ScaleUp of Intein Processes

Although the advantages of intein-based bioseparations are well established at the laboratory scale, only a few very recent reports support their use at large scale (Sharma et al. 2003; Ma and Cooney 2004). Furthermore, the economic viability of inteins in the commercial production of any real therapeutic or commodity protein has yet to be proven directly. However, the practical issues associated with this type of process can be partially assessed through a theoretical analysis with a few basic assumptions. This is typically accomplished by comparing a hypothetical intein process to a well-characterized conventional process for a given product. This analysis can be further simplified through the use of several readily available engineering software tools.

Industrial Applications of Intein Technology 2.1 Conventional Affinity Tag Processes

The development of conventional affinity tags in the mid-1980s provided a breakthrough method for the purification of arbitrary proteins at the research scale (Germino and Bastia 1984; LaVallie and McCoy 1995). With this method, purification of a given target protein is facilitated by the addition of an affinity tag sequence to the target protein gene at the genetic level. The expressed tag binds tightly and specifically to a corresponding immobilized ligand, thus facilitating the purification of its fusion partner through pre-optimized affinity purification protocols. Tags include the maltose- and chitin-binding domains, glutathione S-transferase, polyhistidine, and others, and research in this area continues to deliver new tags with new capabilities. The reliability of this method has made it particularly attractive in the investigation of newly discovered gene products, where it eliminates the need for new purification protocols to be developed for each new target. Furthermore, at the several milligram scale, the convenience of these gene fusions far outweighs the cost of subsequent tag removal.

In principle, conventional affinity tags might be used in the large-scale manufacture of highly purified proteins. Unfortunately, however, the cost associated with removal of the affinity tag is clearly prohibitive. Based on product literature (pMal fusion system, New England Biolabs, Beverly, MA, USA), the proteolytic cleaving of a given affinity tag can cost anywhere from $10,000 (using genenase) to $320,000 (using enterokinase) per gram of product delivered. At manufacturing scales of tens to hundreds of kilograms per year, these costs rapidly exceed the gross annual sales of even the most lucrative blockbuster drugs today. Finally, the addition of protease necessitates an additional purification step, and can complicate drug approval due to the highly bio active nature of these enzymes.

2.2 Intein-Mediated Protein Purification

Modified inteins can effectively render affinity tags self-cleaving in response to temperature, pH, or thiol addition, thus eliminating the requirement for protease treatment in the recovery of native products from tagged fusion proteins. This has led to the development of several intein-mediated purification processes at the laboratory-scale for the recovery of recombinant proteins on a single column (Chong et al. 1997,1998; Southworth et al. 1999; Wood et al. 1999; Xu et al. 2000; Zhang et al. 2001). The simplicity by which the cleavage reaction can be induced suggests that this technique may be appropriate for large-scale bioseparations. Furthermore, it has been demonstrated in at least one case that this method is feasible for proteins expressed under high cell-density conditions (Sharma et al. 2003), and a self-cleaving tag has also been successfully incorporated into a pilot-scale vortex-flow affinity capture scheme (Ma and Cooney 2004). Thus, the potential for economic feasibility of a large-scale intein-mediated purification system for recombinant proteins does exist.

2.2.1 Modeling Large-Scale lutein Bioseparations

In order to evaluate the intein-mediated system for scale-up, the laboratory-scale process must be considered from an economic prospective. Bioprocess simulators can aid this process by facilitating the economic assessment of a laboratory process at an industrial scale (Rouf et al. 2001; Shanklin et al. 2001; Petrides et al. 2002). To compare the economics of a conventional bioseparations process to one including an intein-mediated step based on the IMPACT system (New England BioLabs), a series of simulations were performed using the commercial software package SuperPro Designer (Intelligen, Inc., Scotch Plains, NJ, USA). The IMPACT system includes a modified Saccha-romyces cerevisiae VMA1 intein, coupled to a chitin-binding affinity tag. Following purification of the fusion precursor protein, cleavage by the intein is induced through the addition of high concentrations of thiol compounds such as dithiothreitol (DTT). For the purposes of this study, recombinant al-antit-rypsin produced in E. coli was used as the hypothetical product protein around which each process was designed. The key difference between the two processes is the replacement of an immobilized-antibody affinity-purification step in the conventional process with an IMPACT-purification step in the intein process (Fig. 1). Both processes were simulated from seed cultures to freeze-dried bulk protein at a scale of 9000 kg/year. This level of production would be anticipated for an industrial enzyme or high-dose chronic therapeutic, such as al-antitrypsin (de Serres 2002). By comparison, the current production level for human insulin is approximately 2000 kg/year (Ladisch 2001).

Several process assumptions were included for accurate analysis. For example, it was assumed that the quantities of recombinant protein expressed by the cells in both cases were equal and accounted for 10% of the total dry cell weight of the cell. The design parameters for the equipment in both processes, such as percent rejection in the filtration units of common components, were also assumed to be similar. For the ion-exchange, gel-filtration and affinity resins, conservative separation efficiency, reusability and cost estimates were used. The binding capacity and replacement frequency of the chitin resin were set according to values obtained from laboratory studies (5 mg/mL and 5 cycles, respectively). Cleavage efficiency of the fusion protein was set to the maximum value observed experimentally.

CONVENTIONAL

CONVENTIONAL

Fig. 1. Block diagram for the production of a recombinant protein. Conventional and intein processes are shown. FR Fermentation; MF microfiltration; HG homogenization; DS disk stack centrifugation; DEF dead-end filtration; UF ultrafiltration; INX ion-exchange chromatography; GFC gel-filtration chromatography; DF diafiltration; AFC affinity chromatography; FDR freeze drying; INT intein purification step

Fig. 1. Block diagram for the production of a recombinant protein. Conventional and intein processes are shown. FR Fermentation; MF microfiltration; HG homogenization; DS disk stack centrifugation; DEF dead-end filtration; UF ultrafiltration; INX ion-exchange chromatography; GFC gel-filtration chromatography; DF diafiltration; AFC affinity chromatography; FDR freeze drying; INT intein purification step

The basic economic rules for this type of comparison are well established and are summarized here. Raw material and consumable costs were estimated from current supplier list prices of the largest quantities available, while capital costs were determined based on factors commonly employed in bio-processing plants (Harrison et al. 2003). The production cost ($/g) was calculated based on the annual revenue ($/year) required to obtain a uniform series disbursement for the total capital investment, with a 20% internal rate of return (IRR) over a 15-year period. Calculations included inflation and depreciation values based on a 2003 year of analysis. The plant construction period was 2.5 years, with a startup time of 6 months and a 75% operating capacity during the first year. Finally, the simulation for each process was performed in "design mode", thus allowing the SuperPro software to automatically size each process unit as required.

2.2.2 Economics of the IMPACT Process Scale-Up

Based on the stated assumptions, the overall product cost for the conventional process was calculated to be $38/g. However, incorporation of the scaled-up IMPACT step raised the cost to $87/g. Despite the fact that this is orders of magnitude cheaper than any conventional affinity-tag process, it is still economically unrealistic compared to most industrial large-scale protein-purification processes. For comparison, the total manufacturing costs estimated for Monsanto's bovine growth hormone are less than $5/g (Swartz 2001).

A detailed analysis of each process indicated that the operating costs were the major factor for the cost difference, with raw materials accounting for the most substantial increase in the cost of the intein process (Fig. 2A). Further analysis confirmed that the purification section of the intein process was the primary reason this process was not economically feasible, and the costs associated with DTT addition and the Tris-HCl buffer system were identified as major contributors to the price increase (Fig. 2B). The most expensive raw material on a mass basis was DTT, which accounted for 29% of all raw material costs for the intein process. The second most expensive raw material on a mass basis was Tris-HCl.

2.3 Economic Optimization of Intein-Based Bioseparations

Although the initial analysis points to DTT and Tris-HCl as the major contributors to the high cost of the intein process, these costs can be significantly inflated by limitations in other areas of the purification. For example, if the chitin-affinity resin has a low binding capacity, then the required scale for that

Conventional

Raw Materials

Labor-Dependent 1.4%

Utilities and Waste Disposal 0.4%

Raw Materials

Labor-Dependent 1.4%

Equipment-Dependent 3.0%

Utilities and Waste Disposal 0.4%

LaboratoryiQ/QA 0.2%

Fig. 2. Cost breakdowns for the conventional and intein processes, a Annual operating cost breakdown. Total annual operating expenditures were $332 and $704 million for the conventional and intein processes, respectively.

Consumables 70%

Equipment-Dependent 3.0%

LaboratoryiQ/QA 0.2%

Consumables 2.5%

Equipment-Dependent ■ 3.5%

Consumables 70%

Utilities and Waste Disposal 0.50%

Intein i \

Dependent

0 6% X fi :: :: :: r:: :: :: :: :: :: :::•:: :: s ^ - - - - ■

Fig. 2. Cost breakdowns for the conventional and intein processes, a Annual operating cost breakdown. Total annual operating expenditures were $332 and $704 million for the conventional and intein processes, respectively.

Raw Materials

stage of the purification will increase. This leads to a "domino effect", where buffer consumption, equipment size and overall plant design can be greatly influenced. A full evaluation of the intein process must therefore include an impact analysis for hypothetical changes in several of its aspects. From this analysis, specific suggestions can be generated for further development and optimization, and realistic expectations for the long-term viability of this technology can be determined. For the intein process, the areas that should be examined have been grouped into three major areas: buffer composition, resin, and alternate intein-cleaving modes.

2.3.1 Buffers

The recommended buffer solutions for use with the IMPACT system at laboratory scale typically include HEPES or Tris-HCl as buffering agents, ED-TA to chelate unwanted metals, and high concentrations of NaCl. The costs of the Tris-HCl and DTT (to induce cleavage) contribute most significant-

b Annual raw material cost breakdown. Total annual raw material costs were $83 and $653 million, respectively. WFI Water for injection

b Annual raw material cost breakdown. Total annual raw material costs were $83 and $653 million, respectively. WFI Water for injection ly to the cost of the purified protein. The DTT cost alone is projected to be over $18/g of purified protein for a 100% efficient process. Therefore, the sensitivity to the price of Tris-HCl and DTT was determined for both the in-tein and conventional processes (Fig. 3A). The simulations indicate that the intein process is more sensitive to the price of all the buffer components than the conventional process, primarily due to the large volumes of buffer required for the intein washing and cleaving step. In addition, the intein process is more sensitive to the cost of Tris-HCl than DTT, again due to the large quantities of Tris-HCl required. A simple extrapolation of the data shown in Fig. 3A suggests that the intein process would become substantially more economical than the conventional process if the overall buffer costs can be cut by more than 70%. This would require inexpensive buffering components, such as phosphate, and possible reductions in EDTA and NaCl. Furthermore, the elimination of DTT will likely be necessary, suggesting the use of alternate cleaving agents or pH-inducible cleaving inteins.

Industrial Applications of Intein Technology 2.3.2 Resins

The two most common laboratory resins used with intein-mediated purification systems are amylose and chitin. The reported binding capacity for these resins is on the order of 2-3 mg of fusion protein per milliliter column volume. This is significantly lower than most commercial affinity resins, which have a binding capacity between 15-20 mg protein per milliliter column volume. However, the cost of the chitin resin is relatively inexpensive compared to most conventional affinity resins, with a small-scale retail cost of approximately $1900/1. The amylose resin is more expensive with a small-scale retail cost of approximately $3750/1. By comparison, traditional affinity chromatography resins can cost up to $6000/1 or more, whereas simpler ion exchange resins usually cost $300/1 (Harrison et al. 2003). These comparisons suggest several approaches for improving the economics of the intein process. In particular, increases in the binding capacity of the chitin and amylose resins were examined, as well as decreases in the size of the affinity tag and intein.

It is expected that the development of a high-capacity chitin resin might substantially increase its cost, and therefore the effects on product cost of resin cost and capacity were simulated over a range of values (Fig. 3B). These simulations indicate that an increase in the binding capacity would result in a significant decrease in the product cost, primarily due to a decrease in the required volume of the IMPACT step. Further, product cost was relatively insensitive to resin cost, even for a tenfold increase. Resin reusability was examined and found to not be a significant factor on the product cost for the more inexpensive chitin resins (data not shown). These results suggest that further research into resins with higher binding capacities would be of great economic benefit to the intein process, even if the resin cost increased significantly.

A significant difference between the simulations for the two processes was that the intein process required significantly more biomass to achieve the same production capacity. While the conventional process required only the expression of target protein, the intein process required the expression of the affinity-tagged intein-fusion protein. The cells were assumed to have a fixed capacity to express recombinant protein, although exact values for a given product protein would have to be determined experimentally. Some proteins have been observed to have higher target protein expression levels when fused to an intein-CDB tag (Chong et al. 1998), while inclusion of an N-termi-nal maltose-binding domain tag is also generally thought to improve expression and solubility (Kapust and Waugh 1999). To determine the effect of target protein size on product cost, the process was simulated with various ratios of product protein size to intein affinity-tag size. In each case, the quantity of recombinant fusion protein was maintained at 10% of the dry cell weight. The simulation results clearly indicate that the product cost is very sensitive to the target protein size (Fig. 3C), and that the use of smaller intein-affinity tags can substantially decrease the overall product cost.

2.3.3 Alternate Intein-Cleaving Modes

The development of inteins that are induced to cleave by a pH and/or temperature shift has the potential to eliminate the need for DTT in intein-based a m)

m 150

b cu

Binding Capacity (mgfmL)

S 300 O

0 0.2 0.4 0.6 0.8 1 1.2 1.4 Target to Intein-Affinity Tag Mass Ratio m 150

% of Base Case Material Cost

Binding Capacity (mgfmL)

0 0.2 0.4 0.6 0.8 1 1.2 1.4 Target to Intein-Affinity Tag Mass Ratio

Fig. 3. Cost analysis for the conventional and intein processes, a Effect of raw material costs on product cost. The buffer cost was reduced by lowering both Tris-HCl and DTT costs, simultaneously. For the conventional process, only the effect on cost due to Tris-HCl was examined, b Effect of chitin resin binding capacity on product cost for the intein process. c Effect of target protein size on product cost for the intein process. The relative size of the target protein compared to the intein-affinity tag was examined to account for various intein-affinity tag sizes as well as various recombinant protein sizes bioseparations (Southworth et al. 1999; Wood et al. 1999). To examine the process economics associated with these inteins, simulations were conducted based on pH-induced cleaving in a phosphate buffer system, with a resin cost of $100/1 and binding capacity of 15 mg/ml. Not surprisingly, the removal of DTT from the process resulted in significant improvements in cost. The overall product cost decreased to $13.2/g compared to the base-case intein process cost of $87/g. An increase in the cost of the chitin resin to $1000/ 1 only increased the product cost by $2.8/g, underlining again the importance of resin capacity and buffer price over resin cost. As part of this series of simulations, the incubation time for the cleavage reaction was varied from 1 to 14h. Surprisingly, however, this did not significantly affect the overall product cost (data not shown), perhaps because the intein-cleaving reaction only represents a small fraction of the overall process time from fermentation to delivery of purified bulk product.

2.4 Economics of an Optimal Large-Scale Intein Process

Based on the results of these sensitivity analyses, several recommendations can be made to economically optimize the scale-up of intein processes. For the basic IMPACT process, these include use of a cheaper phosphate buffer system, increasing the capacity of the affinity resin to 15 mg/ml, decreasing the chitin resin cost to $100/1, and increasing the resin reusability to 100 cycles. The overall product cost for this improved intein process was $21/g compared to the conventional process at $38/g and the base-case intein process at $87/g. The raw material costs would still account for 84% of the annual operating costs. However, the annual operating costs were reduced from $704 to $159 million. DTT and phosphate buffer costs were still the two largest raw material costs at 50 and 18%, respectively, compared to 29 and 61% for the Tris-HCl base case.

In addition to the changes proposed above, the intein process can be modified to include pH-inducible cleaving. In this case, the DTT is eliminated completely from the process, resulting in an additional 50% savings in raw materials cost, and a final product cost of $13.2/g. The purification stage of the resulting process is the cheapest of all the intein processes, and is substantially cheaper than the conventional affinity separation process. However, the other stages of the process are still somewhat more expensive than the conventional process due to the additional biomass capacity required to produce the tagged protein (Fig. 4).

Fig. 4. Capital distribution for the production of a recombinant protein by conventional or intein-mediated processes. The intein pH-in-duced case is also based on the use of phosphate buffer

2.5 Additional Considerations for Intein Process Scale-Up

A major strength of the intein-based purification process is that it retains the flexibility of conventional affinity-tag technology. This aspect is likely to motivate the development and use of smaller affinity tags with strong affinities for inexpensive resins. Smaller inteins have also been shown to improve process efficiency due to decreases in the overall affinity-tag size, and this effect will be accentuated in cases where more expensive eukaryotic expression systems are required. Although structural studies indicate that conventional inteins will never be smaller than approximately 140 amino acids, a precise understanding of the structural and functional aspects of the cleaving reaction may allow much smaller pseudo-inteins to be developed. Ironically, however, the development of high-capacity resins may be more important to the commercial adoption of large-scale intein processes than any future improvements in the inteins themselves.

The combination of strong affinity-binding of the tagged target protein with an irreversible on-column cleaving reaction distinguishes the intein process from any conventional chromatography or adsorption processes. These unique features suggest several opportunities for a variety of optimized large-scale process configurations. In particular, expanded bed adsorption and vortex flow adsorption (Ma and Cooney 2004) have been proposed. These methods minimize the pressure drop found in traditional chromatographic columns, improve contact between the affinity tag and the resin, and are well suited to the use of inteins. In addition, the precise controllability of pH-triggered inteins has allowed the development of predictive models for process time, product peak shape, and product concentration in conventional chromatographic column process configurations (Wood et al. 2000).

A final aspect of inteins that will be critical to their adoption is the availability of highly controllable inteins that can be used in a wide range of expression hosts. Ideally, these inteins will be strongly repressed during protein production and purification, but will be easily induced to cleave through a pH shift or the addition of a cheap and biologically benign small molecule. Although it was not clearly examined here, premature cleaving has a dramatic negative effect on process efficiency. Prematurely cleaved tags occupy sites on the affinity resin, thus decreasing the direct yield of uncleaved precursor and the apparent capacity of the resin itself. This is a significant problem that has not been solved, particularly in the development of pH-inducible inteins. More importantly, no small-molecule-inducible cleaving inteins have yet been developed.

With the development of new inteins and cheaper affinity chemistries, the economic attractiveness of inteins for large-scale applications will be undeniable. However, some obstacles will likely remain. The pharmaceutical industry is notoriously conservative in the adoption of new technologies. Validation of intein processes will require proof that the product protein will have no uncleaved precursors, and that the cleaved tags can be easily and completely removed from the product stream. This represents a new set of challenges relative to the conventional purification of untagged proteins. In the commodity enzyme industry, purity is often not a major concern, and extremely cheap large-scale methods are generally employed to produce acceptable products. In this case, highly optimized intein processes may become economical due to their capability to significantly increase purity at a relatively low cost.

Delicious Diabetic Recipes

Delicious Diabetic Recipes

This brilliant guide will teach you how to cook all those delicious recipes for people who have diabetes.

Get My Free Ebook


Post a comment