Comparative Analysis of Bispecific Antibody and Streptavidin-Targeted Radioimmunotherapy for B-cell Cancers

An estimated 19,970 Americans died of non-Hodgkin Lymphoma in 2015 (1), despite the availability of modern immunochemotherapy regimens and the introduction of novel agents, such as ibrutinib and idelalisib (2, 3). This statistic emphasizes the need for additional improvements in the therapeutic armamentarium for lymphomas. One promising approach is to exploit the specificity of mAbs to target drugs, toxins, or radionuclides to lineage-specific cell surface antigens present on B-cell malignancies (4, 5). Early studies with “first-generation” radiolabeled antibodies targeting the CD20 antigen, such as ¹³¹Iodine-tositumomab and ⁹⁰Yttrium-ibritumomab tiuxetan, induced objective tumor responses in 60%–80% of patients with relapsed or refractory indolent B-cell malignancies (4, 6), and in 80%–100% of patients treated in the front-line setting (7, 8), leading to FDA approval of both radioimmunoconjugates. Despite the unequivocal efficacy and safety of these agents, they have been under-utilized, resulting in withdrawal of ¹³¹Iodine-tositumomab from the commercial market. Although the reasons for the under utilization of radioimmunotherapy (RIT) are controversial, it appears likely that the simultaneous emergence of other promising agents (e.g., bendamustine, ibrutinib, idelalisib), which were logistically easier to administer in the offices of hematologists and oncologists, put RIT at a competitive disadvantage. However, despite their convenience and demonstrated short-term efficacy none of these new agents have curative potential, at least as single agents. First-generation RIT often achieves disease control; however, low tumor-to-normal organ ratios of absorbed radioactivity (1.5:1 for tumor-to-lung and 10:1 for tumor-to-whole body following ¹³¹I-anti-CD20) do not reliably lead to disease eradication (9). The majority of patients treated at conventional doses of anti-CD20 RIT eventually relapse (10). While myeloablative RIT conditioning for autologous stem cell transplant (ASCT) has led to objective remission rates of 85%–95%, at least one-third of patients with NHL still relapse, 3%–5% of patients die of infections or pneumonitis, and nonhematopoietic toxicities can be significant (9, 11–16). Advances in RIT that markedly improve the efficacy, and diminish the toxicity of the approach, would potentially elevate the attractiveness of RIT, despite its logistic challenges. The development and optimization of multi-step “pretargeted” radioimmunotherapy (PRIT) approaches appear particularly promising.

Several PRIT methods have been developed, all of which have been shown to be markedly superior to conventional RIT with directly radiolabeled antibodies (17–22). All of these strategies administer a derivatized lymphoma-reactive antibody in a nonradioactive form, allowing it to localize to tumor sites and accumulate without subjecting the rest of the body to nonspecific irradiation. After maximal accumulation of antibody in the tumor, a low molecular weight radioactive moiety with a high affinity for the derivatized tumor-reactive antibody is administered. The small size of the second reagent facilitates rapid tumor penetration, capture, and retention by the pretargeted antibody. Unbound molecules of the radioactive second reagent are so small that they are rapidly cleared from the blood and excreted in the urine. In some PRIT approaches, a “clearing agent” (CA) is injected shortly before the radiolabeled small molecule to accelerate the removal of residual unbound antibody from the bloodstream, preventing it from complexing with the radiolabeled second step reagent (Fig. 1).

Technologies employed to facilitate high-avidity association of the radiolabeled second step reagent with the pretargeted tumor-bound antibody include streptavidin–biotin (SA-biotin) approaches, bispecific mAbs, complementary hybridization (Watson–Crick pairing) of phosphorodiamidate morpholino DNA oligomers (23, 24), and trans-cyclooctene-modified antibodies (Ab) binding to radiolabeled tetrazine ligands (25, 26). Although each of these PRIT methods has been shown to be superior to “first-generation” RIT with directly radiolabeled antibodies, no head-to-head comparisons have been conducted to discern which of the PRIT approaches is most promising for clinical development. Here, we present a comparative analysis of the biodistribution and therapeutic efficacy of the two most popular PRIT strategies, SA-biotin and bispecific antibody targeting (Fig. 1). As demonstrated in this report, both approaches are highly promising, although biodistributions of radioactivity favor the bispecific antibody approach. Each PRIT method was capable of curing 70%–100% of animals bearing lymphoma xenografts when used under optimal conditions, but the bispecific antibody method produced less hematologic toxicity than SA-biotin PRIT. Expected reduced immunogenicity, and absence of potential interference from endogenous biotin-blocking, also argue in favor of the bispecific antibody approach over SA-biotin PRIT for future clinical trials.

A pDG expression vector containing a 2H7-hIgG1-hRNase gene under the control of the CMV promoter was provided by Jeffrey Ledbetter (University of Washington, Seattle, WA). An anti-yttrium-DOTA 2D12.5 scFv was supplied by Claude Meares (University of California Davis, Davis, CA). A plasmid harboring a C825 ds-scFv gene, an affinity-improved 2D12.5 antibody, was obtained from Dane Wittrup (MIT, Boston, MA; refs. 27, 28). An EcoRV-XbaI fragment of 2D12.5 was cloned into the pDG expression vector generating an O22-3-9 plasmid. An EcoRV-XbaI fragment of C825 was inserted into the plasmid O22-3-9, resulting in an O89-1-6 construct carrying 2H7-Fc-C825.

The AscI-linearized O89-1-6 DNA (250 μg) was mixed with 2 × 10⁷ CHO-DG44 cells and electroporated at 280V, 950 microFarads. Transfected cells were incubated in nonselective media overnight and plated in 96-well plates in Excell 302 complete CHO media (CCM) containing 50 nmol/L methotrexate (Sigma). Clones with the highest expression of the FP, determined by an IgG sandwich ELISA, were further treated using progressively increasing concentrations of methotrexate, (50–500 nmol/L).

One of the highest expressing clones, 16B12/300, was thawed (10⁷ cells) and expanded into 40 T175 flasks with 100 mL per flask CCM plus 300 nmol/L MTX at an initial density of 1 × 10⁵ cells/mL and incubated for 14 days. Supernatants were purified over a 12-mL protein A-agarose column (RepliGen Bio Processing). The fractions containing the FP were pooled, dialyzed, and sterile filtered. A nonbinding control bispecific anti-Tag72 CC49-Fc-C825 FP was produced using the same methods.

Amino dextran 500 kDa, 30.5 mg, (Life Technologies) was reacted with 6.1 mg of DOTA–benzyl-NCS (MW = 697 kDa) and 11.4 μL of triethylamine overnight (28), diluted in sodium acetate pH 5.2 and 100 equivalents of yttrium nitrate (336 mg), and incubated overnight at 37°C. The mixture was dialyzed against 2 L of water for 3 days, dried on a Biotage V10 evaporator, dissolved in 2 mLPBS, passed twice over a Bio-Rad EconoPak 10 DG column, dialyzed against water for 5 days, dried, weighed, and resuspended in saline at 4 mg/mL and sterile filtered.

Conjugates of the 1F5 anti-CD20 mAb and SA were synthesized, purified, and characterized as published previously (29, 30). A synthetic, dendrimeric CA containing 16 N-acetylgalactosamine residues and a single biotin residue per molecule (NAGB) was obtained from Aletheon Pharmaceuticals for use with the 1F5–SA conjugate.

⁹⁰Y-labeling of DOTA-biotin was performed using 12 mg/mL DOTA-biotin, 500 mmol/L ammonium acetate pH 5.3, and ⁹⁰Y heated for 60 minutes at 84°C. After cooling, 100 mmol/L DTPA was added and labeling efficiency determined using avidin–agarose beads.

The human Ramos (Burkitt lymphoma) cell line was obtained from the ATCC (2012). Granta-519 (mantle cell lymphoma; 2012) was obtained from Deutsche Sammlung von Mikroorganismen und Zellkulturen (DSMZ). The EL4 and EL4-CD20 cell lines were gifts from Dr. Martino Introna and J. Golay, (Milan, Italy; obtained from ATCC; 2004). Cells were passaged twice, frozen, and stored in liquid nitrogen. Fresh vials of cells were thawed and grown in log-phase growth in RPMI1640 medium supplemented with 10% FBS in a 5% CO₂ incubator. Cell viability exceeded 95% by Trypan blue exclusion. Mycoplasma testing (indirect DAPI stain and PCR testing) was performed on all cell lines prior to storage; in addition, cell lines maintained in culture were tested on a rotating schedule. The Granta-519^luc cell line expressing firefly luciferase was created by transducing Granta-519 cells with the G-RV-FFLUC-Thy1.1-Neo plasmid as described previously (31).

Ramos cells were concentrated and plated at a density of 10⁶ cells/well in a 96-well plate. Cells were blocked with nonbinding antibody (negative control, HB8181), or an anti-CD20 antibody (1F5) at 50× the concentration of FPs and incubated for 45 minutes on ice. The FPs were added at 14× saturation of the CD20-binding sites, incubated for 45 minutes on ice, washed, and incubated with a 0.5× saturation ⁹⁰Y-DOTA–biotin for 45 minutes. Cells were washed to remove unbound ⁹⁰Y-DOTA-biotin and cell pellets counted on a Perkin Elmer Wizard 2480 gamma counter. The percent bound radioactivity was calculated as bound CPM divided by applied total CPM.

Athymic female mice (6–8 weeks old; Harlan Sprague-Dawley) were housed in the Fred Hutchinson Cancer Research Center animal facility according to the Institutional Animal Care and Use Committee. All mice were placed on a biotin-deficient diet (Purina Mills) 7 days prior to PRIT studies. Ramos, Granta-519, or Granta-519^luc cells (10⁷) were injected subcutaneously in the right flank 7–14 days prior to experiments to produce lymphoma xenografts measuring 6 to 8 mm in diameter. Anti-asialoGM1 antiserum (30 μL, WAKO) was injected intraperitoneally 8 days and 3 days prior to FP injection, and weekly for 100 days to prevent spontaneous tumor regressions.

Groups of 3–5 athymic nude mice bearing Ramos or Granta-519 flank tumors were injected intravenously with 0.14 to 2.8 nmol of 2H7-Fc-C825 followed 23 hours later by either 5.8 nmol of NAGB or 5–32 μg of DOTAY-dextran (DYD) CA. Two micrograms of ⁹⁰Y-DOTA-biotin were injected 1 hour later. Retro-orbital blood sampling was performed at serial time points up to 24 hours. ⁹⁰Y was counted on a calibrated gamma counter and the percent of the injected dose per gram (%ID/g) present in blood was calculated.

Groups of 3–5 mice with similar-sized tumors were injected intravenously with 0.14 to 2.8 nmol of 2H7-Fc-C825, CC49-Fc-C825, HB8181-SA, or 1F5-SA. Twenty-three hours later, mice were injected with 5.8 nmol of NAGB CA (50 μg) or 5 μg of DYD CA, followed 1 hour later by 1.2 nmol DOTA-biotin labeled with 20 to 40 μCi (0.74–1.48 MBq) of ⁹⁰Y. Blood, tumors, and body organs were obtained and ⁹⁰Y activity was measured using a calibrated system.

Mean values of radioactivity in organs and tissues for groups animals in biodistribution studies were plotted against time after injection and integrated to determine the total number of radioactive disintegrations in each major organ or tissue. Time–activity curves were integrated through complete decay of ⁹⁰Y. Radiation-absorbed doses to organs and tissues were calculated from the integrated time–activity curves using the method of Hui, which accounts for organ mass, specific absorbed energy fraction, the beta-emission spectrum of ⁹⁰Y, and the beta-particle absorbed fractions for small organs (32). Average carcass values were estimated by blending the remains after tissue harvest. The results were expressed as radiation absorbed dose (cGy) per unit administered activity (μCi).

The therapeutic efficacies of ⁹⁰Y-PRIT using SA-biotin and bispecific methods were evaluated in groups of 10 mice at each dose level. Groups of mice with similar sized, palpable tumors were randomized for the studies. Mice were given 1.4 or 2.8 nmol of 2H7-Fc-C825, 1F5-SA, or the nonbinding, negative control antibodies, CC49-FC-C825 or HB8181-SA, followed by optimized doses of CA (5.8 nmol for NAGB or 5 μg DYD) 23 hours later. A single dose of 1.2 nmol of DOTA-biotin labeled with 14.8-37 MBq (400–1,000 μCi) ⁹⁰Y was administered 1 hour after the CA. Mice were assessed every 2 days for tumor volume, weight changes, and general appearance. Blood was sampled before therapy and 5, 12, 20, 30, 50, and 150 days after ⁹⁰Y administration to measure the leukocyte and platelet counts as well as hemoglobin and hematocrit values. Serum was collected before therapy, and 12, 20, 30, 50 and 150 days after therapy, to assess changes in aspartate aminotransferase (AST), alanine aminotransferase (ALT), creatinine, and blood urea nitrogen (BUN) levels. Bioluminescent tumor imaging was performed on mice bearing Granta-519^luc xenografts. Mice were injected intraperitoneally with 10 μL/g (15 mg/mL) d-luciferin (Caliper Lifesciences) and imaged 10 minutes later on an IVIS Spectrum device. Mice were euthanized if xenografts exceeded 1,200 mm³, caused obvious discomfort or impaired ambulation, or if mice lost more than 30% of their baseline body weight.

Differences between treatment groups in absorbed radiation dose and blood counts were determined using Student t tests. Tumor growth rate differences were determined using repeated measures ANOVA, and survival differences were plotted by the Kaplan–Meier method (33) and compared using log-rank tests. Analyses were performed using GraphPad Prism 6 and JMP 12.2.0 (SAS Institute).

A bispecific antibody recognizing the B-cell–specific CD20 surface antigen as well the Yttrium-DOTA ligand was prepared by fusing the cDNAs encoding the scFv domain of the 2H7 anti-CD20 mAb with the scFv of the affinity-enhanced C825 antibody, joined by a human IgG1 CH2-CH3 Fc hinge region and an NLG linker, as described above and illustrated in Fig. 2A. The fusion gene construct was transfected into CHO-DG44 cells, and high-expressing clones were selected with methotrexate. The expressed 80-kDa monomeric protein spontaneously dimerized to a 160-kDa product, that was purified from culture supernatants using a protein A column and characterized by SDS-PAGE (Fig. 2B). The purified bispecific antibody bound avidly to CD20-expressing target cells [i.e., EL4 murine T cells transfected with human CD20, (Fig. 2C) and human lymphoma cell lines such as Ramos and Granta (Supplementary Fig. S1)], but not to control cells (EL4) lacking CD20 expression (Fig. 2C). The 2H7-Fc-C825 FP also avidly bound the Y-DOTA ligand in a concentration-dependent fashion, as demonstrated in a sandwich ELISA assay (Fig. 2D).

Pharmacokinetic analysis of the 2H7-Fc-C825 FP was challenged by an inability to radioiodinate the bispecific molecule without impairing its binding function. To circumvent this limitation, in vivo pharmacokinetics and blood clearance were performed by first injecting the unlabeled 2H7-Fc-C825 FP with or without the administration of various doses of DYD CA 23 hours later, followed by injection of 2.4 nmol of ⁹⁰Y-DOTA-biotin 1 hour later. Blood was collected at various time intervals (0.08, 0.25, 0.5, 1, 2, 4, and 24 hours) following injection of radioactivity. The pharmacokinetics of various doses of bispecific antibody (0.14, 0.35, 0.7, 1.4, or 2.8 nmol) labeled with ⁹⁰Y-DOTA-biotin in the absence of CA are shown in Supplementary Fig. S2. Dose-dependent levels of circulating radioimmunoconjugate were observed ranging from 14.7 ± 1.4 %ID/g (23 hours after 2.8 nmol bispecific Ab) to 0.6 ± 0.4%ID/g ID/g (23 hours after 0.14 nmol Ab), with a very slow blood clearance in the absence of DYD CA (t_1/2 beta of ∼44 hours with 1.4 nmol of 2H7-Fc-C825). In contrast, DYD CA was highly effective at rapidly clearing >98% of circulating FP from the bloodstream within 30 minutes of administration at all doses tested (Fig. 3A).

The biodistributions of various doses (0.14, 0.35, 0.7, 1.4, or 2.8 nmol) of 2H7-Fc-C825 bispecific Ab followed by trace-labeled ⁹⁰Y-DOTA were initially assessed in the blood, Ramos tumor xenografts, and in critical normal organs in the absence of CA. Dose-dependent accumulation of radioactivity (up to 11.9%ID/g) was demonstrated in tumor sites 48 hours after injection of 2.8 nmol of bispecific Ab (23 hours after injection of ⁹⁰Y-DOTA; Supplementary Fig. S3) and subsequent experiments optimized the system by assessing the impact of various doses of DYD CA. Athymic mice were injected with 1.4 nmol of 2H7-Fc-C825 followed 23 hours later by 0, 5, 16, or 32 μg of DYD CA, followed 1 hour later by 2.4 nmol of ⁹⁰Y-DOTA-biotin. The greatest amount of radioactivity was present in the tumors of animals that received no CA (5.7 ± 0.6%ID/g, Fig. 3B). However, high levels of radioactivity were also observed in the blood (5.6 ± 0.6%ID/g), lungs (2.9 ± 0.1%ID/g), liver (1.3 ± 0.1%ID/g), and kidneys (1.6 ± 0.2%ID/g, unless DYD was administered (Fig. 3B). All three doses of CA (5, 16, and 32 μg) were highly effective in removing excess FP from the bloodstream (∼98% clearance within 30 minutes) and in markedly diminishing the uptake of the ⁹⁰Y-radiolabeled construct in critical normal organs (Fig. 3B), although the amount of radioactivity in tumor sites also dropped progressively with increasing amounts of CA. Hence, the lowest amount of CA (5 μg) was selected as optimal for further experiments as it afforded the best compromise between effective clearance of bispecific Ab from the blood stream and normal organs, while retaining high levels of radioactivity in target tumor sites (Fig. 3B).

An experiment performed to evaluate the retention of radioactivity in tumor and normal organ sites as a function of time (Fig. 4A) demonstrated selective targeting and retention of ⁹⁰Y in CD20-expressing tumor sites with high levels of radioactivity achieved within 4 hours of ⁹⁰Y-DOTA injection (13.8 ± 1.4%ID/g), which persisted after 24 hours (10.4 ± 1.5%ID/g), 48 hours (6.12 ± 0.5%ID/g), and 120 hours (2.1 ± 0.3%ID/g). Tumor-to-blood radioactivity ratios were 6:1 after 4 hours, 21:1 after 24 hours, 51:1 after 48 hours and 43:1 after 120 hours. Among normal organs, the lung retained the greatest amount of radioactivity (Fig. 4A), presumably because of its high vascularity. Kidneys and liver retained much less radioactivity after 24 hours, (0.49 ± 0.03%ID/g) and (0.53 ± 0.05%ID/g), respectively, with tumor-to-organ ratios of 21:1 and 19:1, respectively. A nonbinding CC49-Fc-C825 control FP showed no preferential retention in tumor sites, with a lower %ID/g in tumor than in lungs at all time points (Fig. 4B). A comparative experiment was subsequently performed to assess the relative targeting of ⁹⁰Y-DOTA-biotin to tumor sites using bispecific 2H7-Fc-C825 PRIT compared with our previous “gold standard” PRIT approach using an anti-CD20-SA conjugate (1F5-SA). Virtually identical tumor targeting was observed with these two PRIT methods (8.4 ± .2%ID/g vs. 8.2 ± 1.0%ID/g, respectively after 24 hours; Fig. 4C). However, the concentrations of ⁹⁰Y in the blood and normal organs were consistently higher with 1F-SA PRIT than with bispecific 2H7-Fc-C825 PRIT (e.g., for blood, 2.1 ± 0.5%ID/g vs. 0.6 ± 0.1%; and for kidneys, 2.1 ± 0.4%ID/g vs. 0.6 ± 0.1%, respectively). Consequently, tumor-to-organ ratios were considerably superior for bispecific PRIT compared with SA-biotin PRIT (e.g., tumor:blood, 14.8 ± 3.5 vs. 4.4 ± 1.6, P = 0.0003; tumor:kidney, 15.3 ± 2.6 vs. 4.2 ± 1.1, P < 0.0001, respectively). Very similar results were obtained in a separate experiment targeting the Granta xenografts (Fig. 4D).

Radiation-absorbed doses to tumors, whole body, and 11 normal tissues were calculated for bispecific antibody PRIT using the methods described above (Table 1). The dosimetry method used accounted for organ self-dose absorbed fractions as well as beta-particle cross-organ dose contributions (32). The absorbed doses (cGy) per millicurie ⁹⁰Y administered, calculated from the ¹¹¹In tracer, showed tumor-to-normal organ ratios of 112:1 for the femur, 29:1 for the whole body, 25:1 for the kidneys, 21:1 for the blood, 6:1 for the liver, and 3:1 for the lung (Table 1).

Therapy studies were performed in athymic mice (n = 10 mice per group) bearing subcutaneous Ramos or Granta-519^luc xenografts, using the optimal reagent concentrations and time points for administration identified in the biodistribution studies. Four consecutive, concordant experiments demonstrated similar high levels of efficacy between bispecific antibody PRIT compared with SA-biotin PRIT when administered at equimolar doses (1.4 nmol). In one representative experiment, escalating amounts of ⁹⁰Y-DOTA-biotin [14.8, 25.9, or 37 MBq (400, 700, or 1,000 μCi)] were administered to mice bearing Ramos xenografts injected 24 hours earlier with 1.4 nmol of 2H7-Fc-C825 bispecific Ab, 1F5-SA conjugate, or control reagents (CC49-Fc-C825 bispecific Ab or HB8181-SA conjugate). Dose-dependent inhibition of tumor growth (Fig. 5A) and prolongation of mouse survival (Fig. 5B) were dramatic with both CD20-pretargeted methods. All 30 mice in the three control groups and the 10 mice treated with 14.8 MBq (400 μCi) ⁹⁰Y-DOTA-biotin group died within 25 days, whereas 60%–70% of mice treated with 25.9 MBq (700 μCi) or 37 MBq (1,000 μCi) of ⁹⁰Y-DOTA-biotin were cured, surviving tumor-free for the 150-day duration of the experiment (Fig. 5A and B, P < 0.0001, comparing any control or the 14.8 MBq group to either the 25.9 or 37 MBq group). The 1F5-SA conjugate possesses four binding sites for ⁹⁰Y-DOTA-biotin, whereas the 2H7-Fc-C825 bispecific Ab possess only two binding sites per molecule for ⁹⁰Y-DOTA-biotin. A subsequent experiment was therefore performed evaluating the potential benefit of administering twice the amount of 2H7-Fc-C825 Ab as 1F5-SA conjugate (2.8 vs. 1.4 nmol), followed by equivalent amounts of ⁹⁰Y-DOTA-biotin [37 MBq (1,000 μCi), Fig. 5C and D]. In this experiment, 100% of animals treated with 2.8 nmol of 2H7-Fc-C825 were cured compared with 50% of those with 1.4 nmol of 1F5-SA (Fig. 5D, P = 0.01), while treatment with 1.4 nmol 2H7-Fc-C825 treatment was no better than with 1F5-SA (P = 0.24). In this experiment, all 40 mice in 4 control groups died of progressive tumor growth before day 25 (Fig. 5D).

Similarly encouraging efficacy of bispecific PRIT was demonstrated in an experiment in mice-bearing Granta-519^luc xenografts where 90%–100% of mice were cured with 25.9–37 MBq (700–1000 μCi) of ⁹⁰Y-DOTA-biotin (Fig. 6A–C), sharply contrasting with the rapid tumor growth and death observed in all 20 mice in two control groups. (Fig. 6A and B, P < 0.0001).

PRIT was well tolerated at doses up to 37 MBq (1,000 μCi) of ⁹⁰Y-DOTA-biotin, with negligible weight loss using either bispecific Ab or SA-biotin methods (Supplementary Fig. S4). No significant evidence of renal or hepatic toxicity was observed by monitoring serial transaminase, alkaline phosphatase, blood urea nitrogen, or creatinine (not shown). Myelosuppression was minimal with bispecific Ab PRIT with very minor changes in the leukocyte, platelet, or red blood cell counts (Supplementary Fig. S5). In contrast, SA-biotin PRIT produced more marked myleosuppression in three toxicity experiments. With SA-biotin PRIT, nadir WBC counts dropped to 10% of baseline, while with bispecific PRIT, the WBC dropped to 49% of baseline (P < 0.0001). Hematocrits dropped to 52% of baseline with SA-biotin PRIT compared with 86% of baseline with bispecific PRIT (P < 0.0001), and platelet counts dropped to 22% of baseline with SA-biotin PRIT compared with no decrement in platelets (100% of baseline) with bispecific PRIT (P < 0.0001, Supplementary Fig. S5).

Despite rapid advancements in the management of B-cell malignancies using chemotherapeutic agents, mAbs, RIT, and kinase inhibitors, approximately 25% of patients with NHL still succumb to the disease (1). Anti-CD20 RIT is an underutilized modality that has a unique mechanism of action and the potential to eradicate disease in patients who are refractory to other forms of treatment when the therapeutic ratio is optimized. RIT with beta-particle emitting radioisotopes such as ⁹⁰Y kills tumor cells by inflicting multiple single strand (and to a lesser degree, double strand) DNA breaks. Multi-step PRIT is designed to overcome limitations associated with conventional directly radiolabeled antibody by facilitating dose escalation and enhancing efficacy. Many investigators have demonstrated the dramatic superiority of PRIT compared with conventional RIT in preclinical models (34–39). However, no comparative studies have been published rigorously comparing competing PRIT technologies to provide objective insight into the approach that should be preferentially translated into human clinical trials. This article provides for the first time, to our knowledge, head-to-head comparisons of the two most widely used systems for pretargeting, namely SA-biotin and bispecific Ab PRIT. The SA-biotin approach has been widely tested and demonstrates striking efficacy, but has been criticized because SA is a highly immunogenic bacterial protein that may limit the ability to administer repeated cycles of therapy (18, 40, 41). This potential limitation may be mitigated, however, in patients with hematologic malignancies, as their immunocompromised status may render them incapable of responding to foreign immunogens. A second potential limitation of SA-biotin PRIT is the presence of endogenous biotin in the blood and tissues of patients, which theoretically could occupy SA-binding sites, blocking binding of subsequently administered radio-biotin compounds. Several approaches have been proposed to circumvent the disadvantages of SA-biotin PRIT, including the genetic engineering of less immunogenic versions of SA (42, 43) and of mutant SA molecules with a lower binding avidity for endogenous biotin, which nevertheless retain high avidity binding for synthetic, divalent radio-biotin ligands (44–46). Despite these refinements to the SA-biotin PRIT approach, alternative methodologies are desirable. Bispecific Ab PRIT approaches appear particularly attractive, as they can easily be generated with nonimmunogenic human or humanized antibodies and are not impacted by endogenous biotin.

In the studies described here, SA-biotin and bispecific approaches were both well tolerated, with minimal weight loss and no evidence of toxicity to normal organs in animals monitored for >150 days after ⁹⁰Y construct infusions up to 37 MBq (1,000 μCi). Initial biodistribution experiments demonstrated tumor-to-kidney uptake ratios that were 5:1 24 hours after ⁹⁰Y-DOTA-biotin pretargeted by 1F5-SA (not shown). This finding was consistent with previous biodistribution experiments reported by our group demonstrating the kidney as the normal organ with the highest nonspecific radiation uptake after 1F5-SA-biotin PRIT (tumor-to-kidney ratios as low as 6:1 at 24 hours after the radiolabeled moiety was infused; ref. 47). While no evidence of long-term renal toxicity has been observed in SA-biotin PRIT therapy studies with infusions up to 44.4 MBq (1,200 μCi; refs. 48, 49), the kidney has represented the organ that would most likely define dose-limiting toxicity in a SA-biotin PRIT clinical trial. In contrast, the biodistribution studies using 2H7-Fc-C825 PRIT described here demonstrate a tumor-to-kidney ratio of 21:1 after 24 hours (Fig. 4A). This relative renal sparing may facilitate dose escalation in future clinical studies.

Myelosuppression was significantly more pronounced with SA-biotin PRIT (Supplementary Fig. S5), presumably a consequence of the somewhat higher levels of blood radioactivity (Fig. 4C and D) compared with that of bispecific PRIT. At day 14, the mice receiving 37 MBq of ⁹⁰Y pretargeted by 2H7-Fc-C825 FP had WBC, HCT, and platelet counts that were 49%, 100%, and 86% of baseline, respectively (Supplementary Fig. S5) while those receiving the same amount of ⁹⁰Y using the SA-biotin approach had counts that were 10%, 22%, and 52% of their baseline values (P < 0.0001 for all comparisons to FP counts). In clinical settings, reduced myelosuppression may translate into lower rates of infection, fewer bleeding complications, and improved outcomes. Moreover, while the incidence of secondary malignancies, including myelodysplastic syndrome, after conventional anti-CD20 RIT is not greater than that observed following other forms of chemotherapy (50) [3.5% following ¹³¹I-tositumomab (50); 2.5% following ⁹⁰Y-ibritumomab tiuxetan (51)], the relative sparing of bone marrow from nonspecific radiation exposure after 2H7-Fc-C825 FP PRIT may reassure treating clinicians concerned about treatment-associated myelotoxicity.

A potential limitation to PRIT treatment regimens is the inherent complexity that is unavoidable in a multi-step process. A potential advantage associated with bispecific PRIT is the relative simplicity of CA synthesis. Production of the DYD CA (28) involves straightforward chemistry that facilitates inexpensive large-scale production. In contrast, the SA-biotin system requires a synthetic N-acetylgalactosamine CA consisting of a biotin joined through a modified aminocaproyl spacer to a core of a fourth-generation dendrimeric backbone (39) and as consequence, production is more complex and costly.

Taken together, the results presented in this report confirm the highly specific targeting of radioactivity with both PRIT methods, although tumor-to-normal organ radioactivity ratios were superior with the bispecific approach. Therapeutic efficacy was similar with bispecific Ab and SA-biotin PRIT when the targeting first-step reagents were used at equimolar doses (1.4 nmol), with 50%–100% of animals cured with 37 MBq (1,000 μCi) of ⁹⁰Y in four consecutive experiments using two different lymphoma xenograft models (e.g., Fig. 5B). When the dose of 2H7-Fc-C825 was doubled to 2.8 nmol to adjust for the difference in valency between the 1F5-SA conjugate (tetravalent binding to ⁹⁰Y-DOTA-biotin) and 2H7-Fc-C825 (bivalent binding to ⁹⁰Y-DOTA), the therapeutic efficacy was superior with the bispecific approach (Fig. 5D).

In conclusion, both SA-biotin PRIT and bispecific Ab PRIT are safe and highly effective methods of curing mice bearing B-cell lymphoma xenografts, but the bispecific Ab approach is preferred for future clinical trials because of a superior biodistribution profile, less myelosuppression, and improved efficacy when used at equivalent valency.

S.H.L. Frost is a team leader at Roche Innovation Center Zürich. K.D. Orcutt is a senior director translational research at InviCRO and has ownership interest (including patents) in a patent. No potential conflicts of interest were disclosed by the other authors.

Conception and design: D.J. Green, S.H.L. Frost, A.L. Kenoyer, A.K. Gopal, O.W. Press

Development of methodology: D.J. Green, Y. Lin, D.K. Hamlin, S.H.L. Frost, A.L. Kenoyer, A.K. Gopal, K.D. Wittrup, O.W. Press

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): D.J. Green, S.L. Frayo, Y. Lin, D.K. Hamlin, A.L. Kenoyer, M.D. Hylarides, J.J. Orozco, O.W. Press

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): D.J. Green, S.L. Frayo, D.K. Hamlin, S.H.L. Frost, A.K. Gopal, T.A. Gooley, J.J. Orozco, S. O'Steen, O.W. Press

Writing, review, and/or revision of the manuscript: D.J. Green, S.L. Frayo, Y. Lin, S.H.L. Frost, A.K. Gopal, B.G. Till, S. O'Steen, K.D. Orcutt, D.S. Wilbur, O.W. Press

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): S.L. Frayo, K.D. Orcutt

Study supervision: A.K. Gopal

This work was supported by grants from the US NIH NCI-K08CA151682 (D.J. Green); NCI-R01CA076287 (O.W. Press), NCI-R01CA136639 (O.W. Press), NCI-R01CA154897 (O.W. Press), NCI-K23CA154874 (B.G. Till), K24CA184039 (A.K. Gopal), and by the David and Patricia Giuliani Family Foundation.

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.