J Sci Food Agric 1997, 75, 273È280 Maintenance of Rumen Protozoa Populations in a Dual Outflow Continuous Fermenter Laurent-Philippe Broudiscou,1* Yves Papon,1 Michel Fabre1 and Anne F Broudiscou2 1 Station de Recherches sur la Nutrition des Herbivores, Institut National de la Recherche Agronomique, Centre de Clermont-Theix, 63122 St Genès Champanelle, France 2 Laboratoire de Methodologie de la Recherche Experimentale, Universite dÏAix-Marseille III, 52 av Escadrille Normandie Niemen, 13013 Marseille, France (Received 27 September 1996 ; accepted 3 April 1997) Abstract : Using a 25h2 fractional factorial design, Ðve factors were screened for their ability to promote rumen protozoa growth in a dual outÑow fermenter and on their inÑuence on fermentation. The tested factors were the addition of bovine serum (BS, 20 ml litre~1) and of a 100 g litre~1 yeast extract solution (20 ml litre~1) to artiÐcial saliva, the addition of diethystilbestrol (15 mg day~1) to fermentation broths, the reduction of shear stresses by lowering the stirring speed from 260 rpm to 230 rpm, and the improvement of ciliate sequestration by the use of a polyurethane foam belt. In the fermentors, the ciliate population density ranged from 10 kl~1 to 58 kl~1. The genus Isotricha was rarely observed, with a population density estimated at \0É7 kl~1 while Dasytricha was not maintained. All the experimental factors markedly inÑuenced total protozoa numbers. Lowering the stirring speed was the sole beneÐcial treatment. Nutritional supplements, in particular BS, were all detrimental to Entodinium. Their association was characterised by a signiÐcant antagonism. The population size of Polyplastron and Eudiplodinium was lowered only by BS addition, possibly through the changes induced in the fermentation pattern. J Sci Food Agric 75, 273È280 (1997) No. of Figures : 0. No. of Tables : 5. No. of References : 18 Key words : rumen protozoa, continuous culture system, factors of maintenance INTRODUCTION ditions. Their ability to support a varied protozoal ciliate population, however, is uneven (MansÐeld et al 1995). As protozoa are major agents in the rumen microbial community, investigating the best set of general operating conditions to maintain their population at high levels is an important issue. The present work was designed to screen Ðve factors for their ability to promote protozoal growth in a dual outÑow fermentor supplied with an intermediate diet made of 750 g kg~1 pelleted orchard-grass hay and 250 g kg~1 ground barley, and on their inÑuence on fermentation. Three nutritional factors were investigated : the addition of yeast extract, bovine serum and diethylstilbestrol. Two physical factors were also considered : the reduction of shear stresses by lowering the stirring For many years in Europe, the need for reliable alternative techniques to animal experimentation has been urged by ethical and legal prescriptions (Mahouy 1995). In ruminant nutrition, several continuous culture devices have been developed since the 1960s to study the metabolism of rumen microbes. Mastering an artiÐcial situation and closely simulating rumen physiological mechanisms were the main trends in validating in vitro systems. Dual effluent stirred-tank reactors (Hoover et al 1976 ; Merry et al 1987) have proved to be suitable for the accurate control of environmental con* To whom correspondence should be addressed. 273 ( 1997 SCI. J Sci Food Agric 0022-5142/97/$17.50. Printed in Great Britain L -P Broudiscou et al 274 speed, and the improvement of ciliate sequestration by the use of a polyurethane foam belt. EXPERIMENTAL The study was conducted in dual outÑow continuous fermenters adapted from Hoover et al (1976). In these discontinuously stirred reactors, the fermentation broth is characterised by a di†erential turnover of particulate and liquid phases due to the presence of an overÑow and an outlet equipped with a Ðlter (in our apparatus, a 100 km pore size stainless-steel gauze). Intact fermenter contents leave the vessel by the overÑow, mainly during the periodic mixings, while only liquids and small particles can be pumped through the Ðltering outlet. These effluent fractions are currently called displaced effluent and Ðltered effluent, respectively. The fermenters differed from the original apparatus in the daily collection and analysis of fermentation gases, the supply of solid substrates in separate meals rather than on a semicontinuous mode, the use of a marine impeller (Applikon Dependable Instruments, Schiedam, The Netherlands) and of baffles rather than Ñat-blade turbine impellers for fermentation broth mixing. The working volumes of vessels, which are adjustable in our system, were set to 1100 ^ 10 ml. Four wethers, equipped with a rumen cannula and kept indoors in individual boxes, were used as donors of rumen contents. The animals were fed on 900 g day~1 chopped hay and 300 g day~1 ground and pelleted barley, distributed in two meals, and they had free access to water and mineral supplement. Four litres of rumen contents were withdrawn after a 24 h fast. They were immediately strained through a 1 mm stainlesssteel screen and kept at 39¡C under a continuous Ñow of N . Within minutes, the vessels, previously Ðlled with 2 700 ml of artiÐcial saliva and 20 g of pelleted feed, were inoculated with 400 ml of strained rumen Ñuid, Ñushed with N then closed and started up. The fermenters, 2 maintained at 39¡C, were continuously infused with an artiÐcial saliva containing Na HPO 12 H O (6É157 g 2 4 2 litre~1) NaHCO (5É268 g litre~1), KHCO (0É597 g 3 3 litre~1), NaCl (0É305 g litre~1), HClÈcysteine (0É4 g litre~1) and adjusted to pH 10É0 with 1É63 g litre~1 NaOH. The fermentation broth was separately complemented with CaCl (31É7 mg day~1), MgCl 2 2 (47É5 mg day~1) and (NH ) SO (0É755 g day~1). A pel42 4 leted diet made of 750 g kg~1 pelleted orchard-grass hay and 250 g kg~1 ground barley was supplied to the fermenters at the rate of 10 g at 10 :00 and 20 g at 18 :00. The saliva and the Ðltered effluent were pumped to obtain dilution rates of 0É03 h~1 and 0É06(^0É002) h~1, respectively, for particle and liquid phases. All effluents were separately collected in a container kept at ]4¡C. The screening of environmental conditions favouring the maintenance of a wide range of protozoal genera in vitro considered three nutritional factorsÈthe addition of bovine serum (BS, 20 ml litre~1) and of a 100 g litre~1 yeast extract solution (YE, 20 ml litre~1) to the artiÐcial saliva and the addition of diethystilbestrol (DES, 15 mg day~1) to fermentation brothsÈand two physical factorsÈthe stirring speed (SS, from 230 to 260 rpm) and the presence of a polyurethane foam belt (F) disposed on the inner side of the culture vessel. All chemicals were purchased from SIGMA. The polyurethane belt was organised into open cells of 300 km diameter on average. It was 8 cm high, 5 mm thick, for a volume of around 150 ml. We intended to estimate independently all the main e†ects plus the two-way interaction e†ects YE ] BS and YE ] DES. Therefore, we selected a two-level fractional factorial design 25h2 (Haaland 1989) using generators I 4 124 4 235 4 1345. Table 1 shows the corresponding experimental worksheet made of eight combinations of experimental treatments. The run K8 represented the presumed best set of conditions for protozoa maintenance and it was applied Ðve times to give an estimation of experimental error. The 12 runs were randomly assigned to six independent fermenters, identically assembled, which were operated for two seven-day contiguous experimental periods. The fermentation gas volumes and the amount of collected effluents and delivered saliva were measured every day. TABLE 1 Worksheet for the 25h2 experimental design Run Y east extract Stirring speed (rpm) DES Foam Bovine serum 1 2 3 4 5 6 7 8 No Yes No Yes No Yes No Yes 260 260 230 230 260 260 230 230 No No No No Yes Yes Yes Yes Yes No No Yes Yes No No Yes Yes Yes No No No No Yes Yes Maintenance of rumen protozoa in a continuous fermenter After a 5-day adaptation period, the displaced and Ðltered effluents collected on days 6 and 7 were pooled and kept at [20¡C prior to subsampling for volatile fatty acids (VFA) and ammonia nitrogen (NH -N) 3 analysis. On day 6, the fermentation broths were sampled at 16 :00 for pH, Eh, VFA and NH -N determi3 nation. On days 6 and 7 the fermentation gas composition was determined. At the end of the experimental period, 16 h after the last input of solid substrate, the fermenter contents were weighed, an aliquot was Ðltered through a 200 km pore size nylon gauze, then the residues were washed in saline and Ðltered twice. Filtrates were pooled, and a sample was Ðxed by addition of three volumes of 13É3 ml litre~1 glutaraldehyde, 666 ml litre~1 glycerol solution, for determination of the protozoa population in the vessels. In addition to total protozoa counting, the genera Entodinium, Polyplastron, Eudiplodinium, Isotricha and Dasytricha were identiÐed according to Ogimoto and Imai (1981) and enumerated in a Jessen cell on a microscope (Nikon Labophot). VFA were determined as described by Jouany (1982). NH -N was determined as described by Davies and 3 Taylor (1965). Fermentation gases were analysed by gas chromatography (column Ðlled with Carboxen 1000 (Supelco) at 100¡C). Individual gas molar concentration was calibrated using a certiÐed standard (relative accuracy of 2%, Alphagaz no 073562.00). The hydrogen balance was calculated as described by Demeyer and Van Nevel (1975). 275 The results were submitted to linear regression by a SAS procedure (SAS 1990), Data were Ðtted to the following model : Y \ b ] b . Y E ] b . SS ] b . DES ] b . F 0 1 2 3 4 ] b . BS ] b . Y E . DES ] b . Y E . BS (1) 5 13 15 for coded variables, Y E, DES, F, BS : [1 \ absent, ]1 \ present ; SS : [1 \ 260 rpm, ]1 \ 230 rpm. The coefficient estimates of the model were compared to null by StudentÏs t-test. RESULTS The mean ciliate population density in the rumen Ñuid of donor animals was equal to 395É6 kl~1, 16 h after feeding a meal. The population densities of the genera Entodinium, Polyplastron, Eudiplodinium, Isotricha and Dasytricha were respectively 368É0 kl~1 (93É0% of the entire population), 5É8 kl~1 (1É5%), 8É0 kl~1 (2É0%), 3É4 kl~1 (0É9%), 10É4 kl~1 (2É6%). In the fermentors, the ciliate population density ranged from 10 kl~1 to 58 kl~1. The genus Dasytricha was never observed. All the experimental factors markedly inÑuenced total protozoa numbers, mainly by a†ecting Entodinium maintenance (Table 2). Nutritional supplements, in particular BS, were all detrimental to Entodinium. Their association was characterised by a signiÐcant antagonism, as shown by the two-way interaction coefficient TABLE 2 Regression analysis for protozoa population dataa Response Source DF T otal (]106) SS Model Error Total R2 Adj R2 Exp error 7 4 11 2304 39É1 2343 Prob [ F 0É002 0É98 0É95 3É13 Factor Coe† Intercept YE SS DES F BS Y E ] DES Y E ] BS 27É0 [4É7 3É1 [2É1 [6É0 [11É3 4É2 4É8 Entodinium (]106) SS 2046 35É7 2082 Prob [ F 0É002 0É98 0É95 2É99 P-value 0É01 0É04 0É012 0É005 0É0004 0É017 0É010 Coe† 24É0 [4É8 2É8 [2É0 [5É5 [10É5 3É9 4É6 Poly ] Eudiplo (]106) SS 10É01 2É53 12É54 Prob [ F 0É22 0É80 0É45 0É79 P-value 0É009 0É049 0É12 0É005 0É0005 0É018 0É010 Coe† 2É89 0É08 0É37 0É05 [0É33 [0É91 0É36 0É04 P-value 0É77 0É24 0É85 0É28 0É027 0É25 0É88 a Abbreviations : Adj R2 : adjusted R2 ; P-value : conÐdence level for t-test of null-hypothesis ; the model coefficients are calculated for coded variables, Y E, DES, F, BS : [1 \ absent, ]1 \ present ; SS : [1 \ 260 rpm, ]1 \ 230 rpm. 276 TABLE 3 Regression analysis for fermentation pattern dataa Response Source Model Error Total DF pH SS 7 4 11 0É0363 0É0119 0É0482 R2 Adj R2 Exp error 0É75 0É32 0É055 Factor Coe† Intercept YE SS DES F BS Y E ] DES Y E ] BS 7É00 [0É06 [0É01 0É01 0É02 0É02 [0É01 0É01 Eh (mv) Prob [ F 0É31 SS 371É5 61É7 433É2 Prob [ F 0É12 0É86 0É61 3É93 P-value 0É034 0É60 0É72 0É30 0É41 0É78 0É63 Coe† [380É4 1É8 [0É2 [0É4 [2É7 [3É6 4É3 2É0 N-NH (mg litre~1) 3 SS Prob [ F 155826 2298 158124 0É0016 0É99 0É96 24É0 P-value 0É25 0É89 0É77 0É11 0É052 0É032 0É20 Coe† 311É5 41É6 4É4 [6É7 0É4 104É3 [4É6 3É2 Acetate (mM SS 7783 572 8355 M~1 V FA) Prob [ F 0É033 0É93 0É81 12É0 P-value 0É007 0É61 0É45 0É96 0É0002 0É59 0É71 Coe† 625 0 [4 [1 [6 [23 0 2 Propionate (mM SS 3077 1785 4863 M~1 V FA) Prob [ F 0É54 0É63 0 6É7 P-value 0É96 0É36 0É74 0É20 0É004 0É95 0É71 Coe† 179 [2 [1 0 5 1 [1 1 Butyrate (mM SS 1031 64 1095 M~1 V FA) Prob [ F 0É025 0É94 0É84 4É0 P-value 0É33 0É77 0É98 0É078 0É79 0É72 0É77 Coe† 115 [1 5 [2 [2 8 2 [1 P-value 0É63 0É029 0É24 0É31 0É004 0É27 0É53 a Abbreviations : Adj R2 : adjusted R2 ; P-value : conÐdence level for t-test of null-hypothesis. L -P Broudiscou et al Response Source DF Acetate (mmol d~1) SS Model Error Total R2 Adj R2 Exp error 7 4 11 423É8 78É8 502É6 Prob [ F 0É15 0É84 0É57 4É44 Factor Coe† Intercept YE SS DES F BS Y E ] DES Y E ] BS 76É45 5É41 1É21 0É48 [1É51 1É93 [0É51 [0É14 Propionate (mmol d~1) SS 65É7 3É8 69É5 Prob [ F SS Prob [ F 0É022 4É44 0É07 4É52 0É002 0É94 0É85 0É98 P-value 0É022 0É46 0É76 0É37 0É27 0É75 0É93 Coe† 21É79 1É31 0É38 0É03 0É13 1É42 0É05 0É25 Isobutyrate (mmol d~1) 0É98 0É96 0É136 P-value 0É016 0É31 0É94 0É71 0É013 0É89 0É49 Coe† 1É66 0É20 0É07 [0É01 [0É05 0É53 0É02 0É05 Butyrate (mmol d~1) SS 42É8 4É1 46É9 Prob [ F SS Prob [ F 0É052 9É061 0É078 9É138 0É0006 0É91 0É76 1É01 P-value 0É012 0É18 0É89 0É32 0É0003 0É63 0É37 Coe† 15É22 0É80 0É68 [0É04 0É09 1É35 [0É01 [0É19 Isovalerate (mmol d~1) 0É99 0É98 0É139 P-value 0É080 0É12 0É91 0É80 0É017 0É98 0É60 Coe† 2É08 0É29 0É07 [0É03 [0É08 0É77 0É05 0É06 V alerate (mmol d~1) SS 6É896 0É118 7É014 Prob [ F SS Prob [ F 0É002 1É982 0É192 2É174 0É053 0É98 0É95 0É172 P-value 0É003 0É21 0É55 0É16 0É0001 0É31 0É24 Coe† 2É83 0É38 0É07 0É07 0É02 0É55 0É00 0É02 Caproate (mmol d~1) 0É91 0É76 0É219 P-value 0É003 0É30 0É30 0É71 0É0007 0É98 0É74 Coe† 1É75 0É13 0É04 0É23 0É06 0É21 [0É07 0É00 Carbon (mmol d~1) SS 16820 1001 17821 Prob [ F 0É023 0É94 0É85 15É8 P-value 0É16 0É62 0É033 0É46 0É045 0É39 0É98 Coe† 323É2 22É8 7É5 2É7 [2É4 23É6 [1É0 0É3 Maintenance of rumen protozoa in a continuous fermenter TABLE 4 Regression analysis for VFA production data P-value 0É013 0É23 0É64 0É68 0É01 0É86 0É95 Abbreviations : Adj R2 : adjusted R2 ; P-value : conÐdence level for t-test of null-hypothesis. 277 L -P Broudiscou et al 278 TABLE 5 Regression analysis for gas production dataa Response Source Model Error Total R2 Adj R2 Exp error Factor Intercept YE SS DES F BS Y E ] DES Y E ] BS DF 7 4 11 V olume (litres d~1) SS Prob [ F 2É927 0É461 3É388 0É12 0É86 0É63 0É339 Coe† 3É85 0É35 0É21 0É05 [0É05 0É03 0É05 0É10 CH (mmol d~1) 4 CO (mmol d~1) 2 SS SS 219É2 48É3 267É5 Prob [ F 0É19 0É82 0É50 3É48 P-value 0É037 0É14 0É69 0É71 0É83 0É68 0É42 Coe† 38É9 3É42 1É67 0É40 [1É30 [0É63 0É73 0É98 2945 416 3362 Prob [ F 0É098 0É88 0É66 10É2 P-value 0É043 0É23 0É75 0É33 0É62 0É57 0É45 Coe† 115É1 10É6 6É6 1É5 [0É5 1É7 1É3 3É0 P-value 0É036 0É12 0É68 0É90 0É65 0É73 0É43 a Abbreviations : Adj R2 : adjusted R2 ; P-value : conÐdence level for t-test of null-hypothesis. estimates. The use of a polyurethane foam belt appeared to be disadvantageous as well, while lowering the stirring speed to 230 rpm was slightly beneÐcial. The population size of larger protozoa (Polyplastron and Eudiplodinium, of the subfamily Diplodininae) was merely lowered by BS addition. The genus Isotricha was observed in the experimental runs K2 and K3 (Table 1) with a population estimated at less than 8 ] 105. Table 3 shows regression analysis for fermentation pattern data. The pH and redox potential were not signiÐcantly modiÐed by any experimental factor. BS and YE supplementation greatly increased NH -N concentration, with 3 no signiÐcant interaction. The VFA concentration in the fermentation broth averaged 78 mM. The variation in VFA concentration was poorly explained by the statistical model (F-value of 0É79, R2 of 0É47, experimental error of 7É5). In the absence of nutritional supplements, the molar proportions of acetate, propionate and butyrate were, respectively, 654, 175 and 116 mM M~1 VFA. The addition of BS tended to decrease acetate proportion, in relation with an increase of butyrate. As shown by Table 4, the factors Y E and BS clearly increased the amount of VFA produced per day, with no signiÐcant interaction. Both additives led to the same increase in the amount of carbon recovered in effluents in the form of fermentation end-product. BS, however, had an e†ect mainly on the production of VFA containing four or more carbon atoms, in particular branched-chain VFA, while Y E acted more on smaller VFA (acetate and propionate). This di†erence was conÐrmed by the regression analysis of fermentation gas production data (Table 5). Supple- mentation with Y E was the most active factor and it led to a 20% increase in the volume of fermentation gases and in the amount of methane and carbon dioxide collected per day. The variation in the amount of hydrogen collected was poorly explained by the statistical model. The hydrogen balance was satisfactory in all experiments, with a high recovery rate estimated at 98%. DISCUSSION The physical characteristics of fermentation broths were favourable for ciliate growth in all experimental runs. A pH of 7É0 and a redox potential lower than [350 mv were optimal values to maintain protozoa (Quinn et al 1962 ; Clarke and Hungate 1966). In a previous experiment with the same diet and artiÐcial saliva, the osmotic pressure of the broth was equal to 210 mOsmol litre~1. Anaerobiosis was satisfactory, with very low proportions of oxygen detected in fermentation gases. All experimental factors markedly inÑuenced the ciliate population size in fermenters. The expected antagonisms between nutritional factors were also observed. Williams and Coleman (1991) stressed the poor biosynthetic ability of ciliates and the need for a number of organic compounds in the medium to maintain these microorganisms in vitro. YE supplies aminoacids and water-soluble vitamins and it is a common component of culture medium. BS is a complex mixture of albumins, globulins and growth promoters (Blood et al 1989). According to Clarke and Hungate (1966) the addition of BS or protozoa extract to holotrich culture Maintenance of rumen protozoa in a continuous fermenter medium was necessary. The e†ect of DES on ciliates in ruminants is still uncertain (Williams and Coleman 1991). However, in dairy cows fed on alfalfa hay, barley and soya, complementing the diet with DES (8 mg litre~1) increased rumen ciliate population density three-times (Ibrahim et al 1970). Subsequently, Itabashi and Kurihara (1972) noted a favourable action of BS (20 ml litre~1) and DES (10 mg litre~1) addition to saliva on protozoa growth in single-outÑow stirred-tank fermentors supplied with alfalfa hay, crushed barley and concentrate pellets. Using similar incorporation rates, we did not observe such e†ects in our experimental conditions. All the nutritional additives tested hindered ciliate growth, in accordance with the observations of Williams and Coleman (1991) on pure cultures of protozoa. We can primarily conclude that the substrates supplied by the diet and the bacterial biomass in our fermentors efficiently supported protozoal growth. The mechanisms by which BS and Y E were detrimental are unclear. They can be attributed to a number of abiotic or biotic factors. The amounts of carbon atoms recovered in the form of VFA when Y S or BS was used imply that both supplements were extensively fermented. Individual VFA and gas production data indicate that they were metabolised in two distinct patterns. BS fermentation was characteristic of aminoacid catabolism while the addition of Y E also led to end-products from carbohydrate catabolism. The speciÐc e†ect of BS on the genera Eudiplodinium and Polyplastron could be related to this di†erence. Ammonia and VFA concentration in the fermentation broths, though, did not reach detrimental values (Williams and Coleman 1991). The supply of bacteria with soluble organic substrate may also have increased bacterial competitiveness for energy resources, leading to a limitation of ciliate growth. The choice of a resolution III fractional factorial design limited the number of experiments but it allowed major e†ects to be identiÐed with a number of two-way interaction e†ects and with higher order interaction terms. In particular, the main e†ect of SS was identiÐed with the two-way interaction between BS and DES. The main e†ect of F was similarly identiÐed with the threeway interaction between Y E, BS and DES. The F and SS coefficient estimates reÑect, to some degree, the complex interplay of nutritional factors. Nevertheless, one can reach the conclusion that a synthetic foam belt did not improve Entodinium maintenance. The foam belt was supposed to enhance ciliate entrapment in the vessels, as advocated by Abe and Kurihara (1984). In our fermentors, the solid phase turn-over rate was low enough to allow the development of a particulate bulk and prevent an extended ciliate washout, in accordance with Crawford et al (1980). Lowering the stirring speed appeared slightly beneÐcial, probably by limiting the disruption of shear-sensitive protozoa. In a dual effluent culture system equipped with an overÑow and discon- 279 tinuously agitated, both factors are partly determined by the stirring device properties. The calculation of solid phase turn-over rate assumes that the distribution of solids is homogeneous within the vessel. In practice, the stirring device must satisfactorily mix the particulate upper layer formed after the last agitation cycle, before solids Ñow out of the fermentor. A marine propeller engenders a better axial Ñow, leading to a better initial mixing of solids, and creates lower shear than the Ñatblade turbines commonly used in similar fermenters. The best combination of our experimental factors, ie a stirring speed of 230 rpm with neither supplement nor foam belt, allowed the maintenance of the major protozoal genera, in numbers large enough to signiÐcantly compete with and prey on bacteria. Taking into account the di†erences in substrate inÑows, the protozoan population in the fermentors represented approximately a third of the initial population in the rumen of sheep. A subsequent test run under similar operating conditions in all the fermentors conÐrmed the present results : a mean ciliate population density of 49É3 kl~1 (n \ 12, SD \ 8É3) was observed, compared to a predicted value of 57É8 kl~1. CONCLUSIONS Applying a 25h2 fractional factorial design provided enough information to conclude that, in our dual outÑow continuous fermentors, the use of a synthetic foam belt and supplementation with Y E, BS, DES were unnecessary and that lowering the stirring speed from 260 to 230 rpm promoted protozoa maintenance. These operating conditions were chosen for future in vitro essays. ACKNOWLEDGEMENTS The authors are grateful to Mrs J Rohlion for her skilled technical support. REFERENCES Abe M, Kurihara Y 1984 Long-term cultivation of certain rumen protozoa in a continuous fermentation system supplemented with sponge materials. J Appl Bacteriol 56 201È 213. Blood D C, Radostits O M, Arundel J H, Gay C C 1989 V eterinary Medicine. Baillière Tindall, London, UK, pp 1463È 1464. Clarke R T J, Hungate R E 1966 Culture of the rumen holotrich ciliate Dasytricha ruminantium Schuberg. Appl Microbiol 14 340È345. Crawford R J, Hoover W H, Knowlton P H 1980 E†ects of solids and liquid Ñows on fermentation in continuous cultures. I. Dry matter and Ðber digestion, VFA production and protozoa numbers. J Anim Sci 51 975È985. 280 Davies A W, Taylor K 1965 Application of the autoanalyser in a river authority laboratory. Symposium T echnicon pp 294È300. Demeyer D I, Van Nevel C J 1975 Methanogenesis, an integrated part of carbohydrate fermentation and its control. In : Digestion and Metabolism in the Ruminant, eds McDonald I W & Warner A C I. University of New England Publishing, Armidale, pp 366È382. Haaland P D 1989 Experimental Design in Biotechnology. Marcel Dekker Inc, New York, USA, pp 85È112. Hoover W H, Crooker B A, Sni†en C J 1976 E†ects of di†erential solidÈliquid removal rates on protozoa numbers in continuous cultures of rumen contents. J Anim Sci 43 528È 534. Ibrahim E A, Ingalls J R, Stanger N E 1970 E†ect of dietary diethylstilbestrol on populations and concentrations of ciliate protozoa in dairy cattle. Can J Anim Sci 50 101È106. Itabashi H, Kurihara Y 1972 Maintenance of rumen microbial populations in an improved artiÐcial rumen. Jap J Ecol 22 262È265. Jouany J-P 1982 Volatile fatty acid and alcohol determination in digestive contents, silage juices, bacterial cultures and anaerobic fermentor contents. Sci Aliments 2 131È144. L -P Broudiscou et al Mahouy G 1995 Legislation et reglementation de lÏexperimentation animale. In : L ivre Blanc sur lÏExpe rimentation Animale, eds Carnaud C, Mahouy G, Marsac J, Maurin-Blanchet H, Schmoll E & Tambourin P. Les editions INSERM, CNRS Editions, Paris, France, pp 1È12. MansÐeld H R, Endres M F, Stern M D 1995 Comparison of microbial fermentation in the rumen of dairy cows and dual Ñow continuous culture. Anim Feed Sci T echnol 55 47È66. Merry R J, Smith R H, McAllan A B 1987 Studies of rumen function in an in vitro continuous culture system. Arch Anim Nutr 37 475È488. Ogimoto K, Imai S 1981 Atlas of Rumen Microbiology. Japan ScientiÐc Societies Press, Tokyo, Japan. Quinn L Y, Burroughs W, Christiansen W C 1962 Continuous culture of ruminal microorganisms on chemically deÐned medium. II. Culture medium studies. Appl Microbiol 10 583È592. SAS 1990 UserÏs Guide, Release 6.04. SAS Institute Inc, Cary, NC, USA. Williams A G, Coleman G S 1991 T he Rumen Protozoa. Springer-Verlag, New York, USA, pp 145È164.