The promise of highly efficient conversion processes coupled to a "green" technology is now universally appealing to industry and governmental policymakers. In general, hydrolytic enzymes offer depolymerization of naturally occurring polymers in high yield, with few, if any, by-product disposal problems, unlike acid-based hydrolysis processes. More than a decade ago, advances in the production of fungal cellulase preparations rekindled interest in enzyme-based biomass conversion processes. However, there is some evidence that this technology has now reached its zenith. Also, cellulase biochemistry has reached an enabling phase of development, in which combined efforts in biochemistry and molecular biology may be able to deliver improved cellulase systems for industrial application. Many factors that govern cellulase component action on crystalline cellulose (cellulase synergism, crystallographic structures, cellulase active site structure/fiinction relationships) have been established (or at least preliminarily elucidated), so design of engineered recombinant cellulase systems by means of advanced concepts, such as enzyme component selection and site-directed mutagenesis (SDM), is now at hand. To be commercially viable, engineered cellulase systems must ultimately produce highly active cellulases with improved specific activities, protein yield, and/or production cost relative to current submerged culture fimgal preparations. Once these goals are achieved, the ability to "tune" recombinant systems should permit access to new process feedstocks and markets.

Cellulosic Biomass and Cellulase Action. Biomass feedstocks most commonly considered for conversion to bioethanol in the near term are waste wood, agricultural

wastes, and the paper fraction of municipal solid waste. The fermentable fractions of these feedstocks include cellulose (p-l,4-linked glucose) and hemicellulose. Although it is an abundant biopolymer, cellulose is unique because it is highly crystalline, water insoluble, and highly resistant to depolymerization. Three features of the cellulose in pretreated biomass make it extremely resistant to enzymatic hydrolysis.

First, the p-l,4-glycosidic linkage renders cellulose oligomers (cellodextrins) with a chain length in excess of six glucose residues extremely insoluble in aqueous systems. Cellulose chain lengths in plant cell walls are very much longer, and therefore, completely insoluble in water.

Second, in plant tissues, cellulose is organized by extensive hydrogen bonding and hydrophobic interaction into semicrystalline bundles of parallel cellulose chains, called microfibrils. This structural feature contributes to the rigidity and strength exhibited by wood and simultaneously protects most of the glycosidic bonds from attack by cellulases.

Third, acid pretreatment alters the chemistry of biomass to yield a substrate totally unlike materials encountered in nature. Specifically, the acidic cooking of biomass at temperatures above the phase transition of lignin (about 160°C) leads to redistribution of lignin during cool-down, perhaps with the effect of "coating" residual polysaccharide fibers. It is reasonable to assume that this redistribution of lignin inpedes enzyme action through strong, nonproductive binding. This hypothesis follows, because several studies have shown a strong tendency for lignin in wood to interfere with cellulase action (58-59). Pretreated hardwood pulp also harbors a weak net negatively charged surface not associated with native wood, because regions of hemicellulose adjacent to 4-<9-Me — glucuronate branches are protected during dilute acid hydrolysis and probably remain in the fiber {60-61) and glucuronate content in cellulose increases as a function of normal wood oxidation (62). The result of the alteration of wood surface chemistry following pretreatment most certainly alters interactions between cellulases and the biomass surface.

As a consequence of these properties, converting cellulose to glucose requires substantially larger ratios of enzyme to substrate than for otherwise similar processes which convert starch to glucose. Despite this fact, it is remarkable that an amount of cellulose approximately equivalent to that synthesized during the yearly growing season is recycled on a global scale during the same period of time in the environment.

Cellulose is enzymatically degraded to glucose by the synergistic action of three distinct classes of enzymes: the "endo-l,4-p-glucanases" or 1,4-p-D-glucan 4-glucanohydrolases (EC 3.2.1.4), which act randomly on soluble and insoluble 1,4- p — glucan substrates, the "exo-l,4-p-D-glucanases," including both the 1,4-p-D-glucan glucohydrolases (EC 3.2.1.74), which liberate D-glucose from 1,4-p-D-glucans and hydrolyze D-cellobiose slowly, and 1,4-p-D-glucan cellobiohydrolase (EC 3.2.1.91), which liberates D-cellobiose from 1,4-p-glucans, and the "p-D-glucosidases" or p-D — glucoside glucohydrolases (EC 3.2.1.21), which act to release D-glucose units from cellobiose and soluble cellodextrins, as well as an array of glycosides. The concepts of exo-endo and exo-exo synergism are shown diagrammatically in Figure 1.

Synergism between exo — and endoglucanases is best explained in terms of providing new sites of attack for the exoglucanases. These enzymes normally find available chain ends at the reducing and nonreducing termini of cellulose microfibrils. Each random internal cleavage of surface cellulose chains by an endoglucanase provides two additional sites for attack by cellobiohydrolases. Therefore, each hydrolytic event by an endoglucanase yields both a new reducing and a new nonreducing chain end. Thus, logical consideration of catalyst efficiency dictates the presence of exoglucanases specific for reducing and nonreducing termini, which has now been confirmed for Trichoderma reesei CBH I and CBH II (63). Exo-exo synergism may be explained by considering the tendency of cellulose chains to "reanneal" or return to the crystallite surface following hydrolysis if unimpeded by the presence of an exoglucanase. It is possible that maximal initiation of the processive hydrolytic process catalyzed by exoglucanases occurs only if both exoglucanases (reducing and nonreducing specific) are present at the site of internal bond hydrolysis immediately following endoglucanase action. In order to reduce the effects of steric hinderance at this site, both exoglucanases must bind to their respective substrates with high precision.

As of early 1996, glycosyl hydrolases have been grouped into 56 families of related proteins, based on amino acid sequence homology (64-66) and hydrophobic cluster analysis (67) of catalytic domains. Cellulases are included in 11 of the 56 glycosyl hydrolase families so far elucidated by these methods. Some cellulase families include endo — and exoglucanases from either fungal or bacterial systems. The mechanism of cellulase action is a major feature common to all members of a family (e. g., single displacement, leading to inversion of configuration at the anomeric carbon, or double displacement, leading to retention of configuration at the anomeric carbon (68, 69). Subsets of these glycosyl hydrolase families have been grouped into super-families, or clans, on the basis of conservation of enzyme mechanism and tertiary structure of the molecule (70, 71).

Applications and Economics of Cellulase Production. The ultimate goal of advanced cellulase development research is to develop a robust and easily integratable cellulase production system that can produce adequate amounts of highly effective enzyme to satisfy process requirements at a cost compatible with overall ethanol process economics-probably in the range of $0.05 to $0.10/gallon ethanol produced. For comparison, starch-based ethanol processes currently consume approximately $0.04 to 0.05 in enzyme per gallon of ethanol produced (Miller, C, personal communication, 1995). Current assumptions dictate that the enzyme production system must be able to produce vast quantities of enzyme. For example, at an enzyme loading of 25 FPU/g cellulose, a bioethanol process will require about 11 million FPU (19 kg, 42 lbs) of cellulase to process 1 ton of biomass (1000 lb cellulose) to 84 gallons of ethanol. This amount of enzyme is equivalent to about 143 L of a commercial preparation that contains

80.0 FPU/L. Therefore, a single 2,000 ton/day bioethanol plant would require a staggering 15,000 ton/year cellulase, which is conservatively one-fourth the entire 1994 U. S. market for all industrial enzymes (72). Put another way, at 25 FPU/g cellulose,

136.0 FPU (approximately 1.8 L of commercial enzyme) of cellulase will be required to produce 1 gallon of ethanol.

Most industrially useful enzymes are produced using fermentative processes, which involve capital — and labor-intensive production plants that use large stainless steel tanks, huge volumes of media, expensive energy, and material inputs (agitation, oxygenation, media, steam sterilization, etc.). Commercially available cellulase preparations are currently employed on a large scale almost entirely in non-biomass conversion applications, such as textile processing ("biostone"-washed jeans), detergents, and food processing. Each of these markets commands a much higher price for cellulase than can be afforded by any projected bioethanol process in the United States today, which is due, in large part, to the low cost of gasoline and the high cost of feedstock materials.

Given that cellulases are a critical component of lignocellulose conversion technology and that commercially available cellulase preparations are currently far too expensive for use in bioethanol processes, the alternatives are many and can be summarized briefly.

• Develop an on-site liquid or solid-state fermentative process for producing cellulase with optimized microbial sources (e. g., filamentous fungi, such as T. reesei, Aspergillus nigerf or Humicola insolens).

♦ Develop a genetically engineered cellulase production system that uses a bacterial or fimgal host to express and secrete effective cellulase preparation, or some other genetically engineered cellulase production system that uses a non-microbial host organism, such as insect cells, crop plants, or lactating mammalian systems.

Developing cost-effective cellulase production methods, designed for complete and efficient hydrolysis of cellulose in relevant feedstocks is therefore essential. Besides the cost that results from losses in overall ethanol yield when enzyme is produced from biomass, high cellulase production costs are also due to the intrinsic costs of fermentative processes, including capital-intensive tankage, agitation, and sterilization equipment, and costly media and contamination control chemicals.

Prospective for T. reesei Cellulase Use in Near-Term Bioethanol Plants. T. reesei mutants are generally recognized to be the best strains currently available for the industrial production of cellulases (73). Yet, the cost of bulk quantities of cellulase to the ethanol from biomass process has remained an area of uncertainty. Although few detailed economic studies are available, an estimate for the cost of cellulase production from lactose for an advanced bioethanol plant based on Iogen technology was proposed as $0.53/gallon ethanol (74). More recently, an estimate of $0.30 to $0.81 per pound cellulase protein was proposed for on-site cellulase production based on Army-Natick data (75). (These values may be cautiously converted to a cellulase cost of approximately $0.11 to $0.30/gallon ethanol assuming a specific activity of 600 FPU/g protein and that

100,0 FPU are required to produce 1 gallon ethanol). Cellulase cost data from the 1993 study by Hayn (76) proved to be in agreement with the earlier Iogen data and were based on actual separate hydrolysis and fermentation (SHF) pilot plant results, which provided a cellulase cost estimate of approximately $0.68/gallon ethanol. Because no detailed pilot — scale studies of cellulase production from pretreated woody biomass are readily available, production parameters based on small-scale T. reesei growth and induction studies are critical.

Enzyme Technology for Next-Generation Bioethanol Plants and Beyond. Because the cost of producing the enzymatic catalysts for the SSF process is a critical issue, the available enzymatic activity must be maximized. This requirement can be met by ensuring that the enzymes used are obtainable at minimal cost and have the highest specific activity and the highest possible stability at the pH and temperature of the intended application. Esterbauer and co-workers (77) caution that, "In retrospective, we and others feel that cellulase production by Trichoderma has its limitations and a significant further improvement cannot be expected. In the future, efforts should also be focused on other cellulolytic microorganisms, both bacteria and fungi."

We feel that cellulase systems capable of greater productivities and carbon conversion efficiencies than those possible from fungi are required for the success of bioethanol plants targeting objectives for the next decade (i. e., $0.67/gallon ethanol) (78,3). Preliminary technoeconomic analyses of the bioethanol process at NREL show that the cost of on-site cellulase production is keenly sensitive to delivered feedstock cost ($/ton feedstock), moderately sensitive to the carbon conversion efficiency of cellulase production [FPU produced/gram carbon consumed from feedstock = (gram cellulase/gram carbon) x (FPU/gram cellulase)], and less sensitive to enzyme loading (FPU/gram cellulose content in SSCF) (Glassner, D., personal communication, 1996). Thus, the key to this strategy is to increase the specific activity, thus deriving more FPU activity per gram of protein produced and to increase both carbon conversion efficiency and effective enzyme loading. The degree to which the specific activity of the cellulase system can be increased is not known. A related issue of equal importance is the development of an expression system that can produce large quantities of recombinant enzymes at low cost, preferably from low-value processing plant streams.

Strategies for Improving Cellulase Systems. The first major research goal for advanced cellulase technology is to increase the specific activity of cellulase systems. We propose approaching this challenge through the use of at least four distinct strategies. First, the component enzymes constituting known cellulase systems can be isolated and recombined to create new, non-natural systems to be evaluated for possible improvements in activity-to a large degree, work initiated at NREL (79) and elsewhere (80-83) has shown that significant, but probably limited, increases in system efficiency can be expected from this approach. One important discovery from two-component enzyme mixing studies at NREL was that a significant improvement in degree of synergism (DSE) and reducing sugar (RS) release (i. e., as much as 40% DSE and 30% RS release) can be found when mixing enzymes produced from diverse organisms, in this case a hot-spring bacterium and a filamentous fungus (79). Our progress in demonstrating the potential for new cellulase system engineering by assessing the efficacy of mixed origin and native binary and ternary systems is shown in Table 1. Although admittedly not as active on Sigmacell as the native T. reesei ternary system, the new ternary system based on Acidothermus cellulolyticus El, Thermomonospora fusca E3, and T. reesei CBH I is competitive and offers the potential for another important advantage for bioethanol process cellulases, thermal tolerance and increased process half-life.

Engineered Cellulase Systems at NREL. A. cellulolyticus is a thermo tolerant, cellulolytic bacterium which was originally isolated from a Yellowstone National Park hot spring (84). One component enzyme from this cellulase system, El, has been purified, characterized, crystallized, and subjected to x-ray crystallographic analysis (85). El is a 60 kD endoglucanase that is highly thermostable, demonstrates high specific activity on carboxymethylcellulose (86, 87), and is a member of the glycosyl hydrolase Family 5, Subfamily 1 (88). Because of its membership in this family, El is expected to show retention of the stereochemistry at the anomeric hydroxyl following catalysis.

Cellobiohydrolase I (CBH I), a 52 kD exoglucanase produced by the filamentous fungus, T. reesei, is the most abundant component of that organism’s cellulase system. T. reesei also produces another exoglucanase, CBH II, and at least three endoglucanases, referred to as EG I, EG П and EGV. CBH I is produced in extremely large amounts from a single genomic gene and may represent up to 60% of the protein secreted by strains of T. reeseitbax can produce extracellular protein at concentrations up to 40 grams/L (89).

Table 1

Non-Native Endoglucanase/Exoglucanase Mixtures Tested at
10% Total Saccharification for Degree of Synergistic Effect (DSE)
and Reducing Sugar (RS) Release

%Max. % Max. Enzyme Mixtures________________________ DSE______ DSE RS Release RS Release Cellulase Binary Systems* A. cellulolyticus El & T. reesei CBH I 2.75 100 40 100 T. fusca E5 & T. reesei CBH I 2.16 79 30 75 T. reesei EG I & T. reesei CBH П 1.72 63 31 75 T. reesei EG I 8c T. reesei CBH I 1.58 57 27 68 Cellulase Ternary Systems6 T. reesei EG I & T. reesei CBH I & T. reesei CBH II 2.51 100 55 100 A. cellulolyticus El & T. reesei CBH I & T. reesei CBH II 2.16 85 50 91 A. cellulolyticus El & T. reesei CBH I & T. fusca E3C 1.87 75 44 80 M. bispora endo A 8c T. reesei CBH I & T. fusca E, 1.79 71 36 65 Reducing sugar (as mM glucose) from Sigmacell 20 measured after 120 h at 50°C and pH 5.0 according to Baker and co-workers (79). aData collected at an endo/exo ratio of 20/80 (79). b Data collected at an endo/exoR/exoNR ratio of 20/20/60. cData for this study collected at an endo/exoR/exoNR ratio of 20/30/50.

CBH I is not thermotolerant, acts processively from the reducing end of a cellulose substrate, demonstrates retaining-type product stereochemistry (63), is a member of glycosyl hydrolase Family 7 (90), and exhibits a synergistic activity in combination with every endoglucanase that has so far been tested, including El (79).

E3 is a nonthermotolerant exoglucanase secreted by the actinomycete T. fusca (80). This enzyme is about 60 kD in molecular weight, demonstrates an inverting-type product stereochemistry, and is a member of glycosyl hydrolase Family 6 (Wilson, D., personal communication, 1996).

We feel that when acting on pretreated biomass, engineered enzyme mixtures can be formulated that are more effective than the most active native mixtures, at least at the binary enzyme level. Based on the notion that an engineered mixture of activities is desirable, and that an effective cellulase system must have at least one highly active endoglucanase, one cellulose reducing terminus-specific exoglucanase, and one cellulose nonreducing terminus-specific exoglucanase, we have proposed the selection of the

A. cellulolyticus El, T. reesei CBH I, and T. fusca E3 as components for our basic system.

Second, it should be possible to improve the kinetic efficiency of cellulases that work on pretreated biomass by using known principles of enzyme engineering (new strategies may also be required and are proposed for future work). We have proposed, for example, a two-phase approach to this problem by first modifying the native structure of El by targeted amino acid replacement to ensure optimal enzyme-cellulose (biomass) surface interaction and by improving the catalytic efficiency of the active site, also by SDM. A similar approach will be used to improve the action of the exoglucanases on pretreated biomass, with the additional goal of improving thermal tolerance.

Third, the usefulness of "accessory" glycosyl hydrolases for the enhanced saccharification of pretreated biomass substrates should be investigated. For example, the new countercurrent pretreatment methodologies described earlier produce somewhat higher levels of cellulose hydrolysis; however, the material left insoluble is more resistant to conventional cellulase action. Also, the liquid waste streams from complete hydrolysis tend to harbor enhanced levels of xylooligodextrins, as well as other, as yet unidentified, oligosaccharides. An important example of the effectiveness of removing substituent chemical inpediments to cellulases action was shown by Kong and co-workers (97) when they demonstrated that the cellulase digestion of aspen wood was substantially accelerated by prior chemical deactylation. This result could presumably also be achieved by the action of xylan acetyl esterases (92).

The overall strategy for developing engineered cellulase systems is depicted in Figure 2. The obvious benefit of routine reassessment of the specific enzyme candidates chosen for system membership is apparent, but resource intensive. A fundamental dilemma is encountered when limited resources dictate that choices be made between attempting to improve the operational characteristics of competitive enzymes (by SDM or mutation/selection) or simply returning to new rounds of screening from known cellulase producing organisms. A secondary benefit for the enzyme engineering approach is the acquisition of fundamental knowledge concerning cellulase action and structure relationships, which should improve chances of future success.

Relevant Cellulase Assays for Bioethanol Process Applications. The activities described above will be supported by the critical determination of the effectiveness of new enzymes using a novel assay method. Although cellulase enzymes are widely sold, and

If resources permit

Improve by SDM using biomass

Commercial scale expression system

their industrial utilization estimated, on the basis of the FPU of activity, the traditional "filter-paper assay" is severely limited as a predictor of cellulase performance in the extensive saccharification (80%-90%-plus) of actual industrial lignocellulosic substrates. These limitations are traceable both to the chemical and physical differences between filter paper and industrial substrates, and to the nonhomogeneous nature of most cellulosic substrates (filter paper included), which means that assays run to very limited extents of conversion (such as the 4% conversion target in the filter-paper assay) measure the digestibility of only the most easily digestible fraction of the substrate, and reveal little about the convertibility of the bulk of the substrate. Actual performance of cellulases is estimated better by assays that utilize the actual application substrate, and are run to the extents of conversion required in the process.

Because of the inhibitory nature of the products of cellulase action (primarily glucose and cellobiose), such high-conversion assays encounter the problem of significant product inhibition, if run in "closed" systems as simple saccharifications (93, 94). A new saccharification assay has been devised at NREL in which a continuously buffer-swept membrane reactor is used to remove the solubilized saccharification products. The diafiltration saccharification assay (DSA) serves as a reliable predictor of the performance of combinations of cellulase and substrate under simulated SSF conditions but retains the analytically more direct and accurate nature of a saccharification reaction. This assay will be used to compare the effectiveness of commercial T. reesei and specially engineered (cloned) cellulase preparations in the saccharification of standard and novel dilute acid pretreated substrates.

For decades, cellulase biochemists have proposed that enzyme-secreting microorganisms should produce somewhat different hydrolytic enzyme systems when presented complex biomass substrates, than when grown on simple, soluble sugars. We have recently shown that a T. reesei mutant archived at NREL produces a cellulase system considerably more effective at hydrolyzing pretreated hardwood sawdust when grown in the presence of the same substrate (95).

Expression of Engineered Cellulase Genes. Our initial goal was to identify the cellulase genes that will be incorporated into the first-generation cellulase production system. This has been done (at least provisionally). The next goal is to select a host organism most likely to be able to produce a high specific activity enzyme preparation in a cost-effective manner. Initially, single-gene expression strains will be constructed to maximize expression of active gene product. Strains shown to express different target genes at desired levels will be crossed to combine two or more cellulase genes in the same strain. It is essential to maximize the specific activity (activity/g protein) of the genetically engineered cellulase preparation, which requires specific molar ratios of each particular combination of endoglucanases, exoglucanases, and p-glucosidases. Multi­gene expression strains will eventually be constructed to maximize expression of balanced cellulase expression systems.

Once a host system has been selected, the cellulase coding sequences must be incorporated into genetically engineered artificial gene constructs, which will be recognized and readily expressed in that organism. Because the cellulase genes targeted for expression in the desired host may have originated from either a bacterium or a fungus, the standard approach will be to use the coding sequence from the selected cellulase gene and place it downstream from a host-derived promoter known to express its native product at very high levels. Downstream from the coding sequence will be spliced appropriate transcription termination signals (and polyadenylation signals, if required) known to function effectively in the chosen host organism, possibly from the same gene as the promoter was derived. The exact DNA sequence context and proximity of the host promoter sequence and the foreign coding sequence is critical in any chimeric gene construction, because a single nucleotide change in sequence or distance can make a huge difference in the behavior of the gene. How changes in promoter sequence context will affect expression of chimeric genes cannot be accurately predicted for any organism at this time (except perhaps E. coli).

The cellulase coding sequences that have been selected for incorporation into the initial cellulase production system include:

• A. cellulolyticus El endoglucanase (gene of bacterial origin; 61% G+C content)

• T. reesei CBHI (cDNA clone of fungal origin; ~50% G+C content)

• T. fusca E3 (gene of bacterial origin; high G+C content)

• A. cellulolyticus p-glucosidase (gene not yet cloned)

Initial work in the area of gene cloning took place in the early 1970s using the gram­negative bacterium E. coli as the host organism. E. coli is still by far the most well — characterized organism in terms of its molecular genetics and the biochemistry of its genetic machinery. Within a short period it became possible to clone and express foreign DNA in other bacteria, such as Bacillus subtilis, Streptomyces lividans, and others, as well as brewer’s yeast (Saccharomyces cerevisiae). During the past 10 years, cloning and expression of foreign DNA in fungi has become fairly routine (Neurospora, Aspergillus, and Trichoderma, for example). The ability to introduce foreign DNA and control its expression in a wide variety of higher organisms has also been achieved only recently (including many plant species, insects, and mammalian lactation system).

The most technically approachable systems for expressing foreign DNA at high levels include bacteria (E. coli, S. lividans, B. subtilis, B. brevis, B. stearothermophilis, and B. licheniformis), yeasts (S. cerevisiae, Pichia pastor is, and P. stipidis), filamentous fungi (A. niger, A. oryzae, A. nidulans, and A. awamori), higher plants (tobacco, alfalfa, and Arabidopsis thaliana), insect cells and larvae (Baculovirus), and mammalian milk. Two of the three bacterial systems listed have already been explored at NREL for their potential to express and secrete functional foreign cellulase gene products at high levels. Although perfectly adequate for producing reagent quantities of functional cellulases, both E. coli and S. lividans are inadequate as cellulase production systems on a scale required by the bioethanol process.

(1) . Escherichia coli. E. coli is still the workhorse of modem molecular biology. As a genetic system, it is extremely well characterized. All cellulase genes obtained from various microorganisms under the sponsorship of the ethanol project during the past few years were first cloned and expressed in E. coli. Efforts to improve recombinant bacterial expression of cloned cellulases quickly gave way to other hosts, including S. lividans and

B. brevis after it became clear that heterologous products were being degraded by proteases and incorporated into insoluble inclusion bodies in E. coli, and that E. coli has a limited capacity for secreting foreign proteins into the medium (87).

Although E. coli can synthesize a foreign protein to levels of 20%-30% of total cell protein, because of limitations of cell density in batch liquid culture, this amounts to only modest volumetric yields of protein (e. g., 1-2 g/L) (96). Fed-batch and continuous culture methods can achieve significantly higher specific volumetric yields (e. g., g/L/h). We employ E. coli only as a host for cloning and expressing new cellulase genes, for constructing expression plasmids which will be used in other host systems, and for producing reagent quantities of individual cellulases.

(2) . Streptomyces lividans. Reputedly "strong" promoters isolated from various Streptomycetes were used at NREL to construct expression vectors for use in S. lividans, including tipA (a thiostrepton inducible promoter) and STI-II (S. longisporus soybean trypsin inhibitor II). Using these promoters, we have constructed various expression vectors in hopes of achieving g/L quantities of secreted, functional recombinant El in S. lividans. We have successfully secreted fully active El from most of these constructs, but none exceeds the production level of that produced by the native gene expressed in S. lividans. On the other hand, efforts to increase the expression level of a p-glucosidase cloned from Microbispora bispora successfully increased the expression level by a factor of 3-4, to more than 200 mg/L (Xiong, X., unpublished results, 1996). Additionally, Wilson reports that he has successfully increased the level of expression of one T. fusca exoglucanase gene, E3, approximately five fold using the STI-II promoter (i. e., 150-200 mg/L) (Wilson, D., personal communication, 1996). Despite these successes, it is unlikely that the S. lividans system will be capable of much more than about 1 g/L production levels in batch culture.

(3) . Saccharomyces cerevisiae. Despite being genetically well characterized, which provides the genetic engineer with numerous potential host strains, selectable markers, well-characterized promoters, and several vector alternatives, baker’s yeast is not a good choice as a host for a cellulase expression system (97). Because of its notorious capacity for hyperglycosylation of foreign proteins and its limited capability for synthesizing and secreting foreign proteins, S. cerevisiae has not been considered a serious candidate for a cellulase production system at NREL.

(4) . Pichia pastoris. P. pastoris can generate very high densities of protein-rich biomass in simple defined media, using methanol as a carbon source. For this reason, P. pastoris was first exploited commercially in a methanol-based process to produce single-cell protein (SCP) for use as a protein supplement in animal feeds. After characterization of the biochemical pathways for the metabolism of methanol in this and similar organisms, the two alcohol oxidase genes from P. pastoris were cloned and sequenced. Efficient transformation and expression systems were also developed for this organism, and are based on the methanol-inducible promoter from the alcohol oxidase 1 (AOX1) gene. The vectors used for P. pastoris transformation integrate into the yeast genome by homologous crossover at the AOX1 locus. Because P. pastoris does not heavily glycosylate secreted proteins, as is common in S. cerevisiae, it is useful in the production of human pharmaceuticals that require glycosylation for biological activity. The state of the field of heterologous gene expression in P. pastoris has been very recently reviewed (98).

Several examples show that Pichia can routinely achieve percentage yields (5%-40% of total cell protein) much higher than baker’s yeast, and often equivalent to E. coli or baculovirus (99,100). Because Pichia is able to grow to much higher densities in liquid culture than any of the aforementioned systems, it can produce much higher volumetric yields (g/L). Scale-up of Pichia culture to extremely high cell density is simple and has resulted in enormous volumetric yields (e. g., 12 g/L for tetanus toxin fragment C [101] and >3 g/L secreted human serum albumin [102]).

We have tested a strain of P. pastoris designed to express and secrete of the A. cellulolyticus El endoglucanase. In this construction, the mature El coding sequence was joined in the same translational reading frame to the yeast alpha factor signal sequence present in pPIC9. Transformants have not yet been analyzed for gene copy number, but trial fermentations have already yielded 1.5 g/L of EI. Approximately 50% of the EI produced in P. pastoris is secreted into the medium and the remainder is found intracellularly (Thomas, S., unpublished results, 1995).

(5) . Extracellular Production of Heterologous Fungal Proteins in Fungal Hosts. Several investigations concerned with the efficiency of secretion of heterologous proteins have focused on fungal enzymes. The results of published experiments to express foreign genes in filamentous fungi have recently been summarized by van den Hondel {103). Most of these enzymes are important for the industrial production of foodstuffs, animal feeds, and detergents. A large number of studies have been carried out using Aspergillus awamori, A. niger and A. oryzae, for which extensive experience in fermentation and downstream processing has been established.

The initial level of production of fungal proteins in heterologous hosts is usually in the range of 10 to 50 mg/L. Under nonoptimized conditions, similar production levels are observed for efficiently secreted homologous proteins. However, after optimization of the production process, levels of at least 3 g/L have be obtained. Clearly production yields can usually be improved considerably through the use of modified hosts, media optimization, use of appropriate large-scale fermentation conditions, and classical strain improvement procedures.

(6) . Extracellular Production of Bacterial Proteins in Fungal Hosts. The

literature reveals only a few reports of studies that deal with the expression and secretion of bacterial proteins in filamentous fungi are available to date. The Cellulomonas fimi endoglucanase, being a cellulase, is particularly relevant to the matter at hand. The 5′ region of the inducible A. nidulans ale A gene was employed to direct expression of the

C. fimi endoglucanase in an A. nidulans host strain previously modified to overproduce the alcR gene product, which positively regulates alcA transcription. This promoter is repressed in media that contains glucose, but can be induced when the carbon source is switched to ethanol. This work demonstrated production of approximately 20 mg/L of functional, secreted endoglucanase in shake flask cultures growing sub-optimally in minimal media at 37°C for 48 h {104).

Alternatively, Turnbull and co-workers {105) produced E. coli enterotoxin subunit В (LTB) at low levels (2 ng/|ig soluble protein, or 24 ig/g wet weight of mycelia) in

A. nidulans using an "up-regulated" inducible amdS promoter. This promoter can be induced by either acetate or acetamide. The fact that the рге-LTB product was properly processed to remove the bacterial signal peptide but not secreted into the medium indicates some sort of incompatibility with the fungal secretory apparatus, or that secreted protein was rapidly degraded by extracellular proteases.

(7) . Trichoderma reeseu T. reesei can produce remarkable amounts of extracellular protein (20 to 50 g/L). Even so, until now, the use of filamentous fungi as general production hosts has been restricted mostly to research and industrial laboratories with special interest in these organisms. Compared with E. coli and S. cerevisiae, which serve as model organisms for basic research and are widely used in molecular biology, the efforts so far invested in the development of filamentous fungi as production hosts have been very limited. The major cellulase, cellobiohydrolase I (CBH I), which is produced from a single copy gene, represents -50% of the total protein secreted. Thus, the cbhl promoter is extremely strong. The excellent synthesis and secretion capacity of the organism, together with established fermentation conditions, prompted development of

T. reesei as a host for production of heterologous proteins. T. reesei has the advantage of possessing a eukaryotic secretory machinery, and, most likely, similar protein modification properties (e. g., high mannose type N-glycosylation; 106) to mammalian systems.

Recently, in a continued search for powerful promoters that are active in the presence of glucose-containing media, the group at VTT (Espoo, Finland) has screened a cDNA library for sequences that are highly abundant (707,108), It has then used those cDNAs to isolate the corresponding genomic clones and the promoters for those genes. This level of expression proved to be 20 to 50-fold higher than that of the pgkl gene. Promoters for two of these genes have been isolated and used to drive expression of the homologous EG I coding sequence in T. reesei strain QM9414 growing in glucose — containing medium.

(8) . Higher Plants. There is sound reasoning behind the approach to produce bulk industrial proteins in crop plants. According to Pen and co-workers, "In terms of cost — effectiveness for producing biomass, the growing of crops in the field can generally compete with any other system. It is inexpensive, it can be done in bulk quantities, and it requires limited infrastructure. These observations suggest that the exploitation of arable crops for the production of food, feed, or processing materials would be very attractive" (109), Aside from various academic laboratories around the world, at least three plant genetic engineering companies are actively pursuing the expression of relatively low value bulk industrial proteins in crop plants of one sort or another: MOGEN International (Leiden, The Netherlands), human serum albumin (HSA) and a-amylase in tobacco and potato; Ciba-Geigy (Research Triangle Park, NC), cellulases in maize; Calgene, Inc. (Davis, CA), cellulases in tobacco.

The idea for production of industrial enzymes and other bulk proteins in plants is not unique to NREL. For example, others have already demonstrated expression of Bacillus licheniformis a-amylase in tobacco (109, 110), Clostridium thermocellum xylanase (XynZ) in tobacco (111), and HSA in potato (109). These examples involve the stable transformation of plant nuclear DNA via Agrobacterium-mediated gene transfer. In all three cases, no effect on cell growth and development was observed, indicating the foreign DNA expression did not adversely affect the transformed plants. A patent suggests that a variety of proteins from many sources can be used as target proteins for expression in barley endosperm (772).

B. licheniformis a-amylase was expressed at 0.5% of total protein in tobacco. Alpha- amylase purified from tobacco exhibited a slightly higher molecular weight than the native enzyme (64 versus 55 kDa), and was entirely due to glycosylation in the plant system. Nevertheless, despite glycosylation, the a-amylase expressed in tobacco was active, secreted, stable at 95-100°C, and otherwise completely indistinguishable from the native enzyme. Perhaps just as important is that the starch content of transformed leaves (in chloroplasts) was unaffected by the cytoplasmically expressed a-amylase. This is an important point because it clearly demonstrates the ability to isolate a transgenic protein from its potential substrate by compartmentalization, thus avoiding potentially detrimental effects on the plant.

Herbers and co-workers (111) recently expressed a truncated version of a thermostable C. thermocellum xylanase in transgenic tobacco plants. The authors speculate that these plants might be useful for producing xylanase, which has numerous applications in the paper industry and agriculture. The xylanase was synthesized as a 37 kDa polypeptide and correctly targeted to the intercellular space by means of a proteinase inhibitor П signal peptide. The xylanase was one of the most abundant proteins in total extracts (4.1 +/- 1.6%) and represented more than 50% of protein present in the intercellular fluids. The transgenic plants, grown under greenhouse conditions, were not affected by the foreign enzyme, possibly because of the high temperature optimum of the xylanase and the low levels of xylan in tobacco cell walls.

In 1993, the highest published level of expression of a foreign protein by nuclear transformation in a plant system was 1.5% of soluble protein {113). Differences in the level of foreign gene expression seem largely gene dependent and may be due to efficiencies in transcription/translation or stability of the gene products. Further improvements in expression levels seem likely, given the modest effort expended in this area so far.

Another new approach to plant transgenic expression systems involves integrating the target gene into the tobacco mosaic virus RNA genome, followed by infecting tobacco and other solanaceous species (e. g., tomato). As the infection spreads systemically, the plant becomes a dedicated bioreactor for expressing the foreign gene (114-116). TMV — based vectors can produce heterologous proteins in tobacco at levels between 5% and 40% of total cell protein (della-Cioppa, G., personal communication, 1995).

Despite progress in the development of gene expression technology, significant problems remain in the manufacture of many complex proteins. Many post-translational modifications performed by animal cells can be performed by green plants which, like animals, are complex, eukaryotic organisms. Green plants are a promising, underexploited system for expressing new proteins, including cellulases. In addition, green plants are photoautotrophic, requiring only carbon dioxide, water, nitrogen, sulfur, phosphorus, and trace amounts of other elements for growth.