Genomics:GTL

Technologies for Production & Characterization of Proteins and Molecular Tags

Technologies for production and characterization of proteins and molecular tags will provide scientists with an understanding of genome components by using DNA sequence to make and characterize proteins and reagents for interrogating their functions in cells.

Scientific and Technological Rationale

Systems biology requires that we understand the proteins that make up a cell and the mechanisms of their function. Individual proteins encoded in the genome are the basic building blocks for biological functions potentially useful in DOE missions. Virtually every cellular chemical reaction and physical function necessary for sustaining life is controlled and mediated by proteins generally organized into macromolecular complexes or "molecular machines," which might contain proteins, RNAs, or other biomolecules. A typical microbial genome has 2000 to 5000 genes that encode thousands of proteins and regulatory regions that control their expression. The challenge of understanding these workhorse molecules is technically complex and necessitates production and analysis of very large numbers of them. Experimental analysis has determined the functions of only a few thousand of the millions of proteins encoded by the collective genomes on this planet--and even that understanding is incomplete.

We currently have insufficient data and conceptual insights to assign at least one function to about half the proteins found in even the most intensively studied microorganisms. Functional assignments for proteins in unculturable or less-studied organisms often occur by inference from a homologous protein's putative role in an intensively studied organism. A comprehensive understanding of cellular behavior will require experimental data for a significant portion of an organism's proteins. We must have the ability to produce and characterize, as needed, essentially all the thousands of proteins encoded in many single genomes and in metagenomes to support functional gene annotation and, ultimately, mechanistic understanding. We also need to be able to produce and screen numerous variants of individual proteins or molecular machines so they can be used for DOE applications.

Having full-length and active forms of proteins in hand for biochemical and biophysical analysis can serve many purposes critical to the next generation of biology. These proteins provide an opportunity for discovery and a starting point for optimizing complex cellular processes from their components and molecular mechanisms. Providing rigorous and comprehensive characterizations for these proteins is invaluable to researchers and frees them to confidently pursue creative experimentation. "Molecular tags" or "affinity reagents" can be produced only by working from the proteins or via protein modification. These tags are critical for detection and potential quantitation of individual proteins and molecular machines in living systems.

The study of microbes, and especially those of DOE relevance, presents a special challenge. Microbial community systems that we must understand possess millions of genes as opposed to the tens of thousands of even the most-complex higher organisms. The readily available genome sequences and even metagenome sequences of microbial communities have provided our first look into microbes' many functions. Most of the recently sequenced microbial genomes and metagenomes, however, show that roughly 40% of the genes are of unknown function, and, further, the microbes themselves either are not available or are "unculturable." Roughly 200 microbes had been sequenced by 2005, resulting in a catalogue of unknown genes that now contains 200,000 to 400,000 candidates for investigation. The ability to create and gain insight into proteins from genomic information alone is a crucial first step to understanding these microbial systems. Eventual culture-dependent experimentation on an important subset of microbes will be facilitated greatly by the availability of basic information on proteins and their respective affinity reagents.

Protein production currently is limited by economic and technological constraints and is a widely dispersed and inefficient "cottage industry." While substantial technology exists for generating the easy-to-produce (i.e., small, soluble) proteins, the ability to readily produce large multidomain proteins, membrane proteins, proteins with cofactors, and many other critical proteins is only emerging. For comprehensively understanding microbial systems, access to all proteins in metabolic, signaling, and regulatory pathways and networks is important. The most difficult proteins often are the very ones most vital to cellular function (e.g., those associated with essential transmembrane molecular machines, such as the photosystems in a photosynthetic microbe). Protein production and characterization technologies are especially required for these hard-to-produce, but critically important, proteins. The GTL program will enlist the research community to help develop needed methods.

To understand the physical and functional properties of proteins, a substantial suite of high-throughput, automated, and increasingly sophisticated characterization assays must be performed on proteins. Protein production and characterization both will benefit as the transition is made from widely dispersed efforts focused on easy proteins to the economy of scale made possible by developing automated high-throughput technologies capable of producing any desired protein with an accompanying database of reliable characterizations. The situation is somewhat analogous to genomic sequencing as it transitioned from dispersed, somewhat unreliable sequence data to higher-quality, lower-cost data at high-throughput, automated sequencing centers.

Automated high-throughput protein and affinity-reagent production will have several important impacts, including the following, that will enable the expeditious systemic study of chemical and physical interactions of proteins that underlie biology:

• Rigorous production environments will establish the necessary standards, diagnostics, control, and quality to develop and execute the demanding protocols for readily and repeatedly producing difficult proteins.
• Production capabilities will support a comprehensive and sophisticated array of characterization methods, most unavailable to the individual researcher, that can be applied to production diagnostics and protein characterization.
• Large-scale robotics, miniaturization, and automation will greatly enhance throughput and reduce costs.
• Creating a mechanism to pool resources to make material and data products available to all scientists will leverage the investment to reach a larger community, whose work will facilitate further production, characterization, and understanding.
• Unlike the current situation, in which only selected portions of labor-intensive data are accessible, a strong computational and data infrastructure will facilitate mining of both successful and unsuccessful metadata associated with production attempts for each protein.

Value of Proteins for Research

Ready and economic availability of proteins and affinity reagents will provide the foundation for the next generation of biological research, building on the national investment in genome sequencing. Having widespread access to cutting-edge technology in protein production will level the playing field, increasing the availability of proteins and protocols and creating a broader biotech industry. Proteins form the starting point for biochemical and biophysical functional studies, for eventual protein engineering, and for creating chimeric or new (optimized) biochemical pathways or even reactions or pathways that work in reverse directions (e.g., carbon dioxide to formate to methane). They offer the ability to study low-abundance proteins such as important regulatory proteins. Many variants (mutations) can be produced and studied for functional analysis. For nonculturable organisms, proteins can be produced from sequence alone to provide a shortcut to functional genome annotation and allow determination of quantitative biochemical binding or reaction constants. Comparative analyses of the structure and functions of protein families can be used to determine design principles. Proteins are reagents for studying metabolomics, post-translational modifications (substrate identifications), biosynthesis of metabolites and intermediates, binding-partner identification, and affinity-reagent generation. Functional proteins are the starting material for reconstituting molecular complexes, making quantitative and qualitative three-dimensional spectral and structural analyses, and mapping molecular interactions (with DNA, metabolites, and other proteins). They also can serve as mass and spectral standards for enhancement of mass spectrometry (MS) data analysis. Proteins, affinity reagents and other molecular tags, and data are needed to capture molecular machines for MS and other analyses and to identify the machines' components. They also are needed for cellular-imaging studies and verification of models (see tables: Analysis of Technology Options for Protein Production and Roadmap for Development of Technologies to Produce Proteins).

Value of Protein Characterization for Research

Automated high-throughput and high-quality biophysical and biochemical characterizations of proteins will provide a more rigorous assignment of gene function, resulting in first insights to a mechanistic understanding of microbial capabilities. Ultimately, comprehensive reliable data on thousands of proteins should be available to analysts. Researchers can use high-throughput screening to characterize many proteins simultaneously under widely varied, controlled conditions. First-generation analyses will focus on characterizations to determine basic biochemical function and biophysical information (e.g., solubility and insolubility in multiple solutions, multimeric state, presence of metals, and ordered and disordered domains). As technologies mature, the nature and sophistication of these characterizations will expand to determine more complex functions of individual proteins and molecular complexes (see table: Summary of Characterization Needs and Methods).

Value of Molecular Tags for Research

Two types of molecular tags are discussed here: Affinity reagents and fusion tags. Affinity reagents comprise proteins, peptides, nucleic acids, and small chemical molecules that bind targets of interest with high specificity and affinity. They commonly are used to detect where particular proteins are localized in cells, recover the protein and its associated molecules from cell lysates, and quantitate protein amounts in complex mixtures. Antibodies, popularly used as affinity reagents, can be generated by immunizing rodents or rabbits with the protein target and harvesting immunoglobulins (e.g., IgM and IgG) from the serum several months later. With the advent of various in vitro methods termed "display technologies," antibody fragments (i.e., scFv, Fab, VH, VL, and VHH) can be isolated from naïve libraries in several weeks' time without the use of animals. In addition to modifying antibody-based molecules, scientists are altering other proteins (e.g., lipocalin, ankyrin, fibronectin domain, and thioredoxin) to bind to specific targets of interest. This is accomplished by modifying the open reading fragments through mutagenesis and selecting among the resulting library of randomized proteins for those that bind targets specifically. Finally, affinity reagents can be selected from libraries of combinatorial peptides, nucleic acids (i.e., aptamers), or small organic molecules (see tables: Analysis of Technology Options for Affinity Reagent Production, Roadmap for Development of Technologies to Produce Affinity Reagents, and Examples of Affinity Reagents and Their Applications).

The following items focus on affinity reagents:

• Production of affinity reagents must be designed around their many applications. When proteins are in structured environments, some surfaces are exposed while others are hidden because they are in contact with other proteins or molecules. To deal with this contingency, multiple affinity reagents for each protein will ensure that any exposed surface or epitope can be accessed.
• Affinity reagents are needed that either disrupt or preserve protein activity. They can be used to manipulate proteins, including fabrication of biosensors; map post-translational modifications; determine spatial distributions; array targets in a unique spatial configuration; disrupt protein-protein interactions; promote crystallization of proteins; and stabilize membrane proteins.
• Affinity reagents can be used to assess biodiversity and in diagnostic tools for energy-production processes. They are critical for affinity purification of proteins and complexes, for identifying binding surfaces and mapping interactions in protein complexes, and for characterizing functional states (by targeting epitopes unique to active or inactive forms of the proteins). Finally, they are valuable in flow cytometry to sort cells from mixtures and for use in nanotechnology to anchor proteins during fabrication of novel biohybrid materials.

Another type of molecular tags—fusion tags—are short peptides, protein domains, or entire proteins that can be fused at the genetic level to proteins of interest. The target protein then is imparted with the fusion tag's biochemical properties. In general, the type of fusion tag used is dictated by its application. Short peptide tags (e.g., six-histidine, epitopes, StrepTag, calmodulin-binding peptide) regularly serve to permit facile purification of the recombinant protein, allow detection of the fusion protein, or direct the recombinant protein's interaction with other proteins or inert surfaces. Larger fusion partners such as protein domains (e.g., chitin-binding domain) or proteins (e.g., cutinase, GFP, GST, MBP, and intein) usually are employed to promote folding, solubility, purification, labeling, chemical ligation, or immobilization of the recombinant protein. If desired, the fusion tag can be detached from the protein of interest by cleaving a linker region with a site-specific protease that does not affect the protein (see table Examples of Fusion Tags and Their Applications).

Technology Description

The GTL program will bring together comprehensive technologies for high-quality mass production and characterization of proteins produced directly from sequence data or other genetic sources such as gene variants or clones. These technologies also will generate specific capture and labeling affinity reagents for each protein. To derive insights into gene function and assess the best and most cost-effective protein-production strategies, a key capability will be computational comparison of genomic sequences of unknown organisms against the comprehensive GTL Knowledgebase. The program will integrate the basic research and technology development necessary to enable continued advancement of capabilities by working with investigators and technologists in academia, national laboratories, and industry.

Protein and Affinity Reagent Production and Characterization Outputs

Researchers will need the following products:

• Expression vectors (clones) for targeted genes
• Milligram quantities of purified, full-length, functional proteins
• Multiple affinity reagents for each protein, as well as chips with arrayed affinity reagents
• Proteins with a variety of fusion tags
• Initial biophysical and biochemical characterizations of each protein
• Production protocols so researchers and commercial concerns can readily produce proteins for research and biotechnology applications
• Comprehensive production and characterization databases and computational analyses referenced to the subject genome or classes of proteins

Instrumentation and Infrastructure Requirements

The scale and challenge of protein production and characterization require the development of extensive robotics for efficient sample production and processing and suites of highly integrated instruments for sample analysis and characterization of proteins and affinity reagents.

Production will require instrumentation for production of large numbers of different DNA molecules, including cloning and insertion into expression vectors and, eventually, gene synthesis capabilities; production of proteins from any biological source; purification; quality assessment; and production of protein variants [e.g., isotopically labeled proteins, post-translationally modified proteins, proteins with novel cofactors, proteins incorporating nonstandard amino acids, and site-specific mutant arrays (high-throughput mutagenesis)]. Imaging and other experimentation will necessitate production of multiple affinity reagents for each protein; production of membrane proteins and multiprotein complexes; multimodal protein biophysical and biochemical characterization; and combinatorial capabilities to screen for complexes under multiple defined conditions. Methods should include cellular or cell-free expression and chemical synthesis. Onsite DNA sequencing will be required for several steps in any production process. Informatic capabilities must be available to track each gene or clone, protein, affinity reagent, and the associated data. Quality control should include mass spectroscopy (MS) and a range of other biophysical and biochemical analyses.

Automation and computationally based insights are key to achieving high throughput at steadily declining costs, just as they were in DNA sequencing. Over time, as the GTL Knowledgebase matures, the GTL computational infrastructure will enable use of DNA sequence to predict the following for each protein: Efficient and successful production methods, likely binding partners, appropriate assay conditions, and, ultimately, information about the functions of each gene. Achieving this goal will require experience and the data created from production and characterization of tens of thousands of proteins.

Development of Methods for Protein Production

Proteins have wide variability in their structure and stability. No single production method and characterization scheme will be applicable to every protein. Thus, several methods must be developed simultaneously, including all appropriate variations on cell-based, cell-free, and chemical synthesis.

Whichever method is selected, nearly all protein production is based on transcription from DNA obtained via cloning or possibly direct chemical synthesis of the gene encoding the desired protein. In cases where only gene sequence is available, chemical synthesis alone will be required. Working with genome databases, protein production will be facilitated by the development of a sequence-verified library of publicly available protein-coding microbial genes. This library could be available for translation into protein or for use in transformational studies by the other facilities or the larger scientific community.

For the scale of production challenges, technologies should be scalable, economic, and sufficiently robust to work in a high-throughput mode. At least 50% of all proteins are anticipated to pose significant problems for any current method, so development work will be required. Some genes have evolved to generate only very small amounts of protein products. Most proteins are idiosyncratic with respect to conditions; for example, some proteins are not readily soluble or they are relatively unstable and require discovery of special conditions for storage, handling, and use. Others will function only in a properly reconstituted assembly and may need to be produced with their partners under specialized conditions. Consequently, a significant challenge will be research into new methods of protein production. In addition, many DOE-relevant systems may require techniques compatible with anaerobic or other extreme conditions. The strategy for success includes high-throughput parallel processing to allow exploration of a very large number of conditions and protocols specific to each protein.

Improved techniques are needed to predict from genome sequence the production and purification approaches most likely to succeed with each protein. Also needed are methods to identify all DNA sequences in a genome that should encode proteins. Thus, computation and informatics is an integral component. Algorithms based on data from successful and failed protein expressions are expected to improve future protein-production and -characterization efficiencies.

Disorder and the Formation of Molecular Machines. Proteins must be produced in their functional state. Disorder is emerging as an increasingly important factor in protein function, particularly in the assembly of protein partners into molecular machines. This key process very often is mediated by disorder-to-order transitions at the binding interfaces as the disordered regions of two proteins become ordered by their interaction. R&D must be carried out to develop characterization methods that will, among other things, allow their general structure (whether ordered or disordered) to be defined and mapped. Whereas disordered protein regions are a hindrance in crystallization for classic protein crystallography techniques, our goal is to allow protein disorder to become a useful tool to predict binding partners and aspects of protein function.

LIMS. A laboratory information management system (LIMS) will provide for machine learning from failures and successes of all production attempts, the larger experimental and theoretical program, and other similar efforts. Experience-based decision making will allow selection of optimal expression, purification, storage, and characterization routes based on bioinformatics. Identification of domains that do and do not inhibit activity and strategies for affinity reagent production will be revealed. Inventory tracking and provenance records will become more essential as systems biology experimentation becomes more extensive. Development will include better integration of instrument data files for generation of provenance records. For more information on LIMS and other computational and information technologies, see Creating an Integrated Computational Environment for Biology.

Production Targets

The initial numbers of proteins required are large by any current standard and certainly will increase over time with ongoing progress in metagenomic and other analyses and increased sophistication of experimentation and program goals. In addition, each protein probably will require exploration of a wide range of conditions to define successful production and characterization protocols. Several independent factors drive the need:

• Producing encoded proteins and characterizing them in a low-cost and high-throughput mode will make tractable and affordable the exploration of large numbers of unknown genes from sequenced plants and microbes.
• Metagenomics is becoming more important as a methodology for studying natural systems critical to DOE mission environments. These studies are revealing millions of genes with the recurring 40% unknown ratio. Although more-sophisticated computational analyses can reduce the numbers that must be produced for analysis and for uncovering culturing techniques for some discovered microbes, potentially millions of proteins could or should be beneficially investigated through extensive protein production.
• Understanding and eventually optimizing such critical microbial functions as redox processes, cellulose degradation, hydrogen production, and all the ancillary metabolic and regulatory pathways will entail screening potentially thousands of naturally occurring variants of hundreds of protein families. Exploring intentional modifications to understand function and to optimize properties could involve very large multiplicative factors on identified targets. Gene shuffling can involve thousands of modifications.
• Exploring microbial function and incorporating non-natural or isotopically labeled amino acids will be beneficial with or without various fusion tags (e.g., six-His, FlAsH tag, and biotin).
• Engineering microbial systems or biobased cell-free systems for energy or environmental applications will require significant exploration of rationally engineered primary and ancillary proteins, machines, and pathways in a concerted and comprehensive way.
• Providing a source of proteins and their characterizations from gene sequence alone would produce a rapid and cost-effective alternative to historical culturing techniques and an important knowledgebase for possible culturing experiments.

As technologies mature, production will proceed at multiple scales; the first exploratory pass to determine optimum successful production protocols should be at the smallest and most rapidly executable scale, followed by scaleup of interesting ones accordingly. Three examples follow.

• Screening mode: Microgram quantities, semipure, >10⁴ to 10⁵proteins/year
• Macroscale: Milligram quantities, >90% pure, >10⁴/year
• Large scale: Hundreds of milligram quantities, >95% pure, >10²/year

Material and data products must be accompanied by protocols that define optimal parameters for production, activity, storage, and use of proteins. The challenge is to use various technologies in appropriate ways to cover production needs for all proteins, including small soluble proteins, membrane proteins, multiple domain proteins, and multiprotein complexes. Detailed comparisons of these available options will be a key part of creating of production modalities. Table 1 provides a summary of technology options for protein production. Table 2 is a simplified technology development roadmap covering the necessary research, pilot, and production phases of the R&D process. Each technology application has its own set of challenges. For the easy, soluble proteins, the challenge is scaleup, while the more difficult proteins and complexes require exploration of methods to produce and stabilize them. Due to the extreme diversity of proteins being discovered, continued exploration of new techniques for protein production will be needed.

Specifications for Proteins and Comparisons of Their Production Methods

Methods eventually must be capable of cost-effectively producing on demand all the proteins coded in any microbial genome for which we have sequence, including the ability to coexpress proteins and purify or reconstitute protein complexes, difficult proteins such as membrane and multidomain proteins, metalloproteins, and proteins that cannot be overexpressed in host cells. Proteins must be properly folded and active, incorporate correct cofactors and metals, and have correct post-translational modifications. Eventually, optimized versions of proteins should be available on demand, requiring screening of only dozens rather than hundreds or thousands of candidates. Three key methods for protein production and purification are described below.

Comparison of Cell-Based Expression Systems

Large-scale cell-based expression systems have been used worldwide in structural genomics centers and elsewhere, with Escherichia coli as the mainstay system. Yeast and other eukaryotic expression systems have been developed for proteins that fail in E. coli-based systems. Their use is not as readily automated as with cell-free systems. Various alternatives are contrasted and compared in the three paragraphs below.

E. coli. Use of E. coli for protein production is a robust technology (numerous vectors, strains, extant instrumentation infrastructure) that is relatively inexpensive. Bacterial cultures are a renewable resource (from small- to fermenter-sized cultures), and transformants can be stored indefinitely as DNA or frozen cells. Bacterial hosts can be engineered to coexpress certain proteins or chaperones. Shortcomings include scalability (the number of cultures and culture volume required); difficulty in predicting yields and solubility; product subjectability to proteolysis; costly labeling with certain isotopes; possible absence of necessary cofactors or chaperones; and necessarily large freezer storage capacity (and tracking) of transformants. Development needs include miniaturization of cultures for screening and production; improvements in methodologies and strains; and improvements for generating membrane and other difficult-to-produce proteins.

Alternative Hosts. Use of alternative hosts (yeast, Pichia, Aspergillus, insect cell lines) may permit better expression of particular proteins, but they have less-developed vector systems and strains and are more costly than bacterial and cell-free methods. In addition, they have slower growth rates compared to E. coli, codon-usage differences, and possibly missing cofactors or chaperones. These methods require investment in heterologous host systems and improvements for producing membrane and other difficult proteins.

Homologous Hosts. Use of homologous hosts has the advantage that cofactors, accessory proteins, modifying enzymes, and chaperones are present, and codons are optimized for open reading frames. These systems are less developed, however, with uncertain scalability, slow growth rates, low yields, nonexistent or difficult genetics and transformation, and the absence of selectable markers. Furthermore, they are not feasible for proteins from currently unculturable microbes. Development needs include defining optimal growth conditions, development of vectors and transformation protocols, and improvements in producing membrane and other difficult-to-produce proteins.

Cell-Free Systems

Cell-free expression systems, such as those based on wheat germ or E. coli extracts, hold the greatest potential for full automation and hence lower costs and higher throughput. Successful efforts in Japan using these extracts have yielded hundreds to thousands of proteins per year. Having the ability to automate the systems and the potential to incorporate labeled or nonstandard amino acids adds to their value. However, these methods have not yet seen widespread use or application. A broader experience base needs to be established.

Cell-Free Methods. Amenable to robotics (and microtiter plates), cell-free methods can have either small sample-reaction volumes (30-µL reaction volumes, 30-µg yields) or large. Cell-free proteins can be produced from PCR-amplified DNA templates, eliminating extensive cloning steps and simplifying rapid testing of many construct variations, thereby making this an attractive method for high-throughput screening. Produced protein molecules exist in simpler mixtures, sometimes permitting functional assessment without purification. Multiple proteins can be coexpressed to assemble complexes. Cofactors and detergents can be added, and certain isotopes can be cost effectively incorporated. Shortcomings include relatively expensive application, although this is expected to decrease substantially as the method becomes more widely used. Disulfide bonds must form spontaneously when reducing agents are removed. Development needs include advances in directed disulfide bond formation, replacement of cell lysates with recombinant proteins and ribosomes, and improvements in generating membrane and difficult-to-produce proteins.

Chemical Synthesis

Solid-state chemical synthesis is a possible approach for important proteins that fail in all DNA-based expression systems. Currently, this method can produce peptides up to 50 amino acids in length, but longer peptides are made at ever-diminishing efficiencies. Full-length proteins might be synthesized through chemical ligation of multiple peptides. This currently is a costly procedure, and refolding into active protein remains a major problem. The technique has the advantage of producing milligrams of proteins labeled by incorporation of isotopes, chemical modifications, unnatural amino acids, or other chemical groups.

Chemical Synthesis Methods. Requiring no DNA, chemical synthesis can have large yields (>50 mg) for small proteins. There is no contamination by cellular proteins, and incorporating unnatural amino acids, labels, and post-translational modifications is easy. Chemical synthesis currently is not high throughput, and it is labor intensive. It is limited to proteins shorter that 200 amino acids, and the product typically requires refolding. Development needs include cheaper production of thousands of peptides, expansion of peptide ligation sites, reliable refolding, and improvements for generating membrane and difficult-to-produce proteins.

Protein Purification

Protein purification after expression presents a number of challenges, particularly in a high-throughput environment. Substantial reliance is placed on experience-based informatics methods to guide the purification strategy for each protein, with the expectation of achieving significant improvement as the database expands. Automated protocols aimed at eliminating centrifugation should be developed since this step accounts for the major bottleneck in current protein-production protocols.

Purification Methods. Methods based on affinity-purification tags permit generic protocols for purification, but tags can interfere with structure or function and tag removal may be required. Current methods are not high throughput, contaminants may be hard to eliminate, and activity may be lost during purification (i.e., loss of cofactors, denaturation). Development needs include improved instrumentation for high throughput, and the special problems of purifying and storing native membrane proteins should be addressed.

Development of Methods for Protein Characterization

Key to understanding the structure and function of unknown proteins is stabilization and extensive characterization of each produced protein under well-defined conditions.. Given the investment in each expressed protein and its scientific value, investigators should subject each to a substantial suite of assays. Measurements for thousands of proteins will need to be generated robotically under standardized conditions, producing voluminous data. Assays must be rapid and inexpensive, requiring miniscule protein quantities to allow data collection from a broad range of conditions. Technologies such as microfluidics and other lab-on-a-chip methods eventually will provide the required versatility and sensitivity, with attendant sample economies and speed. Some of these protocols should reveal additional functional, structural, biological, chemical, and physical insights. Serving several purposes, characterization first supports production by validating that the right protein has been produced (without sequence or translation errors), that the protein is stable and nominally folded, and that conditions necessary for long-term stabilization and storage have been met. Since no single measurement provides all the answers, suites of techniques will need to be employed as they are feasible and required (see Table 3).

Once we are assured that validated and stable proteins are produced, a more complete set of biophysical and biochemical characterizations can be made as required by the particular research problem and system. Various parameters that might be measured are listed below.

• Screen, identify, and measure enzymatic or binding activity, cofactor state and requirements, and the effect of affinity reagents on proteins (e.g., epitopes, inhibitory or noninhibitory for selected activities)
• Identify agonists and antagonists
• Identify binding partners and determine affinities (dissociation constants) under a suite of conditions, including salts, buffers, pH, temperature, and aerobic or anaerobic
• Identify monomeric or multimeric state
• Identify reconstitution conditions, intermolecular interactions
• Probe the folding landscape, establish structure
• Identify motifs, folding stability, thermodynamics, ordered and disordered regions
• Discover substrates (orphan enzymes)
• Identify cofactors (metals, NADH, ATP, ligands)
• Elucidate biological effect of post-translational modifications
• Identify DNA/RNA binding and sequence motifs

Specific biochemical functions and sensitivities pertinent to DOE applications (e.g., metal reduction, proton or electron transfer, carbon reduction) will be critical. Many of these measurements can be made before the proteins have been purified and thus done in screening mode during the production process. Some measurements could be done with proteins produced to contain sensitive fluorescent probes designed to facilitate inexpensive, high-throughput characterizations with miniscule quantities of protein.

For a set of proteins selected for their unique and mission-relevant properties (e.g., hydrogen and biofuel production, carbon cycling, contaminant immobilization, sensors), the ultimate characterization suite should allow determination of structure at the highest-possible resolution (primary, secondary, tertiary, and quaternary). Approaches could utilize state-of-the-art national synchrotron, neutron, NMR, and electron microscopy facilities and lab-based molecular techniques. These measurements will allow the establishment of structural-activity relations and the understanding of design principles. Computation will be a key part of such analyses.

Meeting DOE mission goals will require refinement and redesign of proteins and affinity reagents for a diverse suite of energy and environmental applications. Researchers must be able to produce and characterize the effects of a wide range of modifications to understand design principles and optimize performance. This includes design of affinity reagents spanning several approaches, not all of which may be proteins or even cellular; protein and molecular-machine redesign or refinement; pathway redesign; and the engineering of biofunctional materials into nanomaterials and devices for energy and environmental applications and research.

As the study of these systems matures, emphasis will shift from supporting production methods to more advanced characterizations that provide finer detail on structure and function and elucidate design principles.

Requirements, Specifications for Functional Characterization Techniques, Data

Methods should be sensitive enough to work with screening-mode levels of proteins where possible and should include cost-effective and high-throughput biochemical and biophysical measurements. Individual measurements should be very inexpensive so they can be repeated under a variety of conditions to reflect salt, pH, buffer concentration, cofactors, ligands, and temperature. They also should have a low coefficient of variation to permit statistical analysis. They should be highly parallelized and scalable and provide QA/QC with feedback to the production process. Computational support will include algorithms for cherry-picking samples for retesting and optimizing activity conditions.

Much of the needed instrumentation is laboratory based but some measurements could benefit from remote instruments like a high-brightness synchrotron or neutron source. For example, at a synchrotron, high-throughput systems (flow or robotic enabled) could be developed and evaluated as a means to provide a cost-effective platform for making certain types of valuable measurements on protein samples [e.g., small-angle X-ray scattering (SAXS) or extended range circular dichroism (or UV-CD)]. Results of such developments could be evaluated for their usefulness in the context of meeting DOE mission goals. To take advantage of such an approach, methods would need to be developed for transporting and automating sample handling, data logging and processing, and comparison of results obtained by these methods. Results would need to be integrated with other laboratory-based measurements.

Development of Approaches for Affinity-Reagent Production

Production of multiple high-affinity, high-specificity affinity reagents and suitable fusion tags for each protein presents enormous challenges (see Molecular Tags: Fusion Tags and Affinity Reagents). Several promising approaches are under development worldwide, although none has yet emerged as an economical and reliable solution to systems biology's high-throughput needs. Overcoming this obstacle is therefore a major target for GTL pilot studies (see Table 4, Table 5, Table 6, and Table 7).

High-throughput systems must be capable of producing numerous affinity reagents that recognize different domains of each protein. This will require multiple new libraries of affinity reagents from which members with desired affinity and specificity to each target protein can be selected. Different and complementary approaches are under development, including phage and yeast display systems and aptamers. When full proteins cannot be produced, these tags might be created for appropriate epitopes that can be determined by computational analyses. In addition, computational insights eventually might recommend the best affinity reagent approach for particular proteins. These techniques will require substantial development.

Further developmental areas include improved reagent stability and specificity; improved multiplex screening protocols; and rapid, high-throughput affinity-maturation techniques. Reagents also will be evaluated to determine where they bind to their protein targets and whether they disrupt the target's function, thereby dictating how different affinity reagents can be used. Development of modular affinity reagents also would be extremely useful; selected binding domains could be generated rapidly for such different purposes as protein isolation or live-cell imaging.

In many cases, the most useful affinity reagents may be proteins themselves. They can be produced and characterized using technologies already developed for bacterial proteins. They will be standardized reagents, however, so processes can be developed to allow for their rapid and large-scale production.

Specifications for Affinity Reagents and Their Production

Affinity reagent production technologies must be rapid, cost-effective, and amenable to high-throughput automation; they should be capable of being based on antibody fragments, engineered protein scaffolds, combinatorial peptides, and aptamers as the need dictates. They should work with targets that have reduced cysteines or are cell toxic. A computationally based decision process is needed for selecting proteins or epitopes of proteins to serve as targets for affinity-reagent generation. Affinity reagents should bind either individual proteins or complexes, and the collection should recognize three to five different epitopes on a protein and be amenable to epitope subtraction and existing target-detection strategies. The process should identify reagents best suited for particular applications (i.e., Western blot, pulldown, coimmunoprecipitation, staining, complex disruption, inhibited catalytic activity, and inhibited protein-protein interactions).

Affinity reagents should bind their target with modest to high affinity, have lowest-possible failure rate (cross-reactivity, low affinity), be obtainable in reasonable amounts (5 mg, >90% pure) in a cost-effective manner, and be stable and storable. They should be formattable on chips with excellent shelf life and available in fluorescent, biotinylated, or enzyme-linked forms; and formattable for affinity chromatographic methods to purify individual proteins or protein complexes from cells. Ideally, they should be expressible inside cells where they can bind their target and be made conditional or regulatable.

Just as for proteins, no single method will work equally well for producing all affinity reagents, so several methods will be needed. Operationally, methods must be capable of generating reagents from small target amounts (tens of micrograms). They must readily screen diverse libraries with targets and select out the best binders applicable under a variety of conditions; have the capability to screen libraries of more than 109 members in a rapid manner for hundreds of targets per day; validate binding to specific target protein; and be amenable to affinity maturation.

Material and data products must be accompanied by protocols that define optimal parameters for production, activity, storage, and use. The challenge is to use various technologies in appropriate ways, including phage display, yeast display, ribosome and puromycin display, DNA or RNA aptamers, and immunization of animals. Table 4 provides a summary of technology options for production of affinity reagents.

Table 5 is a simplified technology development roadmap covering the necessary research, pilot, and production phases of the R&D process. Each technology application has its own set of challenges, so continued exploration of new techniques will be needed.

Technologies for Affinity-Reagent Production

Phage Display. This technology can use libraries of combinatorial peptides, antibody fragments, and engineered protein scaffolds. Phage display is amenable to high-throughput screening with robotics; it is protein based, so functionality is added easily by creating fusion proteins with different functional domains; and it has been used for in vivo and subtractive selections. The resulting output, however, may have to go through a second round of evolution as it tends to isolate weak and strong binders at the same time. In addition, candidates should be sorted according to differences in affinity, specificity, epitope overlap, stability, storage, and application, and the output may be misleading about the strength of binding due to multivalent display. The technology may require different scaffolds, depending on the application. Development needs include the optimization of scaffolds and screening methodologies.

Yeast Display. Capable of using libraries of combinatorial peptides, antibody fragments, and engineered protein scaffolds, the yeast display technology can discriminate affinities by flow cytometry, permitting fast assessment and identifying downstream candidates. Good for directed-evolution experiments (enhanced affinity, specificity, expression, or stability) and for epitope identification, yeast display is protein based, so functionality can be added easily by creating fusion proteins with different functional domains. It may need to go through a second round of evolution, however, and its libraries tend to be less diverse than other display formats. Candidates may require sorting by affinity, specificity, epitope overlap, stability, storage, and application. Yeast grow slower than phage, taking more time and effort and needing larger volumes per screening cycle, so making this technology high throughput is more difficult. Yeast display requires different scaffolds, depending on the application. Development needs include optimization of scaffolds and screening methodologies.

Ribosome and Puromycin Display. These methods can work with very large libraries (i.e., 10¹² members), and monovalent display leads to selection of the best binders. Ribosome- and puromycin-display technologies can incorporate mutagenesis during screening and enhance binding during the general selection process. They are protein based, so functionality can be added easily by creating fusion proteins with different functional domains. They are more expensive than phage- and yeast-display technologies, however, and large libraries require more rounds of screening. Candidates need to be sorted by affinity, specificity, epitope overlap, stability, storage, and application; they require different scaffolds, depending on the application. Development needs include optimization of scaffolds and screening methodologies and automation.

DNA or RNA Aptamers. Use of DNA or RNA aptamers is amenable to very large libraries (i.e., 10¹² members) and high-throughput screening with robotics. Synthesizing large amounts of individual aptamers is relatively expensive, however, and large libraries require more rounds of screening than phage or yeast libraries. Aptamer candidates should be sorted by affinity, specificity, epitope overlap, and application, and they are limited to DNA/RNA. Development needs include optimization of screening methodologies.

Immunization of Animals. This traditional, well-established approach requires animals and large amounts of antigen. Repeated injections are necessary, so it is slow. This is a nonrenewable resource unless hybridomas are generated, so the method is expensive; it is limited by the immune response because common epitopes cannot be subtracted. Development needs include DNA immunization and improvements in hybridoma production (see Table 4 for strengths and weaknesses and development roadmap).

Development of Data Management and Computation Capabilities

Each step and process in protein production and characterization and subsequent affinity reagent production will involve very large numbers of biological samples that need to be tracked appropriately through the automated systems. Sophisticated bioinformatic analysis will be greatly needed at all steps so insights can be gained from both successes and failures. Processes will generate vast amounts of valuable data on clones and proteins and their characterization. These and other data should be captured properly and disseminated to the scientific user community. Implementation of appropriate LIMS and data-mining capabilities will be absolutely crucial to achieving high-throughput, cost-effective clone and protein production as well as to enable the use of these materials in contributing to the goals of GTL and the Department of Energy. These criteria will require large computing resources and development of the best scientific tools to properly mine the invaluable data being produced. For more details, see Table 8. Computing Roadmap.

Workflow Process

Conceptual diagrams depict prospective major equipment layout, process flow, and production targets. The process begins with genomics, which includes comparative genomic analyses against the GTL Knowledgebase to (1) gain insight into an unknown genome and identify its protein production targets and (2) produce clones or synthesized genes. Protein production first is pursued in a high-throughput, low-volume screening mode using appropriate microtechnologies, followed by full-scale production with successful protocols and robotics. Characterization is carried out for QA/QC, for initial biophysical and biochemical analyses, and for in-depth studies as needed. With applicable technologies, affinity reagents to selected proteins are produced using pipelines very similar to those for protein production. Computing and information technologies will support and inform all processes and provide protocols, supporting data, and characterizations to the scientific community. Capabilities will include data and sample archives and distribution.

This Webpage adapted from Genomics:GTL Roadmap, DOE/SC-0090, October 2005. See References PDF.