γ-secretase inhibitors augment efficacy of BCMA-targeting bispecific antibodies against multiple myeloma cells without impairing T-cell activation and differentiation

We here defined the impacts of γ-secretase inhibitors (GSIs) on T-cell-dependent BCMA-specific multiple myeloma (MM) cell lysis and immunomodulatory effects induced by bispecific antibodies (BisAbs). GSIs-induced membrane BCMA (mBCMA) accumulation reached near maximum within 4 h and sustained over 42h-study period on MM cell lines and patient MM cells. GSIs, i.e., 2 nM LY-411575 or 1 μM DAPT, robustly increased mBCMA densities on CD138+ but not CD3+ patient cells, concomitantly with minimum soluble/shed BCMA (sBCMA) in 1 day-culture supernatants. In ex vivo MM-T-cell co-cultures, GSIs overcame sBCMA-inhibited MM cell lysis and further enhanced autologous patient MM cell lysis induced by BCMAxCD3 BisAbs, accompanied by significantly enhanced cytolytic markers (CD107a, IFNγ, IL2, and TNFα) in patient T cells. In longer 7 day-co-cultures, LY-411575 minimally affected BCMAxCD3 BisAb (PL33)-induced transient expression of checkpoint (PD1, TIGIT, TIM3, LAG3) and co-stimulatory (41BB, CD28) proteins, as well as time-dependent increases in % effector memory/central memory subsets and CD8/CD4 ratios in patient T cells. Importantly, LY41157 rapidly cleared sBCMA from circulation of MM-bearing NSG mice reconstituted with human T cells and significantly enhanced anti-MM efficacy of PL33 with prolonged host survival. Taken together, these results further support ongoing combination BCMA-targeting immunotherapies with GSI clinical studies to improve patient outcome.

However, membrane BCMA (mBCMA) receptor molecule is constantly cleaved by γ-secretase (GS), an intramembrane multisubunit protease complex, and its extracellular portion with part of the transmembrane domain is shed to cell culture media or the circulation to form soluble BCMA (sBCMA) [17,18]. The release of sBCMA may contribute to MM immunodeficiency by sequestering B-cell activating factor and a proliferation-inducing ligand [19]. Elevated sBCMA levels are found in serum samples of MM patients than healthy individuals [20] and further associated with myeloma burden and poorer survival [20][21][22][23]. These clinical results also indicate that sBCMA could serve as a trap to compete binding of BCMA-targeting drugs, thereby reducing their therapeutic promises. Moreover, decreased expression or loss of the antigen BCMA have been reported in patients relapsed from BCMA CAR T treatments [22,24,25].
We thus here defined the therapeutic potencies of small molecule GS inhibitors (GSIs) to increase mBCMA and counteract sBCMA inhibition in T-cell dependent cytotoxicity (TDCC) of BCMAxCD3 BisAbs against MM cell lines and patient MM cells in multiple preclinical human MM models in vitro and in vivo. A single GSI treatment rapidly enhanced mBCMA and eliminated sBCMA, leading to an improved BCMAxCD3 BisAb-induced MM cell lysis while sparing patient T-cell activation and differentiation in ex vivo co-cultures. GSI significantly augmented the in vivo anti-MM activity of a single sub-curative concentration BCMAxCD3 BisAb and prolonged host survival, further supporting ongoing combination clinical trials to improve patient outcome.

MATERIALS AND METHODS Cell lines and primary cells
All MM cell lines (American Type Culture Collection and DSMZ German Collection) express various levels of BCMA [14,15]. CRISPR/CAS9 technique was used to knock down BCMA expression to generate BCMA knock-down (KD) H929-KD [14] and U266-KD cells.
Samples from normal donors and MM patients were obtained after informed consent was provided, in accordance with the Declaration of Helsinki and under the auspices of a Dana-Farber Cancer Institute (DFCI) Institutional Review Board approved protocol.
Culture supernatants of MM cell lines and BMMCs of MM patients in 96-well culture plates (10 5 cells per well) were collected for sBCMA measurement. Various concentrations of recombinant BCMA (sBCMA) (R&D Systems) were used to test its impact on MM cell killing by BCMAxCD3 vs. control BisAbs.

Flow cytometric (FC) data acquisition and analysis
Protein expression levels of mBCMA and mCD138, live vs. dead cell confirmation, redirected T-cell-dependent cytotoxicity (RTCC or TDCC) assays under mono-or combination treatment settings, autologous patient MM cell lysis, degranulation (CD107a mobilization) and intracellular Th1cytokine expression of T effector cells, T-cell phenotype and memory cell differentiation, as well as CD8/CD4 ratios were evaluated by quantitative FC analysis.
The density of mBCMA molecule on MM cell surface was also quantitated by PE Phycoerythrin Fluorescence Quantitation Kit (BD Biosciences) and shown as Abs bound per cell (Ab binding capacity, ABC).
All data of FC-based investigations were collected by BD FACSCanto™ II and BD LSRFortessa™ flow cytometers, and then analyzed by FlowJo V8.6.6 (BD Life Sciences) and FACS DIVA software (BD Biosciences).

In vivo adoptive T-cell transfer (AdT)-NSG mouse model of MM
Animal study procedures and protocols were performed under approval by the Novartis Institutes for BioMedical Research Institutional Animal Care and Use Committee and in compliance with the Guide for the Care and Use of Laboratory Animals. In brief, NSG mice (Jackson Labs Bar Harbor, ME) received adoptive transfer of human PBMCs (15 × 10 6 /mouse) (AdT) 5 days before KMS11-luc MM cells were implanted subcutaneously into the right flank (5 × 10 6 cells/mouse) (d0). When the tumor volume reached 198 ± 47 mm 3 , mice were randomized into indicated groups (10 mice/ group): controls (tumor only and Tumor + PBMC) or treatments receiving either LY-411575 (3 mg/kg), PL33 (1 or 3 mg/kg), or a combination of both. LY-411575 was administered for 2 consecutive days by oral gavage: 24 h before (d8) and the day when PL33 was administered (d9) as a single IV injection into the lateral tail vein 2 h post the second dose of LY-411575.

Statistics
Unless otherwise specified, all data were analyzed and graphed using GraphPad Prism 9.0.1. P-value < 0.05 was considered statistically significant.
Multiple groups (≥3) were analyzed by one-way analysis of variance (ANOVA), and paired groups were analyzed by two-way ANOVA or Student t-test.
Anti-tumor activity in vivo was evaluated using one-way analysis of variance with the Tukey post test, or an unpaired t-test with assumed similar variance. Kaplan-Meier Survival statistics were analyzed using Logrank test with all pairwise multiple comparison ad hoc (Holm-Sidak method).
The kinetics of GSI activity was next studied over a 42h-period in KMS11 cells cultured with 2 nM LY-411575 (Fig. 1D, left panels) or 200 nM PF03084014 (Nirogacestat) (Supplementary Fig. S1E). Only in the presence of GSI, mBCMA expression was enhanced over time (upper panel) whereas culture supernatant sBCMA levels stayed minimal (lower panel). Without GSIs, sBCMA concentrations steadily increased. GSI upregulated mBCMA expression as early as 1 h and near maximal level was reached at 4-6 h. GSIs continued to increase mBCMA with decreasing sBCMA levels for over 30 h. To determine how long the effect of GSI treatment persists, LY-411575 was washed out after 1 day treatment, and drug-free growth media was replenished (Fig. 1D, right panels). Membrane BCMA density on LY-411575-pretreated KMS11 cells persisted at 12-fold higher levels than untreated cells till 30 h before decreasing. Meanwhile, mBCMA baseline levels in untreated cells remained constant over time. Concentrations of sBCMA remained low in LY-411575-pretreated MM cell culture supernatants while significantly increasing in untreated cell media.
We next showed that % CD138 + cell lysis induced by PL33 vs. ER79 was augmented as early as 4 h with co-treatments with 2 nM LY-411575 or 1 μ DAPT in co-cultures of 2 target MM cell lines with PBMCs (E:T = 6:1) or T cells (E:T = 3:1) from MM patients (n = 3) ( Supplementary Fig. S2A). GSI alone induced neither MM cell lysis without BisAbs, nor MM cell apoptosis (n = 3) in the absence of T cells following 3 day-incubation (Supplementary Fig. S2B, C).
Under similar suboptimal test conditions, PL33 induced earlier and higher % lysis of H929 vs. U266 target cells, associated with increased mBCMA levels (Fig. 2C). When H929-KD or U266-KD target cells were co-cultured with patient T cells (n = 5) (E:T = 1:1), co-treatments with LY-411575 still significantly enhanced PL33induced MM cell killing as early as 4 h (P < 0.03) and continued to increase at 1 day (P < 0.05), reaching maximal lysis associated with increased % CD107apatient T cells ( Supplementary Fig. S2D). GSIpretreated MM target cells again significantly augmented BCMAspecific MM cell lysis induced by patient T cells (P < 0.05), regardless which BCMAxCD3 BisAb or target MM cells were used ( Fig. 2D and Supplementary Fig. S2E).
GSI treatment for 1 day was next done in additional MM BMMCs (n = 7) with baseline BCMA ABC ranging from 714 to 5311 on CD138 + cell membrane (Fig. 3C,    . Furthermore, concentrations of sBCMA negatively associated with mBCMA levels, being undetectable in~75% culture supernatants treated with GSI (Fig. 3D).
Following 1 day-incubation, PL33-induced patient autologous lysis were next shown using BMMCs from 6 RR MM patients with BCMA ABC ranging from 608 to 5048 on CD138 + PCs (Fig. 3F). As expected, higher PL33 concentrations triggered higher autologous patient cell lysis.
GSI upregulated autologous patient MM cell lysis induced by BCMAxCD3 BisAb MM BMMCs (n = 6) were next pretreated with LY-411575 or DAPT for 4 h and 1 day prior to the addition of three indicated BCMAxCD3 BisAbs vs. ER79 with matched patient PBMCs. These three BCMAxCD3 BisAbs induced autologous CD138 + patient cell lysis, with higher potency of PL33 and BU76 compared with BQ76 (Fig. 4A). GSI significantly enhanced autologous patient MM cell lysis induced by BCMAxCD3 BisAbs as early as 4 h and continued at 1 day. In contrast, pretreatments with GSI did not further change % background lysis. In five more MM patient BMMCs, pretreatment with LY-411575 for 1 day also enhanced % autologous patient cell lysis by PL33 vs. ER79 (Fig. 4B). We next co-treated BMMCs from RRMM patients (n = 8) with LY-411575 or DAPT during PL33 (1 nM)-induced RTCC assays for 1 day and 4 day (Fig. 4C, left). Co-treatment with GSI still significantly augmented PL33-induced autologous patient MM cell lysis at d1 (P < 0. 004) and d4 (P < 0.003). This was associated with increased % CD107 + patient T cells at d1 (P < 0.015) and d4 (P < 0.004) (Fig. 4C, right). MM1S target cells were further co-cultured with T cells from five additional individuals of normal donors or MM patients, for 1 and 3 days at lower E:T ratio of 1:1 and PL33 concentration, with or without LY-411575. PL33-induced MM1S cell lysis was associated with enhanced CD107a degranulation and Th1-type cytokine (IFNγ, IL-2, TNFα) production in T cells from both sources (Fig. 5A,  B). Importantly, PL33-induced MM1S cell killing by patient T cells was further enhanced with LY-411575 co-treatment at d1 and continued at d3, associated with further increased cytolytic markers. Expression of PD1, LAG3, and TIM3, continued to increase in PL33-activated T cells from d1 to d3 (Fig. 5C, P < 0.01). LY-411575 did not reduce % of EM or TEMRA subsets induced by PL33 (Fig. 5D). Moreover, PL33 induced regulatory T (Treg) cell subsets from d1 and continued at d3, while LY-411575 co-treatment did not further increase % Treg, IL-10 + , or TGFβ + T-cell subsets (Fig. 5E).
Further 7 day-co-cultures were done to study the serial changes in expression of key immune checkpoints (PD1, LAG3, TIM3, TIGIT) and co-stimulatory proteins (41BB, CD28) using patient T cells (n = 8) and MM1S cells (E:T = 1:1) treated with PL33, in the presence or absence of LY-411575. PL33 induced transient expression of these immune activation and regulatory proteins (Fig. 6A). Higher induction of these proteins was seen in PL33-activated CD8 vs. paired CD4 patient T cells. LY-411575 did not alter transient expression patterns of these markers induced by PL33 in patient T cells.
xenograft adoptive transfer (AdT) model. Mice reconstituted with human PBMCs were implanted with KMS11-luc MM cells and randomized into seven indicated groups (n = 10 mice per group) when tumors reached an average of~200 mm 3 . LY-411575 was administered 1 day before and on the same day as a single PL33 treatment. All groups received human PBMCs (AdT) except tumor only (ctrl w/o AdT) group. Both untreated control cohorts (ctrl w/o AdT and ctrl groups) showed similar growth kinetics, indicating no allogeneic response (Fig. 7B). There was no anti-MM activity with single agent LY-411575, as tumor progression was comparable to two untreated controls. The mice receiving a single administration of PL33 at sub-curative concentrations also had minimal antitumor effects (% Treatment/Control (T/C) of 77.02 and 78.46, respectively, at d21), which were not statistically significant from two untreated controls or the single agent LY-411575.
Importantly, the 1 and 3 mg/kg PL33 in combination with LY-411575 did not show toxicity in animals (Fig. 7C) and further decreased tumor burden (%T/C 25.63 and 52.80, respectively, at d21) when compared with either agent alone, with the 1 mg/kg PL33 in combination showing stronger response compared to the 3 mg/kg PL33 in combination. The MM growth inhibition in the 1 mg/kg PL33 with LY-411575 showed statistically significant differences from the single agent LY-411575 and PL33 matched doses, as well as the untreated controls (at d21, P = 0.0.0012 for 1 mg/kg combination vs. GSI; P < 0.03 for 1 mg/kg combination vs. ctrl; P < 0.008 for 1 mg/kg combination vs. ctrl w/o AdT). Cohorts for two untreated controls and the LY-411575 single agent reached end tumor volume by d21, but the remaining groups were continued on the study to assess survival advantage (Fig. 7D). The combined 1 mg/kg PL33 with LY-411575 treatment significantly improved animal survival compared to 1 m/kg PL33 alone (P < 0.01). Additionally, at the time that all mice treated with 1 mg/kg PL33 alone (median survival day 21; 1PR + 0CR) had been removed from study due to tumor progression, 50% of animals in the group receiving combination 1 mg/kg PL33 with LY-411575 were still alive (median survival 34 days, 1PR + 4CR). The 3 mg/kg PL33 only and combination with LY-411575 resulted in 3PRs + 1CR and 5PRs + 2CRs, respectively, suggesting an improved treatment effect in the combination, even though statistical significance in host survivals was not realized. The Kapan-Meier Time to Endpoint plot indicated a trend toward prolonged survival in the combination groups over the single agent PL33.

DISCUSSION
We here show that GSIs effectively restored BCMA loss on the MM cell membrane by blocking GS cleavage in releasing sBCMA into culture supernatants of patient MM BMMCs. This potent GSI action led to enhanced anti-MM activity of BCMAxCD3 BisAb, without adverse impact on T-cell activation and memory cell differentiation in ex vivo 7 day-co-cultures. Importantly, a single low-dose GSI rapidly cleared sBCMA in serum samples of NSG mice bearing KMS11-luc myeloma. While two consecutive administrations of LY-411575 did not impact tumor growth, combination LY-411575 with a single sub-curative PL33 treatment significantly increased anti-MM efficacy of PL33 monotherapy, evidenced by extended host survival in the KMS11-luc MM xenograft-AdT mouse model.
The use of GSIs that were originally developed for Alzheimer's disease and cancer is primarily based on the premise that GSIs act by inhibiting the cleavage of GS on Notch ligands, thereby resulting in Notch 1 signaling blockade [34][35][36][37][38]. However, their success in preclinical models has not yet translated into clinical benefit [39]. Despite disappointing clinical performance, our data here further support repurposing GSI use to effectively augment BCMA targeting to achieve more complete and durable MM cell lysis by BCMAxCD3 BisAbs, as recently demonstrated in BCMA CAR T therapies [40].
Because of its highest potency to block BCMA cleavage without direct apoptotic effects on MM cells and T cells, 2 nM LY-411575 (ED 50 of 0.07 nM) was used in our studies. LY-411575 most efficiently retained mBCMA while diminished sBCMA for~2 days. Accumulation of mBCMA molecules by LY-411575 quickly reached near maximum within 4 h and sustained with barely detectable sBCMA for over 42 h. Importantly, LY-411575 depleted sBCMA by >20-50-fold in patient samples, further confirming that GSI blocks BCMA shedding from MM cells in patients. Despite different kinetics seen in MM cells expressing various baseline BCMA levels or % BCMA + patient CD138 + cells, GSIs robustly upregulated mBCMA densities and downregulated sBCMA in all MM cell lines and patient MM cells, regardless of disease and drug resistance statuses. As expected, other MM antigens including CD138, SLAMF7, CD38, and GPRC5D (data not shown) were unchanged since they were not the substrates for GS.
We further confirmed that mBCMA density is correlated with sBCMA concentration due to continuous GS cleavage under normal physiological conditions. Serum sBCMA levels were significantly associated with MM disease burden, supporting sBCMA as a valuable biomarker and highlighting sBCMA as a potential drug sink to abrogate effective BCMA-directed therapies. Indeed, sBCMA at patho-physiological concentrations (12.5-200 ng/mL) observed in MM patients inhibited MM cell lysis by BCMAxCD3 PL33 in a dose-dependent manner. Since RRMM patients have higher disease burden secreting even more sBCMA into the circulation [20,21,23], coupled with impaired effector Tcells, GSI treatment could provide a very promising strategy to overcome sBCMA interference, thereby allowing even efficient MM cell targeting and T effector cell engagement to kill MM cells by BCMAxCD3 BisAbs. Importantly, we showed that GSI potently led to increased autologous patient MM cell killing.
In 3 day-co-cultures, GSI did not further extend PL33-induced Tcell activation and checkpoint marker expression, suggesting that GSI minimally induced T-cell exhaustion. GSI further increased % EM and/or TEMRA T-cell subsets, associated with enhanced MM cell lysis. Consistent with our previous report on another BCMAxCD3 AMG 701 [14], PL33 also induced higher % of Treg, IL10 + , and TGFβ + T-cells at later time points. Deceased % T-cell subsets with inhibitory characteristics were seen in GSI treatment vs. control media cohorts. These findings are consistent with studies reporting that GSI (i.e., LY 411575, DAPT) treatment blocked expression of TGFβ1-induced Foxp3 and its downstream genes, as well as Treg suppression on naive T-cell proliferation [41,42]. Moreover, LY-411575 did not alter transient induction patterns of key co-stimulatory and immune regulatory markers on patient CD8 + and CD4 + T-cells by PL33 in 7 day-co-cultures. It affected neither PL33-induced time-dependent increase in % CM + EM nor CD8/CD4 ratios in patient T cells at d7. Thus, GSI treatment spared BCMAxCD3 BisAb-induced patient effector T-cell function and memory cell differentiation, suggesting that combination therapy can trigger persistent anti-MM activity in the clinic.
Since inhibition of BCMA cleavage by GS is reversible, we performed two GSI doses in consecutive days to prevent tumor  3 . Serum was serially collected following indicated time periods to evaluate sBCMA levels. Shown are fold reduction in sBCMA from baseline (time 0) (upper panel) and sBCMA pg/mL (lower panel). Data are presented as means ± SEMs (error bars). B-D NSG mice were injected with PBMCs (AdT) 5 days prior to KMS11-luc cell implantation on d0 and mice were randomized into their respective groups on d8 when average tumor size was~200 mm 3 . LY-411575 (3 mg/kg, QDx2) was administered PO as a single agent GSI and in combination groups. On the following day (d9), the second dose of LY41175 was administered followed by a single dose of PL33 to the single agent PL33 groups and combination groups at a dose 1 or 3 mg/kg IV. Cohorts included: 3 mg/kg LY-411575 (purple circle), 3.0 mg/kg (dark green triangle) or 1.0 mg/kg (light green triangle) PL33, LY-411575 + 3 mg/kg PL33 (maroon solid square) LY-411575 + 1 mg/kg PL33 (red solid square), tumor only (ctrl, black circle), tumor without PBMCs (ctrl w/o AdT, gray circle). B Shown are mean tumor volumes (mm 3 ) ± SDs (error bars) at following days vs. start of treatment. *P < 0.05 for tumor growth inhibition of single agent and combination treatments (one-way ANOVA followed by Tukey's multiple comparison test and unpaired t-test between single agent PL33 compared to combination. C Weights of mice were followed. D Conditional survival plot with time to endpoint set when tumor burden (TB) reached~1200 mm 3 . Kaplan-Meier and log-rank (Mantal-Cox) analysis followed by multiple comparison tests (Holm-Sidak method) was used to estimate median overall survival of animals (1 mg/kg PL33 with GSI, 34 days; the other 6 groups, 21 days) (P < 0.01). *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. cells still escaping BCMAxCD3 BisAb recognition in vivo. The first dose shifted the dynamic ratio of sBCMA to mBCMA to enhance target antigen on the membrane and simultaneously decrease the sBCMA sink. The administration of the second dose of GSI ensured that this shift is maintained through the BisAb dosing interval. Thus, LY-411575 was administered once one day prior to PL33 treatment to decrease sBCMA sink and the following day to maintain sBCMA suppression. Importantly, compare with PL33 monotherapy, combination LY-411575 with a single sub-curative PL33 treatment significantly enhanced KMS11 MM cell growth and prolonged host survival in the AdT-NSG mouse model.
In conclusion, GSI prevents mBCMA loss from MM cells and overcomes the sBCMA decoy neutralization, thereby augmenting MM cell targeting and efficacy of BCMAxCD3 BisAbs. A single treatment with GSI at~2-log lower drug concentration than used to induce anti-cancer activity maximizes mBCMA density without harming normal cells. This could avoid severe gastrointestinal toxicity of GSIs seen in earlier studies, which limits its potential clinical use [43][44][45][46]. Rationally incorporating GSI into all BCMAtargeting immunotherapy therefore represents a promising novel combination approach to further improve response durability and patient outcome in MM.

DATA AVAILABILITY
The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.