- Email: [email protected]
Accurate determination of process variables in a solid-state fermentation system
Process Biochemistry, Vol. 31, No. 7, pp. 669-678, 1996 Copyright © 1996 Elsevier Science Ltd Printed in Great Britain. All rights reserved 0032-9592/96 $15.00 +0.00 ELSEVIER
P1 I : S 0 0 3 2 - 9 5 9 2
(96)00019-2
Accurate Determination of Process Variables in a Solid-State Fermentation System J. P. Smits, a* A. R i n z e m a , b J. Tramper, h E. E. Schl6sser b & W. K n o l a Division of Agrotechnology and Microbiology, TNO Nutrition and Food Research Institute, PO Box 360, 3700 AJ Zeist, The Netherlands ~Department of Food Science, Food and Bioprocess Engineering Group, Wageningen Agricultural University, PO Box 8129, 6700 EV Wageningen, The Netherlands (Received 11 January 1996; revised manuscript received 24 February 1996 and accepted 24 February 1996)
The solid-state fermentation (SSF) method described enabled accurate determination of variables related to biological activity. Growth, respiratory activity and production of carboxymethyl-cellulose-hydrolysing enzyme (CMCase) activity by Trichoderma reesei QM9414 on wheat bran was used as a model SSE The standard deviation (s) of measured substrate weight loss and glucosamine content, and the overall standard deviation of oxygen consumption rate (OCR) and carbon-dioxide production rate (CPR) after 72 h of fermentation were less than 7% of the mean measured values. A statistical method to estimate the standard deviation of time-dependent data is presented. A closed carbon balance could be set up with the results of the measurements of CPR, elemental compostion and chemical composition, indicating the reliability of the methods used. Measurements of CMC-ase activity were less accurate due to interaction between substrate and product. The measurement of the ATP level was also less accurate. Copyright © 1996 Elsevier Science Ltd
described by Raimbault & Alazard 7 is often used. This system demands, in our experience, high practical proficiency. In theory, the plugflow design of the columns does not allow homogeneous growth of fungi on the substrate. On the other hand, mixed systems do, 8 but might cause damage to hyphae resulting in poor growth. 9 Accurate measurement of variables related to biological activity in SSF is difficult but most important. The lack of good methods of measurement is a disadvantage in research on SSF compared to that on homogeneous, mixed, submerged fermentations. This might also be one of the reasons why mathematical models do not always predict experimental results. 1° For
INTRODUCTION
Solid-state fermentation (SSF) is gaining more and more attention in applied science publications because of its advantages t and its application in upgrading agricultural byproducts 2,3 and in production of fine chemicals and enzymes. 4'5 One important step in the development of an SSF process is the proper description of biological activity. Results of these descriptions can be used for modelling,6 optimization and scale-up. Several procedures are used for measurement of biological activity. The column system *To whom correspondence should be addressed. 669
670
J . P . Smits e t al.
this reason research on SSF lags behind that on submerged fermentations. In the current research programme on SSF Trichoderma reesei QM9414 on wheat bran is used as a model fermentation. Glucosamine is chosen as a measure of fungal biomass, while oxygen consumption, carbon dioxide production, ATP and weight changes are selected as variables for biological activity. Carboxymethylcellulose-hydrolysing enzyme (CMC-ase) is selected as the fermentation product. A simple Petri-dish system, in which variables related to biological activity can be measured accurately, was developed and optimized. This paper describes the system and presents the results of accuracy determinations of weight changes, oxygen consumption rate (OCR), carbon dioxide production rate (CPR), ATP, glucosamine and CMC-ase activity measurements. Furthermore, a carbon balance is set up to show the reliability of the system described.
MATERIALS AND METHODS Microorganism Trichoderma reesei QM9414 was used throughout. Spore suspensions were obtained by growth on malt-extract agar (50 g litre -1, pH 5.4) (Oxoid CM 59, Unipath Ltd, Basingstoke, UK) at 28°C in Roux bottles and harvesting the spores after 1 week with 0.1% w/w Tween-80. Glycerol (10% w/w) was added to the suspension as cryo-protectant before storage in 4.5 ml cryo-tubes at -80°C. The final suspension contained 3.1 x 107 (s = 1.2 × 107, n = 12) viable spores per ml after thawing, counted as colonyforming units on plate-count agar, potato-dextrose agar and glucose/yeast agar. Substrate preparation and inoculation A single batch of wheat bran of commercial origin stored at 4°C and containing 9% w/w moisture was used. After mixing the batch thoroughly, 90 g was added to 62 ml water in a 1 litre bottle with a screw cap (Scott Duran, Germany). After sterilization (121°C, 20 min) and cooling to room temperature, 4.5 ml spore suspension, suspended in 18 ml sterile 0.9% w/w NaCI, was aseptically sprayed over the bran and mixed by intensive shaking. More bottles of the same size and of similar contents were used if more than 174.5 g inoculated wheat bran was
needed. The moisture content of the inoculated wheat bran was 53.0% w/w.
Incubation The inoculated wheat bran was aseptically divided into 5 g portions in Petri dishes. Weights of Petri dishes were determined before and after filling. The Petri dishes were placed in a fixed climate incubator (VEA-Instruments, Houten, The Netherlands). On each of the three racks stacked inside, 20 Petri dishes were placed side by side. The climate in the incubator was set at 26.0_+0.3°C and 97_+1% RH. Temperature and humidity were registered by a computer every 5 min. The relative humidity measurement was calibrated in advance, using saturated salt solutions. 1~ Sampling Sampling was carried out from the incubator and empty ones to minimize fermentation of changes within the incubator.
by taking Petri dishes replacing them with the influences on the in local air streams
Wet and dry-matter weight Change in weight was determined by measuring the weight of a Petri dish before and after incubation. Dry-matter content was determined by measuring the change of weight of approximately 0.4 g fermented wheat bran after 16 h drying at 106°C. Dry-matter weight was calculated from the total weight of remaining bran and dry-matter content. Oxygen consumption rate and carbon dioxide production rate Oxygen consumption rate and carbon dioxide production rate were measured simultaneously in the set-up shown in Fig. 1 and which contained a measurement chamber, a tube pump (Masterflex 7014; Cole Parmer, Chicago, IL, USA), a flow meter, a paramagnetic O2 analyser (Servomex Series 1 1 0 0 ; Servomex, Zoetermeer, The Netherlands) and an infrared CO2 analyser (Servomex Series 1400). The measurement chamber was placed in the same incubator as described earlier to assure maximum similarity in conditions of fermentation and measurement. Up to three Petri dishes were placed in the measurement chamber. If fewer Petri dishes were used, the remaining space was filled with Teflon discs to decrease
Accurate determination in S S F
671
fixed-climate incubator
26"C 97 % RH
flow meter Pebi-dish
C0 2analyser
--© pump
..... I
0 2analyser
t ,/
..... ( ..... ;,:
[
\
magnetic stirring rod measurement chamber recorder
Fig. 1. Apparatus for measurement of CPR and OCR.
the gas volume in the system. The air in the chamber was additionally mixed by a magnetic stirring rod. Air from the measurement chamber was pumped through gas-tight Masterflex Tygon tubes (Cole Parmer, Chicago, IL, USA) at an air flow rate of 214 ml min -1. Calibration of the analysers was carried out using air and calibration gases containing 1.77 and 18.11% v/v CO2 in N2 (Hoekloos, Dieren, The Netherlands). Decreases in 02 concentration and increases in CO2 concentration were measured as volume percentages (% v/v) and recorded simultaneously for at least 20 min. OCR and CPR were calculated from the change in 02 and CO2 concentration per second (A%/At), respectively. The gas volume in the system (V m3), atmospheric pressure (p N m -2, as registered by the Koninklijk Nederlands Meteorologisch Instituut, De Bilt, The Netherlands), the gas constant (R = 8.314 J mol -~ K I), temperature (T = 299 K) and the amount of fermented wheat bran per measurement (W kg) were taken into account: A% r=-At
1 100
V.p 1 -----mols R.T W
-lkg
1
(1)
The gas volume was corrected for the volume occupied by the Petri dishes and wheat bran.
The density of inoculated wheat bran was 0.92 g m1-1, estimated by measuring the volume of hand-pressed wheat bran. During the time of a single measurement p was assumed to be constant. ATP From the fermented bran, 0-4 g was extracted by adding 4.0 ml 20% w/w trichloroacetic acid (TCA) in 4 mM EDTA and mixing for 30 s on a Vortex. The extract was diluted to at least 1:5000 with filter-sterilized HEPES buffer (2.5 mM EDTA in 25 mM HEPES, pH 7"75). The ATP level was measured with the Lumac Biocounter M 2500 (Lumac, Landgraaf, The Netherlands), which is based on the light generated with luciferin and firefly luciferase. Solutions of 100, 500 and 5000 nM ATP (Boehringer Mannheim, Germany) in HEPES buffer were used for calibration. Glueosamine A modified version of the method described by Lin & Cousin 12 was used for hydrolysis of the samples for glucosamine analyses. Approximately 0.1 g fermented bran was hydrolysed with 2.5 ml 4 M HC1 for 3 h at 100°C in a sealed tube. The hydrolysate was diluted to 10 ml with distilled water after cooling. The concentration
672
J.P. Smits et al.
of sugars was then measured by HPLC ionexchange chromatography (CarboPac PA-1 column with guard column; Dionex, Sunnyvale, CA, USA) with pulse amperometric detection at 25°C using o(+)-glucosamine hydrochloride (Sigma, St Louis, MO, USA) as the reference solution. 18 mM NaOH was used as eluent. CMC-ase
Considine et al. 2 and Chaha113 have reported an extraction efficiency between 90 and 93% of the total amount of recoverable cellulolytic enzyme activity from several lignocellulose substrates within 2 h of extraction with 20 mM sodiumacetate buffer (pH 5) or 0.1% Tween 80. The procedure followed in this paper was based on both reports. Samples of 1 g fermented wheat bran were extracted twice in 10 ml 0.1% w/w Tween-80 for 1 h at 4°C. The two supernatants per sample, obtained after centrifugation (10 rain, 4000 g), were pooled and dialysed in a 10 kDa cut-off cellulose-acetate dialysis tube against 2 litres 50 mM citric acid (pH 4.8) for 4 h at 4°C, to reduce the influence of the free sugars originating from the fermented bran on the measurement. The dialysed extracts obtained were stored at - 2 0 ° C prior to further analysis. CMC-ase activity was measured in undiluted dialysed extracts according to the method described by Wood & Bhat, 14 with dinitrosalysilic acid solution (DNS) as colorimetric agent for reduced sugars. A 2% w/w suspension of carboxymethyl cellulose (CMC, Aldrich, Brussels, Belgium) in 50 mM citric-acid (pH 4.8) was used as substrate. The undiluted dialysed extracts were incubated (1:1) with the 2% w/w CMC suspension for 30 min at 50°C. The reaction was stopped by adding DNS. After 5 min in a boiling water bath and after dilution with water, the absorbance at 540 nm was measured in a spectrophotometer. Activities of CMC-ase were corrected for blanks, which were incubations with citric acid buffer instead of substrate. A calibration curve was made using glucose solutions in citric acid buffer. All activities are expressed as units (U) per actual gram of fermented wheat bran. 1 U releases 1 pmol glucose per minute from the substrate. Carbon balance
A fermentation was carried out and analysed for CPR, weight changes and elemental and
chemical composition in order to set up a carbon balance. Data on changes in dry-matter weight and composition of fermented wheat bran were used to calculate the CO2 production, using the C-based elemental composition of cellulose (CH1.67Oo.83),starch (CH1.67Oo.83), free sugars (CH20), fat (CH1.84Oo.114) and wheat bran protein (CH1.94Oo.56No.27). 15 The results were compared with CO2 production estimated by integration of the curve fit through the CPR results. Samples were taken every 8 h and analysed for CPR and weight changes. Samples taken at 0 and 72 h of fermentation were analysed for chemical and elemental composition using the folowing methods for foodand feed-composition analysis. Protein was measured by the Kjeldahl method (N 6.25). Fat was measured after acid hydrolysis according to the method described by Schormiiller. 16 Ash was measured after heating to 550°C. Total carbohydrate content was measured as reduced sugars after boiling in water, digestion with pancreatin and acid hydrolysis according to the method of Van de Kamer. 17 Free sugars were analysed by using HPLC with RI detector. 18 Starch content was calculated from the total carbohydrate content minus the free-sugar content. Fibre content was determined by colorimetry using the Englyst Fiberzym Kit (Novo Nordisk Bioindustries UK Ltd, Farnham, UK). The methods used cover statistically approximately 94% of the total material analysed. Statistics and curve fit
In order to claculate the standard deviation, s, a set of data independent of time is obligatory. These were only obtained for wet and drymatter weight, ATP and CMC-ase activity. Due to the method of sampling and measurement, results of OCR, CPR and glucosamine were not obtained at the same time since the measurements of OCR and CPR took at least 20 min. To exclude a time-dependent change, the next procedure was followed. The results measured were plotted against time and a curve was fitted through these results using the method of least squares. The standard error of fit, Se, was calculated using the method described in the Appendix. To calculate the standard deviation within a time fragment T around t, in which samples were taken, eqn (2) was used
Accurate determination in SSF
673
Table 1. Mean value ~ and s of wet and dry-matter weight
ST = /./~(yi-~i)2 ~/
(2)
with ieT
after 72 h of fermentation, expressed per g initial weight
nT--m
Wet weight
The denominator in (2) represents the degree of freedom, with nT being the number of samples, and m being the number of coefficients of the curve fit in fragment T. The results obtained at around tea, t45 and t7o were not obtained completely independently of each other since, although in random order, the same 10 sets of Petri dishes were used at each measurement time. Curve fits and statistical calculations were done with the computer software SlideWrite for Windows V3.00 (Advanced Graphics Software, Carlsbad, CA, USA).
RESULTS AND DISCUSSION Wet and dry-matter weight An incubation, consisting of 60 Petri dishes each containing approximately 5 g wheat bran divided over the three racks in the incubator, was carried out to measure the standard deviation (s) in wet and dry-matter weight after 72 h of fermentation. In Fig. 2 wet and dry-matter weight are shown, expressed per g initial weight of inoculated wheat bran to facilitate statistical calculations and comparisons. Mean value (.f) and s are given in Table 1. Of both wet and drymatter weight, s was less than 5-0% of the mean values. The location of the Petri dish in the incubator influenced the weight. Wet weights of Petri dishes located on rack 2 were higher than those on racks 1 and 3. In contrast, results of
• q-s
Dry-matter weight .f+_s
0.828 + 0.041 0-895 +__0-028 0.806 + 0.040
0.408_ 0.020 0.400 + 0.009 0.409 + 0.012
(g)
Rack 1 Rack 2 Rack 3
(g)
dry-matter weight measurements did not differ significantly. The differences in wet weight found on rack 2 could be explained by differences in accumulation of water droplets, which were observed on the inside of the cover of the Petri dishes. This formation of condensate droplets might be caused by small differences in air currents, and therefore by small differences in temperature and diffusion. The influence of the condensate droplets seemed to be negligible. To check the influence of the amount of inoculated wheat bran per Petri dish on the drymatter weight changes, 56 Petri dishes, initially containing between 1.5 and 11.5 g wheat bran, were incubated for 72 h. The results are shown in Fig. 3. The dashed line represents initial drymatter weight per g wheat bran, calculated from the dry-matter content of the wheat bran and the amount of water added to it during substrate preparation and inoculation. The results showed that below approximately 4 g initial
0.6
t.20
Plate
Plate 2
1
Plate
3
A o) v
6.5
1.oo 000
00o 0
°0o000 o0
O)
OOOC 0
o Oo° ooO°o
0.80 °°o°°°° o°°°°°°°°°l
o
o
°° °°
0.4
0.60
E
0.40 Pooeeo° ° ° ° e ° ° ° ° ° ° l
"0
O0000QOUOOOOOOOOOOOt19O°oOQ°OOOOoOOOOOO
;;.-
....
,o
6o
,,. . " ' v
o o
o
o O~
.... ,%- . . . . . . . . . . . . =. . . . . . . . . . . . . . . . . . . . .
,,¢ ;'.='..!
O.3
0.20
0.00
~ 5
I g
I
I
13
17
L
21
25
I
I
I
29
33
37
41
I
I
I
I
45
49
53
57
Petri-dish
Fig. 2. Wet (©) and dry-matter (e) weight after 72 h of fermentation, expressed per g initial weight of inoculated wheat bran. The dashed lines represent the initial values.
02o0
20
66
,oo
,zo
initial total weight (g)
Fig. 3. Influence of the initial amount of wheat bran per Petri dish on dry-matter weight after 72 h of fermentation, expressed per g initial weight of inoculated wheat bran. The dashed line represents the initial value.
J . P . Smits et al.
674
weight scatter increased slightly. The s in drymatter weight appeared to be less than 3.5% of the mean value in Petri dishes initially containing more than 4 g of wheat bran. Thus, high initial weights up to 11.5 g have no significant effect on the change of dry-matter weight during fermentation. Throughout the rest of the experiments between 4 and 8 g of wheat bran per Petri dish was used.
Oxygen consumption rate and carbon dioxide production rate The respiration activities, OCR and CPR, were measured in 10 sets of three Petri dishes after approximatelly 25, 45 and 70 h of fermentation. One single measurement with three Petri dishes was carried out shortly after t = 0. The order of Petri dishes in the measurement chamber did not influence the results of the measurement (data not shown). The change in concentration of oxygen and carbon dioxide, A%/At, was nearly linear throughout the measurement time of 20 min. Results of OCR measurements, presented as dots, and a fitted curve through the sets of measured OCR values are presented in Fig. 4. Based on the arbitrary choice of the logistic curve to describe the growth of fungi in S S F , 9 the differential of the logistic curve was used as curve-fit equation. The Se of the OCR measurement was shown to be 1.05 x 1 0 - 6 mol 02 s -~ kg -1, equal to 6-8% of the predicted OCR value at 70 h of fermentation. The Se of the CPR was shown to be 0.77 x 10 - 6 mol CO2 s -1 kg -~, which is 5.6% of the predicted CPR value at 70 h of fermentation. Table 2, in addi-
25 --"- 2o
tion to the Se, gives the standard deviation ST of the sets of data around t = 25, 45 and 70 h of fermentation. The values of s T are presented as a percentage of the OCR and CPR values predicted by the curve fit at t = 25, 45 and 70 h of fermentation, respectively. The fact that the values of ST appear to be larger than the overall Se is caused by the difference in degrees of freedom and the fact that values are presented as relative values. Se gave the standard error of curve fit over a period of a little more than 70 h of fermentation. The values of se might give the wrong idea that the scatter in OCR and CPR was constant over time. In fact, it fluctuated in time as can be seen by the values of the ST. Time dependence of s r is improbable, but it could not be concluded whether or not the scatter in OCR and CPR depended on the magnitude of OCR and CPR. The increase in scatter, visible during the increase of biological activity, might be caused by the influence of the cultivation method used, or by the natural variation in growth of fungi 19 due to the complexity of the substrate or to genetic determination.
Glucosamine Samples used for OCR and CPR measurements at around t - - 7 0 h were used for determination of the accuracy in glucosamine measurements. Results of the glucosamine measurement also depend therefore on time of sampling. Figure 5 shows the results of the glucosamine measurement in which the time dependence is visible. The time dependence has, like in OCR and CPR measurements, to be excluded before accuracy can be determined. However, in this case an overall curve could not be fitted to calculate se since data were only available around the tTo. course of glucosamine in time was expected to be a sigmoidal c u r v e . 6 Fragments of this curve could be described by the secondorder polynomial y
"6
= ao-l-al
"x+a2 "X2
(3)
F: 10
c rY o 0
. - t " ,. - -i"
10
i
i
2O
3O
i
40
I
5O
i
I
6O
i
/
70
i
80
fermentation time (h)
Fig. 4. O C R m e a s u r e d after t = 0, 25, 45 and 70 h of f e r m e n t a t i o n (o), and the curve fit (- - -).
which replaced the sigmoidal fit within each fragment. Figure 5 shows the second-order polynomial fit through the measured glucosamine values. The st_-_70 calculated using (3) is 0.194 mg g - i (n = 10, m = 3), which was 4.5% of the glucosamine content at t7o, predicted by (3).
Accurate determination in SSF
675
Table 2. Estimated standard error of curve fit, se, presented as a percentage of values of OCR and CPR predicted by curve-fits at t70, and standard deviation, st, presented as a percentage of values of OCR and CPR predicted by curve-fits at t25, t45 and t70 OCR ~e (% of predicted rate) Fermentation time t (h) ~r (% of predicted rate)
6.8 45 12.5
25 9.8
ATP Measurements of ATP were carried out in eight samples taken from one Petri dish and in 10 samples of wheat bran taken from 10 Petri dishes after 70 h fermentation. The correlation coefficient of the calibration curve was 0.999. The s in the ATP content of the eight samples appeared to be 5.2% and of the 10 independent samples 9.9%, of the mean ATP level. This last figure was significantly higher than the s for weight measurement, the Se for OCR and CPR and the s t = 7o for glucosamine measurements. In contrast to glucosamine, ATP is a measure of biological activity. It therefore seems to have a high potential for use in SSF control. The amount of ATP is described as a useful parameter for kinetic modelling 2° and production control, 21 but the amount present in microbial cells depends strongly on the physiological status of the cells and hence on the available substrate and sample treatment. In a complex substrate like wheat bran, large variation in ATP is therefore not unlikely. The measurement of ATP is fast but demands practical skilfullness. In the current research, extremely high inaccuracies were
° .
r,s
°
•
. . . . . . . . . II Otl .
llo"°ll " " II e •
-
g~ 3 41 t.-
8 0'1
time (h)
Fig. 5. Results of glucosamine measurement at t7o (0) and the second-order polynomial fit (- - -).
CPR 70 7.6
25 4.8
5.6 45 6.1
70 9.4
observed in the ATP measurement in samples of fermented wheat bran taken during the course of the fermentation, which forced the ATP measurement to be rejected as a tool in this SSF research. CMC-ase Measurement of activity of cellulolytic-enzymes activity with DNS as colouring agent is widely used in a broad range of research disciplines. 22-2s The DNS method has, however, a large spread. 24 The spread in the measurement of CMC-ase activity was, in our experience, caused by the interaction between substrate and product, since the presence of CMC largely increased the spread of standard glucose solutions (results not shown). The spread in CMC-ase activity measurement in dialysed extracts was 14% (n = 10) of the measured activity level of 7 U per gram wheat bran. C-balance The amount of C O 2 produced during 72 h of fermentation could be predicted using the elemental composition of the dry matter and the change in dry-matter weight calculated as C-mol substrate. CO2 production could also be predicted based on the changes in chemical composition of the bran, that is, changes in amount of starch, free sugars, fibre and fat, and their elemental composition. Components originating from biomass (cellular substances and enzymes) are present in the fermented bran and are therefore included in these components unless they belong to the missing 6% of the total weight which could not be determined. Both predictions can be compared with the CO2 production as calculated from the integral of the curve fit through the CPR values. The two predicted and the calculated values for CO2 production are given in Table 3. The elemental and chemical composition of wheat
J.P. Smits et al.
676
Table 3. Chemical and elemental composition of fermented wheat bran after 0 and 72 h of fermentation, and predicted and calculated values of CO2 production. Values are expressed per g initial dry matter
Composition
Chemical Composition Starch + fibres + free sugars Protein Fat Ash Rest Total (C-tool) Predicted CO2 production Elemental composition Elemental formula Dry matter Predicted CO2 production CPR Calculated COe production
(g) (g) (g) (g) (g) (g)
C02 production
t=O h
t=72 h
0.58 0.18 0.05 0.07 0.12 1.00 31-4 x 10 3
0.43 0.18 0.04 0.07 0.15 0.86 24.7 x 10 -3 6"7 x 10 -3
(mol) CH1.74700.680N0.053
(g) (tool)
1.00
CH1-75oOo.643No.o63 0-86
(mol)
bran at 0 and 72 h are also given in Table 3.
CONCLUSION The present system, in which Petri dishes containing inoculated wheat bran were incubated at constant temperature and humidity, was shown to be accurate. By taking 10 samples, an overall s could be obtained which was less than 7% of the mean measured value of wet weight, drymatter weight, OCR, CPR and glucosamine. Thus, within the same experimental circumstances, there was a certainty of more than 95% that the real value of a measured parameter, A, is A + 15%. For the purpose of modelling in our SSF research this is an acceptable range. Measurement of the ATP level is less accurate. The application of ATP measurement also has another drawback; the level of ATP depends largely on the physiological state of the microorganism and is thus related to the complexity of the substrate and handling of the samples. Wheat bran might be too complex a substrate for accurate measurement of ATP. CMf-ase measurements are even more inaccurate. The high s is mainly caused by the combination of CMC and glucose within the DNS method. Measurements of standard deviations of OCR, CPR and glucosamine were obstructed by time-dependent changes. These changes could be excluded by a curve fit through the sets of
4-9 x
10 - 3
5"2 x
10 - 3
measured values. The calculated s e can be used as an estimation of the actual time-independent standard deviation. The results of CPR measurement were shown to be in good agreement with those of elemental and chemical composition of fermented wheat bran, indicating a good reliability of the results of measurements. ACKNOWLEDGEMENTS This research was financially supported by Bavaria BV, Gist Brocades, Unilever Research Laboratorium, Quest International and TNO. The authors wish to thank Eric Schoen and Henk van Sonsbeek for their comments on this paper, Cees Verbeek for glucosamine analysis, members of the Organic Chemistry Department of the Agricultural University Wageningen, Netherlands, for executing elemental analyses, and members of the Division of Analytical Sciences of TNO Nutrition and Food Research Institute for composition analysis. Elemental analysis for oxygen was executed by Ets Gordinne & Co. N.V., Rotterdam, The Netherlands. REFERENCES 1. Pandey, A., Recent process developments in solidstate fermentation. Process Biochem., 27 (1992) 109-17.
Accurate determination in SSF 2. Considine, P.J., Mehra, R.K., Hackett, T.J., O'Rorke, A., Comefford, F.R. & Coughlan, M.P., Upgrading the value of agricultural residues. Ann. New York Acad. Sci., 469 (1986) 304-11. 3. Paredes-Lolbez, O. & Alpuche-Solis, A., Solid substrate fermentation - - a biotechnological approach to bioconversion of wastes. In Bioconversion of Waste Materials to Industrial Products, ed. A.M. Martin. Elsevier Scientific Publishers, London, 1992, pp. 117-45. 4. Trejo Hernafidez, M.R., Lonsane, B.K., Raimbault, M. & Roussos, S., Spectra of ergot alkaloids produced by Claviceps purpurea 1029c in solid-state fermentation system: influence of the composition of liquid medium used for impregnating sugar-cane pith bagasse. Process Biochem., 28 (1993) 23-7. 5. Fukushima, D., Koji as an important source of enzymes in the orient and its unique composite systems of proteinases and peptidases. In Use of Enzymes in Food Technology, ed. P. Dupuy. Technique et documentation Lavoisier, Paris, 1982, pp. 281-388. 6. Kim, J.H., Hosobuchi, M., Kishimoto, M., Yoshida, T., Taguchi, H. & Ryu, D.D.Y., Cellulase production by a solid-state culture systems. Biotechnol. Bioeng., 27 (1985) 1445-50. 7. Raimbault, M., Alazard, D., Culture method to study fungal growth in solid fermentation. Eur. J. Appl. Microbiol. Biotechnol., 9 (1980) 199-209. 8. Siiman, R., Enzyme formation during solid-substrate fermentation in rotating vessels. Biotechnol. Bioeng., 22 (1980) 411-20. 9. Mitchell, D.A., Do, D.D., Greenfield, P.F. & Doelle, H.W., A semimechanistic mathematical model for growth of Rhizopus oligosporus in a model solid-state fermentatiomn system. Biotechnol. Bioeng., 38 (1991) 353-62. 10. Smits, J.P., Janssens, R.J.J., Knol, W. & Bol, J., Modelling of the glucosinolate content in solid-state fermentation of rapeseed meal with fuzzy logic. J. Ferment. Bioeng., 77 (1994) 579-81. 11. Young, J.F., Humidity control in the laboratory using salt solutions - - a review. J. Appl. Chem., 17 (1967) 241-5. 12. Lin, H.H. & Cousin, M.A., Detection of mold in processed foods by high performance liquid chromatography. J. Food Protection, 48 (1985) 671-8. 13. Chahal, D. S., Production of Trichoderma reesei cellulase system with high hydrolytic potential by solid-state fermentation. In ACS Symposium Series, ed. J. Comstock. American Chemical Society, Washington DC, 1991, pp. 111-22. 14. Wood, T. M. & Bhat, K. M., Methods for measuring cellulase activities. In Methods in Enzymology, ed. W.A. Wood & S.T. Kellogg. Academic Press Inc., San Diego, 1988, pp. 100-2. 15. Pomeranz, Y. & Shellenberger, J. A. In Bread Science and Technology, ed. Y. Pomeranz & J.A. Shellenberger. AVI Publishing Company Inc., Westport, 1971, pp. 224-45. 16. Schormiiller, J. In Handbuch der Lebensmittelchemie band IV. Springer Verlag, 1969, pp. 423-5. 17. van de Kamer, J.H., Chemisch Weekblad, 38 (1941) 286-8. 18. Richter, K. &Woelk, H.U., Die Stiirke, 29 (1977) 273-7. 19. Lonsane, B.K., Saucedo-Castaneda, G., Raimbault, M., Roussos, S., Viniegra-Gonzalez, G., Ghildyal, N.P., Ramakrishna, M. & Krishnaiah, M.M., Scale-up
20.
21. 22. 23. 24. 25.
677
strategies for solid-state fermentation systems. Process Biochem., 27 (1992) 259-73. Wang, H.-H., Wu, T.-Z., Hsu, J.-P. & Tsai, Y.-S., A kinetic analysis of sorghum brewing - - a typical solidstate fermentation. Nippon Shokuhin Kogyo Gakkaishi, 38 (1991) 716-21. Cochet, N., Tyagi, R.D., Ghose, T.K. & Lebeault, J.M., ATP measurement for cellulase production control. Biotechnol. Lett., 6 (1984) 155-60. Sinha, S.N., Ghosh, B.L & Ghose, S.N., Detection of cellulase inhibitor in the wheat bran culture of Asperg~llus terreus. Can. J. MicrobioL, 27 (1981) 1334-40. Sim, T.S. & Oh, J.C.S., Spent brewery grains as substrate for the production of cellulases by Trichoderma reesei QM9414. J. Ind. Microbiol., 5 (1990) 153-8. Bailey, M.J., Biely, P. & Poutanen, K., lnterlaboratory testing of methods for assay of xylanase activity. J. BiotechnoL, 23 (1992) 257-70. Olama, Z.A., Hamza, M.A., E1-Sayed, M.M. & AbdelFattah, M., Purification, properties and factors affecting the activity of Trichoderma viride cellulase. Food Chem., 47 (1993) 221-6.
APPENDIX
The value of a measured dependent variable y is determined by the relationship between y and the independent variable x in combination with a random error e. A curve fit describes this relationship by approach. If, for instance, the real relationship is y = flo+fl,x+e,
(4)
in which /~o and fl~ are the coefficients describing the relation between x and y, and in which is the random error, then the curve fit approaches this relationship by the linear equation (5)
~9 = a o + a l x
Here )~ can be seen as the approximated mean value of y dependent on x. The coefficients ao and ai are the estimated values of //o and ill. These coefficients are calculated such that
SSE = E (Yi _~i)2
(6)
is as small as possible (method of least squares). In fact eqn (6) is the sum of squares of the deviation of y from )3. The corresponding standard deviation, s~, is the standard error of the curve fit. 2
se =
--xl
(7)
n -m
se is an estimation of e, the random error. Here,
the number of degrees of freedom ( n - m )
is
678
J.P. Smits et al.
determined by the number of samples (n) and the number of coefficients in the curve fit (m). In the case of a linear fit, the number of coefficients m = 2. The curve fit describes the variation in y due
to variations in x, while s e describes the variation of y occurring independent of variations in x. By calculating se, the standard deviation of a set of data, which cannot be sampled independent of variable x, is obtained.
P1 I : S 0 0 3 2 - 9 5 9 2
(96)00019-2
Accurate Determination of Process Variables in a Solid-State Fermentation System J. P. Smits, a* A. R i n z e m a , b J. Tramper, h E. E. Schl6sser b & W. K n o l a Division of Agrotechnology and Microbiology, TNO Nutrition and Food Research Institute, PO Box 360, 3700 AJ Zeist, The Netherlands ~Department of Food Science, Food and Bioprocess Engineering Group, Wageningen Agricultural University, PO Box 8129, 6700 EV Wageningen, The Netherlands (Received 11 January 1996; revised manuscript received 24 February 1996 and accepted 24 February 1996)
The solid-state fermentation (SSF) method described enabled accurate determination of variables related to biological activity. Growth, respiratory activity and production of carboxymethyl-cellulose-hydrolysing enzyme (CMCase) activity by Trichoderma reesei QM9414 on wheat bran was used as a model SSE The standard deviation (s) of measured substrate weight loss and glucosamine content, and the overall standard deviation of oxygen consumption rate (OCR) and carbon-dioxide production rate (CPR) after 72 h of fermentation were less than 7% of the mean measured values. A statistical method to estimate the standard deviation of time-dependent data is presented. A closed carbon balance could be set up with the results of the measurements of CPR, elemental compostion and chemical composition, indicating the reliability of the methods used. Measurements of CMC-ase activity were less accurate due to interaction between substrate and product. The measurement of the ATP level was also less accurate. Copyright © 1996 Elsevier Science Ltd
described by Raimbault & Alazard 7 is often used. This system demands, in our experience, high practical proficiency. In theory, the plugflow design of the columns does not allow homogeneous growth of fungi on the substrate. On the other hand, mixed systems do, 8 but might cause damage to hyphae resulting in poor growth. 9 Accurate measurement of variables related to biological activity in SSF is difficult but most important. The lack of good methods of measurement is a disadvantage in research on SSF compared to that on homogeneous, mixed, submerged fermentations. This might also be one of the reasons why mathematical models do not always predict experimental results. 1° For
INTRODUCTION
Solid-state fermentation (SSF) is gaining more and more attention in applied science publications because of its advantages t and its application in upgrading agricultural byproducts 2,3 and in production of fine chemicals and enzymes. 4'5 One important step in the development of an SSF process is the proper description of biological activity. Results of these descriptions can be used for modelling,6 optimization and scale-up. Several procedures are used for measurement of biological activity. The column system *To whom correspondence should be addressed. 669
670
J . P . Smits e t al.
this reason research on SSF lags behind that on submerged fermentations. In the current research programme on SSF Trichoderma reesei QM9414 on wheat bran is used as a model fermentation. Glucosamine is chosen as a measure of fungal biomass, while oxygen consumption, carbon dioxide production, ATP and weight changes are selected as variables for biological activity. Carboxymethylcellulose-hydrolysing enzyme (CMC-ase) is selected as the fermentation product. A simple Petri-dish system, in which variables related to biological activity can be measured accurately, was developed and optimized. This paper describes the system and presents the results of accuracy determinations of weight changes, oxygen consumption rate (OCR), carbon dioxide production rate (CPR), ATP, glucosamine and CMC-ase activity measurements. Furthermore, a carbon balance is set up to show the reliability of the system described.
MATERIALS AND METHODS Microorganism Trichoderma reesei QM9414 was used throughout. Spore suspensions were obtained by growth on malt-extract agar (50 g litre -1, pH 5.4) (Oxoid CM 59, Unipath Ltd, Basingstoke, UK) at 28°C in Roux bottles and harvesting the spores after 1 week with 0.1% w/w Tween-80. Glycerol (10% w/w) was added to the suspension as cryo-protectant before storage in 4.5 ml cryo-tubes at -80°C. The final suspension contained 3.1 x 107 (s = 1.2 × 107, n = 12) viable spores per ml after thawing, counted as colonyforming units on plate-count agar, potato-dextrose agar and glucose/yeast agar. Substrate preparation and inoculation A single batch of wheat bran of commercial origin stored at 4°C and containing 9% w/w moisture was used. After mixing the batch thoroughly, 90 g was added to 62 ml water in a 1 litre bottle with a screw cap (Scott Duran, Germany). After sterilization (121°C, 20 min) and cooling to room temperature, 4.5 ml spore suspension, suspended in 18 ml sterile 0.9% w/w NaCI, was aseptically sprayed over the bran and mixed by intensive shaking. More bottles of the same size and of similar contents were used if more than 174.5 g inoculated wheat bran was
needed. The moisture content of the inoculated wheat bran was 53.0% w/w.
Incubation The inoculated wheat bran was aseptically divided into 5 g portions in Petri dishes. Weights of Petri dishes were determined before and after filling. The Petri dishes were placed in a fixed climate incubator (VEA-Instruments, Houten, The Netherlands). On each of the three racks stacked inside, 20 Petri dishes were placed side by side. The climate in the incubator was set at 26.0_+0.3°C and 97_+1% RH. Temperature and humidity were registered by a computer every 5 min. The relative humidity measurement was calibrated in advance, using saturated salt solutions. 1~ Sampling Sampling was carried out from the incubator and empty ones to minimize fermentation of changes within the incubator.
by taking Petri dishes replacing them with the influences on the in local air streams
Wet and dry-matter weight Change in weight was determined by measuring the weight of a Petri dish before and after incubation. Dry-matter content was determined by measuring the change of weight of approximately 0.4 g fermented wheat bran after 16 h drying at 106°C. Dry-matter weight was calculated from the total weight of remaining bran and dry-matter content. Oxygen consumption rate and carbon dioxide production rate Oxygen consumption rate and carbon dioxide production rate were measured simultaneously in the set-up shown in Fig. 1 and which contained a measurement chamber, a tube pump (Masterflex 7014; Cole Parmer, Chicago, IL, USA), a flow meter, a paramagnetic O2 analyser (Servomex Series 1 1 0 0 ; Servomex, Zoetermeer, The Netherlands) and an infrared CO2 analyser (Servomex Series 1400). The measurement chamber was placed in the same incubator as described earlier to assure maximum similarity in conditions of fermentation and measurement. Up to three Petri dishes were placed in the measurement chamber. If fewer Petri dishes were used, the remaining space was filled with Teflon discs to decrease
Accurate determination in S S F
671
fixed-climate incubator
26"C 97 % RH
flow meter Pebi-dish
C0 2analyser
--© pump
..... I
0 2analyser
t ,/
..... ( ..... ;,:
[
\
magnetic stirring rod measurement chamber recorder
Fig. 1. Apparatus for measurement of CPR and OCR.
the gas volume in the system. The air in the chamber was additionally mixed by a magnetic stirring rod. Air from the measurement chamber was pumped through gas-tight Masterflex Tygon tubes (Cole Parmer, Chicago, IL, USA) at an air flow rate of 214 ml min -1. Calibration of the analysers was carried out using air and calibration gases containing 1.77 and 18.11% v/v CO2 in N2 (Hoekloos, Dieren, The Netherlands). Decreases in 02 concentration and increases in CO2 concentration were measured as volume percentages (% v/v) and recorded simultaneously for at least 20 min. OCR and CPR were calculated from the change in 02 and CO2 concentration per second (A%/At), respectively. The gas volume in the system (V m3), atmospheric pressure (p N m -2, as registered by the Koninklijk Nederlands Meteorologisch Instituut, De Bilt, The Netherlands), the gas constant (R = 8.314 J mol -~ K I), temperature (T = 299 K) and the amount of fermented wheat bran per measurement (W kg) were taken into account: A% r=-At
1 100
V.p 1 -----mols R.T W
-lkg
1
(1)
The gas volume was corrected for the volume occupied by the Petri dishes and wheat bran.
The density of inoculated wheat bran was 0.92 g m1-1, estimated by measuring the volume of hand-pressed wheat bran. During the time of a single measurement p was assumed to be constant. ATP From the fermented bran, 0-4 g was extracted by adding 4.0 ml 20% w/w trichloroacetic acid (TCA) in 4 mM EDTA and mixing for 30 s on a Vortex. The extract was diluted to at least 1:5000 with filter-sterilized HEPES buffer (2.5 mM EDTA in 25 mM HEPES, pH 7"75). The ATP level was measured with the Lumac Biocounter M 2500 (Lumac, Landgraaf, The Netherlands), which is based on the light generated with luciferin and firefly luciferase. Solutions of 100, 500 and 5000 nM ATP (Boehringer Mannheim, Germany) in HEPES buffer were used for calibration. Glueosamine A modified version of the method described by Lin & Cousin 12 was used for hydrolysis of the samples for glucosamine analyses. Approximately 0.1 g fermented bran was hydrolysed with 2.5 ml 4 M HC1 for 3 h at 100°C in a sealed tube. The hydrolysate was diluted to 10 ml with distilled water after cooling. The concentration
672
J.P. Smits et al.
of sugars was then measured by HPLC ionexchange chromatography (CarboPac PA-1 column with guard column; Dionex, Sunnyvale, CA, USA) with pulse amperometric detection at 25°C using o(+)-glucosamine hydrochloride (Sigma, St Louis, MO, USA) as the reference solution. 18 mM NaOH was used as eluent. CMC-ase
Considine et al. 2 and Chaha113 have reported an extraction efficiency between 90 and 93% of the total amount of recoverable cellulolytic enzyme activity from several lignocellulose substrates within 2 h of extraction with 20 mM sodiumacetate buffer (pH 5) or 0.1% Tween 80. The procedure followed in this paper was based on both reports. Samples of 1 g fermented wheat bran were extracted twice in 10 ml 0.1% w/w Tween-80 for 1 h at 4°C. The two supernatants per sample, obtained after centrifugation (10 rain, 4000 g), were pooled and dialysed in a 10 kDa cut-off cellulose-acetate dialysis tube against 2 litres 50 mM citric acid (pH 4.8) for 4 h at 4°C, to reduce the influence of the free sugars originating from the fermented bran on the measurement. The dialysed extracts obtained were stored at - 2 0 ° C prior to further analysis. CMC-ase activity was measured in undiluted dialysed extracts according to the method described by Wood & Bhat, 14 with dinitrosalysilic acid solution (DNS) as colorimetric agent for reduced sugars. A 2% w/w suspension of carboxymethyl cellulose (CMC, Aldrich, Brussels, Belgium) in 50 mM citric-acid (pH 4.8) was used as substrate. The undiluted dialysed extracts were incubated (1:1) with the 2% w/w CMC suspension for 30 min at 50°C. The reaction was stopped by adding DNS. After 5 min in a boiling water bath and after dilution with water, the absorbance at 540 nm was measured in a spectrophotometer. Activities of CMC-ase were corrected for blanks, which were incubations with citric acid buffer instead of substrate. A calibration curve was made using glucose solutions in citric acid buffer. All activities are expressed as units (U) per actual gram of fermented wheat bran. 1 U releases 1 pmol glucose per minute from the substrate. Carbon balance
A fermentation was carried out and analysed for CPR, weight changes and elemental and
chemical composition in order to set up a carbon balance. Data on changes in dry-matter weight and composition of fermented wheat bran were used to calculate the CO2 production, using the C-based elemental composition of cellulose (CH1.67Oo.83),starch (CH1.67Oo.83), free sugars (CH20), fat (CH1.84Oo.114) and wheat bran protein (CH1.94Oo.56No.27). 15 The results were compared with CO2 production estimated by integration of the curve fit through the CPR results. Samples were taken every 8 h and analysed for CPR and weight changes. Samples taken at 0 and 72 h of fermentation were analysed for chemical and elemental composition using the folowing methods for foodand feed-composition analysis. Protein was measured by the Kjeldahl method (N 6.25). Fat was measured after acid hydrolysis according to the method described by Schormiiller. 16 Ash was measured after heating to 550°C. Total carbohydrate content was measured as reduced sugars after boiling in water, digestion with pancreatin and acid hydrolysis according to the method of Van de Kamer. 17 Free sugars were analysed by using HPLC with RI detector. 18 Starch content was calculated from the total carbohydrate content minus the free-sugar content. Fibre content was determined by colorimetry using the Englyst Fiberzym Kit (Novo Nordisk Bioindustries UK Ltd, Farnham, UK). The methods used cover statistically approximately 94% of the total material analysed. Statistics and curve fit
In order to claculate the standard deviation, s, a set of data independent of time is obligatory. These were only obtained for wet and drymatter weight, ATP and CMC-ase activity. Due to the method of sampling and measurement, results of OCR, CPR and glucosamine were not obtained at the same time since the measurements of OCR and CPR took at least 20 min. To exclude a time-dependent change, the next procedure was followed. The results measured were plotted against time and a curve was fitted through these results using the method of least squares. The standard error of fit, Se, was calculated using the method described in the Appendix. To calculate the standard deviation within a time fragment T around t, in which samples were taken, eqn (2) was used
Accurate determination in SSF
673
Table 1. Mean value ~ and s of wet and dry-matter weight
ST = /./~(yi-~i)2 ~/
(2)
with ieT
after 72 h of fermentation, expressed per g initial weight
nT--m
Wet weight
The denominator in (2) represents the degree of freedom, with nT being the number of samples, and m being the number of coefficients of the curve fit in fragment T. The results obtained at around tea, t45 and t7o were not obtained completely independently of each other since, although in random order, the same 10 sets of Petri dishes were used at each measurement time. Curve fits and statistical calculations were done with the computer software SlideWrite for Windows V3.00 (Advanced Graphics Software, Carlsbad, CA, USA).
RESULTS AND DISCUSSION Wet and dry-matter weight An incubation, consisting of 60 Petri dishes each containing approximately 5 g wheat bran divided over the three racks in the incubator, was carried out to measure the standard deviation (s) in wet and dry-matter weight after 72 h of fermentation. In Fig. 2 wet and dry-matter weight are shown, expressed per g initial weight of inoculated wheat bran to facilitate statistical calculations and comparisons. Mean value (.f) and s are given in Table 1. Of both wet and drymatter weight, s was less than 5-0% of the mean values. The location of the Petri dish in the incubator influenced the weight. Wet weights of Petri dishes located on rack 2 were higher than those on racks 1 and 3. In contrast, results of
• q-s
Dry-matter weight .f+_s
0.828 + 0.041 0-895 +__0-028 0.806 + 0.040
0.408_ 0.020 0.400 + 0.009 0.409 + 0.012
(g)
Rack 1 Rack 2 Rack 3
(g)
dry-matter weight measurements did not differ significantly. The differences in wet weight found on rack 2 could be explained by differences in accumulation of water droplets, which were observed on the inside of the cover of the Petri dishes. This formation of condensate droplets might be caused by small differences in air currents, and therefore by small differences in temperature and diffusion. The influence of the condensate droplets seemed to be negligible. To check the influence of the amount of inoculated wheat bran per Petri dish on the drymatter weight changes, 56 Petri dishes, initially containing between 1.5 and 11.5 g wheat bran, were incubated for 72 h. The results are shown in Fig. 3. The dashed line represents initial drymatter weight per g wheat bran, calculated from the dry-matter content of the wheat bran and the amount of water added to it during substrate preparation and inoculation. The results showed that below approximately 4 g initial
0.6
t.20
Plate
Plate 2
1
Plate
3
A o) v
6.5
1.oo 000
00o 0
°0o000 o0
O)
OOOC 0
o Oo° ooO°o
0.80 °°o°°°° o°°°°°°°°°l
o
o
°° °°
0.4
0.60
E
0.40 Pooeeo° ° ° ° e ° ° ° ° ° ° l
"0
O0000QOUOOOOOOOOOOOt19O°oOQ°OOOOoOOOOOO
;;.-
....
,o
6o
,,. . " ' v
o o
o
o O~
.... ,%- . . . . . . . . . . . . =. . . . . . . . . . . . . . . . . . . . .
,,¢ ;'.='..!
O.3
0.20
0.00
~ 5
I g
I
I
13
17
L
21
25
I
I
I
29
33
37
41
I
I
I
I
45
49
53
57
Petri-dish
Fig. 2. Wet (©) and dry-matter (e) weight after 72 h of fermentation, expressed per g initial weight of inoculated wheat bran. The dashed lines represent the initial values.
02o0
20
66
,oo
,zo
initial total weight (g)
Fig. 3. Influence of the initial amount of wheat bran per Petri dish on dry-matter weight after 72 h of fermentation, expressed per g initial weight of inoculated wheat bran. The dashed line represents the initial value.
J . P . Smits et al.
674
weight scatter increased slightly. The s in drymatter weight appeared to be less than 3.5% of the mean value in Petri dishes initially containing more than 4 g of wheat bran. Thus, high initial weights up to 11.5 g have no significant effect on the change of dry-matter weight during fermentation. Throughout the rest of the experiments between 4 and 8 g of wheat bran per Petri dish was used.
Oxygen consumption rate and carbon dioxide production rate The respiration activities, OCR and CPR, were measured in 10 sets of three Petri dishes after approximatelly 25, 45 and 70 h of fermentation. One single measurement with three Petri dishes was carried out shortly after t = 0. The order of Petri dishes in the measurement chamber did not influence the results of the measurement (data not shown). The change in concentration of oxygen and carbon dioxide, A%/At, was nearly linear throughout the measurement time of 20 min. Results of OCR measurements, presented as dots, and a fitted curve through the sets of measured OCR values are presented in Fig. 4. Based on the arbitrary choice of the logistic curve to describe the growth of fungi in S S F , 9 the differential of the logistic curve was used as curve-fit equation. The Se of the OCR measurement was shown to be 1.05 x 1 0 - 6 mol 02 s -~ kg -1, equal to 6-8% of the predicted OCR value at 70 h of fermentation. The Se of the CPR was shown to be 0.77 x 10 - 6 mol CO2 s -1 kg -~, which is 5.6% of the predicted CPR value at 70 h of fermentation. Table 2, in addi-
25 --"- 2o
tion to the Se, gives the standard deviation ST of the sets of data around t = 25, 45 and 70 h of fermentation. The values of s T are presented as a percentage of the OCR and CPR values predicted by the curve fit at t = 25, 45 and 70 h of fermentation, respectively. The fact that the values of ST appear to be larger than the overall Se is caused by the difference in degrees of freedom and the fact that values are presented as relative values. Se gave the standard error of curve fit over a period of a little more than 70 h of fermentation. The values of se might give the wrong idea that the scatter in OCR and CPR was constant over time. In fact, it fluctuated in time as can be seen by the values of the ST. Time dependence of s r is improbable, but it could not be concluded whether or not the scatter in OCR and CPR depended on the magnitude of OCR and CPR. The increase in scatter, visible during the increase of biological activity, might be caused by the influence of the cultivation method used, or by the natural variation in growth of fungi 19 due to the complexity of the substrate or to genetic determination.
Glucosamine Samples used for OCR and CPR measurements at around t - - 7 0 h were used for determination of the accuracy in glucosamine measurements. Results of the glucosamine measurement also depend therefore on time of sampling. Figure 5 shows the results of the glucosamine measurement in which the time dependence is visible. The time dependence has, like in OCR and CPR measurements, to be excluded before accuracy can be determined. However, in this case an overall curve could not be fitted to calculate se since data were only available around the tTo. course of glucosamine in time was expected to be a sigmoidal c u r v e . 6 Fragments of this curve could be described by the secondorder polynomial y
"6
= ao-l-al
"x+a2 "X2
(3)
F: 10
c rY o 0
. - t " ,. - -i"
10
i
i
2O
3O
i
40
I
5O
i
I
6O
i
/
70
i
80
fermentation time (h)
Fig. 4. O C R m e a s u r e d after t = 0, 25, 45 and 70 h of f e r m e n t a t i o n (o), and the curve fit (- - -).
which replaced the sigmoidal fit within each fragment. Figure 5 shows the second-order polynomial fit through the measured glucosamine values. The st_-_70 calculated using (3) is 0.194 mg g - i (n = 10, m = 3), which was 4.5% of the glucosamine content at t7o, predicted by (3).
Accurate determination in SSF
675
Table 2. Estimated standard error of curve fit, se, presented as a percentage of values of OCR and CPR predicted by curve-fits at t70, and standard deviation, st, presented as a percentage of values of OCR and CPR predicted by curve-fits at t25, t45 and t70 OCR ~e (% of predicted rate) Fermentation time t (h) ~r (% of predicted rate)
6.8 45 12.5
25 9.8
ATP Measurements of ATP were carried out in eight samples taken from one Petri dish and in 10 samples of wheat bran taken from 10 Petri dishes after 70 h fermentation. The correlation coefficient of the calibration curve was 0.999. The s in the ATP content of the eight samples appeared to be 5.2% and of the 10 independent samples 9.9%, of the mean ATP level. This last figure was significantly higher than the s for weight measurement, the Se for OCR and CPR and the s t = 7o for glucosamine measurements. In contrast to glucosamine, ATP is a measure of biological activity. It therefore seems to have a high potential for use in SSF control. The amount of ATP is described as a useful parameter for kinetic modelling 2° and production control, 21 but the amount present in microbial cells depends strongly on the physiological status of the cells and hence on the available substrate and sample treatment. In a complex substrate like wheat bran, large variation in ATP is therefore not unlikely. The measurement of ATP is fast but demands practical skilfullness. In the current research, extremely high inaccuracies were
° .
r,s
°
•
. . . . . . . . . II Otl .
llo"°ll " " II e •
-
g~ 3 41 t.-
8 0'1
time (h)
Fig. 5. Results of glucosamine measurement at t7o (0) and the second-order polynomial fit (- - -).
CPR 70 7.6
25 4.8
5.6 45 6.1
70 9.4
observed in the ATP measurement in samples of fermented wheat bran taken during the course of the fermentation, which forced the ATP measurement to be rejected as a tool in this SSF research. CMC-ase Measurement of activity of cellulolytic-enzymes activity with DNS as colouring agent is widely used in a broad range of research disciplines. 22-2s The DNS method has, however, a large spread. 24 The spread in the measurement of CMC-ase activity was, in our experience, caused by the interaction between substrate and product, since the presence of CMC largely increased the spread of standard glucose solutions (results not shown). The spread in CMC-ase activity measurement in dialysed extracts was 14% (n = 10) of the measured activity level of 7 U per gram wheat bran. C-balance The amount of C O 2 produced during 72 h of fermentation could be predicted using the elemental composition of the dry matter and the change in dry-matter weight calculated as C-mol substrate. CO2 production could also be predicted based on the changes in chemical composition of the bran, that is, changes in amount of starch, free sugars, fibre and fat, and their elemental composition. Components originating from biomass (cellular substances and enzymes) are present in the fermented bran and are therefore included in these components unless they belong to the missing 6% of the total weight which could not be determined. Both predictions can be compared with the CO2 production as calculated from the integral of the curve fit through the CPR values. The two predicted and the calculated values for CO2 production are given in Table 3. The elemental and chemical composition of wheat
J.P. Smits et al.
676
Table 3. Chemical and elemental composition of fermented wheat bran after 0 and 72 h of fermentation, and predicted and calculated values of CO2 production. Values are expressed per g initial dry matter
Composition
Chemical Composition Starch + fibres + free sugars Protein Fat Ash Rest Total (C-tool) Predicted CO2 production Elemental composition Elemental formula Dry matter Predicted CO2 production CPR Calculated COe production
(g) (g) (g) (g) (g) (g)
C02 production
t=O h
t=72 h
0.58 0.18 0.05 0.07 0.12 1.00 31-4 x 10 3
0.43 0.18 0.04 0.07 0.15 0.86 24.7 x 10 -3 6"7 x 10 -3
(mol) CH1.74700.680N0.053
(g) (tool)
1.00
CH1-75oOo.643No.o63 0-86
(mol)
bran at 0 and 72 h are also given in Table 3.
CONCLUSION The present system, in which Petri dishes containing inoculated wheat bran were incubated at constant temperature and humidity, was shown to be accurate. By taking 10 samples, an overall s could be obtained which was less than 7% of the mean measured value of wet weight, drymatter weight, OCR, CPR and glucosamine. Thus, within the same experimental circumstances, there was a certainty of more than 95% that the real value of a measured parameter, A, is A + 15%. For the purpose of modelling in our SSF research this is an acceptable range. Measurement of the ATP level is less accurate. The application of ATP measurement also has another drawback; the level of ATP depends largely on the physiological state of the microorganism and is thus related to the complexity of the substrate and handling of the samples. Wheat bran might be too complex a substrate for accurate measurement of ATP. CMf-ase measurements are even more inaccurate. The high s is mainly caused by the combination of CMC and glucose within the DNS method. Measurements of standard deviations of OCR, CPR and glucosamine were obstructed by time-dependent changes. These changes could be excluded by a curve fit through the sets of
4-9 x
10 - 3
5"2 x
10 - 3
measured values. The calculated s e can be used as an estimation of the actual time-independent standard deviation. The results of CPR measurement were shown to be in good agreement with those of elemental and chemical composition of fermented wheat bran, indicating a good reliability of the results of measurements. ACKNOWLEDGEMENTS This research was financially supported by Bavaria BV, Gist Brocades, Unilever Research Laboratorium, Quest International and TNO. The authors wish to thank Eric Schoen and Henk van Sonsbeek for their comments on this paper, Cees Verbeek for glucosamine analysis, members of the Organic Chemistry Department of the Agricultural University Wageningen, Netherlands, for executing elemental analyses, and members of the Division of Analytical Sciences of TNO Nutrition and Food Research Institute for composition analysis. Elemental analysis for oxygen was executed by Ets Gordinne & Co. N.V., Rotterdam, The Netherlands. REFERENCES 1. Pandey, A., Recent process developments in solidstate fermentation. Process Biochem., 27 (1992) 109-17.
Accurate determination in SSF 2. Considine, P.J., Mehra, R.K., Hackett, T.J., O'Rorke, A., Comefford, F.R. & Coughlan, M.P., Upgrading the value of agricultural residues. Ann. New York Acad. Sci., 469 (1986) 304-11. 3. Paredes-Lolbez, O. & Alpuche-Solis, A., Solid substrate fermentation - - a biotechnological approach to bioconversion of wastes. In Bioconversion of Waste Materials to Industrial Products, ed. A.M. Martin. Elsevier Scientific Publishers, London, 1992, pp. 117-45. 4. Trejo Hernafidez, M.R., Lonsane, B.K., Raimbault, M. & Roussos, S., Spectra of ergot alkaloids produced by Claviceps purpurea 1029c in solid-state fermentation system: influence of the composition of liquid medium used for impregnating sugar-cane pith bagasse. Process Biochem., 28 (1993) 23-7. 5. Fukushima, D., Koji as an important source of enzymes in the orient and its unique composite systems of proteinases and peptidases. In Use of Enzymes in Food Technology, ed. P. Dupuy. Technique et documentation Lavoisier, Paris, 1982, pp. 281-388. 6. Kim, J.H., Hosobuchi, M., Kishimoto, M., Yoshida, T., Taguchi, H. & Ryu, D.D.Y., Cellulase production by a solid-state culture systems. Biotechnol. Bioeng., 27 (1985) 1445-50. 7. Raimbault, M., Alazard, D., Culture method to study fungal growth in solid fermentation. Eur. J. Appl. Microbiol. Biotechnol., 9 (1980) 199-209. 8. Siiman, R., Enzyme formation during solid-substrate fermentation in rotating vessels. Biotechnol. Bioeng., 22 (1980) 411-20. 9. Mitchell, D.A., Do, D.D., Greenfield, P.F. & Doelle, H.W., A semimechanistic mathematical model for growth of Rhizopus oligosporus in a model solid-state fermentatiomn system. Biotechnol. Bioeng., 38 (1991) 353-62. 10. Smits, J.P., Janssens, R.J.J., Knol, W. & Bol, J., Modelling of the glucosinolate content in solid-state fermentation of rapeseed meal with fuzzy logic. J. Ferment. Bioeng., 77 (1994) 579-81. 11. Young, J.F., Humidity control in the laboratory using salt solutions - - a review. J. Appl. Chem., 17 (1967) 241-5. 12. Lin, H.H. & Cousin, M.A., Detection of mold in processed foods by high performance liquid chromatography. J. Food Protection, 48 (1985) 671-8. 13. Chahal, D. S., Production of Trichoderma reesei cellulase system with high hydrolytic potential by solid-state fermentation. In ACS Symposium Series, ed. J. Comstock. American Chemical Society, Washington DC, 1991, pp. 111-22. 14. Wood, T. M. & Bhat, K. M., Methods for measuring cellulase activities. In Methods in Enzymology, ed. W.A. Wood & S.T. Kellogg. Academic Press Inc., San Diego, 1988, pp. 100-2. 15. Pomeranz, Y. & Shellenberger, J. A. In Bread Science and Technology, ed. Y. Pomeranz & J.A. Shellenberger. AVI Publishing Company Inc., Westport, 1971, pp. 224-45. 16. Schormiiller, J. In Handbuch der Lebensmittelchemie band IV. Springer Verlag, 1969, pp. 423-5. 17. van de Kamer, J.H., Chemisch Weekblad, 38 (1941) 286-8. 18. Richter, K. &Woelk, H.U., Die Stiirke, 29 (1977) 273-7. 19. Lonsane, B.K., Saucedo-Castaneda, G., Raimbault, M., Roussos, S., Viniegra-Gonzalez, G., Ghildyal, N.P., Ramakrishna, M. & Krishnaiah, M.M., Scale-up
20.
21. 22. 23. 24. 25.
677
strategies for solid-state fermentation systems. Process Biochem., 27 (1992) 259-73. Wang, H.-H., Wu, T.-Z., Hsu, J.-P. & Tsai, Y.-S., A kinetic analysis of sorghum brewing - - a typical solidstate fermentation. Nippon Shokuhin Kogyo Gakkaishi, 38 (1991) 716-21. Cochet, N., Tyagi, R.D., Ghose, T.K. & Lebeault, J.M., ATP measurement for cellulase production control. Biotechnol. Lett., 6 (1984) 155-60. Sinha, S.N., Ghosh, B.L & Ghose, S.N., Detection of cellulase inhibitor in the wheat bran culture of Asperg~llus terreus. Can. J. MicrobioL, 27 (1981) 1334-40. Sim, T.S. & Oh, J.C.S., Spent brewery grains as substrate for the production of cellulases by Trichoderma reesei QM9414. J. Ind. Microbiol., 5 (1990) 153-8. Bailey, M.J., Biely, P. & Poutanen, K., lnterlaboratory testing of methods for assay of xylanase activity. J. BiotechnoL, 23 (1992) 257-70. Olama, Z.A., Hamza, M.A., E1-Sayed, M.M. & AbdelFattah, M., Purification, properties and factors affecting the activity of Trichoderma viride cellulase. Food Chem., 47 (1993) 221-6.
APPENDIX
The value of a measured dependent variable y is determined by the relationship between y and the independent variable x in combination with a random error e. A curve fit describes this relationship by approach. If, for instance, the real relationship is y = flo+fl,x+e,
(4)
in which /~o and fl~ are the coefficients describing the relation between x and y, and in which is the random error, then the curve fit approaches this relationship by the linear equation (5)
~9 = a o + a l x
Here )~ can be seen as the approximated mean value of y dependent on x. The coefficients ao and ai are the estimated values of //o and ill. These coefficients are calculated such that
SSE = E (Yi _~i)2
(6)
is as small as possible (method of least squares). In fact eqn (6) is the sum of squares of the deviation of y from )3. The corresponding standard deviation, s~, is the standard error of the curve fit. 2
se =
--xl
(7)
n -m
se is an estimation of e, the random error. Here,
the number of degrees of freedom ( n - m )
is
678
J.P. Smits et al.
determined by the number of samples (n) and the number of coefficients in the curve fit (m). In the case of a linear fit, the number of coefficients m = 2. The curve fit describes the variation in y due
to variations in x, while s e describes the variation of y occurring independent of variations in x. By calculating se, the standard deviation of a set of data, which cannot be sampled independent of variable x, is obtained.