A Tetravalent Single-chain Antibody-Streptavidin Fusion Protein for Pretargeted Lymphoma Therapy¹ Free

mAbs³directed against the CD20 antigen are finding wide application as a means to target radiation to B-cell tumors of NHL. The CD20 antigen is a nonglycosylated phosphoprotein present on the surface of normal peripheral B-cell lymphocytes, malignant cells of pre-B-cell acute lymphoblastic leukemia, malignant cells of B-type chronic leukemia, and B-cell lymphomas. Compared with carcinomas, lymphomas and leukemias generally provide better access to target antigen, and the circulating disease cells can experience direct cytotoxic effects via the binding of whole antibodies. These qualities, along with the inherent radiosensitivity of hematological malignancies, have made RIT of lymphomas an application with established clinical efficacy(1, 2, 3). The clinical studies of Press et al.(4) and Liu et al. (5) have documented both the steep dose-response curve for radiation in NHL and the benefit of increased radiation dose in prolonging disease-free survival. Pretarget® RIT offers the potential to achieve the benefits of high-dose radiation without the need for stem cell rescue. This is accomplished by injecting the antibody and radiation source separately,thereby decoupling the pharmacokinetics of the antibody from the radionuclide (6).

Pretargeted RIT is a multistep process that exploits the high-affinity interaction between streptavidin and biotin (K_a≅ 10⁻¹⁵ m). In the first step of the process, an antibody linked to streptavidin is administered systemically. After sufficient time has elapsed to enable peak tumor uptake, the excess antibody/streptavidin is removed from the circulation by in vivo complexation with a biotinylated poly(GalNAc) “clearing” agent. The resultant complexes are excreted via the liver. The clearing step is essential to achieve the highest absolute concentration of antibody receptor at tumor sites but still maintain low absolute blood and whole body concentrations. Radiation is delivered to the tumor in a third step by administration of radiolabeled DOTA-biotin. This low molecular weight cytotoxic molecule readily penetrates into the tumor, where it is captured by prelocalized antibody-streptavidin conjugate. Unbound radioactivity is eliminated from the body via the urine. Rapid uptake of the therapeutic isotope at tumor sites (before significant loss of potency attributable to radioactive decay) and efficient renal elimination of excess radioactivity represent the fundamental advantages of the pretargeted approach as compared with conventional RIT, in which the radioisotope is directly conjugated to the antibody. This approach has been tested preclinically in mice, producing cures of human small cell lung, colon,and breast cancer xenografts (6), and clinically in patients with adenocarcinoma (7, 8, 9) and NHL(10).

In contrast to antibody-streptavidin chemical conjugates, a genetically fused targeting agent has a well-defined, homogeneous composition and is simpler and less expensive to manufacture. Here, we report the genetic engineering and in vitro and in vivocharacterization of Escherichia coli-produced B9E9 scFvSA fusion protein for use as a first component in pretargeted RIT.

Total RNA (SNAP kit; Invitrogen) was prepared from B9E9 hybridoma cells expressing murine IgG2a anti-CD20 mAb (Bioprobe BV, Amstelveen,the Netherlands). The cDNAs for the V_L (κ) and V_H of B9E9 were obtained by a reverse transcription reaction using oligonucleotides for the antibody constant regions (5′-TAGCTGGCGGCCGCCCTGTTGAAGCTCTTGACAAT for V_L and 5′-TAGCTGGCGGCCGCTTTCTTGTCCACCTTGGTGC for V_H). The variable regions were PCR-amplified from the cDNA using the constant region primers and degenerate variable region primers (5′-TGCCGTGAATTCCATTSWGCTGACCARTCTC for V_L and 5′-TGCCGTGAATTCGTSMARCTGCAGSARTCWGG for V_H). The DNA sequences of the variable regions have been deposited in the GenBank database (accession numbers AF277091 for V_H and AF277092 for V_L). The PCR fragments were cloned into an EcoRI/NotI-digested vector and sequenced using a BigDye kit (PE Applied Biosystems). The NH₂-terminal amino acids of the variable regions were reconstituted based on the consensus sequence (11),including introduction of a serine residue at position 5 in the V_L gene. Primers used were 5′-TTACGGCCATGGCTGACATCGTGCTGTCGCAGTCTCCAGCAATCCTGTCT and 5′-CACCAGAGATCTTCAGCTCCAGCTTGGTCCCA for V_L and 5′-CGGAGGCTCGAGCCAGGTTCAGCTGGTCCAGTCAGGGGCTGAGCTGGTGAAG and 5′-GAGCCAGAGCTCACGGTGACCGTGGTCCCTGCGCCCCA for V_H.

The variable regions were ordered in the V_H-V_L or V_L-V_H configuration and separated by linkers of various lengths or compositions (Table 1). The streptavidin coding region, leader sequence, and approximately 310 bp of the upstream region were PCR-amplified from Streptomyces avidinii (ATCC 27419; Ref. 12). The streptavidin gene was separated from the scFv gene by a linker encoding amino acids GSGSA. Plasmids contained a lacpromoter, a pBR origin of replication, and a gene encoding eitherβ-lactamase (bla; ampicillin resistant) or aminoglycoside 3′-phosphotransferase (neo from Tn5; kanamycin resistant).

Transformants of E. coli strain XL1Blue (Stratagene, La Jolla, CA) were grown overnight at 30°C in Terrific broth (20 ml;Sigma) containing carbenicillin (50 μg/ml). The culture was diluted 100-fold into fresh medium and grown in a shaking incubator at 30°C. When the culture attained an A_{600 nm}of 0.3–0.5, IPTG (Amersham Pharmacia Biotech, Piscataway, NJ) was added to a final concentration of 0.2 mm, and incubation was continued overnight. Periplasmic extracts were prepared for analysis of the scFvSA expression level. Cells were resuspended in an ice-cold solution of 20% sucrose, 2 mm EDTA,30 mm Tris (pH 8.0), and 2.0 mg/ml lysozyme and incubated on ice for 30 min. Supernatants were analyzed on 4–20%Tris-glycine SDS-PAGE gels (Novex) under nonreducing, nonboiled conditions, and gels were stained with Coomassie Blue.

For fermentation cultures, the primary inoculum (50 ml) was grown overnight at 30°C in a shake flask containing Terrific broth plus kanamycin or carbenicillin (50 μg/ml), depending on the selectable marker of the plasmid. The culture was then diluted 100-fold into the same medium and grown at 30°C for an additional 4–5 h. This secondary inoculum (0.5 liter) was transferred to a 14 liter BioFlo 3000 fermentor (New Brunswick Scientific) containing 8 liters of complete E. coli medium [per liter: 6 grams of Na₂HPO₄, 3 grams of KH₂PO₄, 0.5 gram of NaCl, 3 grams of(NH₄)₂SO₄,48 grams of yeast extract (Difco), 0.25 ml of Mazu DF204 antifoam(PPG Industries Inc., Pittsburgh, PA), 0.79 gram of MgSO₄-7H₂O, 0.044 gram of CaCl₂-2H₂O, and 3 ml of trace elements (per liter: 0.23 gram of CoCl₂,0.57 gram of H₃BO₃, 0.2 gram of CuCl₂-2H₂O, 3.5 grams of FeCl₃-6H₂O, 4.0 grams of MnCl₂-4H₂O, 0.5 gram of ZnCl₂, 1.35 grams of thiamine, and 0.5 gram of Na₂MoO₄-2H₂O)]. The medium contained galactose at an initial concentration of 5 grams/liter as a carbon source plus 50 μg/ml kanamycin or carbenicillin for plasmid retention. The culture was grown at 30°C and induced with IPTG (0.2 mm) at 6 h postinoculation. The pH was maintained at 7.0 by the automatic addition of either phosphoric acid or NaOH. Dissolved oxygen concentration was maintained at ≥30% throughout the run, using agitation speeds of 400–800 rpm and oxygen supplementation as necessary. A galactose solution (50%) was fed to a total of 20–25 grams/liter over a 9-h period after exhaustion of the initial galactose present in the medium. Cells were harvested at 24–26 h postinoculation in a continuous flow centrifuge (Pilot Powerfuge; Carr Separations, Franklin, MA), washed with PBS [10 mm sodium phosphate and 150 mm NaCl (pH 7.2)], and pelleted by centrifugation. A typical fermentation produced 80–90 grams of cells(wet weight) per liter of culture medium.

A rhodamine-biotin assay was used for quantifying the fusion protein in fermentor-grown cells. Cells were washed twice in PBS, resuspended in ice-cold 30 mm Tris and 1 mm EDTA (pH 8.0) to 20% (w/v), and disrupted through two cycles of microfluidization(Microfluidics International, Newton, MA). The fusion protein in centrifuged lysates was complexed with excess rhodamine-derivatized biotin, which was prepared as follows: 5 (and 6-)-carboxytetramethylrhodamine succinimidyl ester (Molecular Probes,Eugene OR) was coupled to biocytin (Pierce, Rockford IL) through the formation of a stable amide bond. The reaction mixture was purified by HPLC using a Dynamax semipreparative C-18 column (Rainin Instrument Co., Woburn, MA). The effluent was monitored at 547 nm, and peak fractions were collected and analyzed by mass spectrometry. Fractions corresponding in molecular weight to biocytin-rhodamine conjugate were pooled and concentrated by roto-evaporation (Büchi,Flawil, Switzerland). An excess of purified biocytin-rhodamine conjugate was added to the supernatant of a centrifuged sample of crude lysate and analyzed by size exclusion chromatography using a Zorbax GF-250 column (MAC-MOD, Chadds Ford, PA) equilibrated in 20 mm sodium phosphate containing 15% DMSO at a 1.0 ml/min flow rate. The effluent was monitored at 547 nm using a Varian Dynamax PDA-2 detector, and the peak area corresponding to fusion protein elution was determined using a Varian Dynamax HPLC Data System (Walnut Creek, CA). The concentration of fusion protein in the crude lysate was calculated by comparison with a standard analyzed under the same conditions. The molar extinction coefficient for the fusion protein standard was calculated using a previously described method summing the relative contributions of amino acids absorbing at 280 nm(13).

The iminobiotin affinity matrix used for the isolation of fusion protein was prepared by reacting epoxide-activated Macro-prep matrix(Bio-Rad, Hercules, CA) with 112 μmol of N-(3-amino-propyl)-1,3 propane diamine (Sigma) per gram of matrix in 0.2 m carbonate buffer. The reaction was stopped after 8 h by filtering the slurry through a sintered glass funnel and rinsing the matrix with distilled water. Residual epoxides were inactivated by reacting the matrix with 0.1 m sulfuric acid for 4 h at 80°C, and the matrix was rinsed again. The amine-derivatized matrix was suspended in PBS, and the pH was increased to 8.5 by the addition of a 10% volume of 0.5 m sodium borate (pH 8.5). N-hydroxy-succinimide-iminobiotin (Pierce) was dissolved in DMSO and added to the suspended matrix at a ratio of 2.6 mg/gram of matrix. After a 4-h reaction, the matrix was rinsed with distilled water, followed by several alternating washes with sodium carbonate buffer (pH 11) and sodium acetate buffer (pH 4) and a final rinse with distilled water. The matrix was stored as a slurry in 20%ethanol.

Cells (650–750 grams, wet weight) from the fermentation culture were washed and lysed at a concentration of 20% (w/v) by microfluidization as described above. A 5% solution of polyethylenimine(M_r 1,200; Aldrich, Milwaukee,WI), adjusted to pH 8.0 with concentrated HCl, was added to the lysate(0.3 ml/g cells) and precipitated overnight at 4°C. The lysate solution was centrifuged (10,000 × g for 90 min) to pellet the cellular debris. The supernatant was decanted and filtered through a 0.2 μm filter. Glycine was added to the clarified lysate solution to a final concentration of 50 mm, and NaCl was added to a final concentration of 0.3 m. The lysate solution was then applied over an iminobiotin affinity column equilibrated in 50 mm glycine (pH 9.2) containing 0.5 m NaCl. The column was washed with 20 column volumes of equilibration buffer and then eluted with 0.2 m sodium acetate buffer (pH 5.0) containing 0.1 m NaCl. The pH of the eluted product was neutralized with 0.5 m Tris buffer (pH 8.0), and then the product was exhaustively dialyzed in PBS at 4°C.

To reduce protein aggregation, the iminobiotin-purified scFvSA fusion protein was treated with 30% DMSO in PBS for 5–7 h at room temperature and dialyzed in PBS at 4°C. Preparations were concentrated to 2–3 mg/ml using a YM30 membrane (Millipore) and filter sterilized for aseptic storage at 4°C.

Purified fusion proteins were analyzed on 4–20% Tris-glycine SDS-PAGE gels under nonreducing conditions. Before electrophoresis, samples were mixed with SDS loading buffer and incubated at either room temperature or 100°C for 5 min. Gels were stained with Coomassie Blue, or the protein bands were transferred to polyvinylidene difluoride membranes(Novex). Immunoblot analysis was performed essentially as described previously (14). Peroxidase-conjugated goat anti-streptavidin polyclonal antibody (Zymed, San Francisco, CA) and TMB substrate (Vector Laboratories, Burlingame, CA) were used for detection.

Size exclusion HPLC (Beckman) was performed on a Zorbax GF-250 column with a 20 mm sodium phosphate/0.5 m NaCl mobile phase and a Varian Dynamax detector set at A_{280 nm}. This system, connected in series with a Varian Star 9040 refractive index detector and a MiniDawn light-scattering instrument (Wyatt Technologies, Santa Barbara, CA),was also used to determine the molecular weight of the intact tetramer and aggregate. A dn/dc value of 0.185 for a protein in an aqueous buffer solution was used in the calculations (15). Automated amino acid sequencing was performed using a Procise 494 sequenator (Applied Biosystems, Inc., Foster City, CA).

For molecular weight determination, liquid chromatographic separation was conducted with a Hewlett Packard series 1100 system fitted with a Jupiter C-18 column (300 Å; 3.2 × 50 mm; 5 μm) and a C-18 “SafeGuard” column (Phenomenex, Torrance, CA) at a flow rate of 500 μl/min. The mobile phase was composed of water/1% formic acid(buffer A) and acetonitrile/1% formic acid (buffer B). The gradient applied was 2% buffer B for 3 min, rising to 99% buffer B within 7 min. B9E9 scFvSA was eluted at a retention time of 8.7 min. The analytical column was interfaced with an electrospray ionization ion trap mass spectrometer (LSQ, Thermoquest, San Jose, CA). The instrument was calibrated with myoglobin and operated in the positive ion mode with the heated capillary set to 200°C and 5.1 kV applied to the electrospray needle. The data were acquired in a full scan mass spectrometry mode [m/z (500–2000 Da/z)]. The molecular weights and amino acid sequences of tryptic peptides generated from B9E9 scFvSA were determined by liquid chromatography-mass spectrometry/mass spectrometry as described by Covey et al. (16).

Relative immunoreactivity was assessed in a competitive binding assay using flow cytometry that measured the binding of fluorescein-labeled B9E9 mAb to the CD20-positive Ramos cell line (Burkitt lymphoma; ATCC CRL-1596) in the presence of various concentrations of unlabeled antibody or fusion. B9E9 mAb was labeled using fluorescein N-hydroxysuccinimidate, and an optimized amount of this conjugate was mixed with serial dilutions (3–200 μg/ml) of B9E9 mAb standard or molar equivalents of B9E9 scFvSA and incubated with 1 × 10⁶ cells at 4°C for 30 min. Samples were washed and then analyzed on a single laser FACSCalibur (Becton Dickinson). After gating on single cells, the geometric mean fluorescence intensity was determined from a histogram plot of fluorescence. The concentration of competitor antibody required for IC₅₀ of fluorescein-B9E9 binding was calculated using nonlinear regression analysis for one-site binding. The percentage of immunoreactivity was calculated according to the following formula: (IC₅₀ mAb/IC₅₀ scFvSA) × 100.

Immunoreactivity of radiolabeled antibodies was determined by a cell binding assay. The B9E9 scFvSA fusion protein was stably radioiodinated using [¹²⁵I]-N-succinimidyl para-iodobenzoyl ester (NEN Research Products) according to the method of Wilbur et al. (17). ¹²⁵I-labeled B9E9 mAb or fusion protein was incubated with a fixed concentration of radiolabeled protein (50 ng/ml)and varying amounts of antigen (0.25–10 × 10⁶ Ramos cells). Nonspecific binding was determined in the presence of excess cold mAb or fusion protein (50μg/ml). After a 2-h incubation at ambient temperature, bound and free antibodies were separated by centrifugation through oil (1:1, dinonyl phthalate:dibutyl phthalate) and counted on a gamma counter. Immunoreactivity was calculated from nonlinear regression analysis of a plot of bound (specific − nonspecific binding)/total (bound + free) versus cell number (18).

Avidity was assessed using saturation binding experiments that measure specific binding of radiolabeled mAb or fusion protein (0.025–50 ng/ml) at equilibrium in the presence of excess antigen(10⁷ cells). Nonspecific binding was determined in the presence of excess cold mAb or fusion protein (50 μg/ml). Mixtures were incubated and centrifuged as described above. The equilibrium dissociation constant (K_d)was calculated from nonlinear regression analysis of nanomolar bound versus nanomolar total radioligand using immunoreactivity-adjusted antibody concentrations (19). The calculated immunoreactivities for ¹²⁵I-labeled scFvSA and mAb were 79% and 67%,respectively.

The rate of biotin dissociation was determined at 37°C in 0.25 m sodium phosphate, 0.15 m NaCl, and 0.25% BSA(pH 7.0) containing 10 μm fusion protein or recombinant streptavidin (control), 0.06 μm[³H]biotin (58 mCi/μmol; NEN Research Products) and 30 mm ascorbate as[³H]biotin stabilizer. After incubation to reach equilibrium, biocytin (4 mm; Sigma) was added to initiate irreversible dissociation of[³H]biotin. Aliquots were withdrawn periodically and diluted 20-fold in PBS containing 0.5% BSA. The samples were split for assessment of total and free[³H]biotin, the latter determined after protein precipitation using sequential additions of ZnSO₄/NaOH (60 μm each). Radioactivity was assessed in a fluoroscintillate using a Hewlett Packard beta counter. Linear regression analysis of a plot of the natural logarithm (fraction bound) versus time yielded a dissociation rate constant.

Biotin binding capacity was determined by incubation of the fusion protein with a 9-fold molar excess of[³H]biotin. The amount of[³H]biotin associated with the fusion protein was determined by liquid scintillation after the removal of uncomplexed biotin using a PD-10 size exclusion column (Pharmacia).

Immunohistology was conducted by PhenoPath Laboratories (Seattle, WA). Fluoresceinated B9E9 scFvSA, C2B8/SA chemical conjugate, B9E9 mAb, or a murine IgG2a isotype-matched control (Sigma F6522) was reacted with cryosections of normal and tumor human tissues, which were peroxidase blocked with 0.3% H₂O₂ and biotin blocked with Biotin Blocking System (DAKO X0590). Binding was detected with horseradish peroxidase-conjugated rabbit anti-FITC mAb(DAKO P0404).

The B9E9 scFvSA fusion protein was labeled with ¹²⁵I as described above. Methods have been described previously for preparation of the synthetic clearing agent,designated as biotin-LC-NM-(GalNAc)₁₆(20), radiolabeled DOTA-biotin (6), and the anti-CD20 C2B8/SA chemical conjugate (10). The murine CC49 scFvSA (V_H-V_L 25-mer)fusion protein, which was immunoreactive to TAG-72 antigen and used as a negative control, was prepared in house.⁴

All animal studies were conducted under the supervision of the NeoRx Animal Care and Use Committee. Disappearance of the fusion protein from the blood was assessed by measuring the amount of radioactivity in serially collected blood samples after i.v. injection of ¹²⁵I-labeled scFvSA (600 μg) in normal female BALB/cBkl mice (B&K Universal, Kent, WA). At 18 h postinjection, some mice received a single i.v. injection of the synthetic clearing agent (100 μg). Blood samples were obtained by retro-orbital bleeding.

Pretargeted RIT studies were conducted in female nude mice bearing well-established Ramos xenografts (100–400 mm³). Tumor-bearing Bkl:BALB/c/nu/nu nude mice were obtained by implanting 5–25 × 10⁶ cultured cells s.c. in the flank 10–25 days before study initiation. Mice were injected i.v. at t = −24 h with ¹²⁵I-labeled B9E9 scFvSA (600 μg; 3.488 nmol; 5μCi) or with the positive control C2B8/SA chemical conjugate (400μg; 1.905 nmol; 5 μCi), followed 20 h later(t = −4) by 100 μg (11.558 nmol) of the synthetic clearing agent. ¹¹¹In-labeled DOTA-biotin (1.0 μg; 1.239 nmol; 10 μCi) was injected into each mouse 4 h after the clearing agent (t = 0).

In the ⁹⁰Y-DOTA-biotin studies, mice received i.v. injections of the unlabeled B9E9 scFvSA (600 μg; 3.488 nmol) or anti-TAG-72 murine CC49 scFvSA (600 μg; 3.488 nmol) and received an injection of 100 μg (11.558 nmol) of the synthetic clearing agent 20 h later. ⁹⁰Y-DOTA-biotin (1.0μg; 1.239 nmol; 100 μCi) was injected into each mouse 4 h after the clearing agent. Groups of four mice per time point were bled and sacrificed at 2, 24, 48, and 120 h after the injection of radiolabeled DOTA-biotin. Whole organs and tissue were isolated and weighed, and radioactivity was assessed in a Packard Cobra-II gamma counter as described previously (6).

The V_H and V_L DNA fragments were amplified from the B9E9 hybridoma cell line expressing the murine IgG2a anti-CD20 antibody. Comparison of the amino acid sequence of the B9E9 variable regions showed high homology with other well-characterized anti-CD20 antibodies, e.g., C2B8 and 1F5(Fig. 1; Refs. 21 and 22). The V_H and V_L cDNAs were constructed as a scFv, which was fused to the full-length streptavidin gene of Streptomyces avidinii (Fig. 2). The scFv-streptavidin gene was designed for expression of soluble,tetrameric fusion protein in the E. coli periplasm, which has an oxidizing environment enabling disulfide bond formation. The tetrameric structure of the protein is derived from using streptavidin,which spontaneously forms a stable homotetramer resistant to proteolysis. The scFvSA gene encodes a mature protein of 421 amino acids with a calculated monomeric weight of M_r 43,400 (most abundant mass),whereas the soluble tetramer that forms in the periplasm is M_r 173,600.

A number of genetic variants were constructed that contained scFv linkers of different lengths and composition or the variable regions in different order (Table 1). These constructs were analyzed first in shake flask cultures for expression, as visualized on Coomassie Blue-stained SDS-PAGE gels. The fusion protein migrated as a tetramer in nonboiled, nonreduced samples and as a monomer in samples boiled for 5 min. The linker size, composition, and order of the variable regions significantly affected the expression level (Table 1). Several clones that showed good expression in shake flask cultures were grown further in an 8-liter fermentor and analyzed for expression level and purification recovery. One scFvSA with a 25-mer Gly₄Ser scFv linker (Fig. 2, L1) in the V_H-V_L orientation expressed at 250–300 mg/liter and was purified from homogenized cell extracts with a 50–60% yield on a single iminobiotin affinity column. This purified fusion protein was analyzed further for protein characterization, immunoreactivity, biotin binding ability,immunohistology, blood clearance rate, and in vivo tumor targeting.

SDS-PAGE demonstrated that the fusion protein was purified to >95%homogeneity after iminobiotin chromatography (Fig. 3,A). The major band migrated at the expected molecular weight of ∼174,000, and minor isoforms were evident. These isoforms were also detected with polyclonal anti-streptavidin antibody on Western gel analysis (Fig. 3 B). However, all bands resolved into a single species of M_r ∼ 43,000 when the protein was boiled before electrophoresis, consistent with a single protein entity dissociable into its homogeneous subunit.

On size exclusion HPLC, the purified protein exhibited a major peak with a retention time appropriate for the tetramer and a minor (∼5%)peak representing a higher molecular weight, aggregated species (Fig. 4). Light-scattering HPLC indicated that the aggregate peak had a molecular weight of ∼350,000, suggesting it was a dimer of the tetramer. The deconvoluted mass spectrum of the monomer showed a molecular weight of 43,402, which is in agreement with the calculated most abundant mass of 43,400. Liquid chromatography-mass spectrometry/mass spectrometry analysis of the tryptic peptides of B9E9 scFvSA identified 15 of 21 predicted peptide fragments with the correct molecular weight and amino acid sequences yielding near complete sequence information. In addition, N₂-terminal sequencing of the purified protein revealed that the leader sequence was cleaved at the expected signal peptidase site adjacent to the first amino acid of the V_H region.

Immunoreactivity was assessed by flow cytometry in a competitive assay with fluorescein-labeled B9E9 mAb for binding to CD20-positive human lymphoma Ramos cells. The tetravalent scFvSA was about twice as immunoreactive (∼185%) as the divalent B9E9 mAb on a molar basis and nearly equivalent (∼93%) to B9E9 mAb when adjusted for its tetravalency (Fig. 5). Scatchard analysis indicated that the fusion protein retained the same relative nanomolar avidity as the B9E9 mAb. K_d values were 9.75 and 12.44 nm, and K_avalues were 1.02 × 10⁸ and 0.80 × 10⁸m⁻¹ for B9E9 mAb and scFvSA, respectively.

B9E9 scFvSA was further characterized for its biotin binding ability. The biotin dissociation rate was identical to that of recombinant streptavidin (t_1/2 = 379 and 364 min, respectively), indicating fully functional biotin binding. Determination of the biotin binding capacity revealed that B9E9 scFvSA had an average of 3.6 of 4 possible sites available for binding.

The binding of B9E9 scFvSA to cryosections of normal and human tumor tissues was compared with that of B9E9 mAb, C2B8/SA chemical conjugate(10), and a murine IgG2a isotype control mAb. All of the CD20 antibodies showed positive signals in a cell membrane pattern on B-cell aggregates, as expected from the known distribution of these cells. Thus, appropriate foci in sections of the intestine (large and small), lymph node, spleen, thymus, and tonsil and in malignant B-cell lymphoma showed strong signals with the B9E9 scFvSA, B9E9 mAb, and C2B8/SA conjugate.

Blood clearance studies were conducted in normal mice to examine the potential of the fusion protein in pretargeted RIT and to compare it with the C2B8/SA chemical conjugate. ¹²⁵I-labeled B9E9 scFvSA had a blood clearance half-life(t_1/2β) of 16 h, which was faster than the 46-h half-life of the C2B8/SA chemical conjugate (Fig. 6). The fusion protein was rapidly removed from the blood by a single i.v. injection of biotinylated poly(GalNAc) clearing agent. The scFvSA concentration dropped from 102 to 6.3 μg/gram of whole blood at 1 h after administration of the clearing agent.

The B9E9 fusion protein was examined in the full pretargeting protocol in female nude mice bearing human lymphoma Ramos xenografts. The dose of fusion protein used in this study was determined empirically by comparison with the blood AUC achieved with the C2B8/SA conjugate. ¹²⁵I-labeled B9E9 scFvSA was injected i.v. at t = −24 h, the synthetic clearing agent was injected at t = −4 h, and ¹¹¹In-labeled DOTA-biotin was injected at t = 0 h. As expected on the basis of the blood clearance results, the ¹²⁵I-labeled scFvSA concentration in blood and in most well-perfused soft tissues was very low (Fig. 7). Liver uptake and retention are functions of complexation with the clearing agent, which binds to asialoglycoprotein receptors in the liver. The kidney and intestinal concentrations are believed to be the result of hepatic metabolism of the fusion protein and subsequent renal or biliary excretion of radiolabel. A relatively high concentration of the labeled scFvSA targeted the tumor and was retained there. The apparent concentration decrease is attributable to the tumor volume doubling time of approximately 24 h in this xenograft model.

Stable delivery and retention of ¹¹¹In-DOTA-biotin at the tumor were also observed(Fig. 8). The highest concentration was at the tumor (both stoichiometrically and relative to blood pool concentration) at all time points. Peak concentrations of ¹¹¹In-DOTA-biotin at the tumor were 17–24% ID/g (mean, 21.66 ± 3.17% ID/g) and occurred by 2 h postadministration. This was a significantly higher peak uptake than that observed with the pretargeted C2B8/SA conjugate at the comparable time point (mean, 8.40 ± 0.62% ID/g; P < 0.05; data not shown). Tumor:blood ratios of pretargeted B9E9 scFvSA increased from ∼90 at 2 h after DOTA-biotin injection to >700 by 24 h after DOTA-biotin injection. ¹¹¹In-DOTA-biotin concentration in the blood and in most well-perfused soft tissues was very low, generally <2% of the injected dose in any of the assayed tissues at all time points. Significant liver uptake of ¹¹¹In-DOTA-biotin was not observed, indicating that the ¹²⁵I-labeled scFvSA had been efficiently internalized by the action of the clearing agent, making it unable to bind subsequently administered radiobiotin. In these experiments, no effort was made to optimize the dose of the fusion protein, clearing agent, or DOTA-biotin, nor was any effort made to optimize the schedule of administration of these components.

The CC49 scFvSA fusion protein, which is immunoreactive to an antigen(TAG-72) not expressed by Ramos cells, was used as a negative control to assess antigen-specific tumor localization. Consistent with our prior studies, targeting with ⁹⁰Y-DOTA-biotin(mean, 19.55 ± 5.86% ID/g) was equivalent to targeting with ¹¹¹In-DOTA-biotin (mean, 21.66 ± 3.17% ID/g). By contrast, there was no targeting of ⁹⁰Y-DOTA-biotin to the tumor by the negative control fusion protein (0.78 ± 0.42% ID/g). Nontarget tissue concentrations of CC49 pretargeted ⁹⁰Y-DOTA-biotin were nearly identical to the B9E9 values in all tissues at all time points, demonstrating the absence of antigen-mediated uptake in normal mouse tissues of either antibody fusion construct. The rapid tumor uptake achieved with B9E9 scFvSA resulted in a 2700% greater tumor dose as compared with the CC49 scFvSA (Fig. 9), whereas the % ID/g-h AUCs of blood showed similar total values (9%ID/g-h versus 18% ID/g-h for B9E9 and CC49 fusion proteins,respectively). Ratios of tumor AUC:blood AUC were >60 for the B9E9 group versus 1.2 for the CC49 group.

The B9E9 scFvSA fusion protein was designed to be a homogeneous and cost-effective bifunctional vehicle to mediate biotin-streptavidin pretargeted RIT of NHL. E. coli was chosen as the expression host because the periplasm allows disulfide formation and contains low levels of free biotin that would otherwise occlude the binding sites of streptavidin. In addition, fermentation conditions for E. coli are readily scaleable, enabling production of the gram quantities needed for clinical trials. Initial efforts to produce high levels of a V_L-V_H B9E9 scFvSA with a 15-mer Gly₄Ser linker were unsuccessful, leading us to generate numerous genetic variants with different scFv linkers and different orders of the variable regions. Two constructs, both V_H-V_L with either a 18-mer or a 25-mer Gly₄Ser linker, were highly expressed(>100 mg/liter) in E. coli as soluble, tetrameric fusion proteins. Both purified proteins had apparently identical functional characteristics in vitro and in vivo and could only be differentiated by their expression level.

Although high level E. coli expression of antibody fragments has been achieved, most scFvSA fusion proteins are poorly expressed or insoluble in the periplasmic space (23, 24, 25, 26, 27). Despite the fact that insoluble proteins in the cytoplasm or periplasm can occasionally be refolded from inclusion bodies, this approach is not practical for large scale or high molecular weight proteins. Multiple factors, such as the leader sequence, the linker used, the order of variable regions, the host strain, and the growth conditions, affect the periplasmic expression of antibody fragments and the degree of aggregation (28, 29, 30, 31). In this study, the scFvSA expression levels were affected dramatically by the order of the variable regions and the length and/or composition of the scFv linker. No variations were made in the leader sequence or the linker between the scFv and streptavidin, and these also may be amenable to further optimization.

The in vitro and in vivo functionality demonstrated by the B9E9 scFvSA is consistent and reproducible. In vitro, the competitive immunoreactivity with the B9E9 mAb showed evidence of increased avidity of the tetramer, a phenomenon already noted with trivalent and tetravalent Fab′ constructs(32). Also exhibited is the intact immunoreactivity of the variable regions of the whole antibody when incorporated into the scFv motif. The biotin binding capacity of the scFvSA was nearly equivalent to that of commercially available recombinant streptavidin, whereas the biotin dissociation rate of the B9E9 scFvSA at physiological temperature was unperturbed. These data indicate that the fusion of B9E9 scFv to the NH₂ terminus of each subunit of streptavidin had a negligible effect on the tertiary and quanternary structural requirements of streptavidin to mediate high-affinity biotin binding. The biochemical uniformity of the purified B9E9 scFvSA alone makes it a superior test agent compared with our first-generation mAb-streptavidin covalent conjugates. In addition, the fusion protein exhibits increased antigen-binding avidity, which should decrease streptavidin dissociation from tumor, a key factor in the mathematical modeling of multistep delivery protocols (33, 34). This could be particularly critical in our pretargeting protocol, which rapidly and dramatically disrupts the equilibrium between tumor-bound and circulating immunoconjugate by administration of a clearing agent.

In vivo, the B9E9 scFvSA exhibited more rapid systemic clearance than our mAb-streptavidin conjugates, which is consistent with its lack of the Fc region of the antibody. However, the greater molecular weight of the scFvSA tetramer (∼174,000) compared with conventional antibody fragments (Fab, Fab′, and scFv) provides the fusion protein with sufficient hydrodynamic radius to avoid direct renal elimination via glomerular filtration. In fact, no intact fusion protein or monomeric subunits can be detected in animals’ urine after systemic administration (data not shown). The B9E9 scFvSA half-life of∼16 h appears to be sufficient to allow extravasation from the blood and efficient binding to the tumor-associated antigen. Dual-label coinjection studies of equimolar amounts of C2B8 mAb and B9E9 scFvSA in mice bearing lymphoma xenografts have shown that peak tumor concentrations of both proteins are equivalent, despite the prolonged,elevated blood concentration of the C2B8 molecule. In general, the absence of the whole antibody Fc region should minimize uptake of B9E9 scFvSA in specialized antibody reservoirs, such as the Brambell receptor (35). In our full pretargeting protocol, B9E9 scFvSA is directed quantitatively from the blood to the liver for hepatic processing in a manner similar to that described for other conjugates and tumor models (6).

In recent years, immunotherapy of relapsed or refractory low-grade NHL with CD20 mAbs has been shown to be effective in inducing an objective tumor response. Clinical studies using unlabeled antibody (C2B8,rituximab) and radiolabeled antibodies (Y2B8 and Bexxar) have been recently reviewed by others (1, 2, 3). The work of Press et al. (4) and Liu et al.(5) using myeloablative doses of ¹³¹I-labeled B1 has proven that dose intensification of I-131 by a factor of ∼8 results in a higher complete response rate and a longer duration of response, albeit at the cost of stem cell rescue. Whether given in nonmyeloablative or myeloablative protocols, RIT continues to show promise for replacing total body irradiation in the treatment of NHL. Our own pilot clinical pretargeting efforts in this indication using C2B8/SA conjugate as a model targeting vehicle yielded promising results in terms of clinical responses, absolute tumor-targeting efficiency, and minimal myelotoxicity, even at high ⁹⁰Y doses(10).

We thank Gina Desfachelles for technical assistance in the mouse studies, Santosh Kumar (University of Washington, Seattle, WA)for NH₂-terminal sequencing, Mark Hylarides for the deaggregation protocol, and Steve Goshorn for cloning of the streptavidin gene and construction of the initial fusion protein cassette.