Protein production has evolved from an artisan activity to more efficient, specialized processes, with new tools helping out along the way.
Imagine the possibilities for scientific discovery if progress were not hindered by the availability of any particular protein reagent. Think of the advantages a researcher would have with the immediate access to any protein in the human genome. The outcome of an experiment would no longer be dependent on one’s ability to express or purify a key protein reagent.
Until fairly recently, protein production has been regarded as an artisan activity performed by a highly experienced scientist capable of taking the job from gene to protein. In the search for greater efficiencies in the production of proteins, the process has fragmented into a number of highly specialized activities. The publication of the human genome and the accompanied increase in demand for recombinant proteins also had the effect of accelerating this division of labor. Probably the biggest transformation in the optimization of protein production is the degree to which parallelization and automation have been incorporated into the process.
Many of the innovations responsible represent both incremental and radical changes in the way proteins are produced. These changes have drawn on both newer technologies and the adoption of more established methods from other disciplines that have been combined to increase the efficiency with which proteins can be made.
Many of the new methodologies now being deployed in protein production, although well-established in other disciplines, have been slow to gain general acceptance. One reason for this resistance is the high degree of skepticism that skilled scientists can be replaced with robots and automated processes. Traditionalists would argue that biological processes are too variable and proteins are too diverse to be subjected to automation. The key to producing a reproducible process requires an understanding of the many factors that can affect the outcome. Once this has been determined, the variation in any operation can be controlled and minimized. This discipline has been achieved through design of experiments (DOE) in mission-critical industries, which despite its origins in genetics, has not, until recently, been utilized in the production of proteins. One barrier to the uptake of DOE has been removed with the availability of new software packages, like JMP (SAS Institute, Cary, N.C.), providing sophisticated data exploration tools to the non-statistician for optimization of protein production.
Traditional methods for optimization would involve the systematic examination of the effect of one factor at a time on the output. This is an extremely time-consuming and expensive process, especially where a large number of factors are involved, and can often result in the wrong conclusion being drawn from the experiment. Now using a high-throughput expression and protein purification platform capable of exploring the effect of multiple factors—in conjunction with DOE—a robust model for protein production and optimization can be established.
Small is beautiful
click to enlarge
Just as miniaturization and microfluidics have transformed the throughput of both enzymology and crystallography (thus allowing automation and the exploration of a greater number of experimental variables), so too is protein expression and purification succumbing to the same changes. Miniaturization of cell growth has been established for all of the most common expression systems. Culture volumes have been significantly reduced, allowing a much larger experiment to be performed in a microtiter format than could be accomplished previously in shaker flasks.
Nowhere is the demand for protein optimization greater than in the creation of protein therapeutics and the need to rapidly develop an effective cell culture-based manufacturing process. To address this issue, BioProcessors Corp., Woburn, Mass., developed the SimCell system for culturing and monitoring mammalian cells.1 The SimCell employs both advanced microfluidics and optical detection in a microbioreactor array consisting of six independent chambers (Figure 1). Each 700 µl volume chamber is flanked with a gas-permeable membrane, thus eliminating the need for spargers or stirrers. The culture conditions inside each chamber can be independently-defined, monitored and controlled for multiple process variables. The SimCell system provides a high throughput platform for mammalian cell growth, in which DOE can be performed using hundreds of different cultures run simultaneously in multiple microbioreactor arrays. Data generated by the SimCell system allows for the development of an effective manufacturing process, as well as providing the in-depth understanding of the process used in the production of protein therapeutics increasingly required by the regulatory authorities.
GE Healthcare’s PreDictor plates can be used to explore a large amount of experimental space in the optimization of protein purification. These filter plates are available in a range of varieties, pre-filled with the different chromatographic media commonly used in purification of proteins and monoclonal antibodies. The results obtained in the low volume PreDictor plates have shown good correlation on scale-up with data obtained by column chromatography.2
No such thing as qualitative data
The new methods for small-scale, high throughput, recombinant protein production combine novel cloning, expression, and purification systems in automated, multi-well configurations. However, current methods of analysis of expressed recombinant proteins are complex and time-consuming, limiting the rate of at which proteins can be assessed.
Figure 2. Microfluidic chip for quantitative analysis of proteins on the LabChip90. The “sipper” allows automatic sampling of proteins from microtiter plates. (Source: Caliper Life Sciences)
The amount of protein, particularly when recombinant, is more often described in qualitative than quantitative terms. Proteins are normally separated and visualized following sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE), enabling an estimation of amount and purity. Another approach to quantification involves a dot blot procedure employing specific antibodies to detect protein bound to a nitrocellulose filter. Although the process can be performed at high throughput, its results can be inaccurate and are open to misinterpretation. For this reason, protein yields are frequently described subjectively as estimates or percentages. This over-reliance on qualitative measurement reflects the difficulties encountered in accurately determining protein quantity. The problem is only aggravated when working in a high throughput production environment where protein analysis and characterization cannot keep pace with the number of samples. This bottleneck in generating quantitative data has now been effectively removed with the LabChip90 protein assay system developed by Caliper Life Sciences Inc., Hopkinton, Mass.3 The protein assay uses a microfluidic chip, which has a “sipper” to allow access to protein samples prepared in microtiter plates (Figure 2). Nanoliter volumes of protein are used for the analysis. The LabChip90 can process over 300 samples and is capable of generating a vast amount of quantitative data on protein yield within minutes.
The quantitative data produced from a high throughput expression experiment can, in turn, be used to rapidly optimize protein production. Optimization of protein expression and purification, a process which could take weeks or months, can now be accomplished within days.4
Another development in the quantitative analysis of protein production is the use of surface plasmon resonance (SPR) biosensor-based assays to screen conditions in expression and purification. David Myszka and coworkers at University of Utah, Salt Lake City, have pioneered many of the most interesting SPR applications in recent years, including a novel high throughput application to screen solubilizing conditions for membrane-associated chemokine receptors.5 The biosensor assay from Biacore, GE Healthcare, is automated, which allows the exploration of maximum experimental space. The assay, at least with the receptors examined, has the ability to quantify the total amount receptor and determine what fraction is active. The implication for the production and purification of functional membrane proteins, one of the most challenging protein families for protein production, is tantalizing.
Supplying the protein content
The changes these innovations are bringing about are somewhat predictable, having been previously observed in the evolution of many other manufacturing production processes. Instead of improving the tools used at each step, it eventually becomes more efficient to mechanize the entire process using a fundamentally different type of tool, the assembly line. This is the challenge AbPro, Cambridge, Mass., has set for itself: to provide the first industrial gene-to-protein platform. By exploiting many of the innovations now impacting protein production, the goal is to provide ready access to the proteome. Many of the innovations described here are already making an impact with the availability of many more proteins; the challenge ahead will be to transform these assets with scientific understanding. Lack of protein content has so often been cited as the main reason for the slow development of the human proteome; it is assumed that its availability will herald a new age of discovery. With the protein expression and purification bottleneck potentially removed, the next obstruction will likely be revealed.
About the Author
Stephen P Chambers, PhD, has 20 years of experience producing recombinant proteins for both the pharmaceutical and biotechnology industries. He has authored papers for over 30 peer-review publications.
Seth P Cohen, PhD, has been involved in the development and application of automation and microfluidics technologies supporting diagnostics, drug discovery, and drug development for more than two decades.
This article was published in Drug Discovery & Development magazine: Vol. 11, No. 8, August, 2008, pp. 31-33
Filed Under: Genomics/Proteomics