Targeted Delivery of Cytotoxic NAMPT Inhibitors Using Antibody–Drug Conjugates

This article is featured in Highlights of This Issue, p. 2493

Antibody–drug conjugates (ADCs) have attracted a great deal of recent interest, with FDA approval of four drugs and promising data from several others in various stages of clinical development (1). Most of the advanced ADCs contain antitubulin drugs such as auristatins and maytansines, or DNA-targeted drugs such as calicheamicins and camptothecins (2–5). While these payloads have been applied to a broad spectrum of tumor types, an important objective within the field has been to extend the technology beyond these drug classes and to identify novel agents that have complementary mechanisms of activity (1). This endeavor has proven to be quite challenging, because most drugs do not have the right profiles of potency, availability, chemical functionalities, and stability required for successful application as ADC warheads. Identification of new payloads will help advance this promising and rapidly advancing therapeutic modality.

We became interested in exploring inhibitors of nicotinamide phosphoribosyltransferase (NAMPT) as ADC payloads due to their chemical simplicity, high potencies on both their biochemical targets and on cells, and their unique mechanisms of activity (6). NAMPT is a key regulator of NAD levels in cells. In addition to serving as an essential redox cofactor to support energy and anabolic metabolism, NAD serves as a substrate for PARPs, sirtuins, and cyclic-ADP-ribose synthases (7). The nicotinamide byproduct from these enzymatic reactions is converted back into NAD via a salvage biosynthetic pathway involving NAMPT as the rate-limiting step. In this way, NAMPT maintains the NAD pool at a level sufficient to support metabolism. When NAMPT is inhibited, NAD levels decline below the level needed for metabolism leading to energy crisis and ultimately cell death. In contrast to many other cytotoxic molecules, the cell death following depletion of NAD proceeds via a nonapoptotic and proinflammatory mechanism termed oncosis (8). These mechanisms of action and cell death distinguish NAMPT inhibitors from current ADC payloads.

In addition, nontargeted small-molecule NAMPT inhibitors advanced to clinical trials but gave disappointing results. Clinical candidates FK-866, CHS-828, and GMX-1777 each encountered dose-limiting toxicities prior to any objective responses (9–11). More recent efforts to advance NAMPT inhibitors as small molecules appear to have met with unsuitable retinal and cardiac toxicities in preclinical species (12, 13), and there are no NAMPT inhibitors approved for clinical use at this time. We envisioned that a delivery approach using ADCs might circumvent the toxicities observed with systemic administration of NAMPT inhibitors, thus enabling the successful use of this drug class in the clinic (14). Here, we report that ADCs comprising NAMPT inhibitors (NAMPTi-ADCs) kill cells in an antigen-specific manner through depletion of intracellular NAD pools, and that the resulting conjugates display significant levels of activity in several tumor models at well-tolerated doses.

FK-866 was purchased from Cayman Chemical. Inhibitors and drug linkers were prepared using standard synthetic chemistry methods as previously described (15); detailed experimental procedures for compounds 2, 4, 5, 8, and the probe molecule (compound S12) used in the fluorescence polarization assay are provided in the Supplementary Methods file. Inhibitors were formulated in DMSO at 10 mmol/L and stored at −20°C. Drug linkers were formulated in dimethylacetamide (DMA) at 20 mmol/L and stored at −20 °C. PBS (pH 7.4) was purchased from Thermo Fisher Scientific (catalog no. 10010-023).

NAMPT bearing a C-terminal His-tag was expressed and purified from E. coli as described previously, except a single round of purification using nickel-agarose affinity chromatography was employed (16). Eleven-point, 2-fold serial dilutions of test compounds were prepared in assay buffer (50 mmol/L HEPES, 50 mmol/L KCl, 5 mmol/L MgCl₂, 125 μmol/L ATP, 0.005% BSA, 0.5 mmol/L β-mercaptoethanol, pH 7.2) starting at 10 μmol/L (2× final concentration). NAMPT and fluorescent probe molecule S12 were prepared at 2× final concentrations of 240 nmol/L and 60 nmol/L, respectively, in assay buffer. In black-walled 384-well plates, 15 μL of test compound and 15 μL NAMPT/probe preparations were mixed in quadruplicate and allowed to equilibrate for 4 hours. Fluorescence polarization was measured on an Envision multilabel reader (Perkin Elmer) using an installed FITC FP dual mirror. Measurements of polarization (milli-polarization units) are defined as (mP) = 1,000 × (S-G × P)/(S+G × P) where S and P represent the parallel and perpendicular background subtracted fluorescence count rates following polarized excitation, and G (grating) is an instrument dependent factor calculated from pure fluorophore solution. Binding data were analyzed using GraphPad Prism software using the four-parameter log(inhibitor) versus response model to provide EC₅₀ values.

The L540cy cell line was provided by Dr. Harald Stein (Institute fur Pathologie, University of Veinikum Benjamin Franklin, Berlin, Germany) and cultured in RPMI1640 with 20% FBS. The A549, HepG2, and Ramos cell lines were obtained from ATCC. The L-428, DOHH-2, and HNT-34 cell lines were obtained from DSMZ. EMEM + 10% FBS was used to culture HepG2 cells; RPMI1640 + 10% FBS was used to culture DOHH-2, HNT-34, L-428, and Ramos cells; F-12K + 10% FBS was used to culture A549 cells. All cell lines were authenticated by short tandem repeat profiling at IDEXX Bioresearch and cultured for no more than 2 months after resuscitation. Cells cultured in log-phase growth were seeded for 24 hours in 96-well plates containing 150 μL RPMI1640 supplemented with 20% FBS. Serial dilutions of antibody–drug conjugates in cell culture media were prepared at 4× working concentrations, and 50 μL of each dilution was added to the 96-well plates in duplicate. Following addition of test articles, cells were incubated with test articles for 4 days at 37°C; for kinetic experiments, cells were incubated for the indicated times (1–5 days). Viability, as indicated by ATP levels, was measured by CellTiter-Glo (Promega). NAD levels were measured using NAD/NADH-Glo (Promega). Luminescence was measured on an Envision Multilabel Reader (Perkin Elmer). The data was analyzed using GraphPad Prism software using the four-parameter log(inhibitor) versus response model to provide EC₅₀ values. EC₅₀ value for cellular assays is defined here as the concentration that results in 50% reduction in assay signal relative to untreated controls.

NAMPT was crystallized as described previously (17) with the notable exception that large diffraction quality crystals were obtained through a method of microseeding using a cat whisker. NAMPT–inhibitor cocrystals were obtained by soaking the previously grown protein crystals with 2 mmol/L compound 2 for 30 minutes before cryoprotection and flash freezing in liquid nitrogen.

NAMPT compound data were collected at data beamline 4.2.2 at the Advanced Light Source (Lawrence Berkeley Labs). Images were indexed and processed using XDS, and structure solution by the difference Fourier method and refinement were performed using the PHENIX package. Model building was carried out iteratively using the Coot software. Data collection and refinement statistics are given in Supplementary Table S1. The atomic coordinates have been deposited in the Protein Data Bank (www.rcsb.org; PDB ID code 6E68).

All antibodies used for this study were of the IgG1 isotype. The αCD30 antibody is clone cAC10/brentuximab (18, 19); the αCD19 antibody is clone hBU12/denintuzumab (20, 21); the αCD123* antibody is clone h7G3ec containing an S239C mutation (22). “IgG” denotes a human mAb with no known binding epitope; IgG* is the corresponding antibody with S239C mutation. Reduction of antibody interchain disulfides and disulfide-capped S239C residues was achieved by incubation with an excess of tris(2-carboxyethyl)-phosphine (TCEP) for 60 minutes at 37°C. Reduced mAb was then buffer-exchanged into PBS + 2 mmol/L EDTA using a NAP-5 desalting column (GE Healthcare). Full reduction was confirmed by reverse-phase chromatography (PLRP 3 μm, Agilent; detailed descriptions of all analytic methods are provided in the Supplementary Methods file). Conjugation was carried out at room temperature by adding 1.15 molar equivalents of drug linker per reduced cysteine. After 20 minutes, the extent of conjugation was assessed by evaluating the mass increase of antibody chains using reverse-phased UPLC (PLRP 3 μm, Agilent) coupled with mass spectrometric detection (Waters Xevo G2-S QTOF). Additional drug linker was added when needed. Once all available cysteines were occupied with drug, excess drug-linker was removed by incubation with activated charcoal followed by buffer exchange into PBS using a NAP-5 desalting column. Aggregation was assessed by size-exclusion chromatography. Homogenous drug loading at interchain disulfides, and S239C mutations sites where applicable, was confirmed by comparing predicted and observed changes in the molecular weights of heavy and light chains using PLRP-MS analysis and confirming absence of underloaded species. Characterization data for all ADCs is provided in Supplementary Table S2. Final ADCs were sterile-filtered through a 0.22-μm centrifugal filter and concentration was determined by the DC Protein Assay (Bio-Rad Laboratories).

ADCs were incubated in sterile rat plasma (Sprague Dawley) at 37°C. Time points were collected at 0 hour, 6 hours, 1 day, 3 days, and 7 days. ADC samples were recovered from plasma by batch purification using Anti-Human Capture Affinity Resin (IgSelect, GE Healthcare) for 2 hours. Bound samples were washed with PBS + 0.5 mol/L NaCl and eluted using 50 mmol/L glycine, pH 3. Eluted samples were neutralized with Tris pH 7.4, deglycosylated using PNGase F (New England BioLabs Inc) and reduced using 10 mmol/L DTT. Each sample was analyzed using reverse-phase UPLC (PLRP 3 μm, Agilent) coupled with mass spectrometric detection (Waters Xevo G2-S QTOF). The drug-to-antibody ratio of each sample was calculated using the relative intensity of the deconvoluted masses of the drug-loaded antibody chains and observed peaks indicating decomposition products.

All experiments were conducted in concordance with the Institutional Animal Care and Use Committee in a facility fully accredited by the Association for the Assessment and Accreditation of Laboratory Animal Care.

Pharmacokinetic studies were performed in naïve female Sprague Dawley rats (Envigo). ADCs with DAR = 8 were evaluated using a radiolabel methodology. To a solution of antibody or ADC in PBS supplemented with an additional 50 mmol/L potassium phosphate (pH 8.0) and 50 mmol/L sodium chloride was added 55 μCi N-succinimidyl propionate, [propionate-2,3-3H]- (Moravek Biochemicals, 80 Ci/mmol, 1 μCi/mL, 9:1 hexane:ethyl acetate solution) per mg of antibody or ADC. The resulting mixture was vortexed and left at room temperature for 2 hours. The mixture was centrifuged at 4,000 × g for 5 minutes and the lower aqueous layer was removed. Unconjugated radioactivity was removed by buffer-exchange into PBS using a NAP-5 desalting column followed by one round of dilution and centrifugation at 4,000 × g using an Amicon 30-kDa MWCO Ultra-15 Centrifugal Filter Units (Millipore). The resulting products were filtered through sterile 0.22-μm Ultrafree-MC Centrifugal Filter Units (Millipore) and the final antibody or ADC concentration was measured using the DC Protein Assay (Bio-Rad Laboratories). The specific activity (μCi/mg) of each product was determined by liquid scintillation counting.

Test articles were administered at 1 mg/kg via intravenous injection (n = 3 per ADC). Blood was drawn into K2EDTA tubes via the saphenous vein at various time points or by cardiac puncture for terminal bleeds. Plasma was isolated by centrifugation for 10 minutes at 10,000 × g. A 10–20 μL sample of plasma from each time point was added to 4-mL Ecoscint-A liquid scintillation cocktail (National Diagnostics) and the total radioactivity was measured by liquid scintillation counting. The resulting disintegrations per minute values were converted to μCi and the specific activity of the radiolabeled test articles was used to calculate the concentration of antibody or ADC remaining in the plasma at each time point. The data were fit to a noncompartmental model to derive the clearance rates using Phoenix WinNonLin v7 (Certara).

The IgG*-8 ADC (DAR = 10) was prepared without radiolabeling. Dosing of ADC and collection of samples was as described above. Total human IgG was detected in plasma using the Gyrolab platform (Gyros AB). Assay standards and quality control samples (QCs) were prepared using IgG*-8 ADC (DAR = 10) diluted in pooled female Sprague Dawley rat plasma. Standards, QCs, and study samples were diluted 10-fold into Rexxip buffer (Gyros AB). Briefly, a biotinylated murine anti-human IgG was captured onto streptavidin-coated beads within the Gyrolab Bioaffy CD. After being captured, human IgG was detected with an Alexa Fluor 647 (Thermo Fisher Scientific) labeled version of the same anti-human IgG. The fluorescence signal (in Response Units) was read at the 1% photomultiplier tube (PMT) setting. Unknown sample concentrations were determined by interpolating against a standard curve fit with a 5-parameter logistic function weighted by 1/y² using the Gyrolab Evaluator Software (Version 3.4.0.24).

Tumor cells, as a suspension, were implanted subcutaneously in immunocompromised SCID mice. Upon tumor engraftment, mice were randomized to study groups (n = 5) once the average tumor volume reached approximately 100 mm³. The ADCs were dosed by intraperitoneal injection at indicated times. For the HNT-34 model, animals were dosed with hIgG 24 hours prior to test article administration to reduce Fc-mediated ADC uptake into cells (23). Tumor volume as a function of time was determined using the formula (L × W²)/2. Animals were euthanized when tumor volumes reached 1,000 mm³. Complete response (CR) is defined here as absence of measureable mass at tumor site during the course of an experiment.

NAD measurements in xenografted tumors were obtained using an adaptation of the in vitro NAD/NADH-Glo assay. Tumor masses were collected from animals 4 days postdose, cut into portions of approximately 50 mg, and suspended in 1 mL 0.1 mol/L perchloric acid in lysing matrix D vials (MP Biomedicals). The tumors were homogenized in a bead beater using three 30-second cycles, cooling in an ice-water bath in between cycles. Homogenates were centrifuged at 16,000 × g for 10 minutes to pellet debris. Supernatant was removed and further diluted 1:5 in 0.1 mol/L perchloric acid. To measure NAD content, 3 μL of diluted tumor lysate was added to 75 μL buffer (0.2 mol/L Tris, pH 7.4) in triplicate, followed by 50 μL NAD/NADH-Glo detection reagent. After 30 to 60 minutes, luminescence was measured on an Envision multilabel reader (Perkin Elmer). NAD concentrations were interpolated from a standard curve of NAD using GraphPad Prism.

In vivo rat safety studies were done in naïve female Sprague Dawley rats (Envigo). Studies had 3 animals/group/timepoint and included a vehicle control group (1× PBS pH 7.4). Animals were administered a single dose of 30, 60, and 100 mg/kg IgG-8 (DAR = 8) via intravenous injection. Animals were necropsied and tissues were collected at 1 and 4 weeks postdose. Blood was sampled under isoflurane anesthesia via the jugular vein for periodic hematology analysis on the Sysmex XT-2000iV. At necropsy, blood was collected from the caudal vena cava for hematology and clinical chemistry panels and organs were collected for histologic evaluation. Clinical chemistry was analyzed on a Beckman Coulter AU680. The eyes were fixed in modified Davidson solution for 24 hours and transferred to 10% neutral buffered formalin. The remaining organs were fixed in 10% NBF. Tissues were processed routinely into 5-μm thick, hematoxylin and eosin (H&E)-stained slides for light microscopic evaluation by the same anatomic pathologist.

Development of a cleavable linker for drug release requires the presence of a reactive heteroatom in the drug, typically in the form of an amine, hydroxyl, or sulfhydryl functional group (24–29). FK-866 is devoid of these groups leading us to pursue new linker-compatible analogues (30). The cocrystal structure of FK-866 with NAMPT (16) shows a clearly defined pharmacophore consisting of (i) a heteroaromatic head group which mimics nicotinamide, (ii) a connecting functional group with H-bonding potential, (iii) a hydrophobic extension, and (iv) a “tail” group which makes weak contacts with the entrance to the binding site. Portions of the tail group are largely solvent-exposed, and we rationalized that modifications in this area might yield inhibitors that retain binding potency while incorporating the necessary chemical handle for linker attachment. For this study, a series of FK-866 analogues were prepared consisting of aniline substitutions at different positions on the aromatic tail (Fig. 1, compounds 1–3). To rank biochemical potency, the analogues were evaluated for binding to NAMPT using a competitive fluorescence polarization (FP) assay where they displayed varying reductions in EC₅₀ as compared with FK-866. The compounds were further evaluated for cytotoxicity against L540cy, A549, and HepG2 cell lines. Compound 2 displayed similar potency to FK-866 in both assays. To validate the binding mode of 2, we prepared crystals of NAMPT with the inhibitor bound (Fig. 1B). The key interactions of the head group pyridine and the acrylamide functional groups seen for FK-866 are maintained (16). Unexpectedly, the newly incorporated aniline group interacts with the backbone amide of Glu376 via a hydrogen bond, potentially offsetting any desolvation penalty for addition of a polar functional group.

Prioritizing the 3-NH₂ modification of the tail group, we next evaluated alternative head groups present within other known NAMPT inhibitors. Analogues of compound 2 were prepared containing the pyridyl-cyanoguanidine group found in the potent inhibitor CHS-828 (31) to give compound 4, and with a pyridyl-squaramide group (32) to give compound 5. These head group substitutions similarly retained strong binding to NAMPT in the FP assay and potent cytotoxicity in the cell-based viability assay (Fig. 1).

On the basis of the strong potency in biochemical and cell-based assays, compounds 2, 4, and 5 were advanced to evaluation as ADC drug payloads using a cleavable linker system (Fig. 2A). To maximize hydrophilicity of the drug linker, we employed a glucuronide trigger, which serves to activate drug release upon cleavage by β-glucuronidase in endosomes and lysosomes (27). For each payload drug, the tail group aniline moiety was joined to the core linker via a carbamate bond, providing a linkage that is chemically stable in plasma but that releases drug rapidly upon cleavage of the glucuronide moiety. Finally, an appended maleimide group provides a handle for bioconjugation to antibodies, with select maleimide formats providing for stabilization of the linkage and/or addition of a polyethylene glycol (PEG) chain to enhance hydrophilicity (33, 34).

Bioconjugation to antibodies was achieved by full reduction of interchain disulfides followed by incubation with the synthetic drug linkers to give ADCs with high degrees of homogeneity having a drug-to-antibody ratio of 8 (“DAR = 8”). In addition to native IgG1 backbones, we examined antibody backbones engineered to include an S239C mutation in the heavy chain (35). Full reduction of these engineered antibodies followed by bioconjugation gave ADCs with a DAR = 10. Both DAR = 8 and DAR = 10 ADCs showed minimal aggregation by SEC (Fig. 2B; Supplementary Table S2). An isotype control ADC using linker 8, containing the self-stabilizing mdpr maleimide, was evaluated for stability upon incubation in rat plasma. Over an 8-day incubation, there was no observed payload loss resulting from carbamate cleavage or linker loss resulting from deconjugation (Supplementary Fig. S1). Pharmacokinetic analyses nonbinding ADCs carrying linkers 6–8 (DAR = 8) were carried out in rats at a dose of 1 mg/kg and showed terminal half-lives in the range of 5–8 days, similar to other ADCs in clinical use (36, 37). A DAR = 10 ADC using linker 8 displayed a reduced terminal half-life of 2.5 days, but after one week still maintained plasma concentrations above the EC₅₀ values determined in in vitro assays (see below). The chemical stability and pharmacokinetic profiles of these ADC formats met our criteria for further preclinical evaluation.

To evaluate biological activity of NAMPT inhibitors delivered as ADCs, we prepared conjugates of candidate linkers using an anti-CD30 antibody. The conjugates were evaluated for activity using two in vitro assay formats (Fig. 3A). In the first, NAD levels were measured providing a direct evaluation of the pharmacodynamic effect of the conjugates. In addition, ADCs were tested in a standard viability assay that measures ATP content as a surrogate for viability. In the L540cy cell line, a model of CD30-positive Hodgkin lymphoma, all three ADCs displayed effective depletion of NAD and ATP. Immunologic specificity of the conjugates was established with HepG2, a hepatocellular carcinoma cell line that is sensitive to NAMPT inhibition but lacks CD30 expression thereby rendering it insensitive to the ADCs in both assay formats.

Because literature precedents suggest that sustained NAD depletion is required for progression to cell death (8, 38, 39), we investigated the temporal aspects of NAD and ATP depletion following ADC treatment. L540cy cells were treated with αCD30-8 and assayed for NAD and ATP levels every 24 hours over five days (Fig. 3B; Supplementary Fig. S2). Two trends in the data were noteworthy. First, doses above the 96-hour EC₅₀ led to substantial decline in NAD (∼10% of untreated) within 24 hours, while ATP levels did not fall below 50% until 72 hours. This delay between NAD and ATP depletion likely reflects a reserve of ATP equivalents that can sustain energy metabolism through transient NAD depletion. Second, we observed that doses of ADC just below the 96-hour EC₅₀ led to initial depletion of NAD at 24 hours, but that levels are restored at later time points and no downstream effect on ATP levels results. This behavior results in a steepening of dose–response curves over time (Supplementary Fig. S2), and suggests that NAD levels must fall below a certain threshold to disrupt ATP synthesis and elicit cell death. While these kinetic and threshold behaviors have been reported for free NAMPT inhibitors, we confirm these apply to ADC delivery and suggest they may be relevant to appropriate application of the platform.

To investigate the pharmacodynamic effect of NAMPTi-ADCs in vivo, we measured intratumoral NAD levels in mice bearing L540cy Hodgkin lymphoma xenografts (Fig. 4A). For our initial in vivo studies, we employed linker 9, which includes a hydrophilic PEG with the potential to maximize pharmacokinetic half-life (33). SCID mice bearing subcutaneous xenografts were treated with αCD30-9 at doses of 3 mg/kg and 10 mg/kg. At 4 days postdose, tumors were excised, homogenized, and assayed for NAD content using an adaptation of the in vitro NAD assay. In untreated tumors, NAD was detected at a concentration of 35–110 μg/g tissue. In tumors treated at either dose, NAD was detected at or near the assay background, thus demonstrating that the ADCs maintain their pharmacodynamic effect in vivo.

Tumor-regressing activity of αCD30-9 was further evaluated in the L540cy xenograft model at doses of 1, 3, and 10 mg/kg. In congruence with the pharmacodynamic data, single doses of 3 and 10 mg/kg led to complete tumor regressions in all animals, with only a single animal dosed at 3 mg/kg experiencing tumor outgrowth after 65 days (Fig. 4B). Mice treated at 1 mg/kg showed tumor growth delay but no complete regressions. Also evaluated in this efficacy experiment was αCD30-8, where drug linker 8 differs from 9 by absence of the hydrophilic PEG moiety; both linkers lead to the intracellular release of inhibitor 2. Because both ADCs showed equivalent activity at 3 mg/kg, we prioritized linker 8 over linker 9 for further study due to its simplified chemical structure and favorable pharmacokinetic properties.

The relative efficacy of the three prioritized payloads containing different head group structures was also compared in the L540cy xenograft model. Anti-CD30 ADCs bearing drug linkers 6, 7, and 8 were administered at the subcurative dose of 1 mg/kg. Despite comparable in vitro activities, αCD30-8 provided the greatest tumor growth delay (Fig. 4C).

To understand the breadth of activity of NAMPTi-ADCs, additional xenograft models for Hodgkin lymphoma, non-Hodgkin lymphoma, and acute myeloid leukemia were evaluated using ADCs targeting CD30, CD19, and CD123, respectively (Fig. 5). In vitro NAD and ATP depletion assays were performed in the selected cell lines with the ADCs prior to xenograft studies (Supplementary Table S3). Interestingly, while all the ADCs demonstrated good activity in the NAD assay, none were able to achieve an EC₅₀ value in the ATP assay.

In the CD30-positive L428 xenograft model, a single dose of αCD30-8 at 10 mg/kg resulted in complete tumor regressions in 4 of 5 treated mice (Fig. 5A). For NHL, the Ramos (Burkitt lymphoma) and DOHH2 (follicular lymphoma) models were selected for evaluation in vivo using αCD19-8. In Ramos, single doses of αCD19-8 at 10 mg/kg gave complete regressions in all animals treated (Fig. 5B). In DOHH2, complete regressions were observed in 3 of 5 mice treated with a single dose of 10 mg/kg (Fig. 5C). Because of the shorter duration of these responses, we extended dosing in this model to 3 weekly doses (q7dx3) of 10 mg/kg ADC and observed complete regressions in all animals with greatly extended durability of response. For AML, the HNT-34 xenograft was evaluated with αCD123*-8 (DAR = 10) using the 10 mg/kg q7dx3 dosing regimen (Fig. 5D). The isotype control ADC IgG*-8 (DAR = 10) showed a transient interruption of tumor growth following the third dose, potentially driven by the ability of AML cells to uptake ADCs via Fc-mediated interactions (23). In contrast, αCD123*-8 induced rapid tumor regression after the initial dose and sustained tumor regression after subsequent doses. Taken together, these various models indicate that the NAMPTi payload can be efficacious in multiple tumor indications when delivered with antibodies that bind various tumor antigens.

In single-dose rat toxicity studies, administration of the IgG-8 ADC was well tolerated at the tested doses of 30, 60, and 100 mg/kg. Toxicities associated with the ADC included acute decreases in circulating lymphocytes, monocytes, eosinophils, red blood cells (RBC) and reticulocytes on study days 4 and 8, with no reduction in platelets (Fig. 6A). There were no significant changes in serum chemistry parameters. The timing of the RBC and reticulocyte nadir (study day 4 and 8, respectively) and histologic appearance of increased pigment (presumed hemosiderin) in the spleen is suggestive of an extravascular hemolytic component (Fig. 6B). Investigations are ongoing to determine the mechanism of this toxicity as well as its translatability between species. Spleenic red pulp hyperplasia and extramedullary hematopoiesis seen at one week post dose are likely compensatory in response to the anemia. Both the test article and physiologic stress likely contributed to the lymphoid depletion present in the lymph nodes, spleen, and thymus.

All findings from hematology and histopathology recovered by study day 29 (4 weeks postdose). In contrast to reports for NAMPT inhibitor small molecules, neither myocardial nor retinal toxicities were evident in the rat toxicity studies (12, 13).

While NAMPT inhibitors are potent cytotoxic molecules, their utility as chemotherapeutic agents in the treatment of cancer has been hampered by toxicity. In this work, we introduce NAMPT inhibitors as a viable drug class for targeted delivery to cancer cells using ADCs.

Structure-guided modification of the prototypical inhibitor FK-866 allowed the identification of new inhibitors, which retained cytotoxic activity in cells and possessed a chemical handle on which to attach an enzyme-cleavable drug linker. Candidate payloads were prioritized as free drugs based on a biochemical binding assay against NAMPT followed by a viability assay using a small panel of cell lines. Three select inhibitors were advanced to drug linker synthesis using a glucuronide-based linker system to maximize the hydrophilic character of the molecules. Ultimately, the combination of payload and linker chemistry delivered ADCs with excellent biophysical properties. Negligible aggregation was observed with ADCs conjugated with 8 or 10 drug equivalents, and pharmacokinetic analysis in rats confirmed favorable exposure and clearance parameters consistent with advanced ADCs.

From an activity standpoint, the resulting ADCs showed antigen-specific NAD-depleting activity in vitro. Secondary ATP-depleting activity is often, but not always, observed in vitro (Fig. 2; Supplementary Table S1). The weak or absent ATP depletion in some cell lines after 96 hours of treatment may be due to the kinetics of ADC internalization and drug release. This hypothesis is supported by the observation that application of free NAMPT inhibitors to cells readily leads to complete ATP depletion within a 96-hour experiment. In addition, we postulate that unknown metabolic parameters may influence the potential of a cell line to demonstrate ATP depletion in vitro. Ongoing studies in our laboratories are aimed at better understanding the determinants of drug sensitivity in different cell lines.

We further observed that insufficient depletion of NAD at early time points can lead to recovery of NAD levels at later time points (Fig. 3B). The mechanism of this rebound effect is unclear, but the findings suggest that rapid and sustained exposure to drug may be necessary for optimal activity. ADC delivery might thus provide an optimal format for NAMPT inhibition, as the long circulating half-lives of ADCs in vivo may provide continual exposure of drug to antigen-positive target cells. At the same time, antigens with high surface expression and/or efficient internalization may provide optimal targets for NAMPTi-ADCs by providing faster intracellular drug exposure and metabolite depletion at the outset of treatment.

The in vivo antitumor activity of NAMPTi-ADCs is demonstrated in multiple xenograft models of Hodgkin lymphoma, non-Hodgkin lymphoma, and acute myeloid leukemia. Tumor regressing activity is observed at single doses as low as 3 mg/kg in the L540cy model of Hodgkin lymphoma through CD30-mediated targeting (Fig. 4). In other models of Hodgkin lymphoma, non-Hodgkin lymphoma, and acute myeloid leukemia, tumor regression is commonly observed with a single ADC dose of 10 mg/kg, while a weekly dosing regimen can greatly extend the durability of regressions (Fig. 5). In vivo activity of the ADCs roughly parallels the in vitro NAD-depleting activity of the ADCs. Weak or absent activity in the ATP depletion assay, a common surrogate for cell viability, does not preclude success in the corresponding xenograft model. This divergence between ATP depletion in vitro and xenograft activity may be explained by the temporal aspects of metabolite depletion described for the in vitro assays, or alternatively by changes in sensitivity to metabolic disruption as cells experience an altered nutrient environment in a tumor as compared with cell culture media. From a practical standpoint, the better predictive value of the NAD depletion assay supports its routine application as a screening tool in a preclinical development program for NAMPT inhibitor payloads, as exclusive reliance on more traditional ATP-based viability assays may lead to exclusion of active ADCs.

In addition to establishing in vivo activity for this class of ADC payloads, a meaningful advance toward a therapeutic can only come with an improvement in the toxicity profile relative to systemic administration of NAMPT inhibitors. Our toxicology data in rodent models indicate that targeted delivery has the potential to improve the therapeutic index. Whereas the tumor-regressing activity of targeted ADCs is typically observed at single doses of 10 mg/kg in our models, single doses of a nontargeted ADC at 100 mg/kg are well tolerated in rats. While mild, dose-dependent cytopenias are observed in acute-phase hematologic and histopathologic evaluations, all findings are recoverable 4 weeks posttreatment (Fig. 6).

Absent from our toxicologic findings are retinal and cardiac degeneration. These toxicities are prominent in rodents and are considered to be class-specific and relevant to the use of NAMPT inhibitors in people. While published data on FK-866 rodent toxicities is limited, comparison of our ADC platform to the small-molecule NAMPT inhibitor GNE-617 is instructive. GNE-617 achieves tumor regression in a U251 glioblastoma xenograft model when dosed at 25 mg/kg orally qdx7, and in additional models when dosed at 10 mg/kg twice daily for 7 days (40, 41). In toxicology studies using SD rats, GNE-617 dosed at 30 mg/kg orally once daily resulted in early animal deaths or euthanasia on study day 4 resulting from myocardial degeneration (12). Retinal degeneration was also observed in this time frame (13). While lower doses of GNE-617 were not reported, these data suggest that efficacious dosing regimens in mouse xenograft models are not well tolerated in the rat model for toxicity. In contrast, the ADC delivery approach shown here achieved tumor regressions at single doses of 10 mg/kg, which is less than 1/10th the MTD in rats (>100 mg/kg). This is in stark contrast to results obtained with GNE-617. Absent from the current characterization of toxicity is thrombocytopenia, a potential dose-limiting toxicity in humans that does not recapitulate in rodent models (9, 42). Understanding of the impact of NAMPTi-ADCs on human platelet biogenesis and survival is a continuing aim of ongoing investigations.

In summary, we have introduced NAMPT inhibitors as novel payloads for ADC technology. Targeted delivery of this drug class provides tumor-regressing activity in multiple xenograft models and a promising therapeutic index in preclinical species. The NAMPTi mechanism of action is distinct from existing ADC payload classes and FDA-approved drugs, starving hyperproliferative cells of an essential metabolic intermediate and inducing a proinflammatory cell death pathway. Continued evaluation of NAMPTi-ADCs in a range of indications is warranted to leverage these unique features in the continued pursuit of therapeutics for clinical application.

No potential conflicts of interest were disclosed.

Conception and design: C.S. Neumann, K.C. Olivas, P.D. Senter

Development of methodology: L.V. Loftus, P.G. Pittman

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): C.S. Neumann, K.C. Olivas, M.E. Anderson, J.H. Cochran, S. Jin, L.V. Loftus, D.W. Meyer, J.C. Nix, P.G. Pittman, M.L. Ulrich, A.B. Waight, A. Wong, M.C. Zaval, W. Zeng

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): C.S. Neumann, K.C. Olivas, J.H. Cochran, F. Li, L.V. Loftus, J. Neale, J.C. Nix, J.K. Simmons, A.B. Waight, M.C. Zaval

Writing, review, and/or revision of the manuscript: C.S. Neumann, K.C. Olivas, J.H. Cochran, L.V. Loftus, J.K. Simmons, P.D. Senter

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): D.W. Meyer, P.G. Pittman

Study supervision: C.S. Neumann, M.E. Anderson, F. Li, J.K. Simmons, R.P. Lyon

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.