Preclinical Evaluation of Multistep Targeting of Diasialoganglioside GD2 Using an IgG-scFv Bispecific Antibody with High Affinity for GD2 and DOTA Metal Complex

Pretargeted radioimmunotherapy (PRIT) offers the promise of greatly improved therapeutic indices (TI) for delivery of internal radiation to solid human tumors, including diasialoganglioside GD2-positive neuroblastoma [GD2(+)-NB]. Successful PRIT of GD2(+)-NB has been demonstrated using streptavidin–antibody fusion constructs, owing to its high antigen density (5 × 10⁶ molecules per NB cell) and the slow internalization rate of the anti-GD2 antibody/GD2 complex from the NB cell surface. A typical PRIT protocol begins with a cold dose of bispecific antibody (BsAb) to prelocalize at the tumor, followed with a clearing agent (CA) to remove circulating off-target antibody and then last, a rapidly clearing radiolabeled small-molecule hapten or peptide.

Previously, we developed an anti-GD2 single-chain antibody–streptavidin fusion protein [5F11-scFv-SA (streptavidin)] for delivery of radiolabeled biotin, observing tumor-to-blood radiation dose ratios as high as 160:1 in mouse models of human GD2(+)-NB (1). During parallel therapy studies with radioiodinated anti-GD2 monoclonal antibody (mAb) 3F8 (¹³¹I-3F8), a tumor-to-blood ratio of 2.7:1 was achieved, demonstrating the dosimetric advantage of PRIT over conventional radioimmunotherapy (i.e., with directly labeled antibodies or antibody fragments). Despite this, additional optimization of PRIT with 5F11-scFv-SA is required to address the high immunogenicity of streptavidin and high kidney exposure due to prolonged retention of 5F11-scFv-SA in the renal cortex.

As an alternative to streptavidin/biotin, other high-affinity antibody–hapten pairs have been developed for PRIT (2). Over two decades ago, Reardan and colleagues pioneered a PRIT system using antibodies against metal chelates (3). The mAb 2D12.5 has nmol/L affinity for low molecular weight (MW) 1,4,7,10-tetraazacyclododecane-1,4,7,10-tetraacetic acid (DOTA) complexes with yttrium (Y) and lutetium (Lu), and is well suited for PRIT in vivo (4). Orcutt and colleagues subsequently affinity matured the 2D12.5 sequence to yield a novel scFv (“C825”) with pmol/L affinity with improved dissociation half-time of the antibody–DOTA complex (from 5.5 minutes to ∼5 hours; ref. 5). Specifically intended for PRIT, multiple IgG-scFv BsAb were developed consisting of an IgG with specificity to a cancer cell surface target [e.g., carcinoembryonic antigen (CEA) or A33] linked to C825 scFv at the C-terminus of the IgG light chains as an IgG-scFv format (6). The IgG-scFv BsAbs were sufficiently functional and stable in vivo to permit highly efficient tumor targeting of the ¹⁷⁷Lu-DOTA hapten in mouse xenograft models of human adenocarcinoma (e.g., tumor-to-tissue uptake ratios of >450 for blood and >20 for kidneys; ref. 7). On the basis of these reports, as well as our aforementioned experience pretargeting GD2(+)-NB, we reasoned that anti–GD2-C825 could offer a viable clinical developmental path for PRIT directed at GD2(+)-NB.

In this report, we describe the initial demonstration of hu3F8-C825 for PRIT of GD2(+)-NB. The novel BsAb was prepared by cloning and expression of the sequences for hu3F8 (8) and C825 as an IgG-scFv format, followed with purification, and extensive biochemical and functional characterization in vitro. Next, the BsAb was radiolabeled with the positron-emitting isotope ¹²⁴I (t_1/2 = 4.18 days; ¹²⁴I-hu3F8-C825) for noninvasive serial positron emission tomography (PET) of GD2(+) tumor targeting and plasma clearance, followed with sacrifice of each animal for harvesting of select tissues for assay of radioactivity uptake and biodistribution. For PRIT of GD2(+)-NB with hu3F8-C825, a three-step strategy was developed consisting of injections of hu3F8-C825, a dextran-based CA, and ¹⁷⁷Lu-DOTA-Bn hapten. Although C825 scFv has affinities in the range of approximately 15 pmol/L for each Lu- and Y-DOTA-Bn (5), ¹⁷⁷Lu was used for this approach primarily because of its theranostic radionuclide properties (β-maximum energy, 0.5 MeV; β-average energy, 0.13 MeV; γ, 208 keV, 11% abundance), as well as reports suggesting that PRIT with ¹⁷⁷Lu may result in lower off-target doses compared with ⁹⁰Y and ¹³¹I (9). Finally, additional PRIT experiments were performed to estimate the radiation exposure to various tissues and to predict dose-limiting organ toxicity, followed with a pilot therapy study to evaluate therapeutic efficacy and overall toxicity.

The human GD2(+)-NB IMR-32, as well as the GD2-negative [GD2(−)] cell lines SK-N-SH (NB) and BT474 (human breast cancer in which GD2 is present on a small subset of stem cells; ref. 10) were obtained from the American Type Culture Collection and maintained pursuant to the manufacturer's recommendations. The luciferase-labeled GD2(+)-NB tumor cell line IMR-32-Luc was generated by stably expressing an SFG-GFLuc vector into the IMR-32 cells (11). All cell lines were initially cultured and cryopreserved in small aliquots to limit passages to less than 3 months, and periodically tested for mycoplasma using a commercial kit (Lonza). IMR-32, IMR-32-Luc, and SK-N-SH culture media were supplemented with 10% FBS and 100 U/mL of each penicillin and streptomycin. BT474 culture media were supplemented with nonessential amino acids, 10% FBS, and 100 U/mL of each penicillin and streptomycin.

The BsAb format was designed similar to that described by Orcutt and colleagues, as a C825–scFv fusion to the C-terminus of the light chain of a human IgG (6). The heavy chain is the same as that of human IgG1 except N297A mutation for aglycosylated form, whereas the light chain is constructed as leader VL-Cκ-(Gly₄Ser)₂-scFv. Nucleotide sequences encoding VH and VL domains from our hu3F8, and the disulfide-stabilized C825 scFv were synthesized by GenScript with appropriate flanking restriction enzyme sites, and were subcloned into our standard mammalian expression vector.

Linearized plasmid DNA was used to transfect CHO-S cells (Invitrogen) for stable production of BsAb. One million cells were transfected with 2.5 μg of plasmid DNA by Nucleofection (Lonza) and then recovered in CD OptiCHO medium supplemented with 8 mmol/L l-glutamine (Invitrogen) for 2 days at 37°C in 6-well culture plates. Stable pools were selected first with 500 μg/mL hygromycin for approximately 2 weeks and single clones were then selected out with limited dilution. BsAb titer was determined by GD2 and BSA-(Y)-DOTA-Bn ELISA, respectively. Two million cells from several high-expression clones were grown in 10 mL media per T25 flask for 4 days, and supernatants were harvested for further analysis and selection of the optimum clone.

The BsAb producer line was cultured in OptiCHO medium and the mature supernatant harvested. A protein A affinity column (GE Healthcare) was preequilibrated with 25 mmol/L sodium citrate buffer with 0.15 mol/L NaCl, pH 8.2, and bound BsAb was eluted with 0.1 mol/L citric acid/sodium citrate buffer, pH 3.9, and alkalinized (1:10 v/v ratio) in 25 mmol/L sodium citrate, pH 8.5. The total yield of BsAb was 5 to 10 mg/L. For storage, BsAb was dialyzed into 25 mmol/L sodium citrate, 0.15 mol/L NaCl, pH 8.2, and frozen in aliquots at −80°C.

Biacore T-100 Biosensor, CM5 sensor chip, and related reagents were purchased from GE Healthcare. The gangliosides GM1 was from ALEXIS Biochemicals (AXXORA L.L.C.) and GD2 from Advanced ImmunoChemical. A BSA-(Y)-DOTA-Bn conjugate was prepared by reacting the p-isothiocyanate–DOTA-(Y) complex (prepared according to Corneillie and colleagues; ref. 12) with the albumin using protocols described in Hermanson (13). Gangliosides were directly immobilized onto the CM5 sensor chip (GD2 and GM1 in a 1:1 ratio for the active surface, and GM1 only for the reference) via hydrophobic interaction (14). Separately, BSA-(Y)-DOTA-Bn (active surface) and BSA (reference surface) were immobilized using the Amino Coupling Kit (GE Healthcare). Purified BsAbs and control antibodies (hu3F8; ref. 8) for GD2 and A33-C825 (6) for BSA-(Y)-DOTA-Bn) were analyzed, and data were fit to a 1:1 binding model (for GD2) or bivalent analyte model (for BSA-(Y)-DOTA-Bn) using the Biacore T-100 evaluation software.

¹²⁴I was either provided in-house by the Memorial Sloan Kettering Radiochemistry and Molecular Imaging Probes Core Facility or purchased commercially (IBA Molecular). Stocks of hu3F8-C825 were labeled with radioactive iodine using precoated IODOGEN tubes (Pierce) according to standard protocols (15, 16) to final specific activities of 121 to 152 MBq/mg. The tracer immunoreactivities were evaluated using cell-binding assays (17) and were 78.8% ± 0.1% (n = 3) for GD2(+)-NB IMR-32 and 4.9% ± 0.4% (n = 3) for GD2(−) SK-N-SH. The chelate p-aminobenzyl–DOTA (DOTA-Bn; Macrocyclics) was radiolabeled with ¹⁷⁷LuCl₃ (specific activity: 170 MBq/(nmoles); PerkinElmer) at a ratio of 37 MBq to 1.1 μmoles DOTA-Bn using previously described methods (9), but without HPLC purification (C825 does not recognize metal-free DOTA-Bn).

All animal experiments were approved by the Institutional Animal Care and Use Committee of Memorial Sloan Kettering Cancer Center (New York, NY), and institutional guidelines for the proper and humane use of animals in research were followed. Athymic nu/nu female mice (6–8 weeks old; Harlan Sprague Dawley) were allowed to acclimate in the vivarium for at least 1 week. For the BT474 tumor model, mice were implanted with estrogen (17β-estradiol; 0.72 mg/pellet 60-day release; Innovative Research of America) by Trochar 3 days before inoculation with cells. Groups of animals were injected subcutaneously with either IMR-32-Luc or BT474 in the left flank with 5 × 10⁶ cells formulated 1:1 with Matrigel (BD Biosciences), and tumors were observed in 3 to 6 weeks (volume, V: 200–500 mm³ for IMR-32-Luc and 150–300 mm³ for BT474 using the formula for the volume of an ellipsoid |$V\, = \,{\rm 4/3}\pi {\rm (}length{/2} \times width{/ 2}\times height{/ 2}$|⁠; ref. 18). All reagents were given intravenously via the lateral tail vein. The PRIT protocol included injection of hu3F8-C825 (t = −28 hours), followed 24 hours later by CA or vehicle [the CA is a 500 kDa dextran-(Y)-DOTA-Bn conjugate, prepared according to Orcutt and colleagues; ref. 7); the substitution ratio of (Y)-DOTA-Bn (in moles) per mole of dextran ranged from 61 to 161 (Y)-DOTA-Bn/dextran; t = −4 hours)], and ¹⁷⁷Lu-DOTA-Bn (5.6 MBq, 33 pmoles) after 4 hours (t = 0 h). The timing between injection of hu3F8-C825 and CA was based on the PET imaging studies with ¹²⁴I-hu3F8-C825, whereas a time interval of 4 hours was initially chosen for the CA and ¹⁷⁷Lu-DOTA-Bn. For ex vivo biodistribution analysis, mice were euthanized, and tumor and selected organs were harvested, weighed, and radioassayed by gamma scintillation counting (Wallac Wizard 3 automatic gamma counter, PerkinElmer). Count rates were converted to activities using a system calibration factor, decay corrected and normalized to the administered activity, and expressed as the percentage of injected dose per gram (%ID/g).

IMR-32-Luc GD2(+)-NB tumor-bearing mice (n = 5) were injected intravenously with 8.5 to 10.2 MBq of ¹²⁴I-hu3F8-C825 (280–500 pmoles) and placed under anesthesia by gas inhalation (1% isofluorane/oxygen) before scanning in a microPET Focus 120 (Concorde Microsystems) at various times from 3 to 64 hours post-injection (p.i.). The in vivo stability of ¹²⁴I-hu3F8-C825 was confirmed by serial blood sampling via the tail-vein and silica-gel–impregnated glass-fiber thin layer chromatography (TLC) paper (Pall Corporation) with 10% trichloroacetic acid (Sigma-Aldrich) elution; ≥90.3% of ¹²⁴I-activity was protein associated at 64 hours p.i. Images were collected and processed as described previously, except with parameters for ¹²⁴I (19). Curve fitting of manually drawn two-dimensional region of interest (ROI) data (as ROI_MAX; %ID/g) was performed with Prism 6.0 (GraphPad) software. Select groups of mice were sacrificed immediately following PET for assay of radioactivity in select organs by gamma scintillation counting. For scintigraphy studies, mice previously injected with optimum doses of hu3F8-C825 and CA (1.75 mg of hu3F8-C825 and 250 μg (14% (w/w)) of 61 (Y)-DOTA-Bn/dextran, respectively), and 11.1 to 11.5 MBq of ¹⁷⁷Lu-DOTA-Bn were placed under anesthesia by gas inhalation before scanning in a nanoSPECT (BioScan) at 4 and 20 hours p.i. for 30 minutes (∼10⁵ counts/image) using a low-energy high-resolution collimator and a window set at 208 keV. Images were reconstructed to a 256 × 256 matrix using BioScan HiSPECT software and uploaded into ASIPro VM for analysis. The whole-body activity (WBA) for each mouse was determined by assay in the dose calibrator immediately following image acquisition.

Two separate biodistribution studies were conducted to account for the different batch of CA compared with what was used during the CA dose optimization efforts [the (Y)-DOTA-Bn/dextran substitution ratio was 161 (Y)-DOTA-Bn/dextran during the CA dose optimization, and 61 (Y)-DOTA-Bn/dextran during the following experiments]. During the first set of experiments, groups of GD2(+)-NB IMR-32-Luc tumor-bearing mice (n = 3–4) were given 1.75 mg of hu3F8-C825, CA (125 μg; 7% (w/w) of 61 (Y)-DOTA-Bn/dextran), and 5.1 to 5.7 MBq (30–33 pmoles) of ¹⁷⁷Lu-DOTA-Bn, and sacrificed at 1, 3, 24, 96, 168, and 336 hours p.i. for biodistribution in normal tissue and tumor. The study was repeated a second time with a higher dose of CA (250 μg; 14% (w/w) of 61 (Y)-DOTA-Bn/dextran), with groups of mice (n = 2–5) sacrificed at 2, 24, 120, and 216 hours p.i. For each experiment, the nondecay-corrected time–activity concentration data were fitted using Excel to a 1-component, a 2-component, or a more complex exponential function as appropriate, and analytically integrated to yield the cumulated activity concentration per unit administered activity (MBq h/g/MBq). The ¹⁷⁷Lu equilibrium dose constant for nonpenetrating radiations (8.49 gm cGy/MBq h) was used to estimate the tumor-to-tumor and select organ-to-organ self absorbed doses, assuming complete local absorption of the ¹⁷⁷Lu β rays only and ignoring the γ-ray and non–self-dose contributions. The biological clearance of activity (corrected for ¹⁷⁷Lu decay) from tumor and various tissues was determined by fitting the time–activity data to exponential functions using Prism 6.0 (GraphPad) software.

Mice bearing established GD2(+)-NB IMR-32-Luc subcutaneous tumors (tumor volume, 240 ± 160 mm³) were randomized into two groups of 5 mice per group. One group was treated with three successive cycles of 1.75 mg of hu3F8-C825 (8.75 nmoles), 250 μg of CA (14% (w/w) of 61 (Y)-DOTA-Bn/dextran), and 11.1 MBq of ¹⁷⁷Lu-DOTA-Bn (∼60 pmoles). A time interval of 68 hours was allowed between each treatment cycle. The second group of mice received no treatment. This fractionated approach was used to allow for more efficient delivery of the dose compared with administration of a single cycle of hu3F8-C825, CA, and 33.3 MBq based on modeling estimates (see Supplementary Fig. S1). All mice were weighed and the tumor burden was assessed by caliper measurement every 3 to 4 days. In addition, luciferase imaging was conducted on the treated group using the Xenogen In Vivo Imaging System Spectrum (Caliper Lifesciences) at day −4 and 12 days after treatment to examine viable tumor cells at the xenograft site. Briefly, mice were anesthetized by gas inhalation (1% isofluorane/oxygen) and injected intraperitonially with 0.2 mL solution of D-luciferin [prepared by dissolving 75 mg of D-luciferin potassium salt (Gold Biotechnology) in 5 mL of PBS, then stored in aliquots at −20°C]. Luciferase images were collected 4 to 5 minutes p.i. using the following parameters: A 60-second exposure time, medium binning, and an 8 f per stop. Luciferase image analysis was performed using Living Image 3.0 (Caliper LifeSciences). Mice were sacrificed if the tumor was >1.0 cm³ or lost >20% of body mass during the course of the study. The treated mice were sacrificed after 28 days after treatment for evaluation of acute toxicity by conducting hematology, clinical chemistry, and necropsy (all by the Laboratory of Comparative Pathology, Memorial Sloan Kettering Cancer Center).

Differences in tissue uptake between cohorts were statistically analyzed with the Student t test for paired data. Two-sided significance levels were calculated and a P value of <0.05 was considered statistically significant.

For antigen GD2, hu3F8-C825 had a k_on of 5.96 × 10⁴ 1/(mol/L)/s, a k_off of 1.68 × 10⁻³ per second, and overall K_D of 28.2 nmol/L; comparable with parental hu3F8 (k_on of 1.79 × 10⁵ 1/(mol/L)/s, k_off of 2.91 × 10⁻³ per second, and overall K_D of 16.3 nmol/L; Fig. 1A). For antigen BSA-(Y)-DOTA-Bn, hu3F8-C825 had a k_on of 2.95 × 10⁴ 1/(mol/L)/s, a k_off of 2.84 × 10⁻⁴ per second, and overall K_D of 9.63 nmol/L; comparable with A33-C825 (k_on of 2.73 × 10⁴ 1/(mol/L)/s, k_off of 9.82 × 10⁻⁵ per second, and overall K_D of 3.60 nmol/L; Fig. 1B). In summary, hu3F8-C825 retained high binding affinity to both GD2 and BSA-(Y)-DOTA-Bn, especially the very slow k_off comparable with its parental hu3F8, a parameter considered critical for therapeutic efficacy.

Biochemical purity analysis of hu3F8-C825 by SDS–PAGE and SE-HPLC is shown in Fig. 1C. Under reducing SDS–PAGE conditions, hu3F8-C825 gave rise to two bands at around 50 kDa, because the C825–scFv fusion to hu3F8 light chain increased the MW to approximately 50 kDa. SE-HPLC showed a major peak (70% by UV analysis) with an approximate MW of 210 kDa, as well as some minor peaks assumed to be aggregates and lower MW protein fragments, removable by gel filtration. Isoelectric focusing (IEF) analysis showed a single band at around pH 9.5 (data not shown). The BsAb remained stable by SDS–PAGE, SE-HPLC, and BIACORE after multiple freeze and thaw cycles (data not shown).

PET-derived tumor and blood activity curves for ¹²⁴I-hu3F8-C825 up to 64 hours p.i. are shown in Supplementary Fig. S2. The blood clearance of hu3F8-C825 was biphasic [t_1/2 α of 2.8 hours and t_1/2 β of 14.6 hours (R² = 0.989)], with longer half-lives compared with 5F11-scFv-SA (t_1/2 α, 0.2 hours; t_1/2 β, 9.3 hours), but shorter than those previously reported for 3F8 IgG (e.g., 33 hours for ¹¹¹In-3F8; ref. 20). Tumor uptake of ¹²⁴I-hu3F8-C825 increased to approximately 5%ID/g within 12 hours p.i. and remained consistent for the following 24 hours p.i. The peak activity was 5.7%ID/g at approximately 22 hours p.i. From 36 to 64 hours p.i., the activity decreased gradually to a minimum of 3.3%ID/g. On the basis of these data, a time interval of 24 hours was allowed between the hu3F8-C825 and CA injections for PRIT. Immediately following PET scanning at 48 hours p.i., select tissues were harvested for biodistribution assay ex vivo. Activities in GD2(+)-NB tumor and blood matched those derived using PET, and low uptake and retention in normal tissue (e.g., <1.5%ID/g for spleen and kidney) was observed compared with tumor (∼2.5%ID/g; Supplementary Fig. S3).

The optimum hu3F8-C825 and CA doses for PRIT were determined using a series of ¹⁷⁷Lu-DOTA-Bn biodistribution experiments in groups of mice bearing subcutaneous GD2(+)-NB IMR-32-Luc tumors. First, hu3F8-C825 was tested at four different dose levels: 0.6 (∼3 nmoles), 1.5, 1.75, and 2 mg per mouse (all n = 5 except for the cohort given 0.6 mg, which was n = 2). After injection with hu3F8-C825, all groups received 24 hours later, a proportional dose of CA [∼12.5% (w/w) of the antibody mass (12.7 ± 3.9%; mean ± SD) of CA (161 (Y)-DOTA-Bn/dextran)] and finally, after 4 hours, equal doses of ¹⁷⁷Lu-DOTA-Bn (5.6 MBq; 30 pmoles) were given. Shown in Fig. 2A, we observed that PRIT with 1.75 mg of hu3F8-C825 resulted in the greatest ¹⁷⁷Lu-DOTA-Bn uptake in tumor after 24 hours (10.3%ID/g; e.g., compared with 1.5 or 2.0 mg: 7.4 and 7.2%ID/g, respectively; P < 0.05). Next, the CA dose was titrated (0–0.875 mg; 0%–50% (w/w) of 161 (Y)-DOTA-Bn/dextran; n = 2 animals for vehicle cohort, and n = 3 for all CA cohorts) following injection with 1.75 mg (i.e., the optimum dose) of hu3F8-C825. Shown in Fig. 2B, all CA doses were shown to be effective at improving T/NT ratios (e.g., for blood and kidney) compared with vehicle by significantly reducing the uptake of ¹⁷⁷Lu-DOTA-Bn in normal tissues. Tumor uptake of ¹⁷⁷Lu-DOTA-Bn was attenuated at high doses of CA (e.g., 50% (w/w) CA; P < 0.05 compared with vehicle). The reduction in tumor activity was presumably due to the CA (or associated metabolites) occupying hu3F8-C825 at the tumor, causing the displacement or the blocking of ¹⁷⁷Lu-DOTA-Bn uptake by the BsAb. Representative planar SPECT images of a mouse given optimum hu3F8-C825 and CA doses and 11.1 MBq (∼60 pmoles) of ¹⁷⁷Lu-DOTA-Bn at 4 and 20 hours p.i. are shown in Fig. 3A.

The GD2 antigen specificity of hu3F8-C825, as well as the tumor uptake of ¹⁷⁷Lu-DOTA-Bn without PRIT, was evaluated in tumor-bearing mice. Studies included: (1) injection of ¹⁷⁷Lu-DOTA-Bn alone into groups of mice bearing GD2(+)-NB IMR-32-Luc tumors (2), PRIT of GD2(−) BT474 tumors with hu3F8-C825, and (3) PRIT of GD2(+)-NB IMR-32-Luc with an irrelevant IgG-scFv (A33-C825). These data are shown in Table 1 and Supplementary Tables S1 and S2. Shown in Table 1, without PRIT with hu3F8-C825 and CA (i.e., injection of ¹⁷⁷Lu-DOTA-Bn alone), ¹⁷⁷Lu-DOTA-Bn exhibited rapid clearance and very low whole-body retention in mice bearing GD2(+)-NB tumors, comparable with previous biodistribution studies in mice (9). PRIT of GD2(−) BT474 tumors with hu3F8-C825 and ¹⁷⁷Lu-DOTA-Bn showed minimal uptake of ¹⁷⁷Lu-DOTA-Bn in tumor, suggesting that hu3F8-C825 failed to prelocalize appreciably in antigen-negative tumor (Supplementary Table S1). A parallel study comparing PRIT in mice bearing GD2(+)-NB with either an irrelevant IgG-scFv or hu3F8-C825 showed reduced ¹⁷⁷Lu-DOTA-Bn at the tumor (2.25 and 8.21%ID/g, respectively) as well as T/NT ratios <1 for normal tissues for the irrelevant IgG-scFv, suggesting specific uptake for hu3F8-C825 (Supplementary Table S2).

For the first set of experiments in which PRIT was carried out with the lower CA dose[125 μg; 7% (w/w) of 61 (Y)-DOTA-Bn/dextran], the estimated absorbed doses of ¹⁷⁷Lu-DOTA-Bn (as cGy/MBq) for blood, tumor, liver, spleen, and kidney were 2.1, 70.0, 6.1, 10.4, and 11.5, respectively (Table 2). When the study was repeated using the higher dose of CA [250 μg; 14% (w/w) of 61 (Y)-DOTA-Bn/dextran], the estimated absorbed doses of ¹⁷⁷Lu-DOTA-Bn for blood, tumor, liver, spleen, and kidney were 0.6, 84.9, 2.1, 2.0, and 3.7, respectively (also in Table 2). The estimated dose to kidney was highest among normal tissues, thus presumed to be dose limiting. Decay-corrected time–activity curves for tumor and various tissues up to approximately 220 hours p.i. obtained with the higher CA dose are shown in Fig. 3B. Average tumor uptake was 10.2%ID/g at 24 hours p.i. and cleared with a t_1/2 of 80.3 hours (R² = 0.994). Peak kidney, liver, and blood uptake was observed at 2 hours p.i. (0.91, 0.27, and 0.18%ID/g, respectively). Activity from kidney and liver cleared at a slower rate than blood (kidney t_1/2: 68.0 hours, R² = 0.968; liver t_1/2: 77.4 hours, R² = 0.994; blood t_1/2: 41.0 hours, R² = 0.973), suggesting prolonged retention and/or nonspecific uptake of ¹⁷⁷Lu-DOTA-Bn (or hu3F8-C825-¹⁷⁷Lu-DOTA-Bn complex) in those tissues.

A description of the treatment schedule, which included three successive PRIT injection cycles (total ¹⁷⁷Lu dose: 33.3 MBq), is provided in Fig. 4A. The tumor response curves for treated and nontreated groups are shown in Fig. 4B. For treated mice, tumors started to respond substantially (i.e., >50% reduction) after the second injection cycle, and complete tumor ablation was observed in 2 of 5 mice within 10 days following the last injection of ¹⁷⁷Lu-DOTA-Bn. All nontreated mice had uncontrolled tumor growth to beyond 1.0-cm³ size requiring euthanasia within 12 days (or ∼40 days following tumor inoculation; shown in Fig. 4C). Luciferase imaging of treated mice at day −4 showed signals at the tumor site ranging from approximately 10⁴ to 10⁶ × photons/s/cm²/sr, whereas at 12 days after treatment showed no signal at the tumor site (data not shown). The treated mice cohort was sacrificed at 28 days after treatment for hematology, clinical chemistry, and necropsy analyses to evaluate gross and histologic toxicities to normal tissues, as well as obtain a histologic assessment of any residual tumor at the subcutaneous xenograft site. In 3 of 5 mice, too-small-to-measure residual masses at the site of the xenograft showed microscopically residual NB in 1 mouse, but only lipid laden macrophages in the other 2. During the course of the study, all mice maintained body mass (i.e., no changes greater than 20% between weighing intervals), and were in good visible health, suggesting low overall treatment toxicity (data not shown). Hematology showed no anemia (Supplementary Table S3) but polychromasia of red blood cells and increased reticulocyte counts in all mice (data not shown), suggesting a regenerative response. In addition, all treated animals showed total white blood cell counts and hemoglobin within normal ranges, but a slight decrease in platelets (Supplementary Table S3). Histologic analysis (data not shown) of bone marrow and spleen revealed myeloid hyperplasia (mild to moderate, subacute) in 3 of 5 mice. For kidney, 1 of 5 was normal, whereas in 3 of 5 mice the vessels were mildly ectatic, with rare cortical tubules lined by slightly plump epithelial cells with weakly basophilic cytoplasm, and in 1 mouse, a kidney showed an infarct of the cortex and medulla, with collapsed to mildly ectatic tubules in the area (data not shown). All serum chemistry values, including those indicating renal function, were within reference ranges (Supplementary Table S3).

In this study, we evaluated PRIT of GD2(+)-NB using a novel BsAb containing sequences for an anti-GD2 mAb hu3F8 (8) and an anti-DOTA(metal) scFv (6). Using the optimum PRIT parameters determined during initial experiments, we showed that a treatment schedule consisting of three successive cycles (total ¹⁷⁷Lu-DOTA-Bn: 33.3 MBq), was sufficient to achieve complete tumor responses in 5 of 5 mice within 10 days after treatment, including tumor ablation in 4 of 5 animals, as well as no tumor recurrence and low overall toxicity up to 28 days after treatment. Complete tumor response after 12 days was also confirmed using luciferase measurements as a secondary endpoint, showing no viable cells at the xenograft site.

For PRIT at current optimized doses of hu3F8-C825 and CA and approximately 5.4 MBq of ¹⁷⁷Lu-DOTA-Bn, absorbed doses for tumor and kidney are approximately 85 and 3.7 cGy/MBq, respectively. To estimate doses delivered in our pilot study, we assumed that the dose estimates obtained with 5.4 MBq of ¹⁷⁷Lu-DOTA-Bn (∼30 pmoles) would be valid for 11.1 MBq (∼60 pmoles); hence, three cycles with 11.1 MBq ¹⁷⁷Lu/cycle (33.3 MBq total) would deliver estimated doses of approximately 3,400 cGy to the tumor (see Supplementary Fig. S1), <∼120 cGy to kidney, and <20 cGy to blood. Although these parameters need to be optimized to reduce bystander radiation especially to liver and kidney, the lack of toxicities in our xenograft studies are encouraging. Our results agree with the prediction of Press and colleagues that, because of the superior TI of PRIT, administration of a ⁹⁰Y dose as high as twice the lethal dose using non-pretargeted RIT was possible with negligible toxicity (21).

The relatively high kidney exposure compared with other normal tissues is presumably due to capture of hapten by BsAb localized at the organ, because renal dose estimates for ¹⁷⁷Lu-DOTA-Bn alone are only 0.01 cGy/MBq (9). Nonetheless, this is a substantial improvement in TI seen for PRIT directed at GD2(+)-NB with 5F11-scFv-SA, which showed higher doses to both tumor and kidney (approximately 150 and 50 cGy/MBq, respectively), and a TI for kidney of approximately 3:1. These data are also promising compared with other PRIT strategies, including those with BsAb against tumor antigens and histamine–succinyl–glycine (HSG) hapten, pioneered by Sharkey and colleagues (22). For example, they showed in preclinical mouse models of CEA-expressing tumors, absorbed radiation dose estimates for PRIT with anti-CEA/HSG BsAb and ¹⁷⁷Lu-hapten (∼23 MBq) were 16.9 Gy for a 6-mm subcutaneous tumor and 2.3 Gy for kidney, corresponding to approximately 73 and 10 cGy/MBq, respectively (TI for kidney >7:1; ref. 23).

GD2(+)-NB IMR-32 is not a particularly radioresistant human NB cell line (24), and we anticipate that some tumors may require higher doses >3,400 Gy to demonstrate tumor response. Our radiation exposure benchmarks for curative PRIT of GD2(+)-NB (i.e., for ablation without recurrence) are based on prior experience to be ≥4,200 cGy in tumor (18) at the expense of a maximum kidney exposure of approximately 2,500 cGy (25). To achieve this tumor dose of 4,200 cGy, we estimate a single-cycle treatment with the higher CA dose and a ¹⁷⁷Lu-DOTA-Bn dose of approximately 50 MBq is required, exposing the blood and kidney to approximately 30 and 185 cGy, respectively (Table 2). In addition to conducting these studies, there are a number of different approaches that we intend to explore to improve the TI of this PRIT strategy, including (1) optimization of hu3F8-C825 tumor uptake and pharmacokinetics, and (2) optimization of the CA step. The relatively high hu3F8-C825 dose of 1.75 mg (8.75 nmoles) was found to be required to achieve high absolute tumor uptake, which is greater than previous PRIT studies directed at GD2(+)-NB (5.2 nmoles for 5F11-scFv-SA), most likely a function of bivalency of hu3F8-C825 versus tetravalency of 5F11-scFv-SA. This is likely a function of the antigen density (0.6–2.0 × 10⁶ molecules per cell) although improving tumor-targeting efficiency and in vivo stability of the BsAb may theoretically decrease the dose required to achieve saturation (26). One approach is to optimize the hu3F8-C825 affinity or half-life in circulation (27). Yazaki and colleagues suggested that despite serum stability of this IgG-scFv BsAb platform, the murine version of C825 resulted in accelerated clearance of the BsAb by the mononuclear phagocyte system in vivo (28). The Wittrup group has developed a humanized version of C825 that could avoid this clearance issue, besides further reducing the immunogenicity of these constructs in humans (28). Second, when hu3F8 was humanized, its affinity for GD2 (especially k_off) was slightly reduced (8). We have since used an affinity maturation procedure by yeast surface display (29) in which affinity of hu3F8 to GD2 was increased by 10- to 20-fold (data not shown). Finally, although the affinity of C825 was already matured to the pmol/L range (5), further improvement should be possible with an “infinite” affinity approach (30).

Improvement of the clearance step may also enhance tumor uptake of ¹⁷⁷Lu-DOTA-Bn, as well as the tumor-to-normal tissue ratios. The tumor-to-blood ratio at 24 hours p.i. was approximately 75:1 with dextran-based CA during our PRIT studies, but this is significantly less than that for PRIT with 5F11-scFv-SA and a biotinylated N-acetylgalactosamine-dendrimer CA, which showed tumor-to-blood ratios of approximately 450:1 as early as 30 hours p.i. (1). This could be attributed to the million-fold difference in affinity between streptavidin–biotin [>10¹⁵ 1/(mol/L)] compared with C825-DOTA [>10⁹ 1/(mol/L)], and the homotetrameric structure of streptavidin and 5F11-scFv-SA in vivo compared with bivalent binding for C825. As an alternative CA to dextran-(Y)-DOTA-Bn, a DOTA–hapten–N-acetylgalactosamine-dendrimer CA could promote clearance via Ashwell receptors present in the liver and enable excretion via the hepatobiliary route. In addition, the interval between administration of CA and ¹⁷⁷Lu-DOTA-Bn could be optimized to account for reticuloendothelial metabolism of CA–hu3F8–C825 complexes and potentially reduce associated metabolites in circulation.

Future studies will also include optimization of hapten dose and formulation, PRIT using ¹⁷⁷Lu-DOTA-Bn or DOTA-Bn radiolabeled with PET nuclides (e.g., ⁸⁶Y or ⁶⁸Ga) for imaging-guided dosimetry and therapy planning, as well additional treatment studies to further refine dose estimates to produce long-term cures in radioresistant GD2(+)-NB and possibly other GD2(+) tumors, and identify chronic toxicities especially if repeated cycles are to be given.

In this report, we show that PRIT using hu3F8 and C825 is a viable alternative to streptavidin-based therapies, overcoming the major limitation of renal uptake. Hu3F8 has low immunogenicity in human trials (clinicaltrials.gov NCT01419834, NCT01757626, and NCT01662804), and with the introduction of the humanized version of C825, the immunogenicity issue might be finally overcome to facilitate the clinical development of anti-GD2 MST.

N.-K. Cheung has ownership interest (including patents) in hu3F8. No potential conflicts of interest were disclosed by the other authors.

Conception and design: S.M. Cheal, H. Xu, P.B. Zanzonico, S.M. Larson, N.-K. Cheung

Development of methodology: S.M. Cheal, H. Xu, S.M. Larson, N.-K. Cheung

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): S.M. Cheal, H. Xu, H.-F. Guo, S.M. Larson, N.-K. Cheung

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): S.M. Cheal, H. Xu, P.B. Zanzonico, S.M. Larson, N.-K. Cheung

Writing, review, and/or revision of the manuscript: S.M. Cheal, H. Xu, P.B. Zanzonico, S.M. Larson, N.-K. Cheung

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): H.-F. Guo, S.M. Larson, N.-K. Cheung

Study supervision: S.M. Larson, N.-K. Cheung

Other: S.M. Cheal is a postdoctoral research fellow in the laboratory of S.M. Larson

The authors thank Dr. Dane Wittrup and his laboratory at Massachusetts Institute of Technology, Cambridge, MA for their generosity and expert advice. The authors also thank Dr. Vladimir Ponomarev for kindly providing the SFG-GFLuc vector. The authors thank Blesida Punzalan, Michael Doran, Sandhya Chalasani, Shoaib Fareedy, and Valerie Longo for their excellent technical assistance with the animal experiments.

This study was supported in part by the following: The Center for Targeted Radioimmunotherapy and Theranostics, Ludwig Center for Cancer Immunotherapy, Memorial Sloan Kettering Cancer Center (MSK; to S.M. Larson), a training grant from the NIH (R25-CA096945; principal investigator H. Hricak; research and salary support to S.M. Cheal), William H. Goodwin and Alice Goodwin and the Commonwealth Foundation for Cancer Research and The Experimental Therapeutics Center of MSK (to N.-K. Cheung), Kids Walk for Kids with Cancer NYC (to N.-K. Cheung), and the Robert Steel Foundation (to N.-K. Cheung). S.M. Larson was also supported in part by P50-CA86438. Technical services provided by the MSK Small-Animal Imaging Core Facility were supported in part by NIH Grants R24-CA83084 (to H. Hricak), P30-CA08748 (to C. Thompson), and P50-CA92629 (to H. Scher). NIH Shared Instrumentation Grant No 1 S10 RR020892-01 (to S.M. Larson), NIH Shared Instrumentation Grant No 1 S10 RR028889-01 (to P.B. Zanzonico), and a Shared Resources Grant from the MSKCC Metastasis Research Center (to P.B. Zanzonico), which provided funding support for the purchase of the Focus 120 microPET, NanoSPECT/CT Plus, and Ivis Spectrum, respectively, are gratefully acknowledged.

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.