Therapeutic anti-VEGF antibodies.

Vascular endothelial growth factor (VEGF-A) is a key cytokine in the development of normal blood vessels as well as the development of vessels in tumors and other tissues undergoing abnormal angiogenesis. Here, we review the molecular engineering of two humanized antibodies derived from a common mouse anti-VEGF antibody--bevacizumab, a full-length IgG1 approved for the treatment of specified cancer indications, and ranibizumab, an affinity-matured antibody Fab domain approved for use in age-related macular degeneration (AMD). In clinical trials and as FDA-approved therapeutics, these two anti-VEGF antibodies, bevacizumab (Avastin anti-VEGF antibody) and ranibizumab (Lucentis anti-VEGF antibody), have demonstrated therapeutic utility in blocking VEGF-induced angiogenesis.

VEGF and the Fab fragment of a humanized neutralizing antibody: crystal structure of the complex at 2.4 A resolution and mutational analysis of the interface.

BACKGROUND: Vascular endothelial growth factor (VEGF) is a highly specific angiogenic growth factor; anti-angiogenic treatment through inhibition of receptor activation by VEGF might have important therapeutic applications in diseases such as diabetic retinopathy and cancer. A neutralizing anti-VEGF antibody shown to suppress tumor growth in an in vivo murine model has been used as the basis for production of a humanized version. RESULTS: We present the crystal structure of the complex between VEGF and the Fab fragment of this humanized antibody, as well as a comprehensive alanine-scanning analysis of the contact residues on both sides of the interface. Although the VEGF residues critical for antibody binding are distinct from those important for high-affinity receptor binding, they occupy a common region on VEGF, demonstrating that the neutralizing effect of antibody binding results from steric blocking of VEGF-receptor interactions. Of the residues buried in the VEGF-Fab interface, only a small number are critical for high-affinity binding; the essential VEGF residues interact with those of the Fab fragment, generating a remarkable functional complementarity at the interface. CONCLUSIONS: Our findings suggest that the character of antigen-antibody interfaces is similar to that of other protein-protein interfaces, such as ligand-receptor interactions; in the case of VEGF, the principal difference is that the residues essential for binding to the Fab fragment are concentrated in one continuous segment of polypeptide chain, whereas those essential for binding to the receptor are distributed over four different segments and span across the dimer interface.

Affinity maturation of human growth hormone by monovalent phage display.

We describe a selection procedure for construction of very high affinity variants of human growth hormone (hGH) for binding to the extra cellular domain of its receptor (called the hGHbp). Five different libraries of mutated hGH genes (each containing approximately 2 x 10(5) protein variants) were created by randomly mutating four different codons at residues that were shown to be important for receptor binding by structural or functional criteria. Mutated proteins were displayed as single copies from their respective filamentous phagemid particles and sorted in vitro for binding to the immobilized hGHbp. Phagemid particles that bound the immobilized hGHbp were eluted and propagated. After three to seven rounds of binding enrichments, hGH variants were isolated that contained 2 to 4 mutations and exhibited three- to sixfold improvements in binding affinity. Because of the limits of DNA transfection efficiency in creating the library we could not sample thoroughly mutations at more than four codons at once. Nonetheless, the free energy effects for these mutations acted cumulatively. Thus, by combining affinity enhanced mutants from libraries independently sorted we created an hGH variant with 15 substitutions that bound approximately 400-fold more tightly to the hGHbp than wild-type hGH. The affinity enhancements occurred predominantly by slowing the off-rate of the hormone (> 60-fold), and partly through increasing the on-rate (up to 4-fold). Residues that were shown to be important for binding by alanine-scanning were most highly conserved after binding selection, and interestingly many of them could be further improved. Thus, we found it most effective to randomly mutate the residues that were shown to modulate affinity by alanine-scanning, and to combine the selectants from separate libraries that exhibit the highest affinities. The selection procedure and mutagenesis strategy provides a framework for affinity maturation of protein-protein complexes.

Selection and analysis of an optimized anti-VEGF antibody: crystal structure of an affinity-matured Fab in complex with antigen.

The Fab portion of a humanized antibody (Fab-12; IgG form known as rhuMAb VEGF) to vascular endothelial growth factor (VEGF) has been affinity-matured through complementarity-determining region (CDR) mutation, followed by affinity selection using monovalent phage display. After stringent binding selections at 37 degrees C, with dissociation (off-rate) selection periods of several days, high affinity variants were isolated from CDR-H1, H2, and H3 libraries. Mutations were combined to obtain cumulatively tighter-binding variants. The final variant identified here, Y0317, contained six mutations from the parental antibody. In vitro cell-based assays show that four mutations yielded an improvement of about 100-fold in potency for inhibition of VEGF-dependent cell proliferation by this variant, consistent with the equilibrium binding constant determined from kinetics experiments at 37 degrees C. Using X-ray crystallography, we determined a high-resolution structure of the complex between VEGF and the affinity-matured Fab fragment. The overall features of the binding interface seen previously with wild-type are preserved, and many contact residues are maintained in precise alignment in the superimposed structures. However, locally, we see evidence for improved contacts between antibody and antigen, and two mutations result in increased van der Waals contact and improved hydrogen bonding. Site-directed mutants confirm that the most favorable improvements as judged by examination of the complex structure, in fact, have the greatest impact on free energy of binding. In general, the final antibody has improved affinity for several VEGF variants as compared with the parental antibody; however, some contact residues on VEGF differ in their contribution to the energetics of Fab binding. The results show that small changes even in a large protein-protein binding interface can have significant effects on the energetics of interaction.

Anticalins versus antibodies: made-to-order binding proteins for small molecules.

Engineering proteins to bind small molecules presents a challenge as daunting as drug discovery, for both hinge upon our understanding of receptor-ligand molecular recognition. However, powerful techniques from combinatorial molecular biology can be used to rapidly select artificial receptors. While traditionally researchers have relied upon antibody technologies as a source of new binding proteins, the lipocalin scaffold has recently emerged as an adaptable receptor for small molecule binding. 'Anticalins', engineered lipocalin variants, offer some advantages over traditional antibody technology and illuminate features of molecular recognition between receptors and small molecule ligands.

Rapid evolution of peptide and protein binding properties in vitro.

A significant bottleneck in protein engineering arises from the problem of identifying particular molecules with new functions from a potentially enormous range of peptide or protein variants. Two areas of emerging technology, phage display and multiple peptide synthesis, provide new means of screening huge libraries in vitro for novel binding properties. This review is also published in Current Opinion in Structural Biology 1992, 2:597-604.

Monomeric variants of IL-8: effects of side chain substitutions and solution conditions upon dimer formation.

IL-8 dimers have been observed in NMR and X-ray structures of the protein. We have engineered IL-8 monomers by mutations of residues throughout the dimer interface, which introduce hindrance determinants to dimerization. These IL-8 variants are shown by NMR to have wild-type monomer folding, but by ultracentrifugation to have a range of dimerization constants from microM to mM, as compared with a dimerization constant of about 10 microM for wild-type IL-8, under physiological salt and temperature conditions. The monomeric variants of IL-8 bind the erythrocyte chemokine receptor DARC, as well as the neutrophil IL-8 receptors CXCR1 and CXCR2 with affinities similar to that of wild-type IL-8. In addition, the monomeric variants were shown to have agonist activity, with similar potency to wild-type, in both Ca(2+)-flux assays on CXCR1 and CXCR2 transfected cells, and in chemotaxis assays on neutrophils. Thus, these variants confirm that monomeric IL-8 is functionally equivalent to wild-type in vitro assays. We have also investigated the effects of various solution conditions upon IL-8 dimer formation using analytical ultracentrifugation. At salt concentrations, temperatures, and pH conditions lower than physiological, the dimerization affinity of IL-8 is greatly enhanced. This suggests that, under some conditions, IL-8 dimer formation may occur at concentrations of IL-8 considerably lower than 10 microM, with consequences in vivo that are yet to be determined.

IL-8 single-chain homodimers and heterodimers: interactions with chemokine receptors CXCR1, CXCR2, and DARC.

Covalent single-chain dimers of the chemokine interleukin-8 (IL-8) have been designed to mimic the dimeric form of IL-8 in solution and facilitate the production of heterodimer variants of IL-8. Physical studies indicated that use of a simple peptide linker to join two subunits, while allowing receptor binding and activation, led to self-association of the tethered dimers. However, addition of a single disulfide crosslink between the tethered subunits prevented this multimer from forming, yielding a species of dimer molecular weight. Crosslinked single-chain dimers bind to both IL-8 neutrophil receptors CXCR1 and CXCR2 as well as to DARC, as does a double disulfide-linked dimer with no peptide linker. In addition, neutrophil response to these dimers as measured by chemotaxis or beta-glucuronidase release is similar to that elicited by wild-type IL-8, providing evidence that the dissociation of the dimeric species is not required for these biologically relevant activities. Finally, through construction of single-chain heterodimer mutants, we show that only the first subunit's ELR motif is the single-chain variants.

Binding protein-3-selective insulin-like growth factor I variants: engineering, biodistributions, and clearance.

Insulin-like growth factor I (IGF-I) is a potent anabolic peptide that mediates most of its pleiotropic effects through association with the IGF type I receptor. Biological availability and plasma half-life of IGF-I are modulated by soluble binding proteins (IGFBPs), which sequester free IGF-I into high affinity complexes. Elevated levels of specific IGFBPs have been observed in several pathological conditions, resulting in inhibition of IGF-I activity. Administration of IGF-I variants that are unable to bind to the up-regulated IGFBP species could potentially counteract this effect. We engineered two IGFBP-selective variants that demonstrated 700- and 80,000-fold apparent reductions in affinity for IGFBP-1 while preserving low nanomolar affinity for IGFBP-3, the major carrier of IGF-I in plasma. Both variants displayed wild-type-like potency in cellular receptor kinase assays, stimulated human cartilage matrix synthesis, and retained their ability to associate with the acid-labile subunit in complex with IGFBP-3. Furthermore, pharmacokinetic parameters and tissue distribution of the IGF-I variants in rats differed from those of wild-type IGF-I as a function of their IGFBP affinities. These IGF-I variants may potentially be useful for treating disease conditions associated with up-regulated IGFBP-1 levels, such as chronic or acute renal and hepatic failure or uncontrolled diabetes. More generally, these results suggest that the complex biology of IGF-I may be clarified through in vivo studies of IGFBP-selective variants.

The stoichiometry of growth hormone-binding protein complexes in human plasma: comparison with cell surface receptors.

The recent demonstration of two independent receptor-binding sites (sites 1 and 2) on human GH (hGH) raises the question of the stoichiometry of circulating GH-binding protein (GH-BP) complexes in human plasma (i.e. is it one hGH per one GHBP or one hGH per two GHBPs?). Previous studies have all assumed 1:1 binding in plasma, based on gel exclusion chromatography and cross-linking data. To address this issue, human plasma was incubated with radioiodinated hGH as well as hGH mutants that had either a Tyr103-->Ala or a Gly120-->Arg substitution in the region of binding site 2. The former mutant retains normal site 2 binding activity even when iodinated; the latter has binding site 2 inactivated. Bound and free hGH were then separated on a Sephadex G-100 column according to a standard protocol for measuring GHBP. In all three cases, more than 90% of the high affinity GH-BP complex eluting from the column was consistent with 1:1 binding. Similar results were obtained when a physiological amount of recombinant or purified natural GHBP was substituted for plasma. However, at supraphysiological concentrations of GHBP, an additional component corresponding to the 2:1 complex eluted from the column; the relative proportions of the 2:1 and 1:1 complexes were dependent on the GHBP concentration. These data suggest that at physiological GHBP levels in plasma, the 1:1 complex predominates, and that small amounts of the 2:1 complex may be difficult to detect because of partial peak overlap with the 1:1 complex, dissociation, and, in whole plasma, coelution with the low affinity GHBP complex. Calculation of the theoretical partition of hGH between 1:1 and 2:1 complexes indicated that at concentrations of GHBP prevailing in plasma (approximately 1 nmol/L), the 1:1 complex predominates, but that at the high receptor concentrations prevailing at the cell surface (60 nmol/L to 6.7 mumol/L, depending on the cell type), virtually all hGH is captured in a 2:1 complex. These findings are consistent with the present and previous experimental data on the size of the circulating high affinity GH-BP complex, as well as with those indicating the importance of GH-induced receptor dimerization for GH action. A functional consequence of the large concentration difference between GHBP in plasma and GH receptors at the cell surface is that the circulating GHBP can serve as a dynamic buffer, modulating bound and free GH and prolonging its half-life, whereas the receptor acts as a dominant force in unidirectional capture of GH.(ABSTRACT TRUNCATED AT 400 WORDS)

Temperature-mediated regulation and downstream inducible selection for controlling gene expression from the bacteriophage lambda pL promoter.

We have examined in detail the effects of various induction temperatures on the expression of a heterologous fusion gene controlled by the bacteriophage lambda PL promoter in a heat-inducible Escherichia coli expression system which utilizes the CIts857 repressor. Experiments performed over a temperature range spanning 29-42 degrees C indicate that, under our conditions, temperatures as low as 29 degrees C may be required to fully repress the CI857-controlled transcription from PL, and that the highest protein yields are obtained after induction at 36 degrees C for 6 h. We cloned the cat reporter gene downstream from a heterologous gene controlled by PL and found that cat expression at a low induction temperature permits the monitoring of productive transcription through the heterologous gene and thus aids in selecting transformants that are capable of producing the heterologous protein in E. coli.

High-level expression of the simian virus 40 genes LP1, VP1 and VP2 as fusion proteins in Escherichia coli.

The complete sequences of the SV40 agnogene (LP1) and the genes coding for the capsid proteins VP1 and VP2 have been cloned into Escherichia coli expression plasmids. High levels of expression were obtained when the SV40 genes were inserted into the coding sequence of the influenza virus NS1 gene, which has previously been expressed in E. coli. The NS1A-LP1 and NS1A-VP2 chimeric proteins consist of the 81 N-terminal residues of NS1 (designated as peptide NS1A) fused to the complete sequence of the corresponding SV40 protein. The NS1A-VP1 chimera consists of NS1A followed by a linker of nine arbitrary residues and the complete sequence of the SV40 major capsid protein. The observed levels of expression vary considerably among the three chimeric proteins, ranging from approx. 70 micrograms/ml in the case of NS1A-LP1 to approx. 5 micrograms/ml in the case of NS1A-VP2. Cyanogen bromide cleavage of the NS1A-LP1 fusion protein produces fragments with Mrs expected for isolated NS1A and LP1 peptides. A plasmid has also been constructed which expresses the NS1A peptide in high yield.

Bacteriophage display and discovery of peptide leads for drug development.

Phage display makes large-peptide diversity libraries readily attainable for identifying novel peptide ligands for receptors and other protein or non-protein targets. This technology kindles enthusiasm for the idea that large and protein-protein interaction surfaces (epitopes) can be distilled down to small pharmacophores. These may be accessible to organic scaffolding, yielding new orally active drugs that might otherwise have taken greater time and effort to be discovered through chemical-library screening. This review, though not comprehensive with respect to the explosive volume of phage display work over the last few years, focuses on recent developments in phage-displayed peptide technology.

On the recognition of helical RNA by cobra venom V1 nuclease.

The V1 nuclease from cobra venom preferentially hydrolyzes double helical RNA and has been used extensively for detecting RNA secondary structure. To increase the utility of this enzyme as an RNA structure probe, we have investigated its properties and substrate specificity, using assays for polynucleotide hydrolysis based on fluorescent polynucleotide derivatives. Enzymatic activity requires both Na+ and Mg2+, with optima at 100 and 0.3 mM, respectively. From the sharp decrease in enzyme activity above 100 mM Na+ we estimate that 3-4 ionic interactions between the protein and polynucleotide phosphates take place. Analysis of products remaining after extensive V1 digestion also shows that the minimum size substrate is 4-6 nucleotides long. Helical RNAs and DNAs have Michaelis constants a factor of 3-10 times lower than most single-stranded RNAs. However, poly(epsilon A) has a Michaelis constant equal to the best synthetic double helices tested and is hydrolyzed at a rate comparable to helical RNA. The major V1 cutting sites in yeast tRNAPhe have Michaelis constants lower than any synthetic polymers. These data suggest that V1 nuclease recognizes any 4-6-nucleotide segment of polynucleotide backbone with an approximately helical conformation, but does not require that the bases be paired in a helix. A few single-stranded V1 cleavage sites are known in tRNA and rRNA, and their structures are consistent with the suggested V1 recognition site.

Total alanine-scanning mutagenesis of insulin-like growth factor I (IGF-I) identifies differential binding epitopes for IGFBP-1 and IGFBP-3.

The bioavailability of insulin-like growth factor I (IGF-I) in the serum and tissues is controlled by members of the IGF binding protein family (IGFBP). These proteins form high-affinity complexes with IGF-I and thereby either inhibit or potentiate its mitogenic and metabolic effects. Thus, understanding the IGF-IGFBP interaction at the molecular level is crucial for attempts to modulate IGF-I activity in vivo. We have systematically investigated the binding contribution of each IGF-I amino acid side chain toward IGFBP-1 and IGFBP-3, combining alanine-scanning mutagenesis and monovalent phage display. Surprisingly, most IGF-I residues could be substituted by alanines, resulting in less than 5-fold affinity losses for IGFBP-3. In contrast, binding of IGFBP-1 was more sensitive to alanine substitutions in IGF-I. The glutamate and phenylalanine at positions 3 and 49 were identified as major specificity determinants for IGFBP-1: the corresponding alanine mutations, E3A and F49A, selectively decreased IGFBP-1 binding by 34- and 100-fold, whereas IGFBP-3 affinity was not affected or reduced maximally 4-fold. No side chain specificity determinant was found for IGFBP-3. Instead, our results suggest that the N-terminal backbone region of IGF-I is important for binding to IGFBP-3. The fact that the functional binding epitopes on IGF-I are overlapping but distinct for both binding proteins may be exploited to design binding protein-specific IGF variants.

Mutational analysis and protein engineering of receptor-binding determinants in human placental lactogen.

Human placental lactogen (hPL) shares 85% sequence identity to human growth hormone (hGH) yet has some very different receptor-binding properties. For example, hPL binds 2300-fold weaker than hGH to the hGH receptor, yet these two hormones have similar affinities for prolactin receptors. We have expressed hPL in Escherichia coli, and we show that, like hGH, hPL requires zinc for tight binding to the extracellular domain of the human prolactin receptor (hPRLbp). In fact, hPL contains virtually the same receptor-binding determinants and zinc ligands (His-18, His-21, and Glu-174) that hGH uses for coordinating zinc in the hGH.hPRLbp complex. As with hGH, mutation of Glu-174 to Ala in hPL reduces the affinity for the hPRLbp by 1400-fold. We can increase the affinity of hPL by over 200-fold for the hGHbp by installing four hGH receptor determinants that are not conserved in hPL. By simultaneously introducing E174A, we produced a pentamutant whose binding affinity for the hGHbp is only 1.6-fold weaker than hGH, but whose binding affinity for the hPRLbp is weaker by greater than 1000-fold relative to wild-type hPL. Thus, we have identified an hPRLbp epitope in hPL, "recruited" an hGHbp epitope into hPL, and produced receptor selective analogs of hPL that are designed to bind tightly to either, neither, or both receptors. Such variants should be important molecular probes to link specific receptor-binding, activation, and biological events.

A novel family of hairpin peptides that inhibit IgE activity by binding to the high-affinity IgE receptor.

A family of structured peptides that bind to FcepsilonRIalpha, the alpha-chain of the high-affinity receptor for IgE, has been identified. Binding selections using FcepsilonRIalpha and polyvalent peptide-phage libraries yielded a dominant 18-residue peptide-phage clone, as well as related sequences that did not resemble fragments of IgE. Synthetic peptides based on these sequences inhibited IgE binding to its receptor, and were found by NMR analysis to adopt a stable beta-hairpin structure in solution. Optimized peptides with micromolar receptor affinity exhibited high stability in biological fluids and inhibited cellular histamine release in an in vitro bioassay of IgE activity. The structure-activity relationships of these peptides, which are less than 1% of the size of IgE, suggest an overlap between their binding site and that of IgE on FcepsilonRI. Thus, the peptides demonstrate that blocking a small epitope on this receptor chain is sufficient to block IgE activity. Such structured peptides represent a possible starting point for the design of novel antagonists, and offer the potential for testing in vivo a new approach for treating allergic disease.