Safety and Feasibility of PARP1/2 Imaging with ¹⁸F-PARPi in Patients with Head and Neck Cancer

Diagnosis and treatment of many malignant tumors have dramatically improved in recent decades, and oral and oropharyngeal squamous cell carcinoma are no exception to this trend. However, most patients with oral and oropharyngeal squamous cell carcinoma still present with advanced disease and regional or distant metastases at the time of diagnosis.

Presence of pathologic regional lymph nodes is the most powerful and consistent predictor of outcome for oral cancers (1), and the ability to accurately assign the exact extent of metastatic spread within the neck lymphatic system is therefore of great significance (1, 2). Complete surgical removal of metastatic disease in the neck has a clear impact on prognosis, but it often remains unclear until histopathologic analysis is complete, if a lymph node was metastatic or not (3, 4).

Clinical palpation of the neck is inadequate, as are the available radiologic investigative tools. This is because many patients present with enlarged lymph nodes due to inflammation around the tumor site. These inflamed lymph nodes may mimic neck metastases on CT scans and are often ¹⁸F-FDG PET avid (5). On the other hand, some metastatic neck nodes may not be enlarged and show no abnormal ¹⁸F-FDG uptake. For these reasons, elective neck dissection or irradiation are often recommended in patients with head and neck cancer for prophylactic treatment of occult metastases. Ideally; however, these procedures, which can lead to increased comorbidities and reduced quality of life due to overtreatment (6, 7), would be avoided if the presence of occult metastases could be definitively ruled out.

PARP PET imaging could potentially provide a solution for this unmet clinical need. Because of its importance for cell survival, PARP is overexpressed in many malignancies, including oral and oropharyngeal squamous cell carcinoma (8, 9). Several radiolabeled PARP inhibitors were tested in preclinical studies, showing correlation of uptake with PARP1 expression, and suggesting that imaging is possible with little unspecific uptake in healthy head and neck tissue (10, 11).

The purpose of this study was to clinically translate ¹⁸F-PARPi, a PARP inhibitor derived from the core scaffold of olaparib (12), and to provide a first step toward validating the tracer as a clinical tool. Other PARP inhibitor–based imaging agents, based on different core scaffolds, were translated earlier (13, 14). In this first-in-human phase I clinical trial in patients with oral and oropharyngeal cancer, we determined the safety and feasibility of ¹⁸F-PARPi imaging. We also correlated ¹⁸F-PARPi uptake in tumors and normal tissue to standard of care ¹⁸F-FDG imaging.

Further description of the experimental procedures and methods, including radiopharmaceutical preparation, tracer formulation and quality control can be found in the Supplementary Materials.

This exploratory, phase I, single-center, open-label, prospective Health Insurance Portability and Accountability Act (HIPAA) compliant study was approved by the Memorial Sloan Kettering (MSKCC) Institutional Review Board and conducted in accordance with the Declaration of Helsinki (Clinicaltrials.gov NCT03631017). Written informed consent was obtained from all patients. The primary objectives of this phase I trial were to evaluate the safety and feasibility of ¹⁸F-PARPi. We report the biodistribution and radiation dosimetry, and describe the tumor uptake of ¹⁸F-PARPi compared with ¹⁸F-FDG. Patients were accrued between January 2019 and September 2019 and referred to the MSK Molecular Imaging and Therapy Service for their newly diagnosed or recurrent oral or oropharyngeal cancer. Twelve patients were enrolled on this study protocol. Of these, 11 patients completed the study (8 patients with oropharyngeal squamous cell carcinoma and 3 patients with oral cavity squamous cell carcinoma). One patient withdrew consent before administration of ¹⁸F-PARPi. No patients were excluded from the analysis. Inclusion and exclusion criteria are summarized in Supplementary Table S1.

All patients underwent clinical examination, baseline vital signs, pulse oximetry, ECG, and blood tests (<2 weeks prior to imaging). No fasting was required prior to ¹⁸F-PARPi imaging. On the day of imaging, two intravenous catheters were placed in each forearm, one for injection of ¹⁸F-PARPi and one for drawing of IV blood samples. ¹⁸F-PARPi was administered to patients at an average activity of 333 ± 44 MBq (9.0 ± 1.2 mCi) by intravenous bolus injection. For the first 6 patients, a dynamic PET scan (with the field of view including the heart, lungs, liver, and kidneys) was acquired to study the biodistribution and clearance of the tracer. For the subsequent 5 patients, the dynamic PET scan was centered on the head and neck region. In all instances, the CT component of the dynamic PET/CT was acquired with a tube current of 40 mA. Immediately after the dynamic study, a static PET/CT scan (extending from skull vertex to upper thighs in 6 patients, and over the head and neck region in 5 patients) was obtained (30 minutes postinjection; CT tube current: 80 mA). Two further static PET/CT scans were taken at approximately 60 and 120 minutes postinjection (CT tube current: 10 mA).

A total of 5 blood samples (at approximately 1, 5, 30, 90, and 150 minutes postinjection) were drawn to quantify blood pool activity and to study ¹⁸F-PARPi metabolites. After imaging, vital signs were obtained and an ECG performed, and blood samples were collected for hematology and blood chemistry analysis. A follow-up phone interview (1–3 days after the imaging study) was conducted to document any side effects occurring after completion of the imaging study.

Paraffin-embedded slides from surgical specimen or core biopsies of the primary tumor were obtained from consented patients and processed at the molecular cytology core facility at MSKCC (New York, NY). PARP1 IHC was performed using the Discovery XT processor (Ventana Medical Systems). The anti-PARP1 rabbit mAb (46D11, Cell Signaling Technology) specifically bound both human PARP1 (0.4 μg/mL). Paraffin-embedded formalin fixed 3-μm sections were deparaffinized with EZPrep buffer, antigen retrieval was performed with CC1 buffer (both Ventana Medical Systems), and sections were blocked for 30 minutes with Background Buster solution (Innovex). Anti-PARP1 antibody was incubated for 5 hours, followed by 1 hour of incubation with biotinylated goat anti-rabbit IgG (PK6106, Vector Laboratories) at a 1:200 dilution. For IHC detection, a DAB detection kit (Ventana Medical Systems) was used according to the manufacturer's instructions, sections were counterstained with hematoxylin and cover-slipped with Permount (Thermo Fisher Scientific). Incubating with a rabbit IgG instead of the primary antibody controlled for nonspecific binding of the secondary antibody. Slides were scanned (Mirax, 3DHISTECH) to allow for digital histologic correlation. Hematoxylin and eosin (H&E)-stained slides were used to determine areas of tumor and areas of normal muscle. Those exact same areas were used for PARP1 quantification using a consecutive slide.

Blood clearance measurements were performed as previously reported (15). Briefly, multiple venous blood samples were obtained between 1 and 150 minutes after intravenous injection of ¹⁸F-PARPi. Activity in whole blood and plasma was measured in duplicate using a calibrated NaI (Tl) Wallac Wizard 2480 automatic γ-counter (Perkin Elmer, Inc.). The measured activity concentrations were converted to percentage injected activity per kilogram (%ID/kg). Metabolite analysis of activity in plasma was performed by reverse-phase HPLC with in-line radiation (Posi-RAM model 4, LabLogic) detection using a Kinetex Biphenyl column (Phenomenex, 150 × 4.6 mm; 5 μm particle size) and a mobile phase gradient of 10%–75% acetonitrile (0.1% TFA) in water (0.1% TFA) over 20 minutes. Intact ¹⁸F-PARPi elutes at 16 minutes and a number of metabolites elute from 7–8 minutes.

Absorbed radiation doses to normal tissues were estimated for the first 6 patients of the study, based on dynamic and static PET/CT images. Activity concentration–time curves were generated by analysis of VOIs generated for liver, kidney, spleen, cardiac blood pool, bone, lung, gallbladder and urinary bladder. Red marrow activity concentration was assumed equal to that of blood. Whole-body time–activity–time curves, generated using the 4 points defined by the administered activity (time zero) and the total activities in the three whole-body PET scans, were used to calculate monoexponential clearance half-times. The area under activity concentration–time curves (AUC) were estimated by trapezoidal integration with a terminal contribution calculated by extrapolation from the last measured value using the shorter of apparent terminal clearance rate or physical decay. Whole-organ AUCs were obtained by multiplying the activity concentration AUC by organ mass. Baseline values of organ mass were taken from the Oak Ridge National Laboratory (ORNL) phantoms of OLINDA/EXM 2.0 (Hermes Medical Solutions) representing standard human. Organ masses were rescaled if body mass differed by more than 15% from the standard value (73.7 kg for males; 56.9 kg for females). Organ residence times were derived by dividing organ AUC values by administered activity. For urinary bladder contents, residence times were estimated by the OLINDA/EXM 2.0 voiding bladder model based on the fraction of activity clearing via urinary bladder, the monoexponential whole-body biological half-time, and an assumed voiding interval of 1 hour. Residence times for the remainder of body were derived by subtracting all the individually estimated residence times from the whole-body residence time. Absorbed radiation doses to the whole body and various organs were calculated using OLINDA/EXM 2.0 with effective doses based on the tissue weighting factors of ICRP Report 103 (16).

All PET/CT images were obtained on a Discovery 710 PET/CT scanner (GE Healthcare), using low dose CT settings (10–80 mA, 120 kV) for CT images that were used for attenuation correction and anatomic correlation. All studies were reviewed using the Hybrid Viewer display and analysis application (Hermes Medical Solutions). ¹⁸F-PARPi PET/CT and ¹⁸F-FDG PET/CT studies were interpreted by two nuclear medicine physicians with at least 10 years of PET/CT experience. Three-dimensional threshold-based volumetric regions of interest (VOI) were placed in reference regions [bilateral submandibular gland, parotid gland, blood pool of neck, contralateral posterior neck muscles (trapezius and semispinalis), genioglossus muscle, bone marrow, mediastinal blood pool, myocardium, normal liver, renal cortex, and spleen] and over all sites of abnormal uptake in lymph nodes, or soft-tissue lesions with reference to the PET/CT images. Abnormal ¹⁸F-PARPi and ¹⁸F-FDG uptake was defined as outside physiologic sites (such as palatine tonsils or skeletal muscle) and of intensity greater than regional background. Uptake of ¹⁸F-PARPi in the soft-tissue lesions and lymph nodes was assessed by measuring the maximum standardized uptake values (SUV_max).

Uptake values are presented as mean ± SE, unless otherwise specified. Distribution of uptake with ¹⁸F-PARPi and ¹⁸F-FDG (on lesions where both were available) were compared using a Wilcoxon Signed Rank test for paired data, while their variances were compared using a Levene test. R version 3.6.0 was used for analysis.

A total of 11 patients with cytologically or histologically confirmed squamous cell carcinoma of the oral cavity or oropharynx completed the study protocol. Supplementary Table S2 lists the patient demographics and key diagnostic parameters, including stage and lymph node status, and Fig. 1 shows a schematic overview of the study workflow. There were 10 males and one female, with a mean age of 64 years. At the time of imaging, 8 patients had newly diagnosed disease (all with the primary lesion in the oropharynx). The 3 patients that had recurrent disease at the time of imaging had previously been diagnosed with oral cavity squamous cell carcinoma. One patient had their primary lesion surgically removed before imaging. Nodal involvement (anatomically abnormal lymph nodes) was present in 9 of the 11 patients, and disease stage ranged from I to IVb (8th edition, AJCC). Lymph nodes were considered to be abnormal when at least one of following criteria was met: central necrosis or inhomogeneous enhancement (contrast-enhanced CT and /or MRI), shortest axial diameter greater than 11 mm in cervical regions and change in shape and/or ill-defined irregular margins in a lymph node. Twenty-seven percent of patients (n = 3) were HPV-negative (all oral cavity squamous cell carcinoma) and 73% were HPV-positive (all oropharyngeal cases).

The radiochemical synthesis of ¹⁸F-PARPi was performed using a synthetic route similar to what we have reported before (Supplementary Fig. S1; ref. 12). We first synthesized para-[¹⁸F]fluoro-benzoic acid as a radiolabeled synthon before conjugation with the 1(2H)phthalazinone targeting group. On average, patients were therefore injected with 290 pmol of ¹⁸F-PARPi, 6.7 orders of magnitude lower than the twice daily administered dose of olaparib (2 × 300 mg) and therefore unlikely to elicit a pharmacodynamic response.

All patients tolerated the injection of ¹⁸F-PARPi well, and no adverse events were recorded related to the ¹⁸F-PARPi injection. One patient died within a 2-week window after completing the study, and one patient experienced grade 1 mucositis over the tumor site, which resolved the following day. The death was considered unrelated to the administered drug. The mucositis was considered possible related to the administered drug. For all patients, clinical chemistry and hematology were determined and an electrocardiogram was performed before and after the administration of ¹⁸F-PARPi (Supplementary Table S3). While some patients presented with abnormal findings before imaging, no clinically relevant changes were observed after radiotracer injection or at follow-up.

Maximum intensity projection PET images of a representative patient injected with ¹⁸F-PARPi (images obtained at 30, 60, and 120 minutes postinjection) are shown in Fig. 2A. At 120 minutes, uptake in the primary tumor had an SUV_max of 4.1, whereas the metastatic lymph node had an SUV_max of 3.6 (radiologic lymph node level 2). Across the entire patient population, the average primary tumor SUV_max was 2.8 ± 1.1 at 120 minutes. The primary routes of excretion were renal and hepatobiliary with most of the tracer excreted renally. Activity in the renal cortex diminished over time (SUV_max = 16 ± 8 at 30 minutes, 9 ± 4 at 60 minutes, and 7 ± 5 at 120 minutes) with commensurate accumulation in the urinary bladder. The maximal observed activity in the urinary bladder corresponded to 20%–38% of the total administered, typically at the 30–60 minute scan times.

Absorbed radiation doses to normal tissues from ¹⁸F-PARPi were estimated on the basis of the tracer biodistribution of the first six patients. Key dosimetry data are plotted in Fig. 2B, and the entire dataset can be found in Supplementary Table S4. The effective dose of ¹⁸F-PARPi was 0.014 ± 0.002 mSv/MBq, calculated with ICRP103. In a typical diagnostic setting, the effective radiation dose is projected to be 3.9 mSv–5.2 mSv.

PARP1 was highly expressed in both oral and oropharyngeal squamous cell carcinoma in our patient population (Fig. 2C and D; Supplementary Fig. S2), corroborating earlier work (8, 9).

Research blood draws were obtained for 10 patients at five timepoints after tracer injection, activity counted, and metabolites analyzed (Supplementary Fig. S3A). Using a two-phase decay curve, we determined the weighted blood half-life to be 4.2 minutes (whole blood, Supplementary Fig. S3B). Only small quantities of metabolites were detected at 1 and 5 minutes (99.2% ± 1.5% and 89.9% ± 11.4% ¹⁸F-PARPi, respectively, Supplementary Fig. S3C). At 30 minutes, and with decreasing blood pool concentration of the injected tracer, we detected a radiometabolite with a retention time of 7–8 minutes (50.9% ± 11.5%; Supplementary Fig. S4).

Investigating the specificity of ¹⁸F-PARPi, we looked at the residence times of the tracer in tumors, metastatic nodes, and healthy tissues (Fig. 3). In spleen and liver, which both express large physiologic amounts of PARP1 and PARP2 (17), initial uptake was high, followed by rapid clearance over the 2-hour imaging period [SUV_max(spleen, 30 minutes) = 6.1 ± 1.3 and SUV_max(spleen, 120 min) = 2.2 ± 0.6, representing a 64% drop]. Similarly, the SUV_max in bone marrow was high in the 30-minute PET/CT scan, but values declined by 53% between 30 and 120 minutes. Comparably, fast clearance was found for physiologic structures within the head and neck region. Uptake in the submandibular and parotid glands decreased by 57% and 56%, respectively. In contrast, tracer retention in tumor and metastatic nodes was significantly longer, with SUV_max values declining by just 13% for both primary and PET-avid lymph nodes [SUV_max(tumor, 30 minutes) = 3.4 ± 0.8 and SUV_max(tumor, 120 minutes) = 3.0 ± 1.1; SUV_max(lymph node, 30 minutes) = 3.3 ± 1.3 and SUV_max (lymph node, 120 minutes) = 2.9 ± 1.1].

To determine whether ¹⁸F-PARPi could be a relevant imaging tracer for the head and neck region, we compared its retention with that of standard-of-care ¹⁸F-FDG (Fig. 4). For both imaging agents, uptake was corroborated with tumor outlines defined by standard-of-care T₁-weighted Gd-MRI imaging (Fig. 4C). Across the patient population, ¹⁸F-FDG had higher average tumor SUV_max values than ¹⁸F-PARPi, but SUV_max values for ¹⁸F-FDG decreased for level 2 and level 3 lymph nodes. A smaller decrease in SUV_max values was found for ¹⁸F-PARPi (66% and 15% for ¹⁸F-PARPi and ¹⁸F-FDG, respectively, when grouping primary/level 1 lymph nodes and level 2/level 3 lymph nodes; Fig. 4D). Uptake ratios [SUV_max(lesion)/SUV_max(trapezius muscle)] for ¹⁸F-FDG were higher than for ¹⁸F-PARPi (median = 10.4 vs. 4.5, P < 0.0001, Fig. 4E). Interestingly, when comparing uptake ratios [SUV_max(lesion)/SUV_max(genioglossus muscle)], we found similar median values for ¹⁸F-FDG and ¹⁸F-PARPi (median = 3.0 vs. 3.3, P = 0.23), although the variance was less for ¹⁸F-PARPi than for ¹⁸F-FDG (1.3 vs. 6.0, P = 0.001, Fig. 4F). This could be of potential relevance in patients with recurrent oral cancers or in the posttreatment setting, leading to asymmetric uptake in the oral cavity or higher uptake in the genioglossus muscle secondary to partial glossectomy, tongue movement, or hypoglossal palsy (18–20).

¹⁸F-FDG and ¹⁸F-PARPi uptake matched with respect to the presence and location of the primary lesion. However, the two tracers had divergent uptake patterns in the lymphatic system (Supplementary Fig. S5). For ¹⁸F-FDG, 34 lymph nodes were PET-avid. For ¹⁸F-PARPi, we observed 40 ¹⁸F-PARPi avid lymph nodes. These 40 lymph nodes included all of the lymph nodes that were detected using ¹⁸F-FDG. No evidence of additional FDG-avid and PARPi-avid adenopathy or distant metastatic disease was seen. Because of protocol regulations, no biopsy material was available for the additional 6 lymph nodes; however, a subset of them resolved after chemoradiation (which, however, does not prove malignancy, Supplementary Fig. S5A and S5B).

Recently, PARP-targeted agents have received considerable attention as imaging agents (10, 21) based on the ubiquitous expression of PARP in many types of cancers, with the promise to serve as an accurate sensor of malignancy where standard-of-care methods currently fail.

Radiation doses associated with ¹⁸F-PARPi PET imaging were relatively low. The overall equivalent dose was 3.9 mSv–5.2 mSv, lower than that reported for ¹⁸F-FDG (8.1 ± 1.2 mSv; ref. 22). This was, in part, due to the selective uptake of the tracer, paired with a short blood half-life and fast clearance. ¹⁸F-PARPi rapidly cleared from the circulatory system and accumulated in the urinary bladder and gallbladder following respective transit through the kidneys and liver.

PARP1 is located in the nuclei of cells. Consequently, the cell membrane and nuclear membrane permeability of ¹⁸F-PARPi has to be high, allowing both fast uptake of radiotracer and clearance of unbound material for image contrast generation. This was seen in the submandibular glands, where the SUV_max at 20–30 minutes was higher than in the primary tumor for 80% (n = 10 evaluable patients) of all patients (Fig. 3). Subsequently, the delivered activity in these healthy tissues cleared quickly, whereas the activity persisted longer in tumor and metastatic nodes (Fig. 3B). At 120 minutes postinjection the situation had inverted, and none of the submandibular glands showed activity higher than the primary tumor. Similar rapid clearance of radiotracer was noted in other healthy organs, including the spleen, genioglossus muscles, parotid glands, and bone marrow. For future studies, imaging with PARP inhibitors after longer time intervals postinjection might yield further improved contrast ratios.

Because this is a phase I study, we were unable to collect unstained histologic slides (for primary and metastatic lesions) from a statistically meaningful number of patients. Consequently, the focus of an upcoming phase II study will be to determine the sensitivity and specificity of ¹⁸F-PARPi in head and neck cancer.

In conclusion, we performed the first-in-human translation of ¹⁸F-PARPi, a PARP-targeted 1(2H)phthalazinone, and the first imaging of PARP in head and neck cancer. Administration of ¹⁸F-PARPi was safe, and aside from a grade 1 mucositis, which was possibly related, no adverse events were attributed to the tracer injection. ¹⁸F-PARPi is a promising new agent for the imaging of head and neck squamous cell carcinoma.

J.S. Lewis reports holding intellectual property interests related to Theragnostics. J.A. O'Donoghue is an employee/paid consultant for Janssen Research & Development, LLC. S.G. Patel is an advisory board member/unpaid consultant for Summit Biomedical Imaging LLC. T. Reiner is an employee/paid consultant for and reports receiving commercial research grants from and holding intellectual property interests in Theragnostics; T. Reiner also reports other commercial research support from Pfizer, Immunogen, and Summit Biomedical Imaging, and holds ownership interest (including patents) in and reports receiving other remuneration from Summit Biomedical Imaging. N.Y. Lee is an advisory board member at Merck, Merck Serono, Pfizer, and Eli Lilly and receives commercial research support from Astra-Zeneca and Pfizer. J.S. Lewis and T. Reiner report that MSK has institutional financial interests in Theragnostics related to licensed technology. No potential conflicts of interest were disclosed by the other authors.

Conception and design: H.M. Schöder, S.K. Lyashchenko, N.Y. Lee, T. Reiner

Development of methodology: E.M. Burnazi, C. Brand, S.K. Lyashchenko, J.S. Lewis, J.A. O'Donoghue, S.G. Patel, T. Reiner

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): H.M. Schöder, P.D.D.S. França, E.M. Burnazi, S. Roberts, C. Brand, M. Grkovski, J.A. O'Donoghue, I. Ganly, S.G. Patel, N.Y. Lee

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): H.M. Schöder, P.D.D.S. França, R. Nakajima, E.M. Burnazi, M. Grkovski, A. Mauguen, R. Ghossein, J.A. O'Donoghue, T. Reiner

Writing, review, and/or revision of the manuscript: H.M. Schöder, P.D.D.S. França, R. Nakajima, E.M. Burnazi, M. Grkovski, A. Mauguen, M.P.S. Dunphy, R. Ghossein, S.K. Lyashchenko, J.S. Lewis, J.A. O'Donoghue, I. Ganly, S.G. Patel, N.Y. Lee, T. Reiner

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): H.M. Schöder, P.D.D.S. França, J.S. Lewis, S.G. Patel, T. Reiner

Study supervision: H.M. Schöder, N.Y. Lee, T. Reiner

This work was supported in part by NIH grants R01 CA204441, R35 CA232130, and P30 CA008748, the Tow Foundation, the MSK Center for Molecular Imaging & Nanotechnology, the MSK Imaging and Radiation Sciences Program, and the MSK Molecularly Targeted Intraoperative Imaging Fund. We acknowledge and thank Stephen Carlin and Kevin Staton for help with clinical radiochemistry; Aisha Shickler and Yorann Roux for blood and metabolite analysis; Ryan Min for patient coordination; Christopher Riedl, MD, PhD, for help with the clinical workflow; and Susanne Kossatz, PhD and Wolfgang A. Weber, MD for helpful discussions and help with preparing the clinical trial.

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.