Lignin is an aromatic network polymer composed of phenylpropane units [53]. It is generally accepted that lignin is the “glue” that binds cellulose and hemicellulose, imparting both rigidity and moisture resistance to the lig — nocellulosic structure. Lignin has also been implicated as an inhibitor of cellulases; therefore, many of the pretreatment methods currently being ex­plored have tried to decrease the lignin content of the solid substrate while minimizing the degradation of carbohydrates [22]. The amount of lignin that must be dealt with by a particular pretreatment and subsequent hydro­lysis depends on the source of biomass. For example, corn fiber has a low lignin content of 7% (w/w) [23] compared to 30% (w/w) in the case of a soft­wood such as Douglas-fir [51]. In addition to the amount of lignin present in a biomass feedstock, the type of lignin differs between agricultural residues, hardwoods and softwoods [54]. Grasses and agricultural residues contain primarily p-hyroxyphenyl units while hardwoods and softwoods contain greater amounts of syringyl and guaiacyl subunits, respectively [54]. Soft­woods lignin also exhibits a greater degree of cross-linking due to an extra linking site provided by the presence of only a single methoxyl group on the guaiacyl aromatic ring [54]. Another factor that must be considered is the ex­istence of lignin carbohydrate complexes (LCCs) that consist of lignin linked to carbohydrates through bonds such as ester, ether or ketal [55]. LCCs have been shown to be particularly problematic for hydrolysis processes, as access to the carbohydrate fraction is restricted by the attached lignin, therefore pre­treatment processes should either expand the pore structure of the substrate or remove the lignin outright [56].

3.1

Following the release and annotation of a genome, the next logical step is to evaluate the messenger RNA expression level on a whole genome scale, referred to as transcriptome analysis. Targeted metabolic engineering relies heavily on the assumption that a genetic perturbation — gene deletion, con­stitutive overexpression, regulated induction, or modulation — will confer a metabolic flux response. This stems from the central dogma of biology: DNA is transcribed to RNA and subsequently translated to polypeptides that give rise to phenotype. Prior to transcriptome analysis, genes were assumed to be expressed followed by post-translational regulation, with little under­standing of interactions across gene loci [81]. In fact, transcriptome profiling of reference strains has provided a first approximation as to which pathways are active and, equally important, inactive, assuming that up-regulated gene expression leads to up-regulated pathway activity. It has since been shown that this is not always true — elevated mRNA levels do not always translate to elevated protein levels or activity. It has also provided significant insight into alternative modes of regulation, such as transcription factor-mediated as opposed to post-translational regulation. This has permitted narrowing of the experimental space that metabolic engineers need to consider, and made available new strategies to consider. Additionally, transcriptome pro­filing provides a quantitative in vivo assessment of several key metrics fol­lowing a genetic perturbation relative to a reference case: (1) what is the net change in mRNA expression levels of the targeted gene(s), (2) what is the net change in mRNA expression levels of non-targeted gene(s), and (3) what is the net change in mRNA expression levels of either reference or constructed strains under specific environmental conditions. These questions aim to iso­late which genes and pathways may serve as targets and/or explanations for observed or induced phenotypes. Measurement of the transcriptome, via readily available microarray technology, has evolved into a routinely meas­ured data set for many industrially relevant organisms, including E. coli and S. cerevisiae, and is playing a central role in both forward and reverse metabolic engineering [63,82,83].

Among the first applications of transcriptome measurements with in­dustrial relevance to bioethanol production was establishing the baseline response of S. cerevisiae to diverse carbon substrates and medium com­positions — essential for optimizing strains to given feedstocks and vice versa. Steady-state chemostat cultures were used to measure transcriptome responses under glucose, ethanol, ammonium, phosphate, and sulfate lim­itations [84]. Results suggested that genes related to high-affinity glucose uptake, the TCA cycle, and oxidative phosphorylation were up-regulated in glucose-limiting conditions, while genes involved in gluconeogenesis and ni­trogen catabolite repression were up-regulated in ethanol-grown cells [84]. In a similar but earlier study, transcriptome measurements were performed of S. cerevisiae grown using glucose-limited chemostats coupled with nitro­gen, phosphorus, and sulfur limitations [85]. In total, 1881 transcripts (31% of the total 6084 different open reading frames probed) were significantly up — or down-regulated between at least two conditions, and a total of 51 genes demonstrated a >tenfold higher or lower expression within a given condi­tion [85]. The transcriptome profiles under each condition have provided genetic motifs that may be recognized and regulated by transcription factors. These may be used in metabolic engineering strategies that could cater to a specific growth medium composition.

With the experimental mechanics of collecting transcriptome data becom­ing common place, attention and focus is now placed on data analysis methods and integration with other x-ome data sets. It has become abundantly clear that transcriptome data alone, unless used for the purposes of environmen­tal screening or quality control (i. e., confirming that an engineered genotype is producing the corresponding transcription profile), provides limited bi­ological insight. Several efforts have emerged coupling transcriptome with metabolome and fluxome data [86-89]. For example, elementary flux modes for three carbon substrates (glucose, ethanol, and galactose) were deter­mined using the catabolic reactions from the genome-scale metabolic model of S. cerevisiae, and then used for gene deletion phenotype analysis. Control — effective fluxes were used to predict transcript ratios of metabolic genes for growth under each substrate, resulting in a high correlation between the theor­etical and experimental expression levels of 38 genes when ethanol and glucose media were considered [90]. This example demonstrates that incorporating transcriptional functionality and regulation into metabolic networks for in silico predictions provides both more biologically representative models and a means of bridging transcriptome and fluxome data.

In another example, the topology of the genome-scale metabolic model constructed for S. cerevisiae is examined by correlating transcriptional data with metabolism. Specifically, an algorithm was developed enabling the iden­tification of metabolites around which the most significant transcriptional changes occur (referred to as reporter metabolites) [91]. Due to the highly connected and integrated nature of metabolism, genetic or environmental per­turbations introduced at a given genetic locus will affect specific metabolites and then propagate throughout the metabolic network. Using transcriptome experimental data, a priori predictions of which metabolites are likely to be affected can be made, and serve as rational targets for additional inspection and metabolic engineering [91]. This algorithm has been recently extended to include reporter reactions, whereby transcriptional data is correlated with the metabolic reactions of the reconstructed S. cerevisiae genome-scale metabolic network model to identify those reactions around which a genetic or environ­mental perturbation conferring transcriptional changes cluster [92].

As more genomes continue to become available, and microarray technol­ogy continues to become more accessible with cost-effective customizable DNA microarrays now available, transcriptome data will continue to increase. Bioinformatics for data handling, integration of transcriptome with other x-ome data, and the development of various network models that rely on tran- scriptome data for biological interpretation will continue to develop. From an industrial biotechnology perspective transcriptome measurements and analysis have played a large role in reverse metabolic engineering; transcrip­tional surveying of a strain constructed either via random mutagenesis or directed evolution [63,82,83,93]. For example, lysine production via C. glu — tamicum has undergone transcriptome and fluxome measurements to elu­cidate greater than 50 years of traditional metabolic engineering (random mutagenesis), providing new targets for improved strategies [94-96]. This ef­fort, applied to other industrial biotechnology processes, is likely to intensify.

3.3

The key to developing an economically viable biorefinery is to employ a holis­tic approach that integrates the unit steps, maximizing the yield at each, while minimizing both capital and operating costs. At each step of the process, from pretreatment to fermentation, effort must be made to minimize any loss in potential ethanol production. In the example in Fig. 2, the production of degraded sugars during pretreatment, incomplete cellulose or hemicellulose

conversion to fermentable sugars during hydrolysis, and fermentation losses due to sugar consumption by the yeast all contribute to lost value in the con­version. If biomass feedstock such as corn stover, purchased at $5/ton, could be converted with perfect efficiency to its theoretical potential of 113 gal­lons of ethanol per ton of stover with an ethanol selling price of $2/gallon, the value of the ethanol would be ~$225/ton, creating an “operating cost window” for depreciation of capital, operation, and profit of ~$220/ton [5]. Losses in any unit step that reduces the overall yield will reduce the value per ton, whether the losses result from reduced enzyme hydrolysis, poor fer — mentability of the hydrolyzate sugars, or reduced fermentation yield. It is also important to note that maximizing the conversion of the two most abundant sugars, glucose and xylose, is important to viable economics. If only cellulose is utilized with no conversion of hemicellulose, the theoretical yield drops 39% to 69 gallons/ton, reducing the cost window to ~ $135/ton. Unless the xylose is utilized to produce something of equal or higher value, it is un­likely that such a process could be viable. Similarly, a pretreatment selected on the basis of a reduced capital cost for installed equipment, but increasing the required enzyme dosage, may reduce the operating cost window significantly.

3

This category includes methods in between, or a mixture of, purely physical and chemical methods. Steam pretreatment is one of the most widely used methods for pretreatment of lignocellulosics. This pretreatment method used to be called steam explosion, since it was believed that an “explosive” action on the fibres was necessary to render a material suitable for hydrolysis. It has been shown that it is more likely that the effect of steam pretreatment is due to acid hydrolysis of the hemicellulose, which is the reason why some cellulosic materials are easier than others to break down [30,31]. In particu­lar, agricultural residues and some types of hardwood contain organic acids, which act as catalysts for the hemicellulose hydrolysis. Using steam pretreat­ment the raw material is usually treated with high-pressure saturated steam at a temperature typically between 160 and 240 °C (corresponding to a pres­sure between 6 and 34 bar), which is maintained for several seconds to a few minutes, after which the pressure is released. During pretreatment some of the raw material, predominantly hemicellulose, is solubilized and found in the liquid phase as oligomeric and monomeric sugars. The cellulose in the solid phase then becomes more accessible to the enzymes. It is in some cases dif­ficult to find conditions that result in high yields of both hexose and pentose sugars, and at the same time also create a cellulose fraction which is easy to hydrolyse to glucose. This may call for steam pretreatment using two steps, where hemicellulose sugars are recovered at lower severity, while the cellulose fraction is subjected to pretreatment at higher severity.

Steam pretreatment can be improved by using an acid catalyst, such as H2SO4 or SO2. The acid increases the recovery of hemicellulosic sugars, and also improves the enzymatic hydrolysis of the solid residue. The use of an acid catalyst in steam pretreatment results in an action similar to dilute acid hydro­lysis but with less liquid involved. It is especially important to use an acid catalyst for softwood, since softwood in general is more difficult to degrade.

Steam pretreatment with addition of a catalyst is the pretreatment method for hydrolysis and enzymatic digestibility improvement that is closest to commercialization. It has been widely tested in pilot-scale equipment, for example, in the NREL pilot plant in Golden, CO (USA) [32] and in the SEKAB pilot plant in Ornskoldsvik (Sweden) [33], and is also used in a demonstration-scale ethanol plant at Iogen in Ottawa (Canada) [34].

Hydrothermolysis, or liquid hot-water (LHW) treatment, involves treat­ment in water at high temperature. This method is similar to steam pre­treatment, but lower temperatures and lower dry matter (DM) content are used, and thus more poly — and oligosaccharides are recovered [35,36]. A cat­alyst, such as an acid, can be added, making the method similar to dilute acid pretreatment. Since the water content is much higher than in steam pre­treatment, the resulting sugar solution is more diluted and thus causes the downstream processes to be more energy demanding. In the range 1-10 wt % DM virtually no difference in ethanol yield was found when bagasse was treated at 220 °C, after which SSF was performed using S. cerevisiae [37].

Wet oxidation pretreatment involves the treatment of the biomass with water and air, or oxygen, at temperatures above 120 °C, sometimes with the addition of an alkali catalyst. This method is suited to materials with low lignin content, since the yield has been shown to decrease with increased lignin content, and since a large fraction of the lignin is oxidized and solubi­lized [38]. As with many other delignification methods, the lignin cannot be used as a solid fuel, which considerably reduces the income from by-products in large-scale production. As discussed in the “Process Economics” chapter, it is extremely important to recover as much as possible of the lignin fraction (Sassner et al., in this volume).

Ammonia fibre explosion (AFEX) is also an alkaline method which, sim­ilarly to the steam pretreatment process, operates at high pressures. The biomass is treated with liquid ammonia for about 10-60 min at moderate temperatures (below 100 °C) and high pressure (above 3 MPa) [39,40]. Up to 2 kg of ammonia is used per kg of dry biomass. The ammonia is recycled after pretreatment by reducing the pressure, as ammonia is very volatile at atmospheric pressure. During pretreatment only a small amount of the solid material is solubilized, i. e. almost no hemicellulose or lignin is removed. The hemicellulose is degraded to oligomer sugars and deacetylated [41], which is a probable reason for the hemicellulose not becoming soluble. However, the structure of the material is changed, resulting in increased water holding cap­acity and higher digestibility. Like the other alkaline pretreatment methods AFEX performs best on agricultural waste, but has not proven to be efficient on wood due to its higher lignin content [42,43]. According to Sun et al. the AFEX process does not produce inhibitors that may affect downstream bio­logical processes [44].

Another type of process utilizing ammonia is the ammonia recycle perco­lation (ARP) method [45,46]. In the process aqueous ammonia (10-15 wt %) passes through biomass at elevated temperatures (150-170 °C), after which the ammonia is recovered. ARP is an efficient delignification method for hardwood and agricultural residues, but is somewhat less effective for soft­wood.

3.4

There is arguably no other industrial enzyme application that poses a greater challenge to the enzyme producer than supplying cost-effective enzymes for biomass utilization. The high enzyme loading required, combined with the low value of the final product, in the form of ethanol, requires not only that the enzymes be as efficient as possible, but that the cost of producing them be as low as possible. To this end, significant effort has been expended over the past 6 years to increase the productivity of the fungal strains used to pro­duce the enzymes, to reduce the cost of the enzyme fermentation process by reducing the cost of carbon and nitrogen sources for the fermentations, and to reduce the complexity of enzyme recovery and formulation.

Improving the host by classical mutagenesis is one way of developing a host strain with improved total protein production and improved activ­ities. This approach has a long and successful history. Montenecourt and Eveleigh [32] isolated RutC30, one of the best existing Trichoderma cellulase mutants, using a combination of ultraviolet irradiation and nitrosomethyl guanidine (NTG). Recently, Toyama, et al. [45] demonstrated a method to screen for increased cellulase production using growth through an overlay of cellulose substrate (Avicel) in Petri plates. In an effort to increase total cellulase productivity, a combination of these methods were utilized on the T. reesei strain currently used to produce Celluclast 1.5 L. Chemical muta­genesis was used to generate mutants that were screened using the method of Toyama [45] with minor changes. Briefly, mutagenized spores were sus­pended in an agar medium, poured into a plate and allowed to harden. The spore-containing layer was then covered with a top layer of agar contain­ing washed, acid pretreated corn stover (PCS) as the sole carbon source. Colonies growing through the PCS layer fastest were isolated and used in a secondary screening. In this, spores from selected fast-growing colonies were inoculated into shake flasks containing cellulase-inducing media. After 5 days of growth, broth samples were tested by robotic assay for produc­tion of reducing sugars from hydrolysis of PCS. Total protein assays were then conducted, and mutants expressing elevated cellulase and/or total pro­tein were then re-grown in 2-L fermentors. Broth from the fermentors was then analyzed again in PCS hydrolysis assays and for total protein. Some mu­tants were identified as having improved PCS hydrolysis and increased total protein secretion compared with the control. Top strains isolated in this man­ner showed significant increases in protein production and secreted cellulase activity.

Another method of improving a cellulase productivity is through increas­ing the expression of target proteins using genetic engineering. In many cases the total cost of supplying a heterologous mix of enzymes can be reduced by creating a single expression host expressing not only the native cellulases and hemicellulases, but expressing additional components, such as the BG and GH61 proteins, without negatively impacting the expression of the na­tive proteins. The introduction of multiple genes into a single host is no easy feat. A significant amount of work was done to identify strong promoters, to identify a number of selectable markers, and to develop a successful trans­formation technique that allows for co-transformation of multiple transgenes. These technological improvements have allowed us to rapidly and efficiently investigate the effect of introducing various enzymes into the T. reesei cellu — lase mix.

In addition to controlling gene expression transcriptionally, by utilizing promoters of different strengths, we have focused on enhancing individual protein yield by optimizing protein secretion. One example is the replacement of the A. oryzae BG signal sequence with a signal peptide from H. insolens Cel45A EG, which improved the BG secretion in T. reesei by two- to threefold relative to the unfused gene (Fig. 10).

As previously mentioned, several GH61 proteins result in a “boost” in PCS hydrolysis when supplemented to Celluclast 1.5 L. In addition, our stud­ies show that increased levels of в-glucosidase are required in our Tricho — derma host. Therefore, numerous co-transformations of T. reesei with various GH61s, A. oryzae в-glucosidase, and other genes of interest were carried out. Those transformants expressing both a GH61 and the в-glucosidase were then screened in PCS hydrolysis assays in order to identify the top strains in true performance assays. Those strains demonstrating the best perform-

ance in PCS hydrolysis were then fermented in 2-L bioreactors and retested in PCS hydrolysis assays. Eventually, a single strain was identified exhibiting im­proved hydrolysis from our original strains and high total protein production (Fig. 11).

5.1

Pretreatment methods such as solvent extraction (organosolv pulping) [57] or ammonia fiber explosion treatment (AFEX) [58] either modify or remove lignin, while a large proportion of lignin remains intact in the solid phase after SO2-catalyzed steam pretreatment [59] or dilute acid pretreatment [60]. Since lignin is intertwined amongst cellulose and hemicellulose, varying the pretreatment method or conditions employed to improve cellulose or hemi — cellulose yields will undoubtedly affect the lignin content [57,59]. For ex­ample, it has been shown that the amount of lignin in the solid fraction of the substrate increases as the severity of SP is increased. Based on 13 severity fac­tors used during SO2-catalyzed SP of corn fiber, it was shown that the amount of lignin in the solid phase increased as the severity of the pretreatment was raised [59]. The lignin was most likely concentrated in the solid fraction due to the solubilization and degradation of the carbohydrates as the severity was raised [59]. It should be noted that the Klason method that is commonly used to estimate the lignin content of the pretreated substrate can result in arti­ficially high values for lignin, as sugar degradation products and entrapped low molecular weight phenolics can also be measured as “lignin”. For ex­ample, it was shown that the Klason lignin contents of steam-pretreated aspen at a series of severities ranged from 6-30%. However, the supposedly corres­ponding methoxyl groups in the samples were only in the 0.8-7.7% range. This possible elevation of lignin values should be taken into consideration when determining Klason lignin content of a steam-pretreated substrate [61]. Realizing that it is difficult to reduce the amount of substrate lignin by fine — tuning SP process conditions, various researchers have explored methods of “post-treatment” to reduce the lignin content of steam-pretreated substrates.

In the past [62-64], we have tried to enhance the removal of lignin from steam-pretreated substrates, and consequentially increase the rate and ex­tent of hydrolysis by cellulases, by applying several chemical post-treatments (Table 1). Oxygen-alkali and hydrogen peroxide post-treatments removed similar amounts of lignin and thus improved hydrolysis. However, of note, we also showed [64] that the removal of only 7% of the lignin from a steam — pretreated Douglas-fir substrate using a cold NaOH treatment resulted in a 30% improvement in hydrolysis yields, indicating that, in addition to the amount of lignin, the location of lignin is also an important factor affect­ing hydrolysis. Palonen et al. [65] have applied an alternative delignification method employing laccase enzymes in combination with mediators to steam — pretreated softwood, resulting in a slight release of aromatics into the system (delignification was not measured) and a corresponding 21% increase in hydrolysis yield. These researchers also showed that the oxidation of lignin surfaces by the application of laccases in the absence of mediators, as shown by others with pulp fibers [66,67], also resulted in a 13% improvement in hydrolysis yield. These results suggest that the modification of lignin surfaces may also play a role in reducing its inhibitory effect on hydrolysis, perhaps affecting non-productive binding of cellulases to lignin. Since most of the studies have been concerned with altering lignin content by affecting the pretreatment conditions or applying post-treatments, there have only been

limited studies linking the various chemical structures in lignin to changes in hydrolytic activity of cellulases.

3.2

Proteomics is the quantitative study of all proteins expressed in a cell under defined conditions. Proteomics represents one of the more challeng­ing x-omes given that analytical methods enabling measurement of all pro­teins with the sensitivity, accuracy, and precision required have only recently been developed [62,72]. Rapid advances in protein analytical technologies, fueled by the addition of mass spectrometry (MS), liquid chromatography (LC), sequence databases, and data handling methods, have made it possible for protein chemists to identify and examine the expression of many pro­teins resolvable by 2DE (two-dimensional gel electrophoresis). The possibility for large-scale protein studies seemed attainable [97]. It was in this context that in 1994, at the first 2DE meeting in Siena, Italy, the term “proteome” was coined [98]. Methods employed in proteomics have since gone on to include two-dimensional differential gel-electropheresis (DiGE), multidimen­sional protein identification technology (MuDPiT), isotope-coded affinity tag technology (ICAT), and quantitative proteome analysis based on MS-MS spe — tra and a multiplexed set of chemical reagents referred to as iTRAQ [99]. Although still slowly emerging, there are clear examples of where proteome analysis has resulted in strain improvement and successful metabolic engin­eering strategies [62,100].

In line with industrial biotechnology applications, results of 2DE analy­sis can identify targets for strain improvement, such as target gene dele­tions [101] or co-expression for product enhancement [102]. Proteome an­alysis may also improve the design and control of industrial fermentation processes. In such a study, the dynamics of the E. coli proteome were recorded during an industrial fermentation process with and without in­duction of recombinant antibody synthesis [103]. The recombinant anti­body fragment CD18 F(ab’)2 was developed as a biopharmaceutical for the treatment of acute myocardial infarction. Proteomic analysis of the above fermentation process suggested co-expression of Phage shock pro­tein A (PspA) with a recombinant antibody fragment in E. coli resulted in improved yields. Further investigation is required to understand why PspA addition resulted in improved yield [104]. Another example, more relevant to bulk chemical manufacturing, is the metabolic engineering of E. coli to pro­duce the biodegradable and biocompatible thermoplastic polymer, poly-(3- hydroxybutyrate), often referred to as PHB, which has numerous applications including serving as a primary feedstock for synthesis of enantiomerically pure chemicals. Specifically, the proteome of the metabolically engineered E. coli XL-1 Blue for PHB intracellular accumulation was compared to the reference strain, noting that PHB accumulation is not observed in the refer­ence strain. It was revealed that 2-keto-3-deoxy-6-phosphogluconate adolase (Eda) plays a pivotal role in supplying glycerol-3-phosphate and pyruvate to further increase the flux to acetyl-CoA. A larger acetyl-CoA and NADPH de­mand is consistent with cells that produce a large amount of PHB. These conclusions were based on identification of protein spots on 2DE using matrix-assisted laser desorption ionization-time of flight (MALDI-TOF) mass spectrometry [105].

Among the most recent examples of proteomics applied to industrial biotechnology process development is the recent reporting of the com­plete proteome of Mannheimia succiniciproducens [100]. M. succinicipro — ducens MBEL55E is a capnophilic Gram-negative bacterium isolated from bovine rumen, which produces large amounts of succinic acid under anaer­obic conditions (0.68 g-succinic acid/g-glucose), and was first reported in 2002 [106]. Succinic acid is a C4 organic acid, traditionally produced via petrochemical conversion of maleic anhydride. It promises to be a strategic building block chemical to be produced by industrial biotechnology, due to its use as the primary feedstock in the synthesis of key products including bu — tanediol, tetrahydrofuran, y-butyrolactone, and poly-amides [107,108]. Nu­merous groups are exploring production of succinic acid in different host or­ganisms, including E. coli [109], Anaerobiospirillum succiniciproducens [110], Actinobacillus succinogenes [110,111], Aspergillus niger, and Saccharomyces cerevisiae. In M. succiniciproducens using 2DE coupled with MS-MS identi­fication and characterization lead to identification of 200 proteins, with 129 proteins from the whole cell proteome, 48 proteins from the membrane pro — teome, and 30 proteins from the secreted proteome. Characterization of cell growth and metabolite levels in conjunction with proteome measurements during the transition from exponential to stationary growth was carried out.

Two interesting conclusions could be drawn from such analysis that was not possible a priori. First, a gene locus previously annotated as the succinate dehydrogenase subunit A (sdhA) is likely to be the fumarate reductase sub­unit A (frdA), based on comparative proteome analysis supported by physi­ological data. Second, two novel enzymes were identified as likely metabolic engineering targets for future improvements in succinic acid production. PutA and OadA are enzymes responsible for acetate formation and conver­sion of oxaloacetate to pyruvate, respectively, and their deletion is likely to induce higher flux towards succinic acid through minimization of byprod­uct formation [100]. This is a clear example of where proteome measurement and analysis not only provided novel information for future metabolic engin­eering strategies, but also served as a quality-control check for two critical assumptions: (i) that genome annotation is error-free, and (ii) that mRNA expression directly correlates with protein expression and activity.

As discussed previously, acquisition of large bodies of genomic sequences has prompted development and application of tools such as cDNA/oligo — nucleotide microarrays, which in turn has made possible global analysis of cellular processes. As powerful as this approach is proving to be, much of the regulation of physiological processes occurs post-transcriptionally. Thus, measurement of mRNA levels provides an incomplete picture of cellular ac­tivity and regulatory control points that may yield themselves as preferred metabolic engineering targets. Methods and techniques developed to meas­ure the global expression, localization, and interaction of proteins fall within the domain of proteomics. By integrating various data sources with known biological function about individual genes and proteins, one starts uncover­ing underlying mechanisms leading to the creation and analysis of static and dynamic models of regulatory networks and pathways.

A recent study has shown the value of this union of data as an experimen­tal strategy to gain insights into cellular physiology [87]. In this study, both transcriptional and proteomic data were collected from S. cerevisiae and all of the known components of the galactose induction pathway were systemati­cally perturbed. The different data were integrated into a mathematical model that included enzymatic reactions, membrane transport, transcriptional acti­vation, protein activation, and protein inhibition. The model predicted pre­viously unknown intra-pathway interactions, and inter-pathway interactions of the galactose induction pathway and other cellular processes. Several of these predictions were then verified experimentally [87]. The galactose sig­naling pathway is of particular industrial relevance as one of the classical and best-understood promoter and induction systems used for protein expres­sion. This example further highlights that even such an extensively studied pathway will manifest new mechanisms for control and manipulation using x-omic approaches.

Related directly to bioethanol process development, several groups are evaluating proteomes of production organisms under defined environments that are of immediate industrial relevance. For example, Salusjarv et al. (2003) performed a proteome analysis of metabolically engineered S. cerevisiae strains cultured on xylose as compared to glucose under aerobic and anaero­bic carbon-limited chemostats [113]. Lignocellulosic feedstocks are abundant and renewable; however, are also composed of xylose — the most abundant pentose sugar in hemicellulose, hardwoods, and crop residues, and the sec­ond most abundant monosaccharide after glucose [114]. S. cerevisiae fails to consume pentose sugars efficiently, compared to glucose, and therefore sig­nificant research has occurred in metabolically engineering such strains (see Sect. 3.5 for further discussion). Proteome analysis of xylose fermentations revealed 22 proteins that were found in significantly higher concentrations relative to glucose fermentations. Such proteins included alcohol dehydroge­nase 2 (Adh2p), acetaldehyde dehydrogenases 4 and 6 (Ald4p and Ald6p), and DL-glycerol-3-phosphatase (Gpp1p) [113]. As will be revealed in the fluxome discussion, this protein expression profile is indicative of the redi­rection of metabolic fluxes believed to occur under xylose fermentation. Pro — teome analysis bridges the gap between genetic engineering, transcription profiles, and observed metabolism by identifying that over — or underexpres­sion of specific proteins (i. e., enzymes) are pushing targeted (or untargeted) metabolic fluxes in desired (or undesired) directions.

Proteomics is a rapidly developing area of research, whereby new technolo­gies are often developed and validated in model systems such as S. cerevisiae. Compared with genomics, however, proteomics is still limited because it is strongly biased towards highly abundant proteins and, therefore, does not yet provide the genome-wide coverage obtained by other x-ome technologies. Additionally, the proteome world is possibly the most complex of all x-omes because of its highly dynamic nature and complexity resulting from splice — variants, isoforms, and protein post-translational modifications. For some proteins, in excess of 1000 variants have been described [104]. It is evident that there is an ongoing need for improvement in (quantitative) proteomics technologies, whereby yeast will likely have its role again as the benchmark model system. Proteomics, largely absent in bioethanol development, is at the infancy of finding key roles in industrial products. Those products are likely to be targeted as co-products for bioethanol-based biorefineries. Succinic acid has already been considered as a potential added value co-product that could diversify the product portfolio of a biorefinery where the high-volume, low — value product will be bioethanol [115,116].

3.4

The amount and types of enzymes required for the saccharification of cel­lulose and hemicellulose are strongly dependent on the biomass being hy­drolyzed and the type and severity of pretreatment. Ultimately the selection of biomass feedstock will be based on local availability and economy of sup­ply. In the early stages of commercial development, feedstocks with the great­est potential for demonstrating economic viability of an integrated process are likely to be developed first. These likely will include processing residues that are already available at processing plants such as corn fiber, soybean hulls, sugarcane bagasse, wood waste, and paper mill waste. The selection of both desizing and pretreatment processes may also be strongly influ­enced by local economics, especially with regard to co-location with existing wood, coal, or municipal solid waste-burning power plants, where inexpen­sive power and steam are available. With a diversity of potential substrates, different thermochemical pretreatments will be utilized to balance accessibil­ity to enzymatic attack with destruction of valuable sugars. Variations in the severity of the pretreatment (pretreatment severity is defined as the combined effect of temperature, acidity, and duration of treatment) must also be varied to maximize both sugar and fermentation compatibility. For example, an acid pretreatment, run at high temperature, high pressure and for long periods of time is considered more severe than a neutral pH water pretreatment run under the same temperature and pressure conditions. A low severity pretreat­ment will solubilize less of the hemicellulose fraction, increasing the amount of hemicellulase enzymes required, but may also reduce the production of by-products toxic to the fermentation, increasing the ethanol yield from the fermentation.

3.1

Biological pretreatment can be performed by applying lignin-degrading mi­croorganisms, such as white — and soft-rot fungi, to the lignocellulose mate­rials [44,47]. The method is considered to be environmentally friendly and energy saving as it is performed at low temperature and needs no use of chemicals. However, the rate of biological pretreatment processes is far too low for industrial use, and some material is lost as these microorganisms to some extent also consume hemicellulose and cellulose, or lignin [42]. Never­theless, the method could be used as a first step followed by some of the other types of pretreatment methods.

4

The total production cost for cellulosic ethanol must still be substantially re­duced to enable large scale commercialization, and at least a portion of this reduction must come from enzyme cost. Realistically, enzyme cost targets in the range of $0.30/gallon at the commercial scale should be achievable in the near future by avoidance of transportation and formulation costs. In such a scenario, on-site or near-site enzyme production is essential, where enzymes are produced using reduced-cost feedstocks, transported short dis­tances, and not stored for extended periods of time. The least expensive alternative in this situation involves the direct use of whole fermentation broth (including cell mass) to circumvent expensive cell removal and en­zyme formulation steps. To investigate this possibility, we compared the use of whole fermentation broth and cell-free broth as catalysts for PCS hydro­lysis in microtiter-scale reactions at 50 °C, pH 5.0, for up to 120 h. The results of this study strongly suggest that both preparations, dosed at equal volumes, give comparable yields of reducing sugars from PCS, suggesting that costly cell removal may not be required.

6

Conclusions

The development of cost-effective enzymes for the widespread utilization of lignocellulosic biomass will require continued research and development to be successfully deployed. Although great progress has been made in identi­fying new enzyme mixes with improved catalytic efficiency, improvements in enzyme yield, and improved enzyme production economics, much work remains. There are thousands of organisms involved in the natural decom­position of plant material in our biosphere, and only a fraction of those have been isolated or investigated. Since these organisms work collectively to de­grade biomass, better enzymes, with greater synergies, will be uncovered with additional work. Future efforts will also likely require the use of directed evolution techniques to collectively optimize enzymes to perform under con­ditions more compatible with the fermentation organisms used to produce ethanol and other products. In the short term, there are also great strides to be made in the area of process integration. Here, closely coupling the steps of pretreatment, hydrolysis, and fermentation has the potential to significantly increase overall process efficiency and reduce cost.

Acknowledgements Our thanks to the members of the National Renewable Energy Lab­oratory who kindly supplied acid pretreated corn stover, numerous methods of analysis, and many helpful discussions over the course of our collaboration and to the US Depart­ment of Energy for funding much of the work described.