Sandra T. Merino • Joel Cherry (И)

Novozymes Inc., 1445 Drew Ave., Davis, CA 95618, USA JRoC@novozymes. com

1 Introduction……………………………………………………………………………………………….. 96

2 Lignocellulosic Biomass to Ethanol Process Overview……………………………………. 97

2.1 Minimizing Yield Loss and Cost…………………………………………………………………… 99

3 Impact of Process Steps on Enzyme Dosage and Cost…………………………………. 100

3.1 Impact of Substrate Selection on Enzyme Cost……………………………………………. 101

3.2 Impact of Pretreatment Selection……………………………………………………………….. 102

3.3 The Impact of Process Integration on Enzyme Requirements………………………… 104

4 Enzyme Discovery: Catalytic Efficiency and Productivity…………………………….. 106

4.1 T. reesei Cellulases: The Current Industry Standard………………………………………. 106

4.2 Searching for Synergy………………………………………………………………………………… 107

4.2.1 P-Glucosidase………………………………………………………………………………………….. 108

4.2.2 Glycosyl Hydrolase Family 61 109

4.2.3 Synergistic Hemicellulases………………………………………………………………………… 111

5 Producing Enzymes Economically………………………………………………………………. 115

5.1 Reduced Enzyme Recovery………………………………………………………………………… 117

6 Conclusions………………………………………………………………………………………………. 118

References ……………………………………………………………………………………………………. 118

Abstract Enzymes play a critical role in the conversion of lignocellulosic waste into fu­els and chemicals, but the high cost of these enzymes presents a significant barrier to commercialization. In the simplest terms, the cost is a function of the large amount of enzyme protein required to break down polymeric sugars in cellulose and hemicellulose to fermentable monomers. In the past 6 years, significant effort has been expended to re­duce the cost by focusing on improving the efficiency of known enzymes, identification of new, more active enzymes, creating enzyme mixes optimized for selected pretreated substrates, and minimization of enzyme production costs. Here we describe advances in enzyme technology for use in the production of biofuels and the challenges that remain.

Keywords Biomass • Enzymes • Hydrolysis

1

Introduction

The utilization of lignocellulose for the production of fuels and chemicals has the potential to change the world economically, socially, and environ­mentally. Today roughly 87% of the energy used in the world is derived from non-renewable sources such as oil, natural gas, and coal, with total en­ergy consumption increasing at approximately 4% per annum. About 28% of that energy consumption is in the form of liquid transportation fuels, de­rived almost entirely from petroleum [1]. The long-term cost of continued use of these finite fuel sources can already be seen in increased conflict over their control and distribution, climate change linked to increased greenhouse gas emissions, and increasing prices, all of which negatively impact people around the world every day. Lignocellulosic biomass, in the form of plant materials such as grasses, woods, and crop residues, offers a renewable, geo­graphically distributed, greenhouse-gas neutral source of sugars that can be converted to ethanol or other liquid fuels via microbial fermentation.

In the past 30 years, ethanol produced from corn starch and sugarcane has been established as an economically viable supplement to gasoline. In the USA over the past 5 years, production has increased from 175 million gallons per year to nearly 4.5 billion gallons last year, and is growing at more than 25% per year. In the near future, the use of sugar and starch as feedstocks for fuel production is expected to face increasing competition with their direct use as food and animal feed, impacting both availability and price. Current estimates suggest that in the USA, starch-based ethanol output will reach a maximum of between 12 and 15 billion gallons per year [2]. To significantly impact the use of petroleum in the USA, which uses approximately 140 bil­lion gallons of gasoline per year, additional sources of fermentable sugar for ethanol production will be required.

Lignocellulosic biomass has the potential to become a major source of these fermentable sugars in the future. It is estimated that in the USA alone, more than one billion tons per year of biomass could be sustainably harvested in the form of crop and forestry residues, replacing as much as 30% of the total US gasoline consumption [3].

To turn the prospect of replacing a significant proportion of the current liquid fuels into reality, the conversion of lignocellulose to ethanol must be­come less expensive in both operating cost and capital investment. Current estimates suggest that the cost of producing cellulosic ethanol is $1.80/gal — lon or higher, or almost twice as high as the cost of producing ethanol from starch [4]. Part of this high cost results from a significantly higher esti­mated capital investment for the construction of cellulosic plants compared to starch-based production facilities. Cellulose-to-ethanol plants in current design scenarios require more unit operations, must be larger to accom­modate more dilute sugar streams, and in some cases require acid-resistant construction materials, which in sum are projected to increase the invest­ment more than fourfold relative to current dry milling starch-based ethanol plants (from $1.10/gallon installed capacity to $4.70/gal) [4]. On the operat­ing cost side, equipment replacement may be more frequent due to processing materials that are more abrasive than seed, enzyme cost will be significantly higher due to the increased complexity of the substrate and higher enzyme dosage required to release the sugars, and higher water consumption may be required to remove compounds that interfere with the hydrolysis and fermen­tation processes.

Starch is present in plants as an energy source for growing seeds, while lignocellulose is present as a structural cell wall component to give the plant rigidity; therefore it should be no surprise that the latter is much more resis­tant to enzymatic attack. On a protein weight basis, it takes anywhere from 40-100 times more enzyme to break down cellulose than starch, yet the cost of enzyme production is not substantially different (Novozymes, unpublished data).

In 2001, Novozymes was awarded a research subcontract by the US Depart­ment of Energy with the goal of reducing the cost of cellulases for ethanol production from biomass. This effort, called the Cellulase Cost Reduction Project, was administered by the National Renewable Energy Laboratory (NREL), with Novozymes providing expertise for enzyme improvement and production, and NREL contributing expertise in biomass pretreatment and enzyme evaluation. The stated goal of the project was to achieve a tenfold reduction in the cost of enzymes for the conversion of acid pretreated corn stover to ethanol in laboratory-scale testing. At the beginning of this work, the cost of providing a commercial cellulase preparation for the conversion of 80% of the cellulose in acid pretreated corn stover to fermentable glucose was estimated to be $5.40/gallon ethanol produced. During the course of the contract, significant advances were made in improving the efficiency of the cellulases, increasing the yield in production, and reducing the cost of pro­duction. In addition, work focusing on other enzyme activities required for effective enzymatic hydrolysis of lignocellulosic substrates other than acid pretreated corn stover was successfully conducted. In this manuscript, we highlight some of those efforts that have contributed to making enzymes for lignocellulose hydrolysis more affordable.

2

Lignocellulosic raw materials contain much less L-arabinose than D-xylose, and solving the problem of xylose fermentation has been prioritized. The relative amounts of the sugars strongly depend on the raw material. For example, corn stover contains of 19% xylan and 3% arabinan, whereas wheat bran contains 19% xylan and 15% arabinan [58]. As a consequence, L-arabinose-utilizing strains of S. cerevisiae have been developed only re­cently. Furthermore, the conversion of L-arabinose into intermediates of the PPP requires more enzymatic reactions than the conversion of xy­lose (Fig. 2). In many bacteria, such as E. coli, L-arabinose is utilized via L-arabinose isomerase (AraA), L-ribulokinase (AraB), and L-ribulose-5- phosphate 4-epimerase (AraD) [59]. Xylulose-5-phosphate is then further metabolized via the PPP. Enzymatic activities of alternative bacterial arabi — nose and xylose utilization pathways have also been described [60-62].

The fungal arabinose utilization pathway consists of four alternating reduction-oxidation reactions (Fig. 2), where L-arabinose is converted to

D-xylitol via L-arabi(ni)tol and L-xylulose [23,63-65]. D-Xylitol is then further metabolized by XDH and XK, resulting in the PPP intermediate D-xylulose-5-phosphate. The first two complete fungal arabinose utilization pathways were recently kinetically characterized for Candida arabinofermen — tans PYCC 5603r and Pichia guilliermondii PYCC 3012 [65]. The fungal xylose and arabinose utilization pathways share the enzymes XR, XDH, and XK, since XR also reduces L-arabinose [65-68]. Indeed, all arabinose-utilizing yeast and fungi also utilize xylose, whereas not all xylose-growing yeasts uti­lize arabinose [66,69]. Similar to the fungal xylose pathway, the cofactors of the enzymes in the fungal arabinose pathway cannot be regenerated within the pathway but require oxygen or an external electron acceptor for regener­ation (Fig. 2).

3.2

Corn stover is another potentially interesting feedstock for ethanol produc­tion, especially in the USA. The fermentation characteristics of S. cerevisiae RWB 218 on corn stover hydrolysate were tested under industrially relevant fed-batch conditions (W. de Laat, unpublished data). The corn stover pulp obtained after steam explosion (190 ° C, 5 min, ENEA, Italy) was diluted with water to 150gL-1 dry matter and subsequently hydrolysed with 10g cellu — lase protein (kg hydrolysate dry matter)-1 (GC220, Genencor, 96 h at 50 °C). After filtration, the resulting sugar solution contained 40 g L-1 glucose, 9 g L-1 D-xylose and 4 g L-1 acetic acid.

Fermentation experiments were initiated by a 32 h batch phase on mo­lasses medium (containing 100gL-1 sucrose, pH 4.8, 32°C) in a volume of 200 mL. Subsequently, 455 mL of corn stover hydrolysate was added during a 16 h fed-batch phase. During the fed-batch phase, glucose was almost com­pletely consumed. However, only 40% of the D-xylose fed to the culture was consumed during this phase (Fig. 9). After 16 h, the fed-batch phase was terminated and the culture was allowed to consume accumulated sugars. Con­version was complete after 35 h. At a biomass concentration of 1.0—1.5 gL-1, this corresponded to a D-xylose fermentation rate of 0.5 mmolg-1 h-1 dur­ing this latter phase. The overall ethanol yield on total sugars was 0.46 gg-1, which corresponds to 90% of the theoretical maximum yield on glucose and D-xylose. Consistent with the wheat straw hydrolysate fermentations, xylitol formation was not observed.

8

Enzymatic hydrolysis can be facilitated by chipping, milling and grinding the biomass into a fine powder to increase the surface area of the cellulose. In most cases the power consumption is forbiddingly high to reach high digestibility in the enzymatic hydrolysis. It can be even higher than the theor­etical energy content that is available in the biomass [21]. However, physical treatment in an extruder combined with heating and addition of chemicals could be an interesting option [22]. Another method that has been suggested is irradiation of cellulose by gamma rays, which cleaves the ^-1,4-glycosidic bonds, thus giving a larger surface area and a lower crystallinity [23]. This method is, however, far too expensive to be used in a full-scale process. It is also doubtful that it can be used in combination with technologies supposed to be environmentally friendly.

3.2

In nature, microbes can efficiently degrade biomass by secreting an array of synergistic enzymes, including cellulases, often from numerous microbes in­termingled in their growth. In an effort to identify new proteins that could work synergistically with those secreted by T. reesei, we conducted mixing ex­periments by supplementing Celluclast 1.5 L with broth from a wide array of cellulolytic fungi grown under cellulase-inducing conditions. By comparing the saccharification of acid pretreated corn stover using equal protein load­ings of either Celluclast alone or mixtures of Celluclast with these broths, fungi secreting components that could work synergistically with the T. ree­sei cellulases could be detected. In Fig. 7, an example of such an experiment shows that a mixture of T. reesei broth and Thielavia terrestris broth has the same level of hydrolyzing activity as twice as much T. reesei or T. terrestris broth alone. These results suggested that some activity present in the T. ter — restris broth was working in synergy with the cellulases present in T. reesei broth to more efficiently degrade the cellulose in the corn stover.

In order to identify the protein or proteins responsible for the observed synergism with the T. reesei cellulases, the T. terrestris broth was fraction­ated and individual fractions were assayed for synergism similarly. Fractions displaying synergism were separated on one — and two-dimensional polyacry­lamide gels, individual proteins were isolated by removal from the gels and subjected to sequencing by tandem mass spectrometry. Several independent chromatographic fractions contained proteins with homology to glycosyl hydrolase family 61A, a protein previously identified in a number of cellu­lolytic fungi. When purified to homogeneity, a number of these proteins were demonstrated to significantly enhance the activity of the T. reesei cellulases in synergism assays. Inclusion of these proteins at less than 5% of the total enzyme dose in some cases could reduce the required cellulase loading by as much as twofold. These results suggested that the GH61 family proteins were the major components responsible for the enhancement of Celluclast 1.5 L activity by crude T. terrestris fermentation broth.

We also tested the cellulase-enhancing effect of GH61 proteins on a var­iety of other lignocellulosic substrates from a variety of pretreatments when

combined with T. reesei cellulases. Those GH61 proteins capable of enhanc­ing hydrolysis of acid pretreated corn stover also enhanced hydrolysis of other substrates, although they differed in their effectiveness by varying amounts. None of the GH61 proteins were able to enhance the hydrolysis of pure cel­lulose in the form of filter paper. This lack of enhancement was also shown with other pure cellulose substrates such as Avicel, phosphoric-acid swollen cellulose, and carboxymethyl cellulose.

The GH61 proteins by themselves showed no significant detectable hy­drolytic activity on PCS or any other lignocellulosic substrate tested, indi­cating that the enhancement was not likely to be the result of any intrinsic endo — or exoglucanase activity of the GH61 proteins. The hydrolytic activity of several GH61 proteins was tested on a variety of model cellulose and hemi — cellulose substrates, but little or no activity was found. These results suggest that the enhancement of cellulolytic activity by GH61 is limited to substrates containing other cell wall-derived material such as lignin or hemicellulose, although there is no clear correlation between the proportions of these ma­terials and the degree of enhancement observed. These findings could be of significant interest for not only the elucidation of the physiological functions of the GH61 protein family, but also the development of a viable enzymatic system to convert biomass to simple platform sugars.

Several of the GH61 genes were transformed into T. reesei, resulting in transformants expressing GH61 at various levels, depending on the number of inserts and site of integration. Fermentation broths produced by these trans­formants were assayed for PCS hydrolysis at various protein loadings to assess their improvement in specific performance relative to control strains not ex­pressing non-native GH61 proteins. The results confirmed that certain GH61 proteins expressed at relatively low levels are capable of significantly enhanc­ing the hydrolysis of cellulose in PCS. For example, expression of T. terrestris GH61B in T. reesei allows for a reduction in protein loading of 1.4-fold to reach 90% conversion of cellulose to glucose in 120 h. The protein loading reduction made possible by GH61 addition becomes more pronounced at longer incubation times and higher levels of hydrolysis, and higher solids loadings.

4.2.3

The problem of cofactor regeneration has also been addressed by engineering reactions distant from the xylose utilization pathway, as demonstrated by dif­ferent approaches to introduce a transhydrogenase function in S. cerevisiae. Heterologous expression of a bacterial transhydrogenase [118] in S. cerevisiae carrying XR and XDH reduced xylitol formation, but also increased glycerol, rather than ethanol formation [116] (strain TMB3254, Table 2), indicating that the transhydrogenase reaction did not proceed in the direction favorable for ethanolic xylose fermentation [116,118].

Intracellular cofactor concentrations have also been altered, introduc­ing in S. cerevisiae the NAD(P)+-dependent glyceraldehyde-3-phosphate dehydrogenase (GAPDH) from Kluyveromyces lactis [119] (strain H2673, Table 1). When the ZWF1 gene was simultaneously deleted, the expression of GAPDH improved ethanol formation [115] (strain H2684, Table 1). Similarly, when a NAD(P)+-dependent nonphosphorylating GAPDH from Streptococ­cus mutans was overexpressed in an XR-XDH-XK-carrying strain, increased ethanol formation was observed [120] (strain CPBCB4, Table 3). The result suggested that less carbon was lost as carbon dioxide when NADPH was formed outside the oxidative PPP and that NAD+ consumption in the lower glycolysis was simultaneously reduced.

Engineering the ammonium assimilation pathway [121] has also been used to modify the intracellular cofactor concentrations. Based on the as­sumption that NADH would be used for ammonium assimilation to generate NAD+ for the XDH reaction, the NADPH-dependent glutamate dehydroge­nase gene GDH1 was deleted, and an NADH-dependent isoenzyme (GDH2) was overexpressed. Reduced xylitol formation and higher ethanol forma­tion were observed [121] (strain CPB. CR4, Tables 1 and 3). Alternatively, the GS-GOGAT complex coded by the genes GLT1 and GLN1 was overex­pressed, which only affected xylose fermentation in a continuous fermenta­tion setup [121] (strain CPB. CR5, Tables 1 and 3).

4.6

2.1.1

Engineering Scheme

The development of ethanologenic E. coli has included a combination of dir­ected engineering and metabolic evolution; the overall scheme is summarized in Fig. 3. The Z. mobilis homoethanol pathway (PET operon) was introduced

into E. coli in plasmids and these derivatives produced ethanol as the main fermentation product [13-16]. The PET operon was stably integrated into the chromosome at the pfl locus along with an antibiotic resistance marker; spon­taneous mutants exhibiting high ADH activity and high antibiotic resistance were selected to ensure high PET activity. Side reactions that drain carbon away from ethanol were eliminated either by mutation (frd — succinate) or physiologically (differences in Km for pyruvate) (Fig. 2). The resulting strain KO11 produced ethanol at a yield of 95% in complex media [17]. While it was originally reported that KO11 was derived from E coli B, it has recently been discovered that E coli W is the parental strain (Jarboe and Ingram, unpub­lished).

While the rate of ethanol production by KO11 is as high as yeast, the ethanol tolerance is lower than the commercially employed yeast strains. In complex media, KO11 shows a complete lack of growth in the presence of 35 gL-1 ethanol and only 10% survival from 30 s of exposure to 100 gL-1 ethanol [18]. Using strain KO11 as a starting point, mutant strains with sig­nificantly increased ethanol tolerance were isolated. The 3-month metabolic evolution consisted of alternating periods of selection in liquid media for in­creased ethanol tolerance and selection on solid media for increased ethanol production. The final product of this evolution, strain LY01, was able to grow in the presence of 50 gL-1 ethanol and had greater than 80% survival from 30 s of exposure to 100 g L-1 ethanol. The method of metabolic evolution used to derive LY01 from KO11 has proved to be successful and has been applied to the design of other ethanologenic biocatalysts and to the production of other commodity products, as described in Sect. 5.

2.1.2

By selectively choosing pretreatment conditions it should be possible to cre­ate substrates with characteristics (hemicellulose and lignin content, particle size, available surface area etc.) that greatly enhance subsequent enzyme — mediated hydrolysis. For example we can adjust the time, temperature and SO2 concentration of steam pretreatment with consequential effects on over­all product recovery, ease of hydrolysis, etc. In addition to adjusting pretreat­ment parameters, it has been shown that the inherent physical properties of the biomass can be used to anticipate the performance of the pretreated sub­strate in subsequent hydrolysis experiments [45-47]. The particle size, purity, moisture content, and the internal variations in the biomass feedstock, such as the presence of compression or tension wood can all have significant ef­fects on the efficiency of pretreatment. As mentioned earlier, a reduction in the substrate particle size prior to pretreatment is a common practice used in some pretreatment processes [22]. However, size reduction through milling or grinding requires a substantial input of energy, and adds significantly to the total cost of the pretreatment [47]. Therefore, in the case of woody biomass, it is desirable to utilize a biomass particle size that can be produced economically at a large scale with existing equipment, such as the wood chips used in the pulp and paper industry [48].

For the operation of a large-scale bioconversion process a suitable wood chip size should be selected based on a compromise between the energy/cost of producing the chips and the subsequent effectiveness of the pretreatment and product recovery. Some of our previous work has shown that chip size and moisture content have a profound effect on the ease of hydrolyzability of the resulting substrate [46]. For example, by increasing the chip size of Douglas-fir (Pseudotsuga menziesii (Mirb.)) from 0.42 mm2 to 5 cm2 prior to SO2-catalyzed steam pretreatment, greater recovery of the solids and a re­duction in the production of inhibitors could be observed. This could be attributed to a decrease in the “relative severity” of pretreatment undergone by the larger chip at equivalent pretreatment conditions. It was also shown that the lignin present in the larger chips (5 x 5 cm) experienced less con­densation and was therefore more amenable to subsequent alkaline peroxide delignification. This material from the larger chips consequently exhibited a 10-15% increase in hydrolysis yield over that obtained with the mate­rial originating from the smaller chips. Similarly, by increasing the moisture content of the chips from 12 to 30% an improved recovery of glucose and hemicellulose-derived sugars could be achieved. The increased recovery of sugars by raising the moisture content of the chips could be explained by a similar mechanism as it was observed with the increase in chip size. This was due to the additional moisture adsorbing the heat applied during steam pretreatment, resulting in a decrease in the severity of the treatment under­gone by the chips.

In addition to the inherent variations in the properties of incoming biomass, pretreatment schemes must also deal with the fluctuating “purity” of the lignocellulosic substrate, as woody biomass can be expected to contain “con­taminants” such as bark, needles, leaves, branches, etc, that differ significantly in their chemical composition from “white wood” [49]. It is likely that future wood-based bioconversion facilities would involve large amounts of biomass being sent to a chipping unit without careful control of debarking or branch removal. Similarly, it is unlikely that higher-value wood chips will be used extensively in commercial bioconversion facilities due to competition from traditional pulp and paper mills. Therefore a bioethanol facility’s feedstocks are likely to be either coppiced whole plants such as willow or, in the short term, residues from saw and pulp mills such as sawdust, shavings and hogfuel that contain significant amounts of bark, ash and lipophilic extractives. Tree thinnings such as branches have been pretreated using dilute acid during inves­tigations assessing their viability as a biomass feedstock for bioconversion [50], where a two-stage pretreatment was required to hydrolyze 50% of the cellulose while the remaining cellulose was readily hydrolyzed by cellulases.

In similar work, utilizing feedstocks with high bark content, the liquid stream from the steam pretreatment (SP) of a Douglas-fir chip furnish con­taining 30% bark (SP-DFB) was shown to contain lower amounts of total sugars, furfural and 5-hydroxymethylfurfural compared to the liquid stream (prehydrolyzate) isolated from the pure whitewood (SP-DF) [51]. Although the concentration of lipophilic extractives increased in the bark-laden water — soluble stream, subsequent fermentation by Saccharomyces cerevisiae re­sulted in a complete utilization of the hexose sugars within 48 h with compa­rable ethanol yields regardless of whether bark was present or not. Although these results looked encouraging, with regard to the “robustness” of the over­all SP process, the solid fraction of the pretreated Douglas-fir containing bark showed significant differences when compared to the whitewood. The addition of 30% bark to a Douglas-fir chip furnish prior to SP resulted in a significant increase in the lignin detected in the water-insoluble fraction. The SP-DFB solids fraction also contained 56% lignin compared to only 44% lignin in the case of the pure whitewood (SP-DF) sample. Consequently, the lignin content decreased to only 18% for the SP-DFB fraction as compared to 9% in the case of SP-DF upon subsequent alkaline peroxide delignifica — tion [52]. Although the alkaline peroxide delignified SP-DFB had a higher lignin and phenolic extractive content, it resulted in a similar hydrolysis yield to the SP-DF fraction, most likely due to the removal of surface lignin dur­ing the alkaline peroxide delignification stage [53]. Although hydrolysis was not performed in the absence of the delignification step, it can be expected that, due to its higher lignin content, the SP-DFB fraction would be more re­sistant to hydrolysis than was the SP-DF substrate. It is apparent that, during SP of lignocellulosic feedstocks such as Douglas-fir, the inherent properties of the wood have a significant effect on the downstream partitioning of the cellulose, hemicellulose and lignin components.

3

The tested thermostable fungal enzymes, classified as cellobiohydrolases (CBHs) or endoglucanases (EGs) based on the activity determinations, were chosen by preliminary screening and characterisation. Several thermostable CBHs from various thermophilic organisms were purified and characterised (Voutilainen et al, manuscript in preparation). The gene of the most poten­tial CBH isolated from Thermoascus aurantiacus was fused with the T. reesei CBHI cellulose binding domain (CBM). In addition, an EG from Acremo — nium thermophilum, a в-glucosidase and a xylanase from T. aurantiacus were used to compose the thermostable mixtures. Fermenter supernatants pro­duced in pilot scale were used to obtain the thermostable cellulase mixtures. The optimal ratio of EG to CBH amount (measured as protein of the enzyme mixtures) was determined on the basis of FPU activity of the preparations. The highest FPU activity was obtained by an EG to CBH protein ratio of 3 : 10, which corresponded well to the respective ratio of the native T. ree — sei enzymes. This ratio also gave the highest sugar yields in the hydrolysis of the steam pretreated corn stover substrate (results not shown) and was used as the standard basis for various mixtures. Three different mixtures were used in this work, differing with respect to the xylanase activity (Table 3). The xylanase-free preparation (TM 1) was first used for the spruce substrate

and the xylanase-containing preparations (TM 2 and TM 3) for the corn stover substrate. As it has frequently been observed that xylanases enhance the hydrolysis of lignocellulosic substrates containing even low amounts of re­sidual xylan [9], preparations with xylanase activity were later used for both substrates.

7

Although S. cerevisiae cannot grow on D-xylose as the sole carbon source, its genome does contain genes that code for a non-specific NADH-dependent al­dose reductase (GRE3) and for a xylitol dehydrogenase (XYL2). It has been shown that over-expression of these native S. cerevisiae genes using endoge­nous promoters enabled a specific growth rate of 0.01 h-1 on D-xylose in shake flasks [64]. However, in these shake-flask cultures this engineered yeast strain converted D-xylose into xylitol with a yield of 55%. Under anaero­bic conditions, precluding respiratory NAD+ regeneration, the strain over­expressing the endogenous enzymes was unable to utilise D-xylose [64].

In addition to this metabolic engineering approach, the presence of en­dogenous genes for D-xylose-converting enzymes has been used in recent experiments by Attfield and Bell (2006), describing a non-recombinant S. cere- visiae strain that grows on D-xylose as the sole carbon source in aerobic shake flask cultures. In their study a combination of population genetics and evolutionary engineering [5,60] resulted in an increase in growth rate from extremely low, barely measurable growth rates to a specific growth rate of around 0.12 h-1 (a doubling time of less than 6 h) over a period of 1400 days. Apparently, this S. cerevisiae strain had evolved in such a way that the very low “background” xylose reductase and xylitol dehydrogenase activities, which were previously described as insufficient for growth on D-xylose [8], increased to levels that did enable growth. Indeed, subsequent analysis of the evolved strain showed that xylose reductase activity had increased fourfold and the xylitol dehydrogenase activity 80-fold relative to the parental strain. The ac­tual genes that underwent mutation have not yet been characterised. Although this very interesting study underlines the tremendous potential of evolution­ary approaches, the selection procedure inevitably resulted in a yeast strain displaying the characteristics of redox imbalance, such as xylitol production.

1.4