Therapy of Myeloid Leukemia using Novel Bispecific Fusion Proteins Targeting CD45 and ⁹⁰Y-DOTA

Despite recent advances, therapies for acute myeloid leukemia (AML) most often result in poor outcomes, with 25% or less of patients alive 5 years after diagnosis (1). Moreover, the increased incidence of AML among older patients often makes treatment challenging and limits the ability to deliver intensive therapy. Selectively targeting therapeutic radionuclides to malignant cells may address these challenges by improving treatment efficacy while reducing associated toxicity (2–4). Radiolabeled antibodies appear particularly promising for treating AML, given the highly radiosensitive nature of disease and that many leukemia antigens have been well-characterized (5–7). CD45 has been an important target for radioimmunotherapy (RIT) of hematologic malignancies as this antigen is highly expressed on the surface of nearly all hematopoietic cells, but has limited expression on nonhematopoietic tissues (8). Because of the targeted radiation to hematopoietic cells, anti-CD45 RIT has been clinically studied in the context of hematopoietic cell transplantation (HCT); RIT targeting CD45 followed by HCT has resulted in more than 40% survival at 1 year among patients with AML who were largely considered ineligible for standard HCT studies (9–12).

Despite the potential of improved outcomes with RIT for patients with high-risk disease, challenges remain for using an antibody directly conjugated to a radionuclide. Prior to target localization, the circulating radioimmunoconjugate results in radiation exposures in nontargeted cells. To address this problem, two-step pretargeted RIT (PRIT) approaches have been developed to separate the delivery of the radionuclide from the delivery of the antibody. One PRIT method has employed first-step as unlabeled antibody conjugated to streptavidin (SA) delivered to target cells. After 24 to 48 hours, to allow for maximal accumulation at target sites, unbound Ab-SA conjugate can be cleared from circulation by infusion of a clearing agent. The radiolabeled reagent, DOTA-biotin, that has a high affinity for the pretargeted Ab-SA conjugate can subsequently be delivered. The radiolabeled DOTA-biotin ligand binds quickly to streptavidin localized at disease target sites, and unbound radiolabeled DOTA-biotin is readily excreted from the body because of its smaller size, minimizing nonspecific radiation exposure (13–16). This PRIT approach has demonstrated significantly improved biodistribution of the therapeutic radionuclide compared with directly labeled radioimmunoconjugates (17, 18). However, the large intact antibody may hinder tumor penetration in different clinical scenarios, while interference from endogenous biotin and immunogenicity from streptavidin raise concerns that could limit the effectiveness of this strategy. We have, therefore, developed new step 1 reagents, bispecific antibodies that bind both CD45 and yttrium-DOTA (Y-DOTA) as an alternative to SA-biotin PRIT. In this article, we show that these bispecific constructs effectively deliver radiation to CD45⁺ cells and improve survival in murine leukemia models.

The anti-murine CD45 and anti-Y-DOTA bispecific antibody were produced starting with two Pfuse Plasmid Constructs (Invitrogen), pFUSE2ss-CHIg-hG1 and pFUSE2ss-CLIg-hK, carrying the heavy and light chain genes, respectively, of the anti-murine CD45 30F11 antibody (19). The light chain construct also has DNA encoding for the variable heavy and light chains of the C825 single-chain variable fragment (scFv) radio metal trap cloned downstream of the anti-murine CD45 light chain. Additional bispecific antibody expression, production, and purification details are described in Supplementary Materials and Methods (available online). Control bispecific antibodies, LDL-Fc (targeting LDL, but without the C825 scFv radiometal trap for Y-DOTA), and CC49-Fc-C825 (targeting the irrelevant adenocarcinoma antigens, TAG-72 and Y-DOTA), were generated as described elsewhere (20).

A gene for the bispecific fusion protein targeting human CD45 and Y-DOTA was constructed following a different bispecific antibody design to facilitate production. The anti-CD45 and anti-Y-DOTA scFv fusion gene coding for BC8-Fc-C825 was produced as described in Supplementary Materials and Methods (available online). For comparison with SA-biotin PRIT, anti-CD45 BC8-SA and nontargeting anti-bovine herpes virus-SA (BHV1-SA) conjugates were generated as described previously (17).

HEL (human AML), Ramos (human Burkitt lymphoma), or EL4 cells (murine T-cell lymphoma) were purchased from the ATCC and used at low passage number (less than 8 weeks), while murine leukemia SJL cells were harvested from serially transplanted mice as described previously (21, 22). Cell lines were tested for pathogens by PCR (IMPACT III Profile) at IDEXX. Female athymic Foxn1^nu mice (Envigo), 7–10 weeks of age, were used in HEL xenograft studies. Female B6SJLF1/J mice (The Jackson Laboratory), 8–12 weeks of age, were used in SJL (murine AML) disseminated syngeneic studies. Mice were housed at the Fred Hutchinson Cancer Research Center (Seattle, Washington) Animal Care Facility in a pathogen-free environment under protocols approved by the Institutional Animal Care and Use Committee. Study mice were placed on a diet containing uniprim (trimethoprim/sulfadiazine) antibiotic (irradiated, 4,100 ppm from Harlan Laboratories).

DOTA-biotin was produced and radiolabeled as described previously (17). Where experiments called for different radioactivity amounts, the same amount of DOTA-biotin was used to maintain consistent DOTA-biotin concentrations. Variable-specific activities were pursued to maintain similar biodistribution that was largely impacted by the concentrations of reagents. Although bispecific fusion proteins bind Y-DOTA, we used ⁹⁰Y-DOTA-biotin as the radioligand for both bispecific antibody and Ab-SA approaches to maintain similar circulation time for the radioligand and facilitate comparative studies between the two approaches. Radiolabeled product was assessed for radiochemical purity via an Avidin-Bead Assay (Sigma) and injected at >90% purity. The sarcosyl-biotin (NAGB) clearing agent used for Ab-SA groups (23) and the DOTAY-dextran clearing agent used for bispecific antibody groups were synthesized as described previously (20, 24).

Binding of bispecific antibody to target cells was confirmed by flow cytometry (20). PE-conjugated-SA secondary reagent was used to detect binding of ⁹⁰Y-DOTA-biotin to bispecific antibody bound to murine CD45⁺ EL4. Fluorophore-labeled anti-IgG1 secondary antibody was used to detect BC8-Fc-C825 fusion protein binding to human CD45⁺ Ramos cells. Blocking antibody (30F11 or BC8) was used in excess (100:1 or 10:1) before assaying with secondary antibody to show binding specificity by flow cytometry. Bispecific BC8-Fc-C825 or negative control bispecific fusion protein, LDL-Fc, was applied to BSA-Y-DOTA conjugate–coated (1 μg/mL in PBS) 96-well plates at 16 μg/mL and serial dilutions for ELISA capture studies (20).

All mice were placed on a biotin-free diet for at least 5 days prior to injections of first-step reagents for all biodistribution and therapy studies to minimize the impact of endogenous biotin. Groups of 5 athymic mice received 1 × 10⁷ HEL cells subcutaneously to the right flank. Approximately 7 days later, as human xenograft AML tumors became palpable (∼50–100 mm³), mice received either intravenous 1.4 or 2.8 nmol BC8-Fc-C825, 1.4 nmol BC8-SA, or 1.4 nmol CC49-Fc-C825. Different amounts of bispecific fusion protein and Ab-SA constructs were evaluated in anti-human CD45 studies to maintain equivalent radioligand binding capacity, as the Ab-SA construct can potentially bind up to four biotin-DOTA ligands, whereas the bispecific fusion protein can only bind up to two Y-DOTA-biotin ligands. Twenty-two hours after the first-step reagent injection, mice received intraperitoneal DOTAY-dextran (5 μg, 9.1 pmol) or 50 μg (5.8 nmol) NAGB clearing agent to remove unbound circulating antibody, followed 2 hours later with 1.85 MBq (50 μCi) ⁹⁰Y-DOTA-biotin. For clearance studies, mice were bled via the retro-orbital plexus at serial timepoints (5 minutes to 20 hours). Blood radioactivity was assessed on a Packard Cobra II Gamma Counter (PerkinElmer) and percent injected dose per gram of tissue (% ID/g) was calculated for each sample. For biodistribution studies, normal organs and tumors were harvested at 4, 24, 48, and 96 hours after injection of ⁹⁰Y-DOTA-biotin. Biodistribution data collected at four timepoints were used in standard Medical Internal Radiation Dose Methods modeling (25), from which radiation absorbed doses for organs were calculated as described previously (26–28).

For studies targeting murine CD45, groups of 5 B6SJLF1/J mice received intravenous 1 × 10⁵ SJL murine AML cells and 24 hours later were given intravenous 1.4 nmol 30F11-IgG1-C825 or CC49-Fc-C825. Twenty-two hours after injection of each fusion protein, 5 μg (9.1 pmol) DOTAY-dextran clearing agent was administered, and 2 hours after administration of clearing agent, mice were given 1.85 MBq (50 μCi) ⁹⁰Y-DOTA-biotin. Tissues were harvested 6 and 24 hours after injection of DOTA-biotin, and gamma counts were obtained for all harvested tissues (18, 19) from which % ID/g was calculated.

In syngeneic disseminated murine CD45⁺ leukemia studies, groups of 10 B6SJLF1/J mice were given intravenous 1 × 10⁵ SJL cells, and 24 hours later were injected with 1.4 nmol 30F11-IgG1-C825 or CC49-Fc-C825. For human leukemia subcutaneous xenograft models, therapy studies used groups of 10 athymic mice that were subcutaneously administered with 1 × 10⁷ HEL cells expressing human CD45. When tumors were approximately 50–100 mm³, mice were then given intravenous 1.4 nmol of BC8-Fc-C825, CC49-Fc-C825, BC8-SA, or BHV1-SA. Twenty-two hours after injection of conjugate or fusion protein, mice received 50 μg (5.8 nmol) NAGB clearing agent (Ab-SA–treated mice) or 5 μg (9.1 pmol) DOTAY-dextran (bispecific antibody–treated mice), followed 2 hours later by 29.6–55.5 MBq (800–1,500 μCi) ⁹⁰Y-DOTA-biotin. For all studies, weights of mice and tumor volumes were monitored with electronic calipers up to 170 days postinjection.

Biodistribution studies yielded % ID/g for each timepoint, with the mean and SD error bars from the 5 mice per group graphed in figures. Two-sample t tests were used to compare tumor uptake (% ID/g) at different organs. Survival was graphed with the survival data tables, and comparisons were made with log-rank test using Prism (GraphPad Software).

Bispecific antibody (30F11-IgG1-C825), recognizing murine CD45 and Y-DOTA as a mAb fusion protein, was produced by cloning the regions coding for the light chain variable region (V_L) and heavy chain variable region (V_H) fragments of the Y-DOTA capturing C825 disulfide–stabilized (ds-scFv) scFv gene onto the light chain of the parent anti-murine CD45 antibody, 30F11 (Fig. 1A). The vectors carrying the light or heavy chain sequences (Fig. 1B) were cotransfected into HEK293T cells and bispecific antibody was purified from supernatant. Functional binding of the 30F11-IgG1-C825 bispecific antibody was verified by incubating the fusion protein with mCD45⁺ EL4 cells, followed by Y-DOTA-biotin, with or without 30F11 blocking in excess, and then assayed for capture of Y-DOTA-biotin by flow cytometric analysis with fluorophore-labeled-SA secondary reagent (Fig. 1C). 30F11-IgG1-C825 showed binding to CD45⁺ cells and Y-DOTA-biotin as detected by secondary antibody. Binding of 30F11-IgG1C825 to target cells could be blocked by preincubating with 30F11 (Fig. 1C). CD45⁺ EL4 cells incubated with Y-DOTA-biotin alone (no secondary) or incubated with just 30F11 without streptavidin conjugation did not bind any secondary antibody as expected. Thus, constructs were functional as they bound to both CD45⁺ target cells and Y-DOTA.

Having verified functional binding of 30F11-IgG1-C825 to both mCD45⁺ cells and Y-DOTA in vitro, biodistribution studies were performed using mice modeling disseminated minimal residual disease leukemia. Groups of 5 mice per timepoint were given 1.4 nmol of 30F11-IgG1-C825 or CC49-Fc-C825 as a nontargeting bispecific control 2 days after SJL tumor cells were injected, followed 22 hours later by 5 μg (9.1 pmol) DOTAY-dextran clearing agent. Mice were then given 1.85 MBq (50 μCi) ⁹⁰Y-DOTA-biotin 24 hours after injection of the bispecific reagents, and murine tissues were harvested at 6 and 24 hours after ⁹⁰Y-DOTA-biotin. CD45⁺ target spleen and marrow tissues exhibited highest specific uptake by 6 hours after injection of ⁹⁰Y-DOTA-biotin (21.7 ± 6.7 and 10.3 ± 1.5% ID/g, respectively) and was still detected at the 24-hour timepoint (9.0 ± 1.5 and 8.1 ± 1.2% ID/g, respectively; Fig. 1D). All nontarget organs, including kidneys and lungs, had lower uptake (<0.5% ID/g) at each timepoint. There was minimal uptake in any tissue at either timepoint for the control nontargeting bispecific antibody, CC49-Fc-C825. Because the anti-murine CD45 biodistribution studies only had two timepoints, no absorbed radiation doses were calculated.

After establishing the specific biodistribution of 30F11-IgG1-C825 to mCD45⁺ tissues, we performed therapeutic studies using the 30F11-IgG1-C825 bispecific antibody construct in a syngeneic, disseminated murine leukemia model. Groups of 10 mice received intravenous 1 × 10⁵ SJL murine leukemia cells, followed 2 days later by 1.4 nmol 30F11-IgG1-C825 or nontargeting control bispecific antibody, CC49-Fc-C825. Twenty-two hours after construct injection, mice were given 5 μg (9.1 pmol) DOTAY-dextran clearing agent, followed by either 29.6, 44.4, or 55.5 MBq (800, 1,200, or 1,500 μCi) ⁹⁰Y-DOTA-biotin. These activity doses were chosen based on our prior PRIT studies using activities more than 37 MBq (1,000 μCi) ⁹⁰Y with significant toxicity, and early deaths observed at activities more than 44.4 MBq (1,200 μCi) ⁹⁰Y (29). SJL leukemia–bearing mice treated with 30F11-IgG1-C825 and 29.6 MBq (800 μCi) of ⁹⁰Y-DOTA-biotin had a therapeutic benefit, with a median overall survival of 43 days, compared with untreated disease control mice (median 30 days; P < 0.0001; Fig. 1E). All mice treated with 29.6 MBq (800 μCi) ⁹⁰Y-DOTA-biotin eventually died from leukemia, with splenomegaly observed on necropsy (spleen weight, 0.4–0.6 g and normal spleen weight ≤0.1 g). In contrast, mice treated with the higher doses of 44.4 or 55.5 MBq (1,200 or 1,500 μCi) ⁹⁰Y-DOTA-biotin experienced substantial treatment-related toxicity, as assessed by weight loss below euthanasia threshold, lethargy, and anemia with pale extremities and thrombocytopenia manifested by subcutaneous hemorrhage on necropsy. Median survival for mice treated with 30F11-IgG1-C825 followed by 44.4 or 55.5 MBq (1,200 and 1,500 μCi) of ⁹⁰Y-DOTA-biotin was 13 and 17 days, respectively.

We also designed an anti-human CD45 bispecific fusion protein for assessment in a clinically relevant model of human leukemia (Fig. 2A). A bispecific fusion protein, BC8-Fc-C825, recognizing both human CD45 and Y-DOTA, was produced by engineering the region coding for the scFv of the BC8 antibody and the ds-scFv of the DOTAY-specific C825 antibody onto an hIgG1 Fc hinge (Fig. 2B). This bispecific design was pursued to simplify production by using a single recombinant fusion protein, instead of the murine 30F11-IgG1-C825 antibody fusion protein that requires expression of separate plasmids for heavy and light chains. We attempted to produce the murine bispecific antibody in scFv-Fc-scFv format used for BC8-Fc-C825, but in the sc-Fv-Fc-scFv format, the murine CD45 bispecific did not express well despite multiple optimization attempts. After production and purification of the fusion protein, functional binding of BC8-Fc-C825 fusion protein was assessed via flow cytometry for binding to CD45-expressing Ramos human Burkitt lymphoma cells (Fig. 2C). Ramos cells alone or those incubated with BC8 antibody only did not show any detectable binding of BC8-Fc-C825 by fluorophore-labeled-anti-IgG secondary antibody. BC8 is a murine-derived antibody and thus, is not recognized by secondary antibody. Binding of BC8-Fc-C825 fusion protein to Ramos cells was blocked by incubating with competing BC8 antibody at 10:1 (Fig. 2C), given the common epitope specificity. Functional recognition of Y-DOTA was confirmed by increasing amounts of bound BC8-Fc-C825 serially diluted onto a BSA-Y-DOTA–coated plate, whereas the negative control bispecific fusion protein (LDL-Fc lacking the C825 Y-DOTA moiety) was not captured in ELISA at any Y-DOTA concentration tested (Fig. 2D).

A more established PRIT method has employed the SA-biotin approach. We thus, performed comparative blood clearance studies with BC8-Fc-C825 bispecific antibody and BC8-SA. After bispecific antibody infusion, DOTAY-dextran clearing agent was infused that functions via the nonradiolabeled Y-DOTA moiety binding to the C825 portion of circulating unbound bispecific antibody. Removal of unbound BC8-Fc-C825 was studied by clearance assessment of the radiolabeled second step ⁹⁰Y-DOTA. A cohort of 5 athymic mice received 1.4 nmol of BC8-Fc-C825 per timepoint, with or without 5 μg DOTAY-dextran clearing agent delivered 22 hours later, followed 2 hours later by 1.85 MBq (50 μCi) ⁹⁰Y-DOTA-biotin. A separate group of 5 athymic mice received 1.4 nmol BC8-SA per timepoint and 22 hours later were given 50 μg (5.8 nmol) NAGB clearing agent, and subsequently received 1.85 MBq (50 μCi) ⁹⁰Y-DOTA-biotin 3 hours later. Blood was obtained at 5, 10, 15, and 30 minutes, and 1, 2, 4, and 20 hours after injection of ⁹⁰Y-DOTA-biotin. Clearance of BC8-Fc-C825 without DOTAY-dextran clearing agent was similar to BC8-SA clearance when clearing agent was employed, with 0.4 ± 0.2% ID/g and 0.7 ± 0.5% ID/g remaining in the blood 20 hours after delivery of the ⁹⁰Y-DOTA-biotin, respectively (Fig. 3). Importantly, BC8-Fc-C825 administration followed by delivery of the clearing agent led to faster clearance, with only 0.1 ± 0.02% ID/g persisting in circulation at the 20-hour timepoint.

Biodistribution and dosimetry studies with ⁹⁰Y-DOTA-biotin were performed to compare tissue distribution of BC8-Fc-C825 bispecific antibody with BC8-SA and the nonbinding bispecific control, CC49-Fc-C825. Because blood clearance of circulating BC8-Fc-C825 was improved with clearing agent, subsequent studies included the intermediate step with clearing agent injections. To find an appropriate antibody dose, either 1.4 or 2.8 nmol of BC8-Fc-C825, control bispecific antibody, or BC8-SA was injected into groups of 5 athymic mice harboring palpable human leukemia HEL tumor xenografts. Twenty-two hours later, 5 μg (9.1 pmol) of DOTAY-dextran clearing agent (or NAGB clearing agent for BC8-SA groups) was delivered to remove unbound bispecific antibody, then followed 2 hours later with 1.85 MBq (50 μCi) ⁹⁰Y-DOTA-biotin. Tissue was harvested 4, 24, 48, and 96 hours after injection of ⁹⁰Y-DOTA-biotin. As early as 4 hours after ⁹⁰Y-DOTA injection, uptake by the leukemia tumor xenograft was nearly 75% of maximal uptake in the group receiving 1.4 nmol of BC8-Fc-C825 (5.7 ± 0.01% ID/g), with peak uptake 24 hours after ⁹⁰Y-DOTA-biotin injection (7.8 ± 0.02 % ID/g; Fig. 4A), consistent with the kinetics of a relatively small radioligand. Uptake in nontarget tissues was minimal, with peak uptake of 0.6 ± 0.3% ID/g in the kidneys seen at the 24-hour timepoint. Mice that received a higher amount (2.8 nmol) of BC8-Fc-C825 also had rapid tumor uptake, peaking at 6.3 ± 0.01% ID/g 24 hours after ⁹⁰Y-DOTA-biotin injection, with relatively low uptake seen in the kidneys (1.0 ± 0.3% ID/g) at the same timepoint (Fig. 4B). Nonbinding bispecific antibody, CC49-Fc-C825, showed minimal uptake in all tissues at each timepoint assessed (Fig. 4C). The group receiving BC8-SA prior to 50 μg (5.8 nmol) NAGB clearing agent, followed by ⁹⁰Y-DOTA-biotin, displayed similar rapid uptake in tumor, compared with the BC8-Fc-C825 groups, with minimal uptake in nontarget tissues. BC8-SA achieved peak tumor uptake of 6.7 ± 0.01% ID/g at 24 hours after ⁹⁰Y-DOTA-biotin injection compared with the peak uptake of 1.4 ± 0.5% ID/g in the kidneys seen 4 hours after ⁹⁰Y-DOTA-biotin injection (Fig. 4D).

Dosimetry calculations were conducted to determine absorbed radiation doses to individual tissues after administration of the pretargeted first-step fusion protein followed by clearing agent and ⁹⁰Y-DOTA-biotin. Using BC8-Fc-C825 for PRIT maximized the radiation doses per unit administered activity to human CD45⁺ leukemia xenografts (3.9 and 3.5 cGy per μCi ⁹⁰Y-DOTA-biotin injected with 1.4 and 2.8 nmol of BC8-Fc-C825, respectively), which was comparable with doses delivered to tumor with BC8-SA (4.7 cGy/μCi; Table 1). As expected, absorbed doses achieved in tumor (0.4 cGy/μCi) with the nontargeting bispecific antibody, CC49-Fc-C825, were similar to that seen in the blood (0.6 cGy/μCi). Tumor-to-normal organ dose ratios using 1.4 nmol of BC8-Fc-C825 were 7.2 for liver, 15.9 for kidney, and 49.8 for the whole body. Radiation targeted to tumor by bispecific antibody was greater than that achieved in BC8-SA–treated mice that resulted in tumor-to-normal-organ dose ratios of 5.2 for liver, 7.8 for kidney, and 23.4 for total body. Tumor-to-blood ratios of absorbed radiation doses were 6.7 for BC8-Fc-C825 (at 1.4 nmol) and 8.8 for BC8-SA.

We performed therapy studies using the anti-human CD45 and anti-Y-DOTA bispecific construct, BC8-Fc-C825, in a human leukemia xenograft model. Groups of 10 mice bearing palpable subcutaneous HEL xenografts were treated with 1.4 nmol of either BHV1-SA, CC49-Fc-C825 (nonbinding Ab-SA and bispecific antibody, respectively), BC8-SA, or BC8-Fc-C825 first-step constructs. Twenty-two hours after administration of the antibody conjugates, the mice were injected with a NAGB clearing agent for the Ab-SA groups or a DOTAY-dextran clearing agent for bispecific antibody treatment groups. Approximately 2 hours after injection of the clearing agent, mice were given 51.8 MBq (1,400 μCi) ⁹⁰Y-DOTA-biotin and followed for tumor response and toxicity. The initial starting activity dose was chosen, as we have previously used, up to 59.2 MBq (1600 μCi) ⁹⁰Y-DOTA-biotin with anti-CD45 PRIT to treat HEL xenografts (30). Mice treated with BC8-Fc-C825 had a significant reduction in tumor growth compared with control groups (Fig. 5A). Three of 10 mice treated with BC8-SA and 4 of 10 mice treated with BC8-Fc-C825 died from toxicity by day 28 after ⁹⁰Y-DOTA injections with severe weight loss and lethargy. The median survival for BC8-SA–treated mice was 112 days, and the median survival was 67 days for BC8-Fc-C825–treated mice (Fig. 5B). There was no statistically significant difference in long-term survival between mice treated with BC8-SA or BC8-Fc-C825 at 180 days after injection of ⁹⁰Y-DOTA (P = 0.38; Fig. 5B). Mice that received BC8-Fc-C825 had a significantly better median survival of 67 days compared with mice that received irrelevant control CC49-Fc-C825, which had a median survival of 13 days (P = 0.024). Groups of untreated mice and those that received CC49-Fc-C825 were all euthanized because of excessive tumor growth by day 27 after ⁹⁰Y-DOTA-biotin injection (Fig. 5B).

We refined the ⁹⁰Y-DOTA-biotin radioactivity to maintain the efficacy of the PRIT approach and minimize the toxicity from delivery of high doses of the therapeutic radionuclide by evaluating different radioactivity doses. Mice bearing subcutaneous xenografts received 1.4 nmol of BC8-Fc-C825 or CC49-Fc-C825 bispecific fusion protein, followed by DOTAY-dextran clearing agent 22 hours later. Mice were given 37, 44.4, or 51.8 MBq (1,000, 1,200, or 1,400 μCi) of ⁹⁰Y-DOTA-biotin 24 hours after injection of each bispecific construct. After treatment, tumors were undetectable at all three doses of ⁹⁰Y-DOTA-biotin in BC8-Fc-C825–treated groups (Fig. 5C). By day 170 after ⁹⁰Y-DOTA-biotin injection, no mice were euthanized because of tumor growth in any of the BC8-Fc-C825 treatment groups. Although each BC8-Fc-C825 treatment group had early deaths due to excessive weight loss, 60% of mice in the 51.8 MBq (1,400 μCi) group and 70% of mice in the 44.4 and 37 MBq (1,200 and 1,000 μCi) groups survived disease free at least 170 days after ⁹⁰Y-DOTA-biotin injection (Fig. 5D). Observed toxicities included excessive weight loss and lethargy. To address the possibility of hematologic toxicities, optimization experiments had mice receive syngeneic hematopoietic stem cells (1 × 10⁷ i.v.) 3 days after ⁹⁰Y-DOTA-biotin (31) without any change in early toxicity rates. Untreated mice and mice that received nontargeting bispecific antibody, CC49-Fc-C825, were all euthanized because of tumor growth by days 26 and 32, respectively.

Our studies with preclinical leukemia models suggest that bispecific antibody constructs targeting CD45 and Y-DOTA are effective therapeutic agents for myeloid leukemia. First, biodistribution studies showed that targeting using both murine and human bispecific constructs exhibited the highest uptake in tissues with corresponding CD45 antigen expression, including high uptake of anti-human CD45 BC8-Fc-C825 in the HEL human tumor xenografts and anti-murine CD45 30F11-IgG1-C825 in the spleen and marrow of animals with disseminated murine leukemia. Second, dosimetry calculations for BC8-Fc-C825 confirmed expected high absorbed doses delivered to HEL xenograft tumors. Finally, BC8-Fc-C825 was therapeutically as effective as the SA-biotin PRIT approach in HEL tumor–bearing mice (Fig. 5D). More importantly, while different fusion protein formats were used for human (scFv-Fc-scFv) and murine (Ig-scFv) leukemias, the major conclusions remain; bispecific antibodies, targeting CD45 and ⁹⁰Y-labeled radioligand treated leukemia. The results with bispecific antibody are consistent with prior studies targeting sites of hematologic malignancies while sparing delivery of therapeutic radiation to normal organs via other SA-biotin PRIT strategies in various models (20, 29, 32, 33). The complimentary models from these studies were utilized to examine the potential of the new bispecific antibody in very straight forward models of HEL xenografts, where human CD45 target is only on the xenograft cells, and more clinically relevant SJL model, where relevant murine CD45 on-target, off-leukemia delivery of radiation could be assessed. The small therapeutic effect in the group treated with the nonbinding irrelevant control, CC49, at high doses was likely due to the exquisite radiosensitivity of hematologic malignancies to high radiation doses, although the nonspecific effect was not significant as there were no long-term survivors in this group. These properties and potential advantages of the bispecific constructs suggest promise for translation to the clinic.

The favorable biodistribution observed with lower activity of ⁹⁰Y-DOTA-biotin is likely preserved at therapeutic doses. We have shown stable binding, targeting, and efficacy in hematologic malignancy models treated with PRIT employing up to 59.2 MBq (1,600 μCi) ⁹⁰Y-DOTA-biotin (24, 30). Therapy studies use the same amount of first-step antibody conjugate as in biodistributions for each respective construct, and any unbound, circulating first step reagent is removed by clearing agent. One would then expect constant distribution of first-step reagent in both biodistribution and therapy studies in mice given the same amount of first-step antibody conjugates. In addition, while the amount of radioactivity dosed may be different, the total amount of DOTA-biotin that was radiolabeled was the same for their biodistribution and therapy studies, to minimize any second-step concentration differences altering the biodistribution.

Binding affinities of anti-CD45 and anti-DOTA moieties may not have heavily influenced favorable biodistributions given uptake at target organs. The picomolar affinity of C825 to DOTA chelates (34, 35), and the nanomolar binding affinity of BC8 to human CD45 (36), have been characterized in different contexts, but it would be difficult to gauge how binding affinities change when incorporated into the bispecific fusion protein without repeat affinity measurements. The fusion protein backbone is not likely to alter biodistribution as much as target specificity, for which human IgG1 for the hinge or connection regions was engineered into the BC8 bispecific fusion protein given the impetus to minimize immunogenicity of carrier proteins.

Although treatment with bispecific antibody and SA-biotin PRIT resulted in comparable survival, the fusion protein resulted in better tumor size control, posing the notion that bispecific antibody may be more effective than SA-biotin PRIT. These studies were neither designed nor powered to discern any potential differences between bispecific and SA-biotin anti-CD45 PRIT approaches. However, a similarly constructed bispecific antibody targeting CD38 and ⁹⁰Y-DOTA for treatment of multiple myeloma and non-Hodgkin lymphoma in mice resulted in more convincing data that bispecific antibody may be more effective than Ab-SA PRIT (20). While the size, Fc effector function, and design of the bispecific antibody all have the potential to impact penetration capabilities in solid tumors, none of these studies were tailored to address these variables. In addition, these variables may not be as impactful for leukemia control, where clearance may be more relevant for the nonsolid tumor leukemia setting.

One option to improve leukemia outcomes might be to strategically employ PRIT approaches in the appropriate disease setting. In these studies, ⁹⁰Y-based PRIT with bispecific antibody to treat leukemia in mice with subcutaneous HEL human xenografts resulted in long-term survivors, while mice with disseminated syngeneic disease experienced an improvement in overall survival without long-term cures. These results support the premise that bulky disease settings are best treated with radionuclides with relatively longer effective path lengths, like the beta-emitter ⁹⁰Y with an effective path length of 2.7 mm. On the other hand, minimal residual disease settings with isolated tumor cells and clusters of disease may best be treated with radionuclides with shorter effective path lengths from alpha-emitters. This helps explain why some of the hematologic toxicities observed at the higher doses (44.4 MBq or 1,200+ μCi) were not seen in the subcutaneous models, as the radioimmunoconjugates were sequestered away from radiosensitive organs, such as the marrow.

Another possibility to improve survival rates in preclinical models of disseminated leukemia could be to increase the amount of radioactivity delivered, but this may be difficult to achieve given the toxicity observed at necropsy. As the primary objective of these studies was to assess therapeutic efficacy of the novel agents, we did not include formal toxicity studies. Because the radioligand in PRIT is more quickly cleared compared with directly labeled antibody, off-target radiation exposure may be less than directly labeled RIT. ⁹⁰Y-anti-CD45 RIT targeting murine leukemia via 30F11 in an immunocompetent mouse model showed only transient myelosuppression that normalized by week 4 after RIT injection, without any impact on renal or hepatic function (28), although milder effects were seen with directly labeled RIT targeting ⁹⁰Y or ¹⁷⁷Lu to CD20 (29). PRIT with bispecific antibody was also likely limited by myelotoxicity given subcutaneous hemorrhages and anemic organs observed at necropsy from mice with disseminated leukemia dying within 2 weeks of treatment with the highest doses of ⁹⁰Y. Myelotoxicity was also appreciated when utilizing higher doses of SA-biotin PRIT and bispecific RIT approaches for solid tumor malignancies (37–39). Furthermore, when ⁹⁰Y-anti-CD45 RIT was clinically used before reduced intensity HCT for patients with high-risk AML or myelodysplastic syndrome, there was minimal off-target toxicity as no grade 4 nonhematologic toxicity was observed (12). Other published trials using ¹³¹I-anti-CD45 BC8 RIT prior to allogeneic HCT also support precise radiation delivery to target organs, without significant toxicity to other normal organs (10, 11). Formal comparative toxicity and immunogenicity studies of bispecific antibody and Ab-SA conjugates thus will be pursued in future studies.

In summary, these studies utilized two different bispecific constructs, a monoclonal antibody fusion protein for the murine CD45 (30F11-IgG1-C825) leukemia model, and a Fc fusion protein in the human CD45 (BC8-Fc-C825) model; both showed a therapeutic effect. Bispecific antibody that target CD45 and Y-DOTA may be considered for future treatment of AML.

J.J. Orozco reports grants from NCI (K01 CA 234623), NCI (R01 CA138720), NCI (P01 CA009515), NCI (R01 CA076287), NCI (R01 CA154897), NCI (K24 CA184039), NCI (K23 CA154874), Robert Wood Johnson Foundation (RWJF-ASH AMFDP), and ASBMT (NIA), grants and other from David and Patricia Giuliani Foundation (grant) and Frederick Kullman Foundation (grant) during the conduct of the study, as well as has a patent for 62/698632 pending (for Fred Hutchinson Cancer Research Center). Y. Lin reports grants from FHCRC during the conduct of the study. E.R. Balkin reports grants from NIH during the conduct of the study and other from U.S. Dept. of Energy (DOE, current employer) outside the submitted work. K.D. Orcutt reports a patent for engineered proteins with high affinity to DOTA chelates issued. D.J. Green reports grants from NIH during the conduct of the study, other from Juno Therapeutics (clinical trial support), personal fees from Seattle Genetics, GSK, Janssen Pharmaceuticals, and Legend Biotech, and other from Bristol Myers Squibb (clinical trial support) outside the submitted work. A.K. Gopal reports grants from NIH during the conduct of the study, grants, personal fees, and nonfinancial support from Janssen, Seattle Genetics, Pfizer, Takeda, I-Mab, Astra Zeneca, Merck, and Gilead, personal fees and nonfinancial support from BMS, personal fees from ADC, Karyopharm, Actinium, TG Therapeutics, MorphoSys, Asana, Aptevo, BRIM, and Amgen, and grants and nonfinancial support from IgM outside the submitted work. B. Sandmaier reports grants from NIH during the conduct of the study, personal fees from Actinium Pharmaceuticals and Bristol-Meyers Squibb, and other from Bellicum (clinical trial support), AbbVie (spouse has consultant fees), AnaptysBio (spouse has ownership interest), OncoResponse (spouse has ownership interest), Inipharm (spouse has ownership interest), and Blaze Bioscience (spouse has ownership interest) outside the submitted work. O.W. Press reports grants from NIH during the conduct of the study and outside the submitted work. J.M. Pagel reports other from Actinium (consultant), Gilead (consultant), AstraZeneca (consultant), and BeiGene (consultant) outside the submitted work. No disclosures were reported by the other authors.

J.J. Orozco: Conceptualization, formal analysis, supervision, investigation, writing-original draft, writing-review and editing. A.L. Kenoyer: Data curation, validation, investigation, methodology. Y. Lin: Conceptualization, resources, validation. S. O'Steen: Conceptualization, data curation, formal analysis, validation, methodology, writing-review and editing. R. Guel: Resources, data curation, investigation, methodology. M.E. Nartea: Resources, data curation, investigation, methodology. A.H. Hernandez: Resources, data curation, investigation, methodology. M.D. Hylarides: Conceptualization, resources, data curation, investigation, methodology. D.R. Fisher: Resources, formal analysis. E.R. Balkin: Resources, formal analysis. D.K. Hamlin: Resources, data curation, methodology. D.S. Wilbur: Conceptualization, resources. K.D. Orcutt: Resources. K.D. Wittrup: Resources. D.J. Green: Conceptualization, data curation, formal analysis. A.K. Gopal: Conceptualization, data curation, formal analysis. B.G. Till: Conceptualization, resources. B. Sandmaier: Conceptualization, resources, data curation, supervision. O.W. Press: Conceptualization, resources, supervision, funding acquisition. J.M. Pagel: Conceptualization, resources, data curation, formal analysis, funding acquisition, writing-original draft, writing-review and editing.

This work was supported by grants from NCI K01 CA 234623 (to J.J. Orozco), R01 CA138720 (to J.M. Pagel), P01 CA009515 (to O.W. Press), R01 CA076287 (to O.W. Press), R01 CA154897 (to O.W. Press), K24 CA184039 (to A.K. Gopal), and K23 CA154874 (to B.G. Till); RWJF-ASH AMFDP and NIA from ASBMT (to J.J. Orozco), the David and Patricia Giuliani Foundation (to O.W. Press), and the Frederick Kullman Foundation (to J.M. Pagel).

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.