Leveraging Bioorthogonal Click Chemistry to Improve ²²⁵Ac-Radioimmunotherapy of Pancreatic Ductal Adenocarcinoma

Alpha-emitting radionuclides are of great interest in nuclear medicine due to their short particle range (50–100 μm) and high linear energy transfer (∼80 keV/μm), traits that allow for the delivery of highly cytotoxic radiation to cancer cells (1, 2). Actinium-225 (²²⁵Ac; t_1/2 = 10.0 d; 5.8 MeV α) has attracted a large share of this attention due to its long physical half-life and its “nanogenerator” status, producing a total of 4 α and 2 β⁻ particles in its decay chain (Fig. 1; ref. 3). These advantageous properties were explored for the treatment of acute myeloid leukemia in phase I/II clinical trials with a CD33-targeted ²²⁵Ac-radioimmunoconjugate (HuM195; Lintuzumab), studies that underscored the potential of ²²⁵Ac-radioimmunotherapy (RIT) for the treatment of systemic disease (4–6). In addition, recent studies highlighted the potential of ²²⁵Ac-RIT for the treatment of bulky tumors such as glioblastoma by targeting the vascular endothelium of the tumor (7, 8). The radionuclide has also been conjugated to targeting vectors with more rapid pharmacokinetic profiles, including small-molecule inhibitors of prostate-specific membrane antigen (PSMA). To wit, salvation therapy with ²²⁵Ac-PSMA-617 (100 kBq/bimonthly) has produced drastic complete response in patients with metastatic castration-resistant prostate cancer (9, 10). Yet although ²²⁵Ac-based radiotherapy is undeniably promising, its success is restricted by toxicities associated with the irradiation of healthy tissues following target engagement (on-target toxicities; e.g., xerostomia after ²²⁵Ac-PSMA-617 therapy) or the nonspecific accumulation of the radiopharmaceutical (off-target toxicities; e.g., clearance organs). Compromises between antitumor activity and toxicity must be made (10). There is therefore an urgent unmet need for new strategies that can reduce the inherent toxicity of ²²⁵Ac-RIT.

The inverse electron-demand Diels–Alder reaction (IEDDA) between tetrazine (Tz) and trans-cyclooctene (TCO) holds promise for a variety of applications, including radiochemistry (11, 12). Our laboratories reported its potential for the synthesis of ²²⁵Ac-labeled radioimmunoconjugates (13). The selectivity and rapidity of this reaction were investigated for in vivo pretargeting using TCO-bearing antibodies and Tz-containing radioligands (14). This approach consists of 4 steps: (i) the administration of a TCO-bearing immunoconjugate; (ii) an interval period during which the immunoconjugate accumulates at the tumor and clears from the blood pool; (iii) the administration of a Tz-based radioligand; and (iv) the in vivo click reaction between the two components followed by the rapid clearance of excess radioligand. In essence, pretargeting seeks to combine the targeting properties of antibodies with the pharmacokinetic properties of small molecules. Pretargeted SPECT and PET imaging experiments yielded improved tumor-to-background activity concentration ratios as well as reduced absorbed radiation doses to normal tissues (15–17). This approach was successfully applied to the pretargeted radioimmunotherapy (PRIT) of both pancreatic ductal adenocarcinoma (PDAC) and colorectal carcinoma with ¹⁷⁷Lu, a β⁻-emitting radionuclide (18, 19).

The central hypothesis of this study is that leveraging the Tz/TCO click ligation for ²²⁵Ac-PRIT could ultimately mitigate toxicities while improving therapeutic indices. Herein, we report the use of the IEDDA reaction for ²²⁵Ac-PRIT targeting carbohydrate antigen 19.9 (CA19.9; ref. 20), a cell surface antigen overexpressed in PDAC. CA19.9 is one of the most widely studied pancreatic cancer biomarkers because it can be detected in patient serum. In addition, levels of serum CA19.9 can be correlated with the stage of the cancer as well as prognosis or treatment response (21–23). With a 5-year survival rate lower than 8%, PDAC is uniformly lethal and lacks therapeutic options. The fully-human 5B1 antibody demonstrates exceptional affinity and high specificity for an extracellular epitope of CA19.9. Staining with 5B1 revealed a negative reactivity for most normal tissues, whereas samples of pancreatic tumors showed positive reactivity with a diffuse cytoplasmic staining and distinct cell membrane staining (24). Early in 2016, the clinical potential of CA19.9 as a therapeutic target was investigated in patients with pancreatic cancers or CA19.9-positive malignancies (NCT02672917) via the weekly intravenous administration of 5B1 (MVT-5873). Two clinical trials with 5B1-based radioimmunoconjugates are currently ongoing: one in which the antibody is labeled with ⁸⁹Zr for PET imaging and a second in which the antibody is labeled with ¹⁷⁷Lu for RIT (NCT02687230 and NCT03118349). Therefore, the development of a 5B1-based approach to ²²⁵Ac-PRIT could offer a new therapeutic approach for PDAC and has the potential to offer an interesting complementary perspective to these clinical investigations.

Details on antibody functionalization, radiolabeling, cell lines, and xenografts models are provided in the Supplementary Information. All animals were treated according to the guidelines approved by the Research Animal Resource Center and Institutional Animal Care and Use Committee at Memorial Sloan Kettering Cancer Center.

Acute biodistribution studies were performed using healthy athymic nude mice bearing subcutaneous BxPC3 (CA19.9-positive) and MIAPaCa-2 (CA19.9-negative) xenografts (right flank, ∼100 mm³). For the pretargeting strategy, ²²⁵Ac-DOTA-PEG₇-Tz or ²²⁵Ac-DO3A-PEG₇-Tz [18.5 kBq, 0.4 nmol (molar activity, MA = 46 kBq/nmol); in 150 μL of 0.9 % NaCl + 1.0 % BSA] were injected 72 hours following the administration of 5B1-TCO (200 μg; 1.32 nmol; in 200 μL of PBS). BSA is used in the formulation of the radioligands to prevent adsorption to the syringe plastic. Mice (n = 5) were euthanized via CO₂ asphyxiation, and tissues of interest were collected. Each sample was counted for up to 10 min (24 hours after collection when secular equilibrium was reached) using a Wizard² automatic gamma counter set up to a 150 to 600 keV energy window. The calculation of %ID/g values is detailed in the Supplementary Information.

Imaging studies were performed on athymic nude mice bearing BxPC3 and MIAPaCa-2 tumors on the right and left flanks, respectively. Mice received an initial injection of 5B1-TCO (200 μg; in 200 μL of PBS) 72 hours prior the injection of ²²⁵Ac-DOTA-PEG₇-Tz [1.85 MBq, 26.0 nmol (MA = 71 kBq/nmol); in 150 μL of 0.9 % NaCl + 1.0 % BSA]. In this case, a large amount of the radiolabelled Tz was injected simply to achieve high Cerenkov signal. Mice were imaged up to 4 days after the injection of the radioligand and then euthanized. Organs were imaged ex vivo at two time points: 5 minutes after the organ collection and once secular equilibrium was reached (24 hours after sacrifice). Image acquisition, analysis, and quantification are described in the Supplementary Information.

Athymic nude mice xenografted with BxPC3 tumors were injected with either ²²⁵Ac-DOTA-PEG₇-5B1 [18.5 kBq, 8.6 μg (MA = 308 kBq/nmol); in 150 μL of 0.9 % NaCl + 1.0 % BSA] or ²²⁵Ac-DOTA-PEG₇-Tz [18.5 kBq, 0.4 nmol (MA = 46 kBq/nmol); in 150 μL of 0.9 % NaCl + 1.0 % BSA] at 72 hours post administration of 5B1-TCO (200 μg; in 200 μL of PBS). Mice were euthanized 1, 5, and 7 days after injection of the radiopharmaceutical. Mouse kidneys and tumors were embedded and sectioned during the 20 minutes that followed the animal sacrifice. Kidney and tumor sections were placed in a film cassette against a phosphor imaging plate (Fujifilm BAS-MS2325) for a first exposure time of 1 hour at −20°C. A second exposure (6 hours) with the same sections was performed once secular equilibrium was reached (24 hours after first exposure). Kidney sections were later stained with hematoxylin and eosin (H&E) for the differentiation of the cortex and medulla. Image analysis and quantification is described in the Supplementary Information.

Dosimetry calculations are described in the Supplementary Information. Two hypotheses were used to take into account the redistribution of the ²²⁵Ac daughters. In the first, ²²⁵Ac and daughters decay at the site of ²²⁵Ac decay (no redistribution). In this case, the total mean organ absorbed dose is then calculated as the sum of the mean absorbed doses calculated for ²²⁵Ac, and each daughter multiplied by the weighting factor for the relevant decay branch (i.e., 2.2% for ²⁰⁹Tl, 97.8% for ²¹³Po, and 100% for all others). In the second hypothesis, we postulate that the ²²¹Fr and ²¹⁷At daughters decay at the site of ²²⁵Ac decay due to their short physical half-lives, but that the extended half-life of ²¹³Bi allows significant redistribution. Here, as a conservative (worst-case) dose projection, the whole-body residence time of the daughters from ²¹³Bi to ²⁰⁹Pb is assumed to occur entirely in the kidneys. The total mean absorbed dose in all organs except the kidneys is calculated as the sum of mean absorbed dose calculated for ²²⁵Ac and its daughters up to ²¹⁷At. The total mean absorbed dose in the kidneys is calculated as stated previously.

Tumor volumes were monitored with a Peira TM900 imaging device (Peira, Belgium). After the initial tumor measurement, mice were randomized into cohorts (n = 8) so that the mean tumor volume of each cohort was approximately equal. The study was then blinded so that the volume measurements were not biased. One day after randomization, mice receiving PRIT were injected with 5B1-TCO (200 μg; in 200 μL of PBS). Seventy-two hours later, mice were injected with the appropriate radioactive payload: either ²²⁵Ac-DOTA-PEG₇-5B1 (9.25, 18.5, or 37 kBq) for conventional RIT or ²²⁵Ac-DOTA-PEG₇-Tz (9.25, 18.5, or 37 kBq) for PRIT. The control cohorts were injected with either vehicle (0.9 % sterile saline), 5B1-TCO alone (200 μg; in 200 μL of PBS), ²²⁵Ac-DOTA-PEG₇-Tz alone (37 kBq), or ²²⁵Ac-DOTA-PEG₇-IgG (37 kBq; IgG from human serum, Sigma). Tumor volume and body weight were measured biweekly until 120 days or until mice reached the set volume endpoint (>2,000 mm³). Mice were monitored for outward signs of toxicity, including lethargy, loss of appetite, or disseminated intravascular coagulation.

Pancreatic tumor burden was monitored by in vivo bioluminescence imaging (BLI). Fifteen days after implantation, the radiance in each tumor was measured. Mice were randomized in five different cohorts (n = 10) so that the average radiance was almost equal between the cohorts. One day after randomization, mice receiving pretargeted RIT were injected with 5B1-TCO (200 μg; in 200 μL of PBS). Seventy-two hours later, mice were injected with the appropriate radioactive payload; either ²²⁵Ac-DOTA-PEG₇-5B1 (18.5 or 37 kBq) for conventional RIT or ²²⁵Ac-DOTA-PEG₇-Tz (18.5 or 37 kBq) for PRIT. The control cohort was injected with vehicle (0.9 % sterile saline). Tumor radiance and body weight were measured weekly until 103 days. Mice were monitored for outward signs of toxicity, including lethargy, loss of appetite, or disseminated intravascular coagulation. Mice were sacrificed once the tumor burden covered >50 % of the abdomen or when mice showed significant weight loss.

Blood samples (∼50–100 μL, n = 3–5) were collected weekly via retroorbital blood draws and analyzed with an Hemavet 950 (Drew Scientific). Twenty parameters were recorded, including red blood cell count, hematocrit, white blood count, and platelet counts. Mice kidneys were collected when the mice reached endpoint and kept in 10% neutral-buffered formalin at room temperature. Once the therapeutic study was closed, kidneys were processed routinely for histology, embedded in paraffin, sectioned, stained with H&E, and submitted to a board-certified veterinary pathologist for evaluation.

All data are represented as mean value ± standard deviation (with n = 5–10, unless otherwise noted). The sample sizes were selected taking into account both statistical consideration and the exigencies of funding. The significance analyses were performed using GraphPad Prism software 7.0, employing unpaired two-tailed t tests, multiple t tests, and logrank Mantel–Cox test as detailed in figure legends. A P value of <0.05 was considered significant. When appropriate, a conservative correction (Bonferroni) was made for the multiple hypotheses being tested to control the false-positive rate and provide confidence that positive findings are true signals.

To evaluate the potential of ²²⁵Ac-PRIT, we first performed biodistribution studies and compared the in vivo data with that of a directly-labeled radioimmunoconjugate (Fig. 2A). For conventional RIT, 5B1 was radiolabeled with ²²⁵Ac using a two-step protocol published by our groups (13). For the pretargeting technique, 5B1 was first functionalized with TCO via the incubation of the antibody with an NHS-bearing variant of TCO (18) and then injected 72 hours prior to the administration of an ²²⁵Ac-labeled Tz radioligand: ²²⁵Ac-DOTA-PEG₇-Tz (Fig. 2A).

The biodistribution of ²²⁵Ac-DOTA-PEG₇-Tz was assessed in healthy nude mice without the preinjection of 5B1-TCO (Supplementary Fig. S1A). As expected, the radioligand is rapidly excreted. The liver (3.7 ± 0.6 %ID/g) and kidneys (1.9 ± 0.2 %ID/g) display the highest activity concentrations 4 hours after administration. A comparison between the biodistribution of the directly-labeled ²²⁵Ac-DOTA-PEG₇-5B1 (18.5 kBq, 0.06 nmol) radioimmunoconjugate and the 5B1-TCO/²²⁵Ac-DOTA-PEG₇-Tz (1 nmol/18.5kBq, 0.4 nmol) pretargeting approach was performed in mice bearing subcutaneous BxPC3 (CA19.9-positive) and MIAPaCa-2 (CA19.9-negative) xenografts (Fig. 2B; Supplementary Fig. S1C and S1D). Both methods produce high activity concentrations in the CA19.9-positive tumors up to 10 days after injection of the ²²⁵Ac-constructs. At early time points, the pretargeting approach results in lower tumoral uptake (4.6 ± 3.3 %ID/g; 4 hours after injection of ²²⁵Ac-DOTA-PEG₇-Tz) compared with the directly labeled antibody (15.4 ± 3.5 %ID/g; 4 hours after injection of ²²⁵Ac-DOTA-PEG₇-5B1). However, the in vivo ligation between circulating 5B1-TCO immunoconjugates and ²²⁵Ac-DOTA-PEG₇-Tz allows for the accumulation of ²⁵⁵Ac-DOTA-PEG₇-5B1 conjugates in the blood and their subsequent delivery to the tumor site. Consequently, 3 days following the administration of the ²²⁵Ac-labeled constructs, the tumoral uptake values of the two methods are comparable: 31.1 ± 21.4 %ID/g for ²²⁵Ac-DOTA-PEG₇-5B1 and 29.6 ± 6.6 %ID/g for 5B1-TCO/²²⁵Ac-DOTA-PEG₇-Tz. Importantly, no significant decrease in tumor volume was observed over the course of the biodistribution experiments, confirming that any changes in activity concentration were the result of the differential uptake of the radiopharmaceuticals themselves and not changes in the tumor size (Supplementary Fig. S1B). The activity concentrations in the MIAPaCa-2 xenografts are far lower compared with the BxPC3 xenografts: 5.3 ± 2.7 %ID/g for ²²⁵Ac-DOTA-PEG₇-5B1 and 2.2 ± 0.6 %ID/g for 5B1-TCO/²²⁵Ac-DOTA-PEG₇-Tz. This result highlights the specificity of the 5B1 antibody for CA19.9 and reinforces the sensitivity of pretargeting in vivo.

Critically, the PRIT approach yields significantly lower activity levels in the blood (7.7 ± 1.7 %ID/g) at 4 hours after injection compared with the directly-labeled radioimmunoconjugate (16.7 ± 2.3 %ID/g). Moreover, the PRIT strategy produces significantly lower activity concentrations in healthy organs, including the liver, spleen, and bone. To wit, the liver activity concentrations are 3-fold lower for the PRIT (1.5 ± 0.1 %ID/g; 10 days p.i.) compared with the directly-labeled antibody (5.8 ± 2.6 %ID/g; 10 days p.i.). Tumor-to-organ activity concentration ratios confirm the superior performance of PRIT (Supplementary Tables S2 and S3). A higher tumor-to-liver activity concentration ratio is observed 10 days following the administration of ²²⁵Ac-DOTA-PEG₇-Tz (27.0 ± 15.8) compared with ²²⁵Ac-DOTA-PEG₇-5B1 (5.6 ± 3.1). Similarly, higher tumor-to-spleen and tumor-to-bone ratios are noted as well. Biodistribution studies with ²²⁵Ac-labeled radioimmunoconjugates were also performed using a different chelator (DO3A; Supplementary Fig. S2A and S2B) as well as a different vector: the huA33 antibody that targets the A33 antigen, a biomarker of colorectal cancer (Supplementary Fig. S2C and S2D). Comparable biodistribution trends are observed in both cases, clearly illustrating the modularity of this approach. Our data with the 5B1 antibody are particularly promising, because the pretargeting method yields tumoral uptake values similar to those produced using the directly-labeled antibody while reducing the activity concentrations in the blood and healthy organs. This is especially compelling in the context of RIT, because these results suggest that PRIT could significantly reduce nonspecific toxicity.

Both the weight and the uptake of the radiolabeled constructs in the spleen were monitored as indicators of myelosuppression and especially leukopenia (Supplementary Fig. S3). The splenic activity concentrations for both methodologies decrease over the course of the biodistribution study, indicating good clearance of ²²⁵Ac-constructs. However, in the case of the conventional RIT, the weight of the organ significantly decreases from 4 hours to 10 days after administration (P ≤ 0.01). In other cases, decreases in the spleen size were observed in mice administered immunosuppressive treatments (25). It follows that this could be an early sign of myelosuppression in the mice treated with the conventional RIT. No statistically significant decrease in spleen weight is observed with the pretargeting strategy.

Cerenkov radiation emanates in response to the decay of a large variety of radionuclides. In a phantom imaging study evaluating the yield of Cerenkov radiation with a range of positron-, β⁻-, and α-emitting radioisotopes—including ¹⁸F, ⁶⁴Cu, ⁸⁹Zr, ¹²⁴I, ¹³¹I, and ²²⁵Ac—the latter produced the most intense radiation (26). ²²⁵Ac-based Cerenkov emissions are postulated to arise from the β-decay of three of its daughters: ²¹³Bi, ²⁰⁹Tl, and ²⁰⁹Pb. Due to the limited tissue penetration of optical-frequency photons, Cerenkov imaging (intraoperative settings aside) has limited utility with radionuclides that decay via “imageable” emissions (e.g., positrons or gamma rays); however, it can be valuable for monitoring the biodistribution of radiopharmaceuticals labeled with nuclides (e.g., ²²⁵Ac) that cannot be imaged effectively using traditional technologies. Cerenkov imaging was used to study the biodistribution of the α_vβ₃ integrin–targeted radiotherapeutic ²²⁵Ac-DOTA-(RGDyK) in a mouse model of human glioblastoma (27). Yet in the aforementioned study, Cerenkov imaging limitations were clearly on display, as the injected dose (1.9 MBq) was well over the determined MTD, and the mice were only imaged up to 24 hours after injection. Cerenkov imaging with ²²⁵Ac-radioimmunoconjugates is therefore a major challenge due to the high risk of lethal toxicity. However, the improved biodistribution profile obtained using ²²⁵Ac-PRIT allowed us to attempt Cerenkov imaging in vivo.

Mice bearing bilateral BxPC3 and MIAPaCa-2 tumor xenografts were injected with 5B1-TCO (1 nmol) and then—72 hours later—²²⁵Ac-DOTA-PEG₇-Tz (26 nmol). To ensure that sufficient signal would be observed, a high dose of ²²⁵Ac-DOTA-PEG₇-Tz was injected (1.9 MBq). Cerenkov imaging performed up to 4 days following the administration of ²²⁵Ac-DOTA-PEG₇-Tz (Fig. 3A) revealed persistent and increasing radiance (p/sec/cm²/sr) in BxPC3 tumors (10,600 ± 1,000 at 4 hours p.i. and 22,100 ± 3,600, at 96 hours p.i.), whereas the radiance in the remaining body (33,200 ± 4,000 at 4 hours p.i. and 13,400 ± 1,300 at 96 hours p.i.) and the MIAPaCa-2 xenografts (7,100 ± 1,200 at 4 hours p.i. and 3,500 ± 600 at 96 hours p.i.) both decrease over time (Supplementary Fig. S4A). Not surprisingly, Cerenkov imaging confirms the biodistribution profile of PRIT as well as the specificity for CA19.9.

Ex vivo Cerenkov imaging was performed at two time points: immediately following the sacrifice of the animals and again following the establishment of secular equilibrium. In a radioactive decay chain in which the parent radionuclide (i.e. ²²⁵Ac) has a much longer half-life than its radioactive daughter, secular equilibrium is defined as the limiting case (in time) in which the rate of decay of the daughter and parent becomes equivalent (Supplementary Fig. S5). In the case of ²²⁵Ac, secular equilibrium is reached 24 hours after disruption of the equilibrium state. Imaging immediately following the euthanasia of the animals shows higher activity concentrations of the radioligand in the BxPC3 xenografts (36,200 ± 6,900 p/sec/cm²/sr) compared with the MIAPaCa-2 tumors (4,900 ± 800 p/sec/cm²/sr; Fig. 3B). At this time point, Cerenkov radiation is also observed in the liver and kidneys. However, once secular equilibrium is reached—24 hours following euthanasia—a clear decrease in the kidneys radiance is observed. The radiation quantification in regions of interest drawn around the organs confirms that the kidneys are the only organs with a significant change in radiance once secular equilibrium is reached: 16,400 ± 2,300 p/sec/cm²/sr at 4 days after injection compared with 4,900 ± 800 p/sec/cm²/sr at secular equilibrium (P ≤ 0.001; Fig. 3C). This phenomenon results from the presence of nonequilibrium daughters in the kidneys. The strong recoil energy associated with the emission of α-particles (about 100 keV) is higher than the binding energy of any chemical bond. As a result, ²²⁵Ac decay daughters are released from the chelator and can redistribute within the body (28). Depending on the daughters' physical half-lives as well as their affinity for certain organs, this redistribution can result in toxicity to healthy tissues. In the case of ²²⁵Ac, two decay daughters—²²¹Fr and ²¹³Bi—with intrinsic affinity for the kidneys have half-lives sufficiently long to allow for their redistribution. We therefore hypothesize that the higher kidney radiance immediately following necropsy relative to that observed at secular equilibrium arises from the presence of free ²²¹Fr, ²¹³Bi, as well as their β-emitting daughters.

The evaluation of ²²⁵Ac daughters' redistribution in the different regions of the kidney was further pursued via autoradiography following the administration of both the conventional RIT and PRIT strategy. To this end, two autoradiographic exposures of kidney sections were obtained 1, 5, and 7 days after the injection of the radiopharmaceuticals. The first exposure was performed in the 20 minutes that followed euthanasia, and the second was performed 24 hours later (once secular equilibrium was reached). A qualitative comparison of the autoradiographic exposures shows stronger signal intensity in the kidney cortex at secular equilibrium compared with the first exposure, indicating the accumulation of ²²⁵Ac in this kidney area (Fig. 3D). This trend is observed for both ²²⁵Ac-DOTA-PEG₇-5B1 and 5B1-TCO/²²⁵Ac-DOTA-PEG₇-Tz. As noncoordinated actinium mainly distributes to the liver and bone (29), and unreacted ²²⁵Ac-DOTA-PEG₇-Tz is excreted in a few hours via the kidneys, we hypothesize that the ²²⁵Ac observed in the kidneys 3 days following the injection of the radioligand is present in the form of its antibody conjugate. In contrast, the qualitative analysis of the autoradiography shows a decrease in signal intensity between the two exposures in the renal medulla, highlighting the presence of nonequilibrium daughters. Quantitative analysis of these images confirms the qualitative observations (Supplementary Fig. S4B). The difference between the cortex-to-medulla mean intensity ratio at the first (2.5 ± 0.7) and second (4.1 ± 0.9) exposure of the slices obtained 7 days after injection of ²²⁵Ac-DOTA-PEG₇-Tz further reinforces the presence of nonequilibrium daughters in the medulla (Supplementary Fig. S4B). It is important to note that both administration methods lead to the same scenario: neither method stands out in term of limiting the redistribution of ²²⁵Ac's daughters. Similarly, two autoradiographs of tumor sections were collected 3 days after injection of PRIT and RIT. A qualitative analysis of the signal localization and intensity between the postmortem and the secular equilibrium exposure revealed a good correlation (Supplementary Fig. S4C). This observation highlights that within the tumor, the activity is present in the form of ²²⁵Ac and its equilibrium daughters.

The redistribution of ²²⁵Ac's daughters has obvious implications for dosimetry. However, such calculations require the quantitative assessment of the time-dependent biodistribution of each daughter, and the measurement thereof is currently met with well-recognized limitations (e.g., short physical half-life, limited detection technique for α-particle). The Committee on Medical Internal Radiation Dose has recommended that a potential “conservative (worst-case) scenario” be assessed in such circumstances (30). Here, we adopt this approach for two different hypothetical scenarios. In the first, all of ²²⁵Ac's daughters decay at the site of the original ²²⁵Ac decay. Given this hypothesis, ²²⁵Ac-DOTA-PEG₇-5B1 and 5B1-TCO/²²⁵Ac-DOTA-PEG₇-Tz produce similar mean absorbed doses in the tumor: 1377 Gy/MBq and 1994 Gy/MBq, respectively (Table 1; Supplementary Table S4). For all organs, the pretargeting method results in higher therapeutic indexes (>21). Indeed, the spleen, liver, and bone have therapeutic indexes 4, 5, and 7 times higher, respectively, with the PRIT strategy compared with the conventional RIT (Table 1). In this scenario, a 50 kBq dose of ²²⁵Ac-DOTA-PEG₇-Tz would result in a mean absorbed dose of >100 Gy at the tumor site, above the threshold for a meaningful response in the case of solid tumors (31). This injected activity would also result in a kidney mean absorbed dose of 2.8 Gy, below the radiosensitivity threshold of this organ (15–20 Gy; refs. 31, 32). With mean absorbed doses of 120.3 Gy/MBq and 82.6 Gy/MBq for RIT and PRIT, respectively, the blood is expected to be the dose-limiting organs in the case of the first hypothesis. In the second (worst-case) hypothesis, we postulate the redistribution of all the residence time contributed by ²¹³Bi and its progeny to the kidneys. Put differently, we assume that all of the ²¹³Bi nuclei immediately relocate to—and subsequently decay in—the kidneys. In this case, we notice a decrease of 32% in the tumor mean absorbed dose of both administration methods. Consequently, the kidney therapeutic index is affected as well, dropping to 0.9 and 1.9 for ²²⁵Ac-DOTA-PEG₇-5B1 and 5B1-TCO/²²⁵Ac-DOTA-PEG₇-Tz, respectively. In this scenario, the injected activity for therapeutic studies should be <18 kBq to ensure that the kidney mean absorbed dose remains below their radiosensitivity threshold (20 Gy).

The extrapolation of these data to humans consolidates the superior dosimetric profile of the pretargeting approach. Indeed, the mean absorbed doses to most major organs are lower for the pretargeting method (Table 2; Supplementary Table S5–S7). Depending on the hypothesis, the dose-limiting organ for PRIT is either the bone marrow (hypothesis #1) or the kidneys (hypothesis #2; Table 2). Assuming an MTD of 1 Sv_RBE5 for the red marrow and 27 Sv_RBE5 for the kidneys, the maximum-tolerated administered activity would be approximately 10.4 MBq for the first hypothesis and 5.3 MBq for the second hypothesis (Table 2; ref. 33). Dose-limiting organs are similar with conventional RIT (Supplementary Table S8).

The next step in our investigation was clear: longitudinal therapy studies. Based on the aforementioned dosimetry calculations, the maximum injected activity was set to 37 kBq (Table 1). With this amount of injected activity, the dose delivered to the tumor with conventional RIT and PRIT was estimated to be between 63 and 75 Gy, slightly below the threshold (100 Gy) for a meaningful response in the case of solid tumors, while simultaneously limiting kidney toxicity. Mice bearing subcutaneous BxPC3 tumors (n = 8 per cohort) were injected with 3.7, 18.5, and 37.0 kBq of conventional RIT and PRIT to establish a dose-dependent response. Control groups include mice injected with saline (vehicle), 5B1-TCO alone, ²²⁵Ac-DOTA-PEG₇-Tz alone, and ²²⁵Ac-DOTA-PEG₇-IgG. Survival analysis suggests therapeutic efficacy in cohorts receiving 18.5 and 37 kBq compared with control groups (P values < 0.01; Supplementary Fig. S6 and Supplementary Table S9). Median survival for the group treated with 37 kBq of ²²⁵Ac-DOTA-PEG₇-5B1 (107.5 days) and the cohort treated with 37 kBq of 5B1-TCO/²²⁵Ac-DOTA-PEG₇-Tz (114.5 days) suggests a similar therapeutic efficacy between conventional RIT and PRIT at the highest dose (P = 0.17; Fig. 4A; Supplementary Table S9).

The body weight progression for both groups receiving the highest dose of conventional RIT and PRIT deviates from the control group from day 3 to day 21 after the administration of the radiotherapeutic (Fig. 4B). However, the maximum average weight loss did not exceed 14%. All mice recovered their original body weight over a 3-week time period. One mouse out of the 8 treated with 37 kBq of ²²⁵Ac-DOTA-PEG₇-5B1 was euthanized after 18 days due to disseminated intravascular coagulation. The pathologic evaluation of this mouse revealed marked depletion of all the hematopoietic precursors in the bone marrow (Fig. 4C). Acute hemorrhages were observed on the level of the skin and correlate with the marked thrombocytopenia due to the absence of megakaryocytes in the bone marrow. No such observations were made with the RIT and PRIT cohorts that had been injected with the lower doses. Hematoanalysis performed on separate mice cohorts (n = 5) receiving the highest dose treatment (37 kBq) highlights the higher hematotoxicity of RIT compared with PRIT and correlates with the anticipated higher absorbed dose to the blood uncovered in the RIT mouse dosimetry. Marked decreases in hematocrit and red blood cell count are observed between days 10 and 66 with ²²⁵Ac-DOTA-PEG₇-5B1 (37 kBq); during this same period, the values for the cohorts treated with PRIT remained similar to those of the control groups (Fig. 4B). For the cohorts treated with ²²⁵Ac-DOTA-PEG₇-5B1 (37 kBq) and 5B1-TCO/²²⁵Ac-DOTA-PEG₇-Tz (37 kBq), white blood cell (WBC) counts show a maximum decrease between days 3 and 30 of up to 29% and 40% of the original value, respectively (Fig. 4B). However, the PRIT-treated mice recovered their WBC counts faster than those receiving conventional RIT: 45 days versus 66 days (Supplementary Fig. S7). Platelet counts likewise show a marked decreased in the case of conventional RIT, with a recovery to original values 66 days after treatment administration (Fig. 4B). A decrease in platelets has been shown to correlate with bone-marrow radiation absorbed dose in a study with a ¹⁷⁷Lu-labeled antibody specific for prostate-specific membrane antigen (34). In light of this, our results suggest that PRIT is a promising strategy to reduce radiation-induced myelotoxicity.

To establish this finding in a more clinically relevant tissue microenvironment, therapy studies were performed in mice inoculated orthotopically with BxPC3-luc cells. Twenty days following inoculation, the mice were treated with conventional and pretargeted RIT (18.5 kBq and 37.0 kBq). Both cohorts treated with the highest doses (37 kBq) show prolonged median survival compared with the control groups (Fig. 5A). PRIT (37 kBq) produces a significantly superior median survival of 67.5 days compared with 28.5 days for the vehicle-only control group (P < 0.001). Despite a median survival of 60.0 days, conventional RIT (37 kBq) does not demonstrate a significant prolonged survival compared with the vehicle-only control due to the loss of four animals at the beginning of the study: two from rapid tumor progression and two from disseminated intravascular coagulation. Surprisingly, treatment with the lower dose (18.5 kBq) results in significant prolonged survival (P < 0.05) in the case of conventional RIT (46 days) but not in the case of PRIT (32 days; Supplementary Fig. S8A and Supplementary Table S10). The survival of individual animals was shown to correlate with the initial bioluminescent signal (radiance) of its tumor (Fig. 5B). The mice with initial tumor radiance below 100 × 10⁷ p/sec/cm²/sr all show survival superior than 50 days, highlighting the strong therapeutic potential of ²²⁵Ac for small neoplasms.

As previously observed with the subcutaneous model, hematotoxicity was observed in the 37 kBq conventional RIT cohort. In this cohort, 2 mice out of 10 (20 %) had to be sacrificed at day 8 due to disseminated intravascular coagulation. No such observation was made in the PRIT cohorts. Complete blood count analysis confirmed a more severe decrease in the WBC and platelet counts between days 3 and 45 after the administration of conventional RIT compared with PRIT (Fig. 5C). Long-term nephrotoxicity was evaluated via the pathologic evaluation of the kidneys of the 6 longest surviving mice treated with the higher doses. For each administration method, three kidneys out of six (50%) demonstrated mild cortical tubular degenerative changes and/or evidence of minimal to mild cortical tubular necrosis (Fig. 5D; Supplementary Table S11). These changes are hypothesized to be the result of the radiopharmaceuticals, as such observations are not typically reported in mice of this strain and age. However, the lesions were minimal to mild, affected a very small proportion of the tubules, and likely did not affect renal function: there was no significant evidence of elevation in urea and creatine in the blood chemistry performed (P > 0.05; Supplementary Fig. S8B). As these observations were only reported in the kidney cortex, we hypothesize that these morphologic changes were not related to ²²⁵Ac's daughters redistribution, because it was previously demonstrated that these radionuclides primarily localize in the kidneys medulla. These pathology findings correlate with the comparable absorbed doses to the kidney obtained with both administration methods and highlight the accuracy of our dosimetry methodology.

Herein, we have demonstrated that PRIT with 5B1-TCO and ²²⁵Ac-DOTA-PEG₇-Tz is as effective as conventional RIT while simultaneously reducing hematotoxicity. Biodistribution studies carried out with both RIT and PRIT highlight the promise of ²²⁵Ac-based pretargeting in terms of tumoral uptake and clearance. Previous pretargeting studies carried out with short-lived, positron-emitting isotopes were often associated with low tumoral activity concentrations (≤ 6 %ID/g; refs. 15–17). In contrast, pretargeting with the long-lived isotope ²²⁵Ac produced tumoral activity concentrations as high as 41.7 ± 24.0 %ID/g 10 days after the radioligand administration. This increase in tumoral uptake is most likely the result of click reactions between circulating 5B1-TCO and the ²²⁵Ac-labeled Tz, forming ²²⁵Ac-labeled radioimmunoconjugates that slowly accumulate within tumor tissue. Our group has already reported similar trends in ¹⁷⁷Lu-PRIT, with tumor activity concentrations approaching 20 %ID/g (18). It is important to note that the structure of Tz-based radioligands can strongly influence their pharmacokinetic profiles and uptake in the tumor and other tissues (35); this is the most likely cause for the differences observed between the tumoral activity concentrations produced by ¹⁷⁷Lu- and ²²⁵Ac-PRIT. Importantly, ²²⁵Ac-PRIT resulted in increased tumor-to-tissue activity concentration ratios, a trait which is critical in limiting nonspecific toxicity. Finally, the modularity of this PRIT strategy is reinforced by biodistribution studies carried out with another chelator (DO3A) and another antibody-antigen system (huA33/A33).

One major limitation of α-radiotherapy preclinical development is the lack of imaging techniques. However, ²²⁵Ac's daughters' Cerenkov radiation as well as improved biodistribution profile facilitated by pretargeting allowed us to attempt ²²⁵Ac-based Cerenkov radioimmunoimaging. Although in vivo Cerenkov imaging confirmed the pharmacokinetic profile of our pretargeting system, ex vivo imaging provided us with valuable insight into the redistribution of ²²⁵Ac's nonequilibrium daughters. Imaging of the organs postmortem at two different time points highlighted the presence of nonequilibrium daughters in the kidneys, a result that had only been previously reported using γ-ray spectroscopy and autoradiography (36). The short half-lives of the daughters of ²²⁵Ac can make the data acquisition with these latter techniques cumbersome; Cerenkov imaging, in contrast, enabled the screening of 8 tissues in the 5 minutes following euthanasia and therefore constitutes a valuable tool for the preclinical evaluation of ²²⁵Ac-based radiopharmaceuticals.

The accumulation of ²²⁵Ac's daughters in the kidneys medulla was later confirmed by autoradiography. The radionuclides' redistribution is the result of the strong recoil energy associated with α-particles emission (about 100 keV) and constitutes a major concern for the clinical translation of α-therapies. Strategies to control the fate of these daughter radionuclides include the internalization of the nanogenerator (3), the local administration of the radiopharmaceutical (37, 38), and ²²⁵Ac encapsulation within nanocarriers (39, 40). Unfortunately, these methods come with limitations, such as low internalization rates and poor biodistribution profiles. Another approach, which was translated clinically, relies on the use of diuretics to accelerate the excretion of the daughters (41). Along these lines, furosemide and spirolactone are currently used in clinical trials with both ²²⁵Ac- and ²¹²Pb-labeled radioimmunoconjugates (6, 42). Generally speaking, efficient pretargeting strategies employ noninternalizing antibody-antigen systems; however, this necessarily facilitates the redistribution of ²²⁵Ac's daughters. It was therefore critical to ensure that the redistribution of ²²⁵Ac's daughters to the medulla of the kidney was not generating off-target toxicities. The therapy studies carried out in our study focused particularly on late nephrotoxicity and its correlation with the daughters' redistribution. Mild cortical tubular degeneration and necrosis was observed with both administration methods but did not result in the impairment of the renal function. These morphologic changes can be attributed to the ²²⁵Ac-labeled constructs but unlikely related to ²²⁵Ac's radioactive daughters redistribution because only the kidney cortex was affected.

Dosimetry calculations based on two different hypotheses were performed in order to account for the redistribution and decay of daughter radionuclides. The extrapolation of murine data to humans confirms the superiority of the pretargeting approach, which boasts improved mean absorbed doses in most organs. Indeed, the MTDs determined with the pretargeting approach are higher or in the same range as the currently administered activities in the clinical trials focused on ²²⁵Ac-Lintuzumab (18.5 kBq/kg; ref. 6) and ²²⁵Ac-PSMA-617 (100 kBq/kg; ref. 9).

The mouse dosimetry data allowed us to determine a range of suitable doses for longitudinal therapy studies. As a proof of concept, therapeutic studies were first conducted in mice bearing subcutaneous CA19.9-positive PDAC xenografts. Therapeutic efficacy was observed in the cohorts receiving the two highest doses of both RIT and PRIT. The median survival times of the two administration methods were similar at the highest injected activity (37 kBq). However, hematotoxicity was significant in the case of conventional RIT, with one mouse dying of disseminated intravascular coagulation and marked decreases in hematologic parameters such as RBC, HCT, WBC, and PLT across the entire cohort. Similar observations were made in an orthotopic tumor model. Hematotoxicity was therefore the limiting factor for conventional RIT, a finding that correlates with the higher absorbed dose to the blood extrapolated from our biodistribution. In this study, the MTD was not reached with ²²⁵Ac-PRIT, and higher injected activities should be investigated to increase the therapeutic benefit of PRIT. The use of an immunocompetent mouse model should also be considered in order to maximize benefits from the radiation-induced immune response (the abscopal effect).

In sum, the ²²⁵Ac-based strategy for PRIT reported here facilitates (1) the delivery of radioactive payloads to tumor sites while reducing the mean absorbed dose to healthy tissues and (2) enables prolonged survival and reduced hematotoxicity in subcutaneous and orthotopic models of PDAC compared with conventional RIT. The development of PRIT for α-therapy complements the two on-going 5B1-based clinical trials for the diagnosis (⁸⁹Zr) and β-therapy (¹⁷⁷Lu) of PDAC and has the potential to extend the population of patients eligible for CA19.9-targeted radiotherapy to include those with residual metastatic lesions as well as patients that develop resistance to β-therapy. Future studies will assess the ability of ²²⁵Ac-PRIT to delay tumor growth and evaluate the MTD with fractionated doses. A comparison of the potential of α- and β-PRIT in different mouse models will also allow us to further tailor this treatment according to tumor burden and phenotype.

W.W. Scholz holds ownership interest (including patents) in MabVax. J.S. Lewis reports receiving other commercial research support from MabVax and is a consultant/advisory board member for Clarity Pharmaceuticals. No potential conflicts of interest were disclosed by the other authors.

Conception and design: S. Poty, L.M. Carter, B.M. Zeglis, J.S. Lewis

Development of methodology: S. Poty, B.M. Zeglis, J.S. Lewis

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): S. Poty, K. Mandleywala, R. Membreno, D. Abdel-Atti, A. Ragupathi, W.W. Scholz

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): S. Poty, L.M. Carter, R. Membreno, J.S. Lewis

Writing, review, and/or revision of the manuscript: S. Poty, L.M. Carter, W.W. Scholz, B.M. Zeglis, J.S. Lewis

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): S. Poty, K. Mandleywala, D. Abdel-Atti, J.S. Lewis

Study supervision: B.M. Zeglis, J.S. Lewis

The authors gratefully acknowledge the Radiochemistry and Molecular Imaging Probes core facility and the Laboratory of Comparative Pathology (Sebastian Monette and Sara Santagostino), which were supported in part by NIH grant P30 CA08748. They gratefully acknowledge William H. and Alice Goodwin and the Commonwealth Foundation for Cancer Research, the Center for Experimental Therapeutics of Memorial Sloan Kettering Cancer Center (J.S. Lewis), the NIH (F32 EB025050, L.M. Carter; U01 CA221046, J.S. Lewis and B.M. Zeglis; R00 CA1440138, B.M. Zeglis), and the François Wallace Monahan Fellowship from the JLM Benevolent Fund (S. Poty). Dr. Michael R. McDevitt is acknowledged for the fruitful discussions and advice he gratefully provided regarding this project. The authors thank Dr. Kristen Cunanan for reviewing the biostatistical analysis.

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.