Biodistribution Properties of ¹¹¹Indium-labeled C-Functionalized trans-Cyclohexyl Diethylenetriaminepentaacetic Acid Humanized 3S193 Diabody and F(ab′)₂ Constructs in a Breast Carcinoma Xenograft Model

Carcinoma of the breast is the most common malignancy affecting female patients and is the second most frequent cause of death from cancer (1). Although an increasing number of cases are surgically cured as a result of early diagnosis and effective systemic adjuvant chemotherapy, some patients present with advanced disease, and many others develop recurrent and/or metastatic cancer. These patients may be palliated using a variety of treatments including hormonal and cytotoxic therapy; however, response is usually temporary, and it is difficult to demonstrate prolongation of survival (2). In view of this, efforts to identify new active drugs and treatment regimens continue. Conventional radiotherapy and chemotherapy target both normal and neoplastic cells, relying upon the enhanced sensitivity of rapidly dividing cancer cells to achieve preferential killing (3). In contrast, mAb² constructs can be directed more specifically against tumor-associated antigens, which may be expressed on the surface of the tumor cells (4).

A number of tumor antigens have been shown to have high expression on malignant breast cancer cells, including HER2/neu, MUC, CEA, and Le^y (5). Le^y is a member of the H-type II blood group-related antigens and is a difucosylated tetrasaccharide that is carried on both glycoproteins and glycolipids, with chemical structure Fucα1→2Galβ→4(Fucα1→3)GlcNAc (6). The Le^y antigen is located on the epithelial cell surface and is expressed in >70% of epithelial cancers including breast, prostate, pancreatic, colon, ovary, and lung cancer. Expression is also observed on 60% of endometrial and gastric carcinomas (7, 8, 9). The Le^y tumor-associated antigen is an attractive target for immunotherapy because it is broadly applicable to many types of carcinomas, its expression is homogeneous in both primary and metastatic lesions, and there is high antigen density on the tumor cell surface (6, 8). The parental 3S193 mAb was produced using standard hybridoma technology after immunization of BALB/c nude mice with Le^y-expressing MCF-7 breast cancer cells. 3S193 has been humanized (hu3S193) by CDR grafting, and reactivity for Le^y was confirmed to be similar to the murine version (10). Recent studies have extended these early studies and confirmed that hu3S193 can be produced and formulated on a large scale as a potential immunotherapeutic reagent against selected solid tumors (11).

This report explores the use of new smaller molecular weight multivalent analogues of the parent hu3S193 antibody. Antibody fragments including F(ab′)₂ and scFv molecules such as diabody and triabodies [in which the variable heavy (V_H) and variable light (V_L) regions of the construct are stabilized by a flexible linker], are advantageous in that they offer improved clinical pharmacokinetics because of their smaller size and their reduced immunogenicity (because of the absence of an Fc domain; Refs. 12, 13, 14). They may also have more homogeneous tumor penetration, which may circumvent problems of inaccessibility/poor penetration into the tumor mass, and display improved specific tumor retention associated with multivalency (4, 15, 16). However, their rapid clearance from the blood often results in a lower absolute uptake in the tumor (15, 17, 18, 19). Hence, these smaller molecules are often more suitable for diagnostic imaging, rather than as therapeutic reagents, particularly because their rapid clearance from background tissue permits early imaging of patients/tumors after administration of the radioconjugate. In this study, the bivalent diabody and F(ab′)₂ versions of hu3S193 were characterized in vitro for their binding, affinity, and stability properties. The in vivo biodistribution of the ¹¹¹In-CHX-A"-DTPA-labeled hu3S193 diabody and F(ab′)₂ was also evaluated using a BALB/c nude mouse breast carcinoma xenograft model to ascertain their potential in targeting Le^y-expressing solid tumors.

MCF-7, a Le^y-expressing human breast adenocarcinoma cell line (20), was obtained from American Type Culture Collection (Rockville, MD). SW1222, a Le^y-negative human colonic cancer cell line, was obtained from the New York branch of the Ludwig Institute. Both MCF-7 and SW1222 cells were cultured in RPMI supplemented with 10% FCS (CSL Biosciences, Melbourne, Victoria, Australia) and 50 units/ml penicillin-streptomycin (Life Technologies, Inc., Grand Island, NY). The hu3S193 mAb was obtained from the Biological Production Facility (Ludwig Institute for Cancer Research, Melbourne, Victoria, Australia; Ref. 11).

The heat-inducible bacterial expression vector (pPOW3) was used to synthesize diabodies into the periplasm of Escherichia coli Top 10 cells (Invitrogen, Carlsbad, CA) according to conditions described previously (21, 22). In brief, the hu3S193 diabody comprised two identical scFv molecules in which V_H and V_L domains were linked via a five-amino acid residue (Gly₄Ser) linker. A FLAG octapeptide affinity tag (DYKDDDDK) at the COOH terminus of the V_L domain assisted with protein purification. Diabody production was performed in 10 liter fermenter batches; the cells were then harvested by centrifugation at 4°C for 5 min at 6000 × g and then stored frozen at −80°C. The pellet was thawed and processed to yield a soluble periplasmic protein fraction of ∼3 liters. For hu3S193, the protocol produced 1 mg/liter of purified, soluble diabody of the expected size (M_r ∼54,000) with <1% of higher molecular weight aggregates detectable by HPLC size exclusion chromatography.³ The hu3S193 diabody could be stored frozen at 1 mg/ml and thawed without aggregation in Immunopure Gentle Ag/Ab Elution Buffer (Pierce, Rockford, IL).

The parental hu3S193 mAb (10 mg/ml) was digested by incubation with pepsin (Sigma Chemical Co., St. Louis, MO) in 0.2 m sodium acetate buffer (pH 3.5). The digestion was performed at room temperature for 2 h using an optimal antibody:pepsin ratio of 200:1. After digestion, the reaction mixture was neutralized to pH 7.0 with 1 m NaOH. Antibody purification was performed using Protein A-Sepharose affinity chromatography (Amersham Pharmacia, Uppsala, Sweden), followed by size exclusion chromatography on a Sephacryl S-200 column (Amersham Pharmacia) to separate the F(ab′)₂ fragments. The purified protein was concentrated using a YM30 Amicon ultrafiltration membrane (Amicon, Beverly, MA) to a concentration of 5.4 mg/ml. Protein purity was assessed by SDS-PAGE and FPLC analyses.

Proteins were radiolabeled either directly with ¹²⁵I or via a bifunctional metal ion chelating agent, CHX-A"-DTPA, for ¹¹¹In. Antibody constructs were labeled with ¹¹¹In (NEN Life Science Products, Boston, MA), using a modification of a method described previously (23). In brief, proteins were placed in 8-mm Spectra/Por cellulose ester membrane (Spectrum, Houston, TX) with a molecular weight cutoff of 10,000 and dialyzed against 0.1 m borate buffer containing 2.5 g/l of Chelex 100 resin (Bio-Rad, Hercules, CA; pH 8.0) to remove heavy metals. CHX-A"-DTPA was added in a molar excess of 3:1 and incubated at room temperature overnight, protected from light. Excess unbound antibody was removed by dialysis for 8 h with 20 mm sodium acetate buffer containing 0.15 m NaCl (pH 6.4). ¹¹¹In was bound to CHX-A"-DTPA antibody conjugates under mildly acidic conditions (pH 5.5) for 20 min, and then the pH was raised to 7.0 by the addition of small aliquots of 2.0 m sodium acetate, followed by 10 mm EDTA for quenching unbound ¹¹¹In. The radiolabeled mixture was purified by centrifugal desalting on a Sepharose G50 column (Pharmacia, Uppsala, Sweden) equilibrated with PBS. Radioiodination with ¹²⁵I (NEN Life Science Products, Boston, MA) was performed via a modified chloramine-T reaction (24), using a chloramine T:protein ratio of 4:1. Chloramine-T (Merck, Darmstadt, Germany) was dissolved in 0.5 m potassium phosphate buffer (pH 7.0); after a brief 2-min incubation period, the reaction was terminated by the addition of a 5-fold excess of sodium metabisulphite (BDH Chemicals, Melbourne, Victoria, Australia), again dissolved in 0.5 m potassium buffer. The product was purified through a desalting P6DG column (Bio-Rad) equilibrated in PBS.

The immunoreactive fraction of the radiolabeled constructs with MCF-7 cells was determined according to Lindmo et al. (25). Briefly, 20 ng of ¹¹¹In-CHX-A"-DTPA diabody or F(ab′)₂ were added to an increasing concentration of MCF-7 cells (antigen excess) or SW1222 control cells. The assay was incubated for 45 min at room temperature with continuous mixing to keep the cells in suspension. Cells were washed three times to remove any unbound antibody, and pellets were measured in a dual gamma scintillation counter (Cobra II, Model 5002, Auto-gamma; Packard Instruments, Canberra, Australian Capital Territory, Australia) that is capable of measuring the cpm of two isotopes simultaneously. The percentage of binding of hu3S193 to MCF-7 cells was calculated by the formula: (cpm cell pellet/mean cpm of radioactive antibody standards) × 100. Immunoreactivity was obtained by plotting the double reciprocal plot for binding against cell concentration using SigmaPlot for Windows (Jandel Scientific, San Rafael, CA).

Scatchard analysis was used to determine the affinity constant (K_a) and number of antigen binding sites/cell (25). Unlabeled hu3S193, at concentrations ranging from 0.02 to 200 μg/ml, was mixed with 15 ng of labeled ¹¹¹In-CHX-A"-DTPA diabody or F(ab′)₂. The antibody mixture was added to 15 × 10⁶ MCF-7 cells and incubated for 45 min at room temperature with continuous mixing. Specific binding (X axis) was plotted against the ratio of specific binding to concentration of free radioligand (Y axis). The intercept at the abscissa represented the binding capacity of the cell (antibody molecules bound/cell). The apparent association constant can be derived from the slope of the line. A detailed explanation of the calculations can be obtained from Clarke et al. (26).

Serum stability was assessed by incubating 2 μg of each radiolabeled protein in 200 μl of healthy donor human serum at 37°C for a 5-day period. The radiochemical purity of the samples was analyzed by FPLC analysis at 1, 4, 24, and 48 h after incubation, and single-point immunoreactivity assays at 0 (no incubation), 0.5, 1, 2, 4, 16, 24, and 48 h after incubation, with additional time points of 72, 96, and 120 h after incubation for F(ab′)₂. F(ab′)₂ analysis was extended to 120 h to account for the longer half-life of the molecule.

Twenty ng of each radiolabeled construct were incubated with 10 × 10⁶ MCF-7 cells alone or in the presence of 300 μg of competing unlabeled, intact hu3S193 mAb. The 45-min incubation was performed on a rotary mixer at room temperature. Cells were washed three times to remove any unbound antibody, and pellets were measured in a dual gamma scintillation counter (Packard Instruments). Duplicate preincubation radioconjugate standards were measured concurrently with the cell pellet, and binding was expressed as a percentage compared with duplicate preincubation radioconjugate standards.

The constructs were analyzed by FPLC before and after labeling to determine the integrity of the radiolabeled proteins. The samples were eluted isocratically on a Superdex 200 HR10/30 column (Amersham Pharmacia) using 50 mm phosphate, 150 mm NaCl (pH 7.0) at a flow rate of 0.5 ml/min. Eluted radiolabeled antibodies were detected by UV absorbance at 280 nm by a Beckman System Gold detector (Beckman Instruments, Fullerton, CA), and radioactivity was measured by Packard Radiomatic Flo One (Packard Instruments). For serum stability analysis, 60 1-min (0.5-ml fractions) were collected using the Pharmacia Radi-Frac fraction collector (Amersham Pharmacia) on the column described above. The amount of protein in each fraction was detected by UV absorbance at 280 nm using a Life Science UV/Vis Spectrophotometer DU 530 (Beckman Instruments), and radioactivity was measured in a dual gamma scintillation counter (Packard Instruments).

Transchelation with serum proteins was analyzed by incubating ¹¹¹In-CHX-A"-DTPA diabody (0.24 μCi radioactivity/4 μg protein) and ¹¹¹In-CHX-A"-DTPA F(ab′)₂ (2.9 μCi radioactivity/2 μg protein) in normal human serum for 4 h at 37°C. Samples were separated by 4–20% Tris-HCl SDS-PAGE under nonreducing conditions (Bio-Rad) at 200 V for 1 h. Protein bands were visualized with Coomassie blue staining and autoradiography using Hyperfilm MP (Amersham Life Science, Buckinghamshire, England) and image intensifying screens at −70°C.

To establish MCF-7 human breast cancer xenografts, BALB/c nude mice obtained from the Walter and Elisa Hall Institute (Melbourne, Victoria, Australia) were supplemented with exogenous estrogen. A slow release estrogen pellet (0.72 mg of 17β-estradiol; Innovative Research of America, Sarasota, FL) was implanted s.c. in the scapular region, using a small incision closed by a single stitch using silk 5/0 nonabsorbable sutures (B. Braun Surgical, Melsungen, Germany) at the same time as injection of tumor cells. Mice received 20 × 10⁶ MCF-7 cells s.c. into the left inguinal mammary line, and 2 weeks later, 3 × 10⁶ of the more rapidly dividing SW1222 cells were injected into the right inguinal mammary line (Le^y-negative control cell line). The establishment of tumors was monitored by tumor volume measurement [(L × W²)/2], where length was the longest axis and width the measurement at right angles to length (26).

Seventy-five mice (a combination of two studies) with a mean ± SD MCF-7 tumor volume of 457 ± 197 mg were divided into two groups. All constructs were administered i.v. via the retro-orbital plexus. The first group of 29 mice was injected with ¹¹¹In-CHX-A"-DTPA diabody (total of 5 μg of antibody; 3.9 μCi of radioactivity). The second group of 46 mice received ¹¹¹In-CHX-A"-DTPA F(ab′)₂ (total of 1.6 μg of antibody; 3.3 μCi of radioactivity). At the designated time points p.i., groups of mice (n = 2–5) were sacrificed by Ethrane anesthesia, mice were bled by cardiac puncture, and tumors and organs [skin, liver, spleen, small intestine, stomach, kidney, brain, bone (femur), lungs, and heart] were resected immediately. All samples were counted in a dual gamma scintillation counter (Packard Instruments). Triplicate standards prepared from the injected material were counted at each time point with tissue and tumor samples enabling calculations to be corrected for the physical decay of the isotopes. Results of the labeled antibody distribution over time were expressed as %ID/g. The tissue distribution data were calculated as the mean %ID/g ± SD for the diabody (n = 4) and F(ab′)₂ (n = 5), per time point.

Whole body imaging of mice was performed at 1, 4, and 24 h after radioconjugate injection. Mice were anesthetized at these time points by i.m. injection of 150 mg/kg ketamine (Troy Laboratories, Smithfield, New South Wales, Australia) and 5 mg/kg diazepam (Bayer Australia, Sydney, New South Wales, Australia) and placed under a Trionix Biad gamma camera (Biad Trionix Research Laboratories, Twinsburg, OH). Twenty-min images were obtained at each time point. Images were acquired in a 256 × 256-bit matrix, and a standard of known concentration was included in the field of view.

Pharmacokinetics (T_1/2α and T_1/2β) were determined for ¹¹¹In-labeled radioconjugates using a standard curve-fitting program (SAAM II; University of Washington, Seattle, WA), assuming a four-parameter, two-compartment model to calculate pharmacokinetic results.

FPLC size-exclusion chromatography showed that both the diabody and F(ab′)₂ were stable proteins during freeze/thaw cycles (data not shown). Proteins were thawed when required and radiolabeled separately with ¹¹¹In and ¹²⁵I. Intact hu3S193 was radiolabeled with ¹¹¹In (35.4% labeling efficiency) as a control. The labeling efficiency of the diabody and F(ab′)₂ was 14.4 and 44.1% with ¹¹¹In and 5.0 and 50.0% with ¹²⁵I, respectively. The constructs were analyzed by FPLC before (data not shown) and directly after labeling to verify the integrity of the radiolabeled proteins (Figs. 1,A and 2 A). Only the ¹¹¹In-labeled constructs were used in the biodistribution study.

A Lindmo cell binding assay was performed for each of the immunoconjugates to evaluate their antigen-binding capabilities after radiolabeling. The assay was performed using Le^y-positive MCF-7 breast carcinoma cells, and to assess whether there was any nonspecific binding of the antibody constructs, control Le^y-negative SW1222 colon carcinoma cells were included. The ¹¹¹In and ¹²⁵I diabody radioconjugate binding curves reached saturation as expected (Fig. 3,A). The ¹²⁵I diabody had a 2-fold higher immunoreactivity than its ¹¹¹In-labeled counterpart (83.5% compared with 41.3%; Fig. 3,A). The ¹¹¹In and ¹²⁵I F(ab′)₂ binding curves did not appear to reach a plateau and displayed marginally higher immunoreactivity with ¹¹¹In (¹¹¹In, 58.6% and ¹²⁵I, 51.1%; Fig. 3 B). No binding to SW1222 cells was observed, indicating that the binding was specific to Le^y antigen (data not shown).

Scatchard analysis was used to calculate the apparent association constant (K_a) and number of antigen binding sites/cell (Table 1). The affinity of both the diabody and F(ab′)₂ was lower than that of the intact hu3S193 antibody, the affinity of which is within the range expected for binding the Le^y carbohydrate antigen (27). The affinity of the diabody was ∼10-fold lower than the parental hu3S193 intact antibody. The F(ab′)₂ displayed a slightly higher affinity than the diabody, most likely attributable to the orientation of the binding domains, in addition to the increased stability and immunoreactivity observed for this molecule. The number of binding sites/cell was similar for both constructs.

The single-point immunoreactivity assay was used to evaluate the stability of the diabody and F(ab′)₂ after incubation in normal human serum for up to 5 days (Table 2). The ¹¹¹In-labeled diabody displayed a gradual reduction in antigen-binding capacity with time. A loss of one-third immunoreactivity was observed within 30 min of incubation in serum, followed by over two-thirds reduction in binding ability at 2 and 4 h after incubation. By 16 h after incubation in serum, all immunoreactivity was lost for the diabody construct. This loss in immunoreactivity correlates with the FPLC chromatographic profiles of the diabody (Fig. 1). FPLC analysis indicated high radiochemical purity of the diabody directly after labeling with ¹¹¹In, with most of the radioconjugate eluting at the expected time of 30 min (a smaller molecular weight species also appeared to label); however, this was followed by the formation of a high molecular weight aggregate (eluting at 18 min, in a similar position to thyroglobulin). The aggregate was evident as a small peak at 1 h after incubation (Fig. 1,B) but increased markedly by 4 h incubation in serum. After 48 h incubation, most of the diabody was present as the high molecular weight species (Fig. 1,C). The ¹¹¹In-labeled F(ab′)₂ was more stable in serum. A loss of one-third immunoreactivity was observed between 48 and 96 h after incubation, with a further reduction (50% loss of initial immunoreactivity) observed at 120 h (Table 2). The radiolabeled hu3S193 F(ab′)₂ indicated no aggregation, with the majority of ¹¹¹In eluting with F(ab′)₂ protein at the expected time of 26 min at all time points investigated (Fig. 2). SDS-PAGE analysis of radiolabeled diabody and F(ab′)₂ incubated in serum for 4 h examined the nature of the aggregates further. The radiolabeled constructs were visualized by autoradiography. The radiolabeled diabody migrating as a single band at the expected size of M_r ∼30,000, indicating no transchelation to serum proteins. The F(ab′)₂ migrated as a single band at the expected size of M_r ∼120,000, with a minor band also present at M_r ∼50,000, indicating a labeled breakdown product (data not shown).

The in vivo targeting potential of the ¹¹¹In-labeled diabody and F(ab′)₂ was assessed in BALB/c nude mice bearing MCF-7 Le^y-positive tumors and control SW1222 Le^y-negative tumors. The tumor, blood, and normal tissue retention of each radiolabeled construct was determined at 10 and 30 min and 1, 2, 4, 8, 24, and 48 h, with additional time points of 72 and 96 h for the F(ab′)₂. The diabody study was only extended to 48 h because of the very fast clearance of this construct, which had been observed in preliminary experiments. After injection, the diabody displayed a rapid equilibration phase (T_1/2α, 0.45 h) and subsequent slower elimination phase (T_1/2β, 2.7 h) from the circulation; the F(ab′)₂ displayed slightly longer elimination (T_1/2α, 1.76 h) and equilibration (T_1/2β, 12.6 h) phases (Fig. 4). In contrast, specific and prolonged localization of the radioconjugates to the Le^y-expressing MCF-7 tumor was observed as demonstrated by the tumor:blood ratios (Table 3) and the blood clearance and tumor localization (Figs. 4,5,6; Table 3). Maximal uptake of radiolabeled diabody in the MCF-7 experimental tumor was seen at 1 h p.i., 4.7 ± 0.6%, whereas the F(ab′)₂ peaked at 8 h p.i., 14.2 ± 2.4% ID/g. Binding to the control SW1222 tumor was negligible for both constructs (Figs. 4,5,6). The biodistribution analysis of both conjugates showed high kidney uptake (Figs. 5 and 6), without significant localization to the other normal tissues investigated (Figs. 5 and 6).

Localization of ¹¹¹In-labeled diabody and ¹¹¹In-labeled F(ab′)₂ to MCF-7 tumors in BALB/c nude mice over time is presented in Figs. 7 and 8. At 1 h p.i., no tumor localization was evident for either of the constructs, with the image showing generalized blood pool activity. By 4 and 24 h p.i., minimal uptake was observed in the control tumor, but definite specific localization of radioconjugate was observed in the MCF-7 xenograft for both of the constructs. Some cardiac blood pool activity and kidney uptake was also observed at these time points (Figs. 7 and 8).

Recombinant scFv, multivalent molecules, and antibody fragments are emerging as promising new molecules for potential use in tumor diagnosis and therapy (17, 28, 29). These constructs have a number of important features, including a smaller size that may permit more effective and homogeneous tumor penetration, increased avidity and retention time of antigen:antibody interaction associated with multivalency, and a potentially reduced immunogenicity (18, 30, 31). However, they also demonstrate a rapid clearance from the circulation, resulting in a low absolute tumor uptake, and the in vivo behavior of the new construct designs remains to be fully established (15, 32, 33). The purpose of this study was to evaluate the potential of radiolabeled diabody and F(ab′)₂ constructs of the humanized antibody hu3S193 and to determine the in vivo behavior and targeting potential to Le^y-expressing human breast carcinoma xenografts. Both constructs were characterized for their in vitro binding, affinity, and stability properties, as well as their in vivo biodistribution and pharmacokinetic characteristics. No prior study has rigorously examined diabody and F(ab′)₂ constructs concurrently in the same tumor model.

The differences observed in the radiolabeling efficiency between the diabody and F(ab′)₂ constructs may be attributable to the reduced number of accessible lysine and tyrosine residues present in the smaller hu3S193 diabody molecule. The scFv of hu3S193 contains seven lysine residues as potential sites for chelation and labeling with ¹¹¹In, compared with ∼100 residues found in the intact antibody. A lysine residue is located in CDR two of both the variable heavy and light chains. Binding of CHX-A"-DTPA chelate at these sites may have contributed to the decreased binding capacity of the diabody. Hence, site-directed mutagenesis experiments are planned in an attempt to improve the labeling efficiency of the ¹¹¹In-labeled diabody compared with the iodinated diabody construct. Additional lysine residues are planned to be inserted at the NH₂ or COOH terminus, or alternatively lysines will be substituted in the binding domains to avoid potential obstruction during the chelation process.

The differences seen in the immunoreactivity between the two isotopes may be attributable to the nature of the constructs and the different ways in which these isotopes are catabolized within the cell. Furthermore, the Scatchard results show a slightly decreased affinity for the diabody compared with the intact and F(ab′)₂ hu3S193 (Table 1), and the lower affinity may reduce the immunoreactivity of the diabody (29, 34). The in vitro plasma stability analysis of the diabody indicated that there was loss of stability with time, with a complete loss in immunoreactivity observed by 16 h incubation at 37°C (Table 2), and the subsequent formation of a larger molecular weight species observed by FPLC (Fig. 1, B and C). However, given the fast serum clearance of this construct (T_1/2β, 2.7 h; Fig. 4,A), the stability and immunoreactivity of the diabody at 4 h p.i. and beyond may not be as critical. This is also supported by the gamma camera images obtained at 4 and 24 h p.i. for this construct (Fig. 7). The F(ab′)₂ retained its immunoreactivity for a longer period of time (120 h), without any apparent degradation or aggregation being observed (Fig. 2). The stability of the bond between the CHX-A"-DTPA chelate and the ¹¹¹In radiometal has been rigorously investigated previously (23, 35, 36, 37), and the FPLC and binding data for the F(ab′)₂ stability analysis suggest that transchelation of the isotope does not occur, because the majority of the radioisotope elutes with the antibody construct. SDS-PAGE analysis demonstrated no transchelation of ¹¹¹In from the chelated hu3S193 diabody construct to serum proteins (results not shown), indicating that the aggregates observed by FPLC analysis were not attributable to transchelation.

¹¹¹In was selected for the in vivo biodistribution study because of the higher labeling efficiency observed compared with ¹²⁵I, its suitability for gamma camera imaging because of its favorable gamma emissions (173 and 247 keV), and its shorter half-life, which complements the rapid clearance of the constructs (38). Hence, the ¹²⁵I isotope was inappropriate for this study, particularly as we were evaluating the ability of the diabody molecule for potential as a diagnostic imaging reagent. It is important that isotope selection is tailored to match the clearance of the antibody. A recent study using ²¹³Bi CHX-A"-DTPA conjugated to an anti-HER2/neu diabody illustrated that the physical half-life of ²¹³Bi (47 min) was too short for optimal pairing with the diabody, resulting in a lack of demonstration of specific uptake in the tumor (39). In contrast, other scFv studies involving longer lived isotopes such as ¹²³I and ⁹⁰Y demonstrated a greater specific tumor accumulation (18, 40, 41). In addition, ¹¹¹In was selected because superior tumor targeting was observed previously with the intact hu3S193 antibody labeled with ¹¹¹In-CHX-A"-DTPA compared with the ¹²⁵I-labeled construct in the same MCF-7 xenograft mouse model used for the diabody and F(ab′)₂ experiments (30.1 ± 4%ID/g with ¹¹¹In-labeled hu3S193 compared with 10 ± 3.9%ID/g for ¹²⁵I-labeled hu3S193 at 48 h p.i.; Ref. 26). Preliminary experiments have also indicated that the hu3S193 mAb is internalized into MCF-7 cells and directed to the lysosomes (data not shown), as has been reported with other murine antibodies directed to the same antigen (42). Therefore, a radiometal such as ¹¹¹In, which is trapped within the lysosomal compartment after endocytosis, is a more desirable choice of isotope than ¹²⁵I, which is extruded rapidly from the cell (43).

The in vitro studies were extended to a comprehensive in vivo biodistribution study, using an established Le^y-positive breast cancer xenograft model in BALB/c athymic nude mice (26). The blood clearance was very rapid for the ¹¹¹In-labeled diabody, and specific localization of the radioconjugate was greater to the Le^y-expressing MCF-7 tumor compared with control SW1222 tumor (Figs. 4,5,6). Maximum tumor uptake for the hu3S193 diabody was observed at 1 h p.i. (4.7 ± 0.6%ID/g). Investigations with ¹²³I-labeled anti-CEA (8-amino acid linker) and ¹²⁵I- labeled anti-HER2/neu (5-amino acid linker) diabodies reported maximal tumor uptake at 2 h p.i. (13.68 ± 1.49%ID/g) and 4 h p.i. (10.1%ID/g), respectively (18, 30). The higher xenograft uptake observed in these two studies may reflect the higher affinity of these constructs, as determined by surface plasmon resonance (K_a of the anti-CEA diabody, 8.19 × 10¹⁰ m⁻¹; K_a of the anti-HER2/neu diabody, 4 × 10¹⁰ m⁻¹) and the lower affinity associated with antigens such as Le^y (K_a of hu3S193 diabody, 1.68 × 10⁶ m⁻¹). The low uptake of radiolabeled diabody in tumor xenografts may also be explained by the stability properties of this construct in serum, which indicated a rapid loss in binding accompanied by the formation of a large molecular weight aggregate between 0 and 4 h after incubation. The hu3S193 F(ab′)₂ displayed similarly quick serum clearance and MCF-7 tumor-specific uptake (Figs. 4,B and 6), also illustrated by the high resolution gamma camera images acquired at 4 and 24 h p.i. (Fig. 8). Maximal tumor uptake for the F(ab′)₂ was observed at 8 h p.i. (14.2 ± 2.4%ID/g). This agrees with other studies investigating F(ab′)₂ molecules, which reported similar or lower uptake. A study by Brown et al. (44), evaluating an anti-B72.3 F(ab′)₂ labeled with both ¹²⁵I and ¹¹¹In, observed maximal uptake at 6 h p.i. for both radioconjugates (¹²⁵I, 8.74 ± 1.7%ID/g; ¹¹¹In, 10.4 ± 1.8%ID/g). Two other studies with anti-CEA F(ab′)₂ constructs observed greatest uptake in tumor xenografts at 12 h p.i. [11.1 ± 2.3%ID/g (LoVo cells) and 11.2 ± 3.2%ID/g (Co112 cells); Ref. 45] and at 6 h p.i. (12.8%ID/g; Ref. 19).

Tumor uptake of diabody was observed through 48 h p.i. (2.36% ID/g), and high levels of F(ab′)₂ in tumors were observed for the 96 h of study (4.27% ID/g). The excellent tumor:blood ratios reflect the specific retention of both constructs in the target tumor and avidity of the constructs for the Le^y antigen (Table 3). At 4 h p.i., the tumor:blood ratios for the genetically engineered hu3S193, anti-CEA (18), and anti-HER2/neu (30) diabody constructs were 5:1, 5.8:1, and 1.5:1, respectively, and increased to 30:1, 48.7:1, and 9.5:1 at 24 h p.i., respectively. The tumor:blood ratio for the F(ab′)₂ at 4 h p.i. was 2:1 and increased to 55:1 at 24 h p.i.; this result is impressive when compared with other F(ab′)₂ molecules. For example, Pedley et al. (19) observed a tumor:blood ratio of 3.7:1 at 24 h p.i., whereas Vogel et al. (45) reported tumor:blood ratios ranging from 3.3 ± 0.3: 1 for LoVo cells and 3.8 ± 0.2: 1 for Co112 cells in similar studies involving an anti-CEA F(ab′)₂ construct. Our results are favorable in comparison with these other studies, particularly in view of the reduced binding avidity associated with antigens such as Le^y (27). The pharmacokinetic results also concur with the biodistribution observations illustrating a rapid terminal clearance from the blood.

The pharmacokinetic properties observed for the hu3S193 diabody and F(ab′)₂ constructs in this study are highly comparable with other reports investigating diabody and F(ab′)₂ molecules. Investigations with the anti-CEA diabody reported a T_1/2α of 0.25 ± 0.02 h and T_1/2β of 2.89 ± 0.74 h (18), whereas the anti-HER2/neu ¹²⁵I-labeled C6.5 diabody displayed a T_1/2α of 0.67 h and a T_1/2β of 6.42 h (30). The pharmacokinetic profile of the anti-CEA diabody very closely resembles the profile observed for the ¹¹¹In diabody (T_1/2α, 0.45 h; T_1/2β, 2.7 h). Investigations with an ¹³¹I-labeled anti-CEA F(ab′)₂ observed a T_1/2 of 12 h (46), which compares favorably with the pharmacokinetic results of the ¹¹¹In-labeled hu3S193 F(ab′)₂ (T_1/2α, 1.76 h; T_1/2β, 12.6 h). The pharmacokinetic profiles observed are typical of antibody-based molecules with sizes above the renal threshold for first-pass clearance.

The tissue biodistribution and gamma camera images for both conjugates showed prominent kidney uptake because of the clearance of the constructs and greater retention of the ¹¹¹In metabolite in the renal system (Figs. 5,6,7,8.). Localization of radiolabeled conjugates to other normal tissues investigated was nonspecific and attributable to blood pool activity. The uptake of small molecular weight constructs in the kidney presents an obstacle for therapy studies because it may result in potentially greater renal toxicity (47). Therefore, experiments aimed at blocking kidney uptake by the pretreatment or coadministration of cationic amino acids, such as d-lysine with radiolabeled antibody conjugates, which have been shown to reduce kidney uptake, need to be assessed (48, 49). Further improvements in pharmacokinetics can be achieved by increasing the surface negative charge or modifying the radionuclide chemistry, both strategies for reducing the renal uptake observed frequently in small engineered antibodies (49, 50).

Dimer molecules similar to the hu3S193 diabody are also useful reagents for delivering radioisotope to tumors and, similar to diabodies, are potentially most useful for imaging and immunodiagnosis because of their rapid clearance from the circulation (51, 52). However, theoretically, diabodies (M_r 60,000) will be more stable than scFv dimers, and because diabodies also provide more rapid tumor penetration and clearance than F(ab)′₂ (M_r 100,000), they may therefore be the preferred reagent for radioimaging with short half-life radionuclides such as ^99mTc and ¹²³I (gamma emitters) or ¹⁸F (positron emitter; Ref. 38). Conversely, for longer half-life radiolabels designed for immunotherapy (e.g., ⁹⁰Y and ¹³¹I), the slower blood clearance of F(ab)′₂ compared with diabodies generates a higher tumor:blood ratio, higher total tumor uptake, and higher radiotherapeutic index. Following this trend, we would expect that larger multivalent radiolabeled triabodies (M_r 90,000) and tetrabodies (M_r 120,000) would outperform diabodies in total tumor uptake because of higher avidity and slower blood clearance (12, 48), which makes these constructs attractive for further development and evaluation. The hu3S193 F(ab′)₂ showed a high maximal tumor uptake and is an interesting molecule to pursue in therapeutic studies, possibly in combination with other antibody constructs, targeting different tumor compartments, including the stroma and vasculature. The optimal construct size is therefore one that balances the rapidity of blood clearance, the absolute tumor uptake, and the tumor:blood ratios for the desired diagnostic or therapeutic application.

We thank Angela J. Mountain, Angela Rigopoulos, and Cathrine M. Hall for technical advice and their support with the FPLC analysis, cell binding assays, and animal handling, respectively.