The pretreatment of corn stover has been investigated in a large number of studies, as it is an abundant agricultural residue, primarily in the USA but also in Europe. In an extensive study undertaken in the USA, where the same batch of corn stover was pretreated using various pretreatment methods (acid hydrolysis by dilute acid, AFEX, ARP, lime treatment and LHW treatment) and then subjected to standard evaluation techniques, the yields of sugars were found to be more or less the same [63]. This study is commonly re­ferred to as the CAFI study. Total sugar yields—after pretreatment followed by enzymatic hydrolysis—of around 90% or more were reached (see Table 2).

Teymouri et al. pretreated corn stover using AFEX [40], which resulted in 96% glucose yield and about 78% xylose yield after enzymatic hydrolysis of washed material, corresponding to a glucan concentration of 1 wt % after 168 h with 15 FPU/g cellulose Spezyme CP loading. The ammonia loading was 1: 1 (equal amounts, by weight, of ammonia and dry corn stover) and the maximum sugar yield was obtained at 37 wt % moisture in the raw material.

Continuous ammonia pretreatment (ARP) can be used either by itself or preceded by percolation with hot water, in order to hydrolyse the hemicel — lulose under milder conditions and thus prevent hemicellulose loss. Kim et al. [64] performed low-liquid ARP and reported a glucan yield in en­zymatic hydrolysis of 88% using Spezyme CP at an activity of 15 FPU/g glucan. The glucan recovery following APR was, however, not reported. Kim and Lee [65] also performed a two-step percolation using water in the first step and ammonia in the second. This resulted in 83% xylose recovery after pretreatment and 85% glucan yield in enzymatic hydrolysis, again using a Spezyme CP loading of 15 FPU/g glucan.

Several other types of alkaline solutions have been used for pretreatment of corn stover. Kaar and Holtzapple [66] used alkali pretreatment with lime to

Table 2 Summary of studies on corn stover Method Catalyst Time (min) & temp. (°С) Dry matter (wt%) Enzymatic hydrolysis conditions Glucose yielda (%) Xylose yield3 (%) Refs. AFEX Cone NH3 5, 90 63 1% glucan, washed, 50 °С, 15 FPU/g cellulose 96.0 77.7 [39] ARP NH3 10, 170 23 d 1% glucan, washed, 50 °С, 15 FPU/g cellulose 90.0 41.lb [63] Alkali Ca(OH)2 4 weeks, 55 na 1% glucan, washed, 50 °С, 15 FPU/g cellulose 92.0 52.8b [66] Dilute acid hydrolysis-1 H2S04 0.49% 20, 160 5 1% glucan, washed, 50 °С, 15 FPU/g cellulose 91.6 91.2 [69] Dilute acid hydrolysis-2 H2SO4 5% 90, 120 10 3% solids, 50 °С, 15 FPU/g solids, 72 h 54.6 100c [68] Steam-1 H2SO4 5, 190 9.5 5% solids, washed, 50 °С, 25 FPU/g solids, 48 h 73.6 61.0 [71] Steam-2 S02 5, 190 35 2% solids, washed, 40 °С, 15 FPU/g solids, 96 h 90.0 84.0 [72] Liquid hot water Water 15, 190 i6 1% glucan, washed, 50 °С, 15 FPU/g cellulose 85.2 26.3b [70] Wet oxidation O2, N3.2CO3 15, 195 6 2% solids, washed, 50 °С, 25 FPU/g solids, 24 h 74.0 53.7C [74]

a Based on monomer sugar obtained in the hydrolysate after pretreatment and in the enzymatic hydrolysis of pretreated material. b A large fraction of xylose was released in oligometric form.

If this is taken into consideration the xylose yield increases to 76.4% for Alkali, up to 90% for ARP and 81.7 for Liquid hot water. c This value is for xylose + arabinose as the sum was presented in the reference. d Based on a liquid throughput of 3.3 mL of liquid per g of corn stover.

facilitate enzymatic hydrolysis. Pretreatment using 0.075 g Ca(OH)2 per g dry biomass resulted in 88 and 87% yields in enzymatic hydrolysis for glucose and xylose, respectively, after 7 days hydrolysis. However, a rather high enzyme loading of 25 FPU/g dry biomass was used (which is about 42-50 FPU/g cel­lulose, assuming that the cellulose content in the pretreated material is about 50-60% of the total). At an enzyme loading of 23 FPU/g cellulose, which is more in the same range as that used for the data in Table 2, the glucose and xylose yields dropped to 60 and 47%, respectively. Higher yields were achieved when lower pretreatment temperatures (55 °C) and longer residence times (4 weeks) were used [67] (see Table 2). The pretreatment was in this case performed in an excess of lime, 0.5 g per g raw material with aeration, although only about 0.08 g lime was consumed per g raw material.

Varga et al. [68] reached high sugar yields with alkaline pretreatment using 10 wt % NaOH at 120 °C for 60 min. At these conditions more than 95% of the lignin and about 88% of the hemicellulose was removed from the solid mate­rial. After enzymatic hydrolysis, at 50 °C for 48 h, of the solid material diluted to 2 wt % WIS and using 25 FPU/g DM, the overall amount of released sug­ars reached 63.7 g per 100 g corn stover, which corresponds to a yield of about 82% of the theoretical based on the amount of hemicellulose and cellulose present in the raw material. This high overall sugar yield was also obtained using a considerably lower and, from an economic standpoint, more feas­ible alkali concentration of 0.5 wt % NaOH, after pretreatment at 120 °C for 180 min.

Dilute acid pretreatment is probably one of the most investigated pretreat­ment methods. The addition of acid enhances the yield of hemicellulose sug­ars significantly. Acids are also good catalysts during pretreatment. Kalman et al. [69] used dilute sulphuric acid pretreatment and obtained a 55% over­all glucose yield after enzymatic hydrolysis with 15 FPU/g dry biomass. Lloyd and Wyman [70] optimized the conditions for pretreatment of corn stover after soaking in H2SO4 at a dry matter content of 5 wt % solids and a H2SO4 concentration of 0.49 wt %. The pretreatment was performed in a reactor with indirect heating. The highest overall sugar yield, i. e. considering both glucose and xylose, was obtained for pretreatment at 160 °C for 20 min resulting in

91.6 and 91.2% glucose and xylose yield, respectively. The high liquid to solid ratio is very beneficial to prevent hemicellulose sugar degradation. They also reported 100% xylose solubilization in the pretreatment.

LHW pretreatment with controlled pH (the pH-adjusting chemical was not reported) has been performed by Mosier et al. [71]. In this study an enzyme loading of 15 FPU/g cellulose was used in the hydrolysis. An overall glucose yield of 91% and an overall xylose yield of 82% were obtained after 48 h hydrolysis.

Varga et al. [72] investigated steam pretreatment with H2SO4, i. e. by using direct steam at a higher dry matter content. The highest overall sugar yield (glucose, xylose and arabinose), 56.1 g/g raw material corresponding to 73% of the theoretical, was obtained after pretreatment at 190 °C for 5 min with 2 wt % H2SO4 and enzymatic hydrolysis of 5 wt % solids using an enzyme loading of 25 FPU/g dry matter. At these conditions the overall glucose yield was about 74%.

Other acid catalysts have a similar effect on the hydrolysis of various ma­terials. Ohgren et al. [73] also performed steam pretreatment, but instead of H2SO4 used SO2 as acid catalyst at a concentration of about 2-3 wt % and at a higher dry matter content of 35 wt %. The highest overall sugar yield, 90% of theoretical for glucose and 84% for xylose, was obtained for pretreatment at 190 ° C for 5 min.

In a study where the dry matter content was higher, i. e. 40 wt %, steam pretreatment of corn stover after impregnation with SO2 was evaluated [74]. Pretreatment of SO2-impregnated corn stover, with a dry matter content of 40 wt %, at 200 °C for 5 min resulted in a glucose yield of 92% of the theoret­ical and a xylose yield of 66%. The maximum xylose yield was 84%, obtained with pretreatment at 190 °C for 5 min. Under these conditions the glucose yield was 90%.

High enzymatic conversion of cellulose in enzymatic hydrolysis can also be achieved by wet oxidation [75]. The recovery of cellulose after wet oxidation of corn stover at 6 wt % WIS with 2 g/L Na2CO3 and 12 bar O2 as catalysts, and at 195 °C for 15 min, was 85.1%. The enzymatic conversion was 83.4% and the overall glucose yield was 74% after enzymatic hydrolysis at 50 °C using 25 FPU/g dry pretreated corn stover. A decrease in enzyme activity to 5 FPU/g dry pretreated material decreased the overall yield to 63.4%. The overall yield of hemicellulose sugars was about 54%, which indicates a rather high degradation.

This shows that rather high sugar yields from corn stover can be obtained using a variety of pretreatment methods. Thus, it can be concluded that corn stover is easily hydrolysed using the enzymatic process. The overall sugar yields can come close to what is theoretically possible.

However, in the comparison shown in Table 2, hydrolysis was performed on washed material in most cases, and at low solids concentrations. In a full — scale process, the whole slurry from pretreatment would probably be used and at high solids concentrations. The hydrolysis yield alone is not an indi­cator of a successful pretreatment. The fermentability of the liquid fraction after pretreatment is, for example, an equally important parameter. Also, the concentration of sugars after hydrolysis must be high enough to result in an acceptable ethanol concentration. The duration of enzymatic hydrolysis required to reach the desired yield is another important factor as regards pro­cess economics, since longer reaction times imply larger reaction vessels for a certain production capacity.

Several studies have indeed been performed, both on SHF and SSF, at higher dry matter levels, but these are more difficult to compare. Varga et al. [76] performed SSF of wet-oxidized corn stover at high dry matter con­tent, up to 15 wt % DM. However the wet oxidation was performed at 6 wt % DM under the following conditions: 2 g/L Na2CO3, 12 bar O2, 195 °C, 15 min. The liquid was then removed and the concentrated pretreated solids were added back to a smaller amount of the liquid in a fed batch SSF. The high­est yield in the SSF (performed at 30 °C using baker’s yeast and 30 FPU/g dry pretreated stover for 120 h) was 83% of the theoretical based on the glucose content in the pretreated material. Considering that the recovery of cellulose in the pretreatment was 86%, the overall ethanol yield was 71% of the the­oretical based on the glucose content in the raw material. A decrease of the enzyme loading in the SSF to 15 FPU/g dry pretreated corn stover resulted in a decrease in the overall ethanol yield to 63% of the theoretical.

Ohgren et al. (2006) [77] performed SSF on the whole slurry from corn stover, at 11 wt % WIS, after pretreatment of corn stover impregnated with 2 wt % SO2 at 200 °C for 5 min. The SSF was performed as fed batch at 35 °C for 96 h using 5 g/L of a xylose-utilizing yeast, TMB3400, cultivated on pre­treatment liquid and a cellulase loading of 15 FPU/g WIS. The overall ethanol yield was 92% of the theoretical, based on the glucose content in the raw ma­terial, and 59% based on the content of both glucose and xylose. The ethanol concentration was 36.8 g/L. However, a major part of the xylose was still left in the broth. The conclusions were that the cultivated yeast was tolerant to the inhibitors present in the pretreated slurry, but that it is necessary to develop a better feeding strategy to ferment all xylose.

The method of assessment of the pretreatment is crucial. In a study per­formed by Ohgren et al. (2007) [78], steam pretreatment of corn stover was assessed by enzymatic hydrolysis using ordinary cellulases supplemented with xylanases and also after partial removal of lignin. The pretreatment was performed either without any impregnation or with impregnation by SO2. The conditions and overall sugar yields after pretreatment and enzymatic hydrolysis are given in Table 3.

The addition of small amounts of xylanases had a major impact on the sugar yield. The overall glucose yield after enzymatic hydrolysis increased from around 83% to near theoretical and the xylose yield from 71 to 74%, based on the content in the raw material for the pretreatment with catalyst. For the less severe pretreatment using auto-hydrolysis (i. e. 190 °C, 5 min, no catalyst), the addition of xylanases had an even higher effect resulting in an increase of the overall xylose yield from 74.6 to 85% of theoretical. The glu­cose yield increased even more from 69 to 94%. It should be noted that the addition of xylanases had a higher effect on the improvement of cellulose hydrolysis than on the increase of hemicellulose sugars.

It should be emphasized that the assessment with addition of xylanases was performed under pretreatment conditions that were optimized based on assessment with cellulases only in previous studies. Assessment with addition of xylanases during the optimization of the pretreatment might have resulted in less severe pretreatment conditions.

In order to compare pretreatment methods the whole optimized process, including process integration and the preparation and use of co-products, which may differ between different pretreatment methods, has to be assessed. These data are not available, not even for corn stover, which is one of the most investigated materials.

Eggeman et al. [79] investigated the pretreatment cost in ethanol produc­tion from corn stover for the five different pretreatment methods included in the CAFI study described above. The pretreatment design was based on experimental data [80] from the various research groups in the CAFI study, and was implemented in the Aspen Plus model for a full-scale bioethanol plant previously developed by NREL [81]. The model was based on a corn stover feed rate of 2000 dry metric tons per day. The process configuration was based on pretreatment, SSF, ethanol recovery and internal production of heat and electricity from the syrup and solid residue from the process. The process configuration was identical for all processes except for the pre­treatment step. The dilute acid pretreatment process resulted in the lowest ethanol production cost, 0.26 US$ per litre for the base case alternative where oligomers released in the pretreatment and hydrolysis steps were not consid­ered for ethanol production. However, it should be emphasized that a fairer comparison would require optimization of each process alternative taking into consideration the specific features of the pretreatment method used.

4.2

Plant cellulose exists in a highly crystalline form. In addition, it is associated with hemicellulose and surrounded by lignin, which may also be covalently bound to hemicellulose. Pretreatments aim at increasing the surface area of cellulose by either removing lignin or solubilising hemicellulose, disrupting the crystallinity and/or by increasing the pore volume. Hydrolysis of cellu­lose requires the co-operation of three classes of cellulolytic enzymes, namely cellobiohydrolases (CBH, EC 3.2.1.91), endo-|5-1,4-glucanases (EG, EC 3.2.1.4) and в-glucosidases (BG, EC 3.2.1.21). The CAZY (carbohydrate active en­zymes) [16] classification system collates glycosyl hydrolase (GH) enzymes into families according to sequence similarity, which have been shown to reflect shared structural features. Most of the initial cellulase work was con­centrated on the biochemistry, genetics and process development of the mesophilic fungus Trichoderma reesei. This fungus is one of the most power­ful secretors of extracellular proteins. It is industrially used for the production of various homologous and heterologous proteins. Also several thermostable enzymes have been expressed in this host, as reviewed by Bergquist et al. [8].

Trichoderma reesei produces several cellulases which act synergistically in the degradation of cellulose. Eight major cellulase genes have so far been identified from the T. reesei genome; two cellobiohydrolases (CBH I and II, i. e. Cel7A and Cel6A), and six endoglucanases (EG I-VI, i. e. Cel7B, Cel5a, Cel12A, Cel45A, Cel61A, Cel74A) [24]. All known T. reesei cellulases, with one exception (Cel12A), have a two-modular structure. They consist of a cata­lytic module and a carbohydrate binding module (CBM) connected with a linker region. Cel7A (CBHI) is the major cellulose produced by T. reesei; it has been reported to hydrolyse solid cellulose and constitutes about 60% of the cellulases expressed [51, 73]. It has been shown that Cel7A hydrolyses the cellulose chain from the reducing end and it is believed that the chain is hydrolysed processively [3,19,55]. Cel6A, on the other hand, preferably hydrolyses the cellulose chain from the non-reducing end [55,73]. It consti­tutes about 10-15% of total cellulase proteins [51]. The Cel7B is the major endoglucanase, forming about 6-10% of total T. reesei cellulase [51,73]. It has activity against solid and soluble substrates, such as CMC, as well as against xylan, PASC and glucomannans [71]. Also the endoglucanase Cel5A is reported to have activity against solid (Avicel, BMCC) and soluble (CMC, mannan) substrates [31,38,44], but not on xylans. This enzyme comprises about 1-10% of the total cellulases in T. reesei [51,73]. The minor endoglu­canases Cel12A, Cel61A and Cel45A are reported to hydrolyse solid (Avicel, filter paper) and soluble (CMC, glucomannan) substrates with diverse spe­cific activities. Two в-glucosidase encoding genes from T. reesei have been cloned [2,74]. Cel74A [24] is a xyloglucanase having also endoglucanase ac­tivity against в-glucan and CMC [28].

3

In addition to tolerance and robustness, strain stability is a prerequisite when designing yeast strains for industrial use. Strains carrying mul­ticopy plasmids are generally not applicable in industry due to their instability [123,145]. Multicopy plasmids require auxotrophic or antibiotic resistance markers to be retained in the cell, both of which are not ap­plicable in industrial media containing complex nutrients and being used in large volumes. Thus, chromosomal integration is necessary for any genes to be introduced in industrially applied yeast strains. Ideally this re­quires sufficient specific activity of the introduced heterologous enzymes, so that single-copy integration supplies enough activity for metabolic func­tion. Multiple chromosomal integration has also been utilized to gener­ate stable pentose-fermenting strains with high activity of the enzymes introduced [6-8].

Metabolic engineering strategies applied on industrial strains have been limited to the introduction of the initial xylose and arabinose utilization pathways [4,5,8,101]. Only the XR-XDH pathway has been developed in in­dustrial S. cerevisiae strains [4,5,101] (strains A4 and A6, Table 1; strain F, Tables 1 and 4; strain TMB3400, Tables 1, 3 and 4; strain 1400(pLNH32), Table 2). No chromosomally integrated XI constructs have been reported. XI expression in S. cerevisiae seems to require a multicopy expression system to provide sufficient enzyme activity for xylose growth and fermentation [43]. Due to the difficulty of applying complex metabolic engineering strategies in industrial strains, procedures for random strain improvement have been relied upon to improve xylose utilization.

5.3

The same PET operon used in engineering of KO11 has also been used to construct a series of ethanologenic K12-derivatives, designated FBR for the Fermentation Biochemistry Research Unit. These strains were engineered with the goal of maximizing strain stability [43,44]. The most recent strain in this line, FBR5, produced ethanol from a variety of substrates at 86-92% of the theoretical yield [44]. Long-term stability of this strain was demonstrated by the maintenance of ethanol yields over 26 days of continuous culture on glucose or xylose [45]. However, the final ethanol concentration and yield from FBR5 in LB xylose are lower than LY168 in minimal medium (Table 2); additionally, these strains have the disadvantage of rich media dependence and contain plasmids.

2.4

The four categories referred to in Fig. 1 are highly interconnected, and there is significant debate as to the ranking of these categories in terms of priority and impact. Process economics are often estimated by quantitative modeling that

includes major process costs (both operating and capital) and process value, dictated by the product’s estimated market price, demand growth rate, mar­ket share, and any competitive advantages that may exist. For commercial­ization it is reasonable to assume that process economics must be favorable before any further effort can continue on considering and improving pub­lic perception and environmental impact. Sustainability and self-sufficiency are perhaps less well defined, noting that many of the issues considered, in­eluding raw material availability and the potential for process integration into a biorefinery, are likely to be discussed in the context of process economics. It is included in Fig. 1 and discussed as a separate category simply because of the recent focus it has received in the background of significantly increasing petroleum and feedstock prices. Self-sufficiency and sustainability may actu­ally trump process economics in cases where issues such as national security play a role, for example dependence on a foreign state for significant fractions of energy.

We will aim to provide a brief overview of the four major categories (see Fig. 1), with particular attention paid to the economic drivers propelling bioethanol development. In fact, we will demonstrate that those economic drivers are not exclusive to bioethanol, but rather, serve as catalysts for the petrochemical industry’s rapid adoption of industrial biotechnology as a plat­form for producing existing and future products.

1.1.2

The removal of hemicellulose undoubtedly imparts substantial modifications to the structure and accessibility of lignocellulosic substrates. Ideally, the pretreatment should fractionate the cellulose, hemicellulose and lignin so that cellulases can react with pure cellulose. However to obtain the great­est value from biomass feedstocks, it will be necessary to recover all of the components in an exploitable form [22]. Most of the pretreatments that are currently being advocated for their potential application in the bioconversion process employ “optimal” treatment conditions that maximize the recovery of cellulose. However, these conditions differ from those that target max­imum lignin and hemicellulose yields [85]. Of the three main components of lignocellulose, hemicelluloses have been shown to be the most sensitive to changes in pretreatment conditions [59]. Steam pretreatment at high severity (higher temperature, residence time and catalyst concentration) is required to maximize cellulose and lignin yields, while a significant portion of the hemicellulose is destroyed at these conditions [86]. Hemicellulose degrada­tion products such as furfural and hydroxymethyl furfural inhibit subsequent fermentation [87]. Therefore, pretreatment conditions are frequently tailored considering the compromise between separating the lignin and hemicellulose components from cellulose while concurrently maximizing the recovery of all the available carbohydrates. Each pretreatment method approaches the recov­ery of hemicellulose in a distinct manner, as diverse pretreatment methods all have varying effects on the hemicellulose fraction.

Depending on the nature of the pretreatment, either a solid fraction, or combined solid and liquid fractions are generated for subsequent sacchari­fication processes [22]. For example, a pretreatment such as AFEX produces only a solid fraction [88] while pretreatments at an acidic pH such as di­lute acid, steam pretreatment and wet oxidation produce a combined liquid and solid stream [22]. Alkaline pretreatment methods such as AFEX modify and/or remove lignin and leave both hemicellulose and cellulose intact [22]. Acidic pretreatments, such as SO2-catalyzed steam pretreatment or dilute acid pretreatments, depending on the acid catalyst employed, produce a liquid fraction that is composed mainly of hemicellulose, either in monomeric or oligomeric form [59].

The tendency for hemicellulose to be relegated to either the liquid or solid fraction is highly dependent on the severity of the pretreatment (time, temperature and amount of catalyst). In the case of aspen wood (Populus tremuloides (Michx.)), it was shown that varying the steam pretreatment conditions followed by alkaline peroxide bleaching of the solid fraction re­sulted in large variations in the cellulose and hemicellulose content of the solid fraction. Increasing the severity of pretreatment decreased the cellu­lose and hemicellulose molecular weight. For example, it was shown that at low severity the solid fraction contained a xylan content of 7% and a cellu­lose molecular weight of 900 000, while at higher severity the xylose content and cellulose molecular weight in the solid fraction were reduced to 1% and < 40 000, respectively. Without the alkaline peroxide treatment, the xy­lose content in the solid fraction varied from 12.4 to 2.5% as the severity of the steam pretreatment was raised from log10 Ro 2.76 (180.3 °C, 2 min) to log10 Ro 3.62 (189.1 °C, 10 min) [59]. Similarly, during SO2-catalyzed steam pretreatment of corn stover, the solid fraction obtained from pretreatment at 190 °C for 5 min while varying the catalyst dosage from 0 to 3% SO2 was com­posed of almost identical amounts of glucan (56%); however, the amount of xylan differed from 18 to 10%, respectively [89]. Similar results were reported by Boussaid et al. [39] while treating Douglas-fir wood chips as pretreatment at low severity (175 °C, 7.5 min, and 4.5% SO2) resulted in the solubilization of 87% of the hemicellulose component into the water-soluble stream with 80% of the recovered sugars in monomeric form. However, the recovery of hemicellulose was reduced to 43% upon increasing the pretreatment severity (215 °C, 2.38% of SO2, 2.38 min) with a concomitant increase in the produc­tion of sugar degradation products as furfural and hydroxymethylfurfural, which compromised subsequent fermentation to ethanol by Saccharomyces cerevisiae.

Later work by Shevchenko et al. [90] showed that the hemicelluloses solubi­lized into the liquid stream during the steam pretreatment of Douglas-fir wood chips that remained in oligomeric form could be further processed by an acid — catalyzed post-hydrolysis step in order to recover the remaining hemicellulose in monomeric form. Furthermore, it was shown that the partial oxidation undergone by the added SO2 catalyst during steam pretreatment to sulfuric acid provided an acid concentration sufficient to depolymerize the remaining oligomeric hemicelluloses with only minimal production of fermentation in­hibitors. These results show that the steam pretreatment process is capable of recovering a large proportion of the feedstock hemicelluloses in a monomeric form, depending on the pretreatment conditions employed. Although process modifications to improve hemicellulose recovery decrease the production of inhibitory compounds and increase total sugar concentrations, which aid in subsequent fermentation, ultimately, it is the hydrolysis of the solid cellulosic substrate that provides the majority of the glucose for ethanol production. In­creasing the total recovery of sugars while minimizing inhibitor production must be weighed against the negative effect of significant amounts of substrate hemicellulose on the ease of hydrolyzability of the substrate.

4.2

Barbel Hahn-Hagerdal1 (И) • Kaisa Karhumaa1 • Marie Jeppsson1,2 •

Marie F. Gorwa-Grauslund1

Applied Microbiology, Lund University, P. O. Box 124, 221 00 Lund, Sweden Barbel. Hahn-Hagerdal@tmb. lth. se

2 Present address:

GS Development AB, Jagershillgatan 15, 213 75 Malmo, Sweden

1 Introduction……………………………………………………………………………………………… 148

2 Xylose……………………………………………………………………………………………………… 149

2.1 Xylose Utilization Pathways………………………………………………………………………. 158

2.2 Expression of XI in S. cerevisiae……………………………………………………………………. 158

2.3 Expression of XR and XDH in S. cerevisiae……………………………………………………. 159

3 Arabinose…………………………………………………………………………………………………. 160

3.1 Arabinose Utilization Pathways…………………………………………………………………. 160

3.2 Engineering Arabinose Utilization in S. cerevisiae…………………………………………… 161

4 Improving Ethanolic Fermentation by Pentose-Utilizing S. cerevisiae. . . 162

4.1 Sugar Transport………………………………………………………………………………………… 163

4.2 Improving the Conversion of Xylose to Xylulose…………………………………………… 164

4.2.1 Cofactor Dependence……………………………………………………………………………….. 164

4.2.2 Activity of Initial Pentose Pathway Enzymes……………………………………………… 164

4.2.3 GRE3 Deletion…………………………………………………………………………………………. 165

4.3 Xylulokinase……………………………………………………………………………………………. 165

4.4 Pentose Phosphate Pathway……………………………………………………………………… 166

4.5 Engineering the Redox Metabolism of the Cell…………………………………………….. 167

4.5.1 Oxidative PPP…………………………………………………………………………………………. 167

4.5.2 Transhydrogenase and Redox Enzymes……………………………………………………… 168

4.6 Glycolytic Flux…………………………………………………………………………………………. 169

4.7 Other Modifications…………………………………………………………………………………. 169

4.8 Random Methods…………………………………………………………………………………….. 170

5 Industrial Pentose-Fermenting Strains…………………………………………………………. 170

5.1 Inhibitor Tolerance…………………………………………………………………………………….. 171

5.2 Strain Stability…………………………………………………………………………………………… 171

5.3 Fermentation of Hydrolysates……………………………………………………………………. 172

6 Conclusion and Future Outlook………………………………………………………………….. 172

References……………………………………………………………………………………………………. 173

Abstract The introduction of pentose utilization pathways in baker’s yeast Saccharomyces cerevisiae is summarized together with metabolic engineering strategies to improve

ethanolic pentose fermentation. Bacterial and fungal xylose and arabinose pathways have been expressed in S. cerevisiae but do not generally convey significant ethanolic fermen­tation traits to this yeast. A large number of rational metabolic engineering strategies directed among others toward sugar transport, initial pentose conversion, the pentose phosphate pathway, and the cellular redox metabolism have been exploited. The directed metabolic engineering approach has often been combined with random approaches in­cluding adaptation, mutagenesis, and hybridization. The knowledge gained about pentose fermentation in S. cerevisiae is primarily limited to genetically and physiologically well — characterized laboratory strains. The translation of this knowledge to strains performing in an industrial context is discussed.

Keywords Arabinose • Ethanol • Fermentation • Lignocellulose • Xylose • Yeast

Abbreviations

G6PDH Glucose-6-phosphate dehydrogenase

GAPDH Glyceraldehyde-3-phosphate dehydrogenase

mRNA Messenger RNA

PPP Pentose phosphate pathway

RKI Ribose-5-phosphate ketol-isomerase

RPE Ribulose-5-phosphate 3-epimerase

TAL Transaldolase

TKL Transketolase

XDH Xylitol dehydrogenase

XI Xylose isomerase

XK Xylulokinase

XR Xylose reductase

1

Introduction

When in the late 1970s it was discovered independently in two laborato­ries in North America [1,2] that baker’s yeast Saccharomyces cerevisiae could ferment the pentose sugar xylulose to ethanol, it was proclaimed that the development of recombinant xylose-fermenting strains of S. cerevisiae was a task that would be efficiently solved within a couple of years. Still, more than 25 years later, only a limited number of industrial S. cerevisiae strains that ferment pentose sugars have been generated [3-9]. Furthermore, there are relatively few studies on the performance of these strains under industrial conditions in lignocellulosic hydrolysates [6,10-14]. The difficulty in devel­oping efficient pentose-fermenting S. cerevisiae strains is no doubt that the regulation of metabolism in the eukaryotic yeasts is much less understood than that of, for example, the prokaryotic bacterium Escherichia coli. Conse­quently, the research on pentose-fermenting strains of S. cerevisiae has had the spin-off effect of generating more knowledge on the metabolism of this species, not least in relation to other yeasts.

The rationale for developing pentose-utilizing S. cerevisiae strains relies on the fact that this yeast has been used for the industrial production of ethanol and carbon dioxide as long as human history has been recorded. Presently,

S. cerevisiae forms the basis for the world’s largest fermentation industry pro­ducing beer, wine, potable and industrial ethanol, and baker’s yeast. In add­ition, this organism serves as a eukaryotic model organism with an intensely studied cell biology and arrays of genetic engineering tools [15]. However, the most important reason for developing pentose-fermenting S. cerevisiae is the fact that such strains can be integrated into existing ethanol plants al­ready using this yeast. Two independent investigations have estimated that integrated approaches to the production of lignocellulosic ethanol will reduce the production cost by nearly 20% [16,17].

This chapter summarizes the metabolic engineering approaches taken to develop pentose-fermenting strains of S. cerevisiae. Different engineer­ing strategies and their physiological context are described below, and the respective fermentation results from each study are chronologically sum­marized in Tables 1-4. Metabolic engineering for arabinose utilization is reported separately, since engineering L-arabinose utilization in S. cerevisiae has only recently been addressed. As will be detailed below, the fermenta­tion of pentose sugars is governed by carbon catabolite repression and by reoxidation of reduced cofactors. Fermentation results of recombinant S. cere­visiae strains have therefore been summarized in relation to batch (Tables 1 and 2) and continuous culture (Tables 3 and 4), and in relation to anaerobic (Tables 1 and 3), oxygen-limited (Table 2), and aerobic conditions (Table 4). Moreover, the data have been organized in relation to the respective con­trol strain to highlight the relative improvement of a particular engineering strategy. Studies that do not use the four aforementioned experimental con­ditions, or for which information on fermentation parameters is insufficient, have been omitted.

2

Metabolic engineering is defined as the improvement of cellular activities by manipulation of enzymic, transport and regulatory functions of the cell with the use of recombinant DNA technology [6]. After the successful expression of a XI in S. cerevisiae [42], reactions downstream of D-xylulose and the, presumably Gre3-dependent, formation of xylitol were identified as priority targets (see previous section).

As it is unlikely that the high capacity of glycolysis in S. cerevisiae would limit D-xylose fermentation rates; limitations in D-xylose fermentation are likely to reside either in the reaction catalysed by xylulokinase or in one of the four reactions of the non-oxidative pentose phosphate pathway. Modulat­ing the flux through a certain pathway by up-modulation of single enzymes often has little effect, as can be shown by metabolic control analysis [50]. Hence, it was decided to simultaneously increase the levels of all five enzymes. To this end, the S. cerevisiae structural genes encoding xylulokinase (XKS1), ribulose-5-phosphate epimerase (RPE1), transketolase (TKL1), transaldolase (TAL1) and ribulose-5-phosphate isomerase (RPI1) were over-expressed to­gether with the Piromyces sp. E2 XylA gene [43]. Since the non-specific aldose reductase encoded by GRE3 had previously been implicated in xyli — tol formation by S. cerevisiae, this gene was also deleted in the engineered strain [45,66].

Research on pentose metabolism in S. cerevisiae is increasingly impeded by the fact that key biochemical intermediates can no longer be purchased commercially [35,43]. While this precluded enzyme-activity assays for several of the over-expressed genes, mRNA analysis indicated that over-expression, either from strong constitutive promoters inserted in front of chromosomal genes or from plasmid-borne expression cassettes, was successful.

Remarkably, the S. cerevisiae strain (RWB 217) harbouring the six over­expressions and single deletion was directly capable of anaerobic growth on D-xylose as the sole carbon source at a growth rate of 0.09 h-1 [43]. Start­ing with a low-density inoculum, this strain consumed 20 g L-1 of D-xylose within 40 h, with an ethanol yield on D-xylose of 0.43 gg-1. This ethanol yield, which is lower than the theoretical yield of 0.51 gg-1 due to the for­mation of biomass and glycerol, was virtually identical to the ethanol yield found on glucose in exponentially growing, anaerobic S. cerevisiae cultures. Deletion of GRE3 reduced xylitol production to trace amounts (0.4 mM from 20 g L-1 D-xylose), indicating that alternative D-xylose — or D-xylulose reduc­ing enzymes were active at very low rates in this S. cerevisiae background. In the engineered strain, D-xylulose no longer accumulated in the broth, indicating that limitations downstream of D-xylulose had been successfully eliminated.

In an independent study, Karhumaa et al. (2005) expressed the XI gene from T. thermophilus together with the same combination of pentose phos­phate pathway enzymes [35]. In these strains the specific activity of XI was 0.008-0.017 |xmol (mg protein)-1 min-1 at 30 0C. In contrast to the efficient anaerobic growth of the above-described S. cerevisiae expressing the Piromyces sp. E2 XI, D-xylose consumption by the T. thermophylus XI-containing strain (TMB 3045) was not observed under aerobic conditions. After additional se-

lection, a strain capable of aerobic growth on D-xylose at a maximum specific growth rate of 0.045 h-1 was isolated (TBM 3050). Confusingly, although the abstract claims anaerobic production of ethanol, the experimental description and results section describe the production of 0.29 g ethanol (g D-xylose)-1 at a rate of 2.4 mg (gbiomass)-1 h-1 under oxygen-limited conditions [35]. The ethanol production rates, are more than 400-fold lower than observed in the Piromyces XylA-based strain [35,42]. This observation, combined with the interesting observation that TMB 3045 and TMB 3050 display almost identi­cal specific growth rates on D-xylulose, indicates the importance of high-level functional expression of XI for efficient D-xylose fermentation.

In lignocellulosic hydrolysates, D-xylose is generally the second most abundant sugar, with glucose accounting for the majority of the fermentable sugar [24,46,69]. Rapid consumption of glucose-xylose mixtures — either sequential or simultaneous — is therefore crucial for successful industrial im­plementation. When the metabolically engineered strain RWB 217 (described above) was grown in anaerobic batch cultures on mixtures of 20 g L-1 glucose and 20 g L-1 D-xylose (Fig. 5), sequential utilisation was observed. Although both sugars were consumed within 40 h, D-xylose consumption only com­menced when the glucose concentration dropped below 4 gL-1. Instead of increasing exponentially, as anticipated based on the kinetics of D-xylose con­sumption in D-xylose-only cultures, the specific rate of D-xylose consumption decreased over time. Clearly, the kinetics of D-xylose consumption by cells grown in the presence of glucose were sub-optimal. This challenge was ad­dressed by evolutionary engineering.

6

The development of functional genomics has provided new tools and ap­proaches for understanding, mapping, modeling, and manipulating cells. Therefore, the metabolic engineering goal of identifying genes that confer a particular phenotype is conceptually and methodologically congruent with central issues in functional genomics. Functional genomics will not only elu­cidate what the genes do but will also help determine when, where, and how they are expressed as an organized system. The combination of genetics and a wide variety of x-omics data (transcriptomics, proteomics, metabolomics, fluxomics) can also be applied directly in metabolic engineering to iden­tify new targets for improved phenotype. Table 2 provides a summary of genome, transcriptome, proteome, and fluxome data applied to industrial biotechnology process development. However, for industrial systems biology to be further applied in experiments and development efforts, the quality and range of the different x-omics data should be comparable. The implementa­tion of high-throughput, easy to use, platform technologies will be critical in bringing these tools to broad applicability in bioprocess development.

A final point worth touching upon is that industrial partners have often cited that many of the x-omes, particularly the younger disciplines (e. g., pro — teomics), while providing academically interesting research, have not trans­lated into methods or approaches with industrial impact and value. This is a fair assessment, but one that is changing. The momentum of bioethanol de­velopment, and consequently other industrial biotechnology-produced prod­ucts (e. g., 1-3-propandiol), is driving manufacturers to develop better proces­ses with higher yields, titer, productivity, robustness, and efficiencies. The margins and areas for improvement are narrowing, and can only be met with innovative approaches and strategies that may be yet undiscovered. X-ome analysis and data is providing the innovation by developing data sets and tools that are beginning to answer fundamental questions (i. e., Is this path­way’s regulation transcriptional or translational? Is carbon being lost through the citric acid cycle or pentose phosphate pathway?). But perhaps more im­portant, industrial systems biology is leading to new questions not previously considered. In the struggle of how to handle and what do with all this data will emerge the questions that lead to novel and yet unrealized metabolic engineering strategies. Furthermore, these methods are data driven. Even if

manipulation, analysis, and interpretation of the data are not clear, biolog­ically and statistically high quality data are required to drive development. Industry requires this innovation to remain sustainable, and, therefore, must support industrial systems biology in its infancy and development.

4

The process steps of pretreatment, hydrolysis, and fermentation need to be viewed holistically to maximize ethanol yield and overall process cost. As discussed previously, different pretreatments produce different substrates for enzyme action, impacting both the required mix of enzymes, the dosage of those enzymes, and the cost of the hydrolysis step. Similarly, the selection of the fermentation organism determines the pH and temperature optima of the fermentation step, which can affect enzyme performance and loading since hydrolysis and fermentation are often combined at some stage of the hydro­lysis in a single reactor.

The enzymatic hydrolysis can either be done separately from the fermenta­tion (SHF, separate hydrolysis and fermentation) or in combination with the fermentation (SSF, simultaneous saccharification and fermentation). In SHF, hydrolysis is allowed to proceed to a point of completion at reaction condi­tions optimal for enzyme action, which today for T. reesei enzymes is 50 °C and pH 5, then the temperature is lowered to allow survival of the fermen­tation organism (typically 30-40 °C) and the pH is adjusted upwards to pH 5.5-7. The primary drawback to the SHF process is a reduced rate of hydro­lysis caused by product inhibition of the enzymes by the released monomeric and oligomeric sugars. The SSF process for producing ethanol is capable of improved hydrolysis rates, yields, and product concentrations compared to SHF because of the continuous removal of the reaction end products (the sugars) by the yeast, preventing competitive inhibition of some of the com­ponent enzymes, provided the temperature and pH required for fermentation does not drastically slow enzyme action. Ideally we will see organisms and enzymes developed that have similar growth and reaction optima, allowing optimal growth and enzyme action to occur in a single vessel. Currently, com­promises in either or both must be made in the process design since there is a 10-20 °C gap in temperature optima and a 0.5-2 pH unit gap in pH optima.

In hybrid hydrolysis and fermentation (HHF), the hydrolysis and fermen­tation are temporally separated to optimize the combined rate of hydrolysis and fermentation. Hydrolysis is allowed to proceed to a point at which glucose release slows significantly, then the temperature is dropped, the pH increased, and fermentation is initiated by addition of the organism. The development of an economically viable process depends on optimizing the timing of the shift from hydrolysis to fermentation, and is dependent on the enzymes, the organism, and all the factors that contribute to process cost, such as feed­stock cost, hydrolysis/fermentation residence time, solids loading, and capital investment.

It has been established that digestibility of a biomass substrate is highly dependent upon the type of pretreatment, enzyme efficiency, and dosage. Re­cent results indicate that mixing is also an important parameter in integrating pretreatment and hydrolysis [29]. In Fig. 4, acid pretreated corn stover (PCS) and hot water autolysed wheat straw (HWS) were hydrolyzed at Novozymes with a Celluclast/Novozym 188 mix at the same loadings using two different types of mixing: shake flask orbital mixing versus tumbling (lift and drop). While the PCS, a well-pretreated substrate whose cellulose can be wholly hy-

matic hydrolysis. Acid pretreated corn stover (PCS, supplied by the NREL) was washed prior to use. Pretreated wheat straw (HWS) was produced by wet autoclaving at 132 ° C for 30 h as estimated by application of the Arrhenius equation to the data of Garrote (1999) so as to produce minimally pretreated biomass. Residual dry weights were deter­mined as per NREL laboratory analytical procedure (LAP) 012 (NREL procedures can be found at [46]). Cellulose content was estimated from published values (HWS) [47], limit enzymatic hydrolysis, and carbohydrate compositional analysis (PCS). Hydrolysis was performed essentially as per NREL protocol LAP 009 (72 h, pH 5, 50 °C), using either flasks in a rotary shaker at 150 rpm (shaker) or in sealed tubes tumbling free in a rolling tub (tumbling). Analysis of resulting hydrolysis sugars was performed according to NREL protocol LAP 13-15. Calculation of approximate conversion was based on the amount of cellobiose and glucose released as a percentage of the theoretical yield from cellulose. PCS tumbling (A), PCS in a shaker (A), HWS tumbling (•), and HWS in a shaker (□) drolyzed, shows no difference in either the rate or extent of hydrolysis, the poorly pretreated HWS shows a dramatic improvement in hydrolysis from the more disruptive tumble mixing as compared to orbital mixing. Although conversion is fairly low for the HWS compared to the PCS, we saw both an increase in the rate and endpoint conversion of cellulose to glucose with the tumble mixing. These results indicate that the type and vigor of mixing dur­ing hydrolysis may allow less severe pretreatments to be implemented, with the potential to decrease both capital and operating costs during pretreat­ment. In addition, this type of vigorous mixing may allow higher solids levels during pretreatment and hydrolysis, resulting in a more concentrated sugar stream and higher ethanol titers from the fermentation. This has the potential to reduce operating costs in energy consumption used for ethanol distillation. In addition, utilization of higher solids increases plant throughput, reducing the total capital investment required.

4