Mammalian expression and hollow fiber bioreactor production of recombinant anti-CEA diabody and minibody for clinical applications.

Genetically engineered radiolabeled antibody fragments have shown great promise for the radioimmunoscintigraphy of cancer. Retaining the exquisite specificity of monoclonal antibodies yet smaller in molecular size, antibody fragments display rapid tumor targeting and blood clearance, a more uniform distribution in the tumor, and present a lower potential to elicit an immune response. However, one of the factors that has limited clinical evaluation of these antibody-derived proteins has been the difficulty in expressing and purifying the quantities necessary for clinical trials. This study outlines the capability of mammalian expression for the production of recombinant antibody fragments intended for clinical use. Two anti-carcinoembryonic antigen antibody fragments, the T84.66/212 Flex minibody (scFv-C(H)3) and the T84.66 diabody (scFv dimer) have been previously expressed and have shown excellent radioimaging properties in tumor bearing animals. To proceed toward human studies, these high affinity recombinant fragments and a second minibody version, the T84.66/GS18 Flex minibody, were expressed using a high-level mammalian expression system. Production of all three antibody fragments in a small-scale hollow fiber bioreactor resulted in 137-307 mg of crude antibody harvest. A purification protocol that employed ceramic hydroxyapatite and anion exchange chromatography resulted in 50-150 mg of purified T84.66 diabody and T84.66 minibody. The development of this level of research grade material established conditions for clinical production as well as provided material to complete pre-clinical studies and undertake protein crystallization studies. Scale-up for clinical studies produced 3.4 g of the T84.66 minibody in the harvest. A portion of this material was purified yielding 180 mg of highly purified T84.66 minibody intended for pilot radioimmunoscintigraphy studies of carcinoembryonic antigen (CEA) positive disease.

Tumor targeting of radiometal labeled anti-CEA recombinant T84.66 diabody and t84.66 minibody: comparison to radioiodinated fragments.

Recombinant antibody fragments offer potential advantages over intact monoclonal antibodies in the radioimmunoscintigraphy (RIS) of solid tumors. Due to their smaller molecular size, antibody fragments have shown rapid tumor targeting and blood clearance, a more uniform tumor distribution and a lower potential to elicit a human immune response. Previously, we have expressed two genetically engineered antibody fragments, the T84.66 diabody (scFv dimer) and the T84.66 minibody (scFv-CH3 dimer), specific to carcinoembryonic antigen (CEA). When radioiodinated, both antibody fragments exhibited rapid tumor targeting and rapid blood clearance in xenografted mice. To extend and optimize their future clinical RIS utility with radiometals, these antibody fragments were conjugated with the macrocycle 1,4,7,10-tetraazacyclododecane N,N',N' ',N' "-tetraacetic acid (DOTA) and labeled with 111In. Tumor targeting and biodistribution studies were carried out in athymic mice xenografted with a human colorectal tumor cell line, LS174T. The [111In]T84.66 diabody (55 kDa) exhibited very rapid tumor targeting with 12.5 +/- 0.4% injected dose per gram (% ID g(-1) +/- standard error) at 2 h and reached a maximum of 13.3 +/- 0.9% ID g(-1) at 6 h. However, kidney uptake was observed to reached a peak of 183.5 +/- 21.0% ID g(-1) at 6 h, a result similar to that reported by others for other low molecular weight fragments labeled with radiometals. Preadministration of an oral dose of D-lysine resulted in a 59% lowering of the renal accumulation at 6 h, but was accompanied by a 31% reduction of tumor uptake to 9.2 +/- 1.2% ID g(-1). The second recombinant antibody fragment, the [111In]T84.66 minibody (80 kDa), displayed rapid tumor targeting of 14.2 +/- 6.1% ID g(-1) at 2 h, and reached a maximum activity of 24.5 +/- 6.1% ID g(-1) by 12 h. Renal uptake achieved a plateau of 12-13% ID g(-1) which cleared to 7.2% ID g(-1) at 72 h. However, hepatic uptake was elevated and reached a maximum of 26.0 +/- 1.0% ID g(-1) at 12 h in these xenograft-bearing mice. Experiments in nontumor bearing mice showed a reduction of hepatic activity at 12 h to 16.6 +/- 1.5% ID g(-1), indicative of an intrinsic hepatic accumulation of the [111In]DOTA-T84.66 minibody or metabolites. While the anti-CEA [111In]DOTA-T84.66 diabody and T84.66 minibody retain the rapid tumor targeting properties of the radioiodinated form, the normal organ accumulation (kidneys and liver, respectively) of the [111In]DOTA forms appeared problematic for RIS and RIT applications. Development of alternative blocking strategies or new metabolizable chelates are under investigation to enhance the utility of the radiometal form of these and other promising recombinant antibody fragments.

Numerical selection of optimal tumor imaging agents with application to engineered antibodies.

Three analytic indicators were used to compare five members of a monoclonal antibody (Mab) family. The cognates consisted of the genetically engineered intact chimeric IgGI (cT84.66) and related engineered fragments [scFv, diabody, minibody, F(ab')2] reactive against the same epitope of carcinoembryonic antigen (CEA). All analyses were based on radioiodinated Mabs targeting to colorectal xenografts of LS174T tumors in nude mice. Affinity constants were evaluated initially. A second indicator was the imaging figure of merit (IFOM) which determines how rapidly a statistically significant tumor image can be acquired. Finally, deconvolution was used to determine tumor temporal response to an arterial bolus. This last analysis gave the possible tumor accumulation in the absence of normal tissue sequestration. Affinities were all in excess of 10(8) M-1 and were highest for the divalent Mabs. Using the IFOM criterion, an 131I label was best suited as a radiolabel for the intact (IgG) T84.66, while an 123I label indicated optimal imaging with either minibody or F(ab')2. Deconvolution analyses showed that divalent members behaved similarly while the univalent member (scFv) had a tumor residence time smaller by an order of magnitude. The diabody had the largest impulse response function, but renal uptake may limit its present usefulness.

Designer genes: recombinant antibody fragments for biological imaging.

Monoclonal antibodies (MAbs), with high specificy and high affinity for their target antigens, can be utilized for delivery of agents such as radionuclides, enzymes, drugs, or toxins in vivo. However, the implementation of radiolabeled antibodies as "magic bullets" for detection and treatment of diseases such as cancer has required addressing several shortcomings of murine MAbs. These include their immunogenicity, sub-optimal targeting and pharmacokinetic properties, and practical issues of production and radiolabeling. Genetic engineering provides a powerful approach for redesigning antibodies for use in oncologic applications in vivo. Recombinant fragments have been produced that retain high affinity for target antigens, and display a combination of rapid, high-level tumor targeting with concomitant clearance from normal tissues and the circulation in animal models. An important first step was cloning and engineering of antibody heavy and light chain variable domains into single-chain Fvs (molecular weight, 25-27 kDa), in which the variable regions are joined via a synthetic linker peptide sequence. Although scFvs themselves showed limited tumor uptake in preclinical and clinical studies, they provide a useful building block for intermediate-sized recombinant fragments. Covalently linked dimers or non-covalent dimers of scFvs (also known as diabodies) show improved targeting and clearance properties due to their higher molecular weight (55 kDa) and increased avidity. Further gains can be made by generation of larger recombinant fragments, such as the minibody, an scFv-CH3 fusion protein that self-assembles into a bivalent dimer of 80 kDa. A systematic evaluation of scFv, diabody, minibody, and intact antibody (based on comparison of tumor uptakes, tumor:blood activity ratios, and calculation of an Imaging Figure of Merit) can form the basis for selection of combinations of recombinant fragments and radionuclides for imaging applications. Ease of engineering and expression, combined with novel specificities that will arise from advances in genomic and combinatorial approaches to target discovery, will usher in a new era of recombinant antibodies for biological imaging.

High-resolution microPET imaging of carcinoembryonic antigen-positive xenografts by using a copper-64-labeled engineered antibody fragment.

Rapid imaging by antitumor antibodies has been limited by the prolonged targeting kinetics and clearance of labeled whole antibodies. Genetically engineered fragments with rapid access and high retention in tumor tissue combined with rapid blood clearance are suitable for labeling with short-lived radionuclides, including positron-emitting isotopes for positron-emission tomography (PET). An engineered fragment was developed from the high-affinity anticarcinoembryonic antigen (CEA) monoclonal antibody T84.66. This single-chain variable fragment (Fv)-C(H)3, or minibody, was produced as a bivalent 80 kDa dimer. The macrocyclic chelating agent 1,4,7, 10-tetraazacyclododecane-N,N',N", N"'-tetraacetic acid (DOTA) was conjugated to the anti-CEA minibody for labeling with copper-64, a positron-emitting radionuclide (t(1/2) = 12.7 h). In vivo distribution was evaluated in athymic mice bearing paired LS174T human colon carcinoma (CEA positive) and C6 rat glioma (CEA negative) xenografts. Five hours after injection with (64)Cu-DOTA-minibody, microPET imaging showed high uptake in CEA-positive tumor (17.9% injected dose per gram +/- 3.79) compared with control tumor (6.0% injected dose per gram +/- 1.0). In addition, significant uptake was seen in liver, with low uptake in other tissues. Average target/background ratios relative to neighboring tissue were 3-4:1. Engineered antibody fragments labeled with positron-emitting isotopes such as copper-64 provide a new class of agents for PET imaging of tumors.

Truncation of blood curves to enhance imaging and therapy with monoclonal antibodies.

Targeting of monoclonal antibody (Mab) to solid tumor sites is a function of the blood curve of activity versus time. It has been suggested that the blood curve be artificially reduced to approach zero so that the contrast between tumor and blood uptake is maximized. We analyzed tumor uptake as a function of the time tc of blood curve truncation. By using a convolution approach, we were able to find the optimal times for setting the blood curve to zero in either diagnostic or therapeutic animal examples. Two iodinated cT84.66 anti-CEA engineered fragments, diabody and minibody, were considered using previous data from nude mouse studies involving the LS174T colorectal tumor model. Figures of merit (FOMs) were used to compare ordinary and truncated blood curves and their associated tumor accumulations. Using a 1231 label, it was seen that the appropriate time for diagnostic truncation occurred when tumor uptake, as measured, was a maximum. The corresponding point for therapy (with 1311 as a label) was at infinite time. We also demonstrated that the use of traditional indices led to ambiguities in the choice of truncation times. The traditional therapy index, the ratio of the integral of the tumor uptake to the integral of the blood uptake, was found to be a numerical constant independent of tc. This ratio was proved to be the integral of the tumor impulse response function. Use of such convolution techniques to assess truncation of the perfused material is probably also applicable to multistep processes as well as to lesion targeting with other tumor-specific pharmaceuticals.

A secreted and LIF-mediated stromal cell-derived activity that promotes ex vivo expansion of human hematopoietic stem cells.

The development of culture systems that facilitate ex vivo maintenance and expansion of transplantable hematopoietic stem cells (HSCs) is vital to stem cell research. Establishment of such culture systems will have significant impact on ex vivo manipulation and expansion of transplantable stem cells in clinical applications such as gene therapy, tumor cell purging, and stem cell transplantation. We have recently developed a stromal-based culture system that facilitates ex vivo expansion of transplantable human HSCs. In this stromal-based culture system, 2 major contributors to the ex vivo stem cell expansion are the addition of leukemia inhibitory factor (LIF) and the AC6.21 stromal cells. Because the action of LIF is indirect and mediated by stromal cells, we hypothesized that LIF binds to the LIF receptor on AC6.21 stromal cells, leading to up-regulated production of stem cell expansion promoting factor (SCEPF) and/or down-regulated production of stem cell expansion inhibitory factor (SCEIF). Here we demonstrate a secreted SCEPF activity in the conditioned media of LIF-treated AC6.21 stromal cell cultures (SCM-LIF). The magnitude of ex vivo stem cell expansion depends on the concentration of the secreted SCEPF activity in the SCM-LIF. Furthermore, we have ruled out the contribution of 6 known early-acting cytokines, including interleukin-3, interleukin-6, granulocyte macrophage colony-stimulating factor, stem cell factor, flt3 ligand, and thrombopoietin, to this SCEPF activity. Although further studies are required to characterize this secreted SCEPF activity and to determine whether this secreted SCEPF activity is mediated by a single factor or by multiple growth factors, our results demonstrate that stromal cells are not required for this secreted SCEPF activity to facilitate ex vivo stem cell expansion. (Blood. 2000;95:1957-1966)

Expression, purification, and characterization of a two domain carcinoembryonic antigen minigene (N-A3) in pichia pastoris. The essential role of the N-domain.

Carcinoembryonic antigen (CEA) is a 180 kDa glycoprotein expressed on the surface of normal and malignant human colon. The structure of CEA has seven predicted Ig-like domains (N-A1-B1-A2-B2-A3-B3) that are encoded by separate exons and contain independent epitopes that are recognized by monoclonal antibodies. The N-domain mediates homotypic cell adhesion as shown by deletion expression analysis, and may also interact with the A3 domain. Although we have been unsuccessful in expressing these domains in high yields of active protein in either bacterial or mammalian expression systems, we now report high yield expression in Pichia pastoris of a mini-gene (N-A3) comprising the N and A3 domains of CEA, and containing epitopes for the monoclonal antibodies T84.1 and T84.66. N-A3 was constructed by splice overlap PCR from the CEA gene and fused to the yeast alpha-mating factor leader sequence and an N-terminal His6 tag. The secreted protein gave high level expression (20 micrograms/mL) and was purified in two steps using Ni(NTA) affinity chromatography followed by reversed phase HPLC. The purified protein (yield 6 mg from 600 mL of supernatant) had a single N-terminal sequence, the expected amino acid composition, and retained full reactivity to both T84.1 and T84.66 compared to native CEA. BIAcore analysis gave a Kaff of 4.4 x 10(10) M-1 for the binding of N-A3 to T84.1 and 2.2 x 10(10) M-1 for the binding of N-A3 to T84.66. The molecular weight of N-A3 was 37 kDa before and 24 kDa after enzymatic deglycosylation as determined by SDS gel electrophoresis. The average N-glycosyl unit was calculated at 1850 Da (for 7 N-linked sites) suggesting a GN2Man9 oligosaccharide structure. N-A3 migrated as a dimer at 80 kDa and a monomer at 40 kDa on gel filtration analysis performed at pH 7.5, and 4.0, respectively. CEA exhibited the same conversion of dimers to monomers when analyzed by gel filtration at neutral and acid pH. The availability of this highly active CEA mini-gene should enable further structure-function studies including epitope analysis and investigation of monomer-dimer interactions.

The skeletal muscle transverse tubular Mg-ATPase: identity with Mg-ATPases of smooth muscle and brain.

Purified chicken skeletal muscle transverse tubule (T-tubule, TT) membrane preparations contain a very active Ca- or Mg-ATPase (EC 3.6.1.3) previously thought to be a T-system-specific marker enzyme. The function of the Mg-ATPase has not yet been determined although its prominent activity and concentration in junctional complexes supports a possible role in the excitation-contraction cycle. An essential component of the Mg-ATPase has been identified as a M(r) 85,000 glycoprotein (85k-GP). Polyclonal antibodies raised against the TT 85k-GP were specific and exhibited no cross-reactivity with other skeletal muscle proteins on immunoblots. Using this anti-85k-glycoprotein IgG, we have explored other chicken tissues to determine the tissue distribution of the 85k-GP. Antibody reactive polypeptides of M(r) 85,000 were found in gizzard smooth muscle, brain, heart, spleen, and lung tissue. The brain and smooth muscle membrane proteins were further purified and characterized for 85k-GP-associated Mg-ATPase activity. The brain and smooth muscle enzymes exhibited properties indistinguishable from the skeletal muscle TT-specific Mg-ATPase with regard to a series of activators and inhibitors, amino terminal amino acid sequences, and the effects of deglycosylation. The enzyme in all three tissues was inhibited by the diacylglycerol kinase inhibitor R 59022. Identification of the TT Mg-ATPase in gizzard smooth muscle has allowed the investigation of the Mg-ATPase membrane topology using isolated whole smooth muscle cells. The data support an ecto-orientation for the smooth muscle cell enzyme. Although the orientations of the brain and skeletal muscle enzymes have not been conclusively determined, the nearly identical properties of all three enzymes argues for an ecto-orientation of the active sites of these enzymes as well. The responsiveness of the three enzymes to regulatory lipids suggests that the ecto-Mg-ATPase may serve as a master switch controlling extracellular ATP concentrations and ligand accessibility to P1- and P2-purinoceptors. It is also proposed that the ecto-MgATPase may regulate ATP accessibility to ectoprotein kinases in a variety of tissues, and, in brain, the ecto-MgATPase may modulate the neurotransmitter role of ATP.

Amino acid sequence of chicken calsequestrin deduced from cDNA: comparison of calsequestrin and aspartactin.

We have previously reported the amino terminal sequence of adult chicken calsequestrin, an intraluminal Ca2(+)-binding protein isolated from fast-twitch skeletal muscle. The partial sequence showed homology with mammalian calsequestrins contained in the PIR data bank and complete identity with the amino terminus of a putative laminin-binding protein of the extracellular matrix, aspartactin. Based on these data, oligonucleotide primers were synthesized for PCR amplification and direct DNA sequencing. We report herein the primary sequence of chicken calsequestrin, deduced from cDNA. The sequence has been verified by amino acid sequencing of internal tryptic peptides. Importantly, the data show the primary structure of calsequestrin to be identical to the amino acid sequence reported for aspartactin, with the exception of a single amino acid difference (ileu vs. val) which may be animal strain-related. Based on these data, calsequestrin and aspartactin are the same protein.

Calsequestrin, an intracellular calcium-binding protein of skeletal muscle sarcoplasmic reticulum, is homologous to aspartactin, a putative laminin-binding protein of the extracellular matrix.

Calsequestrin was isolated from chicken fast-twitch skeletal muscle, and partial amino terminal sequence was determined. The sequence (NH2) EEGLNFPTYDGKDRVIDLNE shows high identity with known mammalian calsequestrins contained in the Protein Identification Resource data bank (1). Most importantly, this 20 amino acid sequence shares complete identity with the amino terminus of aspartactin, a putative laminin-binding protein of the extracellular matrix (2, 3). The possible relationship of aspartactin to calsequestrin is discussed.

Fate of influenza A virion proteins after entry into subcellular fractions of LLC cells and the effect of amantadine.

Influenza A virus entry into the host cell was studied by monitoring the fate of virion proteins in subcellular fractions of LLC-RMK2 cells under conditions of low multiplicity of infection in the absence of drugs. Adsorption but not entry could be demonstrated at 4 degrees. Restricted entry into a prelysosomal pool could be demonstrated at 20 degrees. At 37 degrees proteins entered the cell in the same relative distribution as present in virions arguing against the fusion hypothesis and for the endocytosis hypothesis. The hemagglutinin and matrix proteins were readily degraded intracellularly at 37 degrees, but the nucleoprotein was relatively resistant to degradation, also consistent with the latter hypothesis. Amantadine blocked neither entry into the host cell nor transfer between the prelysosomal and lysosomal compartments; however, it appeared to block exit from the prelysosomal and lysosomal pools. The addition of the drug reduced degradation of the matrix protein in these pools but not of the hemagglutinin, consistent with inhibition by amantadine of fusion of virion envelope and vesicle membrane.

Role of phospholipase A inhibition in amiodarone pulmonary toxicity in rats.

Amiodarone is effective in the treatment of ventricular and supraventricular arrhythmias. In man its clinical use is associated with the accumulation of phospholipid-rich multilamellar inclusions in various tissues including lung, liver and others. This report presents evidence showing that amiodarone is a potent inhibitor of lysosomal phospholipase A from rat alveolar macrophages, J-744 macrophages and rat liver. When compared with other cationic amphiphilic agents which are known to produce phospholipidosis, amiodarone is one of the most potent inhibitors yet discovered. The subcellular localization of amiodarone has been determined in lung and its distribution was consistent with a lysosomal localization. It is hypothesized that amiodarone causes cellular phospholipidosis by concentrating in lysosomes and inhibiting phospholipid catabolism.

Inhibition of purified lysosomal phospholipase A1 by beta-adrenoceptor blockers.

Inhibition of rat liver lysosomal phospholipases is one of the main events that leads to accumulation of tissue phospholipids during drug-induced phospholipidosis. Drug inhibition of lysosomal phospholipase A may occur by direct effects of drugs on the enzyme (or substrate) or by drug-induced increases in intralysosomal pH. Although beta-adrenoceptor blockers have not been reported to cause lipid storage, they do inhibit lysosomal phospholipase A. To investigate the structural requirements for drug inhibition, we studied the effects of six beta-adrenoceptor blockers on purified rat liver lysosomal phospholipase A1. The agents studied include: propranolol, timolol, metoprolol, practolol, atenolol and the combined alpha and beta adrenoceptor blocking agent, labetalol. The drugs varied by two logs in their abilities to inhibit phospholipase A1 activity. The relative inhibitory potencies were propranolol greater than labetalol much greater than timolol greater than metoprolol much greater than practolol greater than atenolol. Our studies identify drug hydrophobicity as a key determinant for phospholipase A1 inhibition. A strong negative correlation was noted between the octanol/water partition coefficients and IC50 for phospholipase inhibition (r = -0.91). The ability of propranolol to inhibit phospholipase A1 was identical for the d, l and the d and l stereoisomers.

Chloroquine-induced phospholipid fatty liver. Measurement of drug and lipid concentrations in rat liver lysosomes.

A method has been developed to measure the concentration of chloroquine in lysosomes isolated from the liver of rats. It employs 3H2O and [U-14C]sucrose to determine the intralysosomal water volume of purified lysosomes obtained by free flow electrophoresis. Twelve h after a single dose, the concentration of chloroquine in lysosomes was 6.3 mM and at 24 h it rose to 16.5 mM. With continued treatment, lysosomal chloroquine concentrations were 61 and 74 mM at 48 and 72 h. The lysosomal concentrations of chloroquine attained were sufficient to block intralysosomal phospholipase A1 activity. The lysosomal content of phospholipid rises 1.7-fold and 2.6-fold over that of control at 12 and 24 h, respectively. At 72 h, lysosomal phospholipid was 3.7-fold greater than that of control. Lysosomes show an increased negative surface charge with chloroquine administration which is due in part to an increased ratio of acidic to neutral phospholipids in the lysosomal membrane. The phosphatidylinositol content of lysosomes rose rapidly with chloroquine treatment and accounted for the early increase in the ratio. Bis(monoacylglycero)phosphate, an acidic phospholipid synthesized only in lysosomes, increased later in the course of chloroquine treatment and accounted for the continued increase in acidic phospholipids.

The subcellular localization of neutral sphingomyelinase in rat liver.

The subcellular distribution of neutral sphingomyelinase activity has been determined in rat liver. Neutral sphingomyelinase is present in the plasma membrane. This enzyme requires either Mg2+ or Mn2+ for full activity; these cations cannot be replaced by Co2+ or Ca2+. The plasma membrane sphingomyelinase is strongly inhibited by Hg2+. A small amount of neutral spingomyelinase activity appears to be present in microsomes. No neutral sphingomyelinase activity is present in liver mitochondria or bytosol. Lysosomal sphingomyelinase is fully active at pH 4.4--4.8 without added divalent cations. However, between pH 5.0 and 7.5 lysosomal sphingomyelinase activity is stimulated by Mg2+, Mn2+, Co2+, and Ca2+. Below pH 4.8, Mg2+ inhibits the reaction. In contrast to the results obtained with the neutral sphingomyelinase activity of plasma membranes and microsomes, lysosomal sphingomyelinase is unaffected by sulfhydryl inhibitors.