Generating human proteome antibodies via phage display and directed evolution

There are four research areas in the Department of Pharmaceutical Chemistry. Generating human proteome antibodies via phage display and directed evolution is a research challenge within protein and cellular engineering.

The challenge

Discovering, labeling, monitoring, and modulating the myriad proteins in human cells, including:

  • Specific post-translationally modified proteins
  • Enzymes in active and inactive conformations
  • DNA-binding transcription factors and their co-factors
  • Membrane proteins, partly to stabilize them for structural studies

One major approach taps the adaptive ability of antibodies, the Y-shaped proteins generated by immune system B cells that bind to virtually any foreign substance (antigen). The goal is to rationally design and/or select (via competitive assays called directed evolution) antibodies that selectively bind to human proteins of interest even in complex mixtures. These can be used for detecting the presence of their respective proteins in cells or tissues as biomarkers for disease, to modulate function in cells, or even as therapeutics.

Traditionally, monoclonal antibodies have been generated in vivo, by immunizing lab animals with the antigen of interest and then isolating and further engineering the B cells that produce them (fusing them with tumor cells to make them immortal). However, this process is expensive and can take several months. More importantly, one cannot control the state of the antigen (e.g., target proteins may be partially degraded by the host animal) nor precisely tune the specificity of the antibodies, and they are not necessarily renewable (e.g. due to genetic drift as cells replicate).

Phage display methods obviate these problems in that selections are done in vitro, allowing automation which increases the speed and reduces the cost. Importantly, the process allows for precise control of antigen composition and state, and generates the cloned and expressed antibody genes, making them fully renewable.

Examples of our research, methods, and facilities include

UCSF Antibiome Center

The department is home to the UCSF Antibiome Center, a high-throughput robotic platform capable of generating recombinant antibodies of superior quality in a fraction of the time of traditional techniques by combining phage display with automated selection. The Center is one of three NIH-funded laboratories in North America comprising the Recombinant Antibody Network.

The Center’s initial charge is to provide the wider research community with high quality, open source, renewable antibodies selective for the approximately 1,500 human transcription factors—proteins that bind to DNA and regulate gene expression. They are to be used, in particular, for chromatin immunoprecipitation (ChIP) and immunofluorescence studies to determine where those proteins bind and which co-factors they recruit to produce their specific effects.

machine that automates processes

The UCSF Antibiome Center uses automation to increase speed and reduce the cost of recombinant antibody selection.

There are plans to expand this effort to include all cell surface and secreted proteins. A longer-term goal is produce these reagents for the entire human proteome, including for specific post-translational modifications, with potential uses for research, diagnostic assays, and therapeutic agents.

Phage display

Antibody phage display is a technique by which the genes that encode the part of an antibody that binds antigens are inserted into the DNA of a bacteriophage (also known as a phage, a virus that replicates itself inside bacteria) or, at the Center here, a phagemid (a hybrid of phage and a plasmid, a DNA molecule that replicates inside bacteria).

These genes then express fragment antigen-binding (Fab) antibody fragments. These Fab proteins are displayed on the ends of the bacteriophage particles released by the bacteria. (With some additional engineering, Fabs can be made into full-size antibodies.)

bacteriophage particles and recombinant Fab proteins
Schematic diagram of bacteriophage particles (ovals) with inserted genes, displaying different recombinant Fab proteins (colored shapes), which are then tested for binding against a protein target. Effective binders are then amplified via phagemid replication in bacterial hosts for further rounds of competitive selection to find those that bind with the highest affinity and specificity.

In order to engineer antibodies that bind to proteins of interest selectively (high specificity) and tightly (high affinity), scientists here focus on the antigen binding sites at the tips of the Y-shaped immune proteins. Those sites, in turn, have six complementarity determining regions (CDRs) that precisely complement a given antigen’s shape. By substituting different combinations of amino acids at key locations in the CDRs, as well as altering the lengths of their loop-shaped structures, researchers generate recombinant libraries comprised of more than 10 billion different Fab molecules.

CDR loops, amino acid positions

Researchers create synthetic antibody libraries to bind specific natural human protein antigens with high affinity by substituting combinations of amino acids in their complementarity determining regions. In this example, CDR loops are colored and the amino acid positions that were diversified are depicted as numbered spheres.

Automated selection / directed evolution

To select a recombinant antibody for a given protein, the Center’s researchers use an automated robotic platform to fish the best-binding Fab from that collection (library) of 10 billion-plus. The protein targets are labeled with biotin allowing them to be captured on streptavidin-coated paramagnetic beads (biotin and streptavidin form a strong non-covalent bond).  Specialized robots transfer the immobilized proteins-on-beads into the Fab library mixture, allowing the specific antibody molecules to bind to them and be recovered.

During the initial stages of selection there is substantial non-specific binding to the protein, its co-factors, and tags. But iterative rounds of selection for binding to the antigen winnow the Fabs down to only those that bind the target with high affinity and specificity. The Center uses a combination of automated and manual validation assays to determine affinity and specificity of each of the Fab reagents prior to release for the general scientific community.

Our applications of such protein engineering include

Selecting recombinant antibodies for cancer-related proteases

Generating antibodies via phage display that selectively bind to the active forms of certain cell membrane-tethered proteases (matriptase, prostate specific antigen, urokinase-type plasminogen activator) that are dysregulated during various stages of cancer progression. Ongoing research is using immunofluorescence combined with the development of high-throughput multiplex microarrays to monitor the proteases’ activity for the diagnosis and staging of the disease as well as for evaluation of treatment efficacy.

Developing a strategy to make antibodies for phosphorylated peptides

More than 500 kinase enzymes add phosphate groups to up to 30 percent of all eukaryotic proteins, altering their function. This ubiquitous regulatory modification goes awry in many diseases, including cancers and neurodegenerative disorders. Antibodies are needed to detect these specific modifications in cells and tissues to facilitate understanding of signaling and as potential biomarkers.

Researchers here built on their knowledge of how proteins recognize and bind anions, like negatively charged phosphates. They sought out the most commonly employed structural motif—a binding pocket including three consecutive residues where multiple backbone amides form hydrogen bonds with anions—among a database of mAb structures.  They found something similar in a mouse Fab and used it as a scaffold to build humanized phage display libraries.

Structures of antibodies’ phosphorylated-residue binding pockets

Structures of antibodies’ phosphorylated-residue binding pockets recognizing interactions with target modified peptides. Dotted black and yellow lines indicate hydrogen bonds: the top two are phospho-serine; the lower left is phospho-threonine; the lower right is phosphor-tyrosine.

First, they optimized a single Fab CDR’s loop structure (randomizing residues to complete the binding pocket, altering loop length to enlarge it) in order to bind the most common phosphorylated residues – serine, threonine, and tyrosine. Then, using these new Fab scaffolds, they altered other mAb regions (two additional CDRs) generating billions of different mAbs. Via rounds of competitive assays (directed evolution), researchers selected for mAbs that bind phosphorylated residues in context, e.g. a specific site on a protein. These selections led to the identification of 51 phospho-specific monoclonal antibodies.

Department scientists are now seeking to apply this approach to generating renewable antibodies to other common post-translational modifications.

Engineering conformation-selective antibodies

Virtually all enzymes are regulated in cells through conformational switching from on- to off-states. Antibodies not only recognize specific proteins, but also the specific conformational state of that protein. Department scientists have shown it is possible to deliberately isolate such conformationally selective antibodies by first locking the protein in their respective on- or off-states using known or discovered small molecule ligands that trap these states, or by using forms of the protein that are locked in these states naturally. They can then select antibodies for the protein state of interest and against the opposite state using in vitro phage display. This has allowed researchers here to identify antibodies that specifically activate or inhibit proteins by allosteric means.

In one study, this strategy was demonstrated with a protease that plays a key pro-inflammation role in the innate immune response (caspase-1) as a model target. Department scientists used allosteric and active site inhibitors to generate on- and off-state antigens. The resulting antibodies for those states were used to determine where the enzyme is active in cells.

caspase-1 enzyme shown in its on- and off-forms

Researchers use covalent probes to trap the caspase-1 enzyme into its on- and off-forms. The latter were used as antigen “bait” to fish for conformation-specific Fabs generated by phage display.