Membrane proteins are still heavily under-represented in the protein data bank (PDB), owing to multiple bottlenecks. The typical low abundance of membrane proteins in their natural hosts makes it necessary to overexpress these proteins either in heterologous systems or through in vitro translation/cell-free expression. Heterologous expression of proteins, in turn, leads to multiple obstacles, owing to the unpredictability of compatibility of the target protein for expression in a given host. The highly hydrophobic and (or) amphipathic nature of membrane proteins also leads to challenges in producing a homogeneous, stable, and pure sample for structural studies. Circumventing these hurdles has become possible through the introduction of novel protein production protocols; efficient protein isolation and sample preparation methods; and, improvement in hardware and software for structural characterization. Combined, these advances have made the past 10-15 years very exciting and eventful for the field of membrane protein structural biology, with an exponential growth in the number of solved membrane protein structures. In this review, we focus on both the advances and diversity of protein production and purification methods that have allowed this growth in structural knowledge of membrane proteins through X-ray crystallography, nuclear magnetic resonance (NMR) spectroscopy, and cryo-electron microscopy (cryo-EM).

The secretory production of recombinant proteins by the Gram-negative bacterium Escherichia coli has several advantages over intracellular production as inclusion bodies. In most cases, targeting protein to the periplasmic space or to the culture medium facilitates downstream processing, folding, and in vivo stability, enabling the production of soluble and biologically active proteins at a reduced process cost. This review presents several strategies that can be used for recombinant protein secretion in E. coli and discusses their advantages and limitations depending on the characteristics of the target protein to be produced.

We report the development of a system for displaying bivalent antibody fragments on M13 bacteriophage in a manner that effectively mimics the binding behavior of natural antibodies. In the "bivalent display" format, two copies of antigen binding sites are displayed on the coat of a single phage particle. Bivalent display was first achieved by the insertion of a dimerization domain, consisting of an IgG1 hinge region and a homodimerizing GCN4 leucine zipper, between a Fab and the C-terminal domain of the M13 gene-3 minor coat protein. In a phagemid-based display system, the resulting "Fab'-zip-phage" particles display bivalent Fabs that resemble natural IgGs. An important functional consequence of bivalent display is an avidity effect, which results in a greatly reduced off-rate for phage bound to immobilized antigen. The avidity effect improved the capture and retention of bivalent Fab'-zip-phage relative to monovalent Fab-phage both with antigen immobilized on plates and with cell surface antigen. To examine the requirements for bivalent display on phage, we systematically trimmed down the dimerization domain and found that a single cysteine was sufficient to confer the same avidity effect conferred by the complete dimerization domain. Bivalent antibody phage display should be useful for many applications. In particular, the technology should aid in the production of antibodies against difficult antigens, and also, in selections that require dimerization for activity.

F(ab')(2) fragments are desirable structural derivatives of monoclonal antibodies (MAbs) because of their pharmacokinetic properties and bivalent binding to antigen. Production of these fragments, however, has proven difficult because of the variable sensitivity of intact antibodies to proteolytic enzymes, which can result in very low yields and unstable product. To circumvent these problems, we attempted to apply genetic engineering methods to generate stable F(ab')(2) fragments in NSO murine myeloma cells using the glutamine synthase expression system. For these studies, the chimeric MAb, chTNT-3, directed against necrotic regions of solid tumors, was used to generate several F(ab')(2) variants, which contained between one and three cysteine residues at the end of the hinge region. In addition, two different affinity tags (his tag, streptactin tag) were used with each variant to determine the best tag for purification procedures. Stability was measured by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) and by antigen binding studies and the constructs were tested in vivo to measure their pharmacokinetic properties and biodistribution in normal organs and tumor. The results of these studies show that 3 cysteine residues are required to produce stable F(ab')(2) fragments and that either purification tag can be used with this variant to produce suitable reagents for in vivo studies. Those constructs containing one or two cysteines were found to be unstable and broke down to Fab fragments regardless of the purification tag used. These studies demonstrate that stable, clinically useful F(ab')(2) fragments of chTNT-3 can be produced in mammalian cells by genetic engineering methods.

New E. coli vectors based on the pOPE/pSTE vector system [Gene 128 (1993) 97] were constructed to express a single-chain Fv antibody fragment (scFv), a scFv-streptavidin fusion protein and two disulfide bond-stabilized Fv antibody fragments (dsFvs) utilizing different side chain positions for disulfide stabilization. All of these constructs encoded fusion proteins carrying five C-terminal histidine residues preceded by an unpaired cysteine. The influence of this cysteine, which was originally introduced to allow the chemical modification of the fusion proteins, was assessed by exchanging the two amino acids CysIle in front of the carboxy terminal His-tag to SerHis in all constructs. Yield and antigen-binding activity of the antibody constructs were compared after standard lab-scale periplasmic expression in Escherichia coli. The removal of the unpaired cysteine resulted in a significant increase in antigen-binding activity of the crude periplasmic extracts. Further, a three-five fold increase of yield and a significantly improved purity were observed after immobilized metal affinity chromatography (IMAC) with all four constructs.

Fab's with hinges based on the human gamma1 sequence containing 1, 2, or 4 cysteines have been produced by high level Escherichia coli periplasmic secretion, and coupled in vitro by reduction/oxidation to form F(ab')2. We find that the F(ab')2 made with hinges containing 2 or 4 cysteines have a high level (approximately 70%) of multiple disulphide bonds. These F(ab')2 molecules have an increased pharmacokinetic stability as measured by area under the curve compared to those made by direct coupling through a single disulphide bond. One particular molecule containing 4 hinge cysteines has a greater pharmacokinetic stability than a F(ab')2 formed by chemical cross-linking. F(ab')2 made from the Fab' with 4 hinge cysteines is also relatively resistant to chemical reduction in vitro allowing partial reduction to expose reactive hinge thiols. These hinge sequences provide a simple method for producing robust F(ab')2 in vitro, obviating the need to use chemical cross-linkers, and provide a route to hinge specific chemical modification with thiol-reactive conjugates.

Since MRD is the major cause for relapses of malignant diseases, strategies utilizing ITs to target tumor cells surviving conventional treatment have attracted scientific and clinical interest. Many different ITs against various blood-borne as well as solid malignancies have demonstrated specific potent anti-tumor effects in vitro and in animal models. Some of these have already undergone clinical phase I/II-trials. The dose-limiting toxicities of RTA ITs include manifestation of VLS presenting as decreased urinary sodium excretion, hypoalbuminemia, fatigue, hypotonia, myalgia, pulmonary edema, or rhabdomyolysis. Problems encountered clinically include the development of HAMA, HARA, and HACA and the selection of antigen-deficient malignant clones. Most clinical trials performed with ITs so far were conducted in heavily pretreated patients presenting with high tumor burdens. Thus, the responses observed with ITs in these trials are very encouraging and warrant further exploration.

Targeting in immunotherapy has traditionally been achieved by using monoclonal rodent antibodies. Despite gene-engineering, there are many problems and limitations associated with the non-human origin, the targeting specificity and the binding strength of these molecules. Now these issues may be addressed in a more rational way, by designing and then shaping, in vitro, the desired human antibodies. This review addresses how this may be achieved by the selection of monoclonal human antibodies from phage display libraries and the engineering of affinity and specificity thereafter. Phage display of antibody fragments has allowed access to large collections of different phage antibodies, created by cloning antibody V-genes from B-cells. Antibodies against any type of antigen may be derived from such repertoires, by rounds of enrichment on antigen and re-amplification. This review presents the state of the art in rational antibody design and creation. It will highlight the strengths of this increasingly important field, which will aid in the generation of tailor-made targeting entities for immunotherapy.

Multivalency is one of the hallmarks of antibodies, by which enormous gains in functional affinity, and thereby improved performance in vivo and in a variety of in vitro assays are achieved. Improved in vivo targeting and more selective localization are another consequence of multivalency. We summarize recent progress in engineering multivalency from recombinant antibody fragments by using miniantibodies (scFv fragments linked with hinges and oligomerization domains), spontaneous scFv dimers with short linkers (diabodies), or chemically crosslinked antibody fragments. Directly related to this are efforts of bringing different binding sites together to create bispecific antibodies. For this purpose, chemically linked fragments, diabodies, scFv-scFv tandems and bispecific miniantibodies have been investigated. Progress in E. coli expression technology makes the amounts necessary for clinical studies now available for suitably engineered fragments. We foresee therapeutic advances from a modular, systematic approach to optimizing pharmacokinetics, stability and functional affinity, which should prove possible with the new recombinant molecular designs.

Antibodies and antibody domains are ideal agents for targeting the surface of cells, and fusion proteins between cell-targeting domains and cytotoxic proteins may be particularly effective therapeutic reagents. We constructed a family of immunofusion proteins linking the humanized Fab, F(ab')2, or single-chain antibody form of the H65 antibody (which recognizes the CD5 antigen on the surface of human T cells) with the plant ribosome-inactivating protein gelonin. To maximize the product yield and simplify the production process, each fusion protein was linked to a bacterial signal sequence for expression in E. coli as a secreted protein. More than 30 fusion genes were assembled with antibody domains and gelonin in various physical orientations. Each immunofusion accumulated in the bacterial culture supernatant in a properly folded, active form. Bacteria transformed with each fusion gene were then grown in a fermentor, and product was recovered from the cell-free fermentation broth by column chromatography. All of the immunofusion proteins were purified by a single process and each was tested for cytotoxicity toward antigen-positive human cells. A compact cGMP fermentation area was built to manufacture these fusion proteins. Our integrated approach to microbial protein production, including molecular genetics, bacterial fermentation, and initial isolation, is described in detail.

Fusion proteins between cell-targeting domains and cytotoxic proteins should be particularly effective therapeutic reagents. We constructed a family of immunofusion proteins linking humanized Fab, F(ab')2, or single chain antibody forms of the H65 antibody (which recognizes the CD5 antigen on the surface of human T cells) with the plant ribosome-inactivating protein gelonin. We reasoned that such an immunofusion would kill human target cells as efficiently as the previously described chemical conjugates of H65 and gelonin (Better M., Bernhard, S. L., Fishwild, D. M., Nolan, P. A., Bauer, R. J., Kung, A. H. C., and Carroll, S. F. (1994) J. Biol. Chem. 269, 9644-9650) if both the recognition and catalytic domains remained active, and a proper linkage between domains could be found. Immunofusion proteins were produced in Escherichia coli as secreted proteins and were recovered directly from the bacterial culture supernatant in an active form. All of the immunofusion proteins were purified by a common process and were tested for cytotoxicity toward antigen-positive human cells. A 20-60-fold range of cytotoxic activity was seen among the fusion family members, and several fusion proteins were identified which are approximately as active as effective chemical conjugates. Based on these constructs, immunofusion avidity and potency can be controlled by appropriate selection of antibody domains and ribosome-inactivating protein.

Two antibody single-chain Fv (scFv) fragments carrying five C-terminal histidine residues were expressed in Escherichia coli as periplasmic inclusion bodies. Their variable heavy (VH) and light (VL) domains are derived from the mouse monoclonal antibody 215 (MAb215), specific for the largest subunit of RNA polymerase II of Drosophila melanogaster and rat MAb Yol1/34, specific for pig brain alpha-tubulin. ScFv-215 contains an additional cysteine residue near to its C-terminus. After solubilization of inclusion bodies followed by immobilized metal affinity chromatography (IMAC) in 6M urea and a renaturation procedure, scFv monomers, noncovalent dimers, and aggregated antibody fragments were separated by size exclusion chromatography. In addition, a fraction of disulfide-bonded scFv-215 homodimers (scFv')2 was also isolated. The various antibody forms appear to be in equilibrium after renaturation since first peak composed mainly of aggregates could be resolved into a similar pattern of aggregates, dimers, and monomers after repeating the denaturation/renaturation procedure. All fractions of the recombinant scFv-215 demonstrated high antigen-binding activity and specificity as shown by enzyme-linked immunosorbent assay (ELISA) and Western blot analysis. Affinity measurements carried out by competitive immunoassays showed that covalently linked (scFv')2 have binding constants quite close to those of the parental MAbs and fourfold higher than scFv' monomers. ScFv derivatives, specifically biotinylated through the free sulfhydryl group, recognize the corresponding antigen in ELISA and Western blot analysis, thus demonstrating the possibility of using chemically modified scFv antibodies for immunodetection.

Ribosomal inactivating proteins such as gelonin (Gel) and ricin A chain (RTA) conjugated to MoAbs bind to specific target cells, and upon internalization inhibit protein synthesis, ultimately resulting in cell death. We report here that Gel anti-T cell MoAb conjugates are more cytotoxic than RTA conjugates when tested against human peripheral blood mononuclear cells (PBMC). This increased cytotoxicity is observed whether Gel is conjugated to the anti-T cell MoAb or to an anti-mouse immunoglobulin Fab' fragment which then binds to the murine anti-human T cell MoAb. Gel conjugates are not only effective at lower concentrations, but also produce a greater extent of inhibition of cellular proliferation. Moreover, a 10 min exposure to a Gel conjugate is as effective as a 90 h exposure to an RTA conjugate. When part of anti-T cell F(ab')2 or Fab' conjugates, Gel affects the early steps in cellular intoxication more than RTA; Gel conjugates bind more avidly and accelerate the modulation of antigen. In contrast, when part of whole IgG conjugates, Gel does not affect the binding to or modulation of surface antigen compared with RTA, while it does increase conjugate cytotoxicity. These observations suggest that Gel may be delivered more efficiently into the cytosol than RTA. A divergent intracellular pathway for Gel is also supported by the inability of chemical potentiators, which strongly enhance RTA potency, to affect Gel potency. These properties of Gel might also be advantageous for immunoconjugates made with other MoAbs or receptor-binding molecules.

We have produced a single plasmid encoding both the heavy chain Fd domain (VH + CH1) of the anti-interleukin-2 receptor (IL2R) monoclonal antibody anti-Tac, and the anti-Tac light chain fused to PE40, a truncated derivative of Pseudomonas exotoxin. The active immunotoxin anti-Tac(Fab)-PE40 could be recovered from E. coli from either periplasm or renatured inclusion bodies. The double-chain immunotoxin was very cytotoxic toward IL2R-bearing cell lines, human activated T cells and fresh adult-T-cell-leukemia cells. The cytotoxicity was similar to that of anti-Tac(Fv)-PE40, the single-chain recombinant toxin containing only the variable domains of anti-Tac. IL2R-binding affinity was also equivalent to that of anti-Tac(Fv)-PE40, which is one-third that of anti-Tac. The serum half-life in mice was significantly prolonged as compared with anti-Tac(Fv)-PE40, with a beta phase of 430 vs. 57 minutes, but the LD50s were equivalent when the immunotoxins were administered in 3 daily doses. Anti-Tac(Fab)-PE40 was very cytotoxic in vitro toward transfected ATAC-4 carcinoma cells which express IL2Rs. In mice bearing ATAC-4 tumors, anti-Tac(Fab)-PE40 showed significant anti-tumor activity, inducing complete remissions in 80 and 100% of treated animals at approximately 7 and 14% respectively of the LD50. Anti-Tac(Fab)-PE40 was much more effective in vitro and in vivo than chemical conjugates between anti-Tac and truncated PE molecules. The recombinant Fab toxin should be studied further as potential treatment for IL2R-related malignancies, particularly if smaller recombinant immunotoxins have insufficient half-life in humans.

A recombinant chimeric Fab (rcFab), with Fv derived from the monoclonal A5B7 antibody to carcinoembryonic antigen (CEA), and with human CHI and C kappa was cloned into pUC 19 and expressed in Escherichia coli. rcFab (10 to 12 mg per litre) was produced in bacterial culture fluid, and functional purified rcFab was isolated by affinity chromatography (using antibody to human C kappa) and size-exclusion gel filtration. The rcFab did not show reduced affinity for CEA, and reacted with human colorectal tumours showing a typical anti-CEA pattern by immunocytochemistry; it was also stable after iodination. Biodistribution studies in nude mice bearing human tumour xenografts showed no toxicity and good tumour localization. Therapeutic ratios at early time points were better than those obtained with whole murine antibody. The results demonstrate that bacterially produced anti-CEA Fab is of use for tumour targeting.