of Whey Permeate with Kluyveromyces fragilis I: Primary Metabolism

Transcription

1 Production of an Alcoholic Beverage by Fermentation of Whey Permeate: I, pp Volume 106, No. 6, 2000 Production of an Alcoholic Beverage by Fermentation of Whey Permeate with Kluyveromyces fragilis I: Primary Metabolism By Javier Parrondo, Luis A. Garcia and Mario Diaz* Department of Chemical Engineering (IUBA), University ofoviedo, Oviedo (Asturias), Spain Received 25 October 1999 Wltey permeate from the ultrafiltration processing of cheese manufacturing zvas used for alcoholic fermentation with Kluyveromyces fragilis, the objective being the production of an alcoholic beverage of low alcoholic grade. The effect of temperature, initial ph, agitation and initial biomass on yeast groivth and ethatwl production were assayed in batch cultures. In addition, continuous culture behaviour was studied due to interest in a continuous industrial process for a nezv product. An unstructured kinetic model (a Riccati kinetic equation) is proposed; batch data being employed to obtain the kinetic parameters. This model fits the continuous culture results well except for ethanol production, where calculated values were lower than experimental data. When the fermentation temperature ivas changed from 18 to 37 C, a maximum ofypls close to 30 C was observed, ivhich gives an efficiency in the conversion of lactose to ethanol of 88%. The initial ph of the zvhey did not affect yeast groivth significantly. Experiments carried out at different initial biomass concentrations showed that an initial dry weight close to 0.5 g/litre was sufficient to carry out the fermentation. An increase in ethanol concentrations ivas found at higher rates of agitation. In continuous culture, a maximum productivity for biomass and ethanol was attained at a dilution rate of 0.11 h-\ Higher efficiencies (96%) were achieved in continuous culture rather than in batch cultivation mode. Key Words: Kluyveromyces fragilis, whey permeate, ethanol, continuous culture, batch culture, kinetic model. INTRODUCTION USA from 1992 to 1993 was around 69.1 million tonnes22, of which nearly 30 million tonnes was used Cheese whey, a waste stream of significant volume in c,,,, c.' t,..., } for the manufacture of cheese. Thus, around 27 million the dairy industry, has for some time been a subiect of,,..,,,, _,. 1,., tonnes of liquid whey are produced. This amount environmental concern due to the laree fraction of,... '., _..,,.,_,...., ;,.. contains 1.3 million tonnes or lactose. The development lactose present (65% w/w total solids). In 1993, the,.,,,,,,...,. c n - of new uses for lactose is therefore or great interest, European Union produced 114 million tonnes of milk',, and its dairies turned out approximately 26.4 million and has been for cluitc some lime novv- ln the Past' the tonnes of processed milk (pasteurised, UHT, etc.), 5.5 bulk of lhc who>' P«*»uced was simply dumped into million tonnes of cheese, 7.5 million tonnes of fresh rivers. This caused serious problems because of the high produce (yoghurt, etc.), 2.1 million tonnes of milk Biological Oxygen Demand (BOD). The industry has powder and 1.7 million tonnes of butter. Led by developed many uses for whey-lactose and for whey consumer demand, dairies are increasingly turning to itself'» order to resolve the disposal problem. However, the production of cheese (whey is mainly produced in due to the increasing levels of production of cheese the manufacture of cheese) and fresh products like and whey-protein concentrate, the problem remains yoghurts and dairy desserts. More than 80% of cheese unsolved as yet as solutions are required for high production comes from French, German, Italian, Dutch volumes. Large amounts of whey and whey and Danish dairies. The 5.5 million tonnes of cheese permeate need to find a market; estimations made in produced generates around 50 million tonnes of liquid 1992 place the surplus in the USA at 13 to 17 million whey as a by-product. The yearly milk production in the tonnes of whey2. Journal of The Institute of Brewing 367

2 Volume W6, No. 6, 2000 Production of an Alcoholic Beverage by Fermentation of Whey Permeate Currently, the major uses of whey and whey permeate are the manufacture of dried whey powder and refined lactose22. These uses, however, are often aimed at keeping the surplus out of the sewers, rather than at producing a desirable product. The ultimate goal should be to turn whey lactose into a profit-generating feedstock for high value-added products. Examples include chemical or enzymic lactose hydrolysis and the production of chemical derivatives22 such as lactitol, lactulose, lactosyl urea, lactobionic acid and gluconic acid. Whey, whey permeate, the by-product of enhanced cheese manufacture, and lactose can also be used directly as substrates for microbial growth to obtain more valuable products such as single cell protein1422 (SCP), enzymes22, yeast extract16, ethanol22, glycerol18, organic acids22 and oils22. Among these uses, interest in ethanol production19-21 has grown in recent years. Whey-permeate production is increasing because of new manufacturing processes, but major applications have been developed for whey and not its permeate; so new solutions are required for the latter. Khn/veromycesfragilis has been employed in numerous research projects employing Jerusalem Artichoke Juice", cheese whey8 and synthetic broth17, because of its ability to grow in lactose-based media - not many species can metabolise this sugar3 - with high yields and without producing any toxins. It has been authorised by the Food and Drug Administration, Washington, D.C., for use in food products. The aims of the present study were to evaluate the production of an alcoholic beverage from whey permeate; a broth similar to whey in lactose composition but with half its protein concentration or less. Different fermentation temperatures, initial ph's, inoculum levels and rates of agitation were assayed. In addition, continuous experiments were carried out to look into the possibilities of a continuous production process. MATERIALS AND METHODS Organisms, media and culture conditions Kluyveromyces fragilis (Spanish type culture collection, CECT 1123) was the lactose fermenting yeast employed in this work. Cultures of this yeast were kept for less than three months on TYED (triptone 2, yeast extract 1, D-glucose 2 g per litre) plus 2% w/v agar slopes. When the preparation of new slopes was needed, cells were grown from a culture stored in glycerol at -20 C. Inoculum for the experimental fermentations were prepared in two steps. A colony was transferred from a slope to a small amount of synthetic media, previously sterilised (10 ml TYED media in a 50 ml flask) and propagated aerobically for around 24 hours in a New Brunswick Incubator operating at 30 C and 200 r.p.m. This culture was transferred to a fresh medium (100 ml TYED in a 500 ml flask) and propagated during another 24 hours under the same conditions. This culture was used as inoculum in a ratio of 10 volumes per 100 volumes of fresh medium (this gives rise to an initial dry weight of approximately 0.5 g per litre). The medium employed in this study was liquid, sweet whey provided by ARIAS, Vegalencia, Asturias, Spain. This whey is the permeate of an ultrafiltration process used for the recovery of proteins during cheese production. The composition (provided by the dairy) is (% w/v): lactose (approximately) 4.5, whole proteins 0.45, whole dry extract and traces of fats. This medium was sterilised prior to fermentation. In order to develop batch fermentation, sterilisation was carried out employing a cross-flow microfiltration technique; but when continuous culture experiments were conducted, the whey was autoclaved (the ph being previously adjusted to 7.3 to prevent the precipitation of most of the proteins). Cross-flow microfiltration was the initial method of sterilisation chosen. This method was preferred in all experiments, but the culture became contaminated in the continuous assays within three or four days and so the whey permeate for these experiments was autoclaved. The concentration of minerals and lactose did not change when autoclaving. Proteins that precipitate about 50% when autoclaving remain in excess due to the 400% initial excess. Taking these two points into consideration similar behaviour was expected in both broths. A clear medium was needed to allow the measurement of cell dry weight during the course of the fermentation, and so the autoclaved whey was microfiltered through a 0.45 urn WHATMAN filter prior to use. Fermentations Erlenmeyer flasks were employed for cell culturing in batch fermentations. During inoculum propagation, aerobic conditions were maintained by employing a high agitation speed (200 r.p.m.) and low volumes of medium in the flask (100 ml in 500 ml flask). When anaerobic conditions were needed, the agitation was maintained at 100 r.p.m. and higher volumes were employed in the flask (200 ml whey in 500 ml flask) to prevent aeration of the medium. The Erlenmeyer flasks were closed only with cotton bungs. The condition described as 'anaerobic' is therefore not strictly anaerobic although no aerobic biotransformation (ethanol oxidation) was observed. The propagation conditions denoted as 'aerobic' are also not fully aerobic. Diauxie was observed and ethanol was produced initially, being metabolised in the latter part of the process. This yeast is Crabtree negative (aerobic respiring), so the production of ethanol means mixed conditions, between fully aerobic and anaerobic. In all experiments, the ph was 368 Journal of The Institute of Brewing

3 Production of an Alcoholic Beverage by Fcrwi'iilation of Winy Permeate: I Volume 106, No. 6, 2000 left to evolve freely from the initial value (approximately 6.6 for whey without any treatment). For continuous culture experiments the bioreactor, with a working volume of 270 ml, was equipped with a jacket that allowed the whey to be fermented at a controlled temperature (30 C) and a level control outlet for continuous operation. All the equipment which was to be in contact with the feed was autoclaved prior to use. To avoid oscillations reaching the steady state, the chemostat was loaded during batch mode with a medium of half the composition of the feed medium during continuous operation. The chemostat was loaded with 160 ml of sterilised whey, 130 ml of sterilised distilled water and 33 ml of inoculum (initial dry weight approximately 0.5 g/litre). The culture was allowed to evolve in batch mode for 6 hours, reaching the exponential phase of growth. The outlet was then opened to discharge the surplus volume contained in the vessel (20%), and the peristaltic pump was switched on to start the experiment. When the steady state was reached (at least three residence times), the pump was changed to another position (another dilution rate) and the chemostat was fed until reaching another steady state. The change must be gradual to prevent the washout and/or oscillations reaching the new steady state. Analytical methods Cell growth was determined by measurement of dry weight. 1 ml samples (triplicates) were centrifuged in pre-weighted Eppendorf tubes, washed with distilled water and placed in an oven at 90 C during 24 hours. Lactose was determined using a Waters Alliance HPLC System with differential refractometer detection19 (Waters 410). The chromatographic system control, chromatogram acquisition and processing were performed via computer by the MILLENNIUM v program. The column employed was a SPHERISORB- NH, (Tecknokroma) of 20 cm length, 0.4 cm internal diameter and 5 jim particle size. The eluanl was 0.9 ml/ min acetonitrile/water 11 ITS. The temperatures of internal cells (measurement and reference) and of the column oven were fixed at 45 and 30 C. Proteins in the sample may cause interference in chromatographic detection and so were precipitated from the samples with TCA (10% during 30 min in an ice bath). An internal standard (glucose) was employed to quantify the samples. Ethanol was analysed by gas chromatography3"1021 using a Shimadzu GC.14B chromatograph with flame ionisation detection; a SUPELCOWAX 10 (SUPELCO) capillary column, 60 m X 0.25 mm i.d., 0.25 fim film thickness, being used. The initial oven temperature was set at 40 C, held for 10 minutes, ramped to 80 C at a rate of 4CC per minute, held at this temperature for 10 minutes, and then ramped to 200 C at 35 C per min (cleaning procedure), with a final hold at 200 C for 15 minutes. Split injections of five microlitres were made at an injection temperature of 200 C. The temperature of flame ionisalion detector was 230 C. The helium inlet pressure was fixed at 150 kpa (one millilitre per minute); the split flow employed being 100 ml/min. Detector pressures were as follows: hydrogen 60 kpa, air 60 kpa and make-up (helium) 80 kpa. The septum purge flow was fixed at 5 ml/min. Ethanol was quantified by means of a calibration (peak area versus concentration) performed before real samples were injected. Each sample was injected three times to assure reproducibility. Kinetic modelisation A simple approach, an unstructured kinetic model, was assumed. For yeast batch growth, it was assumed that growth was proportional to cell density and had an inhibition factor proportional to the square of the number of cells: IV rx = = Kv X(l-Kxi X); X(t = 0) = Xo d t This is a Riccati differential equation, which can be integrated to give the following curve: X0exp(Kx0 X = l-kxix,,(1-exp(kxt) Limiting substrate concentration was related to cell growth as follows: ds 1 dx = = qsx dt YVsdt As can be seen in this fermentation, if ethanol concentration (this also occurs with fusel alcohols) is plotted against dry weight, product formation is related to yeast growth. Thus, the production rate of ethanol and of other products is proportional to the increase in yeast dry weight. dp v dx Model parameters were obtained from batch fermentation experiments and used to predict continuous cultivation results. For continuous fermentation, the steady state equations are as follows. For biomass balance, the result is: x.-l-j!- 15] [1] [2] [3] [4] Journal of The Institute of Brewing 369

4 Volume 106, No. 6, 2000 Production of an Alcoholic Beiwrage by Fermentation ofwiwy Permeate Lactose is related to the dilution rate according to the concentration are plotted for each temperature against equation: time. In Table I, specific growth rates, final biomass levels, biomass/lactose yields (Equation 3), specific S = Sn / D [6] lactose consumption rates (Equation 3) and the KXiKx calculated Riccati parameters (Equations 1 & 2) are shown for the six temperatures employed. The specific And for ethanol concentration: growth rate increased slightly when the temperature was increased; the values obtained, between 0.1 and 0.2 rr1, are similar to those reported from previous studies with [7] cheese whey-1. The final dry weight decreased when the temperature was increased, even though the specific growth rate increased; the small decrease in the amount RESULTS AND DISCUSSION of biomass produced per amount of lactose metabolised being a strange phenomenon. This might be related to Batch Fermentations: temperature effect lower oxygen solubility at the higher temperatures. It is Cell groivth and lactose consumption In order to analyse the influence of temperature on cell well known that small quantities of oxygen are necessary as an essential nitrilite in lipid biosynthesis, necessary for the proper development of membrane functions"--0. growth, lactose consumption and ethanol production, Similar patterns were described for Saccltaromyces cerei'isiae six batch fermentations were performed at different grown in a glucose-defined medium1, but because of the temperatures. In Figure 1 the dry weight and lactose high ethanol concentrations reached, this was explained 50 8 i FIG. 1. Growth (O) and lactose (D) consumption during batch fermentations of whey permeate at 100 rpm. 370 Journal of The Institute of Brewing

5 Production ofan Alcoholic Beverage by Fermentation of Whey Permeate: I Volume 106, No. 6, 2000 TABLE I. Specific growth rates, final biomass levels, biomass/lactose yields, specific lactose consumption rates and Riccati parameters at the different temperatures assayed. Temperature, C h-1 Final dry weight, g/litre Y.x/s, g dry weight/ g lactose qs. h"1 Kx, h"1 Kxi, litre/g on the basis of an inhibition model. At the same time, the ethanol produced per unit of sugar rises with temperature, reaching a maximum near to 34 C. The biomass/lactose yield (YX/s) decreased slightly (20%) when the temperature was increased from 18.5 to 37 C, while the specific consumption rate (qs) increase with temperature (Table I). The Ricatti parameters - rx = Kx X(l-KXi X) - are also shown in Table I. This equation is a good fit (determination coefficient around 0.95 except for the 37 C fermentation, where only 0.88 regression was obtained) for the experimental data and was also appropriate for predicting the growth curve if the initial value of dry weight was known. The specific growth rate was adjusted to Arrhenius' equation and an activation energy of 2381 cal/mol and a pre-exponential parameter of 3.2 It1 were calculated. The regression coefficient, 0.93, was not very good, possibly because the calculation of specific growth rates implies some sources of error. Ethanol production The values obtained for ethanol concentration at different fermentation times for four assayed temperatures are plotted in Figure 2. For fermentation temperatures above 21 C, fermentation was rapid, final ethanol concentrations being reached in 24 h or less; while in the 18 C fermentation, the final concentration was obtained at 48 h. 25 i 20 "a,5 J,o S 5 0 J4IX g 20 ^ 15 / J 10 S 5 o' 18.5 X >" 24 Time, h 21.4 X MX 24 Time, h FIG. 2. Evolution of ethanol concentration during batch fermentations of whey permeate at 100 rpm. Table II summarises the final ethanol concentrations, yields (Equation 4), productivities (Equation 4) and efficiencies (a theoretical value of 0.54 g ethanol per g lactose was assumed). The yields of ethanol/biomass and ethanol /lactose (the differences among calculated values were only 6% of the average value, so these could be due to measurement errors) increase with temperature until reaching a maximum. The specific ethanol productivity increased with temperature. The results are similar to others published previously. The following results have been described with cheese whey: 75-80% efficiencies, Y )/s «0.4 and maximum concentration of ethanol of 21 g/litre21. Due to the limited quantities of sugars in the whey permeate - for example Artichoke Juice contains 235 g/litre total sugars7, the final levels of ethanol were much lower than others presented previously In continuous culture with 100 g/litre glucose, a Zymomonas mobilis flocculent mutant strain is capable of producing nearly 50 g/litre ethanol6. The efficiency of the biotransformation has a maximum of 88% close to 34 C. The calculated values of efficiencies between 80 and 90% were a little higher than similar data found in the literature21, the average value being 82 ± 7%. Batch fermentations: initial ph, inoculum and agitation effect Initial ph The ph value may affect growth, substrate utilisation8 and product formation. In order to analyse this effect in the fermentation, a series of batch fermentations were carried out with different initial ph's. Figure 3 presents the plots of the growth curves for the different initial ph's. There was no appreciable effect on growth (consequently, analysis of products was not performed). The CHI test performed between ph 6.1 (as expected value) and the other experiments (as actual values) give a probability of a match of more than The data from this experiment is shown in Table III. The final cell dry weights and hence biomass/lactose (Yx/S) yields were similar in all curves (only 3.5% deviation in cell dry weight and 3.4% for Yx/S). Final ph and its decrease are also shown in this table. The final ph is correlated (correlation coefficient, R2, of 0.99) with initial ph in the range assayed by the following straight line: y = x Journal of The Institute of Brewing 371

6 Volume 106, No. 6, 2000 Production of an Alcoholic Beverage by Fermentation of Whey Permeate TABLE II. Final ethanol concentrations, ethanol/biomass and ethanol/lactose yields, specific productivities and efficiencies at the different temperatures assayed. Temperature, Ethanol concentration reached, YP/x, YP/S, qp, Efficiency, C g/litre g/g g/g h"1 % Q FIG. 3. Growth curves for fermentation of whey permeate at 30DC and 100 rpm at different initial phs: 3.4 (0), 4.1 (x), 4.7 (O), 5.1 (r)) and 6.1 (A). FIG. 4. Growth curves for fermentation of whey permeate at 30 C and 100 rpm for 5',',', (A), 10';;. in), 15% (O) and 20% (0). TABLE III. Initial lactose concentration, initial dry weight, specific growth rate, final dry weight, biomass/ lactose yield and final ph at 30 C and different initial phs. Initial ph Initial lactose concentration, g/litre Initial dry weight, g/litre h' Reached dry weight, g/litrc Yx/s, g/g 0.11 Final ph ph decrease Inoculum optimisation The quantity and quality of the starter culture is most important when a fermentation is to be optimised, and several studies related to inoculum optimisation have been published49. To select the appropriate initial cell dry weight, four batch fermentations with different percentages (5, 10,15 and 20 ml starter per 100 ml fresh medium) of starter culture were carried out. The conditions were 30 C, 100 r.p.m. (orbital incubators); the relationship between the volume of the flask and the volume of media being 2.5 tol. The growth curves are shown in Figure 4. The 15 and 20% starter culture (15 and 20 ml of inoculum per 100 ml of fresh media) fermentations were very similar. The CHI test performed with a 20% growth curve (as expected values) and a 15%- curve (as actual values) produced a probability that the two curves may be the same. The 10%- starter culture fermentation had a lower cell growth rate, the CHI test performed comparing this curve (actual) with 20% (as expected) gives a probability of a match of Due to the lower dilution of the fresh media, a higher biomass level was achieved at the end. The 5% starter fermentation presented the slower growth rate and a relatively long lag phase, so is not viable. The CHI test only produced a matching probability of 0.46 that this curve corresponded to 20%, data growth curve. The optimum percentage of starter culture was 10%,. When this inoculum was used, relatively higher growth rates were obtained. Besides, it has certain advantages with respect to the higher percentages: it economises the amount of starter culture needed and dilutes the fresh 372 Journal of The Institute of Brewing

7 Production of an Alcoholic Beverage by Fermentation of Whey Permeate: I Volume 106, No. 6, 2000 media less, thus allowing more growth (and more ethanol production as this is a growth related product). The 5% starter culture is not recommendable because of the higher fermentation times and increasing possibilities of contamination (low initial dry weights and long lag phase). Agitation effect In order to analyse the effect of agitation on fermentation, three different experiments were carried out at 200 r.p.m., 100 r.p.m. and no agitation, in an orbital incubator; the temperature being maintained at 30 C. Flask volumes of 2.5 to 1 volumes of fresh media were employed and a 10% starter culture was used to inoculate the media (10 ml inoculum per 100 ml of fresh media). The analysed (HPLC) lactose concentration was 43 g/litre. In these experiments, growth was monitored by measuring the optical density at 660 nm. The absorbances were related to cell dry weight employing a previously obtained calibration curve. At 200 r.p.m., the final dry weight was 28% higher than at 100 r.p.m., a lower increase than from 0 to 100 r.p.m. The maximum growth rate was 26% higher at 200 r.p.m. than at 100 r.p.m. The ethanol concentrations for the agitations assayed are given in Table IV. When agitation was increased, more ethanol was produced. The differences between ethanol concentrations when no agitation was supplied and at 100 r.pm. are important; a 91 % increase being observed. When the change was from 100 to 200 r.p.m., the increase was only 19%. Continuous culture Biomass growth and lactose consumption Dry weight and ph variations during fermentation are presented in Figure 6. Steady states for each dilution rate assayed were obtained and biomass dry weights remained constant for at least 24 hours within the range of error of the measurement. The steady state biomass and productivities attained are depicted in Figures 7 and 8. FIG. 5. Growth curves for fermentation of whey permeate at 30 C without agitation (A), at 100 rpm (O) and at 200 rpm (O). FIG. 6. Transient dry weight (O) and ph (A) evolution during continuous fermentation of whey permeate (feed ph 6.7) at 30 C. The growth curves for the assayed agitations are shown in Figure 5. There were clear differences among the three curves with respect to specific growth rates and final biomass levels (data shown in Table IV). The final cell dry weight, when fermentation was carried out at 100 r.p.m., was twice that obtained without agitation. The maximum specific growth rale was approximately the same, but there was a longer lag phase when no agitation was applied. Maximum biomass productivity was achieved for dilution rates between 0.1 and 0.15 It1. Lactose concentrations in the effluent from the initial lactose concentration of 41 g/litre of sterilised whey permeate and steady state values for dry weight are presented in Figure 7. A linear decrease was observed in cell dry weight, in accordance with the model described by Equations 5 and 6 (Riccati), when the dilution rate was increased. Lactose TABLE IV. Initial and final cell dry weights, lag phase length, specific growth rate and ethanol concentration attained at 30 C at different agitation speeds. Agitation, r.p.m. Initial dry weight, g/litre Final dry weight, g/litre Lag phase, h Ethanol concentration, g/litre Without <3 < Journal of The Institute of Brewing 373

8 Volume 106, No. 6,2000 j! ; : --,-.. /'...- I') 376 Journal of The Institute of Brewing