Introduction

Chimeric antigen receptor T cell (CAR-T) therapy has emerged as one of the most innovative and promising advanced immunotherapy treatments in patients with refractory hematological malignancies, leading to the rapid approval of CAR-T cell products for patients with B cell acute lymphoblastic leukemia, non-Hodgkin’s lymphoma, and multiple myeloma.1–5 However, despite a high overall response rate (>80% of CR), many patients relapse with a median progression-free survival (PFS) of 1 year6 and become refractory to virtually all drugs available. The lack of long-term PFS plateau has been associated with several mechanisms, being the poor persistence of CAR-T cells a major limitation for the success of this therapy.

There is much evidence suggesting that T cell replicative potential and differentiation status are important regulators of antitumor activity.7 The high degree of differentiation of T lymphocytes after the CAR-T manufacturing process may be one of the reasons for the poor persistence of CAR-T cells.8 The incorporation of cytokines such as IL7 and IL15 can favor the generation of cell products with higher self-renewal capacity and therapeutic potential.9–11 The in vivo administration of cytokines such as IL2 or IL15 can provide survival and proliferation signals to CAR-T cells, but their therapeutic dosage is associated with systemic toxicities. In an effort to minimize this problem, the fourth generation of CAR-T cells (T-cells Redirected toward Universal Cytokine Killing, TRUCK-T) has emerged as an alternative approach. These modified T cells express the cytokine locally, allowing for a combination of the CAR-T cells direct antitumor attack with the immune-modulating capabilities of the delivered cytokine. The TRUCK concept is currently being explored using a panel of cytokines, including IL7, IL12, IL15, IL18, IL23, and combinations thereof, and is entering early phase trials.12 Among them, IL15 has prompted special interest because of its relevance in the homeostatic maintenance of long-lived CD8+ memory T cells,13 and its capacity to enhance the antitumor activity of CD8 T cells in vivo.14 Constitutive expression of IL15 has been shown to improve the antitumor activity of CAR-T cells specific for CD19, GPC-3, CLL-1, GD2 and IL13Rα2,15–19 likely due to a combination of greater expansion and persistence.

In addition to transgenic expression of secretory IL15, which may still have implications for systemic toxicity,20 the expression of a membrane-bound form of IL15 has emerged as an alternative way to provide prosurvival signals to CAR-T cells.21 In this approach, the native full-length IL15 peptide is joined to the full-length IL15Rα sequence using a flexible linker. This fusion protein mimics the natural presentation of IL15 with IL15Rα, creating an autocrine loop that enhances T cell persistence and produces potent antitumor effects without apparent signs of toxicity.21 However, many of these studies have been tested in immunodeficient murine xenograft models and using human CAR-T cells expressing the human version of IL15 or the tethered IL15-IL15Rα chimeric proteins. Toxicity studies of CAR-T cells expressing human proteins in immunocompromised mice may not provide a complete assessment, as it may fail to detect activation loops and adverse reactions involving cytokines that do not have cross-species reactivity. In this work, we evaluated the antitumor efficacy of a murine anti-CD19 CAR-T that expresses the murine chimeric membrane-bound form of IL15-IL15Rα fusion protein in immunocompetent BALB/c as well as in immunocompromised NSG mice and evaluate their possible toxicity.

Materials and methodsMice and cell lines

BALB/c and NSG mice (NOD.Cg-Prkdcscid Il2rgtm1Wjl/SzJ) were obtained from Harlan Laboratories and The Jackson Laboratory, respectively, and bred in our animal facilities at the Centro de Investigación Médica Aplicada. All animal handling and tumor experiments were approved and conducted under the institutional guidelines of our institutional ethics committee (Comité Ético de Experimentación Animal, Universidad de Navarra, Gobierno de Navarra, Ref: 019-19) and following the European Directive 2010/63/EU. The murine cell line A20 was obtained from ATCC and cultured in complete medium (RPMI 1640 or Dulbecco′s Modified Eagle′s Medium (DMEM) containing 10% fetal bovine serum (FBS), antibiotics, 2 mM glutamine and 50 µM 2-ME). The Platinum Ecotropic (Plat-E, ATCC) cell line was cultured in DMEM supplemented with 10% FBS at 37°C in a humidified atmosphere with 6.5% CO2.

Plasmids and retroviral transduction

pRubiG retroviral vector backbone was employed to build the anti-murine CD19 CAR-T. The construct combined the murine CD8 signaling peptide sequence, the anti-murine scFv (1D3 hybridoma) including the VDJ-H and the VJ-L fragments), the CD8 hinge and transmembrane regions, the costimulatory domain of CD28, the CD3ζ chain and the eGFP (figure 1A). A control vector to express PSMA CAR-T including the scFv from J591 clone specific for human PSMA was also produced. The CD19/mbIL15q pRubiG retroviral vector was prepared introducing the chimeric sequence containing the Igκ leader peptide, followed by the murine IL15 linked to the murine IL15Rα with a Gly-Ser linker and a C-terminal FLAG to monitor the expression of the chimeric protein. This chimeric construct was added to the plasmid CD19 CAR-T removing the GFP-expressing gene and including the F2A self-cleaving peptide sequence (figure 2A). A pRubiG plasmid expressing anti-PSMA instead of anti-CD19 was also generated to be used as control (PSMA/mbIL15q pRubiG). For retrovirus productions, Plat-E cells were transfected with 5 µg of retroviral plasmid DNA along with 2.5 µg pCLEco plasmid DNA using lipofectamine 2000 (Invitrogen) for 6 hours in antibiotic-free medium. Retroviral supernatants were collected at 48 and 72 hours after transfection.

Figure 1

In vitro and in vivo characterization of CD19 CAR-T cells. (A) Schematic representation of CAR constructs. (B) Example of flow cytometric analysis of CAR transduction percentage in murine CD8+ T cells (GFP+ as indicator of CAR+ T cells). Example of CAR detection in the membrane of transduced CD19 CAR CD8+ T cells (double-positive cells for GFP and Anti-Mouse IgG, F(ab')₂ staining were considered CAR+). (C–G) In vitro functional assays after stimulation of CD8+ CAR-T cells with recombinant mCD19-Fc protein-coated plates. (C) CAR-T proliferation assay measured by 3H-thymidine incorporation (cpm), (D–F) IFN-γ production of CAR-T cells in response to recombinant mCD19-Fc protein (D) or to A20 tumor cells (E) measured by ELISA (D–E) or by ELISPOT (F). (G) Relative percentage of A20 tumor cell lysis at different effector CAR-T: target ratios (after substraction of unspecific lysis with irrelevant CAR-T cells). (H–J) In vivo effect of CD19 CAR-T cells in the percentage of CD19+ cells (left y-axis) and percentage of CAR-T cells (right y-axis) in the blood of naïve mice preconditioned with high TBI (H), Low TBI (I) or no TBI (J) at different time points after CAR-T cell administration (n=7 mice per group). Graphs represent mean±SEM. The results are representative of 2–3 experiments performed independently. ****p<0.0001; ***p<0.001; **p<0.01; *p<0.05. TBI, total body irradiation.

Figure 2

In vitro characterization of CD19/mbIL15q CAR-T cells. (A) Schematic representation of CD19/mbIL15q CAR construct. (B) Expression of mbIL15q in the CAR-T cells was measured by flow cytometry using anti-FLAG antibodies (C) pSTAT5 staining in CD4+ and CD8+ CAR-T cells. (D) Phenotypic characterization of CAR-T cells regarding CD44 and CD62L staining measured by flow cytometry. (E–G) In vitro characterization of CD19/mbIL15q CAR-T cells function. (E) IL2 production by CD4+ or CD8+ CAR-T cells in response to recombinant mCD19-Fc protein coated to the culture plate. (F) Number of IFN-γ producing CD4+ or CD8+ CAR-T cells and specific lytic activity (G), in response to A20 tumor cells. Graphs represent mean±SEM. The results are representative of 2–3 experiments performed independently. NL: non-labeled, Irrelevant protein: irrelevant protein (ovalbumin). One-way or two-way ANOVA with Bonferroni multiple test correction was employed for statistical analysis. ****p<0.0001; ***p<0.001; *p<0.05. ANOVA, analysis of variance.

Antibodies and flow cytometry

Samples were tested with a FACS Canto II flow cytometer (Becton Dickinson) and data were analyzed by FlowJo software (TreeStar). Methods for CAR detection, CAR-T viability and phenotyping using fluorochrome-conjugated antibodies are described in online supplemental methods section.

Functional characterization of CAR-T cells

The functional characterization of CAR-T cells, including their ability to recognize CD19, proliferate and produce IFN-γ or IL2 in response to antigen stimulation, or their capacity to lyse A20 tumor cells was carried out as previously described22–24 and in online supplemental methods.

In vivo assessment of CAR-T cell activity

CAR-T cell activity was measured in vivo in naïve mice as well as in mice bearing A20 lymphoma, analyzing CD19+ B cell counts and mice survival. BALB/c mice were sub-lethally irradiated (Total Body Irradiation, TBI) with 1–2 Gy (Low TBI) or with 4 Gy (High TBI) depending on the experiment. Then, mice were treated with a mixture of 6×106 CAR-T cells (4:1 CD4+/CD8+ T cell ratio) by intravenous route as described in online supplemental methods. When indicated, liver, spleen, lung, kidney and heart were collected, embedded in paraffin and analyzed by H&E and by immunochemistry against murine CD3 was performed for toxicity evaluation.

Statistical analysis

Normality was assessed with Shapiro-Wilk W test. Statistical analyses were performed using parametric Student’s t test, two-tailed paired t-test, and one-way analysis of variance with Bonferroni multiple comparison test, as indicated. Mann-Whitney U and Kruskal-Wallis were used for non-parametric analyses. For all tests, a p<0.05 was considered statistically significant. Descriptive data for continuous variables were reported as mean±SEM. GraphPad Prism V.7 (GraphPad Software) was used for statistical analysis; *p<0.05, **p<0.01, ***p<0.001.

ResultsConventional antimurine CD19 CAR-T cells have a low persistence and poor efficacy in immunocompetent mice

A recombinant retrovirus coding for a second-generation CAR against the murine CD19 antigen was prepared (figure 1A). An irrelevant CAR that recognized human PSMA was used as control. These retroviruses were used to generate the murine CD19 CAR-T and the PSMA CAR-T cells. GFP+ expression after 3 days of expansion showed a high percentage of T cell transduction (above 70% in both cases). CAR expression on T cell membrane was confirmed using anti-Mouse IgG, F(ab')₂ staining (figure 1B).

CAR-T cells were tested for their capability to recognize the recombinant murine CD19-Fc protein coated to culture plates. As observed, mCD19 CAR-T but not PSMA CAR-T cells, proliferated and produced IFN-γ in response to mCD19 antigen (figure 1C,D, respectively) in a dose-dependent manner. CD19 CAR-T cells produced IFN-γ on stimulation with the CD19+ lymphoma cell line A20, as quantified using ELISA (figure 1E) or ELISPOT (figure 1F). Furthermore, CD19 CAR-T cells effectively recognized and lysed CD19-expressing A20 tumor target cells (figure 1G).

Once confirmed the functionality of the antimurine CD19 CAR-T cells in vitro, we studied the effectiveness of CD19 CAR-T cells to eliminate CD19 expressing B cells in vivo in immunocompetent BALB/c mice. Mice received a high dose (4 Gy) of TBI before the adoptive transfer. A group of irradiated mice were left untreated to evaluate the impact of TBI in CD19+ cells total numbers (TBI group). High TBI irradiation significantly decreased the total count of CD19+ cells, which returned to normal levels within 2–3 weeks. However, the adoptive transfer of mCD19 CAR-T cells achieved a CD19 cell depletion that lasted for more than 35 days after CAR-T cell transfer. The CD19 CAR-T cells were able to expand transiently after the in vivo transfer and remained detectable until day 35 (figure 1H).

It has been reported that effective CAR-T cell therapy is highly dependent on lymphodepleting preconditioning, which depletes endogenous lymphoid cells removing competition for lymphoproliferative cytokines and space in lymphoid tissues.25 To evaluate the impact of the lymphodepleting regimen in our model, we tested the CD19+ cells depletion efficacy when mice were treated with a low radiation therapy (Low TBI, 1–2 Gy). As shown in figure 1I, there was an initial drop in CD19+ cells in mice treated with CD19 CAR-T cells. However, this drop was also found in the TBI group. Nevertheless, this initial decrease in the B cell percentage was restored, reaching normal levels at day 17. In this setting, the transferred CD19 CAR-T cells were unable to expand. When the adoptive transfer was carried out in mice not receiving any preconditioning regimen (0 Gy), CD19 CAR-T cells were almost undetectable and did not have any significant impact on CD19+ total count numbers (figure 1J). These results suggest that, although the anti-murine CD19 CAR-T cells are functional, the in vivo effect is very limited, probably because of their poor persistence or expansion capacity in vivo under these experimental conditions. These results are in agreement with previous reports showing that it is necessary to a high lymphodepletion regimen (5 Gy TBI) prior to administration of CD19 CAR-T cells to detect their presence in circulation 1 week after the administration.26

The murine IL15-IL15Rα fusion protein enhances the persistence and in vivo efficacy of CD19 CAR-T cells

Since the conventional CD19 CAR-T cells did not efficiently expand in vivo to exert their cytolytic activity, we generated a retrovirus to coexpress CD19 CAR together with a chimeric fusion protein consisting of the murine IL15 linked to its receptor IL15Rα (figure 2A), similar to that previously described membrane-bound human tethered IL15-IL15Rα.21 As a control, we also generated a plasmid expressing this IL15-IL15Rα chimera and the PSMA CAR. This IL15-IL15Rα fusion protein (henceforth mbIL15q) might provide an autocrine signal for T cell proliferation that could improve their in vivo persistence. Thus, by means of retrovirus infection with this plasmid, we generated the CD19/mbIL15q CAR-T cells. CAR expression was confirmed by flow cytometry using anti-Mouse IgG, F(ab')₂ staining. The percentage of T cells expressing the CAR construct varied from 44% to 60% depending on the T cell type and the construct. For this reason, we purified CAR+ T cells by flow cytometry for their further characterization (online supplemental figure 1). The expression of mbIL15q in the CAR-T cells was confirmed by flow cytometry using anti-FLAG antibodies (figure 2B).

The functionality of the mbIL15q chimera in CD4+ and CD8+ CAR-T cells was confirmed measuring the intracellular levels of STAT5 phosphorylation, a downstream signaling event of IL15R activation. PhospoSTAT5 was significantly higher in CD19/mbIL15q CAR-T cells in comparison to CD19 CAR-T cells (figure 2C). Phenotypic analysis indicated a more T central memory (TCM) phenotype (CD44+ CD62L+) in CD19/mbIL15q CAR-T cells (figure 2D).

We then evaluated the functionality of these new CAR-T cells by measuring the IL2 production of both CD4+ and CD8+ T cells expressing the CD19/mbIL15q CAR or the conventional CD19 CAR, in response to stimulation with CD19-Fc coated plates (figure 2E). Although both CAR-T cell types responded to stimulation with CD19 in a dose-response manner, CD19/mbIL15q CAR-T cells produced higher levels of IL2 than CD19 CAR-T cells. These differences were also observed for IFN-γ production when CAR-T cells were stimulated with A20 lymphoma cell lines in CD4+ but not in CD8+ CAR-T cells (figure 2F). There were no significant differences in the lytic activity between the two CAR-T cell groups (figure 2G). Notably, when CAR-T cells were cultured in the absence of antigen stimulation, there was a basal-like proliferation of CD19/mbIL15q CAR-T cells measured by thymidine incorporation (online supplemental figure 2), a result consistent with the proliferative capacity of IL15 signaling through STAT5.27

We carried out a comparative transcriptomic analysis of CD19, CD19/mbIL15q CAR-T and PSMA/mbIL15q CAR-T cells after stimulation with A20 cell line (1:1 ratio). There was a differential expression of 646 genes when comparing CD19/mbIL15q CAR-T vs CD19 CAR-T, and 766 genes when comparing PSMA/mbIL15q CAR-T versus CD19 CAR-T cells (figure 3A). First principal components (PC1) analyses accounted for 30% of the variance between CD4+ and CD8+ CAR-T cells (figure 3B) and revealed also important variation between CD19, CD19/mbIL15q CAR-T and PSMA/mbIL15q CAR-T cells. We represent in a heat map the clustering of genes associated with the presence of mbIL15q in the CAR irrespective of CAR-T activation by the antigen (present in CD19/mbIL15q CAR-T and PSMA/mbIL15q CAR-T) and the clustering by antigen stimulation (common genes between CD19/mbIL15q CAR-T and CD19 CAR-T). It can also be observed a group of genes differentially expressed only in the CD19/mbIL15q CAR-T group, both in CD4+ and in CD8+ compartments (figure 3C). It was found an enrichment in genes associated with glycolysis, IL15 stimulation, IL2 STAT5 and mTORC1 signaling in those CAR-T expressing the mbIL15q (figure 3D) confirming the functionality of the chimera.

Figure 3

RNA-seq transcriptomic analysis in CD4+ CAR-T cells after co-culture with A20 cells. RNA-seq analysis was performed in in vitro samples of PSMA/mbIL15q, CD19 and CD19/mbIL15q CAR-T cells after co-culture with A20 cells at a 1:1 (CAR-cell:tumor cell) ratio. After 48 hours of co-culture, CD4+ and CD8+ CAR-T cells were purified separately for RNA-seq analysis. (A) Venn’s diagram of genes with differential expression (DE), p<0.05. (B) Principal component analysis (PCA) of the RNA seq. (C) Heat map of gene expression across the analyzed samples. (D) Pathways enrichment analysis for the different CAR-T.

After testing the activity of CAR-T cells in vitro, we evaluated the capacity of CD19 and CD19/mbIL15q CAR-T cells to deplete CD19+ B cells in vivo in immunocompetent mice. BALB/c mice were first irradiated with a low dose of TBI (1 Gy) and then treated with CAR-T cells (figure 4A). As observed in figure 4B, there was an initial drop in the percentage of CD19+ cells in blood, from days 2 to 9 after CAR-T cell therapy in all treated groups, including mice treated only with TBI. Nevertheless, from day 16 and on, there was a recovery of these percentages up to control levels in all the groups except the one treated with CD19/mbIL15q CAR-T cells. Indeed, mice treated with CD19/mbIL15q CAR-T cells experimented a rapid drop in the CD19 numbers in the blood with an almost complete CD19+ cell lymphodepletion from day 30 until the end of the experiment at day 58 post-ACT. Similar results were observed in the bone marrow (figure 4C) and the spleen (figure 4D) at day 40 after CAR-T transfer. In comparison to mice treated with CD19 CAR-T, we also observed a decrease in the percentages of neutrophils and monocytes, as well as an increase in myeloid cells in mice treated with CD19/mbIL15q CAR-T cells (online supplemental figure 3). The CD19 depletion efficacy for CD19/mbIL15q CAR-T was equally observed when mice were treated with CD4+ or CD8+ CAR-T cells as monotherapy (online supplemental figure 4). The incorporation of the mbIL15q construct into the CD19 CAR-T cells improved very significantly their in vivo persistence. Indeed, at day 40 after transfer, we could detect around 60% CAR-T among total CD8+ T cells in the spleen of mice treated with CD19/mbIL15q CAR-T as opposed to only 15% in mice treated with CD19 CAR-T (figure 4E).

Figure 4

In vivo characterization of CD19/mbIL15q CAR-T cells. (A) Schematic representation of the experimental setting. (B) In vivo effect of CAR-T cells in percentage of CD19+ cells in the blood of naïve mice preconditioned with Low TBI (1 Gy) at different time points after CAR-T cell administration (n=7 mice per group). (C, D) % of CD19+ cells in the bone marrow (C) and the spleen (D) at day 40 after treatment. A representative example of flow cytometry plots for CD19+ staining for each experimental group is shown. (E) CAR-T cell numbers were measured in the spleen of mice at day 58 after CAR-T cell administration. Graphs represent mean±SEM. The results are representative of two experiments performed independently. One-way or two-way ANOVA with Bonferroni multiple test correction was employed for statistical analysis.****p<0.0001. ANOVA, analysis of variance; TBI, total body irradiation.

CD19/mbIL15q CAR-T cells eradicate established A20 B cell lymphoma

Having established that CD19/mbIL15q CAR-T cells were able to exert their anti-CD19 lytic activity in immunocompetent mice previously treated with a low TBI regimen, we evaluated their antitumor activity in mice bearing A20 B-cell lymphoma (figure 5A). As previously shown in figure 1H, untreated mice maintain a percentage of CD19+ cells ranging the 20%–30% of total CD45+ cells. Mice treated with TBI alone experienced a transient drop in the percentage of CD19+ cells during the first 2 weeks that returned to normal levels during the third week of follow-up. This CD19+ cell drop was a little bit longer in mice treated with CD19 CAR-T but also returned to normality after 5 weeks. However, mice treated with CD19/mbIL15q CAR-T cells showed an almost complete CD19+ cell depletion which did not recover throughout the duration of the experiment (figure 5B). More importantly, and in agreement with these results, only mice treated with CD19/mbIL15q CAR-T cells showed a protective effect against A20 tumor challenge, with four out of seven mice alive (57%) at the end of the experiment (day 100 after tumor challenge) (figure 5C).

Figure 5

Therapeutic efficacy of CD19/mbIL15q CAR-T cells in the A20 lymphoma murine model. (A) Schematic representation of the experimental setting. Mice were challenged intravenously with 106 A20 tumor cells, and, 7 days later, they were treated with low TBI and CAR-T cells (CD19 CAR-T or CD19/mbIL15q CAR-T cells). (B) Percentage of CD19+ cells in the blood of mice preconditioned with Low TBI at different time points after CAR-T cell administration (n=7 mice per group). (C) Mice survival after CAR-T cell therapy. (D) H&E staining of the spleens and macroscopic aspect of the spleen in mice were sacrificed when the symptoms of the disease were appearing (from days 25 to 50). (E) Schematic representation of the experimental setting for (F–I). (F) Number of IFN-γ-producing cells in the splenocytes of mice in response to stimulation with A20 tumor cells (ratio 10:1, lymphocytes: tumor). (G–I) Percentage of IFN-γ+/GrzB+ cells (G) PD1+ cells (H) or CD137+ cells (I) in the tumor-draining lymph node cells of mice treated with the different CAR-T. Graphs represent mean±SEM. One-way or two-way ANOVA with Bonferroni multiple test correction was employed for statistical analysis. ***p<0.001; *p<0.05. ANOVA, analysis of variance; TBI, total body irradiation.

Spleens were collected from different groups of mice when the symptoms of disease were appearing (from days 25 to 50) and a H&E staining was performed. It was found that only CD19/mbIL15q CAR-T treated mice conserved a normal spleen architecture, comparable to naïve control mice. Moreover, mice treated with TBI alone, or with TBI plus CD19 CAR-T, but not with CD19/mbIL15q CAR-T, developed a tumoural spleen (figure 5D). Nevertheless, as opposed to that described for the same strategy using human IL15-IL15Rα expressing anti-human CD19 CAR-T tested in NSG immunodeficient mice,21 significant splenomegaly was observed in our immunocompetent BALB/c mice treated with CD19/mbIL15q CAR-T (figure 5D).

To better characterize the immunophenotype of the CAR-T cell therapy against A20 cells, we generated A20 solid tumors by s.c. inoculation. Ten days after T cell transfer, mice were sacrificed to characterize the phenotype of CAR-T cells (figure 5E). An ELISPOT assay indicated that mice treated with CD19/mbIL15q CAR-T cells had a greater number of IFN-γ-producing cells in response to irradiated A20 tumor cells (figure 5F). Similarly, we found a significantly higher percentage of IFN-γ+GrzB+ CD4+ or CD8+ CAR-T cells in mice treated with CD19/mbIL15q CAR-T that also expressed a significantly higher percentage of PD-1+ (figure 5G,H and online supplemental figure 5).

Toxicity associated with CD19/mbIL15q CAR-T cell therapy in mice

The appearance of splenomegaly in mice treated with CD19/mbIL15q CAR-T prompted us to study in more detail the potential toxicities of this treatment. Thus, mice preconditioned with high (4 Gy) or low (1–2 Gy) doses of TBI were treated with conventional CD19 CAR-T or CD19/mbIL15q CAR-T and followed carefully to evaluate the appearance of symptoms of toxicity. It was rapidly observed that CD19/mbIL15q CAR-T administration in mice preconditioned with a high TBI dose experienced a continuous decline in weight and survival. None of the animals survived beyond 40 days following CD19/mbIL15q CAR-T administration, in contrast to the conventional CD19 CAR-T which did not exhibit any adverse effects (figure 6A).

Figure 6

Toxicity caused by CD19/mbIL15q CAR-T cells in immunocompetent mice preconditioned with high or low TBI. (A, B) Schematic representation of the experimental setting and evolution of mice weight, survival and histological findings in mice treated with CD19 CAR-T cells or CD19/mbIL15q CAR-T cells after preconditioning with high TBI (A) or low TBI (B). (C) TCR Clonal diversity in mice treated with CD19/mbIL15q CAR-T cells. TCR Analysis was done in endogenous T cells (upper pie charts) or in the CD19/mbIL15q CAR-T cells (lower pie charts) isolated from the spleen of mice at days between 50 and 70 after CAR-T administration. One naïve mouse was also included as a representative control for TCR diversity in endogenous T cells. TBI, total body irradiation.

When CD19/mbIL15q CAR-T cells were injected in mice preconditioned with low TBI, no significant changes in total weight were observed during the first 35 days after cell transfer. However, mice survival was dramatically affected in the long term, with 50% death before day 60 (figure 6B). The histological analyses carried out at day 23 (in the case of high TBI) and at day 49 (in the case of low TBI) revealed an important T lymphocyte infiltration in the lung parenchyma in mice treated with CD19/mbIL15q CAR-T cells (figure 6A,B (anti-CD3 immunohistochemistry), and online supplemental figure 6A,B (H&E staining)). These histological changes were more pronounced in mice preconditioned with a high TBI regimen. Indeed, T cell infiltration in the lung was even more pronounced than in the low TBI regimen, and evident T cell infiltration in the liver was observed. Moreover, the normal spleen histology was also affected.

Cytokine levels in the sera of mice after 50 days of CAR-T injection were measured by luminex. We did not detect an extraordinary elevation in any of the cytokines tested. However, a significant higher level of IL6 and TNF-α was observed in those mice treated with CD19/mbIL15q CAR-T cells in comparison to the rest of groups (online supplemental figure 7A). We also measured the cytokine levels at day 17 after treatment with the CAR-T cells in mice challenged intravenous with A20 tumor cells. Except for MCP-1, which appeared to be higher in the CD19 group in comparison to the rest of the groups (online supplemental figure 7B) no significant changes were observed. There were also significant differences between untreated control mice and the rest of the groups regarding MCP-1. However, in general, these cytokine levels were considered low with respect to that observed in patients suffering the CRS after CAR-T cell therapy.28 Despite this, we studied the potential effect that the administration of anti-IL6 or anti-TNF-a antibodies could have in protecting mice from the toxic effect of CD19/mbIL15q CAR-T administration. Antibody administration did not alter the efficacy of CD19/mbIL15q CAR-T cells to deplete CD19 cells. However, neither the blocking of IL6 nor TNF-α could prevent the death of mice following the administration of 6×106 CD19/mbIL15q CAR-T cells in a mouse previously conditioned with high TBI (online supplemental figure 7C). We tried also to manage the toxicity using rapamycin to control CART cell activation. In this case, rapamycin administration began on day 25 post-tumor challenge to halt CAR-T cell expansion, with subsequent doses administered every 3 days. Unfortunately, although we observed a depletion effect on CD19 cells, we were unable to mitigate the toxic effect of CD19/mbIL15q CAR-T cells (online supplemental figure 7D).

Biochemical analyses carried out in serum samples at day 35 of treatment in the low TBI regimen revealed a significant increase in alanine aminotransferase and aspartate aminotransferase liver enzymes and a trend of lower albumin and alkaline phosphatase levels in mice treated with CD19/mbIL15q CAR-T cells (online supplemental figure 7E–L).

We then assessed varying doses of CD19/mbIL15q CAR-T cells (6, 2, and 0.2×106) to determine efficacy and toxicity in mice following high TBI (figure 7A). Mice treated with 6×106 CAR-T cells showed complete depletion of CD19+ B lymphocytes but, as expected, experienced severe toxicity, leading to all animals succumbing by day 35 of treatment. The 2×106 CAR-T cell dose effectively depleted CD19 B lymphocytes but also resulted in toxicity, with all mice dying by day 70 (figure 7B). In contrast, administering 0.2×106 CAR-T cells still achieved a long-lasting CD19+ B lymphocyte depletion with reduced toxicity, as all mice survived at 70 days post-treatment. However, monitoring CD4 and CD8 T cell numbers post-TBI revealed excessive proliferation above normal levels in all cases, being particularly earlier in mice treated with higher CAR-T cell doses (figure 7C). Histological analyses revealed significant lymphocytic infiltration in the lungs of mice at the onset of toxicity symptoms. Interestingly, a high T cell infiltration was also detected in one mouse treated with 0.2×106 CAR-T cells and sacrificed on day 70, before the appearance of clear toxicity symptoms (figure 7D). These results underscore the need for strategies to control CAR-T cell proliferation and enhance safety.

Figure 7

Dose-dependent CD19/mbIL15q CAR-T cell toxicity. Follow-up of BALB/c mice after preconditioning with high TBI and the injection of 6, 2 or 0.2×106 CAR-T cells. (A) Schematic representation of the experimental setting. (B) Evolution of mice weight and survival. (C) Follow-up of the percentage of CD19+ cells and total numbers of CD4 and CD8 T cells in the blood of mice after CD19/mbIL15q CAR-T cell treatment. (D) Histological analyses (anti-CD3 immunostaining) of tissues from mice treated with different doses of CD19/mbIL15q CAR-T cells at days 25 (in the TBI and the group treated 6×106 CAR-T cells), 52 (in the group receiving 2×106 CAR-T cells) and 70 (in the group treated with 0.2×106 CAR-T cells).

To evaluate the role of endogenous immune cells in this toxicity, we performed the same experiment in NSG mice. Mice were injected with 6×106 CD19 CAR-T or CD19/mbIL15q CAR-T. In some groups, we also injected B cells from naïve BALB/c mice to evaluate the role of CAR-T cell stimulation in toxicity (online supplemental figure 8A). Notably, when murine CD19/mbIL15q CAR-T cells were injected into NSG mice we observed a similar toxicity to what was found in immunocompetent mice. This toxicity was observed even in mice that did not receive naïve B cells (online supplemental figure 8B). In fact, all animals died before day 25 after T cell transfer.

TCR clonal diversity decreases in CD19/mbIL15q CAR-T cells

To better understand the potential risks of CD19/mbIL15q CAR-T cells in the long term, changes in the composition of TCR repertoire were studied by flow cytometry in the spleen of mice that survived more than 50 days after CAR-T cell infusion. The analysis in four CD19/mbIL15q CAR-T treated mice showed a clonal expansion of TCR subsets in the CAR-T cell compartment as compared with the classical TCR distribution observed in the endogenous T cells. Pie chart representation in figure 6C shows the results of TCR Vβ clonality, where red colored segments indicate the percentage of less represented TCR specific populations in the CAR-T and in the non-CAR-T (endogenous T cells) compartments for each animal. While the percentage of less common TCRs remained around 20%–30% in the endogenous T cell compartment of both the control and CD19/mbIL15q CAR-T -treated groups, this percentage was significantly higher in the CAR-T cell compartment (three out of four mice). This is particularly relevant in mouse 2, where the percentage of less frequent TCRs is above 80%. We have compared the phenotypes of CART cells with the most common TCRs to those within the “other TCR” category. Through this analysis, we did not find significant differences in phenotype, with over 90% of cells falling under the “effector cells” subpopulation in both cases (online supplemental figure 9). This skewing of the TCR repertoire might suggest an aberrant T cell proliferation of particular clones. To evaluate potential signs of oncogenic transformation, we isolated CD19/mbIL15q CAR-T cells from three mice on the onset of toxicity manifestations (30–40 days post-CAR-T cell infusion). DNA isolated from the CD19/mbIL15q CAR-T cell product before infusion (input, M1) and from the CAR-T cells isolated from these mice (M2, M3, and M4) underwent low-pass whole-genome sequencing (WGS) to assess tumor mutation burden (TMB) and somatic copy number alterations. Utilizing ichorCNA software for tumor fraction estimation in ultra-low pass WGS,29 only a minimal number of exome mutations (<3 nonsynonymous substitutions) were detected compared with input M1, with an estimated TMB of 0.012–0.013 mutations per megabase. The genome-wide copy number analysis and tumor fraction estimation using ichorCNA indicated the absence of tumor cells in the CAR-T cells isolated from the spleen of mice exhibiting prominent signs of toxicity (online supplemental figure 10). These results suggest that toxicity may be linked to the excessive proliferation of CD19/mbIL15q CAR-T cells rather than oncogenic transformation. However, additional experiments are needed to confirm this possibility.

Discussion

The potential of IL15 to promote an antitumoral immune response30 has positioned IL15 as a promising immunotherapeutic tool for treating cancer.31 However, systemic administration of high doses of IL15 can lead to T cell proliferation abnormalities and toxicities. In humans, dysregulated IL15 production or abnormal IL15 signaling has been associated with autoimmune diseases,32 neural toxicity,33 lymphocytic leukemia,34 and T cell lymphomas.35 Similarly, overexpression of IL15 or a membrane-bound IL15 version in mice ha