Escherichia coli, Azotobacter vinelandii, and

JouRNAL OF BACFRoLoGY, Feb. 1973, p. 834-84 Copyright C 1973 American Society for Microbiology Vol. 113, No. 2 Printed in U.S.A. Interactions of Tetrahymena pyriformis, Escherichia coli, Azotobacter vinelandii, and Glucose in a Minimal Medium J. L. JOST, J. F. DRAKE, A. G. FREDRICKSON, AND H. M. TSUCHIYA Department of Chemical Engineering and Materials Science and Department ofmicrobiology, University of Minnesota, Minneapolis, Minnesota 55455 Received for publication 13 October 1972 A study was made of the food web formed from a protozoon, two bacteria, and a glucose minimal medium in chemostat culture. The system was also divided into simpler parts, first by omitting the protozoon to obtain a competition system, and then by omitting one or the other of the bacteria to obtain two food chains. In the competition studies, one bacterium was displaced by the other at all holding times used. In the food chain studies, sustained oscillations of the population densities of predator and prey developed at short holding times, and then changed to damped oscillations at longer holding times. In addition, the level of residual glucose remained high at long holding times. A new model of microbial growth is necessary to explain these results. In the food web studies, predation of the protozoon on the two bacteria stabilized the competition between the latter and allowed their coexistence in the same habitat. Thus, Gause's principle was circumvented. Relatively little quantitative information is tion of living organisms occupying a common available on the dynamics of the processes environment. whereby the mineral components and organic The interactions studied are shown in Fig. 1. matter of the physical environment are incorporated into the protoplasm of an ecosystem's with ammonium sulfate and salts and glucose The system was made up of a minimal medium biota. Microbial forms are involved in these as the limiting substrate for the two bacteria, processes, of course, not only in the mineralization process but also in the initial links of many The bacteria served individually and collec- Escherichia coli and Azotobacter vinelandii. food chains and food webs. tively as prey (i.e., substrate) for the ciliate, Individual links in food chains and food webs Tetrahymena pyriformis. When all three organisms were present, the system was a simple have not been studied in a dynamical context as extensively as might be warranted. Gause food web. When one or the other of the bacteria (5) was a pioneer in this field; he was the first to were omitted, the system was a simple food study the dynamics of interactions between chain. When the predator was omitted, a known microbial populations under laboratory simple competition symbiosis was obtained. It conditions. Since his classical studies, relatively few investigators have entered the field. grows axenically in rich media, it does not grow should be noted that although Tetrahymena Recent papers in the area include those by the in our simple glucose-ammonium sulfate-salts, Bungays (1), Curds and Cockbum (3), Hamilton and Preslan (6), Jannasch (7), Meers and Initially, the various kinds of symbiosis were ph 7.1, medium. Tempest (8), Megee et al. (8a), Proper and isolated into their elements for investigation. Garver (11), Salt (12), Sudo and Aiba (15), and The competition between the two bacteria for Tsuchiya et al. (16). glucose was first studied. Then the predation of The present paper deals with two types of the ciliate on Azotobacter in the absence, and symbiosis: competition between two bacteria later in the presence, of glucose was studied by for a limiting substrate, and predation of a itself. These studies were repeated for the protozoon on one, or the other, or both, of the predation of the ciliate on E. coli. Finally, the bacteria. "Symbiosis" is used here in DeBary's interactions of glucose, the two bacteria, and (4) original sense; namely, as the close associa- the ciliate in the food web were studied. 834 Downloaded from http://jb.asm.org/ on October 3, 218 by guest

VOL. 113, 1973 FOOD WEB IN CHEMOSTAT CULTURE 835 Predator tively. The amplification setting was l/8 on both Third trophic instruments; the aperture current setting was 1 on Tetrahymeno pyriformis level both instruments. However, the matching switch was 32 L on our model B; it was 4K on our model ZB. The gain control was 92 on the model B; it was 1 on the model ZB. Glucose was determined by using the glucostat reagent (Worthington Biochemical Corp.). Growth conditions. The procedures and special Prey I Prey 2 precautions followed for batch and continuous cul- vinelondii tures have been described (16). Cascades of staged Esceherichig /f Azotobocter chemostats were readily assembled by substituting a Second trophic second chemostat in place of the overflow flask of the level first stage. Standard taper fittings are used on our chemostats. Although only two stages were used in this study, cascades involving more have been assembled and used. Limiting Substrate (glucose) FIG. 1. The food web. First trophic level MATERIALS AND METHODS Organisms. E. coli B/r, kindly supplied by K. B. Raper, University of Wisconsin, Madison, was maintained on tryptone, glucose, yeast extract agar slants. A. vinelandii OP, kindly supplied by P. W. Wilson, University of Wisconsin, Madison, was maintained on Burk's medium (2). T. pyriformis W, kindly supplied by M. J. Koroly, Bryn Mawr College, Bryn Mawr, Pa., was maintained in proteose peptone broth medium. Medium. The composition of the BC medium used in both batch and continuous cultures was, in grams per liter: K2HPO4, 5.; (NHJ2SO4, 1.25; MgSO4.7H2,.1; NaCl,.1; CaCl,,.2; NaMoO4.2H2,.1; FeSO4.7H2,.1; ethylenediaminetetraacetic acid,.4; ph 7.1 +.1. Details on sterilization and other procedures dealing with the medium have already been described (16). Glucose, the limiting substrate, was added separately after sterilization. Analytical methods. Bacteria were counted on either a model B or ZB Coulter counter; the protozoon was counted on a model A. Some of the details followed in counting the bacteria and protozoa have been described (16). Advantage was taken of the differences in size of the organisms in counting them in samples of mixed culture. Tetrahymena has a mean cell volume of 2 x 14 ptm; it was counted on a model A Coulter counter with threshold setting of 1 and aperture current of 1. The two bacteria were counted with either a model B or a model ZB Coulter counter with 3-pm aperture tubes. The mean cell volume of E. coli was approximately.3 pm3, ranging from.25 to.5 pm3. The mean cell volume of Azotobacter was approximately 3. pm3, ranging from 1.5 to 6. p&m3. The lower thresholds for both models were set at 9 and 35 for E. coli and Azotobacter, respectively; the upper thresholds were set at 35 and "infinity" for E. coli and Azotobacter, respec- RESULTS Competition between bacteria. Competition between E. coli and A. vinelandii for the single limiting substrate, glucose, was studied in continuous culture. Some results of this study are shown in Fig. 2. To a steady state chemostat culture of Azotobacter into which sterile medium was flowing, a suspension of E. coli cells was added. The time of addition defined "zero hour." The level of glucose in the first chemostat at steady state was about.5 mg/ml. This was the steady state and limiting concentration for Azotobacter. In about 2 days, Azotobacter washed out to less than 1% of its initial density, and the glucose level dropped to a point where it was no longer measurable. Plate counts on the fifth day showed that there were about 15 Azotobacter per ml. Similar data were obtained at other holding times from 5.6 to 23 hr. In all cases, E. coli displaced Azotobacter; this was true even when the E. coli cells were taken from stationary-phase batch cultures. Growth of the predator on nongrowing bacterial cells. Growth of Tetrahymena on nongrowing bacterial cells was studied next. Figure 3 shows some typical results for growth on Azotobacter. Experiments were conducted by growing Azotobacter to steady state densities in the first stage of a cascade of two chemostats; the overflow containing the Azotobacter flowed into the second stage where Tetrahymena fed and grew on the bacteria. The glucose was lowered in the first stage to the limiting level for Azotobacter; hence, the bacteria showed no measurable growth in the second stage. The population of the prey in the second-stage chemostat in the absence of the predator was more than 18 organisms/ml. Data from experiments in which two holding Downloaded from http://jb.asm.org/ on October 3, 218 by guest

836 JOST ET AL. J. BACTERIOL..5 ' E.11 Time, FIG. 2. Results of pure competition between E. coli and A. vinelandii in continuous culture;, holding time in hours; St, glucose concentration in feed. (I, x C.) (I) E. -. 4 7 6 6 5 4 3 cr3 2 4>.cI 7n 'a ; 8:e9.5 hours e= 25 hours 5 1 15 FIG. 3. Chemostat transients for growth of Tetrahymena on nongrowing Azotobacter; population densities in second stage of cascade. times were employed are shown. At a holding time of 9.5 hr, the Azotobacter population in the second stage was on the order of 5 x 16 days organisms/ml; the density of Tetrahymena in the second stage was about 3 x 1' organisms/ ml. At a holding time of 25 hr, the density of the bacterium in the second stage was on the order of 2 x 16 cells/ml; the density of the predator was on the order of about 2 x 14/ml. The data show that after an initial transient period, the densities of both the prey and predator reached stable values in the chemostats. Similar data were also obtained with E. coli and Tetrahymena in cascades of chemostats. In all of these experiments, the prey did not grow in the chemostat containing the predator; instead, it served only as a "substrate" for the predator. Experiments of this type are useful for estimating certain of the kinetic and stoichiometric constants in mathematical models of predation. Food chains. Figure 4 shows data from experiments in which one of the prey species and the predator were grown together in a single chemostat. The holding time was 5.9 hr. In the case shown, growing Azotobacter and Tetrahymena shared a common environment. Oscillations in the concentration of glucose and in the population densities of Azotobacter and the protozoon were now seen, as they were in our earlier study (16) of the Dictyostelium-E. coli system. In these oscillations, glucose concentration was low when the density of Azotobacter was high; it was high when the density of the bacterium was low. The density of Tetrahymena was low when the density of Azotobacter was low. However, the density of the predator increased, slightly lagging in time, behind the rise in population of the prey. When the -g I... 4A z u :z 'O r. Q) r-.:l! Iq Downloaded from http://jb.asm.org/ on October 3, 218 by guest

VOL. 113, 1973 FOOD WEB IN CHEMOSTAT CULTURE 837 r- F 2 6 x 5-4 E3 a 2 1.6- "...4. glucose E C) ~ 5 1 15 2 25 3 FIG. 4. Sustained oscillations in the glucose-azotobacter-tetrahymena food chain; holding time 5.9 hr. density of the prey was decreased by predation, the population of Tetrahymena also dropped. It should be noted that, unlike our earlier system (16), no resistant form of the prey developed during the course of these experiments. Data were also obtained at longer holding times. Figure 5 gives the data when the holding time was 23 hr; Fig. 6 gives the data when the holding time was 4 hr. At a holding time of 23 hr, sustained oscillations (called "periodic oscillations" in our earlier paper [16]) developed after an initial transient period; the periods and amplitudes of these oscillations differ from those observed at the 5.9-hr holding time. At the 4-hr holding time, an initial drop in glucose level and increase in the densities of the two populations was followed by a period in which all of these quantities remained at essentially steady state values for the duration of the experiment. These damped oscillations at the long holding time were unanticipated. The Monod, or saturation, model predicts oscillations with greater periods and higher amplitudes at the longer holding time. We shall comment later on the unexpectedly high residual glucose; this, too, is not predicted by the Monod model. Similar results were obtained with glucose, E. coli, and Tetrahymena at holding times,, of 7.1 and 15.2 hr. Longer holding times could not be run in this case due to the clumping which occurred in the chemostats. A food web. The two food chains previously studied were combined in a single chemostat. The concentration of residual glucose and the population densities of E. coli, Azotobacter, and Tetrahymena were all followed. Figure 7 shows data for a holding time of 7.4 hr. In the absence of the predator, E. coli and Azotobacter did not coexist at any holding time (Fig. 2, for instance); Azotobacter washed out within 3 days to 16 organisms/ml, or 9% washout. However, in the presence of the predator Tetrahymena, the two bacterial species coexisted for the duration of the experiment. The long period but regular oscillations in the concentration of glucose, Azotobacter, E. coli, and Tetrahymena continued for 8 days when the experiment had to be terminated. At about this time, in this and other similar experiments, an explosive onset of wall growth developed on walls of the chemostats in less than 12 hr and rendered further taking of data futile. DISCUSSION The results of the competition experiment can be most easily explained by comparing the Monod (9) model constants for E. coli and Azotobacter in the medium used. The data indicated that the maximum specific growth rates for these organisms were.32 hr-1 and.23 hr-1, respectively. The difference here gives E. coli a slight competitive edge over Azotobacter, but the really important difference is in the saturation (or Michaelis) constants; these were 1 x 1-7 mg/ml and 1.2 x Downloaded from http://jb.asm.org/ on October 3, 218 by guest

838 JOST ET AL. J. BACTFROL., 2 E 11 6 I 21 4 2 I.1.4.6 3.2 _ 2 3 4 5 6 7 8 9 1 I1 12 13 14!5 FIG. 5. Sustained oscillations in the glucose-azotobacter-tetrahymena food chain; holding time 23 hr. E FIG. 6. Damped oscillations in the glucose-azotobacter-tetrahymena food chain; holding time 4 hr. 1-2 mg/ml for E. coli and Azotobacter, respectively. Hence, the maximum growth rate of E. coli occurs at glucose concentrations that are smaller by at least five orders of magnitude than those required to yield the maximum growth rate of Azotobacter. This gives a tremendous competitive advantage to E. coli in the medium used here. The results of the food chain experiments are most conveniently examined in the light of mathematical models that describe the interactions involved and predict what behavior I1 I should be observed. The initial model for the food chain was based on the assumption that the classical Monod model would describe not only the growth of the bacterium on glucose but also the growth of the protozoon on the bacterium. Predictions of the model were worked out and have been given. (16); Fig. 8 of that paper summarizes the model's predictions. One can see from that figure that, if a series of experiments is done in which (i) the feed glucose concentrations are held constant at a reasonably high level, but (ii) the holding time is pro- Downloaded from http://jb.asm.org/ on October 3, 218 by guest

VOL. 113, 1973 FOOD WEB IN CHEMOSTAT CULTURE 839 8 b a.414 7 > ~~~Sfz.49 mg/ml / \ @w8 74 hours 5 \ E. codi \ E i! i1l \A. vinelandii\ -o 2 a 3 4 glucose A A A 5 6 7 8 B + as B' B' - 2B where B and B' are biomasses of two different physiological states of the organism, and a is the stoichiometric coefficient. This gives the familiar plot of the specific growth rate, s, against substrate concentration, S, which is initially first order and then zero order. If the assumptions of the Monod model for growth of organisms are modified to permit the existence of a second intermediate physiologi- FIG. 7. Oscillations in the Azotobacter-E. coli-tetrahymena-glucose food web. gressively increased, then the food chains cal state, B", as: should exhibit undamped (sustained) oscillations of the residual glucose concentration and B + as the population densities of predator and prey at all practicable, i.e., up to 4-hr, holding times. In fact, the model predicts that the periods and amplitudes of the oscillations should increase and that the residual glucose should decrease to undetectable levels as the against S is holding time is increased. These predictions are not in accord with the and,8 experimental data shown in Fig. 4 through 6. From these figures, damped oscillations can be seen to develop at the longer holding times; these might eventually lead to true steady states. Curiously, the level of residual glucose did not fall to undetectable levels but remained high and not too far below the level in the feed. Evidently, the original model does not describe the system considered, and so must be rejected. Clearly, this called for a reexamination of the model used for growth of the bacteria and of the predator. The assumptions of the Monod model may be written as: B' B' + OS BB" B" -- 2B then the plot of the specific growth rate, j, initially sigmoid (instead of first order) and then zero order. In the foregoing, a are stoichiometric coefficients. The kinetic expression for the growth of organisms changes from ABS/(K + S) to ABS2/(K + S)(K' + S). This involves squaring S because of the incremental amount of substrate required to change B' to B", and the addition of a second saturation constant, K', for the incremental substrate. The double substrate kinetics is similar to that reported for dextransucrase (14). We call this the "multiple saturation model" for microbial growth. In the continuum of physiological states that a bacterium must undergo between its birth and subsequent fission into two daughter cells, more than two states may be postulated. However, for the purpose here of describing and predicting, not much is gained thereby. In this work, data on E. coli and Tetrahymena could be most readily accounted for by assuming two intermediate states, B' and B"; the data on Azotobacter were most easily explained by assuming that K' is vanishingly small so that multiple saturation kinetics reduces to the familiar Monod form. Incidentially, multiple saturation kinetics simulate the kinetics for the allosteric systems of Monod et al. (1). The details concerning the mathematical Downloaded from http://jb.asm.org/ on October 3, 218 by guest

84 JOST ET AL. J. BACTFRUOL. consequences of the multiple saturation model on various regions of stability are reported elsewhere. On the operating plane, Fig. 8, in which the holding time is plotted against the feed substrate concentration, predictions based on multiple saturation kinetics differ from those of our earlier work (16) based on Monod kinetics. A region of damped oscillations now bounds that of sustained oscillations at experimentally feasible holding times. This predicts damped oscillations at 4 hr and sustained oscillations at 5.9 and 23 hr. The principal difference between the present system and that reported earlier (16) is in the magnitude of the saturation constant K for the two predators. For predation of Dictyostelium discoideum on E. coli, the saturation constant is 4. x 1" ml - 1, and this is about two orders of magnitude greater than the saturation constant for predation of Tetrahyrrena on Azotobacter. Thus, sustained oscillations of predator and prey population densities are present in the Dictyostelium-E. coli system at all practicable experimental conditions. On the other hand, the smaller saturation constant of Tetrahymena enables these organisms to crop the Azotobacter to much lower levels, so that at long holding times, damped oscillations-and probably a true steady state-prevail. Thus, Tetrahymena follows the kinetics of the multiple saturation model for growth of organisms. On the other hand, Dictyostelium, with its nonexistence of the second intermediate physiological state, follows the kinetics of the Monod model. These results underline and emphasize the importance of a judicious choice of growth models in describing and predicting microbial behavior. Finally, the principle of competitive exclusion of Gause (5) has been well established and w 2 ad as z 11DAMPED I\SILLATK)NS StUSTAINED OSCILLATIONS NO\ OSILATOS V \REDA REAO WASHESOU TOTAL WASH4OUT % OD.2.3.4 O5.6 FE SUBSTRATE CONCENTRArE ing/ml) FIG. 8. Operating diagram for the food chain based on the multiple saturation model for growth of the predator. confirmed. When two, and only two, species compete for a limiting substrate, one must be eliminated, as seen in Fig. 2. However, in the presence of a predator on both species, E. coli and Azobacter, the principle is circumvented. This is in accord with the finding of Slobodkin (13) who found coexistence between populations of competitive hydra by periodic removal of a portion proportional to that of each population's growth rate. ACKNOWLEDGMENTS This work was supported by National Science Foundation grant GI 28874X1. J. F. Drake was supported by Public Health Service training grant Al 9-1 from the National Institute of Allergy and Infectious Diseases. LITERATURE CITED 1. Bungay, H. R., and M. L. Bungay. 1968. Microbial interactions in continuous culture. Advan. Appl. Microbiol. 1:269-29. 2. Burk, D. 193. The influence of nitrogen gas upon the organic catalysis of nitrogen fixation by Azotobacter. J. Phys. Chem. 34:1174-1194. 3. Curds, C. R., and A. Cockburn. 1971. Continuous monoxenic culture of Tetrahymena pyriform is. J. Gen. Microbiol. 66:95-18. 4. DeBary, A. 1879. Die Erscheinung der Symbiose. Trubner, Strassburg. 5. Gause, G. F. 1934. The struggle for existence. Williams & Wilkins Co., Baltimore. 6. Hamilton, R. D., and J. E. Preslan. 197. Observations on the continuous culture of a planktonic phagotrophic protozoan. J. Exp. Mar. Biol. Ecol. 5:94-14. 7. Jannasch, H. W. 1968. Growth characteristics of heterotrophic bacteria in seawater. J. Bacteriol. 95:722-723. 8. Meers, J. L., and D. W. Tempest. 1968. The influence of extracellular products on the behaviour of mixed microbial populations in magnesium-limited chemostat cultures. J. Gen. Microbiol. 52:39-317. 8a. Megee, R. D., J. F. Drake, A. G. Fredrickson, H. M. Tsuchiya. 1972. Studies in intermicrobial symbiosis: S. cerevisiae and L. casei. Can. J. Microbiol. 18: 1733-1742. 9. Monod, J. 1942. Recherches sur la croissance des cultures bacteriennes. Herman et Cie, Paris. 1. Monod, J., J. Wyman, and J. P. Changeaux. 1965. On the nature of allosteric transitions: a plausible model. J. Mol. Biol. 12:88-118. 11. Proper, G., and J. C. Garver. 1966. Mass culture of the protozoa Colpoda steinii. Biotech. Bioeng. 8:287-296. 12. Salt, G. W. 1967. Predation in a experimental protozoa population (Woodruffia-Paramecium). Ecol. Monogr. 37:113-144. 13. Slobodkin, L. B. 1964. Experimental populations of Hydrida, In British Ecological Society Jubilee Symposium, a supplement to J. Ecol. 52 and J. Animal Ecol. 33:131-148. 14. Stringer, C. S., and H. M. Tsuchiya. 1958. Kinetic study of dextransucrase. J. Amer. Chem. Soc. 8:162-1625. 15. Sudo, R., and S. Aiba. 197. Growth rate of Aspidiscidae isolated from activated sludge. Water Res. 6:137-144. 16. Tsuchiya, H. M., Drake, J. F., Jost, J. L., and A. G. Fredrickson. 1972. Predator-prey interactions of Dictyostelium discoideum and Escherichia coli in continuous culture. J. Bacteriol. 11:1147-1153. Downloaded from http://jb.asm.org/ on October 3, 218 by guest