Cellulose

, Volume 16, Issue 4, pp 743–762

Role of pretreatment and conditioning processes on toxicity of lignocellulosic biomass hydrolysates

Authors

    • National Renewable Energy Laboratory
  • Min Zhang
    • National Renewable Energy Laboratory
Article

DOI: 10.1007/s10570-009-9309-x

Cite this article as:
Pienkos, P.T. & Zhang, M. Cellulose (2009) 16: 743. doi:10.1007/s10570-009-9309-x
  • 1.7k Views

Abstract

The Department of Energy’s Office of the Biomass Program has set goals of making ethanol cost competitive by 2012 and replacing 30% of 2004 transportation supply with biofuels by 2030. Both goals require improvements in conversions of cellulosic biomass to sugars as well as improvements in fermentation rates and yields. Current best pretreatment processes are reasonably efficient at making the cellulose/hemicellulose/lignin matrix amenable to enzymatic hydrolysis and fermentation, but they release a number of toxic compounds into the hydrolysate which inhibit the growth and ethanol productivity of fermentation organisms. Conditioning methods designed to reduce the toxicity of hydrolysates are effective, but add to process costs and tend to reduce sugar yields, thus adding significantly to the final cost of production. Reducing the cost of cellulosic ethanol production will likely require enhanced understanding of the source and mode of action of hydrolysate toxic compounds, the means by which some organisms resist the actions of these compounds, and the methodology and mechanisms for conditioning hydrolysate to reduce toxicity. This review will provide an update on the state of knowledge in these areas and can provide insights useful for the crafting of hypotheses for improvements in pretreatment, conditioning, and fermentation organisms.

Keywords

LignocelluloseInhibitorsPretreatmentConditioningHydrolysateBiofuel

Introduction

Pretreatment of biomass plays a critical role in producing materials with acceptable enzymatic digestibility and subsequent fermentability for the production of cellulosic ethanol or other advanced biofuels such as butanol derived from biomass. The various pretreatment processes, in addition to generating suitable substrates for conversion to biofuel, typically produce a range of compounds that inhibit the organisms used for fermentation. These inhibitory components include carboxylic acids, primarily acetic acid, the sugar degradation products furfural and hydroxymethylfurfural (HMF), phenolic compounds, and inorganic salts (Klinke et al. 2004). These components have been shown to inhibit the growth of the fermentation organisms thereby reducing the rate of ethanol production and in some cases, the overall the yield. The presence of inhibitors in hydrolysate and the additional process costs of detoxification or conditioning steps added to improve fermentability are major impediments in the development of an economic process for the production of cellulosic ethanol. Ultimately, it will be necessary to solve the problem of hydrolysate toxicity without adding significantly to process costs.

A search of the literature quickly reveals the complicated nature of the topic of hydrolysate toxicity, brought about by the multitude of biomass feedstocks, pretreatment and conditioning methods, fermentation methods, and fermentation strains tested. Different biomass feedstocks and pretreatment processes generate different combinations of toxic compounds; different fermentation strains have different levels of natural resistance; and changes in the fermentation processes can lead to different levels of resistance. Economic constraints push for process options that increase the impact of hydrolysate toxicity (minimal conditioning, increased solids loadings and water recycle), and at the same time require improvements in key fermentation parameters (cell growth rates and yields, ethanol yields, ethanol production rates and titers) which are most affected by the toxic compounds.

This review will bring together the information regarding hydrolysate toxicity from both peer-reviewed journal articles and from internal National Renewable Energy Laboratory documents. Several excellent reviews have been written that detail the analysis of hydrolysates from a variety of biomass feedstocks using a number of different pretreatment and conditioning regimes (Klinke et al. 2004; Palmqvist and Hahn-Hägerdal 2000a, b; Sun and Cheng 2002; Mussatto and Roberto 2004). The reader is urged to consult those reviews, but whenever possible this review will refer to the primary literature and will attempt to integrate both analytical and biological data.

Pretreatment

Many different pretreatment processes have been developed over the years. These are designed to make the sugars more available for subsequent hydrolysis and fermentation steps through the breakdown of the cell wall, and the degradation of the cellulose, hemicellulose and lignin matrix. A list of characteristics of the “ideal” pretreatment process has been assembled by Mosier et al. (2005). An ideal pretreatment process:
  • Produces a highly digestible pretreated solid.

  • Does not significantly degrade pentoses.

  • Does not significantly inhibit subsequent fermentation steps.

  • Requires little or no size reduction of biomass feedstock.

  • Can work in reactors of reasonable size and moderate cost.

  • Produces no solid-waste residues.

  • Has a high degree of simplicity.

  • Is effective at low moisture content.

In addition to these, an ideal pretreatment process would also be sustainable with regard to energy consumption and ecological impact, and would not adversely impact downstream processes. It should be noted that the severity of the various conditioning methods described below vary widely. While it is true that the less severe methods often result in formation of less toxic hydrolysate, they also may result in reduced hydrolysis of hemicellulose to monosaccharides, requiring the use of additional enzymes during the saccharification process, increasing the overall process costs.

Elander et al. (2002) described the pretreatment options available at the time. This list is summarized in Table 1 and each technology will be described at some length below.
Table 1

List of pretreatment technologies under evaluation

Pretreatment category

Specific pretreatment technology

Base catalyzed

AFEX/FIBEX

Ammonia recycle percolation

Lime

Non catalyzed

Autohydrolysis-steam

Hot water

Hot water pH neutral

Acid catalyzed

Carbonic

Nitric

Sulfuric acid

Solvent based

Organosolv

Chemical based

Wet oxidation (O2, peroxide, ozone)

AFEX/FIBEX (Ammonia fiber explosion/fiber extrusion explosion): Biomass of small particle size is treated with high pressure steam in the presence of liquid ammonia and then allowed to decompress rapidly. This process decrystallizes cellulose, prehydrolyzes hemicellulose, reduces lignin content, disrupts the fiber structure, and provides a nitrogen source for subsequent fermentation. This method has been applied to a large number of biomass feedstocks but appears to be less effective against biomass samples with high lignin content (Holtzapple et al. 1991). AFEX is a batch process; FIBEX is a continuous process.

Ammonia recycle percolation: This process also uses ammonia, percolated though a packed bed of biomass to solubilize hemicellulose and lignin. The ammonia is recovered and recycled to the reactor. Solubilized components, including carbohydrates and lignin fragments, are continuously removed from the solids. The elimination of lignin components enhances subsequent enzymatic hydrolysis of the cellulose fraction.

Lime pretreatment: A number of base-catalyzed pretreatment options have been evaluated, but lime is the obvious choice based on low cost. This is primarily a delignification process with little hydrolysis of hemicellulose. It is much more effective with herbaceous biomass than with woody biomass.

Steam autohydrolysis: Biomass of small particle size is treated with high pressure steam and then allowed to decompress rapidly. Hemicellulose solubilization and degradation and lignin structural destruction makes cellulose and remaining hemicellulose available for hydrolysis. This is a simple and cost effective means to pretreat hardwood and herbaceous biomass but less effective for softwood (Clark and Mackie 1987). There can be significant loss of xylose to furfural and other degradation products. The effectiveness has been enhanced by addition of H2SO4 or SO2 (Su et al. 1980) or CO2 (Grous et al. 1986).

Hot water pretreatment: Unlike steam pretreatment processes, this method uses water under pressure to maintain the liquid state at temperatures above 200 °C. The water is percolated through a packed bed of biomass resulting in high yields of xylose (primarily as oligomers). As with steam autohydrolysis, the lignin components are largely solubilized and removed from the solids, resulting in higher rates of enzymatic hydrolysis of cellulose.

pH Controlled Hot Water Pretreatment: Unlike the previous process, the pH is controlled during pretreatment to maintain in the range of 4–7 to minimize acid-catalyzed degradation of sugars. This method allows for greater enzymatic accessibility by increasing cellulose pore size and reducing crystallinity.

Carbonic Acid Pretreatment: This method is similar to ammonia explosion and steam and steam autohydrolysis except that CO2 is used under high pressure and the formation of carbonic acid is thought to enhance the hydrolysis (Dale and Moreira 1982). Yields are relatively low compared to steam and ammonia explosion, CO2 explosion is more cost effective than ammonia explosion and yields less inhibitory byproducts than steam explosion (Zheng et al. 1998). The resulting hydrolysate does not require neutralization because the CO2 is released upon decompression.

Dilute Sulfuric Acid Pretreatment: This approach is one of the oldest and most studied pretreatment options, having been known since 1819 (Galbe and Zacchi 2002). Dilute acid hydrolysis makes use of lower concentrations of acid (typically 0.5–5% H2SO4 at elevated temperatures and pressures, and typically achieves hydrolysis of hemicellulose and yields enzymatically accessible cellulose. Different process options include, batch, plug flow and flow through. The main drawback to dilute acid hydrolysis is the degradation of C5 sugars to yield furfural and C6 sugars to yield HMF as well as other toxic byproducts. Other issues include highly corrosive reaction environment, formation of gypsum (after neutralization with lime), and high use of steam. Acid-catalyzed hydrolysis of switchgrass and wheat straw biomass was compared with alkaline-catalyzed hydrolysis and no-catalyst hydrolysis for release of sugars at various severities (Nagle et al. 2006). Acid hydrolysis across the severity range was comparable with both sources of biomass, but wheat straw was more reactive than switchgrass under alkaline hydrolysis.

Dilute Nitric Acid Pretreatment: This is an alternative to sulfuric acid pretreatment. Nitric acid is less corrosive than sulfuric acid, and so reactors can be constructed of less expensive materials. Gypsum formation is also eliminated. These considerations are offset by the added expense of nitric acid relative to sulfuric.

Solvent Pretreatment: These processes typically use a solvent such as methanol, ethanol, or acetone along with dilute acid to achieve fractionation of hemicellulose, cellulose, and lignin, resulting in a highly digestible substrate. It is generally thought to be too expensive, because it relies on recovery and recycle of solvents.

Wet oxidation: The main effort makes use of oxygen at high pressure under alkaline conditions (Skammelsen and Thomsen 1998), though treatment with H2O2 in the presence of peroxidase (Azzam 1989) or with ozone have also been explored. This type of pretreatment is effective at lignin degradation as well as hemicellulose solubilization (mainly as oligomers) resulting in a highly digestible pretreated solid fraction. This approach greatly reduces the amount of furfural and HMF produced if pH is controlled during pretreatment (Bjerre et al. 1996).

Analysis of hydrolysates

Biomass feedstocks for ethanol production are made up mainly of cellulose, hemicellulose and lignin in various proportions (Table 2). As noted above, pretreatment of biomass refers to a process that disrupts the cellulose, hemicellulose and lignin matrix, making the carbohydrates accessible to further enzymatic hydrolysis and fermentation of the resulting sugars to ethanol. Pretreatment methods yield both a soluble and insoluble fraction. The soluble fraction is made up of some combination of solubilized hemicellulose and hydrolysis products, including xylose oligomers and monomers, glucose and other C6 sugars, as well as the toxic byproducts (Klinke et al. 2004). The insoluble fraction is greatly enriched for cellulose and lignin, and is and of itself, non-toxic, though a substantial amount of toxic compounds may be present due to soluble material entrained within its matrix (Berson et al. 2006). The pretreatment conditions will affect the ability of cellulases to hydrolyze cellulose in the insoluble fraction, but that aspect of the overall process is beyond the scope of this review. For the sake of this review, we will refer to the soluble fraction as hydrolysate because most pretreatment methods will result in the hydrolysis of some or all of the hemicellulose, and will focus our attention on the toxic nature of this material as well as the efforts to condition or detoxify this material to allow for more effective conversion of the carbohydrate to ethanol.
Table 2

Chemical composition of the three types of lignocellulosic biomass (from Klinke et al. 2004)

 

Cellulose (%)

Hemicellulose (%)

Lignin (%)

Softwood

41–50

11–33

19–30

Hardwood

39–53

19–36

17–24

Herbaceous plants

24–50

12–38

6–29

Each pretreatment method and each biomass feedstock will generate hydrolysates containing different sets of toxic compounds, but in general, three classes of toxic compounds can be found. These are furans, furfural and HMF produced by the degradation of xylose and glucose, respectively; aliphatic acids, especially acetic acid produced by the deacetylation of hemicellulose and lignin, formic acid produced by the degradation of furans, and levulinic acid produced by the degradation of HMF; and phenolic compounds formed by the breakdown of lignin. In addition to these major components, a wide range of aromatic acids, aldehydes and ketones can be found in varying degrees (Klinke et al. 2004).

A great number of analytical methods have been employed to determine the identity and concentration of compounds in hydrolysate. The major components, glucose, xylose, acetic acid, and furans, as well as a wide variety of aliphatic and aromatic compounds can be monitored directly from hydrolysate samples by high performance liquid chromatography (HPLC) using a variety of columns and detectors. Minor components have been extracted from hydrolysate using organic solvents, concentrated by evaporation, and analyzed by gas chromatography (GC) or HPLC. Ranatunga et al. (1997a) used methyl tert-butyl ether (MTBE) to extract dilute acid hydrolysate of mixed sawdust feedstock to remove potentially toxic compounds. They identified gallic acid, furfural, HMF, protocatechuic acid, vanillin, coniferyl alcohol, syringaldehyde, and synapic acid by HPLC analysis of the extract. They did not test the MTBE extract or the aqueous phase for toxicity. Dilute acid hydrolysates of corn stover, switchgrass, and poplar were extracted by ethyl acetate and the extract analyzed by GCMS (Fenske et al. 1998). Fourteen aromatic monomers were identified and quantified (and nearly an equal number of peaks were left unidentified). The total aromatic monomer concentrations for stover, switchgrass and poplar were 112, 141, and 247 mg/L, respectively. These concentrations were roughly proportional to the toxicity for Pichia stipitis fermentations. Syringyl compounds varied the most in woody vs. herbaceous hydrolysates being approximately fourfold higher in poplar than in stover or switchgrass. More recently, Chen et al. (2006) expanded upon the method described above (Ranatunga et al. 1997a), using MTBE extraction of corn stover hydrolysate followed by HPLC analysis using a reverse phase column with UV detector. They compared this extract with a standard mixture made up consisting of 32 representatives of all suspected classes of inhibitors (compounds were chosen based on identification in hydrolysates and commercial availability) including aliphatic and aromatic acids, furans, and phenolics. Fifteen of these were identified in the MTBE extract, as well as a number of unidentified compounds and a broad unresolved mixture of compounds eluted late in the chromatogram. Of the fifteen, the most abundant compounds were aliphatic acids especially lactic and acetic, furfural and HMF and 3-methoxycinnamic acid.

Inductively coupled plasma spectrometry has been used for inorganic ion analysis of hydrolysates (Ranatunga et al. 2000). Fifteen ions were analyzed in a dilute acid hydrolysate of yellow poplar resulting in concentrations ranging from below detection limits for lead and arsenic, 0.09 ppm for copper, 29 ppm for iron, 198 ppm for potassium, and 258 ppm for calcium. All of the ions were thought to come from the wood itself, with the possible exception of iron (arising from reaction vessel).

Additional analyses of hydrolysates coupled with toxicity measurements will be summarized in the following sections.

Manifestation of toxicity

The sheer number of potentially toxic compounds in hydrolysates and the limited ability to identify all of them (Fenske et al. 1998; Chen et al. 2006) greatly complicate any effort to understand the contribution of individual compounds in hydrolysate to the overall toxicity or to detail the biological effect of these components. Because of this, the bulk of work has been done with individual or simple mixtures of model compounds, though some attempts have been made to draw inferences from hydrolysates by relating the inhibition of fermentation with the concentrations of known toxic compounds present. Examples of both approaches will be summarized below, but it will be helpful to make some generalizations about the toxic effect of the major classes of compounds:
  • Furfural and HMF inhibit glycolysis, especially interfering with the activity of dehydrogenases, causing a reduction in growth rates and cell yields (Banerjee et al. 1981).

  • Phenolics partition into membranes and lead to loss of integrity, interfering with cell growth and sugar transport (Heipieper et al. 1994).

  • Acids disrupt cellular energy generation by collapsing pH gradients especially at low pH. The relative toxicity is a function of hydrophobicity because this characteristic determines the ability of the compound to pass through the membrane (Zaldivar et al. 1999).

  • Aldehyde toxicity is also related to hydrophobicity, but aldehydes do not disrupt membrane integrity or cause a collapse of pH gradient (Zaldivar and Ingram 1999).

  • Alcohols are generally less toxic than related acids or aldehydes, but their toxicity is also related to hydrophobicity. They appear to cause a breakdown in membrane structure (Zaldivar and Ingram 2000).

Model compounds

Delgenes et al. (1996) compared the ability of S. cerevisiae CBS1200 and Z. mobilis ATCC10988 to ferment glucose and of P. stipitis NRRL Y7124 and Candida shehatae ATCC22984 to ferment xylose in the presence and absence of a number of inhibitors. Monitoring cell densities at ethanol titres at a specific endpoint, they determined that furfural, HMF, vanillin, hydroxybenzaldehyde, and syringaldehyde were more toxic than acetic acid, although Z. mobilis showed a 24 and 74% reduction in growth levels at 5 and 15 g/L acetic acid, respectively. Z. mobilis 206C(pZB301) was also sensitive to acetic acid in continuous culture; at an acetic acid concentration above >4.5 g/L at pH 5.5, ethanol production dropped, glucose utilization remained constant, xylose and arabinose utilization decreased, and xylitol levels increased (Mohagheghi et al. 2002). C. shehatae and P. stipitis were especially sensitive to phenolics, and P. stipitis was especially sensitive to furfural and HMF. Z. mobilis was relatively resistant to all compounds except for hydroxybenzaldehyde. In general, the ethanol titres were proportional to growth with the exception of effect of acetate on ethanol production from glucose with Z. mobilis. Since there was only one timepoint taken for all cultures, it is unknown whether these strains could overcome the inhibitory effect of these strains to achieve maximal cell densities and ethanol concentrations. The strains were also tested for the ability to metabolize these inhibitory compounds; all four strains could completely metabolize furfural within 24 h. S. cerevisiae could also completely metabolize HMF, vanillin, hydroxybenzaldehyde and syringaldehyde within 32 h, and only C. shehatae showed any metabolic activity toward acetate. C. shehatae NJ23 was also shown to produce a slightly higher ethanol yield in the presence of acetic acid at low concentrations, as was baker’s yeast and S. cerevisiae ATCC 96581 (Palmqvist et al. 1999).

Taherzadeh et al. investigated in detail the effect of furfural on growth of S. cerevisiae CBS8066 on glucose (Taherzadeh et al. 1999). They found that under anaerobic conditions, CO2 evolution, growth, and ethanol production were inhibited by a pulse of furfural, and that furfural appeared to interfere with glycolysis confirming earlier observations (Banerjee et al. 1981). Growth and ethanol production rates recovered when furfural was completely metabolized. Both furfuryl alcohol and furoic acid were detected as major and minor products, respectively, though mass balance analysis indicates that at least one other significant product was formed from furfural metabolism. Under aerobic growth, furfuryl alcohol can be converted in high yield to furoic acid. Furfural does not only inhibit glycolysis, but will also inhibit aerobic growth on ethanol and acetate (Taherzadeh et al. 2000). This inhibition continues until the furfural is metabolized. When ethanol is the carbon source, furfural appears to inhibit acetaldehyde dehydrogenase, leading to potentially toxic levels of acetaldehyde. This effect plus the shift of electrons needed for furfural reduction away from terminal respiration and ATP synthesis explains the toxic effect. Of the three furans tested, furfural is the most toxic, followed by furfuryl alcohol and furoic acid. The difference in toxicity of furfural and HMF was also observed with P stipitis NRRL Y-7124, S. cerevisiae NRRL Y-12632, and S. cerevisiae ATCC 211239 (Liu et al. 2000). When S. cerevisiae CBS8066 was grown in continuous culture on glucose under aerobic conditions, addition of furfural caused rates of glycolysis, TCA reactions, and respiration to go up, and biomass yield on ATP to go down (Horvath et al. 2003). Furoic acid was the main product of furfural metabolism. In anaerobic chemostats, the furfural was reduced to furfuryl alcohol, diverting electrons away from flux to glycerol, resulting in an increase on biomass yield at low furfural concentrations (1–4 g/L). At 5.8 g/L furfural, the biomass yield dropped significantly and nonspecific ATP consumption went up. The decrease in biomass yield was not matched by a comparable decrease in ethanol yield and so the specific rate of ethanol production was highest at 5.8 g furfural per liter. Baker’s yeast, S. cerevisiae ATCC 96581 and C. shehatae NJ23 all showed slight increases in ethanol yield in the presence of furfural at low concentrations, accompanied by decreased production of glycerol (Palmqvist et al. 1999).

Pichia stipitis NRRL Y-7124, S. cerevisiae NRRL Y-12632, and S. cerevisiae ATCC 211239 were tested for growth in synthetic medium amended with varied amounts of furfural and HMF (Liu et al. 2004). Furfural is more toxic on a molar basis than HMF. At intermediate levels, growth inhibition was manifested by prolonged lag phase. All three strains are approximately equally sensitive. Glucose utilization and ethanol production in presence of HMF did not proceed until nearly all the HMF was converted to 2,5-bis-hydroxymethylfuran (the analog of furfuryl alcohol). Other strains of S. cerevisiae (Tembec T-1, adapted for growth in spent sulfite liquor (SSL) and Y-1528, a native strain) were also tested for inhibition of glucose fermentation by furfural and HMF (Keating et al. 2006). Both strains showed dose dependent responses to both furans manifested by decreased growth rates and ethanol productivity. Cells treated with HMF also showed a decreased ethanol yield. Helle et al. (2003) looked at additional strains of S. cerevisiae including the SSL-adapted strain (Tembec T-2) as well as wine making yeast (K1), a native strain (259A) and a recombinant derivative of 259A strain engineered for xylose fermentation (259ST). Furfural inhibition was transient and resulted in lag phases, but growth rates and final ethanol yields were not much affected.

Some of these same strains were also tested for acetic acid inhibition of glucose fermentation. Tembec T-1 and Y-1528 were equally insensitive to acetic acid in terms of growth rate or ethanol productivity (Keating et al. 2006). Acetic acid stimulated ethanol yields at lower concentrations showing toxic affects only at the highest concentrations (~15 g/L). S. cerevisiae strains Tembec T-2, K1, 259A, and 259ST were also tested for growth in the presence of acetic acid which caused reduced growth rates and final cell yields for all strains (Helle et al. 2003). Lag phases were observed which differed from strain to strain. Ethanol yields in 259ST grown on a mixture of glucose and xylose were inhibited by acetic acid at concentrations of 2–6 g/L.

The recombinant ethanologenic strain of E. coli LY01 was tested for its resistance to a series of acids (Zaldivar et al. 1999), aldehydes (Zaldivar and Ingram 1999), and alcohols (Zaldivar and Ingram 2000). Cells were tested with individual compounds at a number of concentrations, and growth inhibition was determined at a single time point. The data were used to calculate minimum inhibitory concentrations (IC100). E. coli was most sensitive to aldehydes and alcohols with MICs on the order of 1 g/L whereas the MICs for acids were an order of magnitude higher. The order of sensitivity for acids was caproic > vanillic = ferulic = 4-hydroxybenzoic > syringic = furoic = formic > acetic > levulinic = gallic > ethanol. Single point measurements of ethanol production followed closely the patterns of growth inhibition with the exception of formic and furoic acids in which ethanol production was inhibited to a greater extent than growth. Inhibition of growth and ethanol production was a function of the hydrophobicity of the compound as well as the culture pH indicating that toxicity was most likely due to collapse of ion gradient. The order of sensitivity for aldehydes was as follows: hydroxybenzaldehyde > vanillin > syringaldehyde > furfural > HMF > ethanol. The effects of aldehydes could be partly overcome by high inocula; at the MIC, growth and alcohol production could reach the control level by 48 h, though higher concentrations were still toxic. Use of higher inocula levels appears to be a general approach to reduce the effects of hydrolysate toxicity (see, for example Jennings and Schell 2005). This topic will be discussed more completely in the section on Enhanced Resistance, below. Toxicity was related to hydrophobicity but there was not indication that aldehydes effected membrane integrity. The order of sensitivity to alcohols was as follows: methylcatechol > hydroquinone = guaiacol = catechol = coniferyl alcohol > vanillyl alcohol > furfuryl alcohol. Higher inocula levels did not reduce the toxic effect of alcohols. Toxicity correlated with hydrophobicity, and the mode of action for alcohols appears to be different than the mechanism observed with acids or aldehydes, most likely loss of membrane integrity. Interactions between individual components will be discussed in the section on Synergy, below, but in summary, it appears that toxicity of hydrolysate is consistent with concentrations and interaction of individual components.

Thermoanaerobacter mathranii was tested for its ability to ferment a hydrolysate generated by alkaline wet oxidation of wheat straw (Klinke et al. 2001). This hydrolysate had been analyzed and found to contain a number of toxic aromatic compounds including 4-hydroxybenzaldehyde, vanillin, syringaldehyde, 4-hydroxyacetophenone, acetovanillone, acetosyringone, 4-hydroxybenzoate, vanillic acid, syringic acid, 2-furoic acid) as well as xylose (Klinke et al. 2002). In synthetic medium growth was unaffected by seven of ten compounds tested at 2 mM but was inhibited by 4-hydroxybenzaldehyde and stimulated by syringic acid. Growth inhibition at 10 mM was much more pronounced in hydrolysate by all except acetosyringone and vanillic acid. Ethanol yields on xylose in medium were not as sensitive to the inhibitors as growth except for 4-hydroxybenzaldehyde and vanillin.

Ezeji et al. (2007) investigated the impact of the degradation products from agricultural residues on growth and butanol fermentation in Clostridium beijerinckii. They found that the dilute sulfuric acid corn fiber hydrolysate was toxic to C. beijerinckii BA101 resulting in poor cell growth and reduced butanol production. Detoxification using overliming improved the cell growth and fermentability. Detailed testing of various compounds present in the hydrolysate showed that sulfate at 13.4 g/L (a product of pretreatment) reduced cell growth by 53% and also significantly reduced fermentation product yields while acetate at 8.9 g/L enhanced the cell growth by 14% compared to the control. Furfural and HMF are not inhibitory to C. beijerinckii BA101, rather they are stimulatory. However, syringaldehyde, ferulic and p-coumaric acids were potent inhibitors of the ABE production by C. beijerinckii BA101. Clearly this butanol producer has very different inhibitory profiles than ethanologens.

Hydrolysates

Efforts to determine the toxicity of various compounds present directly in hydrolysates usually involve differential fermentation performance in hydrolysate before and after various methods of conditioning with inferences drawn by analytical determination of differences in toxic compounds. This approach will be described in detail in the section on conditioning (below). There have been a few cases where the conditions of pretreatment were varied and the relative toxicity of the resulting hydrolysates determined. Larsson et al. generated a series of acid hydrolysates of spruce under conditions of increasing combined severity (CS) attained by varying H2SO4 concentration, temperature and treatment time. The effect of increasing CS on fermentability was tested using baker’s yeast (Larsson et al. 1999). As the CS increased, sugar release increased and then fell. At higher CS, degradation products began to appear with furfural and HMF peaking first followed by formic, acetic and levulinic acid. Ethanol yields and ethanol productivities were constant until the furfural and HMF concentrations reached their peaks and then fell. In model fermentations with concentrations of acetic, formic, levulinic adjusted to represent hydrolysates from conditions of moderate to high CS, the ethanol yields and productivities fell as function of severity. Addition of furfural, and HMF to the medium containing the aliphatic acid mixtures had no additional effect on yields, but further reduced the volumetric productivities, similar to results seen with furfural and HMF alone (Taherzadeh et al. 1999; Liu et al. 2004; Horvath et al. 2003). Similar work was performed with hydrolysates generated from corn stover and aspen wood using carbonic acid and liquid hot water pretreatment carried out under a variety of severities (Yourchisin and Van Walsum 2004). Aspen hydrolysates were more toxic to baker’s yeast than corn stover hydrolysate, and glucose consumption decreased in hydrolysates produced with higher severity pretreatments.

Attempts have been made to test for toxicity of compound mixtures obtained directly from woody biomass. Extracts of yellow poplar, white oak, and red oak were prepared with acetone:water (7:3 v/v) and hot methanol, freeze dried, and used to amend synthetic media to test for toxicity of S. cerevisiae D5a (Ranatunga et al. 1997b) and Z. mobilis CP4(pZB5) (Ranatunga et al. 1997a). The oak extracts contained primarily hydrolysable tannins, and the yellow poplar extracts contained a mixture of alkaloids, sesquiterpenes, and lignans. None of the extracts were toxic to S. cerevisiae even at high loadings, but the red and white oak extracts were especially toxic to Z. mobilis. In these same papers, hydrolysate was prepared from mixed wood biomass, and tested for fermentation by both S. cerevisiae and Z. mobilis. Z. mobilis performed very poorly with both neutralized and overlimed hydrolysates. S. cerevisiae, on the other hand, performed as well with overlimed hydrolysate as with pure sugar control, and also performed reasonably well with neutralized hydrolysate (61% of control). Z. mobilis was tested for inhibition by individual compounds added to a sugar control at the concentrations found in hydrolysate. Acetic acid at 9 g/L completely inhibited ethanol production, and other compounds such as furfural (1 g/L), HMF (0.09 g/L), gallic acid (0.17 g/L), syringaldehyde (0.13 g/L), and vanillin (0.04 g/L) inhibited ethanol production by 20–40%.

Synergy

Synergistic toxic effects are thought to occur when combinations are more inhibitory than the sum of the individual effects. Palmqvist et al. (1999) tested for synergy with combinations of acetic acid, furfural, and p-hydroxybenzoic acid using baker’s yeast. This strain was resistant to acetic at 10 g/L as well as to furfural and p-hydroxybenzoate at 2 g/L, but the combination of acetic and furfural at the same concentrations had a negative effect on growth and ethanol yield. Furfural also demonstrated a synergistic inhibitory effect when used in combination with HMF in fermentations of P. stipitis NRRL Y-7124, S. cerevisiae NRRL Y-12632, and S. cerevisiae ATCC 211239 (Liu et al. 2004). Synergistic inhibition of ethanol yields were observed in S. cerevisiae strains 259ST with combinations of furfural–ethanol, furfural–acetic, and ethanol–acetic, but combinations were additive with regard to growth inhibition (Helle et al. 2003).

Furfural was also shown to manifest synergistic effects with other compounds when tested with the ethanologenic strain of E. coli LY01 (Zaldivar et al. 1999; Zaldivar and Ingram 1999; Zaldivar and Ingram 2000). Combinations of furfural with acetic acid, HMF, 4-hydroxybenzaldehyde, syringaldehyde, vanillin, furfuryl alcohol, and guaiacol were especially toxic. Some combinations were less than the sum of individual components, indicating that one compound could interfere with the toxic action of the other. Examples of these protective interactions are vanillyl alcohol with catechol, coniferyl alcohol, guaiacol, hydroquinone and methylcatechol as well as the combination of furfural with methylcatechol.

Synergistic effects have also been seen in T. mathranii, tested both with compounds added in combination in synthetic media and with compounds added to hydrolysate (Klinke et al. 2001). At 2 mM, syringaldehyde and vanillin did not inhibit growth in synthetic media, but were inhibitory when added at that level to hydrolysate prepared by alkaline wet oxidation of wheat straw. At 10 mM 4-hydroxybenzaldehyde, vanillin, syringaldehyde, 4-hydroxyacetophenone, acetovanillone, acetosyringone, 4-hydroxybenzoate, vanillic acid, syringic acid, and 2-furoic acid all inhibited growth in synthetic medium, but growth inhibition much more pronounced in hydrolysate by all except acetosyringone and vanillic acid. Ethanol yields on xylose in synthetic medium were not as sensitive to the inhibitors as growth except for 4-hydroxybenzaldehyde and vanillin, but in hydrolysate, ethanol yields appeared to be reduced to the same extent as growth inhibition. A similar experiment was performed with steam exploded poplar biomass (Cantarella et al. 2004). No ethanol was produced in a simultaneous saccharification and fermentation reactor with baker’s yeast unless the biomass was washed at least once with water. Adding inhibitors back to washed biomass: HMF, vanillin, and acetic acid (0.5 g/L) increased lag phase for ethanol production. Vanillin also reduced the ethanol yield. Higher concentrations of acetic had larger effect on lag and yields. Formic and levulinic at 1 g/L also had large impact on yields.

Comparative resistance

Several papers have been published which compare the relative resistance of two or more strains to hydrolysates or individual compounds (see for example, Keating et al. 2006; Klinke et al. 2001; Zhang et al. 1995; Brandberg et al. 2004). Evans and Schell (2006) evaluated a number of ethanologens in corn stover hydrolysate in the presence and absence of glucose to 100 g/L (the concentration that would be achieved if all cellulose were converted by enzymatic hydrolysis). In the first set of experiments, they tested four strains for ethanol yields and determined first, that glucose spiking had very little effect on ethanol yields. Further they observed that the two strains capable of utilizing both glucose and xylose (P. stipitis ATCC62970 and Z. mobilis 8b) were much more sensitive to the hydrolysate concentration than were the two glucose fermenting yeasts (S. cerevisiae D5A and the Broin commercial strain of S. cerevisiae). Even when the hydrolysate was overlimed, the performance for Z. mobilis and P. stipitis fell from 79 and 68% ethanol yields in 40% hydrolysate to 23 and 8% ethanol yields in 85% hydrolysate. This is in contrast to D5A and the Broin strain which fell from 84 to 87% in 40% hydrolysate to 82 and 77% in 85% hydrolysate. Even when the hydrolysate was not overlimed but only neutralized with Ca(OH)2, the two S. cerevisiae strains demonstrated high performance with 74 and 79% ethanol yields for D5A and the Broin strain in 85% hydrolysate. Under those conditions, P. stipitis and Z. mobilis yields were only 0 and 2%, respectively. Both xylose and glucose conversion fell in P. stipitis and Z. mobilis with increased hydrolysate concentration, though Z. mobilis was by far the more robust organism. The two S. cerevisiae strains were compared with S. pastorianus (ATCC26602), Kluyveromyces marxianus (ATCC26548, and C. acidothermophilum (ATCC20381). S. pastorianus was as good as or even slightly better than the two S. cerevisiae strains in both neutralized and overlimed 85% hydrolysate, and K. marxianus and C. acidothermophilum also performed well with overlimed hydrolysate, but these strains were alone in demonstrating extreme sensitivity to neutralized hydrolysate spiked with glucose compared to the non-spiked hydrolysate. Finally, when the three best strains were tested in 85% overlimed hydrolysate spiked with 200 g/L glucose, the ethanol yield with S. cerevisiae D5A was lower by more than an order of magnitude compared with the lower glucose conditions, whereas ethanol yields with the Broin strain and S. pastorianus were only reduced slightly (from about 80% to about 70%).

The effects of a wide range of aldehydes and acids on ethanol production in several different fermentation strains were reviewed in Zaldivar and Ingram (1999) and Zaldivar et al. (1999). Data from these publications are summarized below. It is important to note that experimental conditions varied for the different strains tested. Two points worth noting from these tables:
  • Zymonomas mobilis is, in general, much more sensitive to aldehydes and acids than E. coli LY01, P. stipitis and S. cerevisiae, in agreement with the observation that Z. mobilis was significantly inhibited by the presence of high concentrations of corn stover hydrolysate (Evans and Schell 2006).

  • Escherichia coli LY01 is more resistant to acetic acid than S. cerevisiae, but is more sensitive to levulinic acid, syringic and vanillic acid.

It should be noted that the ethanol production data obtained of Z. mobilis in the test by Ranatunga et al., in the table used cultures with very low inocula (OD = 0.2), relative to the other strains listed in Table 3. The inhibition data for this strain, therefore, reflect a combination of inhibition on growth as well as ethanol production. While the other organisms P. stipitis, E. coli and S. cerevisiae listed in Table 3 tested using higher inocula where the inhibition may be manifested more as decreased ethanol production. Understanding the inhibition modes (growth vs. ethanol production, glycolysis vs. membrane integrity vs. pH gradient etc.) by various compounds in the hydrolysate may help to optimize the fermentation process and provide insights for strain improvement. In addition, most of the studies reported so far focused on the inhibition by various compounds on glucose fermentations. Fermentation of xylose in the hydrolysate is much more problematic as compared to glucose. Understanding of the toxic effect of these compounds on xylose fermentation deserves more attention.
Table 3

Inhibition of ethanol production by compounds commonly found in hydrolysate

 

E. coli LY01

S. cerevisiae

Z. mobilis CP4 (pZB5)

P. stipitis

 

Conc. (g/L)

Inhibition (%)k

Conc. (g/L)

Inhibition (%)

Conc. (g/L)

Inhibition (%)

Conc. (g/L)

Inhibition (%)

Furfural

3.5

−38a

4.0

79b

0.95

40c

1.3

9d

5-Hydroxymethylfurfural

8.0

50a

8.0

50e

0.05

16c

4-Hydroxybenzaldehyde

1.0

−38a

1.0

72f

2.0

97a

      

Syringaldehyde

2.5

13a

0.13

33

Vanillin

1.5

−38a

1.0

25f

0.04

38c

0.09

54d

3.0

98a

1.3

50e

    

Acetic acid

15.0

21g

6.0

74h

10.0

95i

30.0

69g

    

10.0

29j

Levulinic acid

20.0

85g

40.0

50e

4-Hydroxybenzoic Acid

2.5

79g

1.0

30f

Syringic acid

5.0

62g

1.0

−17f

0.09

9c

Vanillic acid

4.0

83g

1.0

2f

0.08

3c

  
  

3.7

50e

    

aZaldivar and Ingram (1999)

bPalmqvist et al. (1999)

cRanatunga et al. (1997)

dTran and Chambers (1985)

eClark and Mackie (1984)

fAndo et al. (1986)

gZaldivar et al. (1999)

hPhowchinda et al. (1995)

iFerrari et al. (1992)

jvan Zyl et al. (1991)

kNegative values of inhibition indicate improved levels of ethanol production at the point of measurement

Enhanced resistance

Enhanced resistance refers to improvements in resistance in a single strain brought about by growth selection or through strain engineering efforts. A number of approaches to select for improved hydrolysate resistance through repeated subculture have been described. C. shehatae ATCC22984 was subcultured 72 times, and P. stipitis CBS5776 was subcultured 12 times on SO2 hydrolyzed aspen wood hydrolysate (Parekh et al. 1986). Adapted strains showed improved xylose conversion in synthetic medium and total sugar utilization in aspen hydrolysate. P. stipitis NRRL Y-7124 was shown to grow in synthetic medium supplemented with 20% red oak hydrolysate only after a long lag phase (Nigam 2001). Serial subcultures were carried out over a six week period using hydrolysate concentrations up to 75%. Selected cultures grew and produced much more ethanol than the wild type strain with synthetic hydrolysate containing 5 g/L acetic acid and in medium containing 30% acid hydrolysate. Unlike the parental strain, the adapted strain could produce ethanol in medium containing up to 60% hydrolysate. P. stipitis NRRL Y-7124 was also used in a serial subculture adaptation experiment along with S. cerevisiae NRRL Y-12632 (Liu et al. 2005). Both strains were adapted to grow in the presence of progressively higher concentrations of furfural and HMF by serial subculture. Adapted strains of S. cerevisiae grew in presence of 30 mM HMF with much shorter lag time than the parental strain. They were also able to degrade HMF and produce ethanol from glucose more rapidly than wild type. Similar trends were observed when cells were tested for glucose fermentation in the presence of 60 mM HMF. The adapted strain of P. stipitis was also tested for glucose fermentation in the presence of 60 mM HMF, and found to begin ethanol production with a shorter lag than the adapted strain of S. cerevisiae, though it was not able to metabolize all the HMF even after 150 h. When parental and adapted strains of S. cerevisiae were tested for glucose fermentation in the presence of 30 mM furfural, the parental strain showed very little glucose metabolism and no ethanol production whereas the adapted strain rapidly degraded the furfural (with furfuryl alcohol as product) and began converting glucose to ethanol after a short lag.

The ability of laccase to detoxify hydrolysates (Jonsson et al. 1998), described in more detail in the section on Biological Conditioning, below, was used as a starting point to enhance resistance to hydrolysate toxicity in a rational fashion (Larsson et al. 2001). S. cerevisiae INVSCI was engineered to secrete laccase enzyme from the filamentous fungus Trametes versicolor. Laccase catalyzes one-electron oxidation of phenolics yielding free radicals which tend to polymerize. The recombinant strain could completely eliminate coniferyl aldehyde in synthetic medium. Growth rates and yields as well as ethanol yields were approximately equal for parent and recombinant strains in reference medium with no inhibitor. Addition of coniferyl aldehyde greatly reduced both growth and ethanol yields in the parent, but had no effect on the laccase-producing recombinant. Similarly, the parent was inhibited by dilute acid spruce hydrolysate; growth rates and yields of the recombinant strain were lower in hydrolysate than in synthetic medium but ethanol yields were actually better.

Furan toxicity also led to a number of rational strain improvement strategies. The observation that furfural and HMF can be reduced to the corresponding alcohol which is much less toxic has led a number of investigators to attempt to manipulate furan sensitivity by overexpressing genes responsible for furan reductase activity. Nilsson et al. (2005) showed that S. cerevisiae TMB3000 (a strain resistant to spruce hydrolysate (Brandberg et al. 2004)) was more resistant to furfural in synthetic medium, and was better able to metabolize furfural than a reference strain, CBS8066. In cell lysates, HMF and furfural reduction activity was much higher in TMB3000 than CBS8066. For TMB3000, both furfural and HMF activity were an order of magnitude higher with NADH than NADPH. The same was true for furfural reduction with CBS8066 (though the magnitude of activity was lower), but the cofactor specificity for HMF reduction varies with the growth conditions. Following up on that work, the same group carried out a genome wide transcription analysis of TMB3000 and CBS8066 for known reductase and dehydrogenase genes (18 in all) using cells grown in continuous culture in presence and absence of 0.5 g/L HMF (Petersson et al. 2006). Nearly all genes showed higher expression levels in TMB3000. ADH2 was notable in terms of overall expression level and degree of overexpression from TMB3000 cells grown in the presence of HMF. Crude lysates of recombinant strain overexpressing 15 of 19 of the reductases and dehydrogensases were tested for cofactor preference. Eleven (including ADH2) were approximately equal with both NADH and NADPH. Three (including ADH6) preferred NADPH, and only one (SFA1) preferred NAD. Both ADH6 and SFA1 had high activity against HMF in presence of NADH, but only ADH6 had activity against HMF in presence of NADPH. The rate with NADPH was two orders of magnitude higher than that with NADH. Furfural reduction by ADH6 also showed much higher rates with NADPH than NADH. Overexpression of ADH6 resulted in more rapid HMF reduction (compared to wild type control), but also resulted in increased glycerol and acetate production and slightly reduced biomass yields (Almeida et al. 2008). The gene for a novel E. coli reductase termed furfural reductase was cloned and overexpressed in E. coli LY01 (Gutierrez et al. 2006). Higher furfural reductase activity was detected in the recombinant strain but no effort was made to test for enhanced resistance to furfural or hydrolysate.

Gorsich et al. (2006) employed a systems biology approach to engineer improved resistance to furfural. Using a S. cerevisiae gene disruption library, they screened more than 4,800 separate gene disruption mutants for reduced ability to grow in presence of 50 mM furfural. Eight mutants failed to grow at all and 54 mutants grew with a much longer lag phase and achieved lower cell densities. Of these 62, 11 mutations occurred in primary metabolic genes, eight in DNA replication or RNA processing genes, eleven in protein synthesis or protein processing genes, sixteen in organelle structural genes, three in genes involved in cell division, and eight genes previously uncharacterized. They focused on four genes (zwf1, gnd1, rpe1, and tkl1) encoding glucose-6-phosphate dehydrogenase, 6-phosphogluconate dehydrogenase, d-ribulose 5-phosphate 3-epimerase and transketolase-1, enzymes of the Pentose Phosphate Pathway. Not only were these disruption mutations more sensitive to the presence of furfural, but all demonstrated reduced ability to convert furfural to furfuryl alcohol. Mutations in other genes involved in the PPP (gnd2, tkl2, tal1, and ygr043c) did not have a significant affect on cell growth in the present of furfural. Mutants were constructed which overexpressed zwf1, gnd1, rpe1, and tkl1 genes individually. Of these, only the zwf1 overexpressor demonstrated enhanced resistance to furfural. The adh6 gene has been implicated in furfural reduction (Petersson et al. 2006), but disruption of that gene had little effect on growth in the presence of furfural. Disruption of the adh1 gene resulted in poor growth in the presence and absence of furfural. Disruption of the oar1 gene encoding 3-oxoacyl reductase resulted in enhanced sensitivity to furfural implicating fatty acid metabolism and membrane maintenance in furfural toxicity. No typical stress response gene (encoding heat shock proteins and chaperonins) were listed as playing a role in furfural toxicity.

A transcriptomics approach was taken with S. cerevisiae exposed to HMF in an effort to identify altered gene expression levels (Liu 2006). Significant increases and decreases were observed with a large number of unidentified genes. Some changed and then returned to normal levels and some remained at the altered level. Some genes appeared to respond specifically to the presence of HMF while some changes were related to common stress mechanisms. Some were identified playing a role in pleiotropic drug resistance. No details or gene annotations were provided. The role of stress response proteins in both thermotolerance and ethanol tolerance have been explored in S. cerevisiae (Jeffries and Jin 2000), but there is to date no indication that stress response plays a specific role in tolerance to hydrolysate resistance.

Conditioning

Conditioning refers to methods developed to treat hydrolysate to reduce toxicity and make the sugars more accessible to fermentation. In general, it would be more advantageous in terms of cost savings to modify the pretreatment processes to reduce the formation of toxic compounds rather than add the additional cost of a conditioning step, but as mentioned above, less toxic pretreatment methods typically release less sugar for fermentations and therefore require additional hydrolytic enzymes. There are three categories of conditioning processes: biological, physical, and chemical. The biological and chemical approaches are designed to convert toxic compounds to less toxic products; the physical approach is designed to remove toxic compounds from the hydrolysate. Sequestration of the toxic compounds, followed by recovery will provide some insights into hydrolysate fractionation as well as open up possibilities for generation of biorefinery chemical feedstocks (discussed below).

Biological conditioning

Enzyme treatment is one way to remove toxic compounds for hydrolysate. Willow extract prepared by steam and SO2 treatment was conditioned by treatment with laccase and lignin peroxidase from T. versicolor (Jonsson et al. 1998). This treatment improved the rate of glucose consumption and ethanol production. These enzymes remove monoaromatic phenolics by catalyzing oxidative polymerization. Whole cell systems have also been exploited to condition hydrolysate. A novel isolate of S. cerevisiae YGSDC metabolizes acetate but not sugars (Schneider 1996). This strain could reduce acetic acid from 0.68 to <0.04% in stream stripped hardwood-spent sulfite liquor. After treatment, growth of Pachysolen tannophilus, and P. stipitis increased by approximately an order of magnitude, as did ethanol yields. Isolation of novel soil microorganisms capable of using ferulic acid, furfural and HMF was carried out to identify additional conditioning strains (Lopez et al. 2004). Five bacterial strains and one fungal isolate could utilize at least one of the toxic compounds as sole carbon and energy source, but only the fungus Coniochaeta ligniaria could remove furfural and HMF from stover hydrolysate. Furfural was removed first followed by HMF, but furfural degradation only took place in the presence of HMF. Follow up work evaluated additional fungal strains for the ability to metabolize furans and grow in dilute acid hydrolysate of corn stover (Nichols et al. 2005). Seven of 23 could grow with furfural as sole C source. Subsets of these seven could also grow on HMF, levulinic acid and p-hydroxybenzaldehyde. The best of these, C. ligniaria NRRL30616, removed more than 99% of the furfural, 85% of the HMF, and 20% of the acetate from hydrolysate but also 50% of the glucose and 20% of the xylose. Abatement of corn stover hydrolysate followed by fermentation with Saccharomyces sp. LNH-ST in the presence of cellulases showed a vast improvement over fermentation of unabated hydrolysate.

More recently, a thermophilic bacterium, Ureibacillus thermosphaericus was investigated for the biological detoxification of hydrolysate of waste house wood (WHW) (Okuda et al. 2008). U. thermosphaericus oxidizes furfural and 5-hydroxymethyl furfural to 2-furancarboxylic acid and 5-hydroxymethyl furancarboxylic acid, respectively, which are less toxic to yeast in synthetic hydrolysate than furfural and HMF. When the hydrolysate was treated by this bacterium at 50 °C for 24 h, the ethanol production rate by S. cerevisiae increased markedly and was comparable to that of overlimed. Chromatographic analysis of synthetic hydrolysates containing furfural or 5-hydroxymethyl furfural revealed that U. thermosphaericus degrades these compounds. In real WHW hydrolysates, however, the concentrations of these compounds were not decreased markedly by the bacterium. These results suggest that the bacterium degrades minor but more toxic compounds in the WHW hydrolysates. The combination of bacterial and overliming treatments of hydrolysates minimized significantly the decrease in ethanol production rate by E. coli KO11. Because the bacterium grows rapidly and does not consume sugars, this biological detoxification may be more advantageous than the fungal detoxicification described above.

Chemical conditioning

Many of the pretreatment processes utilize acids to break down the cellulose/hemicellulose/lignin matrix. The pH of the acidic hydrolysates must be raised to the appropriate level for the fermentation organism. It has been known for almost 70 years that adjustment of the hydrolysate to a basic pH, followed by re-acidification to achieve optimal fermentation pH can have significant detoxification effects (Sjolander et al. 1938). When Ca(OH)2 is used in this manner, the method is known as overliming; it is relatively cheap and effective, but it results in the formation of insoluble gypsum when the acid used for pretreatment is H2SO4. This precipitate must be removed by centrifugation before fermentation can be carried out, requiring a costly processing step. It also leads to losses in sugar content and concomitant lost ethanol titres. A great deal of recent work is the evaluation of different alkaline detoxification treatments, testing different bases and process conditions. These will be summarized below.

Four different bases (NaOH, Ca(OH)2, KOH, and NH4OH) were used to treat acid hydrolysate of spruce to either pH10 with a pH reduction back to 5.5 with HCl or H2SO4 or neutralization to pH 5.5 with same bases (Persson et al. 2002a). None of the treatments had much effect on formic, acetic, or levulinic acid or on the concentration of glucose and mannose. The largest changes were observed with ammonia treatment to pH 10, including drops in HMF and furfural, phenol, vanillin, coniferyl aldehyde and cinnamic acid. Neutralization to pH 5.5 with any base had very little effect on the concentration of the toxic compounds measured. The Ca(OH)2 and NH4OH treatments to pH 10 had the biggest improvements on fermentability especially noted in the volumetric productivity at 7 h. These samples were fivefold higher than the untreated hydrolysate and higher even than the reference fermentation. Another series of bases (Ba(OH)2, Ca(OH)2, Mg(OH)2, NaOH, and NH4OH) was tested in subsequent work by this group (Alriksson et al. 2005). All were used to take spruce hydrolysate to pH 10 followed by HCl treatment to pH 5.5. In this case acid levels were found to increase while furans and phenolics declined. Mg(OH)2 caused the largest decrease in furan concentrations, Ca(OH)2 caused the largest effect on phenolics, and NH4OH had an intermediate effect on both classes of compounds. Baker’s yeast fermentations showed that glucose consumption was much higher with NH4OH conditioning than second place Ca(OH)2. This result was accompanied by improved volumetric ethanol productivity and balanced ethanol yields. The ammonia effect did not seem to be due to NH4+ ions because NH4Cl did not improve other hydrolysates.

Attempts to explore higher pH conditioning were carried out with dilute acid hydrolysate of spruce (Nilvebrant et al. 2003). It was shown that at higher pH and temperature, there was increased furan and lignin degradation leading to higher levels of formic, acetic and levulinic acid as well as higher levels of vanillic acid and phenolics. Degradation of furans and lignin was more pronounced with Ca(OH)2 than with NaOH. Similar results were seen with forest residue hydrolysate overlimed to different pH with Ca(OH)2 at various temperatures (Purwadi et al. 2004). High pH and temperatures increased sugar degradation. Baker’s yeast fermentation produced the highest ethanol level with hydrolysate overlimed to pH 10 at 45 °C with a yield of 0.47 of the fermentable sugars, but the rate was slow with this sample and higher rates were supported with hydrolysates treated at higher temperatures and pH. However, as sugar yields went down at pH 11 and 12, the amount of ethanol decreased as well. More detailed evaluation of NH4OH treatment led to optimum conditions of treatment to pH 9 at 55 °C (Alriksson et al. 2006). Under these conditions, furans were reduced by 30%, phenols by 13%, glucose and mannose by 6%, and baker’s yeast fermentation led to balanced ethanol yield of 120% relative to sugar solution control. NaOH treatment was also optimized with best results at 80 °C, pH 9, yielding about the same reductions in furans and phenols but slightly higher sugar losses. Both NH4OH and NaOH generated distinctive performance responses when pH and temperature were varied because of changes in toxic compound removal and sugar losses. When the optimal conditions were used for NH4OH and NaOH conditioning and compared with overliming (at 30 °C, pH 11), all conditioned hydrolysates supported similar levels of ethanol productivity with a range of 90–110% of sugar solution control compared to about 15% for unconditioned hydrolysate.

Zymonomas mobilis 8b was also used to determine the optimum conditions for overliming of dilute acid hydrolysate of corn stover (Mohagheghi et al. 2006). Glucose, arabinose, and xylose sugar losses increase with increased pH with xylose being most sensitive (35% loss of xylose compared to 15% glucose and 20% arabinose at pH 11). Detoxification seems to be highest at high pH because ethanol yield on sugar increases from pH 9 to 11, but overall yields peak at pH 10 because of the increase in sugar losses above that pH. As discussed above, when Z. mobilis 8b was used to ferment overlimed corn stover hydrolysate, ethanol yields fell from 79 to 23% when the hydrolysate concentration was increased from 40 to 85% (Jennings and Schell 2005). When corn stover hydrolysate was conditioned with ammonium hydroxide, ethanol yields were much higher overall, and fell from 90 to 50% when the hydrolysate concentration was raised from 40 to 85% (Dowe et al. 2007). Performance of Z. mobilis 8b in 65% corn stover hydrolysate conditioned with ammonium hydroxide was evaluated in more detail by Jennings and Schell (2006). They found that ethanol production in fermentors was greatly enhanced when ammonia conditioning was employed in place of overliming. Some of that improvement came from enhanced sugar recovery with ammonia conditioning, but the metabolic ethanol yields were higher with ammonia conditioning as well, which indicates that there is a biological effect to ammonia conditioning as well. This may be due to additional nitrogen availability with ammonia conditioning, but it has been shown that increased NH4+ added as NH4Cl had no positive effect on overlimed hydrolysate in baker’s yeast fermentation (Alriksson et al. 2005), and so there may be another explanation. It should also be pointed out that Z. mobilis reached peak ethanol production in half the time with ammonia conditioned hydrolysate than in overlimed hydrolysate (Jennings and Schell 2006). This is another biological performance improvement that relates to reduced toxicity and could have additional impacts on the economics of ethanol production.

Physical conditioning

Physical conditioning involves a fractionation method to physically remove the toxic compounds from hydrolysate rather than carrying out chemical or biological conversions to less toxic products.

Electrodialysis was used to remove acids from mixed wood hydrolysate (Sreenath and Jeffries 2000), but the effect of removal on fermentation performance was not studied in a systematic way. Rather several batches were analyzed and the fermentable sugars (glucose, galactose, mannose, and xylose) ranged from 10–121 g/L; acetic acid from 0.43 to 6.2 g/L; HMF from below limit of detection to 2.2 g/L. Though the different batches supported varied fermentation results with C. shehatae strain PFL-Y-049, it is not possible to draw conclusions regarding the efficacy of electrodialysis because no unconditioned hydrolysate batches were used as controls.

Liquid extractions have been used to remove toxic compounds directly from steam exploded poplar biomass (Cantarella et al. 2004). No ethanol was produced in a simultaneous saccharification and fermentation reactor with baker’s yeast using unwashed pretreated material. A single wash resulted in ethanol production only after a significant lag; a second wash reduced the lag phase. It is also possible to carry out liquid–liquid extractions with hydrolysate to remove inhibitors. An MTBE extract of a yellow poplar acid hydrolysate was shown to inhibit Z. mobilis CP4(pZB5) fermentation when added to a synthetic hydrolysate made up of glucose, xylose, acetic acid and H2SO4 (Ranatunga et al. 2000). The authors did not report whether the extracted hydrolysate supported better fermentation performance.

Supercritical fluid extraction of an acid hydrolysate of spruce removed a number of potentially toxic compounds by varying degree, resulting in improved fermentation yields and productivity with baker’s yeast as fermentation organism (Persson et al. 2002b). Furfural was reduced by 93%, coniferyl aldehyde by 91%, but HMF was only reduced by 10% acetic acid by 19% and levulinic acid by 6%. Even the poorly removed compounds were identified in the extracted material, concentrated by the evaporation of CO2.

Several solid phase extraction methods have been tested for conditioning of hydrolysates. Activated carbon treatment of corn stover dilute acid hydrolysate was shown to reduce acids without a significant reduction in sugar concentration (Berson et al. 2006). Acid hydrolysis of soybean hulls was carried out at various severities; sugar release increased with increased severity of pretreatment maximizing at 125 °C/1.4% H2SO4 (Schirmer-Michel et al. 2008). At that condition, furfural and acetic acid were also high. Treatment of hydrolysate with activated charcoal reduced furfural by 95%, acetic acid by 37%, and phenolics by 75% with minimal reduction of sugars, but had no impact on fermentability by C. guilliermondii NRRL Y-2075; growth was identical in treated and untreated hydrolysate and much lower than that seen in synthetic medium.

Better success was attained with ion exchange resins. Dilute acid cornstover hydrolysate was used with C. mogi ATCC 18364 for production of xylitol from xylose (deMancilha et al. 2003). Cation exchangers removed 40–60% of HMF, 50–80% of furfural, 0% acetic acid, 70–80% of color and less than 8% of xylose. Weak base anion exchanger removed all or nearly all the HMF, furfural, color and acetic acid with only 6% of xylose removed. The strong anion exchangers did not remove any acetic acid but did remove nearly all the furfural and 50–65% of the HMF and 40–70% of the color but less than 4% of the xylose. The weak base anion exchanger and strong cation exchanger were used to condition hydrolysate but performance was not as good as in reference fermentation. Unfortunately, the authors did not test unconditioned hydrolysate, and so it is not possible to determine if detoxification occurred. Nilvebrandt et al. (2001) also tested an anion exchanger (AG1-X8), cation exchanger (AG50 W-X8), and plain resin (XAD-8) to condition dilute acid hydrolysate of spruce for fermentation by baker’s yeast. All resins performed best when used at high loadings. For biomass yield and ethanol productivity the performance of the different resins followed this pattern: AG1-X8 > XAD-8 > AG50-X8. AG1-X8 removed nearly all aliphatic acids as well as more than 60% of the furfural and HMF, 80% of the phenols and 90% of the Hibbert’s ketones. XAD-8 and AG50-X8 removed no acids and very little Hibbert’s ketones but did remove a significant amount of furans and phenols (though less so than the AG1-X8). The AG1-X8 was the only resin to remove glucose from the hydrolysate but it could be recovered in the presence of added sulfate. Following up on that work, additional strong base (Dowex 1X4, Dowex 2X8, IRA458) and weak base (IRA67, IRA92, and Duolite A7) anion exchangers were evaluated (Horvath et al. 2004). All improved fermentation relative to straight hydrolysate, but strong base resin-treated hydrolysates supported better fermentation rates (final yields were all about the same though the fermentation time varied with the rate) than weak base resins, with the Dowex resins best of all (1.9 g/L/h ethanol compared to 0.3 for hydrolysate and 2.2 for reference fermentation. All resins returned nearly 100% glucose and mannose except for slightly lower yields with IRA67. All reduced sulfate to below detectable levels. Dowex 1X4 was best at removing levulinic and acetic and phenolics, and was among the best in removing formic and furfural.

In another example of ion exchange chromatography, a dilute nitric acid hydrolysis of green hybrid poplar was conditioned by anion exchange chromatography with Dowex MWA-1 (weak base resin) (Luo et al. 2002). This treatment resulted in enhanced fermentation by S. cerevisiae D5A. Total ion chromatography run with an ethyl acetate extract of untreated hydrolysate generated more than 70 separate peaks; 15 of these were identified as aromatic compounds, 17 as aliphatic acids, and 3 as furans (furfural was not detected due to losses during vacuum evaporation. The most abundant compound present was p-hydroxybenzoic acid. Analysis of the ion exchange treated hydrolysate demonstrated that all the aromatic acids and much of the aliphatic acids were gone. These included: 2-furancarboxylic acid, 2-furanacetic acid, 5-hydroxymethyl-2-furancarboxylic acid, and ferulic acid, considered to be strong potential inhibitors. Most of the aromatic acids and aliphatic acids were recovered from the regeneration of the anion exchange resin. The compounds removed by ion exchange were recovered and extracted with chloroform. Some compounds (e.g., 2-furancarboxylic acid) remained in the aqueous phase, some (e.g., p-hydroxybenzoic acid) partitioned into both phases and some (e.g., vanillin) were completely extracted into the ethyl acetate phase.

Conclusions

The ability to ferment lignocellulosic hydrolysate sugars is a matter of great importance because it can potentially provide a source of fuel that is not dependent on oil-producing countries, is capable of reducing greenhouse emissions, and does not compete with resources necessary for food production. There are a number of process details that will require further improvements before lignocellulosic ethanol can be produced economically. Several of these relate to the costs of pretreatment and conditioning. These costs include not just process costs such as raw material costs and utility costs related to raising and lowering temperatures but also performance costs such as lower yields due to sugar losses as well as reduced yields and prolonged fermentations times due to toxic effects of hydrolysate. All of these process aspects have been explored to some degree with a goal of improving the overall economics, and significant strides have been made, but further progress is necessary, especially in the development of new methods to measure biological toxicity, new tools to analyze hydrolysates, and new approaches to understand the molecular targets for and metabolic responses to toxic compounds. The information obtained by the application of these new technologies could conceivably feed into a number of new research areas including improvement of pretreatment and conditioning methods, identification of optimum biomass source/fermentation organism combinations, engineering of biomass feedstocks to reduce the toxic compound progenitors, and genetic enhancement of resistance. As this review has demonstrated, these research areas are not new ideas. All of them have been examined and several of them have led to enhanced understanding as well as process improvements. These include:
  • Optimization of pretreatment and conditioning processes to maximize elimination of toxic compounds and minimize sugar losses. Currently dilute acid hydrolysis with ammonium hydroxide conditioning is one of the leading process options. Other conditioning approaches, especially biological and physical (e.g., laccase catalyzed conversion of phenolics and ion exchange sequestration of acids) provide insights into the importance of specific classes of compounds in overall toxicity.

  • Detailed analysis of hydrolysates and identification of scores of potentially toxic compounds including furans, aliphatic acids, phenolics, aliphatic and aromatic acids, aldehydes and ketones. Chromatographic methods are being developed to distinguish dozens of different components in hydrolysates and new detection methods (especially MALDI-TOF mass spectrometry) are being developed to expand the ability to identify additional compounds that may not be available as standards.

  • Toxicity analysis is becoming increasingly detailed, moving from fermentor and shake flask measurement of inhibition of growth and productivity to identification of metabolic targets and mode of action. Toxicity analysis is also becoming increasingly complex as researchers explore inhibitor synergies, moving from single compounds to mixtures in an effort to model the complex mix of compounds found in hydrolysate. Systems biology approaches (i.e., transcriptomics) are being employed to analyze the response of fermentation organisms to challenge by toxic compounds found in hydrolysate.

  • Evaluation of candidate fermentation strains for their ability to perform well in the presence of toxic compounds or hydrolysate has helped determine which strains are most robust and has provided benchmarks for further improvements either in pretreatment and conditioning or in development of higher levels of resistance. Strain improvement efforts include, selection of more robust strains through continuous culture; identification of specific detoxification enzymes and engineering of overproducing strains; overexpression of single genes in whole genomic libraries; and manipulation of specific genes whose expression levels change when the strain is exposed to toxic compounds. Some of these efforts have led to incremental improvements in strain robustness and have provided additional insights into the mode of action of toxic compounds as well as the relative importance of specific compounds in hydrolysates.

  • Analytical methods employing solid phase extraction and HPLC chromatography as well as physical conditioning methods can provide a basis for fractionating of toxic compounds present in hydrolysates. In some cases these methods also provide a way to concentrate samples to allow for more sensitivity in measuring a toxic response.

  • The sum of effects of all toxic compounds in hydrolysate is almost certain to be more than the sum of the parts. Synergies have already been detected in simple combinations, and the ability to test for toxic effects on a high throughput manner will allow for the identification of more complex combinations of individual compounds or fractions. The existence of these synergies implies that alteration of pretreatment and conditioning steps to eliminate a single member of a synergistic combination could have a greater impact than elimination of compounds acting alone. It also helps explain why enhanced resistance to furfural alone, can improve fermentation in hydrolysate.

Acknowledgments

The work was funded by the DOE Office of the Biomass Program.

Copyright information

© US Government 2009