Stimulation of natural killer cells with small molecule inhibitors of CD38 for the treatment of neuroblastoma

High-risk neuroblastoma (NB) accounts for 15% of all pediatric cancer deaths. Refractory disease for high-risk NB patients is attributed to chemotherapy resistance and immunotherapy failure. The poor prognosis for high-risk NB patients demonstrates an unmet medical need for the development of new, more efficacious therapeutics. CD38 is an immunomodulating protein that is expressed constitutively on natural killer (NK) cells and other immune cells in the tumor microenvironment (TME). Furthermore, CD38 over expression is implicated in propagating an immunosuppressive milieu within the TME. Through virtual and physical screening, we have identified drug-like small molecule inhibitors of CD38 with low micromolar IC50 values. We have begun to explore structure activity relationships for CD38 inhibition through derivatization of our most effective hit molecule to develop a new compound with lead-like physicochemical properties and improved potency. We have demonstrated that our derivatized inhibitor, compound 2, elicits immunomodulatory effects in NK cells by increasing cell viability by 190 ± 36% in multiple donors and by significantly increasing interferon gamma. Additionally, we have illustrated that NK cells exhibited enhanced cytotoxicity toward NB cells (14% reduction of NB cells over 90 minutes) when given a combination treatment of our inhibitor and the immunocytokine ch14.18-IL2. Herein we describe the synthesis and biological evaluation of small molecule CD38 inhibitors and demonstrate their potential utility as a novel approach to NB immunotherapy. These compounds represent the first examples of small molecules that stimulate immune function for the treatment of cancer.


Introduction
Neuroblastoma (NB) is a pediatric malignancy that occurs during fetal or early postnatal development. It is the most frequently diagnosed pediatric cancer during infancy and accounts for 15% of all pediatric cancer deaths. 1 Nearly half of all NB patients will be classied as having high-risk disease, which is therapeutically challenging and has a poor prognosis. 2 The advent of anti-ganglioside 2 (GD2) chimeric monoclonal antibody (mAb) immunotherapy for high-risk NB has improved 5 year survival for patients, but overall survival remains unacceptably low at under 50%. 3,4 In high-risk NB patients, anti-GD2 mAbs such as naxitamab induce antibody-dependent cellmediated cytotoxicity (ADCC). However, the overall success of anti-GD2 mAb immunotherapy for NB is highly dependent on the antitumor activity of natural killer (NK) and other effector cells, 5,6 and failure to respond to treatment can be attributed to NB cell resistance 7 or the inability of effector cells to kill tumor cells. 8,9 As a result, the long-term efficacy of anti-GD2 mAb immunotherapy is unveried, and there is a pressing need for novel strategies to overcome resistance. 7 It has been established that high-risk NB patients with increased RNA signatures for activated NK cells and CD8 + T cells experience improved outcomes. 10 This implies that agents that prevent down regulation of immune function in the tumor microenvironment could represent a strategy for overcoming resistance to immunotherapy in NB and other cancers. The ectoenzyme cluster of differentiation 38 (CD38) has emerged as a potential target for immunomodulation 11 and the anti-CD38 immune checkpoint inhibitors daratumumab and isatuximab have been approved for use in diffuse large B cell lymphoma, follicular lymphoma, mantle cell lymphoma and multiple myeloma (MM). 12,13 CD38 is notable for its eccentric expression pattern, with predominant expression occurring in early and late stage T and B cell lymphocyte maturation. 14 In addition to T, B, and myeloid cells, CD38 has been found to be constitutively expressed in NK cells, 15,16 and is also a prognostic factor for multiple cancer types. 17 In addition, CD38 plays a critical role in the homeostatic regulation of cellular energetics. 18,19 By metabolizing the cofactor NAD + , CD38 removes an essential electron acceptor, 20 thus limiting the energetic capacity of a cell. Importantly, CD38 over expression in immune cells and tumor cells within the tumor microenvironment (TME) causes a reduction in NAD + levels, leading to down regulation of the immune response against tumor cells. 21 CD38 is a multifunctional enzyme, exhibiting both hydrolase and cyclase enzymatic activities, and regulates both a dominant and an alternative adenosine (ADO) pathway (Fig. 1). 20,22 The better-known pathway involves the nucleoside triphosphate diphosphohydrolase known as cluster of differentiation 39 (CD39). 22 The optimal pH for CD39-mediated hydrolysis of ATP and ADP is 8.0-8.5, and in normal tissue the CD39 pathway is the predominant source of exogenous ADO. However, it has been suggested that ADO production via CD38 22,23 (Fig. 1) likely predominates in the acidic TME. [24][25][26][27] ADO is considered a crucial mediator of the immune response (Fig. 2), and ADO receptors are known to be expressed in various immune cells, where they mediate the regulation of immune and inammatory responses. 28 Extracellular ADO, which is prominent in the TME, stimulates ADO receptor subtype 2A, (A 2A AR) 29,30 on immune cells, including T cells, natural killer cells, neutrophils, macrophages and dendritic cells, preventing their activation and driving naïve CD4 + T cell differentiation toward a Treg immunosuppressive phenotype. 31,32 In NK cells, ADO suppresses their cytotoxic activity toward tumor cells and their production of IFN-g, tumor necrosis factor (TNF-a) and granulocyte-macrophage colony-stimulating factor (GM-CSF), which are critical cytokines for effective ADCC. 29,30,33,34 In tumor cells that highly express CD38, such as MM, [35][36][37] and acute lymphocytic leukemia (ALL), 41,42 both reduction in NAD + levels and overproduction of ADO lead to immunosuppression. Recent evidence suggests that CD38 up regulation is one of the most important factors in mediating resistance to checkpoint blockade in MM and other cancers. [43][44][45] In addition, resistance to PD-1/PD-L1 blocking antibodies is mediated through up regulation of CD38 and subsequent production of ADO. 43 Numerous small molecule NAD + mimetics have been developed and tested for CD38 inhibition in the context of aging, mitochondria dysfunction, obesity, and diabetes. 18,19,[46][47][48] Notably, some of these inhibitors were successful in enzymatic studies, and in some cases promoted increases in the levels of NAD + in vivo, but none were evaluated for antitumor or immunostimulatory effects. 19,46,48,49 Our group recently reported that enzymatic inhibition of CD38 hydrolase or cyclase activity in activated human peripheral blood mononuclear cells (PBMCs) resulted in an 82% increase in cellular NAD + and a >100-fold increase in interferon gamma (IFN-g) secretion. 20 We now report the design and synthesis of a limited series of quinazoline-dihydropyrimidine-based CD38 inhibitors related to compound, 1 20,50,51 (Compounds 2-13, Table 1) and describe their immunostimulatory and pro-proliferative effects on NK cells. These analogues can be used to enhance the cytotoxic effect of NK cells toward NB cells in vitro.

Correlation of CD38/CD73 expression in NB
To support the contention that the CD38 adenosinergic pathway was prevalent in neuroblastoma, we performed in silico analysis of CD38 and CD73 expression levels in neuroblastoma samples from 786 patients in the Cangelosi neuroblastoma database. 52 As shown in Fig. 3, there is a strong correlation between expression of CD38 and CD73 (p = 5.3 × 10 −27 ) and between CD38 and CD203a (p = 10 × 10 −8 ), suggesting that this pathway is actively producing ADO, leading to immunosuppression. As indicated by the associated heat map, a large percentage of these samples had high expression of all three of these enzymes.
Preliminary structure activity relationship analysis. To facilitate structural optimization of compound 1, a synthetic Fig. 1 CD38-mediated extracellular ADO generation under neutral and acidic conditions. At neutral pH, CD38 hydrolyzes NAD + or cADPR to ADPR. CD203a then hydrolyzes ADPR to AMP followed by CD73mediated conversion of AMP to ADO. At acidic pH in the TME, CD38 converts NADP to NAADP via a base exchange reaction. NAADP is then hydrolyzed by CD38 to ADPRP which is converted to AMP and ADO by CD203a and CD73, respectively. route to 1 and the previously unknown compounds 2 and 3 was completed in 3 steps, as shown in Scheme 1. A modied Skraup synthesis was used to convert 2,4-dimethylaniline 15a to the corresponding 1,2-dihydroquinoline 16a (Sc(OTf)3, acetonitrile, heat). 53 Intermediate 16a was treated with 2-cyanoguanidine 17 to form the biguanide intermediate 18a, 54 which was converted to 1 in the presence of 4-methylpent-3-en-2-one. 55 A structure search in SciFinder® revealed 200 analogues related to 1, which were available in our in-house South Carolina Compound Library (SC 3 ) or purchased (Vitas-M Laboratory, Hong Kong). These compounds were evaluated for CD38 hydrolase inhibition (data not shown) in our previously published assay. 20 Compounds 2-13 (Table 1) proved to be the most potent inhibitors, and were evaluated for structural similarity using the Tanimoto method. 56 The known CD38 inhibitor 14 was included as a positive control in all experiments at  In silico studies. We next created a model of the potential interactions and bonding of 2 with CD38 at the molecular level through in silico docking and molecular dynamics. Using molecular operating environment (MOE) Dock, compound 2 was docked with a CD38 X-ray crystal structure (CD38 E226 , RCB  PDB: 2I66, Fig. 5A and B). 57 Docking parameters were set to allow the receptor and ligand to ex, and to predict enzyme/ ligand affinity. Molecular dynamics were run as described in the Experimental section, revealing several potentially required interactions. The secondary amine at the 2-position of the dihydroquinazoline ring of 2 appears to form a 1.7 Å hydrogen bond with Asp175. Likewise, the nitrogen at the 3-position of the dihydroquinazoline ring forms a 2.4 Å hydrogen bond with Lys178, the sp 2 nitrogen alpha to the gem dimethyl of the dihydropyrimidine ring hydrogen bonds to Trp176 (2.4 Å) and the dihydroquinazoline methoxy oxygen bonds to NAD + (2.0 Å). There also appear to be pi-pi interactions between the dihydroquinazoline aromatic ring and the adenine ring of NAD + (not shown). These data will be useful to facilitate the structurebased optimization of 2.
Effect of CD38 hydrolase inhibition in peripheral blood NK cells. Human peripheral blood (PB) NK cells (StemCell Technologies, Vancouver, CA, purity > 90%) were transferred to a 96well plate at a conuency of 40 000 cells per well in RPMI 1640 + 100 IU mL −1 IL-2 with or without 1.0 mM 2 for 48 hours, and live cells were stained with Hoechst (nal concentration 1 mM) and imaged at 4× magnication using a BioTek Cytation 5 imager ( Fig. 6A-C). Cells in each treatment group were normalized to vehicle control and data was analyzed as PB NK cell area % difference as a function of compound concentration (Fig. 6C). A dose-response relationship for 2 and 14 is shown in Fig. 6D. Treatment with 1.0 mM compound 2 produced a 190 ± 36% increase in nuclear area. Additionally, there was extensive clumping in cells treated with 2 relative to vehicle control. Cell clumping is characteristic of proliferating immune cells. 58,59 These data suggest that treatment of NK cells with inhibitors of CD38 hydrolase experience an expansion-like effect similar to what is observed when NK cells are expanded with IL-2 and/or feeder cells. 59 We previously demonstrated that compounds related to 2 have the ability to increase cellular NAD + and IFNg levels in human PBMCs in vitro. In light of the substantial change in viability/proliferation of PB NK cells following treatment with 2, changes in the ability of treated NK cells to secrete IFNg were measured. 60,61 Human PB NK cells (purity > 90%) were added to a 96-well plate at a conuency of 49 000 cells per well in  Immunocult-XF T cell expansion medium supplemented with 500 IU mL −1 IL-2, 10 ng mL −1 IL-15, and 0.2 mL mL −1 CD2/CD3/ CD28 T cell activator. These cells were exposed to varying concentrations of 2 or 14 for 24 hours, followed by quantication of IFNg using a Lumit IFNg assay kit (W6040, Promega, Madison, WI). Treatment groups were normalized to vehicletreated control and data was expressed as percent change in IFNg. The resulting IFNg data for donors 1-3 are presented in Fig. 7. As oen happens, we encountered a wide variability in the response of human primary cells from different donors. Donor 1 exhibited the most robust dose-dependent increase in IFNg production with an 759 ± 45% increase at 1.0 mM 2. Likewise, donor 2 exhibited a 100 ± 29% dose-dependent increase in IFNg at 0.111 mM 2 while donor 3 exhibited an 18 ± 2% increase in IFNg at 0.037 mM 2. While the amplitude of the change in IFNg varied signicantly from donor to donor, a concentration-dependent response was observed in all donors treated with 2. The known CD38 inhibitor 14 produced increases in IFNg levels that were similaer to the effects of 2 in all 3 donors. Despite the variability in the magnitude of the IFNg response, the trend in all 3 donors was a dose-dependent increase for both 2 and 14. Decreases in IFNg expression at higher doses of both compounds in donors 2 and 3 may be due to modest cytotoxicity.
The TME has the ability to down regulate NK cell activation and function, therefore it is important to discern whether inhibition of CD38 hydrolase activity over an extended period will improve NK cell activation and support a persistent response. 62 To address this question, donor 3 PB NK cells were plated on a 12-well plate at a conuency of 1 × 10 6 cells mL −1 in Immunocult-XF T cell expansion medium supplemented with 500 IU mL −1 IL-2, 10 ng mL −1 IL-15, and 0.2 mL mL −1 CD2/CD3/ CD28 T cell activator. Cells were treated with vehicle control, 2, or 14. Cells were treated with fresh medium every 3 days over the course of 20 days. Cells from each treatment group were counted on the 20 th day of treatment and viability determined using trypan blue. NK cells treated with 2 or 14 were normalized to vehicle control with the percent change in viable cells presented as a function of the treatment condition (Fig. 8A). We observed 43 ± 6% more viable cells in the group treated with 14 and 36 ± 13% more viable cells in the group treated with 2, relative to vehicle control. Additionally, on the 20 th day of treatment the cell supernatant was collected and analyzed for IFNg using the Promega Lumit IFNg assay. Treatment groups were normalized to vehicle control and the data presented as IFNg % difference as a function of treatment (Fig. 8B). We observed 23 ± 2% and 20 ± 1% more IFNg secreted by PB NK cells treated with 14 and 2, respectively, relative to vehicle control. Collectively, this data indicates that long-term treatment with inhibitors of CD38 hydrolase activity produce more viable cells, and subsequently more IFNg, relative to vehicle treated control.
Undifferentiated SHSY5Y NB cells grow in mounding clusters, which is potentially problematic for cell imaging. 63 Therefore, prior to any co-culture experiments, we wanted to determine an optimal cell plating number per well. Since compound 2 had a signicant effect on stimulation of NK cell proliferation, we used this compound to optimize the assay conditions. These experiments suggested that plating at 20 000-30 000 cells per well produced an optimal signal to noise ratio with more precision across all wells (Fig. 9). Again, compound 2 IFNg was quantified using a Promega IFNg lumit assay. Each data point is the average of readings from at least 3 separate wells ± SEM. Data analyzed by multiple comparison two-way ANOVA: *p # 0.05, **p # 0.01, ***p # 0.001, ****p # 0.0001. was employed to optimize assay conditions. As shown in Fig. 10, a concentration-dependent response with ch14.18-IL2 treatment was observed, and 50 ng mL −1 of ch14.18-IL2 was determined to be optimal.
Dinutuximab is the standard of care immunotherapy for high-risk NB treatment. The high cost and unavailability of dinutuximab made it unsuitable for our studies, therefore, we chose an alternative clinically relevant targeted biologic. The fusion protein ch14.18-IL2 contains a chimeric anti-GD2 antibody (ch14.18) tethered to recombinant human IL-2. 64 The IL-2 portion of the immunocytokine activates NK cells via the IL-2 receptor instead of the Fcg receptor as seen with dinutuximab. 65,66 In addition, it has exhibited activity in NB-bearing mice via NK-mediated effects and enhanced antitumor activity when compared to anti-GD2 antibody in combination with IL-2. 67,68 Thus we examined whether combined treatment with ch14.18-IL2 and an inhibitor of CD38 would enhance the cytotoxic effects of NK cells. To conserve PB NK cells, the concentration of ch14.18-IL2 was optimized to elicit a sufficient response using an effector cell : target cell ratio of 1 : 1.
Following optimization, SHSY5Y-GFP cells were plated at 25 000 cells per well and incubated for 24 hours. PB NK cells stained with cell tracker deep red were added to SHSY5Y-GFP cells with 50 ng mL −1 ch14.18-IL2 and treated with 2, 14 or vehicle control. Cells were incubated for 90 minutes at 37°C and imaged using a BioTek Cytation 5 at 20× (Fig. 11A-C). Live SHSY5Y-GFP cells were quantitated by measuring the integrated GFP uorescence intensity. Aer only 90 minutes treatment with a 1.0 mM concentration of 2 and 50 ng mL −1 ch14.18-IL2a caused a 14 ± 3% decrease in SHSY5Y-GFP uorescent area ( Fig.  11B) relative to vehicle-treated control (Fig. 11A). Interestingly, there was not a statistically signicant difference between cells treated with 1 mM 14 (Fig. 11C) and vehicle-treated control. These results are depicted graphically in Fig. 11D. Fig. 12 is a close-up view from a representative well showing the effects of 1 mM 2 combined with 50 ng mL −1 ch14.18-IL2.

Discussion
CD38 up regulation is thought to be one of the most important factors in mediating resistance to checkpoint blockade in MM and other cancers. [43][44][45] CD38-targeted biologics are currently used in the clinic to treat MM, but they do not mitigate the immunomodulating effects of CD38. The CD38-targeting biologic daratumumab has modest inhibitory activity against CD38 cyclase activity and enhances CD38 hydrolase activity. 69 In addition, CD38 mAbs may also mask regions of the CD38 epitope that are necessary for important CD38 receptor functions that promote NK cell interferon secretion and tumor cell cytotoxicity. Thus CD38-targeted biologics like daratumumab would, in theory, enhance extracellular ADO production and may in fact inappropriately propagate treatment resistance. This may account for the observation that a portion of MM   patients do not respond to antibody therapy and nearly all patients will ultimately become refractory to treatment. 70,71 Targeting CD38 enzymatic activity in the TME presents a potential new strategy for combating the immunosuppressive effects that contribute to treatment resistance to antibody therapy in high-risk NB. We postulate that this approach may also be of value in other cancers featuring CD38 expression and disproportionate NK cell populations.
The known CD38 inhibitor 14 increased NAD + levels in vitro and in vivo, but was developed for use in metabolic disease, and was not evaluated in the framework of cancer immunotherapy. 19,20,48,49 We have explored a quinazoline dihydropyrimidine scaffold from a hit molecule, compound 1, previously identied by our laboratory as an inhibitor of CD38-hydolase activity. 20 Preliminary SAR analysis of 12 commercially available or newly synthesized molecules, indicates that the dihydropyrimidine moiety (1-3) or its isosteric equivalent (4, 5 and 7) is optimal for activity. Alkyl substitution at the 6-and 8positions of the quinazoline ring are well tolerated. Furthermore, substitution at the 4-and 6-positions of the dihydropyrimidine ring is also tolerated without severely diminishing activity. This appears to be supported by in silico modeling (Fig. 5), which suggests that the 6-position on the quinazoline ring and the 6-position on the dihydropyrimidine ring are solvent exposed.
In using 14 as a positive control, we were surprised that the immunostimulatory effects of this known CD38 inhibitor (IC 50 78 nM) were signicantly less than those of 2 (IC 50 1.9 mM) despite having a much lower IC 50 value against CD38. One potential explanation for this observation may lie in the ultimate cellular location of the analogues. CD38 is known to function both as an intracellular and extracellular enzyme. 72 We hypothesize that in addition to IC 50 , the activity of our compounds is dependent on the compartmentalization of 2 versus 14. It is possible that 14 penetrates into cells, where inhibition of CD38 would only marginally affect ADO levels in the tumor microenvironment. By contrast, 2 may not penetrate into cells as readily as 14, and thus its CD38 inhibitory activity would be extracellular. The observed effects of 2 on IFNg levels and the associated immunostimulation likely result from a reduction of extracellular ADO levels. When an optimized analogue of 2 becomes available, we will undertake experiments to test this hypothesis.
Among the analogues structurally related to the parent molecule 1, the synthetic analog 2 exhibits the most potent activity against CD38 (IC 50 1.9 mM). Substitution of a methoxy substituent for the methyl group at the 8-position of the quinazoline ring improved potency more than 2-fold. In silico modeling (Fig. 5) suggests that there is a hydrogen bond  interaction between the methoxy group of 2 and the aromatic adenosine moiety on NAD + that may account for the increase in activity. We previously reported 20 that compound 1 inhibits CD38 hydrolase activity via mixed inhibition. This observation is in agreement with data for the known CD38 inhibitor 14, which also exhibits mixed inhibition kinetics. 49 For our in silico model in Fig. 5, we chose to use the PDB X-ray structure 2I66 (CD38 mutated at E226 in complex with NAD + ) rather than PDB 4XJT, where CD38 is mutated at E226 and ADPR is covalently bound to the active site. 73 The docking pose of 2 in PDB 2I66 strongly supports a mixed mechanism of inhibition, since 2 binds partially in the NAD + catalytic pocket and partly outside of it. However, co-crystallization of 2 with CD38 complexed with NAD + would be necessary to conrm this suggested binding pose.
Evaluation of compound 2 in PB NK cells suggest that small molecule inhibition of CD38 hydrolase activity promotes proliferation and prolonged viability of NK cells and increases extracellular secretion of IFNg. The enhanced proliferation of NK cells was statistically signicant aer a 48 hour treatment (Fig. 6), and the observed increase in NK cell proliferation and increase in IFNg secretion persisted throughout a 20 days treatment (Fig. 8). A previous report that CD38 knockout NK cells secreted more IFNg relative to wild type NK cells 74 coupled with our data suggest that CD38 enzymatic activity may be a source of NK cell exhaustion. 74 Inhibition of CD38 hydrolase activity in NK cells might be a useful strategy for invoking an activated NK cell phenotype. Treatment of peripheral blood NK cells with 50 ng mL −1 ch14.18-IL2 and compound 2 exhibited enhanced cytotoxic effects of NK cells toward NB cells. While the mechanism of the observed cytotoxic effect has not yet been elucidated, these preliminary studies indicate that inhibition of CD38 hydrolase activity might be effective in enhancing NB immunotherapy used in clinic. Mechanistic studies, as well as the synthesis and evaluation of additional compounds related to 2 are ongoing concerns in our laboratory.

Conclusions
In this report we describe the discovery, synthesis and biological characterization of a novel series of small molecule CD38 inhibitors for use in the treatment of NB. To our knowledge, these are the rst small molecules designed for the stimulation of immune cells in cancer immunotherapy. Our studies have shown that the CD38 inhibitor 2 (IC 50 1.9 mM) is a potent inducer of IFNg in vitro and that it promotes the proliferation of NK cells. Further, we have developed an in vitro NK/NB coculture assay and demonstrated that 2 promotes a 14% decrease in the number of SHSY5Y NB cells aer only 90 minutes. Importantly, although both 2 and the known CD38 inhibitor 14 produced signicant increases in NAD + levels, 20 compound 2 was superior to 14 in terms of effects on IFNg, NK cell proliferation and production of NK cell-induced cytotoxicity in SHSY5Y NB in vitro. Because the observed increase in NB cytotoxicity was mediated by inhibition of CD38 and the resulting increase in NK cell proliferation, this approach may be useful in other cancers that express CD38. In addition, agents related to 2 may be of use in preventing or delaying the development of resistance to ADCC observed with currently used mAb immune checkpoint inhibitors.

Experimental
All reagents and dry solvents were purchased from Aldrich Chemical Co. (Milwaukee, WI), Sigma Chemical Co. (St. Louis, MO), VWR (Radnor, PA) or Fisher Scientic (Chicago, IL) and were used without further purication except as noted below. Triethylamine was distilled from potassium hydroxide and stored in a nitrogen atmosphere. Dry methanol, ethyl acetate, tetrahydrofuran, dimethyl formamide and hexane were either purchased (VWR) or prepared using a Glass Contour Solvent Purication System (Pure Process Technology, LLC, Nashua, NH). Microwave synthetic procedures were conducted on an Initiator 8 microwave synthesizer (Biotage, Charlotte, NC). Preparative scale chromatographic procedures were carried out using a Biotage Selekt chromatography system (Biotage, Charlotte, NC) tted with silica gel 60 cartridges (230-440 mesh). Thin layer chromatography was conducted on Merck precoated silica gel 60 F-254. Compound 14 was purchased from Selleckchem (Houston, TX), and compounds 1, 4-9, and 11-13 were purchased from Vitas-M Laboratory (Champaign, IL). Compound 10 was obtained from the SC 3 collection from the MUSC drug discovery core. All 1 H and 13 C-NMR spectra were recorded on a Bruker Avance 600 MHz spectrometer, and all chemical shis are reported as d values referenced to TMS or DSS. Splitting patterns are indicated as follows: s, singlet; d, doublet; t, triplet; m, multiplet; br, broad peak. In all cases, 1 H-NMR, 13 C-NMR and MS spectra were consistent with assigned structures, and 13 C peak assignments appear on the spectrum. Mass spectra were recorded by LC/MS on a waters UPLC/MS system with a model QDa mass spectrometer detector. Prior to biological testing procedures, all compounds were determined to be >95% pure by UPLC chromatography (9 : 1 H 2 O: acetonitrile, +0.1% formic acid to 1 : 9 H 2 O/acetonitrile +0.1% formic acid over 8 minutes) using a waters acquity H-series ultrahigh-performance liquid chromatograph tted with a C18 reverse-phase column (Acquity UPLC BEH C18 1.7 M, 2.1 × 100 mm).
Compounds were also tested for inhibition against CD38 cyclase activity at a concentration of 50 mM using the same enzyme assay, where a 2 × 50.0 mM working solution of NGD + was used in place of 3NAD + . No compounds were found to exhibit cyclase inhibitory activity greater than 20%.
Criteria for IC 50 determination. Compounds were evaluated for structural similarity to compound 1 and those with Tanimoto coefficients <0.8 were subjected to preliminary testing at a concentration of 50.0 mM. 20 Any test compounds that exhibited <65.0% remaining CD38 hydrolase activity in preliminary testing were further evaluated for IC 50 's in our recombinant enzyme assay as previously described. 20 Compounds with Tanimoto coefficients >0.8 were not subjected to preliminary testing and were evaluated for IC 50 's.
In silico docking and molecular dynamics. Modeling, simulations and structural visualizations were performed using MOE 2019 (Chemical Computing Group ULC, Montreal, CA) based on RCSB Protein Data Bank structure 2I66. 72 The protein was protonated at T = 310 K, pH 7.0, salt at 200 mM using GB/VI electrostatics. Docking simulations used exible receptor and exed the ligand, while docking targeted the active site. For each docking simulation initial placement calculated 50 poses using triangle matching with London dG scoring, the top 5 poses were rened using forceeld Amber10:ETH and Affinity dG scoring (Escore2). The top pose was used then rened using molecular dynamics. Molecular dynamics used the NPA algorithm and the Amber10:ETH forceeld. Solvent was a water droplet with 0.1 M NaCl and used 9518 solvent molecules. Simulation protocol was an equilibrium step for 100 ps at 300 K and a production step for 500 ps at 300 K with a step time of 0.5 ps.
Cell viability/proliferation assay. Human PB NK cells were plated (50 mL per well) at 40 000 cells per well in RPMI 1640 supplemented with 10% FBS, 1× penicillin/streptomycin, and 100 IU mL −1 IL-2 on a black, clear-bottom, 96-well microplate. Test compounds were dissolved in DMSO and diluted to a 2× working solution with cell culture medium. Working test compound or DMSO vehicle (50 mL) was added to cells and the cells were incubated at 37°C, 5% CO 2 for 48 hours. Following incubation, cells were treated with Hoechst (10 mL, nal concentration 1 mg mL −1 ), and imaged using a BioTek Cytation 5 imager. Cells were kept at 37°C, 5% CO 2 during imaging. Each well was imaged in 4 quadrants at 4× magnication using the DAPI uorescence channel. The uorescence intensity was measured as the sum of integrated uorescence, with treatment wells normalized to vehicle treated controls and data expressed as the percent difference of PB NK nuclear area. Each assay was tested in technical triplicate.
Quantitation of IFNg. Human PB NK cells were plated at 49 000 cells per well (60 mL per well) on a white, clear-bottom, 96well microplate in Immunocult-XF T cell expansion medium supplemented with 1× penicillin/streptomycin, 500 IU mL −1 IL-2, 10 ng mL −1 IL-15, 0.2 mL mL −1 CD2/CD3/CD28 T cell activator. Test compounds were dissolved in DMSO and diluted to a 4× working solution. Working test compound was added to the cells (20 mL) and the cells were incubated at 37°C, 5% CO 2 for 24 hours. The last 3 columns on the microplate were le empty for IFNg standards. The provided human IFNg standard (10 mg mL −1 ) was diluted to 10 000 pg mL −1 using cell culture medium. The 10 000 pg mL −1 solution was serially diluted 3.33fold for a total of 7 standard concentrations. The IFNg standards were plated (80 mL) in the empty columns (in triplicate) following incubation. A 5× solution of both antibodies was prepared by combining the Lumit Anti-hIFN-g-mAb-LgBiT (24 mL) and the Lumit Anti-hIFN-g-mAb-SmBiT (24 mL) with cell culture medium (2.4 mL). The 5× antibody solution (20 mL) was pipetted into each well. The microplate was briey mixed for 15 seconds on a shaker at 250 rpm and incubated for 90 minutes at 37°C, 5% CO 2 . Following incubation, the microplate was allowed to rest at 25°C for 15 minutes. The Lumit detection substrate (160 mL) was diluted with Lumit detection buffer B (3040 mL) and pipetted into each well (25 mL). The microplate was allowed to incubate for 5 minutes at 25°C and then the luminescence was measured using a Molecular Devices Spec-traMax iD3 plate reader. A standard curve with averaged RLU measurements of the IFNg concentrations was determined with GraphPad Prism using 4-parameter logistic curve tting. The test sample IFNg concentrations were interpolated using the standard curve. Data was presented as percent difference relative to vehicle treated control. Each assay was tested in technical triplicate.
Optimization of SHSY5Y-GFP cell culture conditions. SHSY5Y-GFP cells were plated at 20 000, 30 000, 40 000, and 50 000 cells per well with 6 test wells each. Cells were incubated for 24 hours and imaged with a BioTek Cytation 5 at 20× with 4 elds per well using brighteld and GFP channels. The data is presented as the sum of SHSY5Y-GFP cells in imaged elds as a function of cells per well (Fig. 9A). Optimal plating density was determined to be between 20 000 and 30 000 cells per well. Additionally, we wanted to rule out proliferative effects in SHSY5Y-GFP cells treated with 2. Cells were plated at 20 000 or 30 000 cells per well and treated with varying concentrations of 2 or vehicle control for 24 hours and imaged as described above. Data was then expressed as the sum of SHSY5Y-GFP cells in imaged elds as a function of the concentration of 2 (Fig. 9B). A cell number difference of <4% was observed for vehicle-treated cells and cells treated with 1.0 mM 2 at both 20 000 cells per well and 30 000 cells per well.
Optimization of ch14.18-IL2 levels for co-culture assay. To conserve PB NK cells, the concentration of ch14.18-IL2 was optimized to elicit a sufficient response using an effector cell : target cell ratio of 1 : 1. SHSY5Y-GFP cells were plated at 25 000 cells per well and PB NK cells stained with cell tracker deep red were added along with either 50 ng mL −1 or 125 ng mL −1 ch14.18-IL2. Cells were incubated for 90 minutes and the sum of GFP integrated uorescence was measured using a BioTek Cytation 5 imager.
SHSY5Y-GFP and PB NK cell co-culture assay. SHSY5Y-GFP NB cells were plated at a concentration of 25 000 cells per well (100 mL) for 12 hours on a black, clear-bottom, 96-well microplate. PB NK cells were washed with PBS and resuspended in 5 mM cell tracker deep red in RPMI 1640 without FBS for 30 minutes at 37°C, 5% CO 2 . Following incubation, cells were washed with PBS and resuspended in RPMI 1640 supplemented with 10% FBS. Compounds were dissolved in DMSO and diluted to a 4 mM working solution with cell culture medium. Ch14.18-IL2 was diluted in cell culture medium to a 200 ng mL −1 working solution. SHSY5Y-GFP cell medium was aspirated and PB NK cells (50 mL), compound (25 mL), and ch14.18-IL2 (25 mL) were added to each well. Cells were incubated at 37°C, 5% CO 2 for 90 minutes. Cells were imaged at 20×, 4 elds per well, using bright eld, GFP, and Cy 5 channels using a BioTek Cytation 5 imager. The sum integration of GFP area was measured and normalized to vehicle-treated control. Each assay was tested in technical triplicate.

Data availability
All compound characterization data including 1 H and 13 C NMR, mass spectra and UPLC traces are available in the ESI. †

Conflicts of interest
There are no conicts to declare.