[¹⁸F]FluorThanatrace ([¹⁸F]FTT) PET Imaging of PARP-Inhibitor Drug-Target Engagement as a Biomarker of Response in Ovarian Cancer, a Pilot Study

This article is featured in Highlights of This Issue, p. 1375

PARP inhibitors (PARPi) are a newer class of agents that are FDA approved for the treatment of high-grade serous ovarian cancer (HGSOC) in the maintenance and recurrent setting (1–5). PARPi selectively inhibit predominantly PARP-1, PARP-2, and less so PARP-3 enzymes, although the efficacy of these drugs is primarily dependent on PARP-1 inhibition (6).

PARPi have shown efficacy in malignancies with preexisting DNA repair defects, most notably tumors that harbor mutations in homologous recombination DNA repair genes, such as Breast Cancer Susceptibility genes 1 and 2 (BRCA1/BRCA2). These homologous recombination deficient (HRD) tumors lacking BRCA1/2 or other HR-related genes (e.g., ATM, RAD51C/D, CHK2) cannot repair double-stranded DNA breaks that result from PARP inhibition, leading to cell death with less impact on healthy tissue not expressing these mutations (7, 8). This phenomenon, deemed “synthetic lethality,” has been leveraged for the selection of patients for treatment with this class of medications. Currently, there are three PARPi with FDA approval in ovarian cancer—olaparib, rucaparib, and niraparib—for patients with known HRD recurrent disease, or for all patients regardless of HRD status as maintenance therapy after response to platinum-based chemotherapy (1, 2, 4, 9–11). The BRCA1/2 mutation serves as a genomic biomarker that predicts sensitivity to PARPi (1, 2, 4), although increasing evidence supports PARPi sensitivity in non-BRCA HRD tumors (12–16).

Despite defined selection criteria, response to PARPi remains variable with not all patients responding to or benefitting from treatment with these agents. For example, in the recurrent setting, overall response rates to olaparib in BRCA mutant (BRCA^MUT) ovarian cancer is 30% to 40% (17, 18). In the maintenance setting, PARPi are used to prevent recurrence and even though a significant benefit has been shown, not all patients benefit (19, 20). Because of suboptimal response, new combination strategies that synergize with PARPi by targeting DNA-damage response mechanisms (e.g., ATR, WEE1, BET), oncogenic pathways (e.g., PI3K/AKT/mTOR and RAS/RAF/MEK), angiogenesis, or by enhancing endogenous immunity are under clinical development (21). Thus, there is a pressing clinical need to develop additional biomarkers to better select patients with ovarian cancer for treatment with both PARPi monotherapy and combination approaches.

[¹⁸F]FluorThanatrace ([¹⁸F]FTT) is a radiopharmaceutical for PET imaging that is an analog of rucaparib, with pharmacokinetic properties and PARP-specific binding properties similar to approved PARP-inhibiting drugs (22–26). As such, [¹⁸F]FTT provides an in vivo radioligand imaging assay of regional PARP expression in tumors (22, 25, 26), predominantly PARP-1 (23). Other PET compounds for PARP imaging are also under preclinical and clinical development (27, 28). We have previously demonstrated that PET uptake of [¹⁸F]FTT correlates with PARP-1 expression in ovarian cancer tumor tissue, validating its biologic target (25). Kinetic analysis of dynamic imaging revealed biologic insight into the radiotracer uptake mechanism and supported a static imaging protocol suitable for widespread clinical use (26). As a noninvasive marker of regional drug-target binding amenable to serial measurements, we expect that this PET radiotracer could serve as a noninvasive imaging biomarker of PARPi drug-target engagement to guide therapy (29).

In most clinical trials investigating new targeted agents, response to therapy is typically assessed anatomically by imaging (e.g., using the RECIST1.1 criteria; ref. 30). An effective therapy should lead to a decrease in size of target lesions, indicative of tumor cell death (30). However, such anatomic changes often lag months behind the molecular response (31), hampering the ability to quickly adapt treatment early in the course of therapy to avoid futility and unnecessary toxicity. By imaging the molecular pharmacodynamic response to therapy—including drug-target engagement or downstream molecular sequelae—treatment effects can be revealed sooner, facilitating new treatment paradigms. We hypothesized that PARP-1 expression, measured prior to PARPi treatment, and PARPi drug-target engagement measured after a short exposure to PARPi, can be leveraged as a biomarker to predict response to PARPi therapies.

Herein, we describe pilot translational studies of [¹⁸F]FTT as a biomarker of response to PARPi in ovarian cancer. We first explore [¹⁸F]FTT in preclinical models of ovarian cancer using patient-derived xenografts (PDX), followed by pilot data from a phase II clinical trial integrated biomarker study in ovarian cancer subjects.

Olaparib (PARPi, AZD2281) and ataxia–telangiectasia inhibitor (ATRi, AZD6738) were provided by AstraZeneca. [¹⁸F]FTT was produced at the Cyclotron Facility at the PET Center of the University of Pennsylvania (UPENN) as described previously (22). Specific activity ranged from 2,170 to 21,325 mCi/μmol, and radiochemical purity was ≥99.7%.

NSG mice (NOD/SCID IL2Rγ^−/−) were purchased from the Stem Cell and Xenograft Core at the UPENN. All mice were housed according to the UPENN Institutional Animal Care and Use Committee. Studies were performed in accordance with approved UPENN IUCAC protocol (#806772). Tumor was obtained from ovarian cancer debulking surgeries conducted at the Hospital of the UPENN and Pennsylvania Hospital (IRB #702679). PDXs were generated by surgically engrafting three to four pieces (2–3 mm each) orthotopically to the mouse fallopian tube/ovary (32). For preclinical studies, cryopreserved tissue was thawed and retransplanted to the fallopian tube/ovary of 6- to 8-week-old female NSG mice for evaluation of in vivo drug response and microPET imaging of PARPi drug-target engagement. For PARPi-resistant PDX models, once tumor volume reached 40 to 50 mm³, animals were pretreated with olaparib (75 mg/kg/day, oral gavage 6 days weekly) and after a 2-fold increase in volume (confirming PARPi resistance), were randomized to treatment arms. For the microPET imaging cohort, a 7-day PARPi washout period occurred prior to continuation with the study.

We evaluated the efficacy of olaparib and AZD6738 (ATRi), as single-agents and in combination in a placebo-controlled study. Alongside preclinical efficacy studies, we performed serial [¹⁸F]FTT microPET imaging in a cohort of mice with identical treatment groups. Three PDX models were evaluated to compare a PARPi-sensitive model (WO-3, germline BRCA2^MUT [gBRCA2^MUT]) to two PARPi-resistant models (WO-33 [gBRCA1^MUT]; WO-58 [gBRCA1^MUT, CCNE1 amplified copy number 7, efficacy studies reported previously; ref. 33]). Once tumor volumes reached 70 to 100 mm³, animals were randomized to treatment groups. PARPi-resistant models WO-33 and WO-58 were treated at 75 mg/kg olaparib D1–6, 50 mg/kg ATRi D1–5, or combination, whereas WO-3 was treated at 50 mg/kg olaparib D1–6, 25 mg/kg ATRi D1–5, or combination. Body Condition scores evaluating activity level, fur coat, body condition, appearance with score from 1 to 5 (5 being the healthiest) were used to assess for drug toxicity. Weekly ultrasonographic measurements (SonoSite Edge II Ultrasound System), weight assessments, and condition scores were obtained for treatment groups in a blinded manner. Mice in the efficacy cohorts were followed up to 300 days or removed from study due to tumor progression. For the imaging cohort, [¹⁸F]FTT microPET was performed at baseline (prior to treatment) and to measure drug-target engagement 7 days from the initiation of therapy, which corresponds to treatment steady state. Primary measures of response were assessed as the percent change in tumor volume from the first microPET study.

Mouse imaging was performed on X-Cube micro-CT and β-Cube PET scanners (Molecubes). The X-Cube micro-CT was used for both anatomic localization of tumors (voxel resolution of up to 0.05 mm) and for PET attenuation and scatter corrections (34). The β-Cube PET scanner has a 130 mm × 72 mm field of view, spatial resolution of 1 mm, and absolute peak sensitivity of 12.4%. PET images were acquired for 10 minutes starting 52.5 ± 6.0 minutes (±SD) post-injection of 7.26 ± 1.06 MBq of radiotracer (n = 91 including 47 baseline and 44 posttreatment PET scans). PET images were reconstructed via 3D maximum likelihood expectation maximization (MLEM) reconstruction with 20 iterations and 30 subsets with decay, scatter, and dead time corrections to generate cubic voxels with size 0.4 mm × 0.4 mm × 0.4 mm (34).

PMOD image analysis was used for WO-58 (n = 20), WO-3 (12 baseline and 10 posttreatment), and WO-33 mice (n = 15 baseline and 14 posttreatment; PMOD version 3.7, PMOD Technologies Ltd.). Radiotracer uptake by the tumor, reflected by standardized uptake value peak (SUV_peak), was quantified by calculating the average activity concentration in the five highest activity concentration voxels in a hand-drawn tumor volume of interest (VOI). A 5-mm diameter spherical VOI was drawn on normal hind leg muscle on the opposite side of the tumor to calculate tumor-to-background-muscle [¹⁸F]FTT uptake ratios (T/M). T/M (vs. SUV) was chosen for the mouse models as it is more robust to variability from mouse injections and has been shown to correlate with SUV results in subjects (26).

To infer changes in [¹⁸F]FTT uptake seen with drug treatment in the PDX models, similar human cell lines were chosen for in vitro assays of PARP-1 expression levels pre- and posttherapy. JHOS4 (BRCA1^MUT) ovarian cancer cells, representative of PARPi-sensitive tumors, were grown in DMEM/F12 media with 10% FBS, and penicillin/streptomycin. PEO1 PARPi-resistant ovarian cancer cells (PEO1 PR, BRCA2^MUT) were developed by long-term drug exposure (>12 months; 3 μmol/L of olaparib) and grown in RPMI media with 10% FBS, and penicillin/streptomycin supplemented with olaparib. Prior to in vitro studies, PEO1 PR cells were grown without olaparib for 14 days for drug washout. JOSH4 cells and parental PEO1 cells were a generous gift from Dr. Ronny Drapkin, University of Pennsylvania and Dr. Andrew Godwin, University of Kansas, respectively. Cell lines were authenticated by short tandem repeats at the Oncogenomics Core (Wistar Institute) and confirmed mycoplasma negative by end-point PCR (Cell Center Service at UPENN).

Ovarian cancer JHOS4 and PARPi-resistant PEO1 PR cells and select PDX tumors were evaluated by reverse phase protein array (RPPA) analysis. Cells were seeded and treated with 1 μmol/L PARPi, 1 μmol/L ATRi, or both for 72 hours. The cells were collected and lysed with SDS buffer prior to analysis. To infer PARP-1 expression changes in vivo, PARPi-resistant WO-57 [with a BRCA1 reversion mutation (BRCA1^REV)] PDX tumors harvested and banked from a previously reported preclinical study testing the efficacy of 75 mg/kg PARPi or 40 mg/kg ATRi alone or in combination were also assessed (33). Control and combination treatment samples were collected via biopsy 2 weeks posttreatment initiation. Tumors from control, monotherapy, and combination groups were collected at the survival endpoint. Samples were analyzed by the RPPA platform at the MD Anderson Center RPPA core facility as described (35). Antibodies targeting >300 proteins were included in this assay (see for detailed methods and antibody descriptions: http://www.mdanderson.org/education-and-research/resources-for-professionals/scientific-resources/core-facilities-and-services/functional-proteomics-rppa-core/index.html). The results were reported with normalized data at both linear and Log₂ version.

PDX tumors were fixed in 10% buffered formalin then embedded in paraffin. Paraffin blocks were cut into 4- to 6-μm sections and placed onto slides. After deparaffinization and rehydration, antigen retrieval using citrate buffer was performed under pressure. Anti-PARP-1 antibody (RRID:AB_2679240) staining was optimized and set to a titer of 1:600. Slides were incubated with anti-rabbit HRP and developed using 3,3′-diaminobenzidine (DAB) and chromogen. H&E was performed on serial sections using the DAKO CoverStainer (Agilent). Sides were imaged at 20× using a Leica DM 2000 microscope.

Longitudinal analysis of PDX tumor growth was performed by linear mixed-effect modeling on log pre-processed tumor volumes followed by log transformation using the TumGrowth web tool (https://kroemerlab.shinyapps.io/TumGrowth/; ref. 36). [¹⁸F]FTT T/M uptake ratios and percent change posttreatment were calculated for statistical comparisons. Two-tailed Spearman correlation of the percent change in tumor volume compared with the percent change in T/M was performed. Change in T/M across all models were tested for differences between groups using a one-way ANOVA test with a Tukey post hoc test, or Fishers least significant difference test (when specified). In vitro data were analyzed by ANOVA and Turkey post hoc test or two-tailed Mann-Whitney test when appropriate. Statistical significance was set at *<0.05, **<0.01, ***<0.001, and ****<0.0001 (GraphPad Prism v 9.1, RRID:SCR_002798).

This single-center, prospective, cohort study was approved by the UPENN Institutional Review Board (IRB #833685) and registered with the FDA as an expanded investigational new drug (clinical trial identifier: NCT02637934). All subjects provided written informed consent and studies were conducted in accordance with the Declaration of Helsinki statement on ethical biomedical research and the U.S. Common Rule. Women >18 years of age with known or suspected recurrent epithelial ovarian, fallopian tube, or primary peritoneal cancer were eligible for this study. Patients with HGSOC either with germline or somatic mutations in HR genes, or HRD positive tumors by standard CLIA approved assay (e.g., Myriad MyChoice CDx) were selected for enrollment. A baseline [¹⁸F]FTT-PET scan prior to starting therapy (e.g., a PARPi) was required. A second [¹⁸F]FTT-PET scan 1–21 days after initiation of therapy was requested.

Subjects were recruited through the Division of Gynecology Oncology at UPENN, with particular attention to recruiting subjects enrolled in two trials evaluating PARPi in ovarian cancer: the LIGHT trial and the CAPRI trial, or off trial with plan for treatment with a PARPi. The phase II LIGHT trial (NCT02983799) evaluated olaparib treatment in subjects with platinum-sensitive relapsed ovarian cancer stratified by BRCA1/2^MUT and HRD tumor status (37). The phase II CAPRI trial (NCT03462342) evaluated the combination of olaparib with an ATRi (AZD6738) in subjects with recurrent ovarian cancer, including platinum-sensitive, platinum-resistant, and platinum sensitive and PARPi resistant cohorts (38). Off trial patients with germline BRCA1/2 mutations or had tumors positive for HRD (e.g., Myriad My Choice CDx), who were planning to receive treatment with a PARPi treatment, were offered participation in the imaging clinical trial. For subjects enrolled on the [¹⁸F]FTT imaging and CAPRI trials, posttherapy imaging (if obtained) was planned after the initiation of olaparib (5–7 days), but before the ATRi was initiated. Subjects were also recruited for the [¹⁸F]FTT trial, who would be treated with a PARPi as standard-of-care for recurrent ovarian cancer outside of a clinical trial.

Baseline demographic and tumor characteristic data were collected, including germline/somatic alterations (e.g., BRCA1, BRCA2, and other homologous recombination repair gene defects). Progression-free survival (PFS) was calculated at the time from starting a PARPi to the time treatment stopped secondary to progression or death, or based on CA-125 levels, as determined by subjects treating physicians.

Subjects were imaged with [¹⁸F]FTT as described previously (25, 26). Briefly, images were obtained from the skull base to the thighs on an Ingenuity TF scanner (Philips Healthcare) for 3 minutes/bed position for a total imaging time of 20 to 25 minutes, 60 to 90 minutes after intravenous injection with 8 to 12 mCi [¹⁸F]FTT. [¹⁸F]FTT-PET images were analyzed with MIM v7.1 (MIM Software Inc.) in a blinded manner by a board-certified nuclear medicine physician (ARP) and confirmed by a nuclear radiology fellow (SML). SUV_max data, normalized by body weight, were collected for lesions (primary tumors and/or metastases) identified as target lesions by the RECISTv1.1 criteria (30), described below. Only static imaging was used for this analysis, and no correction was made for partial volume correction, noting that SUV_max values at 60 minutes post-injection has been validated as a measure of PARP-1 expression (26). For subjects with follow-up PET imaging, percent change [¹⁸F]FTT uptake was calculated as the percent difference between the sum of SUV_max on follow up compared with the sum at baseline (△SUV), akin to RECISTv1.1 calculations (30).

Anatomic images were evaluated by a board-certified radiologist (DAT) using RECISTv1.1 criteria, blinded to [¹⁸F]FTT results. Target lesions (a maximum five total and a maximum two per organ) were selected on pretreatment studies among all lesions present based on a (i) RECISTv1.1 size criteria for measurability (≥15 mm in short-axis for nodal lesions and ≥10 mm in long-axis for nonnodal lesions), as well as with preference towards lesions with (ii) larger size, (iii) noncystic or necrotic appearance, (iv) well-defined boundaries such that accurate measurement at baseline and on follow-up assessments could be ensured as much as possible, and (v) representative sampling distribution among all involved organs/tissues. The changes in the sum of lesion diameters were used to provide an indication of index lesion response to which [¹⁸F]FTT-PET results were compared. Best overall response (target plus nontarget tumor burden) was recorded for each subject.

Two-tailed Spearman correlation comparing percent change of SUV_max from baseline (△SUV) measurements, percent change in tumor size (target lesion summed diameter) from CT atomic imaging, PFS, and percent change in CA-125 (baseline to nadir). For PFS, CA-125, and best RECIST response correlation analysis, the two subjects currently active on trial were censored and removed as data points. Blue lines in graphs show the best fit line (linear regression). As seven subjects were off trial, investigator determined PFS was used for analysis unless specified. Two-tailed Mann-Whitney test was performed comparing the PFS, change in CA-125 or change in anatomic imaging to the △SUV at ≤ or >50% reduction in radiotracer uptake between scans.

The data generated in this study are mostly available within the article and its supplemental files. Please contact the corresponding author for additional data requests.

PDX studies were used to (i) investigate changes in [¹⁸F]FTT uptake in PDX tumors after treatment with PARPi, ATRi, or combination and (ii) serve as a mechanistic basis for interpreting clinical [¹⁸F]FTT-PET studies. Study design for [¹⁸F]FTT-PET-PDX experiments is depicted in Fig. 1A. For the PARPi-resistant, BRCA1^MUT WO-33 model, in vivo efficacy studies showed the best response in the combination treatment group (75 mg/kg olaparib + 50 mg/kg ATRi) compared with monotherapy (P ≤ 0.0002, Fig. 1B, left). Previously, we reported findings for the PARPi-resistant, BRCA1^MUT/CCNE^AMP WO-58 model which showed modest activity with ATRi monotherapy and similar synergistic tumor regression with combination PARPi+ATRi treatment (33). The PARPi-sensitive, BRCA1^MUT WO-3 model (PARPi treatment naïve model) had a significant reduction in tumor volume with 50 mg/kg olaparib alone and more so when in combination with 25 mg/kg ATRi compared with control and ATRi monotherapy (P < 0.0001, Fig. 1B, right). Also, there was improvement in duration of tumor response for combination therapy compared with olaparib alone (P = 0.013), but there was no significant difference in size between these groups up to 20 weeks, times at which the combination showed clear benefit for the PARPi-resistant PDX models.

PET imaging was performed to assess the impact of PARPi, ATRi, and a combination of both drugs on [¹⁸F]FTT uptake as an indication of drug-target engagement and possible predictor for tumor response. Once tumors had grown to about 70 mm³, PDX models underwent drug wash out (for PARPi-resistant models which were treated with PARPi to confirm resistance), pretherapy PET/CT imaging, and were then randomized to treatment arms. All three PDX models showed PARP-1 expression pretherapy (Supplementary Fig. S1A), allowing for assessment of drug-target engagement with [¹⁸F]FTT. In the PET-imaged cohort, the best response for the PARPi-resistant WO-33 and WO-58 models was observed in the combination treatment arm, similar to the efficacy studies; whereas no difference was observed in PARPi-sensitive WO-3. PET imaging with [¹⁸F]FTT and representative pre- and posttreatment images are shown (Fig. 1C). [¹⁸F]FTT tumor/muscle uptake ratios (T/M) were calculated for each animal to normalize for nonspecific uptake, and the % decline in T/M pretherapy and after 7 days of the selected treatment regimen was compared with % change in tumor volume at the same time. Of note, there was no significant antitumor effects seen by any treatment in all groups at this early time point. The majority of tumors actually increased in volume during the treatment week (Supplementary Fig. S1B). In addition, RPPA analysis suggests that there was no significant upregulation in apoptosis markers such as cleaved caspase 3, 7, or 8 in a PARPi-resistant PDX model (WO-57 BRCA1^REV) after 2 weeks of combination treatment (Supplementary Fig. S1C), indicating that 7-days posttreatment is an appropriate timepoint to measure drug-target engagement with minimal cytotoxic drug effect. The two PARPi-resistant PDX models showed a correlation been % decline in [¹⁸F]FTT T/M and % change in tumor size (WO-33, r = 0.77, P = 0.103; WO-58, r = 0.81, P = 0.022) when analyzed by PDX model (Fig. 1D). A greater decline in the [¹⁸F]FTT T/M posttherapy correlated with greater abrogation of tumor growth. The WO-3 model, a PARPi treatment naïve model, which was especially sensitive to olaparib treatment alone, did not show this correlation (r = −0.018, P = 0.99), indicating that that the change in [¹⁸F]FTT T/M with treatment was not simply a result of the impact of the change in tumor size. Lack of correlation in this model was likely secondary to the marked sensitivity of PARPi-monotherapy without an additive or synergistic effect of ATRi treatment.

To better understand the treatment-related changes in [¹⁸F]FTT, we compared decreases in [¹⁸F]FTT T/M across all models grouped by treatment type (Fig. 1E). Percent changes in [¹⁸F]FTT T/M after olaparib treatment were significantly different across all models (P < 0.0001, ANOVA). Post hoc analysis revealed significant reduction in [¹⁸F]FTT T/M in PARPi monotherapy and combination PARPi+ARTi compared with control (P < 0.005 and P < 0.0001, respectively). The improvement in drug-target engagement with the addition of ATRi to PARPi compared with PARPi monotherapy approached significance with correction for multiple comparisons (P = 0.09). Without correction for multiple comparisons, this comparison was significant (P = 0.02, uncorrected Fisher LSD), suggesting added benefit for these two agents given no significant benefit of ATRi or PARPi alone compared with control in PARPi-resistant models, as studied previously (33). We note, though, that these experiments were not powered to detect all differences between treatment pairs. Thus, although the data are limited by sample size, there appears to be an association between the impact of therapy on [¹⁸F]FTT and efficacy in controlling tumor growth, particularly with combination therapy.

Given [¹⁸F]FTT changes with combined DNA damage response (DDR) inhibitor (PARPi-ATRi) treatment, we sought to better characterize the nature of the changes in [¹⁸F]FTT uptake before and after treatment in the combination arm compared with olaparib control. We evaluated PARP-1 expression by RPPA using two human ovarian cancer cell lines: PARPi-sensitive JHOS4 (BRCA2^MUT) and PARPi-resistant PEO1 PR (BRCA1^MUT) cells, and tumors collected at the survival endpoint of a preclinical efficacy study (PARPi-resistant WO-57, BRCA1^REV). PARPi treatment did not lead to a decrease in PARP-1 expression in any of the models. Only WO-57 tumors showed a significant decline with ATRi, alone (P = 0.0019) and in combination with olaparib (P = 0.01), highlighting the effects on PARP-1 expression in vivo after therapy (Fig. 1F). However, a trending decrease in PARP-1 expression was observed in PEO1 PR and JHOS4 cells with combination PARPi-ATRi treatment after 72 hours. When models were grouped together, there was a clearly significant difference in the impact of PARP-1 expression for PARPi monotherapy (no-change) and combination PARPi-ATRi therapy (significant reduction in PARP-1 expression, P = 0.0035; Fig. 1G). Comparing the changes in [¹⁸F]FTT uptake in vivo and PARP-1 expression for PARPi alone versus PARPi-ATRi combination therapy suggests that the decline in [¹⁸F]FTT uptake represents the combined effect on [¹⁸F]FTT binding by competition for the PARPi binding site with much higher levels of therapeutic PARPi and any reduction in PARPi expression caused by the ATRi combination. This provides an impetus for further study of [¹⁸F]FTT as a tool for understanding the pharmacodynamic impact of combination therapy.

Sixteen subjects with HGSOC are included in the following analysis. Subject details of genetic mutations are provided in Table 1 and baseline characteristics are summarized in Supplementary Table S1. The subjects’ ages ranged from 46 to 72 years, and they had pathogenic or likely pathogenic mutations in HRR genes or had tumors that were HRD by a CLIA approved assay. Of those subjects with genetic testing performed, 31.3% had germline BRCA1/2 mutations (5 of 16), 18.8% had germline RAD51C, RAD51D, or ATM mutations (3 of 16). 41.7% had somatic BRCA1/2 mutations (5 of 12). Somatic mutations detected in other HRR genes, included ATM and CHEK2. Seventy five percent (6 of 8) subjects had HRD tumors by positive assay. The PARPi treatment course and response is delineated per subject in Supplementary Table S2. Eight subjects (50%) enrolled in the FTT trial, received combination PARPi-ATRi, and all others received PARPi alone. Of note, four subjects across both treatment regimens received prior lines of PARPi, namely olaparib or rucaparib.

Clinical trial imaging study design is outlined in Fig. 2A. Briefly, all 16 subjects had baseline [¹⁸F]FTT-PET imaging prior to starting a PARPi (olaparib), and 13 subjects had paired baseline and second [¹⁸F]FTT-PET imaging scans after a short interval after initiation of olaparib, after which they continued olaparib therapy (n = 5) or olaparib therapy plus ATRi (ceralsertib; n = 8; Supplementary Table S2). The time of the second [¹⁸F]FTT-PET after PARPi initiation ranged from 4 to 14 days (median of 7 days and mean of 11 days for all data) with the goal to assess for ability to block uptake of [¹⁸F]FTT. A single subject had imaging at 55 days after baseline imaging secondary to scheduling during the COVID pandemic. The time between [¹⁸F]FTT-PET scans did not impact the change in SUV uptake between scans (Supplementary Fig. S2A), and low-dose CT imaging obtained (for attenuation correction and anatomic localization and not RECIST measurements) with the PET scan did not show a significant change in tumor volume between scans during this interval between first and second scans (Supplementary Fig. S2B).

Target lesions identified on baseline anatomic imaging per RECIST criteria numbered from 1 to 4 (median 1.5) with 8 subjects having one target lesion with partial volume effects mitigated as lesions were selected based on RECISTv1.1 criteria. Two target lesions in two separate subjects identified anatomically were not well seen on baseline [¹⁸F]FTT-PET and excluded from this analysis. Both of these lesions were adjacent to the liver, with physiologic uptake in the liver precluding accurate assessment.

For determination of time-to-progression, the date of the baseline [¹⁸F]FTT scan served as day 0 (also the first day of initiation of PARPi). Target lesions were identified on anatomic imaging obtained prior to treatment, ranging from 47 days before [¹⁸F]FTT scan to 6 days after (median and mean of 15 and 18 days before, respectively). First anatomic follow-up assessment by RECIST occurred at a median of 68 days and a mean of 93 days (range from 57 to 158 days). Best RECIST measurements occurred at a median of 179 days and a mean of 166 days (range from 57 to 351 days). Baseline SUV ranged from 1.68 to 10.23, and follow-up SUV ranged from 1.36 to 3.1. Representative [¹⁸F]FTT-PET/CT imaging for subjects with and without response to PARPi by RECIST are shown (Fig. 3).

For study analysis, two subjects were excluded from CA-125 data due to missing data. Also, subjects 15 and 16 were censored and thus excluded from PFS, best CA-125 response, and best RECIST response analysis as these subjects remained on trial at the time of this analysis (note that subjects are responding to treatment; and are denoted in red when graphed). PFS ranged from 3 to 23 months, with a median of 5.5 months and a mean of 8.5 months (SD of 6.2; Fig. 2B). Combination therapy was associated with a greater than 6-month PFS, inclusive of a subject still on trial at 10 months (P = 0.036, Chi-squared test).

We first studied the ability of the change in [¹⁸F]FTT-PET uptake after PARPi initiation to predict drug response by anatomic/CT imaging using RECISTv1.1. A correlation was observed between RECIST response at first evaluation and △SUV on a per subject (r = 0.60, P = 0.034; Fig. 4A), but not on a per lesion basis (Supplementary Fig. S3A). When assessing the correlation between △SUV (per subject) and the best RECIST response, there was an even greater correlation (r = 0.75, P = 0.01) demonstrated (Fig. 4B). Subjects with a >50% decrease in SUV had a significant reduction in anatomic size with treatment versus those who did not achieve such a ¹⁸F]FTT-PET response (Fig. 4A and B; first response, P = 0.022; best response, P = 0.009, respectively). Using baseline SUV as opposed to △SUV also did not yield associations with first anatomic response (Supplementary Figs. S4A and S4B).

We next assessed the correlation of FTT imaging with the best response by CA-125, a tumor marker indicative of overall disease burden (percent change in baseline to nadir). [¹⁸F]FTT △SUV demonstrated a significant correlation with change in CA-125 (r = 0.73; P = 0.033; Fig. 4C). △SUV threshold again dichotomized the CA-125 response. Subjects with >50% decrease in SUV all had a >50% decrease in CA-125, whereas subjects who did not experience this degree of SUV uptake decrease, did not achieve a CA-125 response (P = 0.016). A threshold of at least a 50% decrease in CA-125 defines a response by the Gynecologic Cancer Intergroup (GCIG), providing clinical relevance for this metric (39). △SUV compared with GCIG criteria for response, resulted in a similar trend (Spearman: r = 0.64, P = 0.07; T test: P = 0.016; Supplementary Fig. S5A).

Similar to RECIST results, a correlation was not seen with this outcome measure and baseline SUV only (r = −0.28, P = 0.40; Supplementary Fig. S4C). These findings again indicate the value of drug-target engagement measured by [¹⁸F]FTT over drug-target expression alone. Change in size by first RECIST response (median 58 days) was also not associated with eventual best CA-125 response (r = 0.58, P = 0.11; Supplementary Fig. S3B), noting that the best CA-125 response could occur anytime during treatment. As expected, the outcome measures of PFS and CA-125 were correlated (r = −0.92, P = 0.0016; Supplementary Fig. S5B).

Next, we studied the ability of [¹⁸F]FTT-PET uptake prior to PARPi treatment and/or change in [¹⁸F]FTT after initiation of a PARPi compared with baseline to serve as a biomarker predictive of PFS. A significant correlation was seen between provider determined PFS (as imaging not obtained consistently off trial) and percent change in SUV_max between baseline and post-PARPi (short interval) imaging (△SUV; r = −0.67, P = 0.027; Fig. 4D). As [¹⁸F]FTT uptake decreased, indicative of drug-target engagement, PFS increased. All five subjects demonstrating a decline in SUV ≤50% had PFS ≤6 months. Conversely, all six subjects with decline in SUV >50% had >6-month PFS. Dichotomizing subjects by this △SUV threshold of 50% (or similarly a PFS of 6 months) yielded significant differences in PFS (P = 0.004; Fig. 4D). PFS by RECIST had a similar trend (r = −0.6, P = 0.056) and comparing >/≤ 50% △SUV revealed similar results (P = 0.0087; Supplementary Fig. S5C). Baseline [¹⁸F]FTT uptake alone did not yield a significant correlation with PFS (r = 0.26, P = 0.37; Supplementary Fig. S4D), suggesting that the level of drug binding to target alone was not sufficient to predict response. PFS and percent change in lesion size by first assessment on CT by subject also demonstrated a significant correlation (r = −0.65, P = 0.035; Supplementary Fig. S3C), noting that the first RECIST assessment was obtained after the post-PARPi [¹⁸F]FTT-PET (median 68 days vs. 7 days, respectively), increasing the clinical relevance of the PET imaging data. Together given RECIST, CA-125, and PFS correlation with [¹⁸F]FTT target engagement, [¹⁸F]FTT-PET may serve to further identify patients who may benefit clinically from PARPi therapy.

Patients with HRD ovarian cancers should receive a PARPi. Most patients benefit but PARPi resistance ultimately often emerges. Two subjects had benefited and then progressed on a PARPi, then underwent assessment with [¹⁸F]FTT-PET/CT (prior to enrollment onto CAPRI for combination PARPi-ATRi; NCT03462342). Subject 13 completed olaparib 300 mg twice daily for 14 months prior to progression and underwent subsequent [¹⁸F]FTT-PET imaging after drug washout. Subject #16 completed two years of rucaparib therapy nearly 6 years before enrollment on the CAPRI trial, with six cycles of carboplatin/taxol also completed in the interim. Both subjects demonstrated [¹⁸F]FTT uptake at baseline followed by abrogation of signal after PARPi run-in (52% and 50% decrease for #13 and #16, respectively) with subject #13 having a relatively long PFS of 9 months. Subject 16 is currently responding to therapy after 3 months and is still on trial. Representative [¹⁸F]FTT-PET/CT imaging is shown for both subjects (#13, Supplementary Fig. S6; #16, Fig. 5). The constellation of findings suggests continued expression of PARP-1 in these previously resistant tumors and therapeutic effect of PARPi combination, advocating for further study for PARPi retreatment and combined approaches in these tumors and the potential utility of [¹⁸F]FTT-PET/CT in PARPi-resistant patients.

Because the initial approval of olaparib in 2014, treatment of HGSOC with PARPi is standard of care. Indeed, three PARPi are currently approved—olaparib, rucaparib, and niraparib—with broadening clinical indications from ovarian cancer and including now other malignancies (pancreas, prostate, and breast cancer). Patient selection for these targeted agents, though, leverages associations of drug efficacy with underlying genetic mutations or prior sensitivity to platinum agents, without direct measurement of tumoral drug-target expression or engagement. PARP-1 is the major target of PARPi with PARP-2 and PARP-3 less so (40). A target-specific biomarker has the potential to guide therapy identifying patients who will benefit with the goal to ultimately improve patient outcomes. In this study, we evaluated the utility of a non-invasive PET imaging agent of PARP-1 expression, [¹⁸F]FTT, to guide PARPi treatment, both as a biomarker for current agents as well as a tool for future drug development, and in particular, combination DDR inhibitor therapies.

Preclinical human PDX studies support the ability of [¹⁸F]FTT to measure the early pharmacodynamic effect of PARPi monotherapy and in combination with ATRi, with PARP-1 protein expression studies providing insight into the molecular mechanism (Fig. 1). As expected, PARPi administration led to a decrease in radiotracer uptake as evidence by measuring [¹⁸F]FTT T/M. The addition of an ATRi led to a further abrogation of tracer uptake, suggesting an additive effect. PARP-1 protein expression studies provided a plausible molecular mechanism for this phenomenon. Unlike single-agent PARPi therapy, the addition of ATRi decreased PARP-1 expression. The combination of blocking and decreased target expression would therefore be expected to lead to a larger decrease in PET signal compared with blocking alone with PARPi monotherapy. The extent of [¹⁸F]FTT signal decrease correlated with anatomic size changes in the two PARPi-resistant tumors after therapy, with clear separation between the effects of PARPi monotherapy and combination PARPi-ATRi. That is, as drug-target engagement increases, either PARPi binding to the PARP-1 binding site or ATRi reduction in PARP-1 expression, seen as a decline in [¹⁸F]FTT uptake pre- and posttherapy, anatomic growth slowed. In a PARPi-sensitive model, such a correlation was not revealed, likely due the robust sensitivity to PARPi monotherapy where there was no added benefit to combination therapy, underscoring the utility for [¹⁸F]FTT in guiding use for combination therapies. Finally, tumor growth curves demonstrated the additive synergistic effects of combination therapy, corroborating prior studies with the xenograft models (33). These results add to the literature investigating the underlying biology of PARPi-ATRi combination therapy (33), a promising therapy to overcome PARPi- and platinum-resistance, and supports the use of [¹⁸F]FTT as an investigational tool to facilitate drug development.

This preclinical data support the ability to assess PARPi-target engagement for combinations of PARPi with other relevant DDR agents. Such preclinical investigation could help guide clinical studies by identifying combination effects in vivo. Indeed, the Phase II CAPRI clinical trial is underway evaluating the combination of ceralasertib (AZD6738), an ATRi, plus olaparib in subjects with recurrent, platinum sensitive (cohort A) and PARPi-resistant (Cohort C and D) ovarian cancer. Results from the platinum-resistant cohort demonstrate some activity of this combination, particularly in HR-deficient subjects (38). More strikingly, the PARPi-ATRi combination demonstrated efficacy in a cohort of subjects who had previously progressed on prior PARPi, representing human translation of the findings in animal models (41).

In the early human studies with [¹⁸F]FTT-PET imaging presented here, the difference of [¹⁸F]FTT uptake post-PARPi compared with baseline, indicative of drug-target engagement, appeared to identify patients most likely to benefit from PARPi therapy only ∼1 week after treatment initiation (Fig. 4). Most strikingly, △SUV and not baseline SUV alone, correlated with first and best response to treatment by RECIST. Moreover, subjects with greater than 50% decrease in [¹⁸F]FTT uptake had a PFS greater than 6 months, whereas those who had less than 50% decrease had a PFS less than 6 months. A similar dichotomy was seen with the best CA-125 response: subjects with a greater than 50% decrease in [¹⁸F]FTT signal also had a greater than 50% decrease in CA-125, with the converse also true. Change in lesion size by CT (RECIST) correlated with PFS, but not CA-125 response (Supplementary Figs. S3B and S3C), again noting that RECIST assessments took place a median of 2 months after the paired [¹⁸F]FTT scans. Importantly, leveraging a PET measure of drug-target engagement to predict response has previously been demonstrated with estrogen receptor imaging agent, [¹⁸F]Fluoroestradiol (FES), and targeted estrogen-therapy (42, 43), providing a model for translation of this paradigm. However, the pretherapy [¹⁸F]FTT uptake, which provides a measure of PARP-1 expression, alone did not predict response to therapy. This finding deserves further study, but suggest that estimates of drug-target engagement using serial imaging may be important in predicting response to PARPi ± combination therapy.

The preclinical reduction of [¹⁸F]FTT-PET signal with the addition of ATRi to PARPi therapy compared with PARPi alone suggests synergistic effects of these agents, which is beginning to be born out in clinical trials (41). Subjects in our human imaging trial treated with combination therapy, though, were imaged after a PARPi run-in and prior to the addition of ATRi, demonstrating predictive value for combined therapy and the identification of patients with preserved PARP-1 expression and tracer uptake blockade with therapeutic PARPi. Future studies at our center will include imaging patients soon after the initiation of PARPi-ATRi combination therapy to study the earlier biomarker application of [¹⁸F]FTT to infer the impact of the combined therapy similar to the mouse studies. Future study might evaluate [¹⁸F]FTT as a tool to select drug dosages in the development of new PARPi. Dosing to pharmacologic endpoint of complete blockade of [¹⁸F]FTT signal, as opposed to intolerable side effects, may prove especially valuable. A similar strategy has been used with FES in the development of new ER-blocking agents (44).

This study has a number of limitations, most noticeably the limited number of mouse models and clinical trial patients. The hypothesis generated here should be studied in larger trials, including multicenter studies. Additional studies should also include [¹⁸F]FTT-PET imaging after combination therapies, which were not performed here, but which are supported by preclinical studies. Decrease in [¹⁸F]FTT uptake decline could also indicate general cytotoxic drug effect resulting in overall less viable tumor or cleavage of the DNA-binding domain of PARP 1 or 2 during apoptosis by cleaved caspase 3/7 or 8, respectively (45, 46). Evidence from our studies including no change in tumor volume between clinical scans (Supplementary Fig. S2B), supporting that viable tumor has not substantially changed. Tumor growth during initial treatment phase in preclinical studies (Supplementary Fig. S1B), and no significant upregulation of apoptosis markers by RPPA in a PDX model after 2 weeks of treatment (Supplementary Fig. S1C), provide compelling evidence that any apoptosis effects are below biologic levels and not significantly contributing to [¹⁸F]FTT uptake decline. Further validation of the lack of cell/tumor dropout as contributing to the decline in [¹⁸F]FTT uptake in tumors could be performed by pairing [¹⁸F]FTT scans with FDG scans, as was done in a prior study of ER expression using [¹⁸F]Fluoroestradiol PET paired FDG-PET (47).

Overall, the preclinical and early human studies described herein provide the first evidence of a PARPi-PET radiotracer to serve as an early biomarker to guide PARP-targeted therapies. Such information, obtained noninvasively and across the burden of disease, has potential to advance precision medicine and improve outcomes with PARPi. Further larger studies are needed to validate these findings.

A.R. Pantel reports grants from Basser Center for BRCA, NIH KL2TR001879, and DOD OC160269 during the conduct of the study as well as personal fees from Blue Earth Diagnostics and GE and grants and personal fees from Progenics outside the submitted work. D.A. Torigian reports employment with Quantitative Radiology Solutions LLC. L.P. Martin reports personal fees and other support from Sutro Biopharma and ImmunoGen and other support from Mersana and AstraZeneca outside the submitted work. G.B. Mills reports grants, personal fees, nonfinancial support, and other support from Amphista, AstraZeneca, Chrysallis Biotechnology, GSK, ImmunoMET, Ionis, Lilly, PDX Pharmaceuticals, Signalchem Lifesciences, Symphogen, Tarveda, Turbine, Zentalis Pharmaceuticals, Catena, HRD assay to Myriad Genetics DSP to NanoString, Adelson Medical Research Foundation, Breast Cancer Research Foundation, Komen Research Foundation, Ovarian Cancer Research Foundation, Prospect Creek Foundation, NanoString Center of Excellence, Ionis (provision of tool compounds), and Genentech during the conduct of the study as well as grants, personal fees, nonfinancial support, and other support from Amphista, AstraZeneca, Chrysallis Biotechnology, GSK, ImmunoMET, Ionis, Lilly, PDX Pharmaceuticals, Signalchem Lifesciences, Symphogen, Tarveda, Turbine, Zentalis Pharmaceuticals, Catena Pharmaceuticals, HRD assay to Myriad Genetics, DSP to NanoString, Adelson Medical Research Foundation, Breast Cancer Research Foundation, Komen Research Foundation, Ovarian Cancer Research Foundation, Prospect Creek Foundation, NanoString Center of Excellence, Ionis (provision of tool compounds), and Genentech outside the submitted work. R.K. Doot reports grants from PARP-Targeting PET Tracer to Guide DDR Inhibitor Therapy and Department of Defense during the conduct of the study. D.A. Mankoff reports other support from Trevarx outside the submitted work as well as membership on a Trevarx Scientific Advisory Board. R.H. Mach reports other support from Trevarx outside the submitted work; in addition, R.H. Mach has a patent issued and licensed to Trevarx and is also the co-founder of Trevarx. L.L. Lin reports other support from Trevarx and grants from AstraZeneca, Pfizer, and Varian outside the submitted work. F. Simpkins reports grants, personal fees, and nonfinancial support from AstraZeneca during the conduct of the study. No disclosures were reported by the other authors.

A.R. Pantel: Conceptualization, resources, data curation, software, formal analysis, supervision, funding acquisition, validation, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing. S.B. Gitto: Data curation, software, formal analysis, investigation, visualization, writing–original draft, writing–review and editing. M. Makvandi: Conceptualization, resources, data curation, software, formal analysis, supervision, funding acquisition, validation, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing. H. Kim: Data curation, investigation, methodology, writing–review and editing. S. Medvedv: Data curation, investigation, methodology, writing–review and editing. J.K. Weeks: Data curation, writing–review and editing. D.A. Torigian: Data curation, software, formal analysis, methodology, writing–review and editing. C.-J. Hsieh: Data curation, software, formal analysis, investigation, methodology, writing–review and editing. B. Ferman: Data curation, investigation, writing–review and editing. N.A. Latif: Investigation, writing–review and editing. J.L. Tanyi: Investigation, writing–review and editing. L.P. Martin: Investigation, writing–review and editing. S.M. Lanzo: Resources, validation, writing–review and editing. F. Liu: Formal analysis, writing–review and editing. Q. Cao: Formal analysis, writing–review and editing. G.B. Mills: Resources, investigation, methodology, writing–review and editing. R.K. Doot: Data curation, software, formal analysis, investigation, writing–review and editing. D.A. Mankoff: Conceptualization, resources, supervision, methodology, project administration, writing–review and editing. R.H. Mach: Conceptualization, resources, supervision, methodology, project administration, writing–review and editing. L.L. Lin: Conceptualization, resources, supervision, funding acquisition, methodology, project administration, writing–review and editing. F. Simpkins: Conceptualization, resources, data curation, formal analysis, supervision, funding acquisition, investigation, visualization, methodology, project administration, writing–review and editing.

The authors would like to acknowledge the University of Pennsylvania PET Center and Cyclotron Facility for clinical support and manufacturing of radiopharmaceuticals used in this study. The authors also acknowledge Christina Dulal for assistance with clinical trial documentation, and the kind gift provided by the Miriam and Sheldon Adelson Medical Research Foundation.

DOD OC160269 (F. Simpkins, R.H. Mach, A.R. Pantel, and M. Makvandi); SPORE 1P50CA228991 (F. Simpkins, M. Makvandi); NIH KL2TR001879 (A.R. Pantel); Basser Center Team Science grant (F. Simpkins, A.R. Pantel, M. Makvandi), The BioTrust Collection of the UPENN Translational Center of Excellence in Ovarian Cancer.

The publication costs of this article were defrayed in part by the payment of publication fees. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 USC section 1734.