High-Linear Energy Transfer (LET) α versus Low-LET β Emitters in Radioimmunotherapy of Solid Tumors: Therapeutic Efficacy and Dose-limiting Toxicity of ²¹³Bi- versus ⁹⁰Y-labeled CO17-1A Fab′ Fragments in a Human Colonic Cancer Model¹

RIT³ is an attractive therapeutic concept, aiming to deliver tumoricidal radiation doses to tumors without causing significant radiation toxicity to normal tissues (1). Indeed, in radiosensitive tumors, such as non-Hodgkin’s lymphoma, RIT has led to long-term remissions and even cures in a high percentage of treated patients (1, 2, 3). In solid tumors, however, success is still limited, probably due to the relatively low specific accretion of the radiolabeled antibody in the tumor target as compared to the normal tissues (1, 4, 5).

More recently, the use of high-LET radiation emitters has been proposed (6, 7, 8, 9). One such approach is the use of Auger/conversion electron emitters, which requires their conjugation to carriers (e.g., antibodies) that internalize selectively into the target cancer cells. Due to their short path length, these low-energy Auger electrons can reach the nuclear DNA only if the antibody is internalized, such as in antigen-expressing tumor tissue but not in the other nonmalignant cells, e.g., the hematopoietic cells in the red marrow. Indeed, higher antitumor efficacy at lower toxicity has been demonstrated in vitro (6) as well as preclinically (7) and clinically (8, 9).

α emitters appear to be an even more attractive therapeutic alternative because, due to their path length of several cell diameters, their high LET radiation does not necessarily require internalization into the target cell (10, 11). Therefore, non-antigen-expressing tumor cells in their neighborhood may also become inactivated. Several α emitters have been proposed for this purpose (reviewed in Ref. 11). Although attractive due its short half-life (7.2 h) and the anticipated similarity of its chemical behavior to the well-understood chemistry of iodine, the halogen ²¹¹At is burdened by the long half-life (38 years) of its β-emitting decay product, ²⁰⁷Bi. Furthermore, directly astatinylated tyrosyl moieties have been shown to be metabolically unstable (10); the use of astatinylated N-succinimidyl benzoate derivatives yields more stable conjugates but requires a more sophisticated chemical synthesis (10, 12, 13). Radium isotopes (mainly ²²³Ra, but also ²²⁴Ra) have been proposed (1, 11), but no chelator has demonstrated sufficient stability in vivo thus far. ²¹²Bi and its parent ²¹²Pb have been chelated successfully to antibodies and other proteins via DTPA or tetraazacyclododecane-N,N′,N″,N″′-tetraacetate derivatives (14, 15); both have already been used for α-RIT, but the decay product of ²¹²Bi, ²⁰⁸Tl is itself a high-energy β⁻ and γ emitter, potentially raising radiation protection problems. Some years ago, ²¹³Bi was proposed for α immunotherapy (16, 17). It can be readily obtained from an “in-house” ²²⁵Ac/²¹³Bi generator (16, 17). ²¹³Bi decays mainly (98%) by β⁻ and 440-keV γ emission with a t_1/2 of 45.6 min to the ultra-short-lived high-energy (8.375-MeV) α emitter ²¹³Po (t_1/2 = 4.2 μs), whereas a direct α decay pathway to ²⁰⁹Tl plays only a negligible role (2% of all ²¹³Bi decays).

Several studies have been published on the therapeutic use of α emitters, and several authors suggested their therapeutic superiority (most recently reviewed in Ref. 11). However, with exception of the study of Bloomer et al. (18) using colloids for therapy of malignant ovarian ascites in mice, to the best of our knowledge, no investigators have attempted a direct comparison of the therapeutic efficacy and toxicity of targeted therapy with high-LET α versus low-LET β emitters in vivo. Our previous work has shown that (a) monovalent antibody fragments (e.g., Fab′) may have therapeutic advantages over bivalent immunoconjugates [e.g., F(ab′)₂ or IgG], which is due to a more homogenous intratumoral distribution and higher intratumoral dose rates (19); (b) potential nephrotoxicity of antibody fragments can be overcome successfully (20, 21); and (c) matching biological half-lives of the carrier and physical half-lives of the radionuclides may optimize their therapeutic efficacy (21). Therefore, we chose, for this study, the monovalent Fab′ fragment of the MAb CO17-1A, to assess the toxicity and antitumor efficacy of RIT with the α emitter ²¹³Bi/²¹³Po as compared to the β emitter ⁹⁰Y, linked to a monovalent Fab′ fragment in a human colon cancer model in nude mice and to correlate the observed biological effects to the calculated radiation dosimetry.

The murine MAb CO17-1A was obtained from GlaxoWellcome (Hamburg, Germany), as described previously (7, 22, 23, 24). It is an IgG2a isotype directed against a M_r 41,000 glycoprotein that is found on human gastrointestinal malignancies and, to a lower extent, on normal epithelia (22, 23, 24). It has an affinity constant of 5 × 10⁷ liter/mol to its antigen (7, 22, 23, 24). After antigen binding, it is readily internalized into the antigen-expressing tumor cells (6, 7). The antihuman-CD3 antibody, OKT3, was used as irrelevant isotype-matched control (7). OKT3, which is a murine IgG2a as well, was obtained from CILAG (Sulzbach/Taunus, Germany).

Fab′ fragments were prepared from complete IgG by pepsin digestion [ImmunoPure F(ab′)₂ preparation kit; Pierce, Rockford, IL] and subsequent disulfide reduction in the presence of 10 mm cysteine for 12 h at 37°C. Fab′ fragments were purified by protein A chromatography. Isothiocyanate-Bz-DTPA conjugates of iodoacetamide-blocked Fab′ fragments were prepared as described previously (21, 25).

For biodistribution studies, ⁸⁸Y was obtained as ⁸⁸YCl₃ in 0.1 m HCl from Amersham Life Science (Braunschweig, Germany). For therapy, ⁹⁰Y was obtained from Pacific Northwest National Laboratory (Richland, WA) as ⁹⁰YCl₃ in 0.05 m HCl.

²¹³Bi was obtained as mixed iodide/chloride salt (mainly in the form of the BiI₅²⁻ anion; Refs. 11 and 26) from an in-house ²²⁵Ac/²¹³Bi generator, as has been described in more detail previously (16, 17). Briefly, ²²⁵Ac, which had purified from a ²²⁹Th source (the latter being a decay product of ²³³U, originally obtained from the Oak Ridge National Laboratory, Oak Ridge, TN; Refs. 16, 17, 27, and 28) was dissolved in 2 m HCl and loaded onto a lead-shielded 4 × 40 mm column of AG MP-50 cation exchange resin (Bio-Rad Laboratories, Munich, Germany). ²²⁵Ac decays by α/γ decay with a 10.0-day half-life via two short-lived α-/γ-emitting intermediates (²²¹Fr and ¹¹⁷At) to ²¹³Bi, which can be eluted in 2-h intervals with 1 ml of 0.1 m NaI-0.1 m HCl. After each elution, the cation exchange column was reequilibrated by washing with five bed volumes of 0.01 m HCl. The 0.1 m NaI-0.1 m HCl eluate containing the activity was buffered with 200 μl of 0.5 m NaOAc (pH 5.5), and the pH was adjusted with 50–70 μl of 1.0 m NaOH to 5.5.

Radiolabeling of the immunoconjugates followed procedures described previously (21). Briefly, the antibody chelate was buffered in 0.5 m NaOAc (pH 5.5). Activity was added to yield a specific activity of 4–5 mCi/mg antibody protein, and the reaction mixture was incubated for 5 min at 37°C. The labeled chelate was purified from unbound radiolabel within 5 min by size-exclusion chromatography on a PD-10 column (Sephadex G-25 prepacked in a 15 × 50 mm column; Pharmacia, Uppsala, Sweden). Excess of DTPA was added to each preparation to chelate any unbound bismuth or yttrium, to assure its rapid renal excretion and to prevent eventual kidney or bone accretion. The quality of each preparation was tested by high-performance liquid chromatography (Bio-Sil SEC-250 column, 300 × 7.8 mm; Bio-Rad Laboratories, Richmond, CA) as well as by measuring its immunoreactivity (21). The amount of unbound isotope was <2%, and 80–85% of the radiolabeled antibodies bound to the GW-39 cells in each preparation. Labeled antibodies were administered within 20 min of their final preparation.

Female nude mice, weighing 19–23 g and 4–5 weeks old, were purchased from Charles River (Sulzfeld, Germany). The human colon carcinoma cell line, GW-39 (29), was serially propagated by preparing a mince through a 40-mesh screen and rinsing with sterile HBSS (ICN Biomedicals, Eschwege, Germany) to yield a 20% cell suspension. Two hundred μl of this suspension were injected s.c. After ∼10 days, tumors reached the size of ∼100–200 mg, which was the size used in this study.

As will be published elsewhere in more detail,⁴ multiple liver metastases of the same human colon cancer cell line, GW-39, were induced by intrasplenic injection of 100 μl of a 10% tumor cell suspension into the spleen of anesthetized nude mice with subsequent splenectomy 2 min later. All animals treated in this manner develop multiple (>250) microscopic tumor colonies in their livers, reaching a size of ∼250–500 μm at 10 days and a size of 1–2.5 mm at 20 days after tumor cell inoculation (see Fig. 4 C).⁴ With high reproducibility, the animals begin to lose weight by 4–6 weeks and eventually die at 6–9 weeks after tumor inoculation.⁴

s.c. tumor-bearing animals were injected into the tail vein with the radiolabeled antibody fragments at a protein amount of ∼100–200 μg. The mice were sacrificed at 10 min, 1 h, 4 h, 24 h, 72 h, and 168 h for ⁸⁸Y- and at 10 min, 1 h, 3 h, and 5 h for ²¹³Bi-labeled conjugates. They were bled by retroorbital puncture. For determining the whole-body retention, the mice were measured in a well counter. After cervical dislocation, the animals were dissected. The amount of activity in the tumors and tissues (liver, spleen, kidney, lung, heart, intestine, bone, and blood) was determined by gamma scintillation counting using an injection standard to account for physical decay, as described previously (e.g., Ref. 21). The number of animals used for each study was typically five per group at each time point, and the tumor sizes were in the same range as those used for the actual therapy experiments.

The biodistribution data were used to generate time-activity curves (7, 19). Time-activity curves were fit to the difference or sum of two exponents for curves with and without an uptake component, respectively. Cumulated activity in each organ was obtained by analytically integrating the resulting expressions over time.

Tumor and organ doses were calculated, accounting for the electrons of ⁹⁰Y or the α particle, electron, and photon contributions of ²¹³Bi and its daughter, ²¹³Po, but neglecting the ²⁰⁹Pb or ²⁰⁹Tl β contributions. These dose calculations were performed according to the geometric model of a 30-g mouse by Yoriyaz and Stabin (30). This model accounts for cross-organ irradiation and the whole-body contribution by using the Monte Carlo transport simulation code MCNP4A (30, 31). In this model, typically, ∼500,000 particle histories were followed and absorbed fractions and S values were generated for sources in all major organs and the “remainder of the body.” Total absorbed doses were also calculated for all important target organs, considering all sources of radiation dose (32). The fractional contribution from cross-irradiation of different organs by electrons was calculated. Normally, this contribution is neglected in internal dose calculations; however, we have shown previously that, in a small animal, it may be significant (7, 21, 32).

Tumor sizes of s.c. tumors were determined by caliper measurement in three dimensions immediately before therapy and at weekly intervals thereafter. Animals were either left untreated (controls) or injected with a single dose of ⁹⁰Y- or ²¹³Bi-labeled CO-1A Fab′ with the activities indicated. Ten to 20 animals were studied in each treatment group. As a nonspecific therapy control, the Bz-DTPA-conjugated Fab′ fragment of the anti-CD3 human T-cell lymphoma MAb OKT3 was used. Body weight was recorded weekly, and survival was monitored. The MTD was defined as the highest possible dose under the respective conditions that did not result in any animal deaths, with the next higher dose level resulting in at least 10% of the animals dying (21). Animals were observed until their death or a loss of >30% of their original weight or until the tumor began to ulcerate through the skin, at which time they were removed from the group and sacrificed (21).

d-Lysine monohydrochloride (Sigma Chemie, Deisenhofen, Germany) was dissolved in PBS at 175 mg/ml. The animals were injected i.p. with 200 μl of this solution (i.e., 35 mg of d-lysine) with four hourly injections, starting 30 min before administration of the radiolabeled antibody, as was established previously as the optimal regimen (20, 21).

Bone marrow was harvested by sterile technique from untreated donor nude mice of the same strain. The marrow cavity of both mouse femurs was rinsed with 0.9% sterile saline using 26-gauge needles, as described previously (33). An inoculum of 10⁷ bone marrow cells was injected i.v. via the tail vein at 24 h after radioantibody injection.

Total and differential leukocyte and thrombocyte counts were determined on the day of therapy and at weekly intervals thereafter. Seventy-five-μl heparinized specimens were collected by retroorbital bleeding. The sample were counted on a Technicon H3 Auto-Analyzer (Bayer-Diagnostik, Munich, Germany). The mean ± SD was calculated for each group. BUN and creatinine as well as glutamate oxaloactetate transaminase and alkaline phosphatase levels were determined on the day of RIT and at weekly intervals thereafter. All these parameters were assayed as described previously (7, 19, 21).

For organ and tumor histology, the animals were sacrificed by cervical dislocation and necropsied at the times indicated. Organs were fixed in 10% formalin, embedded in paraffin, and cut into 5-μm sections, as described previously (7, 19, 21). Staining was performed with H&E or periodic acid-Schiff.

Differences in the tissue uptake values and the biodistribution data were statistically analyzed with the Student’s t test for unpaired data, as described previously (34). Pairwise comparisons were performed with the Wilcoxon rank-sum test (35). Differences in the therapeutic efficacy between the treatment modalities in s.c. tumors were analyzed by assuming an exponential tumor growth pattern; nonlinear regression analysis based on asymptotic approximation was used as described previously (34). Survival analysis was based on the Kaplan-Meier product limit, and groups were compared using the log-rank test. The log-rank test was also used for comparing times to tumor volume multiplication (36, 37).

Fig. 1 shows the biodistribution of bismuth- versus yttrium-labeled CO17-1A Fab′ (Fig. 1,a) as compared to free, unchelated Bi³⁺ or Y³⁺ (Fig. 1,b) in tumor, blood, kidneys and bone of s.c. GW-39 tumor-bearing nude mice. Due to the short physical half-life of ²¹³Bi, biodistribution studies were performed up to 5 h postinjection (i.e., more than six physical half-lives), in contrast to 1 week with yttrium-labeled conjugates. No significant differences were observed between the biokinetics of both ²¹³Bi- or ⁸⁸Y-labeled immunoconjugates. The blood clearance was biexponential, with similar half-lives (t_1/2α ∼ 15 min, t_1/2β ∼ 5 h) to those observed with radioiodinated Fab′ fragments (19). The uptake in the tumor was rapid, reaching its apogee as early as after 1–4 h postinjection (∼4% ID/g). As expected with monovalent Fab′, renal accretion was predominant, with maximum uptake values of up to ≥80% ID/g. As has been published in detail previously (20, 21), application of d-lysine reduced this renal accretion 4–5-fold [Fig. 1 a; compare closed diamonds (control kidneys) and open diamonds (kidneys of lysine-treated animals)]. For both ²¹³Bi- and ⁸⁸Y-labeled CO17-1A Fab′, all other organs (e.g., the heart, lungs, liver, spleen, stomach, or intestine) behaved as typical “blood pool organs,” with their activity clearances paralleling the blood clearance (data not shown).

In contrast, Fig. 1,b shows the biodistribution of the free radiometal ions Bi³⁺ and Y³⁺. Whereas free bismuth is known to exhibit a very high renal accretion (up to 150% ID/g), which, interestingly, cannot be blocked by lysine (Fig. 1,b, top), yttrium ions are well known to be bone-seeking (Fig. 1,b, bottom). Fig. 1 clearly demonstrates the good metabolic stability of the Bz-DTPA conjugates used in this study (note the low bone accretion of Y-Fab, with the kidney uptake of bismuth-labeled conjugates not being significantly different from those of their yttrium analogues, as well as the possibility of blocking their renal accretion by lysine, which is in contrast to the persistent renal uptake of free Bi³⁺, despite amino acid administration).

MTD-finding trials of ²¹³Bi- and ⁹⁰Y-labeled CO17-1A Fab′ were undertaken. For this purpose, varying amounts of activity were injected, starting at 100 μCi and increasing in 10–20% steps. At each activity level, one group of animals (10–20 per group) received the ²¹³Bi- or ⁹⁰Y-Fab′ with no additional support, and a second group was treated with d-lysine. Blood counts (WBCs and platelets) and levels of BUN, creatinine, glutamate oxaloactetate transaminase, and alkaline phosphatase were determined on the day of radioantibody administration and at weekly intervals thereafter. Acute treatment-related death was defined as occurring within 4 weeks post-RIT, whereas later deaths were regarded as chronic toxicity, if these deaths could not be related to tumor growth.

Seven hundred μCi of ²¹³Bi-CO17-1A Fab′ and 250 μCi of its ⁹⁰Y-labeled analogue were tolerated by all animals without acute treatment-related lethality, regardless of lysine administration (i.e., the respective MTDs). A 10% increase of these activities resulted in an at least 10% lethality. Fig. 2 shows the blood cell counts (top) and BUN levels (bottom) at these respective maximum tolerated activities for ²¹³Bi- (Fig. 2,a) and ⁹⁰Y-Fab′ (Fig. 2,b), respectively. The nadirs of leukocyte and thrombocyte counts were reached at 1 week after radioantibody administration, and at both respective maximum tolerated activities, the severity of myelotoxicity was not significantly different, regardless of the radionuclide. The recovery from myelotoxicity, however, seems to be faster with the shorter-lived ²¹³Bi than with ⁹⁰Y; with ²¹³Bi-Fab′, time to complete recovery was 2–3 weeks posttherapy, but it was 3–4 weeks with ⁹⁰Y-Fab′ (compare Fig. 2, a and b, top).

No significant differences with respect to the severity of acute toxicity were observed, whether the animals were given d-lysine or not, and lysine administration alone (without BMT) did not allow for an increase of the MTD, clearly identifying the red marrow as first-line dose-limiting organ. In contrast, profound differences with respect to chronic toxicity were observed between animals given lysine and controls. Fig. 2, bottom, shows that, in non-lysine-treated animals, after a transient (2–3 week) episode of BUN elevation with subsequent normalization, BUN levels began to rise at 6–7 weeks after therapy. As we have observed previously (21), the serum creatinine followed the BUN values but was less sensitive than the latter (data not shown). Surprisingly, the development of chronic renal failure was more dramatic (more steeply rising BUN levels; Fig. 2) in the ⁹⁰Y- than in the ²¹³Bi-treated groups; all animals treated with 250 μCi of ⁹⁰Y-Fab′ without lysine protection died within 3 months after therapy, whereas the lethality of animals given 700 μCi of ²¹³Bi-Fab′ was only 35% after 6 months.

Lysine-treated groups survived without any signs of renal compromise, and there was no induction of renal insufficiency even six months after treatment (Fig. 2). When kidney protection by lysine was combined with BMT (the latter given at 24 h post-radioantibody administration), the animals survived 400 μCi of ⁹⁰Y- and 1100 μCi of ²¹³Bi-CO17-1A Fab′ (i.e., the MTDs with BMT), whereas 10% higher activities again led to an at least 10% lethality. No other signs of second-organ toxicity were observed in any of the treatment groups; e.g., glutamate oxaloactetate transaminase and alkaline phosphatase as indicators of liver function did not show any significant changes, which is in accordance to normal organ doses well below 20 Gy (Table 1).

Fig. 3 shows the therapeutic effects of ²¹³Bi- versus ⁹⁰Y-labeled CO17-1A Fab′ at their respective MTDs without (Fig. 3,a) or with (Fig. 3,b) BMT in s.c. GW-39-bearing animals. Both radioconjugates led to a significant (P < 0.02) growth retardation as compared to untreated controls or animals treated with the same activities of the Fab′ fragment of the irrelevant anti-CD3 antibody OKT3 (the latter not shown). Antitumor effects improved with dose intensification (Fig. 3, a versus b). However, at each MTD (conventional or with BMT) ²¹³Bi-Fab′ was therapeutically superior to its ⁹⁰Y analogue (P < 0.01; Fig. 3), whereas no significant differences were observed between both radiolabels with the irrelevant OKT-3 Fab′ (data not shown).

Fig. 4a shows the survival of animals bearing GW-39 liver metastases. Groups of 20 animals each were left untreated or given 250 μCi of ⁹⁰Y-CO17-1A or irrelevant OKT-3 Fab′ or 700 μCi of their ²¹³Bi-labeled analogues at 10 days after tumor inoculation, each under kidney protection with d-lysine. The figure shows that untreated animals died within 6–8 weeks of rapidly progressing liver metastases. The survival of animals treated with the irrelevant antibody was prolonged only for 1–3 weeks, regardless of the radiolabel. In contrast, the mean survival of animals treated with ⁹⁰Y-Fab′ was 15 weeks, and 20% of mice treated in this manner survived for >30 weeks. Histology could not demonstrate any viable tumor cells in these long-term survivors, so that these animals were regarded as cured. The cure rate of animals treated with ²¹³Bi-CO17-1A Fab′ was, at 95%, highly significantly better (P < 0.001) than with its ⁹⁰Y analogue. This is in accordance to the macroscopic appearance of animals at 8 weeks after tumor cell inoculation. Fig. 4 b shows animals at this very time point, which were left untreated (left), which were given 250 μCi of ⁹⁰Y-labeled (middle) or 700 μCi of ²¹³Bi-labeled CO17-1A Fab′ at a 10-day tumor stage (right). Whereas the liver of untreated animals is grossly enlarged and almost completely replaced by tumor and the livers of animals treated with ⁹⁰Y-Fab′ still clearly contain multiple metastases, no tumor is visible in animals treated with ²¹³Bi-labeled CO17-1A Fab′, which is in accordance to their relapse-free long-term survival.

Fig. 5 shows an external scintigraphic scan of two mice bearing multiple GW-39 liver metastases at 1 h after the injection of 700 μCi of ²¹³Bi-labeled CO17-1A Fab′ without (left) or with d-lysine administration (right). The scan nicely visualizes the effect of lysine administration on the renal accretion of ²¹³Bi-labeled CO17-1A Fab′, and it shows good tumor uptake in multiple metastatic lesions in the liver.

The radiation dosimetry of ²¹³Bi- versus ⁹⁰Y-labeled CO17-1A Fab′ in s.c. GW-39 xenograft-bearing nude mice, based on the biodistribution studies, is summarized in Table 1. Because we have shown previously that, with high-energy β emitters, such as ⁹⁰Y, cross-organ irradiation may play a major role (32, 38), the model of Yoriyaz and Stabin (30) was used. For the dosimetry of ⁹⁰Y-CO17-1A Fab′, the table accounts for cross-organ radiation, and for ²¹³Bi-Fab′, it differentiates into doses resulting merely from the self-absorption of the α particles of ²¹³Po or including the β and γ rays of ²¹³Bi (with potential interorgan cross-fire). As expected for monovalent antibody fragments, the kidneys were the organs with the highest radiation doses. At the conventional MTDs (i.e., 700 μCi of ²¹³Bi- and 250 μCi of ⁹⁰Y-Fab′), the renal doses were 50–70 Gy without and ∼20 Gy with lysine protection. This is in accordance with the observed biological effects with chronic radiation nephropathy without amino acid administration, but no significant renal damage was seen with amino acid administration. Interestingly, the blood doses, being representative of the red marrow (39), were also very similar between the two labels (5–8 Gy at the MTD without and 9–12 Gy at the MTD with BMT), regardless of the LET character of the respective emitter. In accordance to the long path length of yttrium-90’s β emissions (32, 38), significant cross-fire doses occur from the kidneys to the liver and spleen (compare hepatic and splenic doses with and without lysine protection), whereas cross-organ doses are negligible with ²¹³Bi.

Fig. 6 shows the relation between calculated tumor doses and the resulting antitumor efficacy (40). As a result of tumor growth retardation being the major observed effect in all treated groups, the correlation between mean tumor doses and the extent of induced growth retardation was analyzed. Because GW-39 is a very rapidly growing cell line with tumor volume doubling times of <1 week (7, 19, 40), the correlation between tumor doses and the mean time needed for the quadruplication of tumor volume (i.e., two doubling times) was analyzed (40). In this graph, individual points represent the tumor volume quadruplication times for the calculated mean tumor doses (as derived from the respective injected activities of ²¹³Bi- versus ⁹⁰Y-labeled CO17-1A Fab′; Table 1) of different cohorts in separate experiments. With ⁹⁰Y-Fab′, below a tumor dose of ∼5 Gy, no significant tumor growth delay was noticeable. Above this threshold, tumor growth was retarded in a dose-dependent fashion up to 7-fold at ∼20 Gy (the highest possible dose at the MTD with BMT). The resulting nonlinear regression curve (regression coefficient: r = 0.96) resembles a “shoulder curve,” with its shoulder at ∼5 Gy. In contrast, no threshold dose was seen with ²¹³Bi-CO17-1A Fab′; the time to tumor volume quadruplication was prolonged in an almost linearly increasing manner with increasing tumor dose (regression coefficient: r = 0.98). Comparable antitumor efficacy was seen with ²¹³Bi-CO17-1A Fab′ at doses that were approximately half to a third as high as those needed with the ⁹⁰Y label.

α particles are monoenergetic, high-energy helium nuclei (⁴He²⁺) with high LET (11). Although there are ∼100 α-emitting radionuclides, very few possess physical properties that render them potential candidates for medical applications (recently reviewed in Ref. 11). Several authors and studies have postulated or already provided evidence that high-LET radiation emitters may have therapeutic advantages in RIT (reviewed in Ref. 11). The mean LET values of the nuclides tested in this study are 0.2 keV/μm for the β particles of ⁹⁰Y, as compared to ∼100 keV/μm for the α particles of ²¹³Bi/Po (for the decay schemes of both radionuclides, see Table 2; Ref. 11). On the other hand, the mean range in tissue is 3.96 mm for the ⁹⁰Y β⁻ as compared to only ∼0.1 mm for the ²¹³Bi/Po α²⁺ particles.

²¹³Bi has the advantage that it can be readily obtained from an in-house ²²⁵Ac generator, from which it is eluted as (BiI₅)²⁻ anion (16, 17, 26). Although special bifunctional DTPA derivatives (e.g., cyclohexyl-DTPA) have been designed for bismuth chelation (41), our data using “conventional” bifunctional Bz-DTPA show that this chelator is sufficiently stable, at least with the fast-clearing monovalent antibody fragments used in this study, whereas some earlier data with Bz-DTPA-IgG had shown some loss of bismuth out of chelation over time (42). Our biodistribution data in nude mice show no significant differences in the blood clearance, organ or tumor uptake between ²¹³Bi- or ⁸⁸Y-labeled Bz-DTPA-derivatized immunoconjugates, indicating sufficient metabolic stability of these Bz-DTPA-Fab′ fragments in vivo. Incorporation yields were ≥ 90% within a few minutes even at ambient room temperature, requiring <15 min for the preparation of the final radiolabeled immunoconjugate.

As expected, the tumor uptake was very rapid, which is an important prerequisite for therapy with a short-lived radionuclide such as ²¹³Bi. Also as expected with a monovalent Fab′ (20, 21), the kidneys were the organs with the highest physiological uptake. In contrast to the renal accretion of free Bi³⁺, which could not be prevented by lysine administration (Fig. 1), the renal uptake of both Bi- and Y-labeled Fab′ was reduced 5-fold by treatment with cationic amino acids, which is in accordance to our earlier findings (20, 21). The almost identical biodistribution of both Bi- and Y-labeled conjugates confirms, together with the low bone accretion of the Y-conjugates, the fact that bifunctional DTPA derivatives are sufficiently stable for both radionuclides in vivo, although further improved metabolic stability may be gained by using the more rigid, bifunctional cyclohexyl-DTPA backbone mentioned above (41, 43).

As we reported previously for ⁹⁰Y-labeled Fab fragments (21, 38), with respect to acute radiation toxicity, the bone marrow is the first-line dose-limiting organ with ²¹³Bi-labeled Fab′ fragments as well. The nadirs of WBC and platelet counts were similar between both α emitter- and β emitter-labeled conjugates, but interestingly, the peripheral blood counts recovered faster from the bone marrow injury by the high-LET but shorter-lived ²¹³Bi than by the low-LET but longer-lived ⁹⁰Y (Fig. 2). This observation is in accordance with earlier observations of the dependence of bone marrow recovery upon the effective half-life of the radioconjugate in the blood or bone marrow (19, 32). Most surprising, however, is the observation that the maximum tolerated blood doses, which are generally accepted as representative of the red marrow doses (the latter being difficult to assess directly; Ref. 39), were almost identical for both low- and high-LET radiation (Table 1). This is in sharp contrast to the general assumption of a RBE factor of 10–20 for α particles (44, 45, 46). This assumption relies on earlier in vitro data. However, this factor has never been reproducible in vivo, and more recent studies have shown that the RBE in vivo is dependent upon the α particle energy and the LET (47, 48). In a mouse testes spermatogenesis model, Howell et al. (47, 48) demonstrated RBE values between 3 and 9 (the higher the α particle energy, the lower the RBE). Even more important, Charlton et al. (49) modeled the hemopoietic stem cell survival following irradiation by α particles of various energies, using Monte Carlo methodology to calculate passages of α particles, originating in three geometries, through the nuclei of the stem cells. Survival predictions were made from probability mapping of surviving each passage as a function of LET. These authors predicted, for α particles of various energies, RBE values that lie between 1.5 and 2.1 in the human red marrow, depending upon their energy (again, the higher the energy, the lower the RBE), with a predicted RBE of ∼1.5 for the 8.4-MeV α particles of ²¹³Bi/²¹³Po (49). When the maximum tolerated blood doses of ²¹³Bi-Fab′ are compared to those of ⁹⁰Y-Fab′ (cf. Table 1), the MTD with the α emitter appears to be even slightly higher than the one with the low-LET β⁻ emitter. Thus, most surprisingly, our data suggest an in vivo RBE value with respect to bone marrow toxicity of close to one as compared to low-LET β⁻ radiation. Given the relatively long (0.1 mm) path length of the high (8.4-MeV) energy α particles of ²¹³Bi/²¹³Po, it appears unlikely that heterogeneity of the microdistribution may cause these effects, but future studies must address this point in more detail. However, recent human data with the ²¹³Bi-labeled anti-CD33 antibody HuM195 also seem to confirm an in vivo RBE value with respect to bone marrow toxicity of close to 1, as compared to low-LET β or γ radiation (50).⁵

Radiation-induced renal toxicity is a major concern in the therapeutic application of antibody fragments or peptides radiolabeled with metals. We have shown previously (21, 51) that, for ⁹⁰Y-labeled conjugates, two distinct pathological features must be distinguished in this context: kidney doses of more than ∼100 Gy result, within 1–2 weeks, in an acute renal failure which is histologically characterized by an acute tubular necrosis; this form of acute radiation nephritis is lethal within hours or a few days of its onset. In contrast, at doses between ∼60 and 100 Gy, very few or no symptoms are found in the first weeks after the irradiation, but progressive chronic renal failure develops within months; histologically, this pathophysiological entity is characterized by a progressive glomerular sclerosis, tubular atrophy, arteriolar intimal fibrosis, and severe interstitial scarring (21, 51). This more chronic form of radiation nephropathy corresponds to the clinically well known syndrome of “chronic radiation nephritis,” which is characterized by hypertension (moderate to malignant), edema, proteinuria, anemia, and uremia (21, 51). Interestingly, in contrast to earlier studies (21), non-lysine-protected animals showed a short episode of transient slight but significant BUN elevation, whereas BUN levels remained low in lysine-treated animals. However, after ∼4–6 weeks, BUN levels steadily begin to rise in animals treated at the MTD without lysine protection (Fig. 2). The development of terminal renal insufficiency was more rapid and more dramatic with ⁹⁰Y than with ²¹³Bi, which would be in accordance with the radiation dosimetry (70 versus 54 Gy to the kidneys), but again, this result is highly surprising, because one would expect a much higher biological efficacy of the high-LET α emitter. Obviously, α particle RBE values are lower than originally expected in kidneys as well, similar to our observations in the bone marrow (see above). Most importantly however, none of the lysine-protected animals developed signs of radiation nephropathy, and histology also failed to show significant alterations, which holds true for animals given a dose-intensified treatment regimen with BMT as well.

At equitoxic dosing (either at dose levels below the conventional MTD, at the conventional MTD, or at the MTD with BMT), RIT with ²¹³Bi-Fab′ was therapeutically superior to its conventional β-emitting ⁹⁰Y analogue. The same holds true for the liver metastatic model, in which ²¹³Bi-Fab′ resulted in a >90% cure rate, in contrast to only 20% with ⁹⁰Y. Although, due to its long path length, ⁹⁰Y may not be the ideal radionuclide for treating micrometastates, cure rates with the shorter-range ¹³¹I-Fab′ were very comparable to those with ⁹⁰Y-Fab′, as presented here (52),^4,6 supporting our view that the superiority of ²¹³Bi is not just due to the shorter path length of its α particles. Thus, the high-LET α emitter clearly appears therapeutically superior to the conventional low-LET radiation emitters. This is in accordance to the earlier findings of Bloomer et al. (18), who compared the therapeutic efficacy of α-emitting (²¹¹At) and β-emitting (³²P, ⁹⁰Y, and ¹⁶⁵Dy) radiocolloids for the treatment of experimental malignant ascites and demonstrated an unequivocal advantage for the high-LET α emitter.

We have shown earlier that plotting tumor volume multiplication times against tumor doses (Fig. 6) yields a diagram resembling a sort of inverted cell survival curves (7, 19, 40) known from external beam radiation of cells in cell culture (45, 46). This is not surprising, because the tumor volume is a function of the total number of cells, and the change in tumor volume is a function of the number of viable cells that are able to proliferate. Below tumor doses of ∼5–10 Gy, the effects on tumor growth delay were minimal. In contrast, the tumor volume multiplication curve under ²¹³Bi-treatment lacks a pronounced shoulder and displays an almost linearly increasing efficacy with increasing dose. Interestingly, comparable antitumor efficacy with ²¹³Bi results at approximately half to one-third of the doses needed with ⁹⁰Y. This is very similar to the improved antitumor efficacy which we have demonstrated earlier for Auger/conversion electron emitters internalized into the cytoplasm of the tumor cells (7).

In summary, the data of this study show that RIT with α emitters may be therapeutically more effective than RIT with conventional β emitters. Surprisingly, the red marrow and the kidney seem to tolerate similar radiation doses of the high-LET α as compared to the low-LET β emitters, whereas the RBE in the tumor is 2–3-fold higher, providing a clear therapeutic advantage in favor of the high-LET α emitter. Furthermore, radiation nephrotoxicity of the antibody fragments can be overcome by basic amino acid administration. Due to its short physical t_1/2, ²¹³Bi appears to be especially suitable for use in conjunction with fast-clearing fragments; its 440-keV γ emission also can be used for quantitation by external scintigraphy.

We thank Dr. O. Proske (Glaxo Wellcome, Inc., Hamburg, Germany), for providing us with the CO17-1A MAb. The expert advice of Prof. Dr. M. Oellerich (Department of Clinical Chemistry, Georg-August-University of Göttingen, Göttingen, Germany) with respect to blood count determinations and blood chemistry is gratefully acknowledged. Furthermore, we thank Drs. K. Nebendahl and O. Schunck (Animal Care Facility, Georg-August-University of Göttingen) for their valuable advice with respect to animal care.