Imaging of Cancer γ-Secretase Activity Using an Inhibitor-based PET Probe

Abstract

Purpose:

Abnormal notch signaling promotes cancer cell growth and tumor progression in various cancers. Targeting γ-secretase, a pivotal regulator in the Notch pathway, has yielded numerous GSIs for clinical investigation in the last two decades. However, GSIs have demonstrated minimal success in clinical trials in part due to the lack of specific and precise tools to assess γ-secretase activity and its inhibition in vivo.

Experimental Design:

We designed an imaging probe based on GSI Semagacestat structure and synthesized the radioiodine labeled analogs [¹³¹I]- or [¹²⁴I]-PN67 from corresponding trimethyl-tin precursors. Both membrane- and cell-based ligand binding assays were performed using [¹³¹I]-PN67 to determine the binding affinity and specificity for γ-secretase in vitro. Moreover, we evaluated [¹²⁴I]-PN67 by PET imaging in mammary tumor and glioblastoma mouse models.

Results:

The probe was synthesized through iodo-destannylation using chloramine-T as oxidant with high labeling yield and efficiency. In vitro binding results demonstrate the high specificity of this probe and its ability for target replacement study by clinical GSIs. PET imaging studies demonstrated a significant (P < 0.05) increased in the uptake of [¹²⁴I]-PN67 in tumors versus blocking or sham control groups across multiple mouse models, including 4T1 xenograft, MMTV-PyMT breast cancer, and U87 glioblastoma xenograft. Ex vivo biodistribution and autoradiography corroborate these results, indicating γ-secretase specific tumor accumulation of [¹²⁴I]-PN67.

Conclusions

[¹²⁴I]-PN67 is a novel PET imaging agent that enables assessment of γ-secretase activity and target engagement of clinical GSIs.

Introduction

γ-Secretase is an intramembrane protease complex consisting of four obligatory subunits: presenilin (PS1 or PS2), nicastrin, Aph-1, and PEN2 (1). This protease cleaves a variety of type I membrane proteins that are linked with human diseases (2), including the two most well-studied substrates: amyloid precursor protein (APP) and Notch receptor (3). γ-Secretase cleaves APP to release amyloid-beta (Aβ) peptides in various lengths, among which Aβ42 is the most amyloidogenic form that contributes to the pathogenesis of Alzheimer’s disease (AD) (4). Moreover, proteolytic cleavage of Notch receptor by γ-secretase releases the Notch intracellular domain (NICD), which plays a critical role in the transduction of Notch signaling pathways. Dysregulation of the Notch pathway is involved in varying types of cancers (5). Small molecules that include γ-secretase inhibitors (GSIs) and modulators (GSMs) (6–10) have been developed for target-based therapies. After two GSIs, Semagacestat (LY-450139) and Avagacestat (BMS-708163), failed in AD clinical investigation, the major research efforts surrounding GSIs have been focused on their therapeutic potential in cancer treatment. However, the development of γ-secretase based therapies has been hindered by a lack of suitable means to assess γ-secretase activity in vivo and target engagement.

Assaying γ-secretase activity both in vitro and in vivo has been a challenge as only a fraction of γ-secretase complexes is catalytically active, and enzymatic activity does not correlate with the amount of the catalytic subunit, presenilin (11,12). Active site-directed probes have been developed to examine the active complex (7). Recently, a GSM-based probe [¹¹C]SGSM-15606 has been reported to detect γ-secretase in vivo (13). However, GSMs that selectively modulate γ-secretase activity for Aβ42 production do not reflect overall γ-secretase activity in vivo. Therefore, it is crucial to develop probes that enable the detection and monitoring of γ-secretase activity in vivo for the clinical development of GSIs for cancer treatment.

Positron emission tomography (PET) is a powerful tool for probing biochemical processes and enzymes in vivo and plays a vital role in drug discovery. In this study, we have developed a novel PET probe named PN67 based on the scaffold of Semagacestat, the first GSI that was taken into phase III trials (14). Using membrane- and cell-based ligand binding assays, we first demonstrate the high binding potency and specificity of the probe to γ-secretase in vitro. Next, we show the tumor-specific uptake of [¹²⁴I]-PN67 in mouse models of breast cancer and glioblastoma. Our studies indicated that PN67 PET probe can evaluate target engagement of GSIs and facilitate further clinical investigation of γ-secretase based therapies.

Materials and Methods

Chemical synthesis

General Methods: All commercially available reagents and solvents were used without further purification. All reactions were performed in oven-dried glassware with magnetic stirring unless noted otherwise. Nitrogen was used to protect air and moisture-sensitive reactions. Thin layer chromatography (TLC) was performed on EMD Millipore, TLC Silica gel 60 F₂₅₄ 100 Aluminium sheets and visualized with ultraviolet (UV) light (254nm) and/or potassium permanganate (KMnO₄) staining. Silica flash chromatography was performed with RediSep Rf Gold® silica gel flash columns on a Teledyne Isco CombiFlash Rf system.

Nuclear magnetic resonance (NMR) spectra were recorded on a Bruker UltraShield Plus 600 MHz Avance III NMR at 24°C in CDCl₃ unless otherwise indicated. Spectra were processed using Bruker TopSpin or MestReNova software, and chemical shifts are expressed in ppm relative to TMS (¹H, 0 ppm) or residual solvent signals: CDCl₃ (¹H, 7.26 ppm; ¹³C, 77.16 ppm); coupling constants are expressed in Hz. Mass spectra were obtained at the MSKCC Nuclear Magnetic Resonance (Analytical) Core Facility on a Waters Acuity SQD LC-MS by electrospray (ESI) ionization or atmospheric pressure chemical ionization (AP-CI). High resolution mass spectra (HRMS) were obtained on a Waters Acuity Premiere XE TOF LC-MS by electrospray ionization.

(S)-2-hydroxy-N-((S)-1-(((S)-8-iodo-3-methyl-2-oxo-2,3,4,5-tetrahydro-1H-benzo[d]azepin-1-yl)amino)-1-oxopropan-2-yl)-3-methylbutanamide (2, PN67):

LY-450139 (36mg, 0.1mmol) was dissolved in 2ml of mixed solvent (Acetic acid : H₂O : H₂SO₄ = 100 : 20 : 3), followed by adding periodic acid (9.1mg, 0.04mmol) and iodine (25mg, 0.1mmol). The mixture was stirred with magnetic bar and heated to 70°C for overnight. After cooling to rt, the reaction mixture was diluted with EtOAc, washed with 1M NaHCO₃ solution, water, brine, dried over Na₂SO₄, filtered, and concentrated by rotary evaporator. The resulting crude was purified by silica gel flash column on Teledyne Isco CombiFlash system (gradient: 5% - 20% MeOH/DCM) providing the desired product PN67 as a colorless sticky solid (25mg, 51%).

(S)-2-hydroxy-3-methyl-N-((S)-1-(((S)-3-methyl-2-oxo-8-(trimethylstannyl)-2,3,4,5-tetrahydro-1H-benzo[d]azepin-1-yl)amino)-1-oxopropan-2-yl)butanamide (3):

2 (25mg, 0.05mmol) was dissolved in a sealed tube by 1.5 ml of toluene, followed by adding Tetrakis(triphenylphosphine)palladium(0) (6mg, 0.005mmol) and Hexamethylditin (33mg, 1mmol) under argon protection. The reaction mixture was then heated to 100°C and stirred overnight. After cooling to rt, the crude mixture was filtered through a short celite column and purified by HPLC (gradient: 15% - 95% MeCN/H₂O without TFA). The final tin precursor was obtained from lyophilizer as a white powder (9mg, 34%).

Synthesis of [¹³¹I]PN67 and [¹²⁴I]PN67

Tin precursor 3 (10 ug) in Eppendorf tube was dissolved in 20ul of methanol and followed by adding a solution of carrier-free [¹³¹I]-NaI in 0.1N NaOH (1–5 mCi, 20–50 ul, Nuclear Diagnostic Products) or [¹²⁴I]-NaI in 0.1N NaOH (1–5 mCi, 20–50 ul, Radioisotopes and Molecular Imaging Probes core facility at MSKCC). To the resulting solution, 2 ul of chloramine-T solution (2mg/ml in acetic acid) was added and the reaction tube was then placed on 37°C heating plate for 10min.The crude product was purified by a Shimadzu HPLC system with a bioscan flow-count radio-HPLC detector system for detection of radioactivity (gradient 30 – 95% MeCN/H₂O containing 0.1% TFA). The peak eluent was collected and concentrated by the rotary evaporator for further use.

Cell line culture

The following cells were cultured in Dulbecco’s Modified Eagle Medium (DMEM) with high glucose, penicillin, streptomycin, and 10% FBS: ANPP cells (HEK293 cells that overexpress all four components of γ-secretase, kindly provided by Dr. Sam Sisodia); MEF (wild type and PS1/PS2 double knockout, kindly provided by Huaxi Xu); 4T1-Luciferase (mouse breast cancer cell with stable firefly luciferase expression, kindly provided by David A. Scheinberg); A172, U87, and SF295 (human glioblastoma cells, kindly provided by I.K. Mellinghoff). The cell lines were grown and passaged regularly at 70 – 80% confluence every 2–3 days. All cells were cultured at 37°C with 5% CO₂.

In vitro γ-secretase activity assay

Cell membrane was used as the source of γ-secretase and was prepared as previously described (15,16). The in vitro γ-secretase activity assay was performed according to previously reported literatures (17,18). Briefly, recombinant APP substrate CT6 (1uM) was incubated with membrane protein (40 μg/ml) and 0.25% CHAPSO for 3 hours at 37 °C. To measure enzyme inhibition for Aβ cleavage, the activity assay was carried out in the absence or presence of various concentrations of inhibitors. The reaction (20ul) was then combined 1:1 with detection mix (20ul), which includes streptavidin-conjugated Alpha donor beads (PerkinElmer Life Sciences), anti-mouse IgG AlphaLISA acceptor beads (PerkinElmer Life Sciences) and Aβ40 specific antibody G2–10. The sample was incubated in the dark at room temperature overnight, and the AlphaLISA signal was detected with an Envision plate reader (PerkinElmer Life Sciences).

In vitro ligand binding assay

In vitro binding studies were performed using either cell membranes or suspended cells (19). The assay was performed by incubating cell membrane (30ug of the protein) with radioligand [¹³¹I]-PN67 in a total volume of 200 μl PBS buffer for 60min at 37 °C. The incubation was terminated by rapid filtration through the glass microfiber filters (pre-soaked in 0.6% polyethylenimine, Whatman Grade GF/C filters) by using the cell harvester (Brandel M-24 Harvester). The filters were subsequently washed with cold TBS buffer (pH 7.5, 4 °C) three times. Activity in each filter was measured in a gamma counter (WIZARDTM 3” 1480 gamma counter, Perkin Elmer). For the competition binding assay, various concentrations of test compounds (0.3nM to 1μM) were co-incubated with a fixed dose of radioligand (50,000 cpm). For the saturation binding assay, the hot ligand ([¹³¹I]-PN67, 2 × 10⁶ cpm) was mixed with cold ligand (2, 300nM), and the mixture was diluted in different concentrations (0.1nM - 300nM), which were further used for membrane incubation. For each data point, triplicates were performed, and the data was analyzed using Graphpad Prism 7 software.

The cell-based binding assay was performed with the same procedures described above except using living cells. The cells were grown up to 70 – 80% confluence, trypsinized, and collected. For each data point, 1 × 10⁶ suspended cells were incubated with radioligand for 60min at 37 °C in 0.5ml DMEM medium without FBS.

Mice

All animal experiments and procedures were carried out in compliance with the National Institutes of Health’ Guide for Care and Use of Laboratory Animals’ and institutional guidelines from Animal Care and Use Committee at Memorial Sloan Kettering Cancer Center.

For 4T1-luc model: 6–8 weeks female BALB/c inbred mice (BALB/cAnNHsd) were obtained from Envigo. 4T1-luciferase cells were inoculated in the right shoulder by subcutaneous injection of 5 × 10⁶ cells in a 200μl cell suspension of a 1:1 v/v mixture of media with Matrigel. 4T1 tumor size was measured by vernier caliper and calculated using the formula: tumor volume = (length × width²)/2. The average tumor size was 458 ± 259 mm³ (n = 10) before imaging.

U87 glioblastoma model was generated by intracranial injection using athymic nude mice (Foxn1nu, 6–8 weeks old, The Jackson Laboratory). Mice were anesthetized using ketamine (100 mg/kg) and xylazine (10 mg/kg) by intraperitoneal injection and no pain reflex was seen by foot pinch. The head of the mouse was then shaved and cleaned with alcohol and betadine. An incision was made along the midline of the skull, and the injection site was drilled into the skull at 2 mm right lateral of bregma. The mice were then placed in a stereotactic frame, and a Hamilton syringe (10μl) was inserted 3.5 mm into the brain and slowly retracted 0.5 mm to create room for the cells. 2 × 10⁵ U87 cells in 2 μl PBS buffer were injected into the striatum over 2 min, then the needle was allowed to rest for 5 min. The burr hole was closed with bone wax and the wound was sealed with Vetbond Tissue Adhesive. The tumor was allowed to grow for 4–5 weeks and was confirmed by MRI scan. The U87 tumor volume was determined by MRI imaging, and the average size is 26.9 ± 7.7 mm³ (n = 6).

MMTV-PyMT mice on the B6 background were purchased from The Jackson Laboratory (Stock No. 022974). Female mice positive for the MMTV-PyMT allele develop primary mammary tumors around 92 days of age.

Biodistribution studies

The ex vivo biodistribution studies were performed in 4T1-luc mouse model. The mice were injected with 0.1% potassium iodide in saline (100μl, intraperitoneal) for 1h prior to the injection of radioactivity for thyroid blocking of free I-124. [¹²⁴I]-PN67 (20–30 μCi, in 200 μl of 0.9% saline with 10% DMSO) was injected intravenously in the tail. Competitive inhibition (blocking) studies were performed using non-radiolabeled LY-450,139 (1, 30mg/kg) in formulation of [¹²⁴I]-PN67 (20–30 μCi, in 200 μl of 0.9% saline with 10% DMSO). The mice were euthanized by CO₂ gas asphyxiation at multiple time points after injection, and tissues were harvested and weighed. Activity in organs and tumors was then measured with a gamma counter (Perkin Elmer). Count data were background- and decay-corrected, and the tissue uptake was calculated in units of percentage injected dose per gram (%ID/g).

PET imaging

PET images were acquired on an Inveon small-animal micro-PET/CT scanner (Siemens Medical Solutions, Knoxville, USA) or a Focus 120 small-animal PET scanner (Siemens Medical Solutions, Knoxville, USA) using a dedicated quadruple animal scanning platform for simultaneous scanning. Mice (n = 4 per timepoint) were anesthetized 5–10 minutes prior to scanning via inhalation of a 1.5–2% isofluorane (Baxter Healthcare, Deerfield, USA) in oxygen gas mixture. 1h prior to administration of radioactivity, the mice were injected with 0.1% potassium iodide in saline (100ul, intraperitoneal) for thyroid blocking of free iodine-124. Radiotracer (200 uCi, in 200 ul of 0.9% saline with 10% DMSO) was injected intravenously. Competitive inhibition (blocking) studies were performed by co-injecting [¹²⁴I]-PN67 (200 uCi) with LY-450,139 (1, dose range: 30 – 0.2 mg/kg) in formulation of 10% DMSO in saline (total 200ul) or RO-4929097 (10 mg/kg) in formulation of 10% DMSO, 30% PEG300, and 5% Tween-80 in saline (total 200ul). List mode PET data were acquired over 10 minutes with an energy window of 350–650 keV and a coincidence-timing window of 6 ns. Data were sorted into 2-dimensional histograms by Fourier rebinning, and transverse images were reconstructed by 3D ordered subsets expectation maximization maximum a posteriori (3D OSEM-MAP; 4 OSEM iterations, 18 MAP iterations) into a 128 × 128 × 95 matrix. A 1.6 mm FWHM gaussian post-filter was applied. The image data were normalized to correct for nonuniformity of response of the PET, dead-time count losses, positron branching ratio, and physical decay to the time of injection, but no attenuation, scatter, or partial-volume averaging correction was applied. Individual image volumes were constructed using an in-house Python program. Activity concentrations (%ID/g) were obtained by conversion of the counting rates in the reconstructed images using a system calibration factor, derived from a mouse-sized water-equivalent phantom containing 18F. Further image processing was performed using Inveon Research Workplace (Siemens Medical Solutions, Knoxville, USA). The brain PET images were analyzed using VivoQuant software (version 3.0, InviCRO) as described previously (20). Three-dimensional volumes of interest (VOIs) were defined through manual segmentation and activity concentrations were quantified by selecting the mean uptake value per VOI.

In vivo optical imaging

Optical imaging of 4T1-luciferase was performed on an IVIS 200 optical imaging system (Xenogen Corp). Mice were injected with 200ul D-luciferin (30mg/ml in PBS buffer) intraperitoneally 10 minutes prior to imaging. 1.5% isoflurane/air mixture was then used to anesthetize the mice by inhalation. Mice were imaged for 60 seconds in the imaging chamber.

MRI imaging

Magnetic resonance imaging (MRI) was performed by Animal Imaging MRI core at MSKCC using previously published methods (21). Briefly, MRIs were acquired two days prior to radiotracer injection. Mice were anesthetized with 1%–2% isoflurane gas in oxygen and positioned prone in the scanner. The 9.4-Tesla Biospec Scanner (Bruker Biospin Corp.) with a 114-cm Bruker gradient coil (maximum gradient strength, 530 mT/m) was used for mouse brain imaging, and a Bruker ID 4-cm quadrature volume coil was used for RF excitation and detection. First T2-weighted scout images along three orthogonal orientations were acquired. Then T2-weighted mouse brain images along the trans-axial orientation were acquired covering the whole brain with 16 slices using the rapid acquisition with relaxation enhancement (RARE) fast spin-echo sequence. (Acquisition parameters: slice thickness, 0.5 mm; repetition time, 2.5 s; echo time, 33 ms; RARE factor, 8; spatial resolution, 98 × 78 μm)

Autoradiography

Ex vivo autoradiography was performed using 40um thick tissue sections collected at 2 hours after injection of the radiotracer. Anatomy was confirmed by H&E staining. Section-bearing slides were exposed to digital autoradiography films for 4 days (subcutaneous tumor sections) or 12 days (intracranial tumor sections). The film was then scanned using a Typhoon phosphor imager (Amersham). The result images were visualized and analyzed by Fiji-ImageJ software.

Results

Synthesis of PET imaging probe PN67

The synthetic route and radiochemistry of the probe are described in Fig 1. Iodination of 1 gave 2 (PN67) as the major product, which further underwent a palladium-catalyzed stannylation reaction to obtain the stannane precursor 3 for radiolabeling. Radioiodination by electrophilic demetallation has been performed with high labeling yield and efficiency. The labeling reaction was carried out with either [¹³¹I]-NaI for in vitro binding studies or [¹²⁴I]-NaI for in vivo PET studies to give the desired probe (recovery yield > 80%; recovery purity > 99%; specific activity > 2.3 Ci/umol). The identity of the radiolabeling product was confirmed by co-injecting non-radioactive 2 with radioactive fraction onto HPLC (Supplementary Fig. 1).

Figure 1.

Radiosynthesis of [¹³¹I] or [¹²⁴I]-PN67. Reagents and conditions: (a) HIO₄, I₂, acetic acid, 70°C, 4h; (b) Pd(PPh₃)₄, hexamethylditin, toluene, 90°C, overnight; (c) [¹³¹I]-NaI or [¹²⁴I]-NaI, chloramine-T, 37°C, 10min.

Binding affinity for γ-secretase in vitro

First, we compared the inhibitory activities of 1 and 2 to γ-secretase in vitro. Despite bearing an extra iodine group, 2 improved γ-secretase inhibition potency (IC₅₀ = 3.5 nM) compared with its original compound 1 (IC₅₀ = 16.2 nM, Fig. 2A). Next, we determined the binding properties of PN67 to γ-secretase. We conducted the membrane-based radioligand binding assay with [¹³¹I]-PN67 using HeLa cell membranes. Non-specific binding of [¹³¹I]-PN67 was defined in the presence of 1 μM of compound 1. There is little non-specific binding to Hela membrane (<2%, Fig. 2B). Moreover, competition binding assays showed that K_i values of ligand 1 and 2 are 13.6 and 4.3 nM, respectively (Fig. 2C), which have similar potency for inhibition of γ-secretase.

Figure 2.

In vitro validation of [¹³¹I]-PN67 through ligand binding assays and structure basis of LY-450139 and PN67 (docking) interaction with human γ-secretase. (A) The potencies of the 1 and 2 were determined in a dose-titration study with the in vitro γ-secretase alphaLISA assay. (B) [¹³¹I]-PN67 binding with purified HeLa cell membranes was determined under baseline and blocking conditions where 1uM 1 was used to block the specific binding. (C) Results from competition binding assays with [¹³¹I]-PN67 as the radioligand. Radioligand bound is measured in the presence of varied concentrations of cold ligands 1 and 2. The specific binding was analyzed using GraphPad Prism software to give K_i value for each competitor, which is shown in lower table. (D) Saturation binding curve of [¹³¹I]-PN67 to the γ-secretase in purified membranes from HeLa and ANPP cells. The K_d and B_max values were calculated using GraphPad Prism software and shown in lower tables. (E) Radioligand binding was determined in two types of MEF cells (wt and PS1/2^−/−). The ligand bound showed a significant reduction when PS1/2 was knockout. (F) Cryo-EM structure of γ-secretase complex binding with LY-450139 (PDB bank: 6LR4). (G) The specific binding of LY-450139 to PS1; (H) Molecular docking of PN67 to PS1 binding site. Iodine group (purple) faces towards a hydrophobic crevice formed by surrounding residues (Val272, Leu 282 and Ile 287). The electrostatic potential of key residues was shown as transparent surface. The Glide docking was performed by Maestro 12.6 in Schrödinger suite. All data presented are means ± SE of triplicate samples (n=3).

We further characterized the [¹³¹I]-PN67 binding with membrane fractions from HeLa and ANPP cell (A stable HEK293 cell line with overexpression of all four γ-secretase subunits (22)). The maximum number of binding sites (B_max) and dissociation constant (K_d) of the [¹³¹I]-PN67 were determined for both cell membranes by a saturation binding assay (Fig. 2D). Under saturated conditions, ANPP membranes showed much higher γ-secretase (B_max = 2825 fmol/mg) than HeLa membranes (B_max = 979 fmol/mg), while [¹³¹I]-PN67 bound to ANPP and HeLa membranes with similar affinity (K_d = 4.6 and 6.5 nM, respectively).

Finally, we investigated the binding specificity in live cells using two mouse embryonic fibroblast (MEF) cell lines – wild type (WT) and presenilin 1/2 (PS1/PS2) double knockout (PS dKO). PS1/PS2 are catalytic subunits of γ-secretase and the enzyme activity is abolished in PS1/PS2 deficient cells (16). The binding of [¹³¹I]-PN67 to both cell lines was measured after washing. PS dKO cells have little binding activity (< 7%) as compared to WT, confirming that [¹³¹I]-PN67 specifically binds to γ-secretase (Fig. 2E).

The precise binding site of LY-450139 that includes the interaction with catalytic residue Asp385 and other PS1 residues has been reported (23). It offers a molecular basis for improved potency of PN67 compared with LY-450139, because iodine enhances the interaction with the hydrophobic crevice formed by Val272, Leu 282, and Ile 287 (Fig. 2 F–H).

The binding of [¹³¹I]-PN67 to γ-secretase strongly correlates with γ-secretase activity in cell membranes and live cells

To further examine the relationship between the ligand binding and γ-secretase activity, we determined both proteolytic activities (Fig. 3A) and [¹³¹I]-PN67 binding (Fig. 3B) in three different cell lines – ANPP, HeLa, and HEK293 cells. The relation of two sets of data was analyzed after normalization (Fig. 3C) and showed a strong linear correlation (R² = 0.99, *P < 0.05) between two variables. This suggests that [¹³¹I]-PN67 binding is correlated with γ-secretase enzymatic activity, and that the use of [¹³¹I]-PN67 is suitable for the quantification of γ-secretase activity in membranes. After establishing that [¹³¹I]-PN67 specifically binds to γ-secretase and is an indicator of its protease activity, we next examined the relationship between the binding of [¹³¹I]-PN67 to γ-secretase with other GSIs. We selected three GSIs (PF-03084014, MK-0752, RO-4929097) that have been used for preclinical and clinical investigation for their antitumor activity (Fig. 3D) (24). We determined their IC₅₀ for γ-secretase inhibition and K_i values in blocking [¹³¹I]-PN67 binding using Hela cell membrane (Fig. 3E). A strong correlation between their inhibitory potencies for γ-secretase activity and [¹³¹I]-PN67 binding indicates that this probe can be used for target engagement of these GSIs in clinical studies (Fig. 3F).

Figure 3.

[¹³¹I]-PN67 binding is associated with the active form of r-secretase complex in membranes and cells. (A) The enzymatic activity of γ-secretase in three different cell membranes (ANPP, HEK293, and HeLa) was determined by γ-secretase alphaLISA assay. (B) The amount of radioligand bound in three cell membranes (ANPP, HEK293, and HeLa) was measured by binding experiments with [¹³¹I]-PN67 (50,000 cpm/each sample). (C) The enzyme activity and radioligand binding data from A and B were drawn in a scatter plot and analyzed by linear regression (R² = 0.96, ****P < 0.0001). (D) Chemical structures of GSIs used in this study. (E) Potencies of GSIs for γ-secretase inhibition (IC₅₀) and [¹³¹I]-PN67 binding (K_i). (F) The correlation between enzyme inhibition (IC₅₀) and target engagement (K_i) was shown in the scatter plot. The fit curve was determined by linear regression (R² = 0.98, **P < 0.01). All data presented are means ± SE of triplicate samples (n=3).

Biodistribution study in mice models

We next assessed whether PN67 can be used for a PET probe for detection of γ-secretase in cancer cell lines. Our previous studies showed that 4T1 cells and mouse tumor models had a great response to γ-secretase inhibition by GSIs (25). In addition, the 4T1 cell tumor model is frequently used to model late-stage breast cancer and assess preclinical cancer drugs and other therapeutics (26). Notch activation is frequently found in human gliomas and important for glioma stem cells (27). It has been reported that enrichment of Notch signaling components correlates with response to GSIs in some glioma cells (28). GSIs also showed encouraging clinical outcomes to gliomas in several early-stage clinical trials (29,30). Therefore, we characterized [¹³¹I]-PN67 binding in four cell lines: 4T1 mouse mammary tumor cell with luciferase expression (4T1-Luc) and three human glioblastoma cells (A172, SF295, and U87). As shown in Fig. 4A, these cell lines had varying amounts of probe binding. We selected 4T1-Luc and U87 to generate animal models for further in vivo studies as they have relatively higher amounts of ligand binding than the other cell lines.

Figure 4.

Ex vivo biodistribution for [¹²⁴I]-PN67 in 4T1-Luc mice. (A) Cell based binding assays with various cancer cell lines, including 4T1-Luc (luciferase), A172, SF295, and U87 cells. (B) biodistribution of [¹²⁴I]-PN67 in mouse tissues at 2h post injection. (C) Tumor uptake of [¹²⁴I]-PN67 was shown at different time points (0.5h, 1h, 2h, 4h and 8h) post injection. Blocking was carried out by co-injecting of 1 (30mg/kg) with the tracer, and the data at 2h time point was shown with significant reduction (**p < 0.01) in tumor uptake. (D) Tumor to blood ratio of [¹²⁴I]-PN67 uptake in 4T1-Luc mice from biodistribution studies. The highest tumor/blood ratio was overserved at 2h time point. All data presented are means ± SE (binding assay n=3, animal studies n=5). %ID/g = percentage injected dose per gram.

To further examine [¹²⁴I]-PN67 in vivo, we performed ex vivo biodistribution studies in 4T1-Luc allograft model. Animals were sacrificed at different time points (0.5 – 8h) after tail-vein injection of [¹²⁴I]-PN67, and then radioactivity was measured in various tissues upon dissection. [¹²⁴I]-PN67 accumulated strongly at early time points (0.5 – 2h, Fig. 4B, Supplementary Table 1) in the GI tract, liver, and kidney, suggesting fast elimination of the radiotracer by biliary and renal excretion. Additionally, [¹²⁴I]-PN67 also showed a rapid tissue clearance rate – the radioactivity in most tissues decreased to less than 1 %ID/g after 8h post-injection. Notably, the tumor uptake of [¹²⁴I]-PN67 reached its peak within the first hour post-injection (Fig. 4C), but the tumor to blood ratio was the highest at 2 hour post injection (Fig. 4D), which suggests the 2 hour would be ideal time point for in vivo tumor PET imaging. In addition, the probe distribution in wild type and 4T1 tumor mice are very similar (Supplementary Table 2). To demonstrate specificity, we assessed whether tumor uptake of [¹²⁴I]-PN67 could be blocked. Upon co-administered [¹²⁴I]-PN67 with 1, 4T1 tumors showed a significant reduction of tracer uptake (Fig. 4C, Supplementary Table 1).

In vivo PET imaging

We performed PET imaging of [¹²⁴I]-PN67 in BALB/c mice bearing 4T1-Luc tumor implanted subcutaneously on the right shoulder. The tumor growth and location were confirmed by bioluminescence imaging before the PET scan (Supplementary Fig. 2A). [¹²⁴I]-PN67 was administered intravenously both at baseline and under blocking conditions. As shown in Fig. 5A, [¹²⁴I]-PN67 provided a clear tumor accumulation at 2h post-injection (Fig. 5A). Next, we checked the inhibition of [¹²⁴I]-PN67 signal imaging using varying doses of GSIs – LY-450139 (1, 0.2 mg/kg to 30 mg/kg) and RO-4929097 (10 mg/kg, Fig. 5). The PET/CT imaging showed that the [¹²⁴I]-PN67 signal can be blocked in a dose-dependent manner (Fig. 5A and 5C). Furthermore, the similar dose-response was observed by autoradiography using the tumor cryosection samples (Fig 5B, Fig 5D). Finally, we examined the effect of RO-4929097 on PET imaging and found that RO-4929097, a structurally distinct γ-secretase inhibitor, can block [¹²⁴I]-PN67 signal (Fig. 5A–D). Taken together, these biochemical, cellular, and in vivo studies indicate that [¹²⁴I]-PN67 PET probe is able to monitor tumor γ-secretase activity and target engagement.

Figure 5.

In vivo PET/CT imaging studies in 4T1-luc BALB/c mice using various blocking agents and doses. (A) Representative PET/CT imaging of 4T1 tumors using [¹²⁴I]-PN67 under different blocking conditions (LY-450139, from 0.2 mg/kg to 30 mg/kg; RO-4929097 10 mg/kg). The images were taken at 2h post injection. Blocking was carried out by co-injecting [¹²⁴I]-PN67 with blocking agents; (B) Representative autoradiography of tumors from (A) were shown; (C) Tracer tumor uptake in PET imaging was quantified. Statistical analysis was performed using unpaired t-test. (n=4, **P < 0.01, ****P < 0.0001); (D) Radioactivity in autoradiography sections was quantified using ImageJ. Statistical analysis was performed using unpaired t-test. (n=5, ***P < 0.001, ****P < 0.0001). Data are average %ID/g ± SD, %ID/g = percentage injected dose per gram.

In addition, we evaluated the [¹²⁴I]-PN67 in the MMTV-PyMT (mouse mammary tumor virus-polyoma middle tumor-antigen) mouse model of breast cancer, which closely resembles the progression and morphology of human breast cancers (31,32). The PET imaging was conducted at 2h post-administration of [¹²⁴I]-PN67 (Supplementary Fig. 3). Similar to the 4T1 model, higher tumor uptake of the radiotracer at the baseline condition was observed than that in the blocking condition.

Next, U87 glioblastoma mice were generated by intracranial injection of U87-MG cells in athymic nude mice. The tumor was allowed to develop for 4 – 5 weeks, and the growth dynamics were monitored by magnetic resonance imaging (MRI, Fig. 6A, Supplementary Fig. 4). Sham surgery mice were used as a control to investigate radiotracer specificity to the tumor region. [¹²⁴I]-PN67 again displayed greater tumor-specific uptake than in the normal brain tissue and in the sham group at two hours post-injection (Fig. 6A & B). The mice were sacrificed, and whole brains were obtained from the animals in the PET imaging study; brains were fixed and embedded for autoradiography and hematoxylin and eosin staining (H&E staining, Fig. 6A). Radioactivity in brain sections is associated with tumor localization and is absent in sham control groups. It appears that [¹²⁴I]-PN67 was not able to detect endogenous γ-secretase in mouse brain under current conditions. In parallel, we also compared γ-secretase activity in membrane fractions of both mouse brain and U87 cells, and significantly higher enzyme activity was observed in U87 cells (Fig. 6C), indicating that this probe detects relatively high γ-secretase activity in cancer cells. In addition, we investigated the distribution of γ-secretase activity in normal tissues (Supplementary Fig. 5) and showed that γ-secretase activity was generally low in most of the normal tissues, except for the brain.

Figure 6.

PET imaging of γ-secretase in U87 glioblastoma mouse model. (A) The U87 glioblastoma mice and sham surgery control mice were administrated with [¹²⁴I]-PN67, and the PET/CT, autoradiography, H&E staining, and MRI results were shown. The PET/CT was taken at 2h post-injection. Brain tissue was frozen immediately after PET/CT imaging. (B) The tracer uptake (PET) in U87 and sham control mice were quantified using VivoQuant software. The tumor region of interest (ROI) was selected under the guidance of MRI results. (n=4, *P < 0.05) (C) Comparison of γ-secretase enzyme activity in membrane fractions of U87 cells and brain tissues. (n=3, **P < 0.01)

Discussion

One of the reasons for the failure of GSIs in AD trials was the severe adverse effects and systemic toxicity caused by Notch signaling inhibition. However, Notch upregulation has been reported to be associated with various types of cancers, indicating a potential avenue for the repurposing of GSIs for use as antitumor agents. Determination of γ-secretase activity and inhibition in vivo remains a significant challenge that can aid in the development and dose optimization of GSI-based treatments for cancer and other diseases. We designed PET probes based on two GSIs: Semagacestat (LY-450139) and Avagacestat (BMS-708163). Despite BMS-708163 based chemical probes being widely used to investigate γ-secretase and map the binding site (33–35), BMS-708163 based radioligands showed high non-specific binding. However, LY-450139 based probes, reported here, exhibit ideal characteristic properties as a PET probe. First, it has very low non-specific binding in cell membranes and intact cells. Second, it specially binds to γ-secretase as demonstrated by assays run in γ-secretase deficient cells. Third, the probe binding is highly correlated with γ-secretase activity. Finally, PN67 PET probe is able to detect cancer γ-secretase in vivo and the PET imaging is blockable in a dose-dependent manner. Taken together, these findings have demonstrated that PN67 can serve as a PET probe to monitor γ-secretase in vivo.

In mouse PET imaging study, the biodistribution of PN67 is high in the abdominal region, particularly in clearance organs like the liver, small and large intestine. Our data showed that γ-secretase activity was generally low in most normal tissues (Supplementary Fig. 5), suggesting PN67 distribution is related to its in vivo metabolism/elimination pathway, rather than the tissue γ-secretase activity. Based on ex vivo biodistribution and PET imaging, the half-life of PN67 in mouse models is about 2 hours, which means the majority of probe/radioactivity will be eliminated in the first few hours after injection. In biodistribution, we found a radioactivity transfer from the liver (max at 0.5h) to the small intestine (max at 1h), then to the large intestine (max at 2h, Supplementary Table 1), indicating biliary excretion might be the primary route for the tracer elimination in mouse models.

It has been reported that γ-secretase contains multiple sites that interact with various classes of GSIs and GSMs (6). The BMS-708163 binding site has been mapped through chemical biology approaches (35) and confirmed by structural studies (23). However, how the PN67 binding interplays with the pocket occupied by other clinal GSIs candidates is unknown. We have previously shown that the interaction of BMS-708163 can be blocked by LY-450139 (33), suggesting that two compounds have overlapping binding sites or share the same sites. Our current studies indicate that all three clinical GSIs (PF-03084014, MK-0752, RO-4929097) effectively block PN67 binding to γ-secretase. Importantly, the binding affinity of these three GSIs is highly correlated with their potencies in inhibiting γ-secretase. Collectively, these studies indicate that PN67 not only detects cancer γ-secretase activity in vivo, but also can be utilized as a PET probe to image γ-secretase inhibition for other structurally distinct GSIs in clinical studies. In addition, this probe will enable the evaluation of the relationship of newly developed clinical GSIs with PN67 binding. Clearly, PN67 has enormous potential for a general application to determine the target engagement of various GSIs for clinical studies.

It is noteworthy to mention that current studies focus on cancer cells where γ-secretase is highly expressed. γ-Secretase plays an important role in the pathogenesis of Alzheimer’s disease. However, tracer uptake in brain tissue was low in our current studies, probably due to poor blood brain barrier (BBB) penetration. The application and potential of this probe for imaging brain γ-secretase associated with Alzheimer’s disease warrant further investigation.

In conclusion, we have designed, synthesized, and characterized a diagnostic PET probe for imaging cancer γ-secretase. [¹²⁴I]-PN67 specifically detected the active γ-secretase in both in vitro ligand binding experiments and in vivo PET imaging. The high tumor uptake and specific binding of [¹²⁴I]-PN67 suggest that it has the potential to become a novel class of imaging agents to guide the clinical development of GSIs for determining the target engagement and the relationship between γ-secretase inhibition and clinical responses.

Translational Relevance.

γ-Secretase is an appealing therapeutic target for cancer due to its role in regulating the Notch signaling and other pathways associated with malignancies. γ-Secretase inhibitors (GSIs) have been actively pursued for the treatment of varying cancers. However, it is challenging to evaluate the efficacy of GSIs in vivo, and there is an unmet need for specific tools to assess γ-secretase activity and target engagement of clinical GSIs. Here, we report a novel non-invasive GSI based PET imaging probe [¹²⁴I]-PN67 that demonstrates high specificity in vitro and detects γ-secretase in vivo and can be used to monitor the inhibition of other structurally distinct GSIs. Our study suggests that this probe could further guide the clinical development of GSIs for determining the target engagement and the relationship between γ-secretase inhibition and clinical response. In addition, such a tracer could be used for patient stratification at the early stage of GSI development.