Leveraging PET to image folate receptor α therapy of an antibody-drug conjugate

Unless otherwise stated, all chemicals and solvents were used without further purification. Water used for this study was ultrapure (> 18.2 MΩcm⁻¹ at 25 °C). Phosphate-buffered saline (PBS) as well as cell growth medium was purchased from the Media Preparation Facility at Memorial Sloan Kettering Cancer Center (MSK) (New York, NY). Humanized monoclonal antibody recognizing FRα and M9346A, as well as antibody-drug-conjugate IMGN853 were provided by ImmunoGen, Inc. (Waltham, MA) and further purified via a PD10 desalting column (GE Healthcare). Concentrations of solutions containing antibody were determined by using a NanoDrop™ 2000 spectrophotometer from Thermo Fisher Scientific (Waltham, MA). The bifunctional chelator p-isothiocyanatobenzyl-desferrioxamine (DFO-Bz-NCS) was purchased from Macrocycles (Plano, TX). ⁸⁹Zr-oxalate was acquired from 3D Imaging, LLC (Maumelle, AR). HPLC reactions (1.0 mL/min, phosphate-buffered saline) were performed on a Shimadzu UFLC HPLC system equipped with a DGU-20A degasser, a SPD-M20A UV detector, a LC-20AB pump system, and a CBM-20A communication BUS module using a size exclusion column (GE Superdex™ 200, 10/300 GL).

The humanized antibody, M9346A, was functionalized with the bifunctional chelator p-isothiocyanatobenzyl-desferrioxamine (DFO-Bz-NCS) with a 1:3.5 antibody:DFO ratio, as previously described [17, 18]. Specifically, a solution of previously purified antibody M9346A (500 μL, 2.27 mg) in PBS was adjusted to pH = 8.5 with 1 M Na₂CO₃ (1.0 μL). Then, a solution of DFO-Bz-NCS (4.2 mM, 4.0 μL) in dimethylsulfoxide was slowly added and the reaction mixture was incubated at 37 °C for 90 min. Afterwards, the DFO-modified antibody was purified using a PD-10 desalting column and isolated in PBS solution for in vitro and in vivo studies.

A solution of ⁸⁹Zr-oxalate in 1 M oxalic acid (5.0 μL, 85.1 MBq, 2.3 mCi) was added to phosphate-buffered saline (100 μL), and a 1 M sodium carbonate solution (4.8 μL) was used to adjust the pH to 7.1–7.4. Then, a solution of DFO-M9346A (500 μL, 500 μg) was added to the neutralized Zr-89 solution and the reaction mixture was incubated at 37 °C for 1 h. After purification using PD-10 desalting columns (GE Healthcare), [⁸⁹Zr]Zr-DFO-M9346A (85.7 ± 5.7% isolated radiochemical yield and 3.08 ± 0.43 mCi/mg specific activity) was isolated in a PBS solution (2.0 mL) and radiochemical purity (98.0 ± 0.7%) was determined through radio-instant thin/layer chromatography. Then, the resulting ⁸⁹Zr-radiolabeled antibody solution was diluted with PBS to the desired volume for in vitro and in vivo studies.

We chose four cancer cell lines with variable FRα expression levels to correlate imaging findings. The HeLa-derived cancer cell line KB (high expression), the ovarian cancer cell line OV90 (low/medium expression), the human lung cancer cell line H2110 (low/medium), and the human lung cancer cell line A549 (low/no expression) were purchased from ATCC (Manassas, VA). No further cell line authentication was conducted, and the cell lines were expended by passaging 2–3 times, aliquoted, and frozen in liquid nitrogen. For use in in vitro as well as in vivo experiments, the cell lines were grown in medium recommended by American Type Culture Collection (ATCC) and passaged regularly at 70–80% confluence every 3–4 days. All cell lines were cultured at 37 °C and 5% carbon dioxide.

[⁸⁹Zr]Zr-DFO-M9346A was investigated for stability by incubating the radioligand in 1% bovine serum albumin at 37 °C for 3 days. Therefore, radiochemical purity of [⁸⁹Zr]Zr-DFO-M9346A was tested at various time points via radio-instant thin layer chromatography (Eckert & Ziegler, Germany) with ethylendiaminetetraacetic acid (50 mM) as mobile phase.

A Lindmo assay [19] was performed to validate the integrity of the ⁸⁹Zr-radiolabeled antibody M9346A. In detail, serial dilutions of KB cells were incubated with one concentration of [⁸⁹Zr]Zr-DFO-M9346A in Eppendorf vials (1.5 mL) on a shaking rack. After incubation at room temperature for 1 h, cells were washed with ice-cold PBS and the retained radioactivity on cells was measured using a WIZARD² automatic γ-counter from PerkinElmer.

Furthermore, an in vitro binding assay was performed with cell lines expressing different levels of FRα, as previously described [20]. One day prior to the in vitro experiment, the cells lines KB, OV90, and H2110, as well as A549 (each 1.0 × 10⁶ cells per well, > 90% viability for each cell line) were seeded into 6-well plates containing growth medium (2.0 mL) and incubated at 37 °C to form subconfluent cell monolayers. Then, growth medium was removed from the 6-well plates, cells were washed with PBS, and 900 μL of growth medium was added. After 1 h of incubation at 37 °C, a solution of [⁸⁹Zr]Zr-DFO-M9346A (0.5 μCi, 100 ng) in PBS (100 μL) was added to the wells. For blocking studies, cells were pre-incubated with humanized antibody M9346A (10 μg) 5 min prior to the addition of ⁸⁹Zr-radiolabeled antibody (0.5 μCi, 100 ng). To determine the number of cells at the time of the experiment, a separate set (n = 3) of wells for each cell line and experiment was analyzed by detaching the cells with trypsin and counted immediately using a Vi-cell XR cell viability analyzer (Beckman Coulter). After 1 h post-incubation at 37 °C, the supernatant of each well was collected together with washing solution of ice-cold PBS. Then, cells were lysed with sodium hydroxide solution (1 M, 1.0 mL) for 5 min at room temperature. Finally, each cell suspension was collected and all radioactive samples were measured using a WIZARD² automatic γ-counter from PerkinElmer (Waltham, MA). Receptor-specific uptake was determined by correlating cell-bound activity relative to non-bound activity in the supernatant and displayed as a percentage of applied activity per 5.0 × 10⁵ cells.

All in vivo studies and procedures were performed in accordance with an approved protocol from the Institutional Animal Care and Use Committee at MSK. All in vivo experiments were carried out in female, athymic nude mice (Envigo; outbread; 6–8 weeks, 20–25 g). Subcutaneous KB and OV90 xenografts (150 ± 20 mm³) were established as previously described [5, 20]. In detail, suspended KB cells (1.5 × 10⁶, viability > 93%) or suspended OV90 cells (1.0 × 10⁷, viability: > 95%) in a solution containing a 1:1 mixture of Matrigel (Becton Dickinson, Bedford, MA) and cell culture media (no FBS for OV90 xenografts) were subcutaneously inoculated on the right shoulder of anesthetized mice (1.5–2.0% isoflurane (Baxter Healthcare) in medical air (2 L/min)). Prior in vivo studies, KB tumors were grown for 9–10 days post-implantation and OV90 tumors were grown for 21 days.

For in vivo studies, PET images were recorded using a small-animal Inveon® PET/CT system from Siemens (Knoxville, TN) and mice were anesthetized with 1.5–2.0% isoflurane at 2.0 L/min flow of medical air. PET images were analyzed using AsiPro VM™ software (Concorde Microsystems) and Inveon research workplace 4.1 software (Siemens Healthcare). For biodistribution studies, mice were euthanized at pre-determined time points through asphyxiation with carbon dioxide and organs of interest were collected, weighed, and counted using a WIZARD² automatic γ-counter from PerkinElmer. The percentage of tracer uptake stated as percentage injected activity per gram of tissue (%IA/g) was calculated as the activity associated with tissue per organ weight per actual injected dose, decay corrected to the start time of counting. KB tumor-bearing mice (n = 10) were injected with [⁸⁹Zr]Zr-DFO-M9346A (7.75 ± 0.25 MBq, 209.0 ± 6.7 μCi, 50 μg) in PBS (200 μL). PET images were acquired at 4 h, 24 h, 48 h, and 72 h post-injection, and biodistribution studies were performed at 24 h and 72 h (each cohort, n = 5). OV90 tumor-bearing mice (n = 6) were injected with [⁸⁹Zr]Zr-DFO-M9346A (2.02 ± 0.08 MBq, 48.4 ± 2.0 μCi, 50 μg) in PBS (200 μL). PET images were acquired at 24 h and 48 h post-injection, and biodistribution studies were performed at 24 h and 48 h post-injection of radioligand. Each tumor-bearing cohort exhibited expected bone uptake upon release of Zr-89 from the chelator desferrioxamine (DFO) of about 5–10 %IA/g.

All iodine-131 radiolabeling reactions were performed in pre-coated Iodogen (1,3,4,6-tetrachloro-3α,6α-diphenyl glycoluril) tubes (100 μg per tube). For instance, an aqueous solution of Na¹³¹I (37.0 MBq, 1.0 mCi) was added to a phosphate buffered saline solution of IMGN853 (250 μg, 45 μL) and the tube containing the reaction mixture was gently agitated at 1-min intervals at room temperature over a period of 10 min. After purification using PD-10 desalting columns, [¹³¹I]-IMGN853 (32.1 MBq, 0.87 mCi) was isolated in 87% radiochemical yield and high radiochemical purity (99.9%).

Female athymic nude mice (6–8 weeks, n = 5) were injected intravenously with [⁸⁹Zr]Zr-DFO-M9346A (2.3 ± 0.2 MBq, 62.6 ± 4.6 μCi, 25 μg) or [¹³¹I]-IMGN853 (2.7 ± 0.1 MBq, 73.4 ± 2.6 μCi, 25 μg). At predetermined time points (0.5 h, 1.0 h, 3.0 h, 6.0 h, 24 h, 48 h, and 72 h), a sample of blood was obtained from the great saphenous vein of each animal in a heparin-coated capillary glass tube and the weights of the collected blood samples were obtained by using an analytical balance Toledo XS105 from Mettler. The radioactivity of the blood samples was recorded with a WIZARD² automatic γ-counter from PerkinElmer. The residual radiotracer, expressed as a percentage injected dose per gram (%IA/g), was calculated as the activity present in the blood weighed per actual injected dose, decay corrected to the time of counting.

For biodistribution studies, OV90 tumor-bearing nude mice (6–8 weeks, n = 5) were injected intravenously with [¹³¹I]-IMGN853 (1.42 ± 0.02 MBq, 38.3 ± 0.4 μCi) in PBS (200 μL). At 48 h post-injection, mice were euthanized and organs of interest were collected, weighed, and counted with a WIZARD² automatic γ-counter from PerkinElmer.

Tumor volumes for all mice were measured via manual caliper measurements of the longest dimension (x), shortest dimension (y), and height (z), assuming an ellipsoid shape (V = (3.1415/6) × (x × y × z) [mm³]). During tumor measurements, weight for all mice was recorded, and after initial tumor measurements, mice were randomized into three cohorts (n = 3–5) per cohort ensuring all cohorts had tumor volumes of 150–200 mm³. One day after initial tumor volume measurements, mice in the therapy cohort A were co-administered with a solution of ADC (1.25 mg/kg, 31.2 μg) and companion imaging agent [⁸⁹Zr]Zr-DFO-M9346A (3.34 ± 0.04 MBq, 90.3 ± 1.1 μCi, 25 μg) in PBS (200 μL). Cohort B was injected with the companion imaging agent [⁸⁹Zr]Zr-DFO-M9346A (3.68 ± 0.07 MBq, 99.5 ± 1.8 μCi, 25 μg). Cohort C was injected with PBS (200 μL). Two days after administration, mice of cohort A and cohort B were anesthetized with 1.5–2.0% isoflurane at 2.0 L/min flow of medical air and PET/CT imaging was accomplished over 10 min using small-animal Inveon® PET/CT system from Siemens (Knoxville, TN). Then, tumor volumes were determined via caliper measurement every 3 to 4 days up to an endpoint volume of > 1000 mm³. Additionally, all mice were assessed twice per week throughout the study for outward signs of toxicity and decreasing body weight.

Modification of the humanized monoclonal antibody (mAb), M9346A, with the bifunctional chelator DFO-Bn-NCS was accomplished using a previously reported method (Fig. 2a, Additional file 1: Figure S1) [17, 18, 21]. First, desferrioxamine was coupled to lysine-NH₂ groups of the parent antibody (1:3.5 antibody: DFO ratio) at pH = 8.5 over 90 min in phosphate buffered saline, followed by purification using PD-10 desalting columns. Subsequently, radiolabeling of the modified mAb with neutralized [⁸⁹Zr]Zr-oxalic acid solution proceeded in PBS at 37 °C over 60 min, followed by purification using PD-10 desalting columns. DFO-M9346A was radiolabeled reliably with good specific activity (3.08 ± 0.43 mCi/mg) and radiochemical purity (98.0 ± 0.7%). Lindmo assays of the resulting radiolabeled antibody construct confirmed that the immunoreactivity of the construct was established through binding of the antibody to FRα in vitro (83.4 ± 3.5%, Additional file 2: Figure S2). Stability of [⁸⁹Zr]Zr-DFO-M9346A was examined in bovine serum albumin at 37 °C over a period of 72 h and showed that more than 95% of the radioligand remained intact (Additional file 3: Figure S3). Further characterization of [⁸⁹Zr]Zr-DFO-M9346A included an in vitro cell uptake assay using cancer cell lines with various expression levels of FRα (Fig. 2c). We observed highly specific cell-associated uptake and retention in KB, OV90, and H2110 after 1 h of incubation at 37 °C (43.9 ± 0.9%, 17.6 ± 0.5%, and 17.3 ± 0.8%, per 500,000 cells respectively), corresponding to their folate receptor expression levels [5]. The cancer cell line A549 with very low expression levels of FRα displayed negligible uptake (0.4 ± 0.1%).

We first asked whether [⁸⁹Zr]Zr-DFO-M9346A can target a tumor in vivo known to be high in expression of FRα (Fig. 3a and Additional file 4: Figure S4). After tail vein injection of the radioligand, small-animal PET imaging (4 h, 24 h, 48 h, and 72 h) and biodistribution studies (24 h and 72 h) were conducted using KB tumor-bearing mice (female, athymic, n = 10). Serial PET imaging showed clear delineation of the tumor already after 24 h with low uptake in normal tissue. The maximum intensity projections (MIPs) indicated that blood-pool and background activity cleared over time, leading to improved tumor-to-background ratios at 48 h post-injection. Ex vivo biodistribution data corroborated the PET data, as high tumor localization of [⁸⁹Zr]Zr-DFO-M9346A at 24 h (30.1 ± 1.9 %IA/g, n = 5) was observed and tumor uptake increased over time out to 72 h (45.8 ± 29.0 %IA/g, n = 5) post-injection, suggesting an optimal time point for future PET imaging studies without increasing bone uptake over time (Additional file 5: Figure S5).

After the first proof-of-concept experiment using KB cells, we set out to validate [⁸⁹Zr]Zr-DFO-M9346A in a more clinically relevant cancer model using the ovarian cancer cell line OV90. To this end, [⁸⁹Zr]Zr-DFO-M9346A was intravenously injected into female athymic nude mice bearing a subcutaneous OV90 xenograft on the right shoulder. PET imaging and ex vivo biodistribution after 24 h and 48 h confirmed efficient retention of the radioligand in the tumor (24.2 ± 6.3 %IA/g, n = 6) with very limited background uptake (Fig. 3b and Additional file 6: Figure S6).

After successful characterization of the antibody-based PET imaging agent showing very promising performance in vivo, we next set out to validate that the antibody-drug conjugate and companion diagnostic show nearly identical pharmacokinetic profiles. In order to achieve this goal, we performed a head-to-head comparison between both antibody constructs. In order to do that, we radiolabeled the antibody-drug conjugate, IMGN853, through direct halogenation with I-131 (Fig. 4a). [¹³¹I]-IMGN853 was isolated in good isolated radiochemical yields (86.7 ± 13.5%) and high radiochemical purity (99.3 ± 0.7%). Based on previous in vitro studies with [⁸⁹Zr]Zr-DFO-M9346A, we tested whether the radiolabeled ADC retains its ability to bind to FRα and performed an in vitro uptake with using KB cells (Fig. 4b). Then, we went ahead and compared the drugs’ pharmacokinetic behavior by evaluating each tracer’s blood half-life in vivo. Figure 4c shows the correlation of tracer (%IA/g) in the blood at predetermined time points. The blood half-life for both radioligands was determined through serial bleeds in healthy mice, and the obtained data points were compared to each other. We found a good correlation between [¹³¹I]-IMGN853 and [⁸⁹Zr]Zr-DFO-M9346A (R² = 0.9736). Ex vivo biodistribution data revealed very similar biodistribution pattern in all organs compared to the companion diagnostic with slightly lower tumor uptake of [¹³¹I]-IMGN853 of 17.3 ± 5.2 %IA/g (Additional file 7: Figure S7).

Finally, we determined whether [⁸⁹Zr]Zr-DFO-M9346A could be used as a companion diagnostic during ADC therapy to predict therapy outcome. For these experiments (Fig. 5a), mice bearing OV90 tumor xenografts were injected with an imaging dose of [⁸⁹Zr]Zr-DFO-M9346A (90.3 ± 1.1 μCi, 25 μg) as well as with a therapy dose of IMGN853 (1.25 mg/kg, 31.2 μg). PET images were acquired 2 days post-administration, and tumor volume of each mouse was tracked over several days until the end of the study. In general, administration of [⁸⁹Zr]Zr-DFO-M9346A during ADC treatment with IMGN853 confirmed good tumoral uptake with high tumor-to-tissue contrast (Fig. 5b). After quantification of the acquired PET images, we observed differences in tumoral uptake with the lowest uptake of 42.4 %IA/g and the highest of 61.6 %IA/g. Following the tumor volume of individual mice revealed a differential response in ADC treatment (Fig. 5c). At 6 weeks post-treatment, two mice of the therapy cohort receiving IMGN853 (1.25 mg/kg) responded by inhibited tumor growth, whereas three mice showed similar tumor growth curves as the cohorts injected with [⁸⁹Zr]Zr-DFO-M9346A (positive control) and PBS (negative control). No weight losses were observed in any cohort of the therapy study (Additional file 8: Figure S8). Correlation between tumoral uptake (%IA/g) of [⁸⁹Zr]Zr-DFO-M9346A and therapy outcome shows that in this cohort, mice with uptake greater than 50 %IA/g appeared to be responders and mice with lower than 50 %IA/g were non-responders (Fig. 5d). The cutoff level at 50 %IA/g was arbitrarily set and corresponds to the average tumor uptake of [⁸⁹Zr]Zr-DFO-M9346A in this cohort.