Use of UV Absorbance To Monitor Furans in Dilute AcidHydrolysates of Biomass

Alfredo Martinez,†,‡ Maria E. Rodriguez,†,‡ Sean W. York,† James F. Preston,† andLonnie O. Ingram*,†

Institute of Food and Agricultural Sciences, Department of Microbiology and Cell Science, University of Florida,Gainesville, Florida 32611, and Instituto de Biotecnologıa, Universidad Nacional Autonoma de Mexico,Cuernavaca, Mor. 62250, Mexico

A simple method based on UV spectra was developed for the estimation of total furans(furfural and hydroxymethylfurfural) in hemicellulose hydrolysates. UV spectra ofhemicellulose hydrolysate contained a single dominant peak at around 278 nm.Approximately two-thirds of this peak can be attributed to furan absorbance (furfuraland hydroxymethylfurfural). At 284 nm, both furfural and hydroxymethylfurfural haveequal absorbance on a weight basis. A comparison of HPLC determinations for differentsamples of hydrolysate was used to develop a simple equation that allows the accurateprediction of total furans based on the difference in absorbance at 284 and 320 nm.This method may prove useful for quality control applications during the productionof biomass syrups using a dilute acid hydrolysis process and during treatments forthe amelioration of toxins. Although furans represent only a portion of the toxinspresent in hemicellulose hydrolysates, the abundance of furans appears to serve as auseful marker to predict relative toxicity.

Introduction

Lignocellulose represents an underdeveloped resourcefor the production of sugar feedstocks that can beconverted to a wide variety of renewable fuels andchemicals by microbial biocatalysts (1, 2). Approximately20-50% of the dry weight of plant cell walls is hemicel-lulose, a carbohydrate that can be readily hydrolyzed tomonomeric sugars with dilute mineral acids at modesttemperatures (3). During acid hydrolysis, however, acomplex mixture of microbial toxins that retard biocon-version is generated. These toxins can be mitigated bytreatments with lime, anion-exchange resins, enzymes,etc. (4-8). Two compounds, furfural from the dehydrationof pentose sugars and hydroxymethylfurfural (HMF) fromthe dehydration of hexose sugars, are typically producedat significant concentrations (grams per liter dependingon hydrolysis conditions) and have been proposed asbeing of particular importance for toxicity in yeasts (9-11). A recent study using multisensor array technologyhas identified furfural as a predictor of toxicity for yeast-based ethanol fermentations (12). Additional studies inour laboratory have shown that furans in hemicellulosehydrolysate are also toxic to ethanologenic Escherichiacoli LY01 (13). In tests of binary mixtures with other toxiccompounds known to be present in hydrolysates, furfuralalone was found to increase (potentiate) toxicity for bothSaccharomyces cerevisiae (9-11) and E. coli LY01 (13).

Furfural and HMF can be readily measured by chro-matographic methods (14) or by electronic sensors (12).Both compounds are also known to have distinctive UVspectra (15). In this paper, we have investigated the use

of UV spectra and HPLC methods to monitor furancontent in hemicellulose hydrolysates of sugar canebagasse before and after the mitigation of toxins bytreatment with lime and to predict the toxicity of hy-drolysates for fermentation using E. coli LY01.

Materials and Methods

Microorganisms and Media. Ethanologenic E. colistrain LY01 (16) was used in all experiments. Thisorganism can produce ethanol efficiently from all hexoseand pentose sugars in hemicellulose hydrolysates (17).Stock cultures were grown in Luria broth containing (perliter) 5 g of yeast extract, 10 g of tryptone, 5 g of NaCl,and 50-100 g of xylose. For solid media, xylose wasreduced to 20 g/L, and chloramphenicol (600 mg/L) andagar (15 g/L) were added. Stock cultures were maintainedon plates and incubated at 30 °C under argon.

HPLC Analyses. Sugar composition was analyzed byHPLC using a Waters HPLC system (Milford, MA) andrefractive index monitor. Separations were performed ona Bio-Rad (Hercules, CA) Aminex HPX-87P column (300mm × 7.8 mm i.d.) at 85 °C using pure water as themobile phase (0.5 mL/min). Furfural, HMF, phenoliccompounds, and ferulic acid were analyzed using anAminex HPX-87H column (55 °C) and eluted (0.35 mL/min) with a mixture of 0.01 N sulfuric acid (84%) andacetonitrile (16%) (14). Peaks were quantified by usinga variable wavelength UV detector and electronic inte-grator.

Spectroscopy. UV spectra were measured with aBeckman DU 640 spectrophotometer (Fullerton, CA).Samples of hydrolysates were diluted 1:1000 in distilledwater prior to spectral analysis.

Materials. Fourteen different samples of bagassehemicellulose hydrolysates were provided by a com-

* Ph: (352)392-8176.FAX: (352)392-5922.E-mail: ingram@ufl.edu.† University of Florida.‡ Universidad Nacional Autonoma de Mexico.

637Biotechnol. Prog. 2000, 16, 637−641

10.1021/bp0000508 CCC: $19.00 © 2000 American Chemical Society and American Institute of Chemical EngineersPublished on Web 07/06/2000

mercial supplier. These were accumulated over a 3-monthperiod using a high solids reactor containing dilutesulfuric acid as previously described (3, 18, 19). Detailedinformation regarding processing conditions was notprovided. Furfural, HMF, and other chemicals wereobtained from Fisher Scientific (Norcross, GA).

Fermentation. Inhibition of ethanol production wasused as a measure of toxicity. Samples of treated anduntreated hydrolysate were adjusted to pH 6.7 with Ca-(OH)2 or H2SO4 as necessary but were not sterilized. DryLB nutrients (tryptone, 10 g/L; yeast extract, 5 g/L) wereadded (without sterilization) to minimize the dilution oftoxins. Inocula were grown for 16 h in 2-L flasks (720mL LB broth containing 5% xylose) at 35 °C withgyrotary agitation (120 rpm). Fermentations were con-ducted in 250-mL flasks containing 100 mL of broth,incubated for 48 h at 35 °C, and inoculated to providean intial cell density of approximately 0.16 g cell dryweight/L. Ethanol was measured after 24 h using gaschromatography (18).

Treatment of Hydrolysate with Ca(OH)2. Hydroly-sates (3 kg) were warmed to 60 °C and mixed vigorouslyduring the rapid addition of 50-100 g of anhydrous Ca-(OH)2. After 30 min of incubation, hydrolysates werecooled to ambient temperature in a second water bathand used immediately for analyses or in fermentations.

Results and Discussion

Identification of the Prominent UV Peak inHemicellulose Hydrolysate as Furans. The complex-ity of hemicellulose hydrolysates may ultimately preventa complete analysis (12). However, many of the ligninand sugar degradation products absorb in the UV region.Preliminary experiments revealed that acceptable UVspectra of hemicellulose hydrolysate could be obtainedafter a 1000-fold dilution into distilled water, as il-lustrated in Figure 1A. Spectra for all 14 samples werefound to be quite similar in shape. All contained a singleprominent peak at approximately 278 nm, shoulders at320 and 220 nm, and end absorbance continuing off-scaleat around 200 nm. For comparison, Figure 1B showsspectra for four hydrolysate components that absorb inthese regions.

The identity of components that contribute to the 278nm peak in diluted hydrolysate was further investigatedby using HPLC with a variable wavelength detector (278nm). On the basis of elution profiles, approximately two-thirds of the absorbance at 278 nm can be attributed totwo compounds, furfural and HMF (Figure 2A). Theremaining absorbance represents the sum of numeroussmall unidentified peaks and compounds that elute inthe region of hydroxybenzaldehyde and ferulic acid. Onthe basis of the shape of UV spectra for hydrolysates, it

Figure 1. UV spectra of hemicellulose hydrolysate and selected components. (A) Hemicellulose hydrolysates after dilution 1:1000in distilled water. (B) Selected compounds found in hemicellulose hydrolysate (10 mg/L). Abbreviations: HMF, hydroxymethylfurfural;HBH, hydroxybenzaldehyde.

Figure 2. HPLC elution profiles of hemicellulose hydrolysate monitored at 278 nm. (A) Untreated hydrolysate. (B) Hydrolysatetreated with Ca(OH)2 to mitigate toxicity. Approximately 26 g of Ca(OH)2 was added per kg of hydrolysate. which was mixed andheld at 60 °C for 30 min. Abbreviations: HMF, hydroxymethylfurfural; Fur, furfural; HBH, hydroxybenzaldehyde; Fer, ferulic acid.

638 Biotechnol. Prog., 2000, Vol. 16, No. 4

is likely that ferulic acid or related compounds areresponsible for the shoulders at 220 and 320 nm.

Development of a Method for the Estimation ofFurans in Hemicellulose Hydrolysates Using UVSpectra. The prominence of furans in HPLC elutionprofiles at 278 nm and in UV spectra suggested that itmay be possible to develop a rapid spectral method forthe estimation of furans in hemicellulose hydrolysates.Figure 3A shows a family of curves for various concentra-tions of pure furfural and hydroxymethylfurfural. Notethat on a weight basis, the absorbance of both of thesecompounds is equivalent at 284 nm. Figure 3B showsspectra for equal mixtures of these two compounds. Asshown by Figure 3C, the absorbance at 284 nm can beused to accurately estimate the concentration of total

furans in distilled water (furfural, HMF, or a mixture ofboth) with a slope of 0.127 ( 0.001 absorbance units(intercept ) 0.007 ( 0.003; r2 ) 1.0) per ppm furan (mg/L).

Hemicellulose hydrolysates contain many compoundsin addition to furans that absorb in the 284 nm regionand complicate spectral estimates of furan content.Hydrolysates contain a shoulder at around 320 nm, andpure furans have low absorbance in this region. Subtrac-tion of this shoulder was investigated as a correction forunknowns contributing to absorbance at 284 nm. Withpure compounds, spectra generated by subtracting ab-sorbance values at 320 nm did not affect the wavelength

Figure 3. UV spectra of furfural and HMF. (A) Spectra ofindividual compounds at different concentrations (0.05-10.0 mg/L). (B) Spectra of an equal mixture of furfural and HMF atdifferent concentrations (0.05-10.0 mg/L). (C) Standard curverelating A284 and spectral difference (A284 - A320) to furanconcentration. All define a single line (slope ) 0.127).

Figure 4. Development of spectral method to estimate totalfurans in hydrolysate. (A) Relationship between ∆absorbance (A284- A320) and furan content as measured by HPLC. (B) Compari-son of total furans in hydrolysates estimated by spectral method(eq 1) to HPLC determination using the equation defined bythe data in Figure 4A. (C) Comparison of spectral method (eq1) and HPLC estimates of furan content in Ca(OH)2-treatedhydrolysates.

Biotechnol. Prog., 2000, Vol. 16, No. 4 639

(284 nm) at which both compounds have equal absor-bance or the regression line defined by plotting thisdifference versus furan concentration (Figure 3C).

Subtraction of absorbance values at 320 nm was testedas a baseline correction for hydrolysates by plotting thisdifference in absorbance (A284 - A320) as a function offuran content determined by HPLC (Figure 4A). Theresulting data defined a straight line (r2 ) 0.860) with aslope of 0.127 ( 0.01 absorbance units per ppm furan(mg/L), similar to that of the pure compounds (Figure3C). However, this curve did not extrapolate through zero(y intercept ) 0.056). Thus subtracting the absorbanceat 320 nm provided only a partial baseline correction forfurans in hydrolysate, and further correction was needed.A variety of approaches were empirically tested to providea further correction but none were better (r2 e 0.86) thanthe simple equation defined by the data in Figure 4A:

Equation 1 was used to estimate furan content in eachof the 14 hydrolysate samples. The resulting values wereplotted as a function of HPLC determinations (Figure 4B)with excellent agreement (intercept ) 0.000; slope )1.000; r2 ) 0.86).

Quantitation of Furans in Lime-Treated Hemi-cellulose Hydrolysates. Treatment of hydrolysateswith lime is known to decrease the toxicity of hemicel-lulose hydrolysates (5, 20). This treatment caused adecrease in furans that was readily apparent both in UVspectra (Figure 5A) and in HPLC elution profiles (Figure2B). Furans are unstable at high pH after lime treatmentand continue to decline during storage (data not shown).Supplementation of treated hydrolysate with pure HMF

and furfural at levels equivalent to the losses measuredby HPLC restored the main spectral peak in the 278 nmregion (Figure 5A), confirming the quantitative relation-ship between absorbance and furan content.

Treatment of hydrolysate with a series of lime concen-trations resulted in a dose-dependent decline in absor-bance primarily at 278 nm and a decline in furan contentby HPLC. To determine if the equation relating spectrato furans in untreated hydrolysate was also valid forlime-treated hydrolysates, spectral estimates of totalfurans after lime treatments were plotted as a functionof HPLC determinations (Figure 4C). These data defineda straight line that extrapolated near zero (0.034 ( 0.037)with a slope of 1.031 ( 0.037 (r2 ) 0.99), indicatingexcellent agreement between the two methods.

Furan Content as a Predictor of Toxicity inHemicellulose Hydrolysates. Although furans repre-sent only one group of toxins present in hemicellulosehydrolysates, these compounds can be rapidly measuredand may serve as a useful marker for toxicity. Neutral-ized hydrolysates were supplemented with LB nutrientsand variously diluted with LB containing xylose (90 g/L)prior to inoculation with E. coli LY01. The decrease inethanol production was used as a relative measure ofbiological toxicity. Less than 1 g ethanol/L was producedin all undiluted hydrolysates (Figure 6A). Dilutions of2- to 5-fold were sufficient to permit fermentation. Withrespect to toxicity, hydrolysates formed two groups. Threehydrolysates were notably less toxic than the remaining11 hydrolysates.

Figure 5. Use of UV spectra to evaluate destruction of furans.(A) Treatment with Ca(OH)2 (lime). (B) Effect of fermentation. Figure 6. Fermentation of hemicellulose hydrolysates by E.

coli LY01. (A) Ethanol production in diluted hydrolysates.Hydrolysates were adjusted to pH 6.7 with Ca(OH)2 and dilutedwith LB containing 9% xylose. (B) Ethanol production replottedas a function of furan content after dilution.

A284 - A320 ) 0.127 ‚ total furan in mg/L + 0.056 (1)

640 Biotechnol. Prog., 2000, Vol. 16, No. 4

Plotting all data as a function of furan content in thediluted hydrolysates (Figure 6B) defined two sets ofcurves. For 11 hydrolysates, maximal ethanol productioncorrelated with a furan content below 0.2 g/L. For twohydrolysates, however, maximal ethanol production cor-related with furan concentrations below 0.5 g/L. Previousstudies have shown that HMF is less toxic than furfuralfor LY01 (13). In both hydrolysates with low toxicity, theratio of HMF to furfural (0.8 and 1.1) is much higher thanin the remaining 11 hydrolysates (0.15-0.20). Althoughfurans are not the only toxins present in hemicellulosehydrolysates, the level of these compounds appears tocorrelate well with toxicity provided there are not largedifferences in the proportions of HMF and furfural. It isclear that toxicity is caused by other compounds inaddition to furans. When tested alone, HMF and furfuralconcentrations of over 3 g/L were required to inhibitethanol production by ethanologenic E. coli under similarconditions (13).

UV spectra can also be used to monitor the biologicaldegradation of furans (Figure 5B). Previous studies havereported that furfural can be metabolized by E. coli andother enteric bacteria (21). Lime-treated hydrolysate wassupplemented with furfural and HMF to restore theoriginal furan concentrations, amended with nutrients,and inoculated. Within 24 h, the 278 nm peak attributedto furans was eliminated, indicating that both furans hadbeen metabolized by ethanologenic E. coli.

Conclusions

UV spectra can be used as a rapid and convenientmethod to estimate total furans (HMF and furfural) inhemicellulose hydrolysates and to monitor furans infermentation broth. Such measurements may prove use-ful as a quality control method for the production ofhydrolysate, to evaluate the amelioration of toxins bylime treatments, and as a convenient parameter tomonitor fermentation. Under the conditions examined,the accuracy of total furan determinations in bagassehemicellulose hydrolysate based on UV spectra appearsequivalent to those from HPLC analyses for bagassehydrolysate. It is likely that the same or similar equa-tions can be developed for hydrolysates from otherbiomass sources, although reliability must be verified ineach case. Variable changes in the amounts of otherabsorbing compounds will directly impact the accuracyof this method. The spectral method should be regardedas a rapid and efficient tool for estimation and not as aninfallible quantification. Although furans represent onlya fraction of the toxins present in hydrolysate, monitoringthe production and destruction of these compoundsappears to be a useful indicator of biological toxicity forethanologenic E. coli and S. cerevisiae (12).

Acknowledgment

This study was supported by the Florida AgriculturalExperiment Station (publication R-07250), the U.S. De-partment of Agriculture National Research Initiative (98-35504-6177 and 98-35505-6976), and the U.S. Depart-ment of Energy DE-FG02-96ER20222 and XXL-9-29034-01. A.M. and M.E.R. were partially supported by DGAPA-UNAM-Mexico.

References and Notes

(1) Dale, B. E. Biobased industrial products: Bioprocess engi-neering when cost really counts. Biotechnol. Prog. 1999, 15,775-776.

(2) Lynd, L. R.; Wyman, C. E.; Gerngross, T. U. Biocommodityengineering. Biotechnol. Prog. 1999, 15, 777-793.

(3) Grohmann, K.; Torget, R.; Himmel, M. Optimization ofdilute acid pretreatment of biomass. Biotechnol. Bioeng. 1985,15, 59-80.

(4) Frazer, F. R.; McCaskey, T. A. Wood hydrolyzate treatmentsfor improved fermentation of wood sugars to 2,3-butanediol.Biomass 1989, 18, 31-42.

(5) Larsson, S.; Reimann, A.; Nilvebrant, N. O.; Jonsson, L. J.Comparison of different methods for the detoxification oflignocellulose hydrolysates of spruce. Appl. Biochem. Bio-technol. 1999, 77-79, 91-103.

(6) Leonard, R. H.; Hajny, G. J. Fermentation of wood sugarsto ethyl alcohol. Indust. Eng. Chem. 1945, 37, 390-395.

(7) Perlman, D. Production of 2,3-butylene glycol from woodhydrolysates. Ind. Eng. Chem. 1944, 36, 803-804.

(8) Sjolander, N. O.; Langlykke, A. F.; Peterson, W. H. Butylalcohol fermentation of wood sugar. Ind. Eng. Chem. 1938,30, 1251-1255.

(9) Boyer, L. J.; Vega, K.; Klasson, K. T.; Clausen, E. C.; Gaddy,J. L. The effects of furfural on ethanol production by Sac-charomyces cerevisiae. Biomass Bioeng. 1992, 3, 41-48.

(10) Palmqvist, E.; Almeida, J. S.; Hahn-Hagerdal, B. Influenceof furfural on anaerobic glycolytic kinetics of Saccharomycescerevisiae in batch culture. Biotechnol. Bioeng. 1999, 62, 447-454.

(11) Palmqvist, E.; Grage, H.; Meinander, N. Q.; Hahn-Hager-dal, B. Main and interaction effects of acetic acid, furfural,and p-hydroxybenzoic acid on growth and ethanol productiv-ity of yeasts. Biotechnol. Bioeng. 1999, 63, 46-55.

(12) Mandenius, C.-F.; Liden, H.; Eklov, T.; Taherzadeh, M. J.;Liden, G. Predicting fermentability of wood hydrolysates withresponses from electronic noses. Biotechnol. Prog. 1999, 15,617-621.

(13) Zaldivar, J.; Martinez, A.; Ingram, L. O. Effect of selectedaldehydes on the growth and fermentation of ethanologenicEscherichia coli. Biotechnol. Bioeng. 1999, 65, 24-44.

(14) Yuan, J. P.; Chen, F. Simultaneous separation and deter-mination of sugars, ascorbic acid and furanic compounds byHPLC-dual column detection. Food Chem. 1999, 64, 423-427.

(15) Taher, A. M.; Cates, D. M. A spectrophotometric investiga-tion of the yellow color that accompanies the formation offuran derivatives in degraded-sugar solutions. Carbohydr.Res. 1974, 34, 249-291.

(16) Yomano, L. P.; York, S. W.; Ingram, L. O. Isolation andcharacterization of ethanol tolerant mutants of Escherichiacoli KO11 for fuel ethanol production. J. Ind. Microbiol.Biotechnol. 1998, 20, 132-138.

(17) Ingram, L. O.; Aldrich, H. C.; Borges, A. C. C.; Causey, T.B.; Martinez, A.; Morales, F.; Saleh, A.; Underwood, S. A.;Yomano, L. P.; York, S. W.; Zaldivar, J.; Zhou, S. Entericbacterial catalysts for fuel ethanol production. Biotechnol.Prog. 1999, 15, 855-866.

(18) Asghari, A.; Bothast, R. J.; Doran, J. B.; Ingram, L. O.Ethanol production from hemicellulose hydrolysates of agri-cultural residues using genetically engineered Escherichiacoli strain KO11. J. Ind. Microbiol. 1996, 16, 42-47.

(19) Beall, D. S.; Ingram, L. O.; Ben-Bassat, A.; Doran, J. B.;Fowler, D. E.; Hall, R. G.; Wood, B. E. Conversion ofhydrolysates of corn cobs and hulls into ethanol by recombi-nant Escherichia coli B containing integrated genes forethanol production. Biotechnol. Lett. 1992, 14, 857-862.

(20) McMillan, J. D. Conversion of hemicellulose hydrolysatesto ethanol. ACS Symp. Ser. 1994, 566, 411-437.

(21) Wang, P.; Brenchley, J. E.; Humphrey, A. E. Screeningmicroorganisms for utilization of furfural and possible inter-mediates in its degradative pathway. Biotechnol. Lett. 1994,16, 977-982.

Accepted for publication May 17, 2000.

BP0000508

Biotechnol. Prog., 2000, Vol. 16, No. 4 641