Targeted Brain Tumor Radiotherapy Using an Auger Emitter

Glioblastoma multiforme (GBM) is one of the deadliest forms of solid tumors, with a 5-year survival rate as low as 5% (1). Clinical intervention typically consists of maximal safe surgical resection followed by adjuvant chemoradiotherapy. This therapeutic regimen, unfortunately, imparts only limited improvements to survival (2). In addition, the GBM molecular heterogeneity represents a robust challenge in need of better imaging tools that would allow for the monitoring of therapy response and lead to better and more personalized therapeutic plans. Furthermore, the presence of the blood–brain barrier (BBB) biochemically limits the pharmacokinetics of many GBM drug candidates.

A novel approach for treating GBM is therefore urgently needed, one that would address both pharmacodynamic as well as pharmacokinetic hurdles (3, 4). Promising delivery strategies to overcome these obstacles in the clinics include aligning novel targeting schemes with improved drug delivery approaches including convection-enhanced delivery (CED; refs. 5, 6) and intrathecal injection (7, 8).

A known molecular biomarker for most tumors, including GBM, is PARP1 (9–15). PARP1 is recruited to the nucleus of cancer cells and binds DNA as a single-strand break repair enzyme (16). This central role has been successfully leveraged for the development of various PARP inhibitors, both as a monotherapy and in combination with other therapeutics (17, 18). Modified PARP inhibitors have also been widely used for tumor detection and imaging due to their cancer specificity and their high target-to-background contrast (14, 15, 19–25); more recently, they have found theranostic applications (26).

Here, we focus on PARP1-based radiotherapy with a more sporadically utilized type of radioactive emission: Auger radiation. Auger emitters are an extremely potent radioactive source for targeted radiotherapy, characterized by their greater linear energy transfer, incredibly short range, and ability to cause complex, lethal DNA damage as compared with traditional x-rays or β-particles (27–32). Previous attempts to use Auger emitters as cancer therapies have not been successful, due to the limited range of the radiation emitted and the difficulty of reliably delivering the lethal electrons close enough to the DNA target (<100 Å; refs. 31, 33–35).

In this study, we developed and characterized Iodine-123 Meitner-Auger PARP1 inhibitor (¹²³I-MAPi), the first Auger-based theranostic PARP inhibitor able to directly deliver its lethal payload within a 50 Å distance of the DNA of GBM cancer cells (Fig. 1A). This distance is within the Auger radius of action, resulting in an effective preclinical cancer treatment drug and leading to improved survival in a preclinical GBM model. We used the 159 keV gamma ray to image tumor progression using single-photon emission CT (SPECT) imaging and calculate dosimetry and treatment efficacy using SPECT imaging.

Taken together, these results illustrate the tremendous potential of ¹²³I-MAPi as an Auger-emitting PARP inhibitor and a theranostic agent for GBM treatment.

Sodium [¹²³I]iodide in 0.1 N NaOH with a specific activity of 7.14 × 10⁷ GBq/g was purchased from Nordion. 4-(4-fluoro-3-(4-(3-iodobenzoyl)piperazine-1-carbonyl)benzyl)phthalazin-1(2H)-one was synthesized as described previously (36). Olaparib (AZD2281) was purchased from LC Laboratories. PARPi-FL was synthesized as described previously (14, 37). Water (>18.2 MΩcm⁻¹ at 25°C) was obtained from an Alpha-Q Ultrapure Water System and acetonitrile (AcN) as well as ethanol were of high-performance liquid chromatography (HPLC) grade purity. Sterile 0.9% saline solution was used for all in vivo injections. HPLC purification and analysis was performed on a Shimadzu UFLC HPLC system equipped with a DGU-20A degasser, an SPD-M20A UV detector, a LC-20AB pump system, and a CBM-20A Communication BUS module. A LabLogic Scan-RAM radio-TLC/HPLC-detector was used to detect activity. HPLC solvents (buffer A: water, buffer B: AcN) were filtered before use. Purification of 2,5-dioxopyrrolidin-1-yl 3-(iodo-¹²³I)benzoate was performed with method 1 (flow rate: 1 mL/minute; gradient: 15 minutes 5%–95% B; 17 minutes 100% B; 20 minutes 100%–5% B); QC analysis was performed with Method 1. Method 1 was performed on a reversed-phase C18 Waters Atlantis T3 Column (C18-RP, 5 μm, 6 mm, 250 mm). Purification of the final product was performed on a C6 Waters Spherisorb Column (C6, 5 μm, 4.6 mm × 250 mm) with method 2 (flow rate: 1.5 mL/minute; isocratic: 0–30 minutes 35% B.

The human glioblastoma cell line U251 was kindly provided by the laboratory of Dr. Blasberg (MSKCC, New York, NY). Cells were grown in Eagle minimal essential medium (MEM), 10% (vol/vol) heat-inactivated FBS, 100 IU2 penicillin, and 100 μg/mL streptomycin, purchased from the culture media preparation facility at MSKCC (New York, NY). TS543 cells are a patient-derived glioblastoma stem line kindly provided by the laboratory of Dr. Mellinghoff (MSKCC, New York, NY). These cells were grown in suspension in NeuroCultTM NS-A Proliferation Kit with proliferation supplement (StemCell Technologies, catalog no. 05751), 20 ng/mL Recombinant Human EGF (StemCell Technologies, catalog no. 02633), 10 ng/mL Recombinant Human Basic Fibroblast Growth Factor (StemCell Technologies, catalog no. 02634), 2 μg/mL heparin (StemCell Technologies, catalog no. 07980), 1× antibiotic–antimicotic (Life Technologies Gibco, catalog no. 15240-062), and 2.5 mg/mL Plasmocin (InvivoGen, Cat ant-mpp). All cells were tested for Mycoplasma contamination.

Cells were plated on a coverslip slide on the bottom of a 6-well plate and incubated overnight. ¹²³I-MAPi, ¹²⁷I-PARPi, or vehicle controls were added to the media and cells were returned to the incubator overnight. Cells were then fixed in 4% paraformaldehyde/PBS, permeabilized, and stained with anti-γ-H2AX antibody (Millipore Sigma, 05-636) and anti-PARP1 antibody (Invitrogen, PA5-16452). DAPI was used to localize nuclei. Coverslips were mounted on slides for microscopy imaging.

Cells were lysed in RIPA (Thermo Fisher Scientific, catalog no. 89900) buffer containing protease inhibitor at 4°C. Lysates were run on an SDS-Page gel (Bio-Rad). Bound antibodies were detected by developing film from nitrocellulose membranes exposed to chemiluminescence reagent (Thermo Fisher Scientific, catalog no. 34077). PARP1 primary antibody (Santa Cruz Biotechnology, catalog no. sc-7150, 0.2 μg/mL) and goat anti-rabbit IgG-HRP secondary antibody (Santa Cruz Biotechnology, catalog no. sc-2004, 1:10,000 dilution). An anti-β-actin antibody (Sigma, catalog no. A3854, 1:1,000) was used as loading control.

¹²³I-MAPi was obtained similar to synthetic procedures reported before (26). First, 2,5-dioxopyrrolidin-1-yl 3-(iodo-¹²³I)benzoate was obtained by adding N-succinimidyl-3-(tributylstannyl) benzoate (250 μg, 0.5 μmol) in 10 μL of AcN to a solution containing methanol (40 μL), chloramine T (9 μg, 32 nmol), acetic acid (3 μL), and ¹²³I-NaI in NaOH 0.1 mol/L (2.5 mCi). After the reaction solution was driven for 20 minutes at room temperature, the reaction was purified by HPLC (method 1), and 2,5-dioxopyrrolidin-1-yl 3-(iodo-¹²³I)benzoate at 15.1 minutes. The collected purified fraction was concentrated to dryness in vacuum, reconstituted in a solution of 80 μL can, and added to a solution of 4-(4-fluoro-3-(piperazine-1-carbonyl)benzyl)phthalazin-1(2H)-one (0.3 mg, 0.9 μmol) in 20 μL DMSO, N,N,N′,N′-Tetramethyl-O-(1H-benzotriazol-1-yl)uronium hexafluorophosphate (HBTU; 0.3 mg, 0.8 μmol) in 20 μL DMSO, 4-Dimethylaminopyridine (DMAP; 0.1 mg, 0.8 μmol) in 20 μL DMSO and 10 μL triethylamine. The reaction mixture stirred in an Eppendorf ThermoMixer for 2 hours at 65°C (500 rpm). Afterward, the reaction mixture was injected and purified by HPLC (method 2) and ¹²³I-MAPi was collected at RT = 25.5 minutes (RY > 70%; RP > 99%) and concentrated to dryness under vacuum. ¹²³I-MAPi was formulated with 30% PEG300/70% saline (0.9% NaCl) for both in vitro and in vivo assays. Coelution with nonradioactive ¹²⁷I-PARPi reference compound confirmed the identity of the radiotherapeutic. ¹²³I-MAPi was synthesized with molar activity of 3.93 ± 0.10 GBq/μmol.

Uptake of ¹²³I-MAPi was tested in vitro (three replicates). 5 × 10⁵ U251 cells were plated 24 hours prior to the experiment (n = 3). Media was changed and 1 hour later 3.7 kBq of ¹²³I-MAPi were added to the cells. For blocking, cells were incubated with a 100-fold molar excess of olaparib 1 hour before adding ¹²³I-MAPi. Media was removed, and cells were washed with PBS and lysed (1 mol/L NaOH) at different time points. The lysate was collected, and uptake was determined by radioactivity on a gamma counter.

U251 GBM cells were incubated with 0–296 kBq of ¹²³I-MAPi or equivalent dose of olaparib overnight and then washed and incubated for 4 days in normal media (three replicates, n = 3 each). Viability was determined by AlamarBlue assay as indicated by the manufacturer.

U251 GBM cells were incubated with 370 kBq of ¹²³I-MAPi and apoptosis was detected at 1 and 24 hours posttreatment using an in situ Direct DNA Fragmentation (TUNEL) Assay Kit (Abcam, ab66110) and following manufacturer's instructions.

Colony formation assay (CFA) was performed using U251 cells. Cells were plated as single-cell suspensions and left to attach for 8 hours prior to irradiation or treatment (n = 3). Colony formation was measured at 2 weeks post ¹²³I-MAPi treatment and compared with EBIR. Cells were treated adding 0–23 kBq of compound in each well. Colony count was normalized on plating efficiency at 0 kBq for each treatment. External beam irradiation (EBIR) was performed using a Shepperd Irradiator Cs-137 at Memorial Sloan Kettering Cancer Center (New York, NY). Radiobiological parameters of the linear-quadratic model were determined via nonlinear regression within GraphPad software, and relative biological effectiveness was determined by interpolation, using 37% survival as the endpoint.

All in vivo experiments were performed with female athymic nude CrTac:NCr-Fo mice purchased from Taconic Laboratories at age 6–8 weeks. During subcutaneous injections, mice were anesthetized using 2% isoflurane gas in 2 L/minute medical air. 1 × 10⁶ TS543 cells were injected in the right shoulder subcutaneously in 150 μL volume of 50% media/matrigel (BD Biosciences) and allowed to grow for approximately 2 weeks until the tumors reached about 8 mm in diameter (100 ± 8 mm³).

For intravenous injections, the lateral tail vein was used. Mice were warmed with a heat lamp, placed in a restrainer, and the tail was sterilized with alcohol pads before injection.

For orthotopic injections, TS543 cells (5 × 10⁵ cells in 2 μL growth media) were injected 2 mm lateral and 1 mm posterior to the bregma using a Stoelting Digital New Standard Stereotactic Device and a 5 μL Hamilton syringe and allowed to grow for 3 weeks before treatment. All mouse experiments were performed in accordance with protocols approved by the Institutional Animal Care and Use Committee of MSKCC and followed NIH guidelines for animal welfare.

SPECT/CT scans were performed using a small-animal NanoSPECT/CT from Mediso Medical Imaging Systems. For subcutaneous TS543 xenografts, ¹²³I-MAPi (2.3 ± 0.5 MBq in 20 μL 30% PEG300 in 0.9% sterile saline) was administered intratumorally as single injection. At chosen time points postinjection, the mice were anesthetized with 1.5%–2.0% isoflurane (Baxter Healthcare) at 2 mL/minute in oxygen and SPECT/CT data were acquired for 60 minutes.

For orthotopic TS543 tumor-bearing mice, ¹²³I-MAPi (614.2 kBq in 5 μL 30% PEG300 in 0.9% sterile saline) was injected intracranially using the same coordinates as for tumor cell injection (2 mm lateral and 1 mm posterior to the bregma using a Stoelting Digital New Standard Stereotactic Device and a 5 μL Hamilton syringe). At chosen time points postinjection, the mice were anesthetized with 1.5%–2.0% isoflurane (Baxter Healthcare) at 2 mL/minute in oxygen. SPECT/CT data were collected for 60 minutes.

Biodistribution studies were performed in subcutaneous TS543 xenograft-bearing mice. Mice were randomized and divided in two groups (blocked and unblocked, n_total = 6) and ¹²³I-MAPi was administered intratumorally (average injected activity 1.702 ± 0.629 MBq in 20 μL, 30% PEG300 in 0.9% sterile saline). The blocked group was preinjected (1 mg/mouse in 100 μL 30% PEG300 in 0.9% sterile saline) 60 minutes prior to treatment with olaparib (100 mmol/L stock in DMSO). Mice were sacrificed by CO₂ asphyxiation at 18 hours postinjection and counted in a WIZARD2 automatic γ-counter (PerkinElmer). Uptake was expressed as a percentage of injected dose per gram (%ID/g) using the following formula: [(activity in the target organ/grams of tissue)/injected dose].

Mice were inoculated orthotopically at week 0 with TS543 cells (5 × 10⁵ cells in 2 μL growth media). Cells were injected 2 mm lateral and 1 mm posterior to the bregma using a Stoelting Digital New Standard Stereotactic Device and a 5 μL Hamilton syringe and allowed to grow for 3 weeks before treatment. At week 3, mice were randomly grouped into cohorts, and intratumorally injected with ¹²³I-MAPi or 30% PEG300 in 0.9% sterile saline vehicle using the same stereotactic coordinates as for tumor implantation. Mice were monitored daily thereafter. Study endpoint was determined on the basis of animals' sign of discomfort, pain, or significant weight loss.

ALZET Osmotic Pumps were implanted subcutaneously into the mice to slowly deliver an infusion into the brain using the same coordinates as for the tumor cell injection as described previously (26). Control mice received an ALZET Osmotic Pump Model 1003D with Brain Infusion Kit 3 containing 30% polyethylene glycol (PEG)/PBS vehicle. Treatment mice received ¹²³I-MAPi (average pump activity 481 ± 111 kBq in 100 μL, 30% PEG300 in 0.9% sterile saline). Delivery flow was 1 μL/hour, over 5 days. Pumps were surgically removed 5 days postimplantation and mice monitored daily thereafter.

Brain tumor–bearing mice were anesthetized and injected in the intrathecal space (injection site from L3/5 or L4/5 to prevent spinal cord injury) with 2.0 MBq of ¹²³I-MAPi in 50 μL 30% PEG300 in 0.9% sterile saline. ¹²³I-MAPi were injected and the mice imaged 1 hour postinjection.

Brains were collected from mice at the time of death. The collected brains were embedded in Tissue-Tek O.C.T. compound (Sakura Finetek), flash-frozen in liquid nitrogen, and cut into 10-μm sections using a Vibratome UltraPro 5000 Cryostat (Vibratome). Sections were subsequently subjected to hematoxylin–eosin (H&E) staining for morphologic evaluation of tissue pathology.

Mice were injected intracranially at 4 weeks posttumor implantation with either ¹²³I-MAPi or vehicle for control (n = 3/cohort). Mice were perfused 1 hour postinjection and brains were fixed in 4% paraformaldehyde. Brains were then sliced and stained for γ-H2AX as a marker for double-strand breaks (DSB).

Animals (n = 5/cohort) were injected systemically with either ¹²³I-MAPi or Vehicle control. After 24 hours, blood was collected via retro-orbital bleeding. Enzymes levels were determined by the Antitumor Assessment Core Facility at Memorial Sloan Kettering Cancer Center (New York, NY).

Dosimetry calculations are described in detail in the Supplementary Materials.

All data necessary for interpreting the manuscript have been included in the main manuscript or in the Supplementary Materials. Additional information may be requested from the authors.

We previously showed that it is possible to successfully conjugate a PARP inhibitor with radioiodine without altering the high affinity to the target—maintaining an IC₅₀ in the nanomolar range (11 ± 3 nmol/L) and a logPCHI of 2.3 (38). ¹²³I-MAPi is a novel, previously unreported isotopolog, and was synthesized with a final molar activity of 3.93 ± 0.10 GBq/μmol. Radiochemical purity was 99.1 ± 0.9% for all prepared compounds (Fig. 1B).

Pharmacologic properties determined with ¹²⁷I-PARPi suggest that ¹²³I-MAPi (Fig. 1C; Supplementary Fig. S1A), retains the same properties as ¹³¹I-PARPi, which have been shown to be similar to the FDA-approved PARP inhibitor olaparib (26, 38).

In vitro internalization was tested on U251 cells expressing PARP1 (Supplementary Fig. S1B). ¹²³I-MAPi (37 kBq/well) was added to adherent cells in monolayer and uptake was calculated by measuring gamma radiation in cell lysates at different time points. We confirmed rapid cellular internalization, with 50% of the final total uptake being reached after 5 to 10 minutes postincubation and with an uptake plateau reached at 1 hour posttreatment (Fig. 1D). The final total uptake was approximately 8% of added activity, V_max = 8.2 ± 0.2 % compared with blocked uptake, V_max = 3.8 ± 0.1 % (Michaelis-Menten fit, R² = 0.974 and R² = 0.879, respectively). Uptake was blocked with a 100-fold excess dose of olaparib to show target specificity. Na-¹²³I was used as a control and showed significantly lower uptake, V_max = 2.5 ± 0.1. The two dominant targets for ¹²⁷I-PARPi, the stable-isotope labeled form of the molecule, are PARP1 and PARP2 (26), similar to what has been previously reported for olaparib and other modified PARP inhibitors (15).

We tested the cancer-specific efficacy of ¹²³I-MAPi in vitro by treating U251 GBM cells with ¹²³I-MAPi. Comparing ¹²³I-MAPi with the previously published β-emitting ¹³¹I-PARPi, we observed a 16-fold increase in EC₅₀ potency (EC₅₀ of ¹³¹I-PARPi being 1148 ± 1 nmol/L; Supplementary Fig. S1C). ¹²³I-MAPi treatment proved to reduce viability at safe concentrations of ¹²³I-NaI (Supplementary Fig. S1D and S1E). Critically, in the molar concentration range used in these studies, olaparib showed negligible effect in terms of cell viability, suggesting that clonogenic inactivation arises exclusively from I-123 radiotoxicity as opposed to PARP1 inhibition (Fig. 1E). ¹²³I-MAPi proved to be capable of killing cancer cells with a submicromolar EC₅₀ (EC₅₀ = 68.9 ± 1.1 nmol/L, R² = 0.999). At the same concentrations, olaparib did not show any effect in terms of cell viability, suggesting that the observed effect is due to the PARP1 inhibitor–mediated close proximity to the target DNA.

To characterize the induction of DNA damage in cancer cells after treatment with ¹²³I-MAPi, we performed immunofluorescence analysis of the levels of γ-H2AX foci, a known marker for DSBs (39). We treated U251 GBM cells with 740 kBq ¹²³I-MAPi in 20 μL of 30% PEG/PBS added to the media in each well (2 mL) and compared it with the equivalent dose of nonradioactive ¹²⁷I-PARPi and vehicle control (Fig. 1F; Supplementary Fig. S2A). ¹²³I-MAPi–treated cancer cells showed a significantly higher number of γ-H2AX foci in the cell nucleus (P < 0.001, Kruskal–Wallis test; Fig. 1G). We performed a TUNEL assay to detect apoptosis after treatment with ¹²³I-MAPi in vitro and detected increased levels of apoptotic cells at 24 hours posttreatment as compared with control (Supplementary Fig. S2B; P < 0.05).

We tested ¹²³I-MAPi in comparison with externally applied photon radiation, assessing clonogenic survival. CFAs were performed in 6-well plates to compare efficacy of ¹²³I-MAPi in cell killing when compared with standard EBIR. U251 GBM cells were plated and treated with Cs-137 γ-rays (662 keV) or by adding ¹²³I-MAPi to the wells (Supplementary Fig. S3A).

As expected with high linear energy transfer (LET) radiation treatment, cells treated with the Auger-emitting small molecule showed a steep decline in survival, confirming the high efficacy in tumor cell killing when compared with EBIR (Supplementary Fig. S3B). We derived first-order estimates of relative biological effectiveness of on-target ¹²³I-MAPi through cell dosimetry calculations (Supplementary Information), which suggest an increase in therapeutic potency when ¹²³I-MAPi is bound to PARP1 (i.e., when DNA is within reach of its Auger emissions). Radiobiological parameters were calculated after linear-quadratic model fit of the data: ¹²³I-MAPi α = 19.8 Gy⁻¹, β = N/A. EBIR α = 0.269 Gy⁻¹, β = 0.0588 Gy⁻¹. ¹²³I-MAPi D₃₇ = 0.05 Gy, EBIR D₃₇ = 2.41 Gy. Relative biological effectiveness (RBE) was calculated for ¹²³I-MAPi: RBE at D₃₇ = 48.4, RBE (α_I-123/α_EBIR) = 73.4 (Supplementary Table S1).

TS543 patient-derived glioblastoma stem cells were used to grow tumors in athymic nude mice to investigate the biodistribution of ¹²³I-MAPi. Tumors were grown subcutaneously on the animals' right shoulder. Mice were then randomized and divided into two cohorts (n = 3/group) one of which was used for blocking. Blocking was performed systemically with an intravenous injection of olaparib 1 hour prior to ¹²³I-MAPi treatment. ¹²³I-MAPi was injected intratumorally for this model. SPECT/CT images showed that ¹²³I-MAPi tumor uptake was retained at 1, 6, and 18 hours postinjection in nonblocked animals, but was strongly reduced at 6 and 18 hours in blocked animals (Fig. 2A; Supplementary Fig. S4A). Biodistribution at 18 hours was also examined and showed higher tumor uptake (33.4 ± 28.0 %ID/g) compared with the tumor of blocked animals (0.4 ± 0.1 %ID/g; Fig. 2B). Tumor-to-muscle ratios in ¹²³I-MAPi–treated mice versus blocked mice were more than 500 and 5, respectively, with less than 1%ID/g in all clearing organs.

Auger particles are highly cytotoxic when they can directly interact with the DNA and cause complex damage. However, they are significantly less so in the cellular cytosol, where they are beyond the reach of their DNA target (27). As liver clearance is the main route of excretion of ¹²³I-MAPi that could cause dose-limiting problems in the clinic, we investigated potential clinical liver failure by observing specific liver toxicity in our preclinical study. We compared tumor and liver accumulation of PARPi-FL, a thoroughly characterized fluorescent analogue of the same PARP1 inhibitor with comparable biodistribution and tumor uptake after intravenous injection (14, 15, 20, 38, 40). PARPi-FL was injected intravenously (50 μg/mouse) 2 hours before collecting the animals' tumor and liver (Fig. 2C; Supplementary Fig. S5A). Organs were then sectioned, and tissue sections were digitalized. Hoechst (150 μL/mouse of 10 mg/mL) was used for nuclear counterstaining. PARPi-FL accumulated in the nucleus of GBM cells as confirmed by colocalization with Hoechst signal. Livers were collected and imaged with an inverted confocal microscope. PARPi-FL accumulation was observed in the cytoplasm of liver cells (Fig. 2D), suggesting that liver cells are protected from Auger toxicity. To confirm this, we performed an enzyme analysis assay in the blood of mice injected with and without prior injection of ¹²³I-MAPi. 1.09 ± 0.04 MBq were injected in n = 5 mice for the treated cohort, whereas the vehicle control cohort (n = 5) received 150 μL of 30% PEG/PBS. No significant variations were observed in treated mice compared with control (Supplementary Fig. S5B), suggesting a limited systemic toxicity of the injected molecule, likely due to the large distance of the Auger emitter from the DNA (Supplementary Fig. S5C).

To test ¹²³I-MAPi in a more realistic model of GBM, we orthotopically implanted TS543 cells into the right brain hemisphere. This GBM model proved to be very consistent in growth and take rate, as we monitored by MRI imaging of the head. 50,000 cells were injected at week 0, which resulted in rapid disease progression leading to animal death at week 7 (Fig. 3A). On the basis of this data, we decided to deliver ¹²³I-MAPi treatment at week 3. ¹²³I-MAPi was injected intratumorally using the same stereotactic coordinates as for tumor implantation. Animals were then imaged at 1 and 18 hours posttreatment. Full-body SPECT/CT images were acquired, showing retention of ¹²³I-MAPi in the tumor at 18 hours postinjection (Fig. 3B; Supplementary Fig. S6A).

We monitored the clinical potential utility of ¹²³I-MAPi in treating GBM in the above described orthotopic mouse model. ¹²³I-MAPi–treated mice were then monitored daily, using day 98 (the end of week 14), measured from the day of xenografting, as the study endpoint. We injected an intratumoral single dose (0.37–1.11 MBq) of ¹²³I-MAPi for the treated cohort (n = 10) and the same volume of vehicle for the control cohort (n = 12). Survival data confirmed an improved survival for the ¹²³I-MAPi treatment cohort, with a median survival of 58 days as opposed to 40 days observed for the control cohort. Log-rank curve comparison showed a significant difference, with P = 0.009 (Fig. 3C; Table 1). Animals brains were imaged ex vivo with H&E staining and in vivo with MRI showing a diffuse presence of cancer cells in vehicle-treated mice as opposed to the treatment (Fig. 3D). Mice brains (n = 3/cohort) were also stained for the presence of γ-H2AX foci. We found a significantly increased number of DSB at 1 hour postintratumoral injection of ¹²³I-MAPi when compared with vehicle (P < 0.0001, Kruskal–Wallis test; Supplementary Fig. S6B and S6C).

In clinics, to improve biodistribution toward a more effective targeted therapy, the radiopharmaceutical can be administered directly into the tumor compartment. For brain tumors, this is achieved through intratumoral injection, intrathecal injection, or CED, an approach that elevates the injection pressure so as to impel the agent across the BBB. To allow for clinical translation of ¹²³I-MAPi, we first built a preclinical model of CED. We implanted an ALZET osmotic delivery pump with a subcutaneous catheter in tumor-bearing mice. This delivered the content of the subcutaneous reservoir at a flow rate of 1 μL/h over the course of approximately 100 hours through a cannula connected to the mouse brain. Subcutaneous reservoirs were filled with 100 μL of ¹²³I-MAPi for the treatment cohort (n = 8) and with 100 μL of vehicle for the control cohort (n = 8). Five days postimplantation, we surgically removed the pumps, as per manufacturer's instruction, and monitored mice survival. Kaplan–Meier survival plots confirmed the therapeutic efficacy of ¹²³I-MAPi–treated mice presenting a median survival of 72 days as opposed 48 days in the control cohort, a statistically significant difference of 50% (log-rank P = 0.0361; Fig. 4A; Table 2). Dosimetry for the CED experiments (Fig. 4B; Supplementary Fig. S6D) confirmed accumulation in the tumor with minimal absorbed dose to healthy organs not in direct proximity to the implanted pump (Fig. 4C). Importantly, while not feasible in a mouse model, normal organ doses would be significantly reduced in a corresponding clinical scenario where the radionuclide reservoir would be external, shielded, and the administration performed over a shorter timescale.

We then investigated a more broadly used and technically feasible brain drug delivery method for delivering ¹²³I-MAPi across the BBB. Tumor-bearing mice were injected with 50 μL of ¹²³I-MAPi intrathecally at 3 weeks after orthotopic tumor implantation. SPECT/CT imaging was performed at 1 hour postinjection to visualize ¹²³I-MAPi accumulation in the brain (Supplementary Fig. S7A). Imaging confirmed favorable pharmacokinetics and specific tumor uptake, suggesting intrathecal delivery as a feasible technique for our small molecule. Unfortunately, this technique leads to surgery-related high levels of stress in the mice model preventing us from being able to monitor survival.

In this article, we show the results for the first preclinical characterization of an Auger-emitting theranostic PARP inhibitor. We present a functioning workflow for the synthesis of a stable ¹²³I-MAPi compound which proved to be effective in vitro at reducing the viability of GBM cell lines. The lethality of ¹²³I-MAPi suggests that the DNA is in range of the Auger electrons when it is complexed with PARP1. It was further established that GBM cell death is induced through radiogenic damage rather than PARP inhibition at these tracer levels.

In vivo studies with human tumor xenograft models also show ¹²³I-MAPi drug treatment efficacy. The 159 keV gamma emission of I-123 allowed for quantification of drug uptake in tumor-bearing mice by SPECT/CT imaging. The isotope is ideal for a proposed theranostic PARPi clinical agent because of its widespread use in nuclear medicine for thyroid disorders (Na¹²³I) and pediatric tumors (¹²³I-MIBG). Animal studies employing the nonradioactive drug olaparib show prolongation of blood pool clearance and blocked tumor uptake, proving ¹²³I-MAPi–binding specificity.

To illustrate the potential advantages of clinical translation of ¹²³I-MAPi, and to better illustrate the impact of Auger therapeutics, we looked at the agent's potential subcellular biodistribution. In vitro γ-H2AX analysis showed a significantly increased number of DSBs in cells treated with ¹²³I-MAPi when compared with ¹²⁷I-PARPi or vehicle control (P < 0.001, Kruskal–Wallis test; Supplementary Fig. S2D). Furthermore, by looking at the foci morphology, it is possible to observe larger and more clustered foci in the ¹²³I-MAPi–treated samples as compared with spontaneously induced and background foci (Supplementary Fig. S2A–S2C). These observations were also confirmed in vivo, showing elevated levels of γ-H2AX foci post ¹²³I-MAPi treatment in GBM cells when compared with control (Supplementary Fig. S6C). In this study, we did not notice any variation in the levels of PARP1 (green staining) as expected due to a positive feedback loop of self-amplification of PARP1 expression (21). For this reason, we decided to not include a potential PARP1 amplification effect in the kinetic modeling for dosimetry studies. We performed a CFA to test efficacy of ¹²³I-MAPi in U251 cells. The obtained data show a steep drop in survival, which was expected due to the high-LET nature of Auger electrons. Comparison with EBIR allowed us to calculate the RBE, which corroborated the potential of Auger-emitting molecules for targeted radiotherapy, a fast-growing and promising field (41, 42). This proved to be a significant improvement compared with a previously published version of the molecule, ¹³¹I-PARPi, where the RBE of the emitted electron is significantly lower. There has been a prior phase I/II clinical trial in which the therapeutic effects of two antibodies targeting colon cancer were studied [¹³¹I-A33 (43) and ¹²⁵I-A33 (29)]. I-125 is an Auger electron–emitting radionuclide with comparable properties to I-123 (44). This study showed that the maximum tolerated activity for ¹³¹I-A33 was 2.78 GBq/m² (75 mCi/m²) in heavily pretreated patients, whereas for ¹²⁵I-A33, bone marrow toxicity was not seen after administered activities as high as 12.95 GBq/m² (350 mCi/m²). In this study as in the A33 trial, it is expected that the short-range emissions from I-123 will result in far lower normal tissue toxicity (including dose-limiting organs). In vivo, olaparib-based PARP radiotherapeutics are likely going to be cleared via the hepatobiliary pathway (15) and the liver is therefore a potential organ for Auger-specific radiotoxicity. In preparation for clinical translation of ¹²³I-MAPi, we performed studies to investigate liver toxicity. We found near-exclusive extracellular or perinuclear localization of ¹²³I-MAPi in liver cells (i.e., the DNA of the liver parenchymal cells is out of range of most of the emitted low-range Auger electrons) as shown with the fluorescent analog PARPi-FL. This is in stark contrast to the observed nuclear accumulation of ¹²³I-MAPi in GBM cells (Fig. 2C). The cytoplasmic liver accumulation is also corroborated by a liver enzyme analysis we performed on injected mice. We did not find significant changes of enzyme levels when compared with control mice (Supplementary Fig. S5B). The radiobiological effectiveness of Auger-emitting compounds is critically dependent on the proximity of the electron emitter to the cellular DNA (45, 46). This distance-dependent relationship for DSB production has been shown for I-125 (47). Cell-level dosimetry analysis showed a noteworthy increase in cell sterilization as a function of absorbed dose with ¹²³I-MAPi in comparison with external photon irradiation, resulting in high estimates of relative biological effectiveness and suggesting that intranuclearly delivered radiation doses from PARP1-bound ¹²³I-MAPi are nearly 50 times as potent as that from externally directed photons. This estimation, though dependent on the parameters of the Monte Carlo simulation geometry, is in agreement with the expected equivalent dose for Auger electron emitters, which is comparable with that of intracellularly incorporated 5.3 MeV α-particles (35, 48). The dose was calculated to the tumor cells using PARaDIM v1.0, a Monte Carlo method, using as input the time-integrated activity coefficients from measured data (49). These cellular dose estimates were confirmed against standard cellular methods from MIRDcell (50). We do, however, caution that these cell dose and RBE estimates strongly depend on the measured cell size, end point, reference radiation, and model assumptions. A limitation of this study is that the RBE is calculated starting from in vitro observations. This is because the orthotopic mouse model made it challenging to dose-paint the tumor with external beam irradiation without simultaneously irradiating the whole brain. On the basis of the limited literature on the potential use of I-123–labeled pharmaceuticals for therapy (51, 52)—although this could be different in using I-123–labeled PARP agents)—projected in-human dose could start from an imaging dose of 8 mCi and collect biodistribution data for dosimetry calculations, guiding a phase I study in patients.

To force a more favorable biodistribution toward tumor targeting in clinical targeted radionuclide studies directed at brain tumors, investigators have proposed intratumor injection (53), intrathecal injection (54), and more recently CED to administer the antibody across the BBB and into tumor tissue—for example, to treat diffuse intrinsic pontine glioma (55). The above approaches have shown the ability to significantly improve the ratio of the radiation dose delivered to tumor relative to normal dose-limiting tissues. Many clinics do have the capability to perform intratumoral and intrathecal injection with or without a CED system (56–58). This delivery method, albeit not simple, is becoming more widely used as more patients present with tumors growing in the central nervous system, a consequence of the improved efficacy of chemotherapy for the treatment of vascular-accessible disease. To emulate these studies in a preclinical setting, we adopted an orthotopic GBM model in mice and performed a local injection of ¹²³I-MAPi using an intratumoral osmotic pump delivery system. This system allowed for constant and prolonged delivery of the drug directly to the tumor and its microenvironment. Dosimetry estimates based on imaging, tissue specimen counting, and treatment response shown by Kaplan–Meier survival data are suggestive of the high potential clinical impact of this approach.

CEDs and Ommaya reservoirs are emerging as effective chemotherapy and targeted radionuclide delivery methods for patients with previously inaccessible metastatic disease. We have developed a method to perform CED injection of novel radiopharmaceuticals in preclinical models of cancer. We also performed intrathecal injection which would allow tumor imaging and therapy in clinical settings. This technique is routinely performed in most hospitals for chemotherapy drug delivery and allows clinicians to access the brain without opening the patient's skull. SPECT/CT imaging can visualize and quantify radiolabeled drug uptake in tumor and its dispersion throughout the tissues of the body. We have demonstrated this capability with a newly developed theranostic agent, ¹²³I-MAPi. We have shown that it can access the brain when injected intrathecally and using the intratumoral osmotic pump delivery system. While the favorable pharmacokinetics was accompanied by stressful side effects in some mice (on account of the comparatively small volumes of the murine skull), this has not been, nor is it expected to be, a limiting factor for human studies.

To conclude, we characterized the first Auger-emitting PARP inhibitor in vivo which presented promising therapeutic results in a preclinical glioma model. The physical properties of Auger emission, paired with the biological distribution of a PARP inhibitor, make it possible to speculate that a dose escalation in patients could achieve high tumoricidal doses with limited normal tissue toxicity. The 159 keV gamma ray emitted is at the sweet spot for gamma camera imaging in patients and will allow for monitoring of treatment delivery across the BBB, redistribution, and precise dose evaluation, leading to truly personalized treatment plans (Supplementary Fig. S7B and S7C). ¹²³I-MAPi has the potential to be a new and potent clinical agent for the treatment of brain tumors when used in conjunction with new intrathecal/CED administration methods.

S. Kossatz holds ownership interest (including patents) in and is an unpaid consultant/advisory board member for Summit Biomedical Imaging. J.S. Lewis holds ownership interest (including patents) in Summit Biomedical Imaging. T. Reiner is a paid consultant for and reports receiving commercial research grants from Theragnostics; reports receiving other commercial research support from Pfizer and Summit Biomedical Imaging; and holds ownership interest (including patents) in and is an unpaid consultant/advisory board member for Summit Biomedical Imaging. No potential conflicts of interest were disclosed by the other authors.

Conception and design: G. Pirovano, S.A. Jannetti, A. Sadique, J.S. Lewis, J.L. Humm, T. Reiner

Development of methodology: G. Pirovano, S.A. Jannetti, S. Kossatz, T. Reiner

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): G. Pirovano, S.A. Jannetti, A. Sadique, S. Kossatz, N. Guru, P. Demétrio De Souza França, M. Maeda

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): G. Pirovano, S.A. Jannetti, L.M. Carter, J.S. Lewis, J.L. Humm, T. Reiner

Writing, review, and/or revision of the manuscript: G. Pirovano, S.A. Jannetti, L.M. Carter, S. Kossatz, N. Guru, P. Demétrio De Souza França, B.M. Zeglis, J.S. Lewis, T. Reiner

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): G. Pirovano, S.A. Jannetti, A. Sadique, T. Reiner

Study supervision: S.A. Jannetti, B.M. Zeglis, T. Reiner

We thank Dr. Cameron Brennan and the Brain Tumor Center for the GBM cell line model. We thank Dr. Ingo Mellinghoff and Dr. Beatriz Salinas for their helpful discussions and for providing their expertise. We thank Dr. Pat Zanzonico and Valerie Longo for technical support and help with animal imaging. We thank Dr. Elisa de Stanchina and Dr. Vanessa R. Thompson from the Antitumor Facility Core. We thank the Molecular Cytology Core Facility, the Animal Imaging Core Facility, and the Radiochemistry & Molecular Imaging Probes Core Facility at Memorial Sloan Kettering Cancer Center. We thank Dr Ronald A. Ghossein, MD for consultation on pathology slides. We also thank Garon Scott for editing the manuscript. This work was supported by NIH under grants R01 CA204441, P30 CA008748, R43 CA228815, R35 CA232130, and K99 CA218875. The authors thank the Tow Foundation and MSK's Center for Molecular Imaging & Nanotechnology and Imaging and Radiation Sciences Program. L.M. Carter acknowledges support from the Ruth L. Kirschstein Postdoctoral Fellowship (NIH F32-EB025050).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.