Bone Marrow Mesenchymal Stromal Cells Can Render Multiple Myeloma Cells Resistant to Cytotoxic Machinery of CAR T Cells through Inhibition of Apoptosis

Over the past decade, immunotherapy, with a variety of strategies and flavors, has emerged as a promising therapeutic approach in hematologic cancers (1). In multiple myeloma (MM), impressive response rates have been reported for the recently approved monoclonal antibodies, daratumumab, isatuximab (both anti-CD38), and elotuzumab (anti-SLAMF7; refs. 2, 3). Next to these FDA-approved antibodies, bispecific T cell–engaging antibodies, and T cells that are genetically engineered to express chimeric antigen receptors (CAR T cells) are rapidly entering the immunotherapy area as highly appealing strategies. Among the target antigens, BCMA is most extensively studied in the context of MM-targeting therapy, followed by SLAMF7, GPRC5D, CD38, and CD138 (4–8). BCMA-targeting CAR T cell therapy showed high overall response rates, generally above 80%, irrespective of antibody origin (human or murine) or costimulatory domain (CD28 or 4-1BB) implemented in the CAR design. Nonetheless, many patients experience short remissions and develop relapses (9–11). Relapses of targeted immunotherapy also occur after antibody therapies and involve several resistance mechanisms, not limited to target antigen reduction (12). These and many other lines of evidence indicate that MM is able to escape from potentially very powerful immunotherapies, similar to its ability to escape from conventional chemotherapy and proteasome inhibitors. Currently, immunosuppression, immune exhaustion, and target antigen downregulation are considered major mechanisms of cancer immune escape. In addition, over the past years, we and other investigators have provided evidence that MM cells can develop intrinsic resistance against cytotoxic killer mechanisms of immune cells through the intensive cross-talk with bone marrow mesenchymal stromal cells (BMMSC), similar to what has been observed for drug resistance (13–16). Specifically, we have shown that BMMSCs can protect MM cells from lysis by MM-reactive CD4⁺ and CD8⁺ cytotoxic T cells (CTL) as well as by daratumumab-redirected natural killer (NK) cells (17, 18). We demonstrated that this mode of immune escape is induced mainly by direct BM accessory cell–MM cell contact and involves the upregulation of antiapoptotic proteins Survivin and Mcl-1 in MM cells (17, 18).

Here, to gain a deeper insight into the overall impact of this bone marrow–microenvironment (BM-ME)-mediated immune resistance, we evaluated whether BMMSCs would also protect MM cells from CAR T cell therapy. To this end, we tested a panel of 10 different CAR T cells that were reactive to BCMA, CD38, or CD138 antigens with different avidities to the target antigen and that included different costimulatory domains: CD28, 4-1BB, or CD28 supplemented with a 4-1BB ligand. These CAR T cells were tested against a model MM cell line and against patient-derived primary MM cells in the absence or presence of pooled, MM patient–derived BMMSCs. In addition, we further investigated the involvement of antiapoptotic regulator molecules in BMMSC-mediated immune resistance using FL118, a small-molecule inhibitor of Survivin, Mcl-1, and XIAP (19, 20).

BMMNCs from MM patient BM aspirates were isolated by Ficoll-Hypaque density-gradient centrifugation and cryopreserved in liquid nitrogen until use. All patient material was collected after written informed consent according to the code of conduct for medical research developed by The Council of the Federation of Medical Scientific Societies (FEDERA, https://www.federa.org/codes-conduct). All procedures were in accordance with the Declaration of Helsinki and approved by the institutional medical ethical committee.

Diagnostic BM aspirations from MM patients were used to isolate BMMNCs and culture adherent BMMSCs as described elsewhere (21). To minimalize interindividual variation, all experiments were performed with cells from a pool of BMMSCs derived from 12 MM patients at the earliest possible passage (passage 3).

Luciferase (LUC)-transduced MM cell line UM9, previously generated from BMMNCs obtained from a MM patient at time of diagnosis (22), was cultured in RPMI-1640 (Life Technologies), supplemented with 10% HyClone Fetal Clone I serum (GE Healthcare Life Sciences) and antibiotics (penicillin, 10.000 U/mL; streptomycin, 10.000 μg/mL; Invitrogen). Authenticity of the cell line was verified by STR profiling (GenePrint 10 System Promega).

All CAR T cells with different specificities were generated with costimulatory domains 4-1BB, CD28, CD28 supplemented with a separate 4-1BB ligand (28zBBL), or no costimulatory domain, with the technology as described previously (23). CD138-CAR T cells were produced using single-chain variable fragment (scFv) sequences of the published nBT062 monoclonal antibody (6). BCMA-CAR T cells were produced using published scFv sequences derived from C11D5.3 monoclonal antibody (WO 2010/104949 A2; ref. 24) or from BCMA02 CAR (product name bb2121, WO 2016/094304 A2; ref. 25), both of which were used in earlier clinical trials. The generation and functional analysis of CD38-CAR T cells with low (B1) or high (028) CD38 affinities have been described previously (23, 26). More detailed information on CAR constructs and scFv sequences is listed in Table 1 and Supplementary Table S1. CAR T cells were tested 10–14 days after CAR transduction or frozen until testing. After thawing, CAR T cells were cultured for 16 hours in 60 U/mL recombinant human IL2 (R&D Systems)-containing culture medium [RPMI-1640 supplemented with heat-inactivated FBS (Sigma) and antibiotics], before using in experiments. Flow cytometry analysis, to determine the transduction efficiency and phenotypic profile of each CAR T cell, was performed as previously described and illustrated in Supplementary Fig. S1 (23, 26).

3AB11 is an HLA-DP4 restricted, minor histocompatibility antigen (mHag)-specific CD4⁺ cytotoxic T lymphocyte (CTL) that has been previously described in detail (22, 27) and was earlier used to demonstrate BM-ME-mediated immune resistance (18). 3AB11 was expanded using a feeder cell–cytokine mixture and cryopreserved until further use as described previously (27).

BMMSCs were plated in white opaque, 96-well flat-bottom plates in DMEM (Life Technologies) supplemented with heat-inactivated FBS and antibiotics, at 2 × 10⁴ cells/well. After 24 hours, 1 × 10⁴ LUC-transduced UM9 MM cells were added to BMMSC-seeded or medium-only wells. After 24 hours, serial dilutions of immune effector cells and/or FL118 were added. The survival of MM cells after 24 hours of treatment was determined by bioluminescence imaging (BLI), 30 minutes after the addition of the substrate beetle luciferin (125 μg/mL; Promega). Percent lysis of UM9 cells was determined using the following formula: 1 – BLI signal in treated wells/mean BLI signal in untreated wells × 100%.

BMMSCs were seeded in 24-well plates at a density of 1 × 10⁵ cells/well. After 24 hours, 1 × 10⁴ LUC-transduced UM9 cells were placed in 6.5 mm transwells having 0.4 μmol/L pore membrane inserts (Sigma-Aldrich). After 24 hours, CAR T cells were added in the transwell compartment, in which the UM9 cells were present. Percent lysis of UM9 cells by CAR T cells was determined after 24 hours by BLI, after transferring the cells from the transwell compartment to white opaque, 96-well flat-bottom plates.

Cell-free supernatants of MM cell killing assays were stored at −20°C. The IFNγ and granzyme B contents in thawed supernatants were determined using ELISA kits (eBioscience and Mabtech for IFNγ and granzyme B, respectively) according to the manufacturer's protocol.

CAR T cells were incubated alone or with UM9 cells, in the presence or absence of BMMSCs for 24 hours. Mosenin (BioLegend) was added 4 hours prior to termination of the assay. Nonadherent cells were harvested by gentle pipetting and stained for LNGFR (CD271, BioLegend), CD3 (BD), and live/dead marker (LIVE/DEAD Fixable Near-IR; Life Technologies). The cells were fixed in 4% formaldehyde for 5 minutes at room temperature (RT). After fixation, the cells were placed in FACS lysing solution (BD) for 10 minutes at RT. The permeabilized cells were stained for IFNγ and granzyme B (BioLegend) for 30 minutes at RT. Expression levels were analyzed in viable CD3⁺ LNGFR⁺ or CD3⁺ dsRED⁺ CAR T cells using flow cytometry.

LUC- and Green fluorescence protein (GFP)-transduced UM9 MM cells were incubated alone or on a monolayer of 1 × 10⁶ BMMSCs in a 100 mm dish for 24 or 48 hours. Nonadherent cells were harvested by gentle pipetting, first collecting culture medium and then washing with PBS. Adherent cells were harvested using Stem-Pro Accutase (Life Technologies). Both adherent and nonadherent cells were blocked with human immunoglobulin (100 μg/mL, Sanquin) and stained for CD105 (Synobiotechnology) and live/dead marker. The stained cells were fixed in 4% formaldehyde for 15 minutes at RT. After fixation, the cells were placed in ice-cold 90% methanol for 30 minutes on ice and subsequently stored at −20°C. The permeabilized cells were blocked with human immunoglobulin and stained for Survivin and Mcl-1 (Cell Signaling Technology). Expression levels were analyzed in viable GFP⁺ CD105⁻ UM9 cells using flow cytometry.

Cryopreserved BMMNCs derived from MM patients, containing 6%–60% MM cells, were incubated alone or on a monolayer of 2 × 10⁴ BMMSCs in a 96-well flat-bottom plate for 16–24 hours before the addition of CAR T cell, antibodies, and/or FL118, which were incubated for another 24 hours. After addition of flow-count fluorospheres (Beckman Coulter), cells were harvested using Stem-Pro Accutase, blocked with human immunoglobulin, and stained for CD38, CD138, CD45 (Beckman Coulter), and live/dead cell marker to determine absolute numbers of viable CD138⁺ CD38⁺ CD45^dim MM cells using flow cytometry. The percentage lysis was then calculated using the following formula: 1 – absolute number of viable MM cells in treated wells/mean absolute number of viable MM cells in untreated wells × 100%.

Violet Trace–labeled BMMSCs (CellTrace Violet Cell Proliferation Kit, Invitrogen) were plated in a 96-well flat-bottom plate at 2 × 10⁴ cells/well. UM9 MM cells were added at 2 × 10⁴ cells/well and CAR T cells at 2 × 10⁴ or 6 × 10⁴ cells/well. After 24 hours, all cells were harvested using Trypsin-EDTA (0.05%) phenol red (Gibco) and stained with live/dead cell marker to determine viable Violet Trace⁺ BMMSCs by flow cytometry. The percentage cell survival was then calculated using the following formula: viable BMMSCs in treated wells/viable BMMSCs in untreated wells × 100%. In a separate setting, Violet Trace–labeled UM9 cells were used and cultured without BMMSCs to measure the UM9-killing capacity of CAR T cells as control.

Cryopreserved BMMNCs derived from MM patients or plasma cell leukemia (PCL) patients were treated with various concentrations of FL118. After 24 hours, viable CD45⁺ CD3⁺ T cells and CD45⁺ CD56⁺ CD14⁻ CD3⁻ NK cells were enumerated by multiparameter flow cytometry. Additionally, healthy-donor peripheral blood mononuclear cells (PBMC) were seeded in 96-well flat-bottom plates at 1 × 10⁵ cells/well, stimulated with PMA (25 ng/mL) and ionomycin (500 ng/mL; both Santa Cruz Biotechnology) for 24 hours, and subsequently treated with FL118 for 48 hours before determining activated CD45⁺ CD3⁺ CD56⁻ CD25⁺ T cells by flow cytometry. The percentage cell survival was then calculated using the following formula: absolute number of viable immune cells in treated wells/mean absolute number of viable immune cells in untreated wells × 100%.

PBMCs from healthy donors were seeded in 96-well flat-bottom plates at 4 × 10⁴ cells/well and stimulated with anti-CD3/CD28-coated dynabeads (Life Technologies) in a bead-to-cell ratio of 1:3 and 50 U/mL recombinant human IL2. After 24 hours, the PBMCs were treated with serial dilutions of FL118 or a DMSO control, representing the highest DMSO content, for 24 hours. BrdUrd was added 16 hours prior to the termination of proliferation. BrdUrd incorporation was measured by Cell Proliferation ELISA (Sigma-Aldrich) according to the manufacturer's protocol.

Comparisons between variables were performed using two-tailed paired or unpaired Student t test using Prism software (GraphPad Software Inc., v.7). Comparisons between multiple groups were performed using a nonparametric Kruskal–Wallis test. Where indicated, the effector-to-target (E:T) ratio of CAR T cells required to reach half maximal lysis of MM cells was determined by nonlinear regression of lysis values obtained with increasing E:T ratios. Correlation was computed using two-tailed Pearson correlation coefficients after checking for normal distribution. Where indicated, the Chou–Talalay method was used to quantify immunotherapy-FL118 combinatorial effects with combination index (CI) values of < 1 indicating synergy, of 1 indicating additive effects, and of > 1 indicating antagonism (28). When primary MM cells were used as target cells, interaction between CAR T cells and FL118 was estimated with a BLISS model, in which expected lysis values from combinatorial treatments were calculated using the following formula: [(% lysis with immunotherapy + % lysis with FL118) − % lysis with immunotherapy × % lysis with FL118; refs. 29–31]. The null hypothesis of “additive effects” was rejected if the observed values were significantly different from the expected values. P values below 0.05 were considered statistically significant.

We have previously shown that BMMSCs facilitate MM cell resistance against lysis by conventional CD4⁺ and CD8⁺ CTLs (18). To investigate the impact of BMMSC-induced immune resistance against CAR T cells, we used a panel of BCMA-, CD38-, and CD138-directed CAR T cells (Table 1) that showed no significant difference in transduction efficiency, CD4/CD8 ratio, or phenotypic profile (Supplementary Fig. S1). To target BCMA, we included two different CAR T cells with altered scFv sequences. For CD38, we used CAR T cells with different target affinities and various costimulatory domains containing either no costimulatory domain, 4-1BB, CD28, or CD28 plus a separately expressed full-length 4-1BB ligand (28zBBL; Table 1).

In the absence of BMMSCs, CD138-CAR T cells showed a moderate efficacy to induce UM9 cell lysis (Fig. 1A). The efficacy of BCMA-CAR T cells to induce UM9 cell lysis was moderate (BCMA^C11D5.3-CAR T cells) or high (BCMA^bb2121-CAR T cells), depending on the used scFv (Fig. 1B). Finally, the lytic activity of CD38-CAR T cells was, as previously demonstrated, dependent on both their affinity for its target and their costimulatory domains (Fig. 1C; refs. 23, 26). High-affinity CD38⁰²⁸-CAR T cells were highly lytic, independent of their costimulatory domains, while lower affinity CD38^B1-CAR T cells required CD28 costimulation to reach high lytic capacities (28z-CD38^B1- and 28zBBL-CD38^B1-CAR T cells) in comparison with 4-1BB costimulation alone (BBz-CD38^B1-CAR T cell). The cytotoxic activity of a first-generation CD38^B1-CAR T was similar to that of BBz-CD38^B1 CAR T cells, substantiating the beneficial role for the costimulatory domains, especially that of CD28, to increase cytotoxic activity (see also Supplementary Fig. S2). Furthermore, CAR T cells mediated no bystander lytic activity against BMMSCs (Supplementary Fig. S3).

In the presence of BMMSCs, there was little or no BMMSC-mediated inhibition of MM cell lysis when strongly lytic BCMA^bb2121-CAR T cells were used. In clear contrast, BMMSCs inhibited the MM cell lysis up to 60% when moderately lytic CD138- and BCMA^C11D5.3-CAR T cells were used. In the case of CD38-CAR T cells, the BMMSC-mediated protection was more pronounced for the lower affinity CD38^B1-CAR T cells that contained a 4-1BB costimulatory domain or no costimulatory domain (20%–50% inhibition of lysis). Remarkably, the inhibitory effect of BMMSCs was reduced or completely abrogated when lower affinity CD38^B1-CAR T cells contained a CD28 costimulatory domain, either alone (28z-CD38^B1 CAR T cells) or together with the separate 4-1BB ligand (28zBBL-CD38^B1 CAR T cells).

Taken together, these results suggested that the overall killing capacity of CAR T cells in the absence of BMMSCs, which is a functional reflection of the combination of target antigen expression, scFv affinity, and costimulatory signaling, could be an important predictor for the induction of cytotoxic resistance by BMMSCs. Indeed, we observed a strong inverse correlation between the E:T ratios leading to 50% MM cell lysis, depicted as EC₅₀ of the CAR T cells, and the extent of BMMSC-mediated inhibition of lysis (Fig. 1D).

To gain more insight into the possible clinical relevancy of the results generated with the model MM cell line UM9, we subsequently tested a selected panel of CAR T cells on MM cells that are present in BMMNC samples obtained from MM patients (n = 6) in the absence or presence of BMMSCs. We found a comparable association between the level of the CAR T cell lytic capacity in the absence of BMMSCs and the inhibitory effect of BMMSCs in this ex vivo setting (Fig. 2A). Analysis of the expression levels of target antigens on patient MM cells revealed that the BMMSCs did not influence the expression levels of CD138 or CD38, while a notable but insignificant decrease in BCMA expression was observed (Fig. 2B). Nonetheless, for all the CAR T cells tested, the level of target antigen downregulation in the presence of BMMSCs showed no correlation with BMMSC-mediated inhibition of lysis (Supplementary Fig. S4), thus excluding that alteration of target antigen expression was a dominant mechanism of BMMSC-induced resistance against CAR T cells.

We have previously shown that BMMSCs inhibit T cell– and antibody-dependent NK cell–mediated lysis of MM cells, although a certain degree of immune suppression was also observed for CD8⁺ CTLs (17, 18). To better understand how BMMSCs inhibit CAR T cell–mediated lysis, we first measured the IFNγ and granzyme B production by CAR T cells in response to UM9 cell line or to primary MM cells (n = 3) in the absence or presence of BMMSCs, as a reflection of CAR T cell activation (Fig. 3A–C; Supplementary Fig. S5). Using UM9 as target cells, we also determined the proportion of CAR T cells that produce these cytokines by intracellular cytokine staining (Supplementary Fig. S6). The results revealed that CAR T cell–mediated lysis of UM9 cells or patient MM cells by BCMA^C11D5.3-CAR T cells and BBz-CD38^B1-CAR T cells was inhibited by BMMSCs without a reduction in IFNγ or granzyme B secretion. Instead, secretion of IFNγ and granzyme B even increased for patient MM cells, indicating that the BMMSCs did not dampen the activation of these CAR T cells. For CD138-CAR T cells, we observed a reduction in IFNγ secretion in the presence of BMMSCs against two primary MM samples and with increasing E:T ratios against UM9 (Fig. 3B; Supplementary Fig. S5A). BMMSCs did not reduce granzyme B secretion, suggesting a partial or split CAR T cell suppression (Fig. 3C; Supplementary Fig. S5A). Also, the percentage of IFNγ or granzyme B–producing CAR T cells showed no significant changes in the presence of BMMSCs (Supplementary Fig. S6). Yet, to assess a possible involvement of suppressive soluble factors in the inhibition of lysis, we studied the efficacy of CD138-CAR T cells to kill UM9 cells in a transwell system, in which BMMSCs and MM cells were cocultured either in direct contact or separated by micropore membrane inserts (18). CD138-CAR T cell–mediated lysis of UM9 cells was inhibited only when BMMSCs and UM9 were in direct contact (Fig. 3D), which also upregulated Survivin and Mcl-1 expression in UM9 cells (Supplementary Fig. S7). Together, these results suggested that even in case of any immunosuppression, immune resistance could be the major contributor for BMMSC-mediated protection against CAR T cells, possibly through the upregulation of the tumor intrinsic antiapoptotic machinery as we have previously shown for conventional T cells and NK cells (17, 18). Hence, we next investigated whether we could also overcome the BMMSC-mediated protection against CAR T cell therapy by inhibition of antiapoptotic molecules in MM cells.

In our previous studies, the BMMSC-mediated resistance of MM cells against CTLs and ADCC was effectively modulated with the use of the small-molecule YM155, which inhibits antiapoptotic molecules Survivin and Mcl-1 (17, 18). To further investigate the involvement of antiapoptotic molecules in this novel mode of immune escape, we now used FL118, which is structurally different from YM155 and inhibits multiple antiapoptotic proteins, including Survivin, Mcl-1, XIAP, and cIAP2, while inducing proapoptotic proteins Bax and Bim, in a p53-independent manner (19). Because such molecules can also have effects on nonmalignant cells, we explored whether FL118 could modulate BMMSC-induced inhibition of CAR T cell–mediated lysis at concentrations that are nontoxic for immune cells and BMMSCs. Earlier, we had determined that FL118 had no toxic effects on BMMSCs up to a dose of 100 nmol/L (20). Up to this dose, FL118 also showed no toxic effects on resting T cells or NK cells in BMMNC samples obtained from MM or PCL patients, or on PMA-activated CD25⁺ T cells of healthy individuals (Supplementary Fig. S8). Thus, we tested FL118 at doses equal to or lower than 100 nmol/L for its capacity to modulate BMMSC-mediated protection of the UM9 cell line against CAR T cells. FL118 treatment showed a dose-dependent lysis of MM cells, regardless of the presence of BMMSCs (Fig. 4A–C, right). When combined with CAR T cells, FL118 effectively modulated the BMMSC-mediated protection, even at doses as low as 3 nmol/L (Fig. 4A–C, center). The combinatorial activity of FL118 and CAR T cells was synergistic, especially in the presence of BMMSCs, as illustrated by CI values below 1 (Fig. 4A–C, center). To extend the analyses in an ex vivo setting, we tested FL118 at a low dose of 10 nmol/L for its capacity to modulate BMMSC-mediated protection of patient-derived MM cells against the same panel of CAR T cells (Fig. 4D). As expected, BMMSCs readily inhibited the lysis of primary MM cells by the tested CAR T cells at an E:T ratio of 1:1, except in a single new sample where patient MM cell lysis by CD38-CAR T cells was not inhibited. Importantly, although FL118 showed no/minimal anti-MM activity at the tested dose, it significantly enhanced the patient MM cell lysis by CAR T cells in the presence of BMMSCs (Fig. 4D). Comparing the observed lysis values with the calculated expected lysis values, which presume additive combinatorial effects, showed that the combination of FL118 with CAR T cells enhanced MM cell lysis in a synergistic fashion (Fig. 4D).

The ability of FL118 to modulate BMMSC-mediated immune resistance to CAR T cells prompted us to test whether this compound could also modulate the immune resistance toward conventional CTLs and daratumumab-dependent ADCC. Confirming previous results, BMMSCs protected the UM9 cell line against the MM-reactive CD4⁺ CTL 3AB11 (Fig. 5A, left), and FL118 completely abrogated this protective effect in a synergistic fashion (Fig. 5A, center). Furthermore, the profound BMMSC-mediated protection against daratumumab-induced MM cell lysis was also abrogated by FL118 (Fig. 5B, middle, observed values), similar to the results we have previously obtained with YM155 (17). Also, in these ADCC assays, the interaction between FL118 and daratumumab in the presence of BMMSCs was synergistic, as indicated by the significantly higher observed lysis values than the calculated expected lysis values.

CAR T cell–based therapy is a highly promising approach for the treatment of hematologic cancers, including MM. Recent clinical trials in MM patients showed that CAR T cells can achieve unprecedented immediate results but are not able to eradicate tumor cells, indicating immune escape. Currently, various strategies are considered to improve CAR T cell therapy. These include the search for new target molecules, optimization of costimulatory domains, application of dual CAR constructs (32), engineering CAR T cells to secrete immune-stimulatory cytokines (33, 34), or to secrete PD-1 blocking single-chain variable domains to tackle immune suppression (35, 36). In the current study, we approached the quest from a different angle and studied the potential negative impact of BMMSCs on the efficacy of CAR T cells.

To gain a broad insight into the potential impact of BMMSCs on CAR T cells, we used a large panel of CAR T cells with various antigen specificities (BCMA, CD38, and CD138) and target cell affinities. The CD38-CAR T cells also contained different costimulatory domains to cover various flavors of CAR T cell designs as much as possible. To obtain results that represent the clinical setting as closely as possible, we used BMMSCs derived from various MM patients (n = 12) and tested our hypotheses not only with the model MM cell line UM9 but also with >15 BM samples containing primary MM cells obtained from untreated or previously treated MM patients. Overall, our results reveal a remarkable inverse correlation between the lytic capacity of CAR T cells and the capacity of BMMSCs to protect MM cells from CAR T cells, regardless of their antigen specificity. Although the most potent BCMA^bb212- and CD38⁰²⁸-CAR T cells were largely unaffected by BMMSCs, the cytotoxic efficacy of moderately lytic CD138-, BCMA^C11D5.3- and BBz-CD38^B1-CAR T cells was diminished by almost 100% in some MM patient samples in the presence of BMMSCs. Based on our results obtained with BCMA- and CD38-specific CAR T cells with different avidities or affinities, it appears that CAR avidity and affinity are important factors to increase cytotoxic capacities of CAR T cells and to circumvent BMMSC-mediated immune resistance. Furthermore, in agreement with several other studies (23, 37, 38), we found that the presence of CD28 costimulatory domain in the CAR construct provides more efficient cytotoxic capabilities to CAR T cells, hereby competing with the protective capacity of BMMSCs that shields MM cells from cytotoxic attack. Although these latter results were obtained only in the setting of CD38-targeting CAR T cells, they strongly suggest that CAR designs aimed to improve the lytic activity of CAR T cells, through either increasing the CAR affinity or optimizing the costimulatory domains could enable CAR T cells to avoid BMMSC-mediated immune resistance in the tumor microenvironment.

Increasing the lytic capacity of CAR T cells as a strategy to avoid BM-ME-induced resistance is, however, not always desirable because hereby one can also trigger cytokine release syndrome or increase the on-target off-tumor toxicity, especially if the target antigen is not entirely tumor specific, such as in the case of CD38- and CD138-specific CAR T cells. In such cases, and in all other cases where the CAR design cannot be improved, our results provide the possibility to overcome the BMMSC-mediated immune resistance by combining CAR T cells with inhibitors of antiapoptotic molecules such as the small-molecule FL118 (19). Additional research on possible application of FL118 showed that this small molecule also effectively abrogates BMMSC-mediated resistance against MM-reactive, HLA-restricted conventional CTLs and against daratumumab-dependent ADCC. In all these cases, we found that FL118 interacted with the immune effector cells in a synergistic fashion, especially in the presence of BMMSCs. It is important to emphasize that the optimal dose of FL118 to synergize with T and NK cells is, similar to YM155, much lower than the dose that is required to achieve optimal anti-MM effects with single-agent FL118. Although FL118 concentrated up to 100 nmol/L did not affect T and NK cell viability, we have observed that FL118 can mediate a dose-dependent antiproliferative effect on T cells (Supplementary Fig. S9) and hereby may alter the phenotype of CAR T cells or other immune cells, because recent studies indicate the existence of a division-linked differentiation of CD8⁺ T cells into effector versus central memory subsets (39). Thus, although our in vitro and ex vivo assays suggest a safe and effective window of application, and in our earlier studies the deduction of in vivo doses of YM155 from ex vivo assays revealed strong in vivo synergistic effects with CTLs and NK cells (17, 18), any (future) attempts of clinical translation of YM155, FL118, or similar compounds would require careful dose-escalation studies in appropriate immune-competent in vivo models, which also properly recapitulate BMMSC-MM cell interactions in the microenvironment.

Although clinical translation requires more studies, we demonstrate in the current study as well as in prior investigations, using the modulatory activities of FL118 or YM155, that the well-known antiapoptotic effects of BMMSC-MM interactions are the main drivers of BMMSC-mediated immune resistance. Taken all our results together, it is now evident that BMMSC-mediated immune resistance is initiated by cell adhesion, followed by modulation of apoptotic pathways that subsequently induces an intrinsic resistance against the cytotoxic machinery of CAR T cells. This type of immune resistance is not related to several other well-known T cell–inhibitory mechanisms such as T cell suppression, antigen loss, steric hindrance, or disturbance of immune synapses, based on the fact that we found no overt inhibition of T cell activation by BMMSCs, except for some partial T cell suppression in CD138-CAR T cells (see Fig. 3; Supplementary Fig. S3). Because YM155 and FL118 can modulate multiple antiapoptotic molecules such as Survivin, Mcl-1, and XIAP, further studies are still required to elucidate the individual contributions of this mode of immune escape.

A main distinction in the current study from our earlier studies is the exclusive use of patient-derived BMMSCs rather than fibroblast stromal cell lines HS-5 or HS-27a. We refrained from using these stromal cell lines as they deviate in their cytokine profiles and contain multiple mutations that are not present in patient-derived BMMSCs. On the other hand, we did face some technical challenges with the choice of patient-derived BMMSCs. Most importantly, we had to create a BMMSC pool of 12 newly diagnosed MM patients in order to obtain a single batch of BMMSCs. Although this solved several reproducibility issues in our assays, we thereby lost the possibility to evaluate the interindividual variety between patients' BMMSCs. This issue needs to be addressed in future studies.

Another relevant challenge for future studies is the development of advanced in vivo models to provide in vivo evidence for BMMSC-mediated immune resistance and safe modulation thereof. Currently, we test the efficacy of anti-MM therapies in a unique in vivo model, in which MM tumors are grown in human BMMSC-coated subcutaneously implanted scaffolds (40). Our previous studies using this model strongly suggest the negative impact of BMMSCs on the efficacy of immune therapies because, similar to the human setting, and different from other in vivo models which do not contain human BMMSCs (7, 8, 41–43), we never achieve cure by conventional T cells, daratumumab, or CAR T cells (5, 17, 18, 23, 26, 29, 44), even with the strongest BCMA^bb2121-CAR T cells (Supplementary Fig. S10). Nonetheless, this model currently lacks a proper non-BMMSC control and therefore requires improvements to demonstrate the impact of BMMSCs on immunotherapy.

Regardless of these aspects, we think that our results can be considered as an important extension of a large series of studies showing the importance of stromal cells in the tumor microenvironment for the development of therapy resistance. Although our study is, to our best knowledge, the first study addressing the importance of BMMSCs on the efficacy of MM-reactive CAR T cells, a previous study in a solid tumor model already pointed out the importance of stromal cells on the efficacy of CAR T cells by showing that the destruction of tumor stroma with IFNγ can contribute to the eradication of large tumors by HER2-specific CAR T cells (45). The important difference of our study from this earlier study, however, is that CAR T cell efficacy could also be upregulated without the destruction of stromal cells.

In conclusion, we have shown that the BM-ME protects MM cells against low- and intermediate-cytotoxic CAR T cells by mechanisms that we had earlier demonstrated for conventional T cells and daratumumab-mediated ADCC. This protective shield seems to be abrogated by increasing the lytic capacities of CAR T cells or by inhibition of antiapoptotic proteins in MM cells, which provides perspectives for future clinical translation.

L.C. Holthof reports a patent for novel anticancer drug FL118 formulation in combination with immunotherapy for treatment of human cancer pending to Canget BioTekpharma. F. Li reports a patent for PCT/US2020/043153 pending; FL118 and FL118 core structure-based analogues will be further developed in Canget BioTekpharma LLC (www.canget-biotek.com), a Roswell Park Cancer Institute-spinoff company. F. Li is one of the initial investors in Canget for development of FL118 and FL118 core structure-relevant anticancer agents and owns Canget equity. S. Zweegman reports other from Oncopeptides, Sanofi, and BMS, as well as grants and other from Janssen and Takeda outside the submitted work. N.W.C.J. van de Donk reports grants and other from Janssen Pharmaceuticals, Amgen, Celgene, and BMS; grants from Cellectis; and other from Adaptive, Novartis, Roche, Takeda, Bayer, and Servier outside the submitted work. M. Themeli reports grants from Multiple Myeloma Research Foundation and Foundation Cancer Center Amsterdam during the conduct of the study; in addition, M. Themeli has a patent for effective generation of tumor-targeted T cells derived from pluripotent stem cells licensed and with royalties paid from Fate Therapeutics. T. Mutis reports grants from Dutch Cancer Society during the conduct of the study, as well as grants from Takeda, Genmab, Onk Therapeutics, and Janssen Therapeutics outside the submitted work; in addition, T. Mutis has a patent for novel anticancer drug FL118 formulation in combination with immunotherapy for treatment of human cancer pending to Canget BioTekpharma + VUmc. No disclosures were reported by the other authors.

L.C. Holthof: Conceptualization, resources, data curation, formal analysis, validation, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing. J.J. van der Schans: Resources. A. Katsarou: Resources. R. Poels: Resources, data curation. A.T. Gelderloos: Data curation. E. Drent: Resources. S.E. van Hal-van Veen: Data curation. F. Li: Resources, writing–review and editing. S. Zweegman: Conceptualization, supervision, writing–review and editing. N.W.C.J. van de Donk: Conceptualization, supervision, funding acquisition, writing–review and editing. M. Themeli: Conceptualization, writing–review and editing. R.W.J. Groen: Conceptualization, supervision, funding acquisition. T. Mutis: Conceptualization, supervision, funding acquisition, writing–original draft.

This study was supported by research funding from the Dutch Cancer Society (VU2014-6567) and Worldwide Cancer Research (14-1223). The authors would like to thank the Company of Canget BioTekpharma LLC (www.canget-biotek.com), Buffalo, NY, for providing the FL118 drug.

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.