There are numerous organisms that rely on biomass degradation for their survival, often existing in the natural environment as a complex consortia of fungi, bacteria, and protozoa, working synergistically to decay the plant cell wall. All of these organisms are potential sources of enzyme discovery, but current commercial products for biomass treatment are derived from fungi because these organisms produce a complex mix of enzymes at high produc­tivity and catalytic efficiency, both of which are required for low-cost enzyme supply. Unlike most bacteria, which express complexes of many carbohydrate­degrading activities arrayed on molecular scaffolds physically attached to the bacterial cell wall, fungal cellulases are typically secreted into the growth medium, allowing cost-efficient separation of the active enzymes in a liquid form suitable for delivery to a hydrolysis reactor.

4.1

The S. cerevisiae genome contains the gene XKS1 coding for XK [26,27], but the XK activity in wild-type S. cerevisiae is too low to support ethanolic xylose fermentation in strains engineered with a xylose pathway [26,99,100]. It is only when additional copies of XKS1 are expressed that recombinant xylose­utilizing S. cerevisiae produce ethanol from xylose [79] (strain TMB3001, Tables 1-4), [100] (strain H1691, Table 1), [101] (strain 1400 (pLNH32), Table 2; strain H2490, Tables 3 and 4). However, nonphysiological or unreg­ulated kinase activity may cause a metabolic disorder [102]. It was indeed experimentally demonstrated that only fine-tuned overexpression of XKS1 in S. cerevisiae led to improved xylose fermentation to ethanol [103,104]. Sim­ilarly, it was shown that arabinose-utilizing recombinant S. cerevisiae strains expressed a mutated L-ribulokinase gene with lower specific activity, indicat­ing that a low kinase activity had been selected as advantageous for arabinose utilization [71].

4.4

Den Haan et al. [49] calculated that a 20- to 120-fold improvement in CBH ex­pression, as well as simultaneous high-level expression of other cellulase com­ponents, will be necessary for slow growth on crystalline cellulose. This calcu­lation assumes a strain that can grow at 0.02 h-1 has a typical anaerobic yield of 0.1 gbiomass/g substrate or an aerobic yield of 0.45 gbiomass/gsubstrate, that the expressed cellulase has a specific activity which is the same as that of crude T. reesei cellulase on avicel (0.6 U/mg), and that CBH1 would make up the same fraction of total cellulase protein as in the T. reesei system [9]. While techniques for rational design of cellulases for improvement in ex­pression level and potentially specific activity will be important to achieving this goal, techniques involving random natural and/or induced mutation will also play an important role. The well-established success of directed evolu­tion techniques for enzymes and enzyme systems (e. g., see reviews in [132, 133]) can be transferred to engineering organisms for CBP, although this application does present unique challenges due to the lack of a good high — throughput screening technique for activity on insoluble cellulosic substrates. On the other hand, the natural connection between cellulase expression and growth on cellulose for CBP organisms makes whole cell selection-based strategies for improving cellulase production a powerful way to screen very large libraries of candidate cells, mimicking the evolutionary process found in nature.

An assumption for any selection-based improvement for CBP organisms is that mutations can result in increased cellulase activity expression. For total cellulase activity such mutations would increase either the per cell expression level (g cellulase/g cell) or the cellulase specific activity (U/mg cellulase). Mu­tagenesis and screening techniques have allowed researchers to isolate strains of S. cerevisiae with “super-secreting” phenotypes [134-136], and similar techniques for the expression of individual cellulase components have been successful [137]. Also, random mutation has been used to change the prop­erties of cellulase enzymes (e. g., [138-140]; a further review can be found in [141]), although to our knowledge enhanced overall specific activity of cellulase on insoluble substrates has not been demonstrated via directed evo­lution. However, the specific activity of a mixture of cellulases also depends on the relative amounts of cellulase components to achieve the highest degree of enzyme-enzyme synergy [142], as well as other parameters (as yet not elu­cidated) that determine enzyme-microbe synergy [6]. These features could be impacted by mutation and therefore lead to enhanced specific activity of cellulase systems expressed by recombinant cellulolytic CBP organisms.

Earlier in this review (Sect. 3) the relationship between cellulase activity and growth rate was examined from a whole-population perspective, using parameters that are averages for many cells. The relationship between growth rate and cellulase expression for an individual cell, especially a cell harbor­ing mutations affecting cellulase expression, as compared to other cells in the population depends on the diffusion of the soluble reaction products from the point they are created at the cellulose surface to the point they are taken up by a particular cell. When a connection between growth rate and en­zyme production can be established, selection in liquid culture—particularly continuous culture—has the potential to screen many cells. For example, if a continuous reactor had a cell concentration of 1010 cells/L and was operat­ing at a dilution rate of 0.02 h-1, then 108 cells/(L*h) would be screened, and a 100-h continuous culture would screen 1010 cells.

The power of this system has been recognized previously (see [143-145] for reviews) and demonstrated in many examples where the enzyme of in­terest is located intracellularly [146-153], including some cases where the limiting enzyme made up 25% of the total cellular protein after selection, an approximately fourfold increase in expression in both cases [154,155]. In a very recent study, the authors were able to create a strain of S. cerevisiae capable of utilizing xylose as the sole carbon source with a 6-h doubling time without using recombinant genetic techniques—only using selection on xylose minimal media from a strain that could grow only very poorly ini­tially [156]. For secreted enzymes (both cell-associated or extracellularly), far fewer studies have shown improvements via selection in liquid culture. Francis and Hansche [157] were able to isolate a mutant of S. cerevisiae in a well-mixed chemostat with 1.7-fold improvement in acid phosphatase activ­ity, and Naki et al. [158] were able to isolate mutants of Bacillus subtilis with about fivefold increased secretion of protease by growing the cells in a hollow fiber apparatus, which physically separated cells, with bovine serum albumin as the limiting nitrogen source. Therefore, understanding the physical char­acteristics of the cell/enzyme/substrate system and the resulting magnitude of differences in growth rate between mutants is critical to applying selection to this system.

When cells are grown on solid media, with significant space between ini­tial cell colonies, those cells that produce more or better cellulase will retain the products of their hydrolytic reactions, and will form larger colonies. This technique—selection by people judging the size of colonies—has the ad­vantage of maintaining separation between cells. It has the disadvantage of limiting the number of cells that can be screened. It is hard to imagine how more than 109 cells (103 colonies/plate* 106 plates) can be screened in a rea­sonable amount of time, even utilizing high-throughput approaches.

When cells are grown in well-mixed liquid culture, the situation is much different because the products of hydrolysis are free to diffuse. A schematic representation of some of the liquid culture cases relevant to recombinant cellulolytic CBP organisms is presented in Fig. 4. In case A, where cellulase enzymes are secreted away from the cell, cellulases with cellulose binding do­mains will diffuse to cellulose, bind to it, and release soluble hydrolysis prod­ucts. In the final step of the overall hydrolysis reaction secreted в-glucosidase converts soluble glucose oligomers into glucose (an overview of fungal cellu­lase systems can be found in [1]). Lelieveld [159] predicted that in cases such as this, the limiting enzyme will form a gradient in the diffusion boundary layer around the cell, creating a gradient of the limiting nutrient as well. Such a gradient would provide a link between mutations conferring increased en­zymatic activity and supply of the limiting nutrient to the cell. With respect to selecting CBP organisms, when a cell secretes a growth limiting cellulase that binds cellulose, it will not necessarily take up the products of the reaction preferentially compared to another cell in the population. Thus, increased activity of that cellulase cannot be selected for. The remaining question is whether the postulated gradient of в-glucosidase exists, and if so what is the effect of the glucose gradient (ДА) on a mutant’s growth rate compared to a parent strain producing less в-glucosidase. Fan et al. [160] used a 2-D reac — tion/diffusion model to predict that the differences in growth rates between mutants and parents in this case are too small to allow the mutant to outgrow the parent in a reasonable length of time.

Recombinant xylanases and cellulases can also be expressed as tethered enzymes [59,119,130] (Fig. 4, cases B and C). In the case where a cell does not bind to the cellulose substrate (case B) (e. g., cellulases with cellulose bind­ing domains are not tethered to the cell surface), the limiting enzyme reaction is once again в-glucosidase conversion of cello-oligomers to glucose. The в — glucosidase enzyme is concentrated at the cell surface, setting up a larger gradient (ДВ) than in case A. However, in this case Fan et al. [160] found that unless the Monod constant (kS) for the substrate was very low, this gradi­ent would not be large enough to allow mutants to outgrow parents in liquid culture.

Case C presents the situation when cellulases are tethered to the surface and the cell binds to a substrate particle. In this case, the particle acts to trap the hydrolysis products, creating a substantial difference between the glucose

concentration in this gap and the substrate and the bulk fluid. When this cell/enzyme/substrate relationship is operative, Fan et al. [160] predict that differences in enzyme expression level will lead to differences in growth rates between mutant and parent cells, and that this will allow selection-based pop­ulation changes to occur in a reasonable amount of time. To date, the promise of selection for improving cellulase production by recombinant cellulolytic microorganisms has not been realized, and knowledge of the local concen­tration of glucose around such cells is limited to prediction. However, it is known that cellulose hydrolysis by naturally occurring cellulolytic microor­ganisms occurs much faster when mediated by cells adhering to the substrate as compared to nonadherent mutants [161]. It has been suggested that adher­ence confers a competitive advantage associated with first access to hydrolysis products.

7

From the investigations presented above it has been concluded that the max­imum yields of mannose, the main hemicellulose sugar in softwood, and of glucose are not obtained at the same degree of severity. The optimal yield of mannose is obtained at a lower severity than that required for maximum di­gestibility of the cellulose in the subsequent enzymatic hydrolysis step. This suggests two-stage steam pretreatment, in which the first stage is performed at low severity to hydrolyse the hemicellulose, and a second stage, at a higher degree of severity, in which the solid material from the first step is pretreated again [82].

Although there are several studies on two-stage acid hydrolysis of soft­wood, the number of studies on two-stage steam pretreatment is scarcer. Soderstrom et al. [85-87] performed a thorough investigation on the two — stage pretreatment of spruce using either SO2 impregnation or H2SO4 im­pregnation in both steps, as well as H2SO4 in the first stage and SO2 in the second. The highest sugar yields were achieved for two-step pretreat­ment with either SO2 impregnation or H2SO4 impregnation in both steps (see Fig. 2). A wide range of pretreatment conditions resulted in similar sugar yields of about 50 g per 100 g raw material.

The highest sugar yield was 51.7 g per 100 g, corresponding to 80% of the theoretical, obtained for pretreatment conditions of 190 °C for 2 min and 210 °C for 5 min. This yield (in %) is slightly lower than that reported by Nguyen et al. [ 14]. However, the amount of sugar obtained expressed as grams per 100 g dry raw material is higher. Ngyuen et al. stated that they obtained a sugar yield of 82%, which, in their case, corresponds to 46 g/100 g dry raw

material. They used a cellulase activity of 60 FPU/g cellulose, which is more than twice the amount that was used in the study shown in Fig. 2. The max­imum overall sugar yield obtained with two-step pretreatment using H2SO4 in both stages was only slightly lower, 77% of theoretical.

Besides overall sugar yield it is also of importance to investigate the fer — mentability of the pretreated materials. Impregnation with dilute H2SO4 fol­lowed by pretreatment at a high combined severity (i. e. high temperature and/or long residence time) resulted in materials that were not fermentable. Impregnation with SO2, however, was successful in creating fermentable ma­terials for all investigated pretreatment severities.

The two-step pretreatment results in a higher ethanol yield than does the one-step pretreatment, and it has also the advantage of lower requirement of enzymes and water in the SSF step. Major drawbacks are, however, the higher capital cost and the higher energy consumption. In a study by Wingren et al. [90] the overall ethanol production cost was shown to be very much de­pendent on the way the two pretreatment steps are performed, especially if the pressure is released or not between the steps, and also on the dry matter (WIS) content in the second step. The lowest cost estimated for the two-stage process, 3.90 SEK/L, which was about 6% lower than that for the one-stage process, requires a high ethanol yield, high concentration of WIS in the fil­ter cake between the steps, and that the sugars being fed to the second step will not become degraded. The higher yield has been demonstrated experi­mentally, but the two other assumptions still need to be verified on the pilot scale.

5

Conclusions

In conclusion, a large number of pretreatment methods have been investi­gated and developed during the last 10 years, resulting in high recovery of sugars and rather high overall ethanol yields. However, most of the results were obtained in studies using batch-operating equipment on a rather small scale. Enzymatic hydrolysis has also, in most cases, been assessed at low sub­strate concentration.

One problem with the data produced so far is the difficulty in comparing methods, as the assessment is performed in different ways. In most cases the pretreatment is not assessed under realistic process conditions. The whole process must be considered as the various pretreatment methods give dif­ferent types of materials: hemicellulose sugars can be obtained either in the liquid as monomer or oligomer sugars, or in the solid material to various ex­tents; lignin can be either in the liquid or remain in the solid; the composition and amount/concentration of possible inhibitory compounds also vary. This will affect how the enzymatic hydrolysis should be performed (e. g. with or without hemicellulases), how the lignin is recovered and also the use of the lignin co-product.

For agricultural residues a large number of pretreatment methods result in high sugar yields while for wood, and especially softwood, the number of feasible methods is smaller. Acid hydrolysis and steam pretreatment with acid catalyst seem to be the methods that can be used for all types of raw ma­terials, but the drawback is the high equipment cost and the formation of inhibitors. This requires further improvement and also a better integration with the enzymatic hydrolysis development, as improved enzyme mixtures may lead to less severe pretreatment conditions and thereby lower cost and reduce formation of inhibitory compounds.

To verify the technology the next step is to implement all of these improve­ments in a pilot-scale process with all steps integrated into a continuous pilot plant. This will provide better data for assessment and for scale-up to a demo­or full-scale process. It will also give better information on how various pre­treatment conditions will affect all the other process steps, i. e. enzymatic hydrolysis, fermentation, downstream processing and wastewater treatment, as well as product and co-product quality.

The enzymatic hydrolysis of the pretreated raw material and the fermenta­tion of the hydrolysed sugars can be performed separately or simultaneously, commonly referred to as SHF (separate hydrolysis and fermentation) or as SSF (simultaneous saccharification and fermentation). The SSF process con­figuration has been generally considered more favourable for reducing the ethanol production costs [72,81]. The hydrolysis rate in the separate hydro­lysis is strongly inhibited by the accumulation of the end products, cellobiose and glucose [60]. In the simultaneous hydrolysis and fermentation, the end product inhibition is alleviated by the continuous removal of glucose by the fermenting organism. In the separate hydrolysis and fermentation the most severe end product inhibition caused by cellobiose has been overcome by adding an adequately high amount of в-glucosidase. For the same reason, the enzyme dosage needed is obviously lower in the SSF. Other claimed ad­vantages of the SSF are the lower risk of contamination and reduction of investment costs by combined reactors. The low concentration of free glucose and the presence of ethanol make it more difficult for contaminating micro­organisms to take over the fermentation and decrease the ethanol yield. The drawback of the SSF is that the conditions, i. e. the pH and temperature of the hydrolysis and fermentation, are suboptimal in a combined process. The optimal temperature for the enzymatic hydrolysis is clearly higher than that of the presently used fermenting organisms. Another drawback of the SSF is the difficulty in optimising the fermentation of techniques, i. e. by running continuous fermentation or recirculating and reusing the yeast due to the presence of the solid residues from the hydrolysis.

To improve the overall process economics and to achieve a faster hydrolysis rate by using thermostable enzymes, various modifications of the present process configurations can be considered (Fig. 1). After the pretreatment, the temperature of the substrate is high, and is reduced to achieve the operat­ing temperature in the following process stages. In the traditional SSF, the temperature is about 35 °C. In a separate hydrolysis and fermentation pro­cess, the first total hydrolysis stage is carried out at about 45-50 °C with the present commercial enzymes, or above 60 °C with novel thermostable

enzymes. Other options include a partial prehydrolysis at higher tempera­tures, denoted as liquefaction, where the viscosity of the substrate is de­creased using a chosen composition of thermostable cellulases based on one or several enzymes. The liquefaction stage, i. e. an enzymatic treatment improving the rheological properties (improved flowability, reduced viscos­ity) of the slurry, can significantly improve the mixing properties of the substrate slurry [83]. This partial hydrolysis can be carried out even with the limited number of thermostable cellulolytic and hemicellulolytic en­zymes available. Using a set of thermoactive enzymes in the prehydrolysis, it was possible to reduce the viscosity and increase the sugar formation [83]. The high viscosity is a consequence of a high initial substrate consistency, needed to achieve a high final sugar and ethanol concentration and to de­crease the distillation costs [69]. With a theoretical ethanol yield of 25-30% of the raw material, the raw material consistency should be at least 15% (d. w.) to reach an ethanol concentration of 4-5%. Some of the technical obstacles related to high consistency can thus be overcome by a rapid de­crease of viscosity. After a liquefying partial hydrolysis, the saccharification stage using a complete or complementary set of hydrolytic enzymes, ei­ther simultaneously or separately from the fermentation (SSF or SHF), can be carried out. A separate hydrolysis stage (SHF) can be carried out at el­evated temperatures with the complete set of hydrolytic thermostable en­zymes needed for a chosen substrate. Finally, thermostable enzymes could be supplemented to bacterial fermentations using anaerobic, ethanol pro­ducing strains, such as Clostridia, to improve their conversion rate of cellu — losic substrates into sugars (SSF or consolidated bioprocessing). Thus, new thermostable enzymes would allow the design of more flexible process con-

figurations, based on the availability of novel thermostable lignocellulolytic enzymes.

The performance of chosen thermostable cellulolytic enzymes with present commercial fungal enzymes was compared in this paper. The refer­ence enzyme preparations contain the whole set of cellulolytic enzymes, i. e. cellobiohydrolases and endoglucanases, as well as several hemicellulolytic ac­tivities and в-glucosidases. These enzymes work at temperatures up to about 45 °C in long-term hydrolysis conditions and up to 50 °C in short-term con­ditions. New enzyme compositions were designed and tested in the hydrolysis of various steam pretreated raw materials.

5

In conclusion, the domestication of S. cerevisiae for carbon dioxide and ethanol formation from hexose sugars has led to the fact that the metabolism of hexose and pentose sugars in this yeast are fundamentally different. As evidenced by genome-scale transcriptome and proteome analyses of numer­ous recombinant pentose-utilizing S. cerevisiae strains, the difference is not only limited to the initial sugar conversion pathways, but also comprises the central metabolism and the glycolytic pathway. The major future challenge remains to translate the knowledge acquired from laboratory strains to indus­trial production strains.

Acknowledgements The authors acknowledge the financial support from the Swedish En­ergy Agency.

Today, industrial biotechnologists are no longer discussing whether a single product can be produced via biotechnology, but are rather considering di­verse portfolios of products leveraging expertise and resources. The various systems biology toolboxes applied to bioethanol production are now being exploited to develop integrated processes that will form “biorefineries”. The concept of the biorefinery was first defined in 1999, when it was postulated that lignocellulosic raw materials may be converted to numerous biocom­modities via integrated unit processes, and offer competitive performance to traditional petrochemical refineries [27]. Several chapters in this volume will more closely examine the biorefinery as a model and platform for future bio-based processes in terms of policy issues and process integration.

If the biorefinery platform model is to evolve from academic conception to industrial reality it will require two key components. First, the economic and socio-political landscape must support and warrant the significant financial investment, favorable legislative policy, and consumer-driven demand that will be required. Second, the advances and tools developed within systems biology for metabolic engineering must be successfully applied in commer­cial environments. Bioethanol is the first industrial biotechnology product to demonstrate that if these two elements are co-supported, then numer­ous bio-based processes can be developed and integrated into a biomass economy.

2

The assignment of specific substrate factors that render a substrate recalci­trant to cellulase hydrolysis is a controversial subject. Crystallinity, DP and specific surface area have all been thought to undergo significant changes during pretreatment, consequently influencing subsequent hydrolysis [2]. The original work prior to the 1990s, focusing on the physical character­ization of substrates, has been presented in previous reviews [2,7]. More recently, our group and others have published work correlating the physical properties of wood pulp fibers employed in papermaking to their hydrolyz — ability, as cellulases have been explored as a potential means to improve both the drainage of recycled pulps and to enhance pulp fiber properties [101]. Admittedly, pulps are not identical to substrates pretreated specifically for subsequent hydrolysis by cellulases. However, it should be noted that, sim­ilarly to pretreated substrates, pulp fibers also represent a lignocellulosic matrix and that the general principles gained from examining the physical properties of pulp fibers that affect hydrolysis can be cautiously extrapolated to substrates pretreated for bioconversion.

5.1

In naturally xylose-utilizing bacteria, D-xylose is isomerized to D-xylulose [18] by xylose isomerase (XI). Xylulose is then phosphorylated to xylulose 5-phosphate [19], which is an intermediate of the pentose phosphate pathway (PPP). A similar pathway has been found in an anaerobic fungus [20]; how­ever, most naturally xylose-utilizing fungi contain a more complex pathway consisting of reduction-oxidation reactions involving the cofactors NAD(P)H and NAD(P)+ (Fig. 1). Xylose is reduced to xylitol [21-23] by a NAD(P)H — dependent xylose reductase (XR), and xylitol is then oxidized to D-xylulose by a NAD+-dependent xylitol dehydrogenase (XDH) [22,24,25]. As in bacte­ria, xylulose is phosphorylated to D-xylulose 5-phosphate by a xylulokinase (XK) [26,27]. Despite the inability of S. cerevisiae to utilize xylose, the genes encoding the reductive-oxidative xylose pathway enzymes XR, XDH, and XK are present in its genome [26,28,29]; however, they are expressed at too low levels to allow xylose utilization. Even when the genes were overexpressed, no growth on xylose could be detected [30]. Neither was it possible through adaptation protocols to upregulate the expression of these genes to levels high enough to allow significant xylose fermentation [31].

2.2

After the proof of principle of XI expression in S. cerevisiae, not only metabolic engineering, but also evolutionary engineering was applied to im­prove the rate of D-xylose utilisation of a strain solely over-expressing XI [44]. Since improvement of the aerobic consumption rate was initially the target of this selection experiment, serial transfer in a shake flask was chosen as the cultivation condition of this evolution run. Indeed, after 30 serial transfers, the specific growth rate of this culture improved drastically (24-fold) from 0.005 h-1 to 0.12 h-1 (Fig. 8). However, a strain isolated from this selection ex­periment was not yet capable of anaerobic growth. Therefore, an additional ten selection rounds were performed in oxygen-limited batch cultures, fi­nally followed by ten cycles in an anaerobic sequencing batch reactor. From this culture a single colony was isolated (named RWB 202-AFX, for anaero­bic fermentation of D-xylose based on strain RWB 202) and used for further characterisation of the end product of this evolutionary engineering.

It was shown that only the expression of a XI, followed by evolutionary en­gineering for anaerobic growth, can also result in a S. cerevisiae strain that can grow on 2% D-xylose as the sole carbon source, with a growth rate of 0.03 h-1 in anaerobic batch fermentations [45]. However, although this strain displayed a good ethanol yield on D-xylose (0.42 gg-1) and very low pro­duction of xylitol (2.8 mM), the obtained growth rate, and therefore ethanol

production rate, were insufficient to allow economically viable industrial ap­plication. During these batch cultivations, small amounts of D-xylulose (up to 8 mM) were still excreted into the broth, indicating that evolutionary engin­eering alone did not fully overcome the metabolic limitations downstream of this metabolite. This result indicates that although evolutionary engineering is a very powerful tool, it has limitations and, in this case, the combination of knowledge-based metabolic engineering (Sect. 5) combined with evolution­ary engineering (Sect. 6.1) resulted in more desirable attributes and higher ethanol production rates.

7