Tumor Therapy with Targeted Atomic Nanogenerators

See allHide authors and affiliations

Science  16 Nov 2001:
Vol. 294, Issue 5546, pp. 1537-1540
DOI: 10.1126/science.1064126


A single, high linear energy transfer alpha particle can kill a target cell. We have developed methods to target molecular-sized generators of alpha-emitting isotope cascades to the inside of cancer cells using actinium-225 coupled to internalizing monoclonal antibodies. In vitro, these constructs specifically killed leukemia, lymphoma, breast, ovarian, neuroblastoma, and prostate cancer cells at becquerel (picocurie) levels. Injection of single doses of the constructs at kilobecquerel (nanocurie) levels into mice bearing solid prostate carcinoma or disseminated human lymphoma induced tumor regression and prolonged survival, without toxicity, in a substantial fraction of animals. Nanogenerators targeting a wide variety of cancers may be possible.

Alpha particles are high-energy, high linear energy transfer helium nuclei capable of strong, yet selective, cytotoxicity (1). A single atom emitting an alpha particle can kill a target cell (2). Monoclonal antibodies conjugated to alpha particle–emitting radionuclides (213Bi and 211At) are starting to show promise in radioimmunotherapy (3, 4). The conjugates [213Bi]- HuM195 (2) and [213Bi]J591 (5, 6) have been used in preclinical models of leukemia and prostate cancer, respectively, and in a phase I human clinical trial, [213Bi]HuM195 was active against leukemia, with no significant toxicity (3). Astatine-211–labeled antibodies to tenascin (anti-tenascin) have been used clinically to treat malignant gliomas in humans (4) in a phase I trial. For clinical use of213Bi, we developed a therapeutic dose-level225Ac/213Bi generator device, approximately 1 cm by 6 cm in size, capable of producing alpha particle–emitting atoms for attachment to ligands suitable for human injection (7,8). Despite this improvement, the major obstacle to the widespread use of these drugs remains the short 213Bi half-life (46 min), which limits its delivery to only the most accessible cancer cells.

One solution to these constraints is to deliver the atomic generator to the target cell, allowing production of the atoms that will yield potent alpha emissions at or in the cancer cell. For this process to be successful pharmacologically, the device must possess molecular dimensions. At its ultimate reduction, the device therefore consists of a single 225Ac generator atom attached to the delivery vehicle. Actinium-225 has a 10.0-day half-life and decays via alpha emission through three atoms, each of which also emits an alpha particle (9, 10). We developed methods that use bifunctional versions of the chelating moiety DOTA (1,4,7,10-tetraazacyclododecane-1,4,7,10-tetraacetic acid) to stably bind 225Ac to the delivery vehicles, and have described the synthesis, purification, and analyses of the resulting constructs [supplementary note 1 (11)].

The stability of an 225Ac-DOTA-antibody construct was determined in vitro under conditions similar to those expected in vivo. The [225Ac]DOTA-HuM195 construct was compared with [177Lu]DOTA-HuM195 in 100% human serum, 100% mouse serum, and 25% human serum albumin at 37°C for 15 days. The [225Ac]HuM195 displayed stability similar to the177Lu analog with less than a 5% loss of 225Ac from the IgG over 15 days [supplementary fig. 1 (11)]. The stability data in all three conditions were similar. Importantly, there was no loss of 225Ac to other serum proteins in vivo [supplementary note 2 (11)].

We examined the in vitro cytotoxicity of 225Ac constructs specific for HL60 leukemia cells [HuM195 (anti-CD33)]; Daudi and Ramos lymphoma cells [B4 (anti-CD19)]; MCF7 breast carcinoma cells [Herceptin (anti-HER2/neu)]; LNCaP.FGC prostate carcinoma cells [J591 (anti-PSMA)]; and SKOV3 ovarian cancer cells [Herceptin (anti-HER2/neu)] at very small doses of generators (12). The median lethal dose (LD50) of the225Ac constructs ranged from 0.3 to 74 Bq/ml (0.008 to 2 nCi/ml) and were lower than values for corresponding 213Bi alpha particle–emitting antibodies (∼7400 Bq/ml) (2, 5,13). Controls at low specific activities (accomplished by adding excess unlabeled antibody) did not show specific binding of the alpha-particle generators to the targets, and were used to represent nonspecific cytotoxicity. The LD50 values were 10- to 625-fold higher in the controls using excess unlabeled antibody.

Critical to the success of this generator approach is the retention of the daughter alpha-emitting atoms at or in the target cells. To investigate whether the radionuclidic daughters of target cell–internalized 225Ac constructs were retained intracellularly in vitro, we measured the resulting increase or decrease of the 221Fr and 213Bi daughters relative to secular equilibrium values. The antibodies HuM195, J591, and B4 internalize into HL60, LNCaP, and Daudi cells, respectively, after binding (2, 5, 1416), carrying with them the attached 225Ac radionuclide. Analyses of the internalized generators show that initially there were greater than equilibrium levels of 221Fr and 213Bi daughters present inside the cells (Fig. 1A).

Figure 1

(A) Internalization and retention in vitro of [225Ac]J591/213Bi/221Fr in LNCaP cells. Actinium-225 that was internalized or was outside the cell was determined after a 5-hour period when secular equilibrium was established and the 213Bi and 221Fr curves converge. The radionuclidic decay of 225Ac yields two daughter radionuclides, 221Fr and 213Bi, that can be monitored in these experiments by gamma spectroscopy. Values shown are mean ± SD. LNCaP cells (107 cells) were exposed to an antibody-to-antigen excess of [225Ac]J591 at 37°C for 90 min [supplementary note 3 (11)]. (B) Biodistribution in vivo of [225Ac]J591/213Bi/221Fr in several tissues from one representative mouse. Actinium-225 that was tissue-associated was determined after a 5-hour period when secular equilibrium was established and the 213Bi and221Fr curves converge. Pharmacological analysis of the two225Ac daughters was performed in vivo by injecting 12 kBq of [225Ac]J591 or 12 kBq of [225Ac]HuM195 (irrelevant control) intraperitoneally in two groups of male athymic nude mice (n = 12 per group) (Taconic, Germantown, New York) bearing a 3- to 4-week-old LNCaP i.m. tumor xenograft. Mice from each group were killed at days 2 and 3, respectively, and the tumors, blood, and other tissues were removed and immediately counted with a Packard Cobra Gamma Counter using two energy windows [supplementary note 3 (11)].

The in vivo biodistribution of the generator constructs and the radionuclidic daughters was determined in mice bearing a LNCaP carcinoma. Approximately 18 and 21% of the injected dose of [225Ac]J591 was localized in the tumor (per g) at 2 and 3 days, respectively. Tumor samples (average ± SD,n ≥ 3) counted 6 to 12 min after death revealed that221Fr was 88% ± 9% and 213Bi was 89% ± 2% of the 225Ac secular equilibrium levels of the tumor associated radioactivity (Fig. 1B). These measurements, however, represent a composite value of tumor cell–internalized and surface-bound [225Ac]J591 and its decay daughters. The daughters produced on the tumor cell's outer membrane surface might be transferred away to other sites such as the kidneys and intestine.

To investigate the in vivo therapeutic efficacy of the generator construct [225Ac]J591, we used an intramuscular (i.m.) LNCaP tumor model (5) in male nude mice. Serum prostate-specific antigen (PSA) is an important surrogate marker for prostate cancer burden and prognosis in humans (17). It can also be used in animal models with prostate cancer cell xenografts (5, 18). Rising PSA levels predict the appearance of visible tumor and death. The mice in our experimental groups had mean PSA values of 2 to 5 ng/ml on 10 and 12 days after implantation of tumor. At the time the generator was administered on day 12 or 15, the tumors were characterized histologically as vascularized and encapsulated nodules, each containing tens of thousands of tumor cells (5). A single nontoxic (19) administration of the generator construct 15 days after implantation significantly improved (P < 0.006) median survival times of mice relative to mice treated with [225Ac]B4 irrelevant control antibody mixed with unlabeled specific J591 (dual control) or untreated controls [Supplementary fig. 2 (11)]. There was no significant difference in survival times between the dual control–treated animals and untreated controls. The median survival time of untreated controls in this model was 33 days (n = 15). The mean and median pretherapy PSA values measured on day 12 were not significantly different among the three groups of mice. However, on days 28 and 42, the PSA values of [225Ac]J591-treated animals were significantly lower than the PSA values for the dual control–treated animals and untreated controls [supplementary fig. 3 (11)]. There was no significant difference in PSA levels between the dual control–treated animals and untreated controls at either time.

Earlier treatment of mice, on day 12 after LNCaP tumor implantation with a single administration of [225Ac]J591, caused tumor regression and significantly improved (P< 0.0001) the median survival times of these mice to 158 days compared to 63 days in the mice treated on day 15 (Fig. 2A). After treatment, PSA levels decreased from pretherapy levels in many of the mice to low and undetectable levels and remained undetectable in 14 animals, of the 39 treated animals; all 39 animals exhibited prolonged survival (Fig. 2B). These mice survived at least 10 months and had no measurable PSA levels or evidence of tumor at the time of death (293 days). Animals treated with unlabeled J591 (0.004 or 0.04 mg) on day 12 after implantation had no prolongation of median survival (37 and 35 days, respectively;n = 9). The therapeutic efficacy was dependent on antibody specificity, the administration of the225Ac-generator, and the delay time between tumor implantation and initiation of treatment.

Figure 2

(A) Kaplan-Meier plot showing survival of mice bearing i.m. LNCaP tumor xenografts treated intraperitoneally in several therapy/control experiments. The 39 animals that received 7.8 kBq [225Ac]J591 were treated on day 12, and the 13 animals that received 7.2 kBq [225Ac]J591 were treated on day 15. Animals were killed when tumor area was ≥2.5 cm2. Median survival versus time was evaluated using a log-rank test (P < 0.0001). (B) Individual serum PSA values of the 39 mice treated with a 7.8 kBq dose of [225Ac]J591 on day 12 in the therapy experiment with LNCaP model (Fig. 2A). The median was marked with a solid line. (Note the split scale of PSA levels.) PSA values were evaluated using an unpaired t-test with two-tailed P values (95% confidence limit) to analyze differences between study groups.

To determine if other tumor types could be treated with225Ac-generator constructs, we investigated a disseminated human Daudi lymphoma cell mouse model (20), using [225Ac]B4 as the therapeutic agent. Mice were treated 1 day after tumor dissemination with a single administration of specific [225Ac]B4 (three different dose levels), irrelevant control [225Ac]HuM195 (two dose levels), or unlabeled B4. Control mice receiving the irrelevant [225Ac]HuM195 had median survival times from xenograft of 43 days (5.6 kBq) and 36 days (1.9 kBq). Mice receiving 0.003 mg of unlabeled B4 per mouse had a median survival time of 57 days. The mice receiving a single injection of [225Ac]B4 showed dose-related increases in median survival times: 165 days (6.3 kBq), 137 days (4.3 kBq), and 99 days (2.1 kBq) (Fig. 3A). This dose response of [225Ac]B4 was significant (P = 0.05). About 40% of mice treated at the highest dose were tumor-free at 300 days, and the experiment concluded on day 310.

Figure 3

(A) Kaplan-Meier plot showing survival of mice bearing a disseminated Daudi xenograft. In this lymphoma therapy experiment, mice (groups of n = 5) were treated 1 day after tumor dissemination via intravenous injection. Animals were monitored for signs of morbidity or hind-leg paralysis, at which time they were killed. Median survival versus time was evaluated using a log-rank test (P < 0.0001). (B) Kaplan-Meier plot showing survival of mice bearing a disseminated Daudi xenograft. Mice (groups of n = 5) were treated with a single 6.3-kBq dose of [225Ac]B4 on days 13, 6, 3, and 1 after intravenous xenograft. Controls were untreated animals with xenografts initiated day 13 or 1. Animals were monitored and statistical analysis was performed as described above (P < 0.0001).

We also examined the effect of treatment delay in the disseminated lymphoma model (Fig. 3B). Mice (n = 15) that received a single injection of [225Ac]B4 (6.3 kBq) treatment on day 1, 3, or 6 after tumor implantation had a similar prolongation of survival relative to untreated controls (28 or 30 days for controls versus 103, 95, or 108 days for animals treated on days 1, 3, or 6, respectively). Mice (n = 5) that received treatment as late as 13 days after tumor dissemination survived >173 days. Unlabeled B4 was minimally active in mice (n = 5 per group) with median survival of 44 or 40 days for mice treated with 0.002 or 0.20 mg, respectively. Untreated controls (n = 15) had a median survival time of 28 days. Therefore, in this lymphoma model, although specificity and dose level were important factors in efficacy, the delay between tumor dissemination and initiation of treatment was less relevant up to a certain time point, at which it was then inversely related to activity. This observation may be related to the geometry of the emitted radiation of the alpha particle, which may kill a cluster of cells more easily than a single cell.

Previous workers have concluded that therapy with225Ac-constructs might not be feasible because the constructs are unstable and because the radionuclidic daughters present an untenable pharmacological problem (21–24). However, our findings indicate that 225Ac can be used as a safe and potent tumor-selective molecular-sized generator in both established solid carcinomas or disseminated cancers. In part, the enhanced potency of these constructs as compared to the 213Bi analogs can be attributed to the longer half-life (313-fold greater 225Ac half-life) and the four net alpha particles emitted by the 225Ac, but other mechanisms must also be involved, such as more efficient cytotoxicity following intracellular delivery of the generator. Once inside the cell, the geometry of the decay trajectory of the alpha particle favors highly efficient cell killing: each decay must pass through the cell, whereas statistically only 30% of the alpha decays will pass through the cell if the generator is surface bound (2). Selection of tumor antigen systems that internalize the 225Ac generator construct help to retain the daughters and therefore lead to enhanced potency; however, internalization is not required for activity.

The development of synthetic methods to yield stable nanogenerator constructs of [225Ac]IgG in useful quantities, and the demonstration of safe, efficacious deployment against murine models of both solid carcinomas and disseminated cancer, using very small doses of isotope, suggest a pathway to widespread clinical use of such targeted drugs. The 10-day half-life of the225Ac generator constructs would allow the drugs to be manufactured at a central radiopharmacy and shipped throughout the world. Because of the extraordinary potency of 225Ac generators, little radioactivity [possibly sub-MBq (mCi)] would be required for therapeutic human use, allowing for economical outpatient use and safety. In addition, the longer half-life of 225Ac may allow better penetration of larger tumors.

  • * To whom correspondence should be addressed. E-mail: d-scheinberg{at}


View Abstract


Navigate This Article