20 antibody, has been used successfully in the treatment of childhood and adolescent mature B-NHL as well as in adults with diffuse large B-cell lymphoma and Burkitt lymphoma (9–11). However, CD20 could also be used as a target for genetically engineered immune cell–based therapies.

Natural killer (NK) cells are bone marrow–derived cytotoxic lymphocytes that play a major role in the rejection of tumors and cells infected by viruses, even without specific immunization (12, 13). Various activating and inhibitory receptors on the NK-cell surface are engaged to regulate NK-cell activities and to discriminate target cells from healthy “self” cells (14, 15). However, factors limiting NK therapy include small numbers of active NK cells in unmodified peripheral blood and a lack of specificity in tumor targeting (16). Our group and others have successfully expanded active NK cells in vitro by short-term culture with cytokines alone, using cytokines and coculture with irradiated Epstein–Barr virus (EBV)–transformed lymphoblastoid cell lines as feeder cells, or using cytokines and coculture with K562 cells expressing transfected cell-membrane bound IL15 and 4-1BBL (17–20).

Chimeric antigen receptors (CAR) usually include a single-chain variable fragment (Fv) from a monoclonal antibody, a transmembrane hinge region, and a signaling domain such as CD28, CD3ζ, 4-1BB (CD137), or 2B4 (CD244) endodimers (21–24). The advantage of the CAR strategy is that no human leukocyte antigen (HLA) expression on the target cell is required for the epitope to be accessible to CAR⁺ T or NK cells. Thus, it is not limited to only a subset of patients with a specific HLA type (22). Several clinical trials testing CAR⁺ T cells in patients have shown promising clinical outcomes. Porter, Grupp, Brentjens, and their colleagues (23–25) have engineered patients' T cells with a lentiviral vector expressing anti-CD19 CAR and reinfused the anti-CD19 CAR autologous T cells into patients with refractory chronic lymphocytic leukemia and acute lymphoblastic leukemia. However, significant toxicities, including hypotension, fevers, renal insufficiency, cytokine-release syndrome, and B-cell aplasia, have occurred after infusions of CAR T cells (23–26). The hallmark toxicity of CAR T-cell therapies is cytokine-release syndrome, a potentially life-threatening complication of inflammatory symptoms resulting from elevated plasma cytokine levels associated with T-cell activation and proliferation (23–26). In contrast, NK cell–based immunotherapy has not been associated with a cytokine-release syndrome in patients; it has been associated with a significant NK versus leukemic effect in the absence of graft-versus-host disease (GvHD) and a significant decrease in leukemia relapse following haploidentical allogeneic stem cell transplantation in which the donor/recipient NK killer immunoglobulin-like receptor (KIR)/malignant major histocompatibility complex (MHC) class I mismatch may occur (27). Our approach to NK cell–based immunotherapy is to expand and activate NK cells and redirect them specifically against resistant lymphoma/leukemia cells. This approach might circumvent resistance to NK-cell tumor cytotoxicity independent of HLA class I ligand expression on leukemia cells, target NK cells to specific lymphoma/leukemic tumor antigens, enhance NK-cell activation, and provide an alternative to adoptive allogeneic T-cell immunotherapy by circumventing the risk of acute GvHD that commonly occurs.

Here, we investigated in vitro and in vivo activities of anti-CD20 CAR-modified expanded peripheral blood NK cells (exPBNK) following mRNA nucleofection, against CD20⁺ B-NHL.

The human cell lines Ramos, Daudi, RS4;11, U698M, K562, and the T-cell ALL lines, Jurkat, were purchased from the American Type Culture Collection (ATCC). Raji, rituximab-resistant Raji-2R, and Raji-4RH cells were generously provided by Matthew Barth and Myron Czuczman from Roswell Park Cancer Institute (Buffalo, NY; ref. 28). All tumor cells were maintained in RPMI-1640 (Gibco) supplemented with 10% FBS (Gibco) and antibiotics. The K562-mb15-41BBL cell line was generously provided by Dr. Dario Campana (St Jude Children's Research Hospital, University of Tennessee College of Medicine, Memphis, TN; ref. 17).

Leukocytes were obtained after informed consent from healthy adult donors at the New York Blood Center. Peripheral blood mononuclear cells (PBMC) were obtained by Ficoll gradient (Amersham Biosciences) separation. NK-cell expansion and isolation were performed as previously described (17) and are detailed in the Supplementary Methods.

Anti-CD20 CAR mRNA was generated as previously described with modifications (29). Briefly, the anti-CD20-4-1BB-CD3ζ was cut from an MSCV-anti-CD20-BB-ζ-IRES-GFP vector and subcloned into the pcDNA3 vector. Anti-CD20-4-1BB-CD3ζ mRNA was transcribed in vitro using the mMESSAGE mMACHINE T7 Ultra Kit (Life Technologies) following the manufacturer's instructions. The resulting product was dissolved in nuclease-free H₂O. RNA concentrations were measured using a Nano-spectrophotometer (Thermo Fisher) at 260 nm.

Six to 8 × 10⁶ expanded purified NK cells were suspended in 100-μL nucleofection solution (Lonza); anti-CD20-BB-ζ mRNA was added at 80 to 100 μg/mL. The mixture was placed in the Nucleofector Cuvette and nucleofected using the Amaxa Nucleofector II Device with U-001 program. CAR expression was detected using FITC-conjugated or AF647-conjugated goat anti-mouse IgG, F(ab′)₂ fragment-specific antibody (The Jackson Laboratory), and flow cytometry.

Intracellular CD107a expression and IFNγ were measured by flow cytometry, as previously described (20). Intracellular CD107a and IFNγ assays are detailed in the Supplementary Methods.

NK cytotoxic activity was determined by europium release assays with a standard kit (PerkinElmer), as previously described (20). In vitro cytotoxicity assays are detailed in the Supplementary Methods.

Six- to 8-week-old NOD/SCID/γ-chain^−/− (NSG) mice were purchased (Stock #5557; The Jackson Laboratory), bred, and maintained under pathogen-free conditions in-house at the Department of Comparative Medicine at New York Medical College (Valhalla, NY). All protocols were approved by the Institutional Animal Care and Use Committee.

The mammalian expression construct ffLUCZeo-pcDNA3.1 (generously supplied by L. Cooper, University of Texas MD Anderson Cancer Center, Houston, TX) was electroporated into tumor cells (Raji and Raji-2R), and stable clones were selected with Zeocin (Invitrogen). Raji or Raji-2R cells (5 × 10⁵) expressing luciferase (Raji-Luc or Raji-2R-Luc) were i.p. or s.c. injected into the NSG mice (NOD.Cg-PRkdc^scidIl2rg^tmWjl/SzJ; 6 weeks; The Jackson Laboratory). Tumor engraftment and progression were evaluated using the Xenogen IVIS-200 System (Caliper Life Sciences) after i.p. injection of 150 mg d-luciferin/kg/mouse. Photons emitted from luciferase-expressing cells were quantified using Living Image software. After tumor engraftment was verified, freshly prepared 5 × 10⁶ CAR exPBNK, mock exPBNK, or medium was i.p. injected into each mouse once a week for 3 weeks.

Tumor regression and/or progression of xenografted mice were monitored weekly by tumor volume measurement and by in vivo bioluminescent imaging (BLI; refs. 30, 31). Tumor size was estimated according to the following formula (32): tumor size (cm³) = length (cm) × width² (cm) × 0.5. Mice were followed until death or sacrificed if tumor size reached 2 cm³ or larger.

Fresh-frozen tumor samples were sectioned for hematoxylin and eosin (H&E) analysis and immunofluorescence staining as described in the Supplementary Methods.

TruCOUNT tubes (Becton Dickinson) were used to determine the absolute counts of tumor cells and NK cells in the peripheral blood of xenografted mice.

Peripheral blood was obtained via face-vein bleeding and stained for the presence of human CD45, CD20, and CD56 NK cells. After gating on the human CD45⁺ population, the CD20⁺ and CD56⁺ subsets were quantified using TruCOUNT tubes (BD Biosciences) with known numbers of fluorescent beads as described in the manufacturer's instructions.

Raji-2R-Luc cells (5 × 10⁵) were intravenously (i.v.) injected into NSG mice. Mice were sacrificed when hind-leg paralysis was observed. Organs (lung, liver, spleen, and kidney) were collected and soaked in D-PBS with 300 μg/mL d-luciferin for 5 minutes before BLI using the Xenogen IVIS-200 system (Caliper Life Sciences).

Statistical analyses were performed using the INSTAT statistical program (GraphPad). Average values are reported as the mean ± SEM. Results were compared using the one-tailed unpaired Student t test, with P < 0.05 considered as statistically significant. Probability of survival in animal studies was determined by the Kaplan–Meier method using the Prism program 5.0 (GraphPad Software, Inc.).

PBMNCs from healthy donors were cultured with irradiated geK562 feeder cell lines, as described (17), in the presence of 40 IU/mL IL2. As reported previously, CD56⁺CD3⁻ PBNK were significantly increased compared with those grown in media alone at day 7 (mean, 66.87% vs. 8.84%; n = 3; Fig. 1A). There was no significant enrichment of CD56⁺CD3⁺ PB NKT cells compared with those grown in media alone at day 7 (data not shown). CD56⁻CD3⁺ PB T cells were significantly reduced compared with those grown in media alone at day 7 (mean, 19.16% vs. 77.00%; n = 3; Fig. 1B). The absolute NK numbers were enhanced with irradiated geK562 as feeders compared with irradiated K562 and IL2 alone after normalized to the INPUT NK cell numbers (mean, 44.4-fold ±1.47 vs. 12.49-fold ±1.50 vs. 0.239-fold ±0.09; n = 3; P < 0.001; Fig. 1C). We also observed that sustained activation and proliferation by irradiated geK562 induced morphologic changes in NK cell shape and size (Fig. 1D). We examined the expression of CD3 and CD56 in the cells cocultured with media alone, K562, and geK562 by flow cytometry analysis (Fig. 1E). The exPBNK by geK562 were negatively selected with more than 96% purity (Fig. 1E).

The anti-CD20-BB-ζ chimeric receptor (anti-CD20 CAR) was originally constructed in the retroviral plasmid pMSCV-IRES-anti-CD20-BB-ζ (kindly provided by Dr. Dario Campana, St Jude Children's Research Hospital, University of Tennessee College of Medicine; Supplementary Fig. S1A). The expression of anti-CD20 CAR was confirmed in virus-infected Jurkat cells by flow cytometry and by Western blot analysis (Supplementary Fig. S1B and S1C). Anti-CD20 CAR-mediated cytotoxicity was first examined using NK-92 cells and CD20⁺ NHL. NK92 cells expressing anti-CD20 CAR had enhanced cytotoxicity against CD20⁺ NK-sensitive Ramos (74.66% ± 8.01% vs. 49.63% ± 5.10%; P < 0.05) and CD20⁺ NK-resistant Daudi (43.15% ± 4% vs. 16.15% ± 6.4%; P < 0.01) cells at E:T = 10:1 ratio but not against the CD20⁻ pre-B-ALL cell line RS4;11 (40.46% ± 1.14% vs. 41.21% ± 1.02%: P = ns) using europium release assays (Supplementary Fig. S1D).

After functional confirmation, the anti-CD20 CAR fragment was excised with restriction enzymes from the retroviral vector pMSCV-IRES-anti-CD20-BB-ζ and ligated into a T7 promoter–driven plasmid (Fig. 2A). Anti-CD20 CAR mRNA was in vitro transcribed, as described previously (Fig. 2B; ref. 33). Expanded CD56⁺CD3⁻ PBNK purified from the peripheral blood of 10 healthy donors were nucleofected in the presence of anti-CD20 CAR mRNA to generate anti-CD20 CAR exPBNK or in nuclease-free H₂O to generate mock exPBNK. Flow cytometry was used to detect the expression of anti-CD20 CAR in 66.73% ± 6.817% of viable exPBNK after 16 hours of nucleofection (Fig. 2C). Background from the exPBNK nucleofected with nuclease-free H₂O was 2.89% ± 0.947%. As expected, the anti-CD20 CAR expression was transient and was reduced from 94.1% at 20 hours to 60.1% at day 6 and 2.99% at day 11 after nucleofection (Fig. 2D).

CAR mRNA nucleofection, however, did not affect the expression of exPBNK-activating receptors (CD16, CD69, NKG2D, CD244, NKp30, NKp44, and NKp46) or inhibitory receptors (NKG2A, KIR2DS4, CD94, CD158a, CD158b, and CD158e; Supplementary Fig. S2).

We examined the functionality of anti-CD20 CAR mRNA-modified exPBNK against CD20⁺ B-NHL target cells. CD20 expression on the target tumor cells was confirmed (Supplementary Table S1). Anti-CD20 CAR exPBNK cytotoxicity was significantly enhanced compared with that of mock exPBNK against CD20⁺ B-NHL at 10:1 (n ≥ 3): Ramos (97.25% ± 2.61% vs. 82.5% ± 4.058%; P < 0.05), Daudi (71.5% ± 3.26% vs. 36.34% ± 6.31%; P < 0.001), and U-698-M (82.84% ± 1.17% vs. 26.2% ± 0.776%; P < 0.001; Fig. 3A). There was no significant difference against CD20⁻ RS4;11 or Jurkat cells (Fig. 3B). Tumor-cell recovery experiments further confirmed the result. Anti-CD20 CAR exPBNK significantly reduced Ramos cell recovery compared with those of medium and mock exPBNK (P < 0.01; Fig. 3C). These data demonstrate the targeting specificity of anti-CD20 CAR mRNA-modified exPBNK against CD20⁺ tumor cells.

We examined the expression of CD20 in rituximab-sensitive Raji and rituximab-resistant Raji-2R and Raji-4RH cells by flow cytometry analysis (Supplementary Table S2). In addition, the standard europium release assay experiments demonstrated that the cytotoxicity of CAR exPBNK was significantly higher than that of mock exPBNK (P < 0.001) against Raji, Raji-2R, and Raji-4RH cells (Fig. 3D).

Daudi and Raji cells have been demonstrated previously to be resistant to NK-mediated cytotoxicity (34). NK resistance of Raji cells is, in large part, mediated through the interaction of KIRs and histocompatibility class I antigens (HLA; ref. 35). Unlike Raji cells, Daudi cells do not synthesize β2-microglobulin and lack the cell-surface HLA antigens (36). Therefore, resistance of Daudi cells to NK-mediated lysis has not been thought to be mediated through HLA antigens. The mechanism of NK resistance in Daudi cells is currently unknown. We found that our exPBNK express KIR2DL1, KIR2DL2/3, KIR3DL1, and NKG2A (Supplementary Fig. S2 and data not shown). Raji, Raji-2R, and Raji-4RH cells used in our experiments express high levels of HLA class I antigens by flow cytometry analysis (Fig. 3E and Supplementary Table S3). The anti-CD20 CAR exPBNK showed enhanced cytolytic activities against NK-resistant Daudi and Raji cells (Fig. 3), indicating that the expression of anti-CD20 CAR on NK cells can overcome inhibitory signals from KIR–HLA interactions and other unknown mechanisms-mediated resistance.

We also observed similarly enhanced expression of exPBNK-activating receptors (CD69, NKp44, and NKG2D) after incubation with U-698-M, compared with that with medium (Supplementary Fig. S3), and the inhibitory receptors (CD94 and KIR2DL2/3) were unchanged in mock exPBNK and CAR exPBNK.

The lysosomal-associated membrane protein-1 (LAMP-1 or CD107a) is a marker of NK-cell functional activity (37). Intracellular CD107a expression in exPBNK was detected by flow cytometry. CD107a degranulation was enhanced in anti-CD20 CAR exPBNK, compared with that in mock exPBNK, in response to CD20⁺ Ramos (31.47% ± 1.74% vs. 15.2% ± 0.26%; P < 0.001; n = 3) and Daudi (38.9% ± 2.7% vs. 19.73% ± 0.58%; P < 0.001; n = 3) stimulation; however, there was no significant difference in response to RS4;11 (5.69% ± 0.45% vs. 3.9% ± 0.06%; P = ns; n = 3) or medium (4.86% ± 0.066% vs. 4.75% ± 0.59%; P = ns; n = 3; Fig. 4A). Similarly, intracellular IFNγ production was also enhanced in CAR exPBNK compared with that in mock exPBNK in response to CD20⁺ Ramos-specific (3.07% ± 0.42% vs. 1.39% ± 0.26%; P < 0.05; n = 3) and Daudi-specific (5.62% ± 0.52% vs. 1.42% ± 0.43%; P < 0.05; n = 3) stimulation in a short 4-hour incubation (Fig. 4B).

We assessed the antitumor activity of exPBNK nucleofected with anti-CD20 CAR mRNA using a humanized Burkitt lymphoma xenograft NSG model. First, we used a dissected tumor model. The purity and kinetics of BLI signal in Raji-Luc were examined (Supplementary Fig. S4A and S4B). Raji-Luc cells (5 × 10⁵) were injected i.p. into NSG mice. After successful engraftment of Raji-Luc cells in mice at day 7, freshly prepared 5 × 10⁶ mock or CAR exPBNK were injected i.p. into each mouse once a week for 3 weeks (days 9, 16, and 23). Mice that received culture medium were used as controls. We demonstrated that after the third NK-cell injection, the luciferase signals measured in the CAR exPBNK-treated Raji-Luc group (n = 6) were significantly reduced, compared with those in the medium-treated mice (P = 0.0087; n = 5) and in the mock exPBNK-treated mice (P = 0.0128; n = 8; Fig. 5A and B). Consistent with the reduced luciferase signals, the CAR exPBNK-treated mice appeared to be more healthy and active than the medium-treated and the mock exPBNK-treated mice (data not shown). The survival curve of mice was built based on death of mice caused by tumor cells disseminated in the whole body from the injection site or sacrifice if tumor size reached 2 cm³ or larger. We found that the CAR exPBNK-treated Raji-Luc mice had significantly extended survival time with a median of 40 days compared with that of the untreated mice (29 days, P < 0.001) and of the mock exPBNK-treated mice (30 days, P < 0.001; Fig. 5C). There was no significant difference in survival between the untreated Raji-Luc mice and the mock exPBNK-treated Raji-Luc mice with the current scheduled NK doses.

The proliferation of transferred CAR T cells is highly correlated with tumor regression (23–25). To examine whether the antitumor efficacy of anti-CD20 CAR NK cells is correlated with increased proliferation and long persistence of anti-CD20 CAR PBNK compared with that of the mock PBNK in NSG mice, peripheral blood was collected from Raji-Luc xenografted mice after 7 days of the third anti-CD20 CAR NK injection. Circulated tumor cells and anti-CD20 CAR NK cells were quantified with TruCOUNT beads by flow cytometry analysis with anti-CD20-PE and anti-CD56-FITC antibodies (Fig. 5D). CD20 Raji cell counts were significantly reduced in mice receiving CAR exPBNK, compared with medium (P < 0.001) and mock (P < 0.05) treatment groups (Fig. 5D, left). Mock exPBNK also significantly reduced the CD20 Raji cell counts, compared with the medium treatment group (P < 0.001; Fig. 5D, left). However, there was no significant difference in the number of NK cells between mice receiving CAR exPBNK and mock exPBNK (Fig. 5D, right). We examined whether the delayed tumor growth was due to the colocalization of CAR exPBNK with the solid tumor masses. Solid tumor masses were dissected from the medium-treated group, the mock exPBNK-treated group, and the CAR exPBNK-treated group, and confirmed by BLI (Fig. 5E, top). H&E staining was used to examine the morphology of tumor tissues from each treatment group (Fig. 5E, bottom). Tumor burden of the CAR exPBNK-treated group appeared to be less compact with tumor cells compared with the other two groups. A fluorescent immunostaining analysis was conducted to detect the presence of NK cells in the tumor masses. We found a few green fluorescent Alexa Fluor 488–labeled NK cells accumulating around the edge of the tumors (Fig. 5F) in mock exPBNK-treated mice and in CAR exPBNK-treated mice but not inside of the tumor masses.

We used the localized tumor model to examine the antitumor effect of CAR exPBNK against rituximab-resistant Burkitt lymphoma Raji-2R cells. The purity and kinetics of BLI signal in Raji-2R-Luc was examined (Supplementary Fig. S4C and S4D). Raji-2R-Luc cells (5 × 10⁵) were injected s.c. into NSG mice. Mock or CAR exPBNK (5 × 10⁶) were i.p. injected into each mouse once a week for 3 continuous weeks (days 3, 10, and 17). Mice that received culture medium were used as control. After the third NK injection, luciferase signal was significantly reduced in the CAR exPBNK-treated Raji-2R-Luc group, compared with the mock exPBNK-treated mice (P < 0.01; Fig. 6A and B). Tumor size measured in the CAR exPBNK-treated Raji-2R group (n = 7) was also significantly smaller than that in the medium-treated mice (P = 0.0175; n = 6) and the mock exPBNK-treated mice (P = 0.0122; n = 7; Fig. 6C). Thereafter, the survival curve was established based on the endpoint of 2 cm³ tumor size. The CAR exPBNK-treated Raji-2R-Luc (n = 7) mice had significantly extended survival time with a median of 24 days, compared with that of the medium-treated (18 days; n = 6; P < 0.001) and of the mock exPBNK-treated mice (22 days; n = 7; P < 0.05; Fig. 6D).

To study the effect of anti-CD20 CAR exPBNK in tumor migration and multiorgan dissemination in xenografted mice, 5 × 10⁵ Raji-2R-Luc cells were i.v. injected into NSG mice. Seven days later, 5 × 10⁶ mock or CAR exPBNK in culture medium with 200 IU/mL of IL2 were i.v. injected into NSG mice with similar tumor burdens once a week for 2 weeks. In the three pairs of NSG mice treated with mock exPBNK or CAR exPBNK, CAR exPBNK-treated NSG mice had less bioluminescence signal in the pointed organs than that in the mock exPBNK-treated NSG mice (Fig. 7A), suggesting that CAR exPBNK therapy inhibited Raji-2R-Luc dissemination to other organs. Consistent with this observation, DiD (a lipophilic fluorescent dye, C67H103CIN2O3S)-labeled exPBNK were able to migrate to the lung, liver, and spleen after i.v. injection to NSG mice, as confirmed by fluorescent imaging using the IVIS imaging system (Supplementary Fig. S5A) and by flow cytometry analysis (Supplementary Fig. S5B). To confirm this, organs (lung, liver, spleen, and kidney) were collected in D-PBS with d-luciferin for BLI. CAR exPBNK-treated NSG mice had less bioluminescence in the lung, liver, kidney, and spleen than that of the mock exPBNK-treated NSG mice (Fig. 7B).

Adoptive immunotherapy with NK cells has been limited in the past by the low numbers of activated NK cells in the peripheral blood and concomitantly by the lack of tumor-targeting specificity. In this study, we established conditions that expanded activated PBNK ex vivo and modified the exPBNK efficiently with anti-CD20 CAR mRNA in a nonviral, clinically relevant method. The anti-CD20 CAR expression in exPBNK by mRNA nucleofection was associated with a statistically significant increase in CD107a degranulation and INFγ production after stimulation with CD20⁺ B-NHL. Consequently, CAR exPBNK treatment results in significant and specific in vitro cytotoxicity against CD20⁺ B-NHL, including both rituximab-sensitive and rituximab-resistant B-NHL cells. Multiple injections of anti-CD20 CAR mRNA-electroporated exPBNK mediated significant tumor regression and inhibited tumor growth in both disseminated and localized tumor models in NSG mice. More importantly, multiple injections of anti-CD20 CAR-modified exPBNK also extended survival in B-NHL tumor-xenografts and inhibited tumor-cell migration. These results indicate the therapeutic potential of multiple injections of anti-CD20 CAR mRNA-modified exPBNK against CD20⁺ B-NHL in patients.

Gene-engineered T cells with viral methods have been used for cancer therapy with success, but concerns about safety has arisen after X-SCID patients developed cancerous T cells after receiving hematopoietic stem cells transduced with recombinant retrovirus (38), and manufacturing challenges may also hamper clinical progress. The development of safe, efficient, and nonviral methods with a lower tendency to insert near oncogenes will significantly facilitate targeted cellular therapies (33, 39). Synthetic modified mRNA in vitro has great therapeutic potential for transient, targeted protein expression with limited toxicity. In addition, large quantities of RNA can be easily prepared by in vitro transcription, which makes it possible to expedite manufacturing and scaling to current Good Manufacturing Practices (cGMP) products. Successful electroporation of CAR mRNA into primary T cells and NK cells has now been documented for a variety of cancers, such as anti-CD19 CAR against leukemia (29, 39) and mesothelin CAR against pancreatic cancer (clinical trial: NCT01897415). However, these synthesized mRNAs result in only short-lived CAR expression in immune cells. As shown in Fig. 2D, anti-CD20 CAR expression lasted less than 2 weeks after nucleofection in exPBNK. Considering the lifespan of adoptively transferred NK cells is only approximately 10 to 12 days even with cytokine support (Supplementary Fig. S5C; refs. 40–42), transient expression of CAR in exPBNK may not be a significant disadvantage and it does not need a suicide gene to limit the risk of the long-term side effect using retroviral or lentiviral vectors. In our xenograft models, three injections of anti-CD20 CAR mRNA-electroporated exPBNK significantly mediated CD20⁺ tumor regression, extended survival (Fig. 5B and C), reduced tumor size of CD20⁺ tumor xenografted mice (Fig. 6C), and inhibited tumor-cell migration to other organs in NSG mice (Fig. 7). However, under current NK doses and injection frequencies, we did not obtain tumor-free animals. The ratio of CAR NK cells to target cells may be critical to the success, as shown in our in vitro cytotoxicity assays (Fig. 3). The remaining tumor cells in the xenografted mice after three NK injections are still CD20⁺ (data not shown), and the relapse is not likely a consequence of CD20 deregulation to escape anti-CD20 CAR NK cytotoxicity. In the future, NK cells expanded with a new feeder K562-mb21- 41BBL and increased injection frequency of higher doses of CAR NK will be considered to enhance NK in vivo expansion and antitumor effect.

To explore the possible mechanisms of antitumor efficacy of mRNA CAR exPBNK, we first examined whether CAR exPBNK proliferate better and persist longer than mock exPBNK. Within our expectations, CAR exPBNK significantly reduced the number of circulating tumor cells in the peripheral blood of mice, compared with that in the medium (P < 0.001) and mock (P < 0.05) treatment groups (Fig. 5D, left), but there was no significant difference in NK cell numbers among mice receiving CAR exPBNK and mock exPBNK (Fig. 5D, right), which may be due to the lack of proliferative capacities of NK cells. Second, we examined whether mRNA CAR exPBNK colocalize better than mock exPBNK in solid tumor masses to inhibit tumor growth. The tumor burden of the CAR exPBNK-treated group appeared to be less compact with tumor cells than that of the other two groups (Fig. 5E), and the tumor size is smaller in the CAR exPBNK-treated group than that in the other two groups (Fig. 6C). We found a few green fluorescent Alexa Fluor 488–labeled NK cells around the edge of the tumors (Fig. 5F) in both mock exPBNK-treated mice and CAR exPBNK-treated mice. We did not find mock or mRNA CAR exPBNK inside of the tumor masses (data not shown). This is consistent with the recent study in which Singh and colleagues (43) have substantiated the difference between permanent CAR T cells and mRNA-based CAR T cells, and the transiently modified T cells were unable to significantly penetrate inside of the tumor masses. Therefore, anti-CD20 CAR mRNA-modified exPBNK can mediate effective antitumor responses in vivo; however, to achieve complete tumor eradication with mRNA-based CAR exPBNK therapies, we will pursue studies combinating CAR exPBNK therapy with other approaches, such as bispecific antibodies.

Tumor cells may escape from NK-cell immunosurveillance and develop resistance by several mechanisms. One mechanism is associated with the downregulation of natural cytotoxicity receptor expression on NK cells (44, 45). Our ex vivo–expanded human PBNK with geK562 upregulated activating receptors, including CD69, NKp30, NKp44, and NKG2D (Supplementary Fig. S6). These exPBNK can mediate cytotoxicity of allogeneic and autologous cancer cell lines by specifically activating receptor–ligand interactions. Another mechanism of tumor-cell resistance is by upregulating the expression of KIR-inhibitory ligands (46). In our study, we found that the expanded NK cells express KIRs, such as KIR2DL1, KIR2DL2/3, and KIR3DL1, and Raji, Raji-2R, and Raji-4RH cells express high-level HLA-ABC (Supplementary Fig. S2 and Fig. 3E). The KIR receptor–HLA ligand interactions mediate resistance to NK-mediated lysis. Previous studies reported that resistance to NK-mediated lysis can be overcome by preincubation of NK cells with IL2 and the generation of “lymphokine-activated killer cells” (34). Our study demonstrates that anti-CD20 CAR-modified exPBNK have significantly enhanced cytotoxicity against NK-resistant CD20⁺ Raji, Raji-2R, and Raji-4RH cells but not against NK-resistant CD20-RS;11 cells (Fig. 3), demonstrating that anti-CD20 CAR-mediated CD20-dependent signaling can overcome the inhibitory signaling from the interaction between inhibitory receptors and HLA ligands. The third mechanism of tumor cells escaping from NK-cell immunosurveillance is by downregulating the NKG2D activation signals such as shedding of NKG2D ligands from tumor cells (47). We are currently investigating the combination of enhancing NKG2D ligand expression on tumor cells and treating with anti-CD20 CAR-modified exPBNK (48).

Promising clinical results have been obtained using rituximab to treat B-NHL because rituximab was approved by the FDA in 1997 (49). However, approximately 60% of patients who responded to their first rituximab therapy, relapsed following rituximab retreatment (50). One of the relapsed mechanisms is the reduced expression of CD20 on the surface of cancer cells after they are repeatedly exposed to rituximab. Despite significantly reduced CD20 antigen expression in rituximab-resistant cells, compared with that in sensitive cells, CD20 is still relatively high in rituximab-resistant cells (Supplementary Table S2; ref. 28). We attempted to assess whether anti-CD20 CAR-mediated exPBNK cytotoxicity would overcome rituximab resistance. Interestingly, anti-CD20 CAR exPBNK had significantly enhanced cytotoxicity against rituximab-resistant Burkitt lymphoma cells in vitro (Fig. 3D), and anti-CD20 CAR exPBNK significantly inhibited the growth of rituximab-resistant Burkitt lymphoma cells in xenografted NSG mice (Fig. 6), compared with that in mice treated with mock exPBNK. Our results indicate that anti-CD20 CAR NK cells can bind more efficiently to the reduced level of CD20 on resistant tumor cells than rituximab, and the anti-CD20 CAR NK cells can be activated to lyse the resistant tumor cells efficiently. Moreover, we observed that the anti-CD20 CAR exPBNK limited the dissemination/migration of rituximab-resistant Raji-2R tumor cells to multiple organs, compared with the mock exPBNK (Fig. 7).

In conclusion, our data confirm that activated expanded allogeneic PBNK become highly cytolytic, killing resistant CD20⁺ B-leukemia/lymphoma after nucleofection with anti-CD20 CAR mRNA. We propose that the interaction between anti-CD20 CAR on exPBNK and CD20 on the surface of B-leukemia/lymphoma transmits an activating signal back to the exPBNK and significantly stimulates exPBNK degranulation and release of cytokines such as IFNγ to lyse resistant tumor targets (Supplementary Fig. S7). Results from our preclinical studies should assist in designing clinical trials with CAR exPBNK to treat patients with CD20⁺ B-leukemia/lymphoma without the associated safety concerns of integrating viral vectors. In future studies, we will examine the persistence of expanded NK cells in vivo and assess the combined therapeutic effect of anti-CD20 CAR-modified NK cell–based therapies with the new generation of antibody-based therapies.

No potential conflicts of interest were disclosed.

Conception and design: Y. Chu, J. Hochberg, M.S. Cairo

Development of methodology: Y. Chu, J. Hochberg, A. Yahr, J. Ayello

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): Y. Chu, A. Yahr, M. Barth

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): Y. Chu, J. Hochberg, A. Yahr

Writing, review, and/or revision of the manuscript: Y. Chu, J. Hochberg, J. Ayello, C. van de Ven, M. Barth, M. Czuczman, M.S. Cairo

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): Y. Chu, J. Ayello, C. van de Ven, M. Czuczman, M.S. Cairo

Study supervision: M.S. Cairo

The authors thank Erin Morris, RN, for her excellent assistance with the preparation of this article; Dr. Dario Campana (St. Jude Children's Research Hospital) and Dr. Terrence Geiger (St. Jude Children's Research Hospital) for kindly providing anti-CD20 scFv; Dr. Carl Hamby (New York Medical College) for sharing equipment and reagents; and Anne Sollas (Core Histology Lab, Pathology Department, New York Medical College) for technical support.

This research was supported by grants from the Pediatric Cancer Research Foundation (PCRF), Children's Cancer Fund (CCF), and a New York Medical College Intramural Research Award.

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.