CB-NK and peripheral blood T cells were obtained from the Banc de Sang i Teixits de Barcelona (BST) from either cord blood (CB) or peripheral blood from healthy donors after obtaining informed consent and being approved by the BST.

Ramos, RPMI8226, U266 and K562 cell lines were purchased from American Tissue Culture Collection (Manassas, Virginia, USA). ARP1 cell line was kindly provided by Multiple Myeloma Research Center (Little Rock, Arkansas, USA). RPMI8226, K562 and ARP1 were cultured in RPMI with 10% fetal bovine serum (FBS) and 1% penicillin/streptomycin (Pen/Strep) and U266 with 15% FBS. K562-based artificial antigen-presenting cells expressing membrane-bound Interleukin (IL)-21 (K562-mb21-41BBL feeder cells) were cultured in RPMI with 10% FBS and 1% Pen/Strep. Tumor MM cell lines were modified to express GFP-FireFlyLuciferase (GFP-FFLuc) using the plasmids pLV-MSCV_Luc-T2A-GFP (kindly provided by Amer Najjar) coding for GFP-FFLuc, and pMD2.G and psPAX2 coding for VSV-G and gag/pol, respectively. Mycoplasma testing in cell lines was performed every 2 months.

For CAR-T cells, either pCCL-EF1a-CAR19 or pCCL-EF1a-CARBCMA with packaging plasmids pMDLg-pRRE, pRSV-Rev and envelope plasmid VSV were used. For CAR-NK cells, pCCL-MSCV-CAR19 with pMDLg-pRRE, pRSV-Rev and envelope plasmid BaEV-TR (provided by Dr Els Verhoeyen) were used for production of virus particles. Production of virus particles to transduce T cells was done in HEK293T cells using the same protocol previously described by us.14 Production of virus for CAR-NK cells was done in HEK293T cells that were transfected with transfer vector pCCL-MCSV-CAR19 with pMDLg-pRRE, pRSV-Rev and BaEV-TR plasmid. JetPEI (PolyPlus Transfection) was used as a transfection reagent. Transfection was performed using NaCl and JetPei following the manufacturer’s protocol. Viral supernatants were collected at 48 hours and concentrated with Lenti-X Concentrator (ClonTech) following the supplier’s protocol and stored at −80°C until use.

K562-mb21-41BBL feeder cells were used to expand untransduced NK cells. They were irradiated at 100 Gy and added at a feeder cell:CB-NK ratio of 2:1. CB-NK were left to expand for 14 days, with new feeder cells added on day 7.

T cells were obtained from buffy coats by Ficoll and magnetic T cell depletion (Miltenyi Biotec). T cells were expanded in Click’s media (50% RPMI, 50% Click’s (Irvine Scientific), 5% human serum, 1% Pen/Strep), activated with Dynabeads Human T-Activator CD3/CD28 (Thermo Fisher Scientific) and IL-2 (100 IU/mL) every other day. Experiments were performed after 8–10 days of T cell expansion. Viral transduction was performed after 48 hours of the expansion at a MOI of 10. NK cells were selected from CB units by Ficoll and magnetic NK cell depletion (Miltenyi Biotec). To produce CAR-NK (ARI3 cells), NK cells from a CB unit were expanded with NK MACS Medium (Miltenyi Biotec), 1% NK MACS Supplement, 5% human serum, IL-2 (500 IU/mL) and IL-15 (140 IU/mL). On day 5, CB-NK were transduced at a multiplicity of infection (MOI) of 5 with Vectofusin-1 (10 mg/mL) following Vectofusin-1 protocol (Miltenyi Biotec). Cells were washed 6–24 hours after transduction. On day 7, ARI3 cells were coexpanded with feeder cells adding IL2 (400 IU/mL) every other day for the next 7 days. Of note, throughout the manuscript, ‘ARI3 cells’ refers to CAR-NK cells and CB-NK refers to untransduced NK cells.

Immunodeficient NSG mice were purchased from The Jackson Laboratory. Mice used for experiments were of the same sex and 8–12 weeks old. Mice were irradiated on day 1 at a dose of 2 Gy in a biological radiator with two sources of 137 Cs (J.L. Shepherd, model MARK 1, 30 serial number 1199) and received itravenous tumor cells modified to express GFP-FFLuc on day 0. Different models of tumor burden were created for NHL depending on the dose of NHL cells administered. Thus, mice received 1×10⁵, 3.5×10⁵, or 5×10⁵ NHL cells to create the low, medium, or high tumor burden models. Then, they received immune cells on day 7 (for NHL) or day 14 (for MM). The disease was followed weekly by bioluminescence as previously described.20 Bone marrow (BM) and spleen were harvested when mice were euthanized to analyze the presence of CAR-T cells and MM cells by flow cytometry.

Luciferase killing assays were performed by coculturing target tumor cells expressing GFP-FFLuc with effector cells tested in each case in a white-walled 96-well plate. D-luciferin (20 mg/mL) was added 15 min prior to the bioluminescence reading in a Synergy HT Plate Reader (BioTek). The percentage of remaining tumor cells was calculated as (luminescence of sample/luminescence of tumor cells alone) ×100.

Interferon Gamma (IFNγ), tumor necrosis factor alpha (TNFα) and IL-2 were quantified by ELISA (ELISA MAX Deluxe Set, Biolegend) following the manufacturer’s protocol. Luminex assays were conducted with ProcartaPlex multiplex immunoassay kit (Thermo Fisher Scientific).

RPMI8226 cell line modified to express GFP was cocultured with CART cells stained with Cell Tracker Blue CMAC Dye (Thermo Fisher Scientific). Either CB-NK or non-transduced T cells (UT) were stained with CellTracker Deep Red Dye (Thermo Fisher Scientific) and added to the previously mentioned coculture. Images were acquired using a Leica SP5 microscope. 405, 488 and 633 lasers were used for excitation. For time lapse experiments, in vivo image acquisitions were performed every 30 s for 14 hours.

ARP1-GFP-FFLuc (MM) cells were cocultured with CART cells at an effector:target ratio 0.5:1. In the condition with CB-NK, CB-NK were added at half the amount of CART cells, leaving a final ratio of CART:MM(:NK) 0.5:1(:0.25). The remaining tumor cells were assessed by fluorescent microscopy until no GFP was observed, and thus the challenge was over. T lymphocytes were then stained, and the phenotype was compared with that seen in unstimulated T cells (day 0) and CAR-T cells before exposure to tumor cells.

CAR expression was detected with either a recombinant BCMA-Fc protein (Enzo Life Sciences) or a recombinant CD19-Fc protein (Thermo Fisher Scientific) and an antihuman IgG Fc-BV-421 (Biolegend, clone: M1310G05). For T cell phenotyping, PD-1-APC (clone: J105), TIM-3-FITC (clone: F38-2E2), TIGIT-PerCP-Cy5.5 (clone: A15153G), LAG-3-PE (clone: 11C3C65), CXCR3-Alexa-Fluor 488 (Clone 1C6/CXCR3), CCR7-PerCP-Cy5.5 (clone: 150503), CD45RA-APC (clone: HI100), CD27-PE (clone: M-T271), CD28-FITC (Clone CD28.2), CD3-PE (clone SK7), CD4-APC-H7 (clone L200) and CD8-PE-Cy7 (clone: RPA-T8) antibodies were used. For NK cell phenotyping, CD16-Alexa488 (Clone 3G8), NKP30-PeCy7 (Clone P30-15), NKG2D-APC-Cy7 (Clone 1D11), NKG2A-PacificBlue (Clone S19004C), NKG2C-Pe (Clone S19005E), KIR2DL1-PE (Clone HP-DM1), KIR2DL2/L3-FITC (Clone DX27), KIR3DL1-BV421 (Clone DX9) and KIR2DL1/S1/S3/S5-PerCp Cy5.5 (Clone HP-MA4) were all from Biolegend. CD56-APC (Clone REA196, Miltenyi) was used for NK cells. Staining was performed in FACS buffer. For T cell determination in BM or spleen at in vivo end-point, mouse FcR blocking reagent (Miltenyi) was used. All experiments were read on a FACS Canto II (BD Biosciences) and analyzed with FlowJo software (Tree Star, Eugene, Oregon, USA).

Data were presented using GraphPad Prism software, and statistical analysis was performed using SPSS software V.20 using Student’s t-test to compare between two groups.

We first optimized a method to obtain a high number of CAR-NK cells directed against CD19 (ARI3 cells) starting from CB-NK. We used a CAR directed against CD19 with 4-1BB as costimulatory domain previously produced by our institution termed ARI-0001 (ARI1 from now on).21 Plasmids coding for ARI1 and ARI3 differed in the promoter (online supplemental figure 1A) as CB-NK presented higher transduction efficiencies using the MCSV promoter instead of the EF1a promoter. Another difference between ARI1 and ARI3 was the packaging plasmid that was used in the transduction. The use of a Vesicular Stomatitis Virus (VSV) packaging plasmid for virus production led to extremely low transduction efficiencies of CB-NK. Interestingly, Bari et al22 demonstrated that exchanging VSV for Baboon Envelope (BaEV) packaging plasmid improved efficiencies of NK cell transduction. Therefore, VSV and BaEV were compared confirming improved efficiency of CAR transduction with BaEV (online supplemental figure 1online supplemental figure 1B). Then, two different expansion methods were compared with BaEV (figure 1A). In the first method (A.1), CB-NK were expanded with IL2 and IL15 over 5 days. On day 5, CAR transduction was performed. On day 7, K562-mb21-41BBL feeder cells were added to CB-NK, and they were left to expand until day 14 with the addition of IL2 every other day. In the second method (A.2), CB-NK were expanded with IL2 and feeder cells for 5 days. On day 5, CAR transduction was performed. On day 7, CB-NK were left to expand until day 14 with the addition of fresh feeder cells and IL2 every other day. Method A.1 achieved a higher percentage of ARI3+ cells than method A.2 (figure 1B), and therefore, method A.1 was selected for all ARI3 cell productions.

The anti-NHL activity of both CAR-T cells and CAR-NK cells directed against CD19 (ARI1 for CAR-T cells and ARI3 for CAR-NK cells: online supplemental figure 1A was compared in vitro. At higher effector:target (E:T) ratios, ARI3 cells presented higher antitumor activity than untransduced CB-NK or ARI1 cells (figure 1C). However, at low E:T ratios, ARI3 efficacy was significantly lower to that of ARI1 cells (figure 1D), being this difference even more pronounced at longer times (24 hours vs 48 hours). Lack of efficacy of NK cells at longer times might reflect their requirement for IL2, which is not present in these assays. Therefore, we confirmed that addition of exogenous IL2 improved the in vitro efficacy of ARI3 cells, especially at later time points (figure 1D). In terms of cytokine production, at higher E:T ratios, ARI3 cells produced a higher amount of IFNγ than CB-NK, with this production being increased further with the addition of IL2. However, production of IFNγ by ARI3 cells with and without IL2 was always lower than that of ARI1 cells (figure 1E). Regarding TNFα, this cytokine is rapidly produced at early time points and decays fast.14 While ARI3 cells produced higher TNFα levels than CB-NK, the highest TNFα production was detected for ARI1 cells (figure 1F).

ARI3 cells required IL2 at low E:T ratios to kill tumor cells, and interestingly, CAR-T cells produce high amounts of IL2. Moreover, we previously demonstrated that CB-NK can enhance the antitumor activity of untransduced (UT) T cells.20 Therefore, we hypothesized that ARI1 CAR-T cells would boost the activity of ARI3 CAR-NK cells, and vice versa, which would lead to an overall enhanced antitumor efficacy. Thus, combination of ARI1 and ARI3 cells, where the total number of immune cells was the same as ARI1 cells alone (0.5+0.5 vs 1), demonstrated increased anti-NHL activity (figure 1G) without impacting in higher IFNγ (figure 1H) and TNFα production (online supplemental figure 1C). Of note, when the total number of immune cells combining ARI1 and ARI3 was double that of ARI1 cells alone (1+1 vs 1), a higher anti-NHL activity and IFNγ production (figure 1G and H) was observed with hardly any changes for TNFα (online supplemental figure 1C). Last, long-term in vitro cytotoxicity assays performed with a low E:T ratio demonstrated superiority of this combinatorial treatment (ARI1+ARI3) versus ARI1 cells alone when the total number of immune cells was the same in both treatments (figure 1I). Of note, IFNγ levels were not increased (figure 1J).

The impact of IL2 in the activity of ARI3 cells and the combinatorial treatment were further evaluated in vivo in different NHL tumor burden models. In a model of highly aggressive NHL, none of the immune cells administered (CB-NK, ARI1 or ARI3) were able to prevent disease progression at all (online supplemental figure 2A and 2B). Of interest, analysis of mice tissues demonstrated that ARI3 cell treatment maintained the bone marrow (BM) with the lowest levels of disease in comparison with untreated mice (online supplemental figure 2C). However, mice treated with either CB-NK or ARI3 cells presented a higher percentage of tumor cells in the spleen compared with both untreated and ARI1-treated animals (online supplemental figure 2C).

In a model of low tumor burden, treatment with either CB-NK or ARI3 cells alone was not able to stop disease progression (figure 2A–2C). ARI3 cells maintained the BM free of disease (figure 2D), but as previously observed in the high tumor burden model (online supplemental figure 2C), a high number of NHL cells were detected in the spleen (figure 2D). The addition of exogenous IL2 to ARI3 cells was highly detrimental by causing NHL progression, as likely due to the fact that IL2 also induces B cell proliferation,23 with this group showing a dramatic increase in the number of NHL cells in the BM compared with ARI3 cells without exogenous IL2 and a high number of NHL cells in the spleen (figure 2D). These findings were in agreement with the role of IL2 promoting the presence of B cells in the BM and also in the spleen but to a lesser degree.24 In all cases, the high number of NHL cells in the spleen correlated with splenomegaly (figure 2E). Moreover, NK cells were detected in the spleen in all groups treated with NK cells, so the presence of tumor cells in the spleen was not due to lack of NK cell migration (figure 2F). Treatment with either ARI1 cells alone or combined with ARI3 cells did not demonstrate differences in this model of low tumor burden (figure 2A–2D) as both treatments completely prevented NHL progression. However, higher IFNγ production was observed in the combinatorial treatment at 31 days. No increase was seen at 24 days (figure 2G). Finally, even though, in these two groups, there were low numbers of human-derived T cells in the BM and the spleen (figure 2H), the presence of CAR+ cells could be detected in both groups with no significant difference observed between groups (figure 2I).

A third in vivo model of NHL with medium levels of tumor burden was performed to find out whether the combinatorial treatment based on ARI1+ARI3 was better than ARI1 cells alone. In this model of disease, the highest number of ARI1 cells was determinant to prevent disease progression, as groups treated with ARI1 cells alone retained the disease progression for longer time than the combinatorial treatment with half dose of ARI1 cells, leading to a higher survival of the mice (figure 3A–3C). In the peripheral blood (PB), 15 days after CAR treatment, a slight tendency for a higher production of IFNγ was detected for the combinatorial treatment, although it was not significant, and there were no differences in the number of T cells present in the PB, although it must be noted that the combinatorial treatment had half the dose of T cells (figure 3D). Analysis of mice tissues demonstrated that both treatments maintained the BM free of disease. In addition, tumor cells were not detected in the spleens of the ARI1 alone group, but a very low number of NHL cells was detected in the spleen of mice receiving the combinatorial treatment (figure 3E). In both groups, NK cells could not be detected, and the presence of T cells in both the BM and the spleen was extremely low (figure 3F), although CAR-T cells were still present (figure 3G). Of note, mice receiving the combinatorial treatment did not present a clear NHL population in the tissues. However, a population with low expression of GFP and CD19 (figure 3H), suggestive of a dying NHL cell population, was present, which might explain the higher bioluminescence signal detected in the combinatorial treatment (figure 3A and B).

Our results determined that ARI3 cells presented much lower efficacy than ARI1 cells at low E:T ratios, which translated into low in vivo efficacy (figures 1 and 2). Moreover, the combinatorial treatment based on ARI1 and ARI3 cells showed that the number of ARI1 cells and the starting tumor cell burden was determinant to prevent disease progression (figure 3). Nevertheless, we previously determined that CB-NK are able to increase the antitumor activity of UT T cells by bringing UT T cells and MM cells into close proximity.20 Thus, we hypothesized that addition of CB-NK to CAR-T cells would increase the antitumor efficacy of CAR-T cells. Of note, patients receiving CAR-T cells receive a previous lymphodepleting treatment based on fludarabine and cyclophosphamide that removes NK cells. Moreover, NK cells represent around 10% of PB leukocytes, a much lower value than for T lymphocytes.25 Thus, we decided to design a treatment that includes NK cells, specifically with half the dose of CB-NK with respect to that of CAR-T cells. This hypothesis was tested in a model of MM and humanized CAR-T cells directed against BCMA (ARI2h cells). Addition of CB-NK to ARI2h cells increased the anti-MM in vitro efficacy. Of interest, this beneficial impact was enhanced when the E:T (ARI2h:MM(:NK)) ratios were lower (1:1(:0.5)) vs 0.5:1(:0.25)) in figure 4A. Moreover, UT T cells and ARI2h cells did not augment tumor killing to the same extent as with CB-NK when they were added at the same ratios, demonstrating that the highest beneficial effect occurred with the addition of CB-NK (figure 4A), especially at earlier time points (figure 4B). Of interest, IFNγ, TNFα and IL2 levels in the supernatant were not increased with the addition of CB-NK, being in some cases even slightly lower (figure 4C). Specifically, IFNγ production by either CD4+ or CD8+ T cells in these coculture assays did not increase when CB-NK were present (figure 4D). Furthermore, long-term cytotoxicity assays performed at a very low E:T ratio (1:8), confirmed the beneficial impact of CB-NK which even prevented the regrowth of MM cells (figure 4E). Of note, the pool of ARI2h cells that patients receive contains UT T cells. Therefore, we analyzed and confirmed that CB-NK also enhanced the anti-MM activity of UT T cells (figure 4F). Moreover, at low E:T ratios, the anti-MM activity of either UT or CB-NK alone was barely detectable and was much higher when both populations were combined (figure 4F).

The beneficial impact of CB-NK over ARI2h cells was confirmed in an in vivo model for MM in terms of lower disease progression and higher survival (figure 5A–5C). Despite the beneficial impact obtained with CB-NK and ARI2h cells, mice treated only with CB-NK showed a high proportion of MM cells in the BM and the spleen (figure 5D). Moreover, IFNγ levels and total T cells in the PB of mice 28 days after treatment (figure 5E), and the presence of T cells in the BM and the spleen (figure 5F), did not differ between the ARI2h and ARI2+CB-NK groups.

NK cells, as part of the innate immune system, perform their activity at earlier time points than T cells,1 which was concordant with the highest impact of CB-NK over ARI2h cells observed at early time points (figure 4B). Therefore, we performed time-lapse in vivo imaging to visualize differences in the activity of ARI2h cells in the presence or absence of CB-NK over a period of 14 hours. As a control, UT T cells were added at the same number and in place of CB-NK. In vivo imaging showed that CB-NK accelerated the migration of ARI2h cells to MM cells, an effect that was not observed with UT T cells (online supplemental movies, figure 6A–6C). Specifically, CB-NK migrate incredibly quickly towards tumor cells (within the first 22 min in figure 6B) leading to enhanced formation of immune/tumor cell clusters. In turn, this promoted the migration of ARI2h cells to MM cells (figure 6C). Moreover, whereas CB-NK came into close proximity with tumor cells remarkably quickly, they started to distance from these clusters when tumor cells had already been eliminated (from 382 min onward in figure 6B). In parallel, movements of UT T cells added to ARI2h cells were hardly visible (online supplemental movies, figure 6B).

The enhanced migration of CAR-T cells to MM cells due to the presence of CB-NK suggested a faster activation of CAR-T cells at earlier time points. Therefore, CD69 expression, a marker of T cell activation, was measured demonstrating a higher activation of CD4+ T cells in the presence of CB-NK (figure 6D and E). Furthermore, the secretion of cytokines in these cocultures was analyzed at 6 hours demonstrating a different secretome in the presence of CB-NK (figure 6F) and that cytokines involved in T cell migration, such as CCL3 and CCL5, were increased in the presence of CB-NK (figure 6G). Of interest, cytokines related to CRS development, such as TNFα, were not increased in the presence of CB-NK (figure 6F).

Last, we analyzed how the impact of CB-NK on ARI2h cells observed at early time points (figure 6) would influence ARI2h cells exposed to MM cells in long-term challenge assays. These long-term assays were performed at low E:T ratios adding cells highly diluted so that the killing would take longer (figure 7A). These experiments were performed with six different donors for T cells and CB-NK from three different CB units. Of interest, the impact of CB-NK over the activity of T cells was not related to differences observed in the expression of some NK cell receptors (online supplemental figure 5). We observed that after encountering MM cells, the total number of CAR^pos and CAR^neg CD8+ cells was higher in the presence of CB-NK, suggesting either a higher expansion or a lower cell death (figure 7B and C). The transition of memory stages of T cells was analyzed after production of ARI2h cells and after challenging to MM cells. In the naïve/stem cell memory (SCM) compartment, a decrease in the proportion of naïve CAR^neg CD4 and CD8 T cells was observed during the expansion, with no changes observed in the SCM compartment (online supplemental figure 4A).

In the central memory (CM) and effector memory (EM) compartments, at day 0, before production of ARI2h cells, there was a lower proportion of CM in comparison with EM T cells in both CAR^pos and CAR^neg T cells (figure 7D and online supplemental figure 4B). Of note, CM T cells present higher capacity for homing to secondary lymphoid organs, while EM T cells are more cytolytic and express integrins and chemokine receptors necessary for localization to peripheral tissues.26 The production/expansion of ARI2h cells induced an increase in CM and a decrease in EM for CAR^pos CD4 and CD8 T cells (figure 7D). For CAR^neg T cells, the expansion only caused an increase in the CM compartment (online supplemental figure 4B). After exposure to MM cells, the proportion of CM CAR^pos CD4 T cells decreased concomitant with an increase in EM CAR^pos cells (figure 7D). In the presence of CB-NK, these changes were more pronounced (figure 7D), suggesting a more differentiated phenotype and a faster capacity of CAR-T cells to migrate where MM cells are emerging. For CAR^pos CD8 T cells, exposure to MM cells caused an increase in the EM compartment only in the presence of CB-NK (figure 7D). Of interest, no changes were observed in the number of terminally differentiated effector ARI2h cells (figure 7D). However, an increase in effector cells was noticed for both CAR^neg CD4+ and CD8+ T cells that was more pronounced for CD8+ T cells in the presence of CB-NK (online supplemental figure 4B).

Moreover, the fitness of ARI2h cells after exposure to MM was analyzed by looking at exhaustion and immunosenescent parameters. The presence of CB-NK decreased the expression of exhaustion markers PD-1, TIM3 and LAG3 in both CAR^pos and CAR^neg T cells. Furthermore, TIGIT levels were also reduced in CAR^pos CD4+ T cells and both CAR^neg CD4+ and CD8+ T cells (figure 7E).

Loss of CD27 and CD28, known as markers related to the development of immunosenescence, specifically after chronic viral infection,27–29 were also improved in the presence of CB-NK. Specifically, exposure to MM cells induced an increase in the emergence of CD27⁻CD28⁻ cells in CAR^pos T cells (figure 7F), a detrimental impact that was reduced in the presence of CB-NK for the CAR^pos CD8+ T cell population (figure 7G). Senescent cells are also characterized by presenting a higher cell size, an event that impairs cell proliferation.30 Importantly, production/expansion of ARI2h cells induced an increase in both cell size and complexity for all CAR^pos and CAR^neg T cells that continued to rise after exposure to MM cells (figure 7H). CB-NK avoided the increase in cell size and cell complexity of both CAR^pos and CAR^neg CD8⁺ and CD4⁺ T cells after exposure to MM cells (figure 7I) suggesting a delayed emergence of senescence in ARI2h cells due to the presence of CB-NK.