|Home | About | Journals | Submit | Contact Us | Français|
Edited by Peter Hudson
A series of anti-tumor/anti-chelate bispecific antibody formats were developed for pre-targeted radioimmunotherapy. Based on the anti-carcinoembryonic antigen humanized hT84.66-M5A monoclonal antibody and the anti-DOTA C8.2.5 scFv antibody fragment, this cognate series of bispecific antibodies were radioiodinated to determine their tumor targeting, biodistribution and pharmacokinetic properties in a mouse xenograft tumor model. The in vivo biodistribution studies showed that all the bispecific antibodies exhibited specific high tumor uptake but the tumor targeting was approximately one-half of the parental anti-CEA mAb due to faster blood clearance. Serum stability and FcRn studies showed no apparent reason for the faster blood clearance. A dual radiolabel biodistribution study revealed that the 111In-DOTA bispecific antibody had increased liver and spleen uptake, not seen for the 125I-version due to metabolism and release of the radioiodine from the cells. These data suggest increased clearance of the antibody fusion formats by the mononuclear phagocyte system. Importantly, a pre-targeted study showed specific tumor uptake of 177Lu-DOTA and a tumor : blood ratio of 199 : 1. This pre-targeted radiotherapeutic and substantial reduction in the radioactive exposure to the bone marrow should enhance the therapeutic potential of RIT.
To enhance the therapeutic efficacy of radioimmunotherapy (RIT), pre-targeting or two-step strategies to uncouple the antibody's (Ab) slow tumor targeting and clearance properties from the delivery of a therapeutic radionuclide have been developed over the past several decades and have been well reviewed (Goodwin and Meares, 2001, Goldenberg et al., 2006, Sharkey et al., 2012). Separation of these two steps allows for the rapid delivery of a radiotherapeutic dose by a low-molecular-weight hapten, reducing the radiation exposure to normal tissues, particularly the bone marrow. While pre-targeted RIT has shown promise, the clinical efficacy has been limited (Chatal et al., 2009). Major challenges that have blocked clinical utility include immunogenicity of the biological agents, inability to achieve radiotherapeutic levels for solid tumors, toxicity to healthy normal tissue, as well as production of the multi-step reagents.
In this report, we designed a series of anti-carcinoembryonic antigen (CEA) and anti-1, 4, 7, 10-tetraazacyclododecane-1, 4, 7, 10-tetraacetic acid (DOTA) bispecific antibody (BsAb) formats for pre-targeting. To determine the best format, these BsAbs variants were evaluated for the first step of pre-targeting, targeting the tumor.
The anti-CEA humanized hT84.66-M5A (M5A) monoclonal antibody (mAb) has demonstrated high tumor accumulation as a direct radiolabeled agent in preclinical studies (Yazaki et al., 2004) and has recently entered Phase I/II RIT clinical trials. For capturing the radiometal-labeled macrocyclic DOTA chelate, the affinity matured C8.2.5 scFv was used that incorporates an engineered internal disulfide for stability and was selected by directed evolution for 10 and 100 pM affinity to yttrium (Y)-benzyl (Bn)-DOTA and Y-DOTA, respectively (Orcutt et al., 2011). The overall BsAb design employed a modular approach, using the M5A mAb as a scaffold and ligating on the C8.2.5 scFv cDNA at various termini to make tetravalent BsAbs. This BsAb series consisted of the carboxy (C)-terminal (Coloma and Morrison, 1997), amino (N)-terminal, light chain (LC) and dual variable domain immunoglobulin (DVD) (Orcutt et al., 2010). The BsAbs and parental M5A mAb were expressed, radiolabeled and evaluated in an animal xenograft model to assess their tumor targeting, normal tissue biodistribution, and pharmacokinetic properties for pre-targeted RIT.
The anti-CEA M5A mAb served as a human IgG1 scaffold for fusion of the anti-DOTA VH-VL C8.2.5 scFv Ab. The various BsAb cDNA constructs were formed by splice overlap PCR (Horton et al., 1989) with PCR primers that incorporate flexible amino acid linkers. The C-terminal (Coloma and Morrison, 1997) and N-terminal BsAbs fuse the C8.2.5 scFv by a glycine serine (G4S)(1-2) amino acid linker to the 5′ or 3′ end of the M5A IgG1 heavy chain cDNA, respectively. The LC fusion has the C8.2.5 scFv attached to the 3′ end of the human kappa LC cDNA (Orcutt et al., 2010). For the DVD, the cDNA encoding the C8.2.5 scFv cDNA was reformatted into a full-length IgG1. The M5A variable domains were then fused to the 5′ end of the C8.2.5 variable domains by a VKappa linker (Wu et al., 2009). The M5A-C825 BsAb heavy and LC genes were cloned into pEE12/6 dual-gene vector as part of the glutamine synthetase mammalian expression/selection system (Lonza Biologics, Slough, UK).
Transient transfection of the plasmids encoding the BsAbs was conducted using linear 25 kDa PEI (Polysciences, Warrington, PA, USA) into Freestyle™ 293-F cell line following the manufacturer's protocol (Invitrogen, Carlsbad, CA, USA). Five- to seven–day-old culture supernatants were screened for anti-CEA activity by enzyme-linked immunosorbent assay and protein quantified by Protein A as previously described (Yazaki and Wu, 2004).
Purification of the BsAbs used a two-step procedure consisting of Protein A capture and ceramic hydroxyapatite chromatography. Briefly, the culture supernatant was clarified in batch with 5% v/v AG1-X8 resin (Bio-Rad Laboratories, Hercules, CA, USA), sterile filtered and loaded on a Prosep rA high-capacity column (4.6 mm × 100 mm, 1 ml/min; Millipore, Billerica, MA, USA). The column was washed in 0.01 M sodium phosphate, 1 M NaCl, 0.05% Tween-80, pH 7 and eluted with 0.05 M NaCl, 0.01 M sodium phosphate, 0.05% Tween-80, pH 3. Eluted fractions were collected in tubes containing 10% v/v of 50 mM sodium phosphate, pH 8 for neutralization. Pooled fractions were pH adjusted with 25% v/v of 0.1 M MES pH 6.5 buffer and loaded on a ceramic hydroxyapatite CHT® type I column (4.6 mm × 100 mm, 1 ml/min; Bio-Rad Laboratories) pre-equilibrated in 0.05 M MES, 0.01 M potassium phosphate, pH 6.5. A linear gradient from 0.01 to 0.5 M potassium phosphate pH 6.5 eluted the fusion protein in a single peak and the purified material was dialyzed vs. phosphate-buffered saline prior to concentration (Ultracel—30 k, Millipore).
Aliquots of the purified proteins were analyzed for protein quantification by UV absorbance 280 nm. Purity was determined by sodium dodecyl sulfate polyacrylamide gel electrophoresis under non-reducing conditions and high-performance liquid chromatography (HPLC) size-exclusion chromatography (SEC) on a Superdex 200 10/300 column (GE Healthcare, Piscataway, NJ, USA) as previously described (Yazaki and Wu, 2004).
Anti-CEA and anti-DOTA binding affinities were measured by surface plasmon resonance (SPR) using a Biacore X100 instrument (GE Healthcare). For the anti-CEA measurements, biotinylated CEA (~500 RU) was immobilized on a Sensor Chip SA, following the manufacturer's instructions. HBS-EP (0.01 M HEPES, 0.15 M NaCl, 3 mM EDTA and 0.005% surfactant P20) was used as running and diluent buffer (GE Healthcare). Using the kinetic affinity program, serial concentrations (15.6–500 nM) of the BsAbs and parental M5A mAb were injected at a flow rate of 30 µl/min at 25°C. The data were analyzed using the bivalent analyte model on the Biacore X100 BIAevaluation 2.0 software (GE Healthcare). The kd1/ka1 data were used to calculate KD1, to compare the relative binding affinity of the Ab variants. To measure anti-DOTA affinity, p-SCN-Bn-DOTA (Macrocyclics, Dallas, TX, USA) was conjugated to human serum albumin (HSA) at a ratio of 50 : 1, purified by SEC Superdex 75 (GE Healthcare) and loaded with non-radioactive 89YCl metal. The 89Y-DOTA-Bn-HSA (~1000 RU) was conjugated to a CM5 sensor chip following the manufacturer's instructions.
Binding to FcRn was also measured by SPR using a Biacore 3000 as previously described (Andersen et al., 2008). Flow cells of CM5 sensor chips were coupled with recombinant forms of soluble human or mouse FcRn (shFcRn or smFcRn; ~500 RU) using amine-coupling chemistry. The coupling was performed by injecting 1 μg/ml of the protein in 10 mM sodium acetate pH 5.0. Phosphate buffer (67 mM phosphate buffer, 0.15 M NaCl, 0.005% Tween 20, at pH 6.0), or HBS-EP buffer was used as running and dilution buffer. Titrated amounts of the BsAbs and parental M5A mAb (3–100 nM) were injected over immobilized receptor at pH 6.0 with a flow rate of 50 μl/min at 25°C. Regeneration of the surfaces was done using injections of HBS-EP buffer at pH 7.4 In all experiments, data were zero adjusted and the reference cell subtracted. Data were evaluated using the heterogeneous ligand binding model supplied with the BIAevaluation 4.1 software and reported as KD1 and KD2.
The BsAbs were radiolabeled with Iodine-125 (125I), using the Iodogen methodology as previously described (Yazaki et al., 2008). All radiolabeled Abs were purified by HPLC SEC on Superdex 200, 10/300 (GE Healthcare).
For dual radiolabeling, the LC BsAb and M5A mAb were conjugated with NHS-DOTA (Macrocyclics) at a ratio of 50 : 1. Each DOTA-mAb was radiolabeled separately with 111InCl (specific activity = 5.4 µCi/µg) or Na125I (specific activity = 10.1 µCi/µg) and purified by HPLC SEC. The two radiolabeled versions were mixed, 4 µCi 111In-LC-BsAb to 10 µCi 125I-LC-BsAb, and co-injected into athymic mice bearing tumor xenografts (see below). Tissues were counted in a calibrated dual-isotope window.
All radiolabeled antibodies were analyzed for immunoreactivity to soluble CEA by a liquid phase assay incubating the radiolabeled protein with 20 equivalents by mass of purified CEA at 37°C for 15 min. The resultant solution was analyzed by HPLC-SEC using a Superose 6 10/300 GL column (GE Healthcare). Anti-CEA immunoreactivity was determined by integrating the area on the HPLC radiochromatogram and calculating the percentage of radioactivity shifting to higher molecular weights, consistent with binding to CEA (180 kDa).
Protein stability studies were performed on the 125I-LC BsAb incubated in either saline or fresh mouse serum at 37°C. Aliquots were analyzed on an HPLC SEC Superose 6 column at the time points of 0, 4, 24 and 48 h.
All animal handling was done in accordance with protocols approved by the City of Hope Institutional Animal Care and Use Committee. Seven- to eight-week-old female athymic mice (Charles River Laboratories, Wilmington, MA, USA) were injected subcutaneously in the flank region with 106 LS-174T human colon carcinoma cells obtained from ATCC. After 10 days, when tumor masses were in the range of 0.1–0.5 g, 1–2 µg (specific activity ~10 µCi/µg) of the purified 125I-M5A-C8.2.5 BsAbs were injected via the tail vein per animal. At selected time points (0, 4, 24, 48 and 72 h), groups of four to five mice were euthanized, necropsy performed, organs weighed and counted for radioactivity. All data presented are mean values with the standard error (±SEM) and have been corrected for background and radioactive decay from the time of injection, allowing organ uptake to be reported as percentage of the injected dose per gram (% ID/g).
The overall BsAb design was to construct a series of tetravalent BsAbs in a modular manner, using the anti-CEA humanized IgG1 mAb as a scaffold to retain affinity, solubility and expression levels. This series of M5A-C8.2.5 BsAbs is illustrated in Fig. 1. The BsAb series consists of the C-terminal, N-terminal, LC and DVD constructs based on formats previously reported in the literature (Coloma and Morrison, 1997; Wu et al., 2007; Orcutt et al., 2010; Kontermann, 2012).
The C-terminal and N-terminal formats were constructed by ligation of the C8.2.5 scFv cDNA to either termini of the cDNA encoding the M5A IgG1 heavy chain by a G4S linker. Similarly, the LC construct fused the C8.2.5 scFv to the C-terminus of the M5A kappa LC. The DVD BsAb was designed with each immunoglobulin light and heavy chain having two variable domains linked in tandem. For the anti-CEA/anti-DOTA DVD, the M5A variable domains were designed external to the C8.2.5 variable domains, to avoid potential steric hindrance in binding the larger CEA molecule compared with the small DOTA chelate (180 vs. 0.5 kDa, respectively).
The individual BsAb light and heavy chain cDNA transcripts were inserted into a single dual-gene vector and produced a transient mammalian expression system. A two-column purification resulted in highly purified BsAbs; the final yields ranging from 0.5–1.5 mg/L. The purified BsAbs displayed a single peak of the expected molecular size (210 kDa) by HPLC size exclusion chromatography, confirming proper Ab assembly (data not shown). The anti-CEA and anti-DOTA kinetic affinities were measured by SPR (Table I). The BsAbs anti-CEA affinities were sub-nanomolar (95–312 pM), comparable with the parental M5A mAb (219 pM). For the anti-DOTA binding, the BsAbs exhibited a broader range of affinity (13–213 nM) much lower than the 100 pM reported for the C8.2.5 scFv alone (Orcutt et al., 2011). Presumably, this was due to steric interference caused by the conjugation of HSA with DOTA-Bn to enable immobilization on the sensor chip.
To evaluate their tumor targeting potential, the BsAbs and the parental M5A mAb were directly radiolabeled with 125I, purified, and all BsAbs demonstrated a high level of immunoreactivity to soluble CEA, ranging from 89 to 100% (Supplementary Table SI). In a series of individual experiments, tumor targeting and normal organ biodistribution studies were conducted in athymic mice bearing ~10-day-old human colorectal cancer LS-174T tumors. The 125I-BsAbs and parental M5A mAb biodistribution results are shown in Fig. 2, with the actual values and tumor : blood ratios in Supplementary Table S2. All the BsAb formats showed specific high tumor uptake ranging from 12–18% ID/g at the peak uptake time of 24 h accompanied by low normal organ uptake. The most striking observation was that the BsAbs as a group had much lower tumor uptake compared with the 42% ID/g for the M5A mAb. The ranking order of highest tumor uptake was M5A mAb, DVD, LC, C-terminal and N-terminal BsAb. A comparison of the individual blood clearance curves was made and showed a direct reciprocal relationship to their tumor uptake (Fig. 3).
To determine if protein instability was the cause for faster blood clearance, in vitro stability studies were performed. The 125I-LC BsAb was incubated in saline or fresh mouse serum at 37°C. The radiochromatograms showed no significant product breakdown or formation of large-molecular-weight species over the 48 h time period monitored, although a shoulder of apparently larger-molecular-weight species appears in the SEC chromatogram immediately upon incubation with serum (Fig. 4).
An alternative hypothesis to explain the faster blood clearance was the tethered scFv interferes with binding to the neonatal Fc receptor (FcRn) which is responsible for the serum half-life of IgG (Roopenian and Akilesh, 2007). The C-terminal, DVD and LC BsAbs and M5A mAb were analyzed for FcRn binding by SPR. Although there was slightly weaker binding observed for the BsAbs binding to mouse FcRn compared with M5A by SPR, the results showed strong and reversible binding of all four Ab constructs to soluble human and mouse FcRn preparations (Table II and Supplementary Fig. S1). The binding responses of both M5A mAb and BsAbs for mouse FcRn were much stronger than that seen for binding to the human form of FcRn, which is in agreement with previous reports (Ober et al., 2001; Andersen et al., 2011). All Ab formats bound mouse and human FcRn in a pH-dependent manner as expected (data not shown).
A dual radiolabel 125I- and 111In-BsAb biodistribution study was conducted for two reasons: (i) to determine if the radioiodination process caused the altered BsAbs blood clearance and (ii) radiometals are not rapidly released from the cell after internalization, enabling retention of the radiotracer where mAb uptake and internalization has occurred in vivo. A version of the anti-CEA LC BsAb (Orcutt et al., 2010) and the M5A mAb were conjugated with the DOTA chelate. Aliquots of the DOTA-LC-BsAb were separately radiolabeled with 125I or 111In, HPLC SEC purified, and the 125I-DOTA BsAb and 111In-DOTA BsAb co-injected into tumor bearing mice with the tissues being counted using dual-isotope windows. This was repeated for the control 125I- and 111In- DOTA-M5A mAb.
The dual-label 125I- and 111In-DOTA LC BsAb biodistribution exhibited similar blood clearance indicating no difference between the two radiolabeling processes. As previously observed, the BsAb rate of blood clearance was faster than the 125I- and 111In-DOTA-M5A mAb (Fig. 5 and Supplementary Table SIII). The most striking feature was the 111In-DOTA LC BsAb exhibited high liver (29% ID/g) and spleen (14.1% ID/g) accumulation over the 72 h time course, not unexpected due to the faster clearance. In comparison, the 111In-DOTA-M5A mAb had peak accumulation of 8.9% ID/g for liver and 7.7% ID/g for spleen at the 72 h time point. Even with this increased normal organ uptake, the 111In-DOTA LC BsAb displayed peak tumor uptake of 39% ID/g at 48 h. While this level of tumor accumulation was significant, it was approximately one half of that observed for the 111In-DOTA-M5A mAb (71% ID/g at 72 h).
Given the BsAb pharmacokinetics, we moved forward to determine the ability of the pre-targeted LC BsAb to bind 177Lu-DOTA in vivo. In the same animal model, the unlabeled LC BsAb was allowed to pre-target for 24 h, eliminated from the vasculature using 89Y-DOTA-Bz-Dextran as the clearing agent and 177Lu-DOTA administered. One day later, a tumor targeting and biodistribution study of the 177Lu-DOTA was conducted with no BsAb administered serving as a control. The results showed specific tumor uptake of the 177Lu-DOTA reaching 6% ID/g (Fig. 6). There was marginal uptake in the normal tissues, resulting in a tumor:blood ratio of 199 : 1 and tumor : liver ratio of 30 : 1 at 24 h. The no BsAb control (177Lu-DOTA alone) showed only residual kidney uptake corresponding to the renal clearance of the small-molecular-weight hapten.
Pre-targeted RIT is an active area of development to provide an effective radiotherapeutic while reducing RIT's toxicity to the bone marrow eliminating it as the dose-limiting organ. The design and use of BsAbs have been key components for in vivo formation of the Ab-hapten complexes on the tumor surface. However, the development of this imaging and therapy platform has been complicated requiring empirical optimization of multiple steps. Many pre-targeting systems have been evaluated over the past decade; however, typically these studies only measured the second step, accumulation of the radiolabeled effector molecule at the tumor site. We believe this is the first detailed report evaluating a series of BsAb formats for the first step of pre-targeting, targeting the tumor targeting. While the four 125I-BsAbs exhibited specific high tumor uptake, surprisingly the level was only half that observed for the parental anti-CEA hT84.66-M5A mAb which is primarily due to their faster blood clearance. A recent report of a bivalent tri-Fab for pre-targeting observed 3–5% peak tumor uptake and almost complete blood clearance at 24 in murine tumor models, but presumably this was due to the lack of an Fc domain (Frampas et al., 2011).
A series of experiments were conducted to investigate the reason for the faster blood clearance of the BsAbs compared with the parental mAb. In vitro stability studies showed no strong evidence for significant Ab aggregation or proteolysis. An alternative hypothesis was that the binding kinetics for FcRn was affected, but no major differences were detected by SPR.
A dual-radioiodine/radiometal biodistribution study provided insight toward the BsAb's faster blood clearance. Radioiodinated Ab biodistribution studies can be misleading as the radioiodine is released from the protein due to metabolism and rapid excretion once the Ab is internalized. However, this radiotracer is optimal to follow the first step of pre-targeting because if the Ab is internalized it will not be available for the second step using the DOTA effector molecule. The 111In-BsAb high liver and spleen uptake suggests increased clearance by the mononuclear phagocyte system (Lobo et al., 2004; Tabrizi et al., 2006). Given the generality of this result for all of the formats tested, it would seem that the C8.2.5 variable domain itself is to blame, rather than the topology of the bispecific fusion. Perhaps the Fv drives aggregation or interacts inappropriately with a plasma component, driving clearance. Studies are ongoing to replace the mouse scFv with a humanized version that may enhance tumor uptake.
Importantly, a pre-targeting study employing one of the BsAbs showed tumor uptake of 6% ID/g for 177Lu-DOTA after 24 h. This was comparable to a 177Lu-peptide (3–5% ID/g) employed in another anti-CEA pre-targeting system (Frampas et al., 2011), which is currently being evaluated in clinical trials in the USA and Europe. In spite of the lower tumor uptake, the pre-targeting approach offers improved RIT, due to the specific tumor uptake and low normal tissue accumulation. The pre-targeted vs.direct-label mAb's tumor-to-blood ratio (199 : 1 vs. 3 : 1, respectively) substantially reduces the radioactive exposure to the bone marrow, and should enhance the therapeutic potential of pre-targeted RIT.
This work was supported by the National Cancer Institute at the National Institute of Health (grant no. P01 CA43904). J.T.A. was supported by the Norwegian Research Council (grant no. 179573/V40) and South-Eastern Norway Regional Health Authority (grant no. 39375).