NN-01-195, a novel conjugate of HSP90 and AURKA inhibitors effectively targets solid tumors

Abstract

Aurora kinase A (AURKA) regulates cell cycle progression into and through mitosis. As overexpression of AURKA in cancer cells is common and associated with mitotic defects and aneuploidy, small molecule inhibitors of AURKA have been developed as candidate therapies for cancer. However, these have typically low activity in clinical trials, with systemic toxicities limiting dose escalation. To concentrate an AURKA inhibitor in tumors, we exploited the fact that cancer cells in solid tumors selectively express high levels of the chaperone HSP90 to counteract intratumoral stresses, providing a potential targeting moiety. We developed NN-01-195 as a novel chimeric small molecule that combines an AURKA inhibitor related to TAS-119/VIC-1911 with an HSP90-binding moiety related to SNX2112, and evaluated its function. NN-01-195 tightly binds and inhibits both AURKA and HSP90 in biochemical assays. In cancer cells, NN-01-195 causes mitotic arrest and spindle abnormalities, and a profile of signaling changes that closely resembles that of an AURKA inhibitor. ADME assessment indicates moderate metabolism in liver microsomes (T1/2 = 46.7 minutes) and sustained plasma exposure following single I.P. injection. Maximum tolerated repeated dose testing over 5 days indicates no weight loss or toxicity at 80 mg/kg. Importantly, NN-01-195 accumulates in xenografted tumors at higher levels and for longer duration than does an AURKA inhibitor. Further, in combination with an inhibitor of the G2/M checkpoint protein WEE1, NN-01-195 is more potent than VIC-1911 in limiting growth of xenograft tumors. These data support the exploration of NN-01-195 and improved analogs as promising new candidates for therapeutic evaluation.

Introduction

Protein-targeted therapies are essential tools for precision medicine, and have yielded enormous benefits for cancer patients. A critical component of targeted therapies is an enlarged therapeutic window over cytotoxic agents. Many cytotoxic agents are dose-limited because of significant side-effects in non-cancerous tissue; in contrast, therapies targeting proteins or protein combinations that are selectively expressed and required in cancer cells can improve efficacy while limiting non-specific toxicity (1). A complementary strategy to enhance specificity of therapeutic action is to develop strategies to concentrate drugs in tumors, limiting the exposure of non-transformed cells to toxic action (2,3). Such tumor-directed delivery also leverages proteins that are selectively expressed on tumors for targeting, and is most effective when the cancer cells require the target, providing selection for its continued expression (4).

Antibodies or antibody derivatives (e.g bi-specific antibodies (5), nanobodies (6)) targeting oncogenes expressed on the surface of cancer cells have been heavily exploited in precision medicine, with agents such as the EGFR-targeting agent cetuximab or the HER2/ERBB2 targeting agent trastuzumab greatly improving survival for numerous types of cancer (7–9). In the form of antibody-drug conjugates (ADCs), antibodies have also been used for tumor-targeting of cytotoxic chemicals or radioactive compounds (10). Although invaluable in the clinic, biological agents such as antibodies have a number of disadvantages compared to small molecules, including greater cost of production, a more complicated delivery process, and issues with stability, tissue distribution, and pharmacokinetics.

Small molecules have been heavily exploited to directly target numerous cellular proteins, with enormous success for some malignancies (11,12). However, with the exception of folate-binding agents (13), there have been few attempts to use small molecules to target conjugated therapeutic agents to tumors. One intriguing approach has been the proposal of HSP90-inhibitor drug conjugates (HDCs) (14,15). The HSP90 chaperone protein is highly expressed in tumors, due to its roles in resolving the cellular stresses associated with rapid growth in a frequently nutrient- and oxygen-limited environment (16). The one reported HDC, STA-8666, conjugated a derivative of the HSP90 inhibitor ganetespib to SN38, the active metabolite of the cytotoxic agent irinotecan (14). A series of preclinical studies demonstrated substantial effectiveness of this compound in targeting tumors in vivo, with in some cases exceptional activity in controlling growth in xenograft and patient-derived xenograft (PDX) models for aggressive tumors (15,17). Now designated PEN-866, this compound has advanced to a phase 2 clinical trial (NCT04890093).

In this study, we have expanded an HDC approach, with the goal of targeting Aurora kinase A (AURKA). Aurora-A kinase (AURKA) was first characterized as a regulator of mitotic entry and progression that is highly active in the G2 and M phases of cell cycle (18–21), promoting centrosome maturation, coordinating mitotic spindle formation and function, and ensuring proper chromosome alignment (22). AURKA expression is highly elevated in many solid tumors, including head and neck squamous cell carcinomas (HNSCCs (20,23)), lung cancers (24), and others, promoting aneuploidy based on induction of mitotic abnormalities and failure to resolve cytokinesis (21). In studies over the past two decades, we (23,25–32) and others have expanded understanding of the AURKA function in cancer, recognizing that this protein promotes tumorigenesis through multiple neomorphic activities which extend its tumor-promoting activity beyond modulation of cell cycle. Among others, these activities include activation of SOX2, NF-kB, and RALA; stabilization of MYC, and degradation of the tumor suppressor p53 (reviewed in (20)). Together, these activities promote cancer cell growth potential, and provide resistance to a number of cancer therapies.

AURKA inhibitors such as alisertib and VIC-1911 have been in clinical development and testing for two decades, and have shown clinical promise in some settings, particularly in combination. AURKA inhibition has been used clinically in combination with cytotoxic agents and some targeted inhibitors (33–35). In preclinical studies, AURKA inhibition has also shown valuable activity in combination with targeted inhibitors, such as those targeting EGFR in EGFR-mutated lung cancers (36) or the G2/M checkpoint inhibitor, WEE1 (26). However, because AURKA inhibition in normal cells leads to cell cycle defects, side effects are associated with the use of AURKA inhibitors, particularly in target cells with a high proliferation index; these include neutropenia, leukopenia, anemia, and thrombocytopenia, and to a lesser degree stomatitis and fatigue (37). A strategy to selectively target AURKA inhibitors to tumor tissue would be predicted to greatly increase the clinical utility of these drugs.

In this work, we describe the design, synthesis, and validation of NN-01-195, a novel HDC that uses an optimized HSP90-binding moiety to target an AURKA inhibitor to tumors. We show that this compound maintains binding activity for HSP90 and AURKA, and AURKA inhibitory activity, and selectively concentrates in xenograft tumors in vivo. Importantly, in combination with a WEE1 inhibitor, NN-01-195 was well tolerated, and more effective than VIC-1911 in controlling xenograft growth.

Materials and Methods

HTRF competitive binding assays:

HTRF technology was used to determine inhibition of HSP90α and GST-AURKA to VY-11–109 and NN-01–217, respectively. Test compounds were dispensed using the Echo 650 acoustic liquid handler (Beckman Coulter). For the HSP90 binding assay, 5 nM of HIS-Hsp90α (Prospec Bio Cat#: HSP-090) was pre-incubated with anti-HIS-Tb conjugated HTRF donor antibody (Revvity Cat#: 61HI2TLB) in assay buffer (10 mM HEPES, pH 7.4, 100 mM NaCl, 5 mM MgCl2, 5 mM DTT, 0.005% Tween20). Following incubation, assays were initiated by adding 50 nM VY-11–109. For AURKA binding, 1 nM GST-AAK (SignalChem Cat#: A28–18G) was pre-incubated with anti-GST-Tb conjugated HTRF donor antibody (Revvity Cat#: 61GSTTLB) in assay buffer. Following incubation, assays were initiated by adding 1nM NN-01–217. HTRF signal was measured using the ClarioStar Plus plate reader (RRID:SCR_026330) at 340 excitation and 520, 620 nm emission. HTRF ratios were normalized to % inhibition where 100% = HTRF ratio in the absence of His-Hsp90, and 0% = HTRF ratio in the absence of test compound. Normalized data were fit by nonlinear regression to 4 parameter dose response equation using XlFit5 (IDBS).

AlphaScreen Ternary Complex assay:

Various concentrations of test compounds were dispensed in 50 nL aliquots using the Echo 650 acoustic liquid handling system into white, low volume 384-well plates (Greiner Cat# 784075). Equal concentrations of GST-Aurora A kinase (SignalChem) and HIS-Hsp90 (full length, Prospec Bio or HIS-SUMO-AVI-Hsp90b (1–240) or Biotin-Hsp90b (1–240) were mixed in assay buffer (10 mM HEPES, pH 7.4, 100 mM Nacl, 5 mM MgCl2, 0.005% Tween 20, 5 mM DTT (0.1 mg/mL BSA also included) containing 10 or 20 ug/mL of both donor and acceptor AlphaScreen beads (Revvity Cat# 6762003 and Cat# AS114D). Five microliters were added to each well. After 2 hr, AlphaScreen signal was measured on the BMG ClarioStar plate reader(RRID:SCR_026330). Multiple approaches were used to increase confidence. In approaches using a GST donor, 8 nM GST-AAK, 8 nM HIS-Hsp90a (full length, ProspecBio), and a 10 ug/mL glutathione donor were used with nickel chelate acceptor AlphaScreen beads. Alternatively, 25 nM GST-AAK, 25 nM HIS-SUMO-AVI-Hsp90B (1–240), and a 10 ug/mL glutathione donor, were used with nickel chelate acceptor AlphaScreen beads. As a third approach, assays were run using a streptavidin donor, 25 nM GST-AAK, 25 nM Biotin-Hsp90B (1–240), and a 10 ug/mL streptavidin donor were used with anti-GST acceptor AlphaScreen beads.

Nanoluciferase Bioluminescence Resonance Energy Transfer (nanoBret) Assay.

For this, we expressed an HSP90-nanoluciferase protein in the HEK293T (human embyonic kidney)(RRID:CVCL_0063) cell line. We then synthesized an HSP90 cellular tracer, ARK-02–280, representing NN-01–148 conjugated to the cell permeable dye, TAMRA. Treatment of HEK293T-HSP90nanoluc model with ARK-02–280 plus nanoluciferase substrate established a brigh bioluminescence signal is emitted. Competition assays were then performed with unlabeled NN-01–148 and NN-01-195 over a concentration range to confirm that these compounds can enter the cell and efficiently compete with the tracer for binding to HSP90.

Cell culture.

The FaDu (hypopharyngeal squamous cell carcinoma) (RRID:CVCL_1218), Cal27 (tongue squamous cell carcinoma) (RRID:CVCL_1107), Detroit562 (tongue squamous cell carcinoma) (RRID:CVCL_1171), NCI-H1975 (non-small cell lung carcinoma, non-smoking patient) (RRID:CVCL_1171), PC-9 (non-small cell lung carcinoma) (RRID:CVCL_B260), WM3451 (skin-derived human melanoma) (RRID:CVCL_C273), WM3000 (metastatic melanoma) (RRID:CVCL_AP86), and MRC5 (fetal lung fibroblast) (RRID:CVCL_C1XY) cell lines were purchased from ATCC or Fox Chase Cancer Center Cell Culture Facility. FaDu, Cal27, and Detroit562 were acquired from ATCC in 2023. H1975 and PC-9 were acquired from ATCC in 2024. WM3451, WM3000, and MRC5 were acquired from Fox Chase Cancer Center Cell Culture Facility in 2023. Identity of each cell line was confirmed by STR profiling every 6 months. Cell lines were also routinely screened every 6 months for mycoplasma and all pathogens included on the IDEXX Impact 1 List and IDEXX h-Impact List. All screening was performed by IDEXX BioAnalytics. FaDu and Detroit562 cells were cultured in EMEM (Corning Life Sciences, Cat# 10–009-CV). Cal27 cells cultured in DMEM (Corning Life Sciences, Cat# 50–013-PB) supplemented with 1x GlutaMAX (Thermo Fisher Scientific, Cat# 35050061). NCI-H1975 cells cultured in RPMI-1640 (Cellgro, Cat# MT50–020-PB), supplemented with 2mM Glutamine (Thermo Fisher Scientific, Cat# 35050061) All media were supplemented with 10% Fetal Bovine Serum (GeminiBio, Cat# S12450) and 0.1mg/mL penicillin-streptomycin. Cell lines were cultured at 37°C and 5% CO2. Cell lines were typically used within 10 passages of thawing.

Cell growth, cell survival, and cell cycle assays.

For cell viability assays cells were treated with vehicle (0.1% DMSO) or compounds for 72 hours, then 5 μl of CellTiterBlue (ProMega, G8081) added to media for 2 hours at 37C, and absorbance at 574nm determined using a Perkin Elmer EnVision 2102 plate reader. For longer term survival analysis, cells were seeded into 12 well plates and treated with vehicle, VIC-1911, NN-01–148, or NN-01-195 for 6 days. Cells were fixed with 10% methanol and 10% acetic acid in dH2O, followed by staining with 0.2% crystal violet. Total area was calculated using FIJI (Version 1.54f RRID:SCR_002285). For cell cycle analysis, cells were treated with vehicle or compounds for 24 hours, then trypsinized and fixed in 70% ethanol. Fixed cells were stained with propidium iodide/RNAse staining buffer (BD Biosciences, Cat# 550825). Samples were processed using a BD BioSciences FACSymphony A5 (RRID:SCR_022538), and data analyzed using FlowJo V10.8.1(RRID:SCR_008520).

Western Blotting and Immunoprecipitation.

Cultured cells were lysed on ice with CellLytic M (Sigma, Cat#C2978) and 1X Halt Protease and Phosphatase inhibitor cocktail (Thermo Fisher Scientific, Cat# 78429). Lysates were centrifuged at 13,300xG for 30 minutes at 4°C. Protein concentration in supernatants was measured by using the Pierce BCA Protein Assay Kit (Thermo Fisher Scientific, Cat# 23223). 25 μg of total protein was prepared with the addition of NuPAGE LDS Sample Buffer (ThermoFisher Scientific, Cat# NP0007) and loaded into lanes of Bolt Bis-Tris 4–12% Protein gels (Invitrogen, NW04120BOX). For immunoprecipitation, protein lysates prepared as previously described. Protein lysate was incubated overnight with Protein A/G PLUS-Agarose beads (Santa Cruz Biotechnology, Cat# sc-2003, RRID:AB_10201400) and total AURKA antibody (1:100 dilution, Cell Signaling, Cat# 3092, RRID:AB_2061342). Input and immunoprecipitation samples prepared and loaded into Bolt Bis-Tris 4–12% Protein gels (Thermo Fisher Scientific, Cat# NW04122BOX).

Xenograft tissue was lysed with T-PER protein extraction reagent (Thermo Fisher Scientific Inc., Cat# 78510), containing 5X Halt Protease and Phosphatase inhibitor cocktail (Thermo Fisher Scientific, Cat# 78429). Following lysis, sample preparation continues, as previously described. 55 μg of protein was loaded in each well.

All antibodies used were diluted in SuperBlock (Thermo Fisher Scientific, 37515). Primary antibodies used for Western blotting include phospho-Aurora A (Thr288)/Aurora B (Thr232)/Aurora C (Thr198) (1:1000 dilution, Cell Signaling, Cat#2914), AURKA (1:1000 dilution, Cell Signaling, Cat# 3092, RRID:AB_2061342), Actin (1:10,000 dilution, Sigma, Cat# A5316, RRID:AB_476743), HSP60 (1:1000 dilution, Cell Signaling Technology Cat# 4870, RRID:AB_2295614), HSP70 (1:1000 dilution, Cell Signaling Technology Cat# 4872, RRID:AB_2279841), p-PLK1(1:1000 dilution, Cell Signaling Technology Cat# 5472, RRID:AB_10698594), PLK1 (1:1000 dilution, Cell Signaling Technology Cat# 4513, RRID:AB_2167409), phosphorylated-S235/236-S6 ribosomal protein (1:1000 dilution, Cell Signaling Technology Cat# 4858, RRID:AB_916156), S6 ribosomal protein (1:1000 dilution, Cell Signaling Technology Cat# 2217, RRID:AB_331355), p-ERK (1:1000 dilution, Cell Signaling Technology Cat# 9101, RRID:AB_331646), ERK (1:1000 dilution, Cell Signaling Technology Cat# 4695, RRID:AB_390779), p-AKT (1:1000 dilution, Cell Signaling Technology Cat# 9271, RRID:AB_329825), AKT (1:1000 dilution, Cell Signaling Technology Cat# 4691, RRID:AB_915783), TPX2 (1:1000 dilution, Cell Signaling Technology Cat# 8559, RRID:AB_10827636). Secondary antibodies used include anti-mouse IgG HRP (1:10,000 dilution, Cytiva Cat# NA931, RRID:AB_772210), anti-rabbit IgG HRP (1:10,000 dilution, Cytiva Cat# NA934, RRID:AB_772206), anti-mouse IgG AP (1:5000 dilution, Jackson ImmunoResearch Labs Cat# 715–055-151, RRID:AB_2340778), and anti-rabbit IgG AP (1:5000 dilution, Jackson ImmunoResearch Labs Cat# 711–055-152, RRID:AB_2340591). For Western blots using HRP-conjugated secondary antibody, the Immobilon Western Chemiluminescent HRP Substrate (MilliporeSigma, WBKLS0500) was used. For AP-conjugated antibodies, an Immun-Star AP Conjugate Substrate Kit (BioRad, Cat# 1705018) was used. Darkroom film (MidSci, Cat# BX57, Fenton MO) was used to image Western blots. Images were quantified using FIJI (Version 1.54f, RRID:SCR_002285). Band intensity normalized to each respective loading control.

Tandem Affinity Purification for identification of Drug Binding Partners (TAP-DBP) for HSP90.

nTurboID-FKBP12(F36V) expressing 293T cells (RRID:CVCL_0063) were seeded at a density of 5 million cells. Cells were treated with vehicle or 1 μM ARK-02–081 and 50 mM biotin for 6 hours. Cultured cells were lysed with Pierce IP Lysis buffer (Thermo Fisher Scientific, Cat # 87787) and 1x HALT Protease and Phosphatase inhibitor cocktail, as previously described. Protein concentration was determined using Pierce BCA protein Assay Kit, as previously described. Equal amounts of protein were added to 30 μL of pre-washed anti-FLAG magnetic beads (Selleckchem, Cat# B26101) and 1 μM of ARK-02–081. Lysate and beads incubated overnight at 4°C.

Supernatant was withdrawn from beads by magnetic separation. After separation, protein was eluted from beads by addition of 1% SDS. Eluted protein was purified with Zeba Spin Desalting Columns 7K MWCO 0.5 mL (Thermo Fisher Scientific Cat# 89882) following manufacturer’s instruction. The purified protein was added to pre-washed Streptavidin magnetic beads (GenScript Cat# L00936) and rotated at room temperature for 2 hours. Supernatant was withdrawn from beads by magnetic separation. Beads washed with TBST (TBS+0.1% Tween 20) and protein lysate was prepared for Western blot with the addition of 1x Laemmli sample buffer (Bio-Rad Cat# 1610747) containing 50 mM DTT. Supernatant withdrawn from beads by magnetic and 20 μL loaded for Western Blot.

Reverse phase protein array (RPPA).

Cells were lysed using lysis buffer (1% Triton X‐100, 50mM HEPES pH 7.4, 150mM NaCl, 1.5mM MgCl2, 1mM EGTA, 100mM NaF, 10mM Na pyrophosphate, 1mM Na3VO4, 10% glycerol) and 1X Halt Protease inhibitor cocktail (Thermo Fisher Scientific, Cat# 78446). Lysate centrifuged at 13,300 xG for 10 minutes at 4°C. Supernatant was collected, and protein concentration was measured using a Pierce BCA Protein Assay Kit (Thermo Fisher Scientific, Cat# 23223). Samples were adjusted to a protein concentration of 1.5 μg/mL with additional lysis buffer. 4x SDS loading buffer (40% Glycerol, 8% SDS, 10% β-mercaptoethanol, 0.25M Tris‐HCL, pH 6.8) was added to each lysate and stored at -80C. RPPA analysis was performed at the M.D. Anderson RPPA facility (RRID:SCR_016649) using established protocols (38–40). Data normalization and quantitation performed at the M.D. Anderson RPPA facility (41). Data plotted using Graphpad Prism 8 (RRID:SCR_002798).

Immunofluorescence (IF).

Cells were cultured on Poly-L-Lysine coverslips (Neuvitro, Cat# GG-22–1.5-PLL). Coverslips were fixed in 4% paraformaldehyde, followed by permeabilization with 0.25% Triton. Coverslips blocked in 10% goat serum and 1% Bovine Serum Albumin (BSA) in PBS. Coverslips were incubated in primary antibody overnight at 4°C, as follows: β-tubulin (1:100 dilution, Cell Signaling Technology Cat# 2128, RRID:AB_823664), p-^S10Histone H3 (1:100 dilution, Cell Signaling Technology Cat# 9706, RRID:AB_331748). Following primary antibody incubation, coverslips incubated with DAPI (Thermo Fisher Scientific, Cat# 62248) and secondary antibody at room temperature for 1 hour, as follows: anti-Rabbit IgG AlexaFluor 488 (1:100 dilution, Thermo Fisher Scientific Cat# A-11034) and anti-Mouse IgG AlexaFluor 568 (1:100 dilution, Thermo Fisher Scientific Cat# A-11031, RRID:AB_144696). Coverslips were mounted using VectaShield Antifade Mounting Medium (Vector Laboratories, Cat# H-1000) and clear nail polish. Images processed using FIJI (Version 1.54f, RRID:SCR_002285).

Maximum Tolerated Dose (MTD), Tumor Accumulation, and Efficacy.

All studies performed in mice were approved by the institutional IACUC review board at Fox Chase Cancer Center. For all in vivo studies, VIC-1911 and NN-01-195 were resuspended in a vehicle consisting of 10% DMSO, 20% Solutol HS-15, and 70% PBS. For all animal experiments, approximately equal numbers of male and female mice were randomly assigned to each group. Initial bodyweight of female mice was 20.0 – 26.0 grams, while initial bodyweight of male mice was between 25.0 – 31.0 grams. Humane endpoint was determined by visual signs of mouse pain and discomfort, 20% loss in bodyweight, or tumor ulceration. All animal studies approved by IACUC guidelines at Fox Chase Cancer Center.

For MTD studies, C57BL6/J mice from Jackson Laboratories (RRID:IMSR JAX:000664) were injected I.P. daily for 5 days with vehicle, or 10mg/kg, 30 mg/kg, or 80 mg/kg NN-01-195. Mice were between 6–10 weeks old. Mouse behavior and bodyweight were assessed daily. 24 hours after the final dose, mice were euthanized by CO2 inhalation, followed by cervical dislocation. Histology of liver, kidney, and spleen was analyzed by a board-certified pathologist. The pathologist was blinded to the experimental cohorts.

For accumulation analysis, 106 FaDu cells were injected subcutaneously into the flank of NSG mice bred by the Laboratory Animal Facility at Fox Chase Cancer Center (RRID:BCBC_4611). All mice were between 6–10 weeks old. Once tumors reached a volume of approximately 500 mm3, mice were injected I.P. with 10mg/kg of either VIC-1911 or NN-01-195. 3 or 24 hours after injection, mice were euthanized by CO2 inhalation, followed by cervical disloation. Tumors were excised and collected from euthanized mice and analyzed by the Wistar Proteomics and Metabolomics Core Facility (RRID:SCR_010211).

For efficacy studies, adavosertib was resuspended in 40% PEG300, 10% DMSO, 5% Tween-80, and 45% PBS. Adavosertib was administered via oral gavage. 10⁶ FaDu or 10⁶ H1975 cells were injected into NSG mice, as previously described. When FaDu xenograft tumors reached 100–150 mm³, mice were randomized into six treatment groups (n = 6–7 mice per group): (i) control (vehicle only), (ii) VIC-191, 30 mg/kg via I.P., (iii) NN-01-195, 30 mg/kg via I.P., (iv) 60 mg/kg adavosertib via oral gavage, (v) VIC-1911 30 mg/kg and adavosertib 60 mg/kg, (vi) NN-01-195 30 mg/kg and adavosertib 60 mg/kg. The drugs were administrated daily for 2 weeks. When H1975 xenograft tumors reached 100–150 mm3, mice were randomized into six treatment groups (n = 6–7 mice per group): (i) control (vehicle only), (ii) VIC-191, 13.4 mg/kg via I.P., (iii) NN-01-195, 30 mg/kg via I.P., (iv) 60 mg/kg adavosertib via oral gavage, (v) VIC-1911 13.4 mg/kg and adavosertib 60 mg/kg, (vi) NN-01-195 30 mg/kg and adavosertib 60 mg/kg. For H1975 xenografts, drugs were administered daily for 3 weeks. Following euthanasia, tumors were excised and processed for histopathology. During treatment, tumor size and body weight were monitored daily. Tumor volume was calculated using the formula: tumor volume (mm³) = (smallest diameter² × largest diameter)/2.

Immunohistochemistry (IHC).

Tissue collected from in vivo experiments were fixed in 10% buffered formalin. Fixed tissues blocked in paraffin. For immunohistochemistry, slides deparaffinized in xylene and ethanol. Slides were boiled in citrate buffer (pH 6.0, Electron Microscopy Sciences, Cat# 64142–08) for antigen retrieval. Afterwards, slides were washed in 3% hydrogen peroxide to block peroxidases, followed by blocking in 1% BSA (Thermo Fisher Scientiftic, Cat# A2153–100G) and 10% goat serum (Gibco, Cat# 16210–064) in PBS. Following blocking, slides incubated in primary antibody overnight at 4C, as follows: Ki-67 (1:100, Proteintech Cat# 27309–1-AP, RRID:AB_2756525), cleaved-caspase 3 (1:200, Cell Signaling Technology Cat# 9661, RRID:AB_2341188), or phospho(γ)-H2AX (1:200, Millipore Cat# MABE205, RRID:AB_10851746). Slides analyzed using Vectra 3 Automated Quantitative Pathology Imaging System (Akoya Biosciences, RRID:SCR_025828).

Statistical analysis.

Plots and bar graphs depict the mean and standard error mean. GraphPad Prism 8.0 (RRID:SCR_002798) was used for all statistical analyses. One-way ANOVA with Tukey’s method for multiple comparisons or Student’s two-sided unpaired t Test was used for all comparisons. * p < 0.05, ** p < 0.01, *** p <0.001, # p < 0.0001.

Data and Materials Availability.

The data generated in this study, including raw data, are available upon reasonable request to the corresponding author. With the exception of the described novel chemical compounds, all reagents described in this study are available from commercial sources. Access to the novel compounds described is available to qualified users following completion of a Material Transfer Agreement, upon request to the co-corresponding author, J. Salvino.

Results

Design, synthesis, and biochemical affinity of NN-01-195.

As basis for an HSP90i-AURKAi conjugate, we have chemically tethered derivatives of the clinically used AURKA inhibitor VIC-1911 (also known as TAS-119; (20,42,43)) with derivatives of the benzamide HSP90 inhibitor SNX-2112 (44–47). As a first step, we established binding assays to confirm that the modified analogs we were developing would have high affinity towards their respective targets. Therefore, we developed an Homogeneous Time Resolved Fluorescence (HTRF) (48) competitive binding assay. We synthesized a suitable HSP90 acceptor probe, VY-11–109, by modifying the reported amine derivative of SNX2112, compound 29 (47) by coupling to fluorescein isocyanate (FITC) (Fig. 1A-C; Supplementary Data File 1). We identified an optimized HSP90 ligand, NN-01–148, that utilizes a rigid aryl piperazine moiety to replace the cyclohexyl alcohol moiety in SNX2112, to create bi-functional HSP90i-AURKAi conjugates. NN-01–148 was predicted, based on docking the compound into the ligand binding site of HSP90α using the X-ray crystal structure (PDB: 6ltk), to provide an improved trajectory for the linker to exit the binding site and assist the attached ligand to enter the aqueous exposed region when bound to HSP90α. Evaluation of NN-01–148 in the HTRF competitive binding assay confirmed that it had equivalent or slightly better binding affinity to HSP90α compared to SNX2112 (Fig. 1D; IC50 = 1.9 nM +/− 0.047 nM for NN-01–148 versus IC50 = 7.5 nM +/− 1.0 nM).

Figure 1.

A. Synthesis of VY-11–109, an FITC-conjugated HTRF fluorescent acceptor probe for study of HSP90 binding. The HSP90-binding moiety is derived from that of SNX-2112. B. The chemical structure of NN-01–148. C. The chemical structure of SNX-2112. D. Dose response curve for binding affinity of NN-01–148 versus SNX-2112 to HIS-tagged HSP90, based on displacement of the HSP90-FITC probe. E. Synthesis of the HSP90i-AURKAi bifunctional molecule, NN-01-195. VIC-1911 was modified at the carboxylic acid functionality with an ethyl 6-aminohexanoate, then the ethyl ester was then hydrolyzed to provide NN-01–192, which was further coupled to NN-01–148 to provide NN-01-195.

We then synthesized the bi-functional HSP90i-AURKAi conjugate, NN-01-195, by coupling the carboxylic acid motif of the AURKAi ligand, VIC-1911, to the linker, ethyl 6-aminohexanoate, using amide coupling conditions to provide NN-01–190. This was followed by basic hydrolysis to provide the modified VIC-1911 derivative, NN-01–192, which contains an alkyl linker terminating in a carboxylic acid. The HSP90 ligand, NN-01–148, was then coupled to the carboxylic acid via an amide bond to provide the HSP90i-AURKAi conjugate, NN-01-195 (Fig. 1E: extended details in Supplementary Data File 1). An HTRF competitive binding assay to measure AURKA binding affinity was developed by synthesizing a VIC-1911-FITC acceptor probe, NN-01–217 (Fig. 2A). The AURKA probe, NN-01–217, bound to AURKA (K_D ~0.5 nM), and could be displaced by unlabeled VIC-1911(TAS119; IC50 = 0.59 +/− 0.029 nM) (Fig. 2B). The HSP90i-AURKAi conjugate, NN-01-195 was evaluated for binding to AURKA and to HSP90α, and shown to have high affinity to AURKA (IC50 = 3.1 +/− 0.26 nM) (Fig. 2B) and to HSP90α (IC50 = 8.7 nM +/− 0.48 nM) (Fig. 2C). Thus, NN-01-195 potently binds to both AURKA and HSP90, with almost 3-fold greater affinity for AURKA.

Figure 2. Development of an AURKA binding assay.

A. The chemical structure of AURKA inhibitor - FITC conjugated probe, NN-01–217. The AURKA-binding moiety is derived from VIC-1911. B. Dose response curves for binding affinity of VIC-1911 and NN-01-195 to GST-AURKA, based on displacement of the NN-01–217 probe. C. Dose response curves for binding affinity of NN-01-195 to HIS-tagged HSP90, based on displacement of VY-11–019. D. Schematic of drug detection using the nTurboID cassette (49) in TAP-DBP interaction detection. In this system FLAG-HA doubly tagged chimeric protein containing nano-luciferase, the TurboID biotinylating enzyme, and FKBP12(F36V) is combined with a chimeric compound, ARK-02–081, which contains a ligand for FKBP12(F36V), AP1867, fused to the HSP90 ligand NN-01–148. E. Structure of ARK-02–081. Figure created in BioRender. Nguyen, T. (2026) https://BioRender.com/5c7y5iz) F. Interaction of ARK-02–081 with HSP90 allows immunoprecipitation confirming binding to HSP90. G, H. Measurement of cell permeability of NN-01–148 (G) or NN-01-195 (H) via nanoBret assays. I. The ability of NN-01-195 to form ternary structures with AURKA and HSP90 was measured by AlphaScreen assay.

Confirmation of intracellular target engagement.

To confirm that bi-functional conjugates using the HSP90 ligand, NN-01–148, can engage and bind cellular HSP90, we utilized our Tandem Affinity Purification for identification of Drug-Binding Proteins (TAP-DBP) method (49). For this, we used a FLAG-hemagglutinin (HA)-tagged chimeric protein featuring the FKBP12(F36V) adaptor protein and the TurboID enzyme (Fig. 2D). Treatment of the nTurboID-expressing HEK293T cells with ARK-02–081, a bi-functional conjugate of HSP90i, NN-01–148 and the FKBP12(F36V) ligand, AP1867 (Fig. 2E) followed by Flag and streptavidin immunoprecipitation strongly pulled down HSP90 and confirmed that the NN-01–148 ligand effectively binds to cellular HSP90 (Fig. 2F).

To further characterize the intracellular availability of NN-01-195, we used a nanoluciferase bioluminescence resonance energy transfer (nanoBret) cellular target engagement assay, using HEK239T cells expressing HSP90-nanoluciferase. These cells were treated with ARK-02–280, a nanoBret tracer (i.e. NN-01–148 conjugated to the cell permeable dye, TAMRA) which resulted in a bright bioluminescence signal when furimazine, the nanoluciferase substrate was added. Treatment of these cells with unlabeled NN-01–148 or NN-01-195 showed a dose dependent reduction in the ARK-02–280-induced bioluminescence signal (Figs 2G, H) under live cell and permeabilized cell conditions (50) confirming that NN-01–148 and NN-01-195 can efficiently enter the cell with no apparent cell permeability limitations. We also asked whether NN-01-195 was able to engage with both AURKA and HSP90 as a ternary complex, i.e. providing evidence that the compound was present intracellularly in an intact form, rather than metabolizing into HSP90- and AURKA-binding moieties. For this we used an AlphaScreen assay to measure the formation of a stable ternary complex between AURKA and HSP90 which was induced by treatment with the bi-functional compound, NN-01-195, and also show that the AURKAi, VIC-1911 (NN-01–127) and HSP90 ligand (NN-01–148) were incapable of forming a ternary complex. These data indicated NN-01-195 existed as an intact compound in cells, capable of binding both AURKA and HSP90 (Fig 2I).

NN-01-195 elevates apoptosis and reduces cell viability.

AURKA inhibition reduces viability, causing G2/M arrest that is initially controlled by checkpoints but can lead to mitotic defects and elevation of cell death over time (21). Using HPV-negative HNSCC (FaDu, Detroit 562, and Cal27), NSCLC (H1975, PC9), and melanoma (WM3000, WM33451) (Fig 3A and Supp Fig S1A) models, we assessed the ability of NN-01-195 to reduce cell viability, benchmarked to VIC-1911 and to the HSP90 ligand, NN-01–148. The IC50s of NN-01-195 ranged from 0.27 – 2.98 μM, which were 2 – 10 - fold higher than those of VIC-1911, and 6 – 15 -fold higher than those of NN-01–148. In comparison, assessment of these compounds in the immortalized cell model MRC-5 demonstrated IC50 of 0.43 μM to NN-01–148, but relative insensitivity to VIC-1911 and NN-01-195 (28.55 μM and 9.75 μM, respectively) (Supp Fig S1A). Measurement of apoptosis at 72 hours after drug treatment (Fig 3B, Supp Fig S1B) indicated that NN-01-195 induces apoptosis, although at higher concentrations than VIC-1911 or NN-01–148, and not in the Cal27 or Detroit cell models at this time point. In cells plated to allow clonogenic outgrowth over a longer term (Figs 3C, and Supp Fig S1C), NN-01-195 reduced colony growth by 50% or more at concentrations from 0.5 – 1.0 μM in all cell models, in contrast to VIC-1911 at 75 nM, and NN-01–148 at 25 – 50 nM.

Figure 3: NN-01-195 reduces cell viability and promotes mitotic abnormalities.

A. Cell viability assays for FaDu and NCI-H1975 cell lines after exposure to vehicle (DMSO), VIC-1911, NN-01–148, or NN-01-195 for 72 hours. B. Left, western blot analysis for PARP in FaDu or H1975 cells exposed to the indicated concentrations of VIC-1911, NN-01–148, or NN-01-195 for 72 hours. Right, quantification of cleaved PARP expression. C. Relative growth of cells plated at low density and grown in the presence of vehicle, VIC-1911, NN-01–148, or NN-01-195 at indicated doses for 6 days. Data shown is average of 3 biological repeats. D. Cell cycle analysis of FaDu or H1975 cells grown in the indicated drugs for 24 hours or 48 hours, followed by staining with propidium iodide and analysis with flow cytometry. Average of n = 3. E. Immunofluorescence for FaDu and NCI-H1975 cells exposed to either vehicle, VIC-1911, NN-01–148, or NN-01-195 for 48 hours at the indicated concentrations. Cells were stained with DAPI (blue), and antibodies to b-tubulin (green) to visualize mitotic spindle defects. Representative images from n = 3 biological repeats. F. Quantification of cells with defective mitoses (defined as asymmetric, monopolar, or multipolar spindles), among all mitotic cells. All data shown represents an average of 3 biological repeats, plotted with SEM. * p < 0.05, ** p < 0.01, *** p <0.001, # p < 0.0001 for all graphs.

AURKA inhibition typically results in cell cycle arrest in the G2 or M phases of cell cycle, and defective formation of mitotic spindles leading to aneuploidy (21,51,52). Cells treated with HSP90 inhibitors also often typically accumulate in G2/M, although with variation between cell models (53,54). Measurement of cell cycle by flow cytometry following 24 or 48 hours of drug treatment indicated that all three compounds caused a reduction in the 2N (G1) and increase in the 4N (G2/M) populations of cells (Fig 3D and Supp Fig S1D). Direct quantitation of cells stained with antibodies to β-tubulin and phospho-^S10histone H3 (pHH3) (55) or DAPI to visualize mitotic figures of cells after 48 hours exposure to the compounds similarly showed elevated levels of cells with spindle defects (defined as monopolar, asymmetric, multipolar, or disorganized) induced by all compounds (Fig 3E, F, Supp Fig S2). Notably, the pHH3 staining pattern of mitotic cells differed between cells treated with VIC-1911 and NN-01-195 versus those treated with vehicle or NN-01–148 (HSP90 inhibitor), with pHH3 non-aligned with DAPI staining in cells exposed to the AURKA inhibitors. In contrast, for NN-01–148-treated cells, DAPI and pHH3 staining was coincident in mitotic cells, whereas vehicle treated cells showed both patterns (Supp Fig S2). Previous studies have noted that the degree of pHH3 staining alignment with chromatin is variable throughout mitosis (56), and also shown direct AURKA-dependent binding and regulation of pHH3 (57,58). These data suggest arrest at a similar mitotic stage in cells treated with VIC-1911 and NN-01-195, arrest at a distinct stage in those treated with NN-01–148, and no arrest in vehicle treated cells.

Previous studies have shown that combination of an AURKA inhibitor with an inhibitor of the WEE1 G2/M checkpoint kinase enhances cell killing (59). We compared the reduction of viability or longer-term clonal growth (Supp Fig S3A, B) in FaDu and H1975 cells treated with VIC-1911, NN-01-195, or the WEE1 inhibitor adavosertib, alone or in combination. These data confirmed similarly enhanced cell killing by combination of NN-01-195 with adavosertib, with this particularly striking in the FaDu model.

Signaling response to NN-01-195, in reference to AURKA and HSP90 inhibitors.

To more exactly benchmark the target inhibition profile of NN-01-195, we used a Reverse Phase Protein Array (RPPA) analysis (38–40)) to evaluate changes in expression of 152 total or phosphorylated signaling proteins 24 hours after treatment of FaDu cells with IC30 levels of each compound (Figs 4A, B, S3C, and Supp Table S1). First comparing NN-01-195 to VIC-1911 (Fig. 4A), the most strongly elevated protein was total AURKA, which is typically observed following AURKA inhibition (26), as cells upregulate total protein in a compensatory response. Also as expected for AURKA inhibition, levels of ph^S10-HH3 were elevated by both VIC-1911 and NN-01-195, and a number of other proteins were similarly affected by NN-01-195 and VIC-1911, including MEK1, TUFM, phospho-^T281/Y220ERK5, and phospho-^S102YB1. The RPPA data also suggested some differences between VIC-1911 and NN-01-195. VIC-1911, and not NN-01-195, elevated the expression of Cdc6, Cyclin-B1, NDRG1, and PLK1. VIC-1911 also reduced the expression of PUMA, while NN-01-195 did not. Overall, the RPPA profiles of NN-01-195 and NN-01–148 had fewer similarities (Fig. 4B), of which the most significant was induction of MEK1 (in common also with VIC-1911); NN-01–148 did not induce total AURKA. A number of targets of SNX2112 were previously characterized in HNSCC cell line models (although distinct from the models used in this study) (60). In the RPPA analysis presented here, NN-01–148 inhibitory activity differed somewhat from that reported for SNX2112; in specific comparison of drug activity for these targets, NN-01-195 had few similarities with NN-01–148, but paralleled VIC-1911 in activity against p53 and BMK1-ERK5 (Supp Fig S3C).

Figure 4: Signaling changes in NN-01-195-treated cells resemble those induced by AURKA inhibitor VIC-1911.

A., B. FaDu cells were exposed to vehicle, VIC-1911, NN-01-195, or NN-01–148 at the indicated concentrations for 24 hours, and RPPA analysis performed. Protein expression values for each protein were normalized to protein expression values for the vehicle treated control. Graphs overlay scatterplots for VIC-1911 and NN-01-195 (A) or NN-01–148 and NN-01-195 (B), with proteins associated with the AURKA pathway that are induced by both VIC-1911 and NN-01-195 relative to vehicle-treated cells circled and labeled in red; proteins that are induced by VIC-1911 alone are circled in blue. n = 2. Proteins associated with HSP90 that are induced by both NN-01–148 and NN-01-195 relative to vehicle-treated cells are circled and labeled. n = 2. C. Selected proteins relevant to HSP90 were validated via Western Blot analysis in FaDu and NCI-H1975 cell lines. Protein expression normalized to vehicle and quantified. Representative of n = 3. D. pAURKA (T288) protein expression was validated via immunoprecipitation, followed by Western Blot analysis. Protein expression normalized to vehicle and quantified. E. Additional proteins that are known to be inhibited by SNX2112 were validated by Western blot. Representative of n = 3. * p < 0.05, ** p < 0.01, *** p <0.001, # p < 0.0001 for all graphs.

We further probed signaling specificities of the three compounds in FaDu and H1975 cells, (Fig 4C, D). These data confirmed that NN-01–148 induced HSP70 and HSP60, and reduced levels of decreased levels of phospho-^S235/236S6 and phospho-^S240/244S6, phosphorylations that are stabilized by HSP90 (61), while the other compounds did not. In contrast, VIC-1911 and NN-01-195 induced total AURKA in compensation for loss of AURKA phosphorylation, whereas NN-01–148 did not. Together, these data further confirmed that the intracellular activity profile of NN-01-195 was more similar to VIC-1911 than to NN-01–148.

As an additional evaluation for HSP90 versus AURKA inhibitory activity, we evaluated the expression and activity of total and phosphorylated AKT and ERK, both of which depend on HSP90 and have been reported to be reduced following HSP90 inhibition in HNSCC and other cancers (60,62,63). In two HNSCC models, we find that while NN-01–148 strongly reduced levels of total and activated AKT and ERK, VIC-1911 and NN-01-195 did not (Fig 4E and Supp Fig S3D). Taken together, these data suggested that activity of NN-01-195 in cells was more similar to that of an AURKA inhibitor than an HSP90 inhibitor.

ADME/PK and tolerability of NN-01-195.

We evaluated NN-01-195 for its metabolic stability by incubation in mouse liver microsomes (MLM) (0.5 mg/mL) at 37 °C and monitored for remaining parent compound. Under these conditions we found that NN-01-195 was shown to have moderate metabolism in mouse liver microsomes, with a T_1/2 = 46.7 minutes. The HSP90 ligand NN-01–148 had a T_1/2 = 62 minutes in the same assay (see Supp Table S2). A pharmacokinetic evaluation to monitor the plasma concentration over time of NN-01-195 after a single i.p. administration (10 mg/kg) in CD1 mice (Fig 5A) yielded a biphasic plasma clearance curve showing an initial clearance after absorption followed by a reabsorption phase, providing sustained long-lasting plasma exposure measured by the area under the curve, AUC value of 24114 hr*ng/mL. The reported T_1/2 (Fig. 5B) is underestimated due to this re-absorption, since plasma concentrations at the 4-hour time point are still quite high (~5000 ng/mL).

Figure 5: NN-01-195 is well tolerated in mice.

A. Concentration of NN-01-195 in the plasma of CD-1 mice treated with a single 10mg/kg dose of NN-01-195 administered I.P., at times indicated after treatment. B. In-life parameters for NN-01-195. C. Daily change in body weight in mice treated with vehicle or NN-01-195 at the doses indicated, administered I.P. daily for 5 days. D. Representative images of livers from mice used for C, stained with hematoxylin and eosin (H&E) or with antibodies to cleaved caspase-3 and γH2AX.

As a first step to assess the safety of using NN-01-195 in vivo, we next determined the tolerability of various doses of NN-01-195 in mice. Mice (4 groups of 6 mice each) each received a 100 μl intraperitoneal injection of: Group 1: Vehicle (10% DMSO/20% Solutol/70% PBS), or Group 2: 10 mg/kg, Group 3: 30 mg/kg, or Group 4: 80 mg/kg of NN-01-195 formulated in 10% DMSO/20% Solutol/70% PBS. Administration of NN-01-195 for 5 days prior to euthanasia had no effect on mouse body weight and induced no signs of distress (Fig 5C). Histological analysis of FFPE liver, kidney, or spleen tissue did not indicate gross tissue toxicity associated with NN-01-195 (Fig 5D, Supp Fig 4, Supp Table S3). Immunohistochemical assessment of the protein expression of cleaved Caspase-3 and γH2AX (Fig 5D, Supp Fig 4) indicated there was no difference in these toxicity markers between vehicle and NN-01-195 treated mice in the liver or kidney, but some elevation of γH2AX staining in the spleen at the highest dose administered (80 mg/kg). Together, these results indicate that NN-01-195 has was well tolerated in mice in short term toxicity assays.

Tumor accumulation and efficacy of NN-01-195.

Based on the above results, we assessed NN-01-195 accumulation in xenografted tumors in vivo, benchmarked to VIC-1911. To this end, we introduced 10⁶ FaDu cells into the flank of NSG mice and allowed tumors to grow to approximately 500 mm³ in volume. Mice were then administered a single dose I.P. of vehicle, VIC-1911 (10 mg/kg), or NN-01-195 (10 mg/kg). Mice were euthanized after 3 or 24 hours and tumors were collected and analyzed for concentration of either NN-01-195, the drug-HSP90 conjugate, and VIC-1911, drug alone post-dose. The average concentration of NN-01-195 in the tumor tissue at the 3-hour timepoint was 1761 ng/g compared to VIC-1911 which had an average tumor tissue concentration of 294 ng/g, a 6-fold difference. At the 24-hour timepoint, the average concentration of NN-01-195 in the tumor tissue was 355 ng/g, whereas VIC-1911 was below the level of detection (Fig 6A).

Figure 6: NN-01-195 is highly effective at controlling xenograft tumor volume.

A. Abundance of VIC-1911 or NN-01-195 in FaDu cells xenografted onto NSG mice. After tumors reached 500mm³, mice were administered VIC-1911 or NN-01-195 in a single dose. Tumors were excised after 3 or 24 hours, and drug concentration determined via mass spectrometry analysis. B., C. FaDu cells were xenografted onto NSG mice. After tumors reached 100–150mm³, mice were administered the indicated compounds daily for 14 days. Body weight (B) and tumor volume (C) were recorded daily. No changes in body weight were significant. D. Xenograft tumor tissue from C. was stained for Ki-67, and results quantified. E. Representative immunohistochemistry images of xenografted tissue from D. Tissues were stained with Ki-67. F., G. H1975 cells were xenografted, as described in (B). Mice were administered the indicated compounds for 21 days. Tumor volume (F) and bodyweight (G) was measured daily. * p < 0.05, ** p < 0.01, *** p <0.001, # p < 0.0001 for all graphs.

We then compared the ability of NN-01-195 and VIC-1911 to reduce tumor volume in vivo. Previous studies in head and neck cancer xenografts suggested limited effect of the single agent AURKA inhibitor alisertib, but greater activity from a combination of the alisertib with adavosertib, an inhibitor of the WEE1 G2/M checkpoint (26). Using a similar strategy, we injected 10⁶ FaDu cells into NSG mice to establish xenografts. When tumors reached 100–150mm³ in volume, mice were randomized to 6 cohorts as follows: Group 1: vehicle, Group 2: VIC-1911, 30 mg/kg, Group 3: NN-01-195, 30 mg/kg, Group 4: adavosertib, 60 mg/kg, Group 5: VIC-1911, 30 mg/kg + adavosertib 60 mg/kg, or Group 6: NN-01-195 30 mg/kg + adavosertib 60 mg/kg. We note, with the MW of VIC-1911 at 505 Da, and of NN-01-195 at 1128 Da, this represented a 2.3-fold molar excess of VIC-1911 (59.39 μmol/kg for VIC-1911 versus 26.59 μmol/kg for NN-01-195). Compounds were administered daily for up to 14 days, with body weight (Fig 6B) and tumor volume measured daily (Fig 6C). The compounds assessed did not induce any significant change in body weight throughout the course of the experiment.

Administration of single agent VIC-1911 and NN-01-195 yielded statistically non-significant change to tumor volume versus vehicle treatment, and several of the treated mice in each group required euthanasia based on humane criteria before the day 14 experimental endpoint. Adavosertib administered alone or in combination with VIC-1911 did not significantly reduce tumor growth. In contrast, the combination of adavosertib with NN-01-195 resulted in significant reduction of tumor growth versus either agent alone. Histopathological assessment of tumors using Ki-67 staining confirmed more effective inhibition of cell proliferation by drug combinations (Fig 6D, E), whereas staining with antibodies to caspase or γH2AX did not indicate significant differences at experimental endpoints (Supp Fig S5). Western blot analysis of tumor tissues collected at the end of treatment demonstrated continued significant inhibition of phosphorylated AURKA and elevation of TPX2, compatible with response to an AURKA inhibitor (Supp Fig S6A). The phosphorylation of the AURKA substrate PLK1 was also inhibited by VIC-1911 and NN-01-195 treatment (Supp Fig S6B). Interestingly, the phosphorylation of the HSP90 client S6 was inhibited in NN-01-195 treated cells but not in VIC-1911-treated cells, whereas expression of HSP60 and HSP70 was not affected by either compound (Supp Fig S6C). Finally, comparison of the liver, kidneys, and spleen of mice treated for three weeks with vehicle, VIC-1911 + Adavosertib, or NN-01-195 + Adavosertib indicated no changes to γH2AX and cleaved caspase-3 staining versus vehicle-treated control, suggesting minimal toxicity (Supp Fig S7).

We also performed in vivo analysis using a second H1975 lung cancer xenograft model. In this case, we used isomolar doses of VIC1911 and NN-01-195, and due to the slower growth of these xenografts, extended analysis to 21 days. In this model, NN-01-195 performed comparably to VIC-1911, with slightly better quantitative control of tumor growth not reaching statistical significance (Fig 6F). Through this longer time course, the weight of mice treated with compounds was unaffected (Fig 6G). Finally, although good control was observed with the drug combination, immunohistochemical analysis of tumors indicated that VIC-1911 and NN-01-195 had no statistically significant impact on Ki-67 and cleaved caspase-3 staining at the end of treatment, suggesting resistance was developing by this time point (Supp Fig S8).

Discussion

The work presented here demonstrates that NN-01-195 is a promising pilot compound for a new strategy of targeting AURKA. In a defined biochemical system or in cultured cells, NN-01-195 binding activity for AURKA and HSP90 is similar to that of the related precursor compounds, VIC-1911 and SNX-2112, although reduced in affinity. NN-01-195 efficiently inhibited AURKA kinase activity in a defined in vitro system, and was maintained as a bi-specific molecule capable of binding both HSP90 and AURKA in cells and in vivo. In cultured cells, NN-01-195 potency in reducing viability or causing G2/M arrest was reduced versus VIC-1911, and evaluation of signaling in NN-01-195-treated cells indicated an activity profile more similar to VIC-1911 than SNX-2112. NN-01-195 and VIC-1911 caused similar mitotic phenotypes that were distinct from those induced by NN-01–148, including induction of total AURKA expression, with similar synergy observed with the WEE1 inhibitor adavosertib.

Importantly, mouse pharmacokinetic analysis for NN-01-195 showed a favorable profile, with appreciable plasma levels at 4 hours after a single dose by IP administration, suggesting good drug exposure in the plasma over time. Conjugation to the HSP90 ligand did not increase undesired toxicity based on a short term toxicity assessment, which showed excellent tolerability of NN-01-195 at doses up to 80 mg/kg administered for 5 days. The biodistribution study suggests that HSP90 conjugation increased tumor tissue concentration at the earlier time point and helped to retain the HSP90-drug conjugate in the tumor for a longer period of time compared to the unconjugated drug. Importantly, NN-01-195 was more effective than VIC-1911 in controlling growth of FaDu xenografts even when the latter compound was used in 2.3-fold molar excess, and comparable to VIC-1911 in controlling growth of H1975 xenografts. Analysis of signaling in recovered xenograft tumors indicated NN-01-195-treated cells effectively maintained AURKA inhibition over several weeks of in vivo treatment. Overall, these finding support the value of this approach as a platform for concentrating targeted inhibitor of AURKA within solid tumors, as first suggested for STA-8666 (14). A specific benefit of this approach would be the ability to elevate dose levels while minimizing on-target cytotoxic effects (e.g. on the hematopoietic system).

The highly AURKA-specific inhibitor VIC-1911 was selected from among a number of AURKA inhibitor options assessed for modification in part because of a relatively favorable safety and side effect profile compared to other available AURKA inhibitors (43). We note that this technology can readily be extended to other clinically used AURKA inhibitors, such as alisertib or LY3295668, if warranted. This compound differs from a prior HSP90-targeted chimera, STA-8666 in which the HSP90-binding moiety and toxic payload, SN38, were joined via a cleavable linker. We utilized a metabolically stable conjugation strategy to provide a stable bi-functional HSP90-AURKA conjugate, rather than a metabolically cleavable linker used in the pro-drug approach for STA-8666, to reduce complexity in optimizing a bi-functional compound with good pharmacokinetic properties and to facilitate comparison between conjugated versus unconjugated AURKA inhibitors. In addition, a stable conjugate can be more readily evaluated in development for safety and toxicity.

This study has focused on assessing mitotic consequences of AURKA inhibition. However, AURKA has kinase-dependent and -independent effects on numerous other cellular processes, including regulation of TP53 expression (64), centrosome maturation (65), cellular migration (66), the ciliary cycle (25,32), and calcium signaling (67). Further, in the case of AURKA overexpression in cancer, AURKA is expressed throughout the cell, and assumes additional roles, including regulation of MYC expression and function (68,69) and hypoxia signaling (70), both of which support invasion and metastasis. Like VIC-1911, NN-01-195 may affect these additional AURKA activities, influencing its potency; based on the results of the RPPA assay, phosphorylation of both cytoplasmic and nuclear substrates of AURKA were affected. Another question is the degree to which NN-01-195 retains HSP90 inhibitory activity in vivo. RPPA profiling and targeted Western blot analysis generally suggests lower activity in inhibiting HSP90 than AURKA, although these studies also indicate differences between HSP90 inhibitor activity in the cell models in this study versus those used in other reports (e.g. (60)). In vivo, the fact that NN-01-195 reduced expression of S6, but did not elevate HSP60 or HSP70 suggests a possible weak or partial inhibition of HSP90. This is an important consideration, as some studies have suggested a potential concern over the use of pan-HSP90 inhibitors due to the potential for cardiotoxicity resulting from disruption of maturation of the hERG channel and ocular toxicity (71,72).

We had chosen the benzamide SNX2112 HSP90 chemotype since this was clinically demonstrated to provide an improved safety profile in human clinical evaluation compared to pan-HSP90 inhibitors such as 17-AAG (45). In addition, TAS-116 (pimitespib), another HSP90 inhibitor in the benzamide chemotype recently approved in Japan was well tolerated and avoids ocular toxicity (73). The potential for cardiotoxicity also appears to be chemotype-specific, with the benzamide chemotype providing minimal effects on QTc prolongation (74,75). As improved analogs for NN-01-195 are developed and tested, it will be important to investigate longer-term toxic effects. While the in vitro and in vivo analysis in this study suggests greater efficacy against AURKA and its substrates than against HSP90 clients, this work suggests a partial HSP90 inhibitory activity, as well. Further detailed analysis and compound optimization will be important in future studies. In conclusion, this study provides evidence to support the rationale for stable bi-functional HSP90-AURKA conjugates to improve efficacy due to increased drug exposure in the tumor tissue, and to improve safety by minimizing mechanism-based toxicity on normal tissues.