The first xylose-utilizing strains of S. cerevisiae were generated by expressing the Pichia stipitis genes XYL1 [23] and XYL2 [24], encoding XR and XDH, re­spectively [46-48]. P stipitis was chosen as the source of the heterologously expressed enzymes because it produces ethanol from xylose with theoret­ical yield, albeit only under well-controlled oxygen limitation [47,49,50], while most other naturally xylose-fermenting yeasts produce considerable amounts of the by-product xylitol [50]. Xylitol formation is a consequence of the inability of the cell to oxidize reduced cofactors in the absence of oxygen [32]. Contrary to XRs from most xylose-utilizing yeasts, XRs from P stipitis, Pachysolen tannophilus, and Candida shehatae can use not only NADPH but also NADH as a cofactor [21], which permits recirculation of the cofactors between the first two steps of the xylose pathway (Fig. 1).

Nevertheless, the first S. cerevisiae strains expressing the P stipitis XR and XDH produced xylitol, and the ethanol yield from xylose was low [47,48]. This was ascribed to the preference for NADPH over NADH of the XR [23]. Much research has been devoted to developing metabolic engineering strate­gies to improve xylose fermentation by XR — and XDH-carrying strains, often guided by the early suggestions to express either a strictly NADH-specific XR activity [32] or to express a transhydrogenase activity [21]. Both approaches are further discussed in the following sections together with other metabolic engineering strategies. Kinetic modeling estimated that the conversion of xylose to xylulose required a ratio of 1:10 of the initial XR and XDH activ­ities [51], which has been experimentally supported by several independent investigations [51-54]. The higher level of XDH is necessary to “pull” the xy­lose toward central metabolism [55], especially since the equilibrium of the XDH reaction favors xylitol formation [56]. In addition, it has more recently been found that efficient xylose metabolism requires high activity of both XR and XDH [54,57].

3

Arabinose

3.1

Wheat straw is an abundant lignocellulosic crop residue with potential as a feedstock for ethanol production, especially in Canada and Europe. Wheat straw hydrolysate was therefore selected as one of the fermentation feed­stocks for evaluating the fermentation characteristics of S. cerevisiae RWB 218 under industrially relevant conditions (W. de Laat, unpublished data). Wheat straw was pretreated using steam explosion (Sunopta, Canada). The pulp thus obtained was then hydrolysed enzymically at pH 5.0 with cellulases and hemicellulases, yielding a hydrolysate that contained 50 g L-1 glucose, 20 g L-1 D-xylose, 6 gL-1 arabinose and 6 gL-1 of disaccharides (cellobiose, melibiose, maltose and sucrose, indicated as DP-2 in Table 2). The hydrolysate, which

also contained 3 gL-1 acetic acid and 0.3 gL-1 of lactic acid, was supple­mented with 0.4 gL-1 of (NH4)2PO4 as a combined source of nitrogen and phosphate. Fermentations were run at 32 °C, with an initial pH of 4.8.

When batch cultures on the wheat straw hydrolysates were inoculated with 1.5gL-1 of S. cerevisiae RWB 218, most of the available sugars were con­verted within 55 h (Table 2). The yield of ethanol on the consumed sugars was very high and, towards the end of the fermentation, even approached the theoretical maximum yield of 0.51 gg-1. This very high apparent yield might partially be caused by the additional hydrolysis of some oligosaccharides or by the presence of other sugars that were not identified in the analyses. Xylitol formation was not observed.

Even when a much lower initial biomass concentration of 0.1 gL-1 was used, S. cerevisiae RWB 218 reached the same degree of conversion in 80 h. Addition of vitamins, trace elements and/or the anaerobic growth factors Tween-80 and ergosterol [2,3] did not result in a faster fermentation. This demonstrates the modest nutritional requirements of S. cerevisiae during fer­mentation of hydrolysates of lignocellulosic materials, which often contain very low levels of nutrients required for microbial growth.

7.3

A multitude of different pretreatment methods have been suggested during the past few decades. They can loosely be divided into different categories: physical (e. g. milling, grinding and irradiation), chemical (e. g. alkali, dilute acid, oxidizing agents and organic solvents), physicochemical (e. g. steam pre — treatment/autohydrolysis, hydrothermolysis and wet oxidation) and biologi­cal, or combinations of these. In general, it is difficult to place the methods into one category only.

A rough classification of the pretreatment methods can also be made ac­cording to the following:

• Acid-based methods, i. e. pretreatment at low pH, result in hydrolysis of the hemicellulose to monomer sugars and minimize the need for hemicel — lulases.

• Methods working close to neutral conditions, e. g. steam pretreatment and hydrothermolysis, solubilize most of the hemicellulose due to the acids re­leased from the hemicellulose, e. g. acetic acid, but do not usually result in total conversion to monomer sugars. This thus requires hemicellulases acting on soluble oligomer fractions of the hemicellulose.

• Alkaline methods leave a part of the hemicellulose, or in the case of ammonia fibre explosion (AFEX), almost all hemicellulose in the solid fraction. This then requires hemicellulases acting both on solid and on dissolved hemicellulose. An alternative is to perform an acid hydrolysis of this fraction, for instance after removal of the cellulose by enzymatic hydrolysis.

This affects, of course, not only the method that should be used for assess­ment of the pretreatment but also the cost of the overall hydrolysis of the carbohydrates.

3.1

An “efficient” cellulase system requires sufficient в-glucosidase (BG) to hy­drolyze cellobiose produced by the action of the CBHs to prevent their prod­uct inhibition [38]. The addition of BG to a complex cellulase mix such as the Novozymes Celluclast 1.5 L dramatically improves the extent and, during the later stages of hydrolysis, the rate of cellulose saccharification. This is re-

flected in Fig. 6, where the T. reesei strain used to produce Celluclast 1.5 L was compared to the same strain expressing Aspergillus oryzae BG in hydro­lysis assays. Due to relief of the product inhibition at high solids loadings (13.5% w/w in this example), the amount of total enzyme protein required to hydrolyze 80% of the cellulose to glucose was reduced by nearly twofold. At this solids loading, the beneficial effect of BG addition was saturated when it reached ~ 5% of the total enzyme protein, but higher solids would require higher BG levels or a more active BG.

4.2.2

The P. stipitis XR, which converts xylose to xylitol, prefers the cofactor NADPH over NADH by a factor of approximately 100 [23]. In yeast, NADPH is primarily formed in the oxidative PPP converting glucose-6-phosphate to ribulose-5-phosphate. Therefore, genes coding for enzymes in the oxidative PPP were deleted in order to decrease NADPH concentration in the cell and

thus force XR to use NADH instead of NADPH, which was demonstrated by the deletion of ZWF1, coding for glucose-6-phosphate dehydrogenase (G6PDH) [114] (strain TMB3255, Table 2), [115] (strain H2723, Table 1). In­creased ethanol formation at the expense of not only xylitol formation but also the xylose consumption rate was observed [114] (strains TMB3001 and TMB3255, Tables 2 and 3). In a follow-up study, the G6PDH activity was in­stead fine-tuned, which enabled the design of strains with increased ethanol yield and reasonable xylose consumption rate [116] (strains TMB3256 and TMB3037, Table 2, Fig. 4). However, in an industrial context, it is worth notic­ing that the ZWF1 deletion increases the sensitivity toward lignocellulose hydrolysates, possibly due to the limited intracellular NADPH concentration, which is important for inhibitor tolerance [116,117].

4.5.2

Historically, Saccharomyces has served as the main biocatalyst for commer­cial ethanol production. Considering that Saccharomyces and Z. mobilis are naturally ethanologenic, these organisms are obvious candidates for ethanol production. However, both organisms lack the native ability to utilize pentose sugars, the major component of the hemicellulose fraction of biomass [9,10]. Though E. coli lacks the native ability to produce ethanol as the major fer­mentation product, it utilizes both hexose and pentose sugars [11] and the uronic acid constituents of pectin [12]. The breadth of carbohydrates metab­olized, extensive background of knowledge, and ease of genetic manipulation made E. coli an obvious choice for metabolic engineering of a microbial bio­catalyst for production of ethanol from lignocellulose.

2.1

Over the past 20 years, our group has looked at steam pretreatment (SP) with regard to its suitability for pretreating a range of lignocellulosic sub­strates, the subsequent ease of enzymatic hydrolysis of the cellulosic stream and the recovery of most of the hemicellulose sugars and lignin in a use­able form. SP is an attractive pretreatment process as it makes limited use of chemicals, requires relatively low levels of energy and, depending on the conditions employed, results in the recovery of most of the original cellulose — and hemicellulose-derived carbohydrates in a fermentable form [23-28]. As will be discussed in more detail, SP disrupts the lignin barrier [29] and facili­tates access of cellulases to the cellulose fibers [30]. Previous work has shown that SO2-catalyzed SP is an effective pretreatment for softwood [26-34], hard­wood [35-37] and agricultural residues [22] and that impregnation of SO2 prior to pretreatment results in lower treatment temperatures and shorter reaction times, thereby improving hemicellulose recovery and reducing the
formation of sugar degradation products [23]. It has also been shown that SO2 impregnation prior to SP enhanced the carbohydrate hydrolysis rate by increasing the accessibility of cell walls via the formation of fractures and the removal of hemicellulose during the steaming of the substrate [28], while re­ducing the DP of the oligomers and increasing the proportion of monomers

♦

in the water-soluble stream [31,38-40]. The “severity” of SP can be summa­rized by a single factor called Ro (Ro = t exp(T — 100)/14.75) which links the effects of time (t, min) and temperature (T, °C) [41]. Due to the high tem­perature and acidic conditions employed during the SP process, depending on the “severity” (temperature, time, pressure, catalyst dosage) of the treat­ment, a portion of the hemicellulose-derived sugars and solubilized lignin fragments can be degraded or transformed into compounds such as furfural and 5-hydroxymethylfurfural (5-HMF); aliphatic acids, such as acetic, formic, and levulinic acid; and phenolic compounds [42]. It is known that these compounds can inhibit both downstream hydrolysis by cellulases [43] and fermentation of the liberated sugars to ethanol [44]. Therefore, compromised SP conditions have to be defined that provide an easily hydrolyzable cellu — losic substrate, good recovery of the hemicellulose-derived sugars, ideally in a monomeric form, while minimizing the formation of inhibitors. Ideally, a reactive lignin stream, with a higher economic application than its intrinsic fuel value should also be obtained.

It is apparent that the nature of the substrate and the pretreatment method used has at least as much influence on the ease of enzymatic hydrolysis as does the nature and efficiency of the enzyme system used to conduct hydrolysis. As illustrated in Fig. 1, the efficiency of enzymatic hydrolysis of a given lignocellu — losic substrate is the result of interplay of various factors. Although it is evident that substrates such as agricultural residues are generally less recalcitrant than softwood residues, it is recognized that enzyme — or substrate-related factors that govern effective hydrolysis can be controlled to an appreciable extent by the type and conditions of the pretreatment employed.

In the sections below we will describe how pretreatment, specifically SP, in­fluences the characteristics of the substrate and the subsequent recovery of the cellulose, hemicellulose and lignin components. Recent progress in eluci­dating the role of substrate properties such as crystallinity, DP, pore volume, and available surface area in enzymatic hydrolysis will also be discussed using wood pulps as “model substrates”. The final section offers concluding re­marks and outlines the remaining challenges associated with understanding the progress of enzymatic hydrolysis during the bioconversion process.

2

The performance of the thermostable enzyme mixtures was studied in hydro­lysis experiment in test tubes (5 mL). Enzyme mixtures were dosed on the basis of FPU activity to the substrates (10 gL-1 dry matter) suspended in 50 mM sodium acetate, pH 5. The standard enzyme dosage was 10 FPUg-1 cellulose. Triplicate samples were incubated with mixing at 35 °C, 45 °C, 55 °C or 60 °C for 24 h, 48 h or 72 h. Reference samples with inactivated enzymes and corresponding substrates were also prepared.

Chemical Analysis

The release of hydrolysis products was measured as reducing sugars assayed by the DNS method using glucose as standard [10]. The results were cor­rected by taking into account the blank samples containing corresponding amounts of inactivated enzymes and substrate. The mono — and oligosac­charides formed were also analysed by high-performance anion-exchange chromatography on a Dionex 4500i series chromatograph with pulsed amper — ometric detection (HPAEC-PAD), as described earlier [75].

6

In contrast to S. cerevisiae, many yeast species are capable of utilising xy­lose as the sole carbon and energy source for respiratory growth. However, only few of these yeasts are capable of fermenting xylose to ethanol under oxygen-limited conditions, such as for instance Pichia stipitis and Pachysolen tannophilus [65].

Maybe not surprisingly, xylose-metabolising yeasts have predominantly been isolated from wood-related environments. The pathway for D-xylose metabolism used by these yeasts to convert D-xylose to D-xylulose was first described in 1955 [25] and involves a two-step conversion that involves two oxidoreductases (Fig. 1): xylose reductase (EC 1.1.1.21) and xylitol dehydro­genase (EC 1.1.1.9). The xylose reductase has a strong preference for NADPH, whereas the subsequent oxidation of xylitol via xylitol dehydrogenase pro­duces NADH (Table 1).

Clearly, this difference in cofactor specificity can result in redox imbalance. To generate the NADPH for the xylose reductase reaction, part of the D-xylose carbon must be directed through the oxidative pentose phosphate pathway (involving the glucose-6-phosphate dehydrogenase and 6-phosphogluconate dehydrogenase reactions). While this results in a loss of some carbon as CO2,

which goes at the expense of the ethanol yield on D-xylose, it enables the efficient regeneration ofNADPH [16,32,45,69].

However, the cells have to take additional measures to reoxidise the ex­cess NADH generated in the xylitol dehydrogenase reaction. In the presence of oxygen, this excess NADH can be reoxidised by respiration. This will re­quire accurate dosage of oxygen to prevent full respiration of D-xylose. Such accurate control is difficult to envisage in large-scale processes for ethanol production, which should preferably involve a minimum of aeration to reduce costs.

Under anaerobic conditions, reoxidation of excess NADH can be ac­complished via the production of compounds that are more reduced than D-xylose, such as xylitol and/or glycerol. The production of xylitol occurs via xylose reductases, which have a dual co-enzyme specificity and thereby can also use NADH, or alternatively via other aspecific reductases. As this mech­anism involves the consumption of one D-xylose for each NADH generated, it has a tremendously negative impact on the ethanol yield from D-xylose [45]. Glycerol production is a well-known redox sink during hexose fermenta­tion and especially under anaerobic conditions, but requires both carbon and ATP [67].

The preference of xylose reductase for NADPH is not only species — but also strain-dependent (Table 1). The in vivo ratio ofNADPH over NADH utilisa­tion by xylose reductase and the redox balance requirements determine the

requirement for NADH sinks such as xylitol and glycerol (Fig. 2) in anaerobic cultures [14,69]. When this NADPH/NADH ratio equals zero, xylose reduc­tase only uses NADH and thereby consumes all NADH produced in the xylitol dehydrogenase reaction. Since in addition no regeneration of NADPH is re­quired for the xylose reductase reaction, redox-balanced xylose metabolism will occur according to Eq. 1:

Ratio = 0: 6 xylose ^ 10 ethanol + 10 CO2 + 10 ATP. (1)

At a ratio of one (Eq. 2), one out of every two D-xylose molecules can be further metabolised to ethanol, whereas the other is reduced to xylitol to maintain NADH balance. In addition, some carbon has to be redirected for the generation of NADPH, resulting in the formation of only 9 mol of ethanol from 12 mol of D-xylose (45% of the theoretical yield). Following these redox-balance considerations, catabolism via a xylose reductase with a NADPH/NADH-utilisation ratio of one will follow:

Ratio = 1: 12 xylose ^ 9 ethanol + 12 CO2 + 9 ATP + 6 xylitol. (2)

At ratios above one, NADH-dependent xylitol formation cannot compensate for the production of NADH in the xylitol dehydrogenase reaction and glycerol formation becomes essential for redox balancing [32,45,69]. When the xylose

reductase solely uses NADPH (an infinite NADPH/NADH ratio) this would result in the formation of only 0.5 mol ethanol per mol of xylose fermented.

Ratio = to : 6 xylose + 3 ATP ^ 3 ethanol + 6 glycerol + 6 CO2 . (3)

Despite these inherent redox restrictions and ensuing loss of ethanol yield on D-xylose, the expression of xylose reductase and xylitol dehydrogenase has long been the most successful strategy to enable D-xylose consumption by S. cerevisiae (elsewhere in this volume, and [29,32,33,39,63]). Although attempts have been made to change the cofactor specificity of xylose reduc­tase, fermentation properties of a S. cerevisiae strain containing this gene are not available [55]. Similarly, expression of a transhydrogenase in S. cere­visiae, with the aim of converting excess NADH into NADPH, did not result in reduced byproduct formation [51]. The latter result is perhaps not alto­gether surprising as, with NADPH/NADP+ ratios generally being higher than NADH/NAD+ ratios [51], reduction of NADP+ with NADH is thermodynam­ically unfavourable.

Despite the inherent redox constraints of S. cerevisiae strains based on the xylose reductase/xylitol dehydrogenase strategy, this strategy has resulted in many important insights into the kinetics of D-xylose metabolism by en­gineered S. cerevisiae strains. These findings include the benefits of over­expression of xylulokinase [29, 56], the side role of the S. cerevisiae aldose reductase (Gre3) (besides the heterologous dual specificity xylose reduc­tases) in xylitol formation [66], the role of the enzymes of the non-oxidative part of the pentose phosphate pathway [34,43], characterisation of D-xylose transport [27,62] and many studies on the inhibitor tolerance/sensitivity of D-xylose-consuming strains [54]. The latter will be especially crucial for suc­cessful application of D-xylose-consuming S. cerevisiae strains for ethanol production from lignocellulosic hydrolysates (see Sect. 7).

1.3

Systems biology is the quantitative characterization of genetic, transcription, protein, metabolic, signaling and other informational pathway responses to
a clearly defined perturbation of a biological system. More specifically, the perturbation may take the form of a genetic, biochemical, or environmental stimulus. At the core of systems biology is the transformation of quantita­tive, typically large-scale data sets, into in silico models that provide both interpretation and prediction. Systems biology has emerged as a tool applied in different fields, including metabolic engineering, to what many consider to be an independent discipline of study and research [57]. Table 1 provides an overview of commonly used industrial biotechnology strategies, focused on metabolic engineering with specific examples taken from applications

Table 1 Overview of commonly used industrial biotechnology strategies

Industrial biotechnology strategy Examples of application to

bioethanol production

A case study considering the co-production of ethanol and succinic acid suggests significant cost reduction (sales price of ethanol decreases from $0.51 to $0.42/gal.). Pilot plant confirmation pending [115,116].

In silico aided metabolic engineering of S. cerevisiae lead to a 40% reduction in glycerol formation and 3% increase in ethanol yield in vivo [154].

Natural ethanol producing bacterium Zymomonas mobilis metabolically engineered to ferment xylose and arabinose as preferred carbon sources via introduction/expression of E. coli pathway genes [6,155].

Xylose (C5H10O5, significant fraction of lignocelluloses) utilization by S. cerevisiae investigated and optimized via introduction of a Piromyces sp. xylose isomerase (XylA). Further xylose metabolic structural genes were overexpressed. Xylose consumption of

0.

9-1.1 g g-biomass-1 h-1, demonstrated in vivo [156-159].

to bioethanol production. The examples cited exploit toolboxes developed within systems biology.

Therefore, we refer here to industrial systems biology, defined as the appli­cation of experimental or numerical methods developed from systems biol­ogy to improve bioprocess development in terms of final product titer, yield, or productivity, or process robustness and efficiency. In most cases, indus­trial systems biology has been product — or process-specific; however, there are emerging examples of successful commercialization of stand-alone systems biology tools and products for broad application [58].

Recent advances in high-throughput experimental techniques have re­sulted in rapid accumulation of a wide range of x-omics data of various forms (Fig. 3), providing a foundation for in-depth understanding of biolog­ical processes [59-62]. How to integrate, interpret, and apply these data is an area of active research. Bioinformatics has become a well-established and recognized interdisciplinary field. To date, large data sets of transcriptomes, metabolomes, and to lesser degrees proteomes and fluxomes, for multiple organisms have been acquired. Resources are being applied to integrating the various data sets for in silico simulations and creating relevant models that represent in vivo physiological conditions of host cells responding to environmental stimuli. Even though our ability to analyze these x-omic (see “Glossary”) data in a truly integrated manner is limited, new targets for strain improvement can be identified from these global data [63-69].

A sampling of recent developments and applications in the field of sys­tems biology will be discussed in relation to improving the productivity of bioethanol. Examples will be provided on single x-ome approaches and com­bined analysis of these x-ome data for the development of improved strains and enhancement of metabolic engineering strategies.

3.1