Resource competition and community structure in aquatic micro-organisms: experimental studies of algae and bacteria along a gradient of organic carbon to inorganic phosphorus supply

. Two microbial communities were grown in chemostats receiving a low supply of inorganic Phosphorus (P) (10 µM) and different supplies of organic carbon (OC), ranging from 0 to 600 µM, either as glucose or a mixture of organic substrates. One community was a natural assemblage of lake plankton and the other was a model community composed of cultured organisms. As the supply ratio of OC to inorganic P increased, concentrations of dissolved OC increased, concentrations of dissolved P decreased and abundances of phototrophic algae decreased. Abundances of bacteria and phagotrophic organisms did not consistently change with the OC:P supply ratio. The model community was ﬁrst established with a phototroph ( Scenedesmus quadricauda ) and bacteria; the steady states of this community were invasible by the mixotroph Ochromonas danica under all OC:P supply ratios used. When OC supply was high, both microbial communities persisted with higher concentrations of dissolved OC when mixed substrates, rather than glucose, were supplied. Otherwise, the effects of organic substrate composition appeared to be secondary to those of the OC:P supply ratio. These experiments conﬁrm some elements of published theory on resource-based interactions among heterotrophic bacteria and phototrophic algae.


Introduction
The micro-organisms of the plankton display an impressive diversity of phylogenetic history and ecological interactions. Their nutritional strategies range widely. The heterotrophic bacterioplankton rely on dissolved organic matter and mineral nutrients. Some phytoplankton are obligate phototrophs, relying on light for energy while consuming dissolved mineral nutrients. Other phytoplankton are facultatively phototrophic (i.e. mixotrophic), capable of photosynthesis but also of consuming particulate matter, gaining both energy and nutrients through phagotrophy. Still other planktonic micro-organisms are obligate phagotrophs, unable to survive without particulate sources of energy and nutrients. Aquatic ecologists have recently documented the widespread distributions and functional relationships of these various organisms, which appear to co-exist in virtually every well lit aquatic ecosystem. Yet, the ecological circumstances permitting their co-existence remain unclear.
One approach of proven use in untangling the complex relationships of microbial ecology is the chemostat (Jannasch, 1967(Jannasch, , 1969), a laboratory model of an ecosystem with a well developed mathematical representation (Smith and Waltman, 1995). Using this framework, Thingstad and Pengerud (Thingstad and Pengerud, 1985) made a pioneering exploration of the co-existence of different types of planktonic micro-organisms. Their theory explicitly invokes resource competition and externally mediated resource supplies. For example, considering only heterotrophic bacteria and phototrophic algae competing for an inorganic nutrient, theory predicts co-existence if the supply of dissolved organic carbon (DOC) is low relative to that of the nutrient. This condition limits heterotrophic bacteria, otherwise assumed to be superior in competition for the inorganic nutrient. If not limited by organic carbon, bacteria are predicted to exclude algae. When the inorganic nutrient in question is phosphorus (P), experiments support the assumption of superior bacterial competitive ability (Currie and Kalff, 1984a,b,c;Pengerud et al., 1987).
Bratbak and Thingstad elaborated this theory to include the commensal supply of DOC through algal exudation (Bratbak and Thingstad, 1985). Algae nearly always excrete greater or lesser amounts of DOC (Jørgensen, 1986;Zlotnick and Dubinsky, 1989), fertilizing their bacterial competitors and making it almost impossible for algae to exclude bacteria competitively. With or without such commensalism, both theories predict that as the ratio of DOC supply to inorganic nutrient increases, bacteria increase, algae decrease and, at a sufficiently high level, disappear altogether (Grover, 1997).
Thingstad and Pengerud also examined the co-existence of algae, bacteria, and a phagotrophic predator selectively feeding on the latter (Thingstad and Pengerud, 1985). They suggested that all three types of organisms would co-exist over a wide range of resource supply conditions, with phagotrophs restraining bacterial abundance even at high DOC supply, permitting algae to persist despite their poor ability to compete for the inorganic nutrient. This theory considered only obligate phagotrophs, which neither photosynthesize nor consume inorganic nutrients, but not mixotrophs that photosynthesize and take up both dissolved and particulate nutrients. Currently, our understanding of how mixotrophy coexists with other nutritional strategies is incomplete. Rothhaupt theorized that mixotrophs could co-exist with either obligate phototrophs on the one hand, or with obligate phagotrophs on the other (Rothhaupt, 1996a). Such co-existence would depend on resource supply conditions, which he demonstrated experimentally. The most elaborate theory to date considers the nutritional strategies of bacteria, phototrophic and mixotrophic algae, and obligate phagotrophs, when all compete for a single inorganic nutrient (Thingstad et al., 1996). This theory does not predict co-existence of all four nutritional types. This paper applies an experimental approach to studying the co-existence of nutritionally diverse aquatic micro-organisms. It seeks to test theoretical predictions concerning communities of bacteria, phototrophic algae and phagotrophs (Bratbak and Thingstad, 1985;Thingstad and Pengerud, 1985). Chemostats were used to culture both (i) a natural community of micro-organisms and (ii) a model community assembled from cultured micro-organisms, first by growing algae and bacteria together, and then inoculating a mixotrophic flagellate. Inorganic P was chosen as a limiting nutrient for comparability with previous work. Phosphorus often limits algal growth in the natural community used [Joe Pool Lake, Texas (Sterner, 1994)], and often limits algal and microbial growth elsewhere (Schindler, 1977;Morris and Lewis, 1992). Experimental cultures all received the same P supply but received different supplies of DOC, thus creating a gradient of OC:P supply ratios.
Theory predicts several responses to an increased OC:P supply ratio. First, for communities of phototrophic algae and bacteria only, without phagotrophs: (i) bacteria should increase in abundance; (ii) algae should decrease in abundance, and eventually be excluded; (iii) inorganic P should decrease in concentration; and (iv) DOC should increase in concentration. Second, for communities which also have phagotrophs: (i) bacteria should remain constant in abundance; (ii) algae should decrease in abundance, though not be excluded; (iii) phagotrophs should increase in abundance; (iv) inorganic P should decrease in concentration; and (v) DOC should increase in concentration. In addition to imposing a gradient of OC:P supply ratio, two qualitatively different DOC sources were supplied: glucose and a substrate mixture. Theory does not currently address effects of substrate diversity on co-existence of bacteria with other micro-organisms, and previous experiments have used a single substrate. Therefore, the possible effects of substrate mixtures require examination.

Continuous cultures
Commercially available bioreactors (Cytolift ® ; Kontes, Vineland, NJ, USA) were used as culture vessels. These are jacketed, borosilicate reaction vessels with a working volume of 600 ml and polypropylene bases and caps. Caps were fitted with Teflon tubes for sampling, injection and air-sparging. Peristaltic pumps delivered media into vessels, with a balancing overflow to a sterile vented flask. Sparging air was supplied with aquarium pumps, humidified in gas-washing bottles, and filtered (0.22 µm). Water was circulated through vessel jackets at 25°± 0.1°C. Light was supplied with cool-white fluorescent lamps on a 14:10 h light:dark cycle, at a photon flux of 120-130 µmol quanta m -2 s -1 , measured at the position of the culture vessels. The irradiance used here should be adequate to saturate growth of Scenedesmus quadricauda (Ahlgren, 1987) and of many other species (Rhee, 1982). It is also within the range of average mixed layer irradiances found in the habitat where the natural assemblage was obtained [Joe Pool Lake (Sterner, 1994)].
Media reservoirs were vented with 0.22 µm filters to maintain sterility and contained sufficient medium for an entire experiment (10 l). The entire reactor system, from reservoir to outflow flask, was sterilized by autoclaving. After each experiment, sterility of nutrient media in reservoirs was verified by drawing samples aseptically for examination by epifluorescence microscopy. Bacterial contamination of media was never found. Each of the experiments reported here used 10 continuous cultures, arranged in two blocks of five. Each block shared a set of fluorescent lamps, temperature-controlled circulator and a peristaltic pump.
Defined nutrient media were based on Guillard's (Guillard, 1975) WC formulation, as modified by Tilman (Tilman, 1981), with P supplied at 1.0 µM as Na 2 HPO 4 , and various organic carbon supplements. This medium supports growth of many cultured and natural algae, and has been used previously in similar experiments on algal-bacterial interactions (Rothhaupt, 1992(Rothhaupt, , 1996a. As modified, it has N:P [where N is nitrogen (molar)] of 1000. Without additional supplements, it has only 1.66 µM DOC as vitamins (biotin, cyanocobalamin and Resource competition in algae and bacteria thiamine) and 20 µM DOC as EDTA, used to chelate trace metals. Two types of organic carbon supplement were added in the experimental treatments described below-either glucose alone, or a mixture of 11 substrates (Table I). Natural waters contain a variety of DOC compounds, mostly uncharacterized and much of high molecular weight (Stumm and Morgan, 1981). Compounds of low molecular weight, thought to be more available to bacteria, constitute a variable but large fraction of the total DOC [16-94% (Jørgensen, 1986)]. The substrate mixture (Table I) crudely mimics the composition of this material of low molecular weight, which is roughly 50% proteinaceous and 20% carbohydrate, with various other structures making up the balance (Stumm and Morgan, 1981). The mixture contains four amino acids found by Cowie and Hedges (Cowie and Hedges, 1992) to be common in vascular plant sources for aquatic DOC, and a variety of other compounds also common in biological materials, and likely to be present in natural waters.

Natural assemblage experiment
An inoculum was collected from Joe Pool Lake, near Dallas, TX. In this lake, algal growth at water temperatures above 15°C tends to be limited by P, and is often co-limited by nitrogen and trace nutrients (Sterner, 1994). Several samples of surface water and plankton net tows (63 µm mesh) were collected on 23 October 1995, and combined. At the time of sampling, the lake was vertically isothermal at 21°C, and dissolved nutrient concentrations suggested P limitation. Soluble reactive phosphorus (SRP) was 0.05 µM, DOC was 284 µM, dissolved inorganic N was 6.1 µM and dissolved silicate was 92 µM.
Upon return to the laboratory, the combined sample was passed through a 50 µm mesh to remove larger zooplankton. Abundant nauplii passed through the screen, so the sample was rendered anoxic by 2 h of purging with N 2 , which appeared to kill the nauplii. The sample was then split into two aliquots. One of these was enriched with ~10% (by volume) of WC medium with P = 1.0 µM and no DOC supplement, and the other was enriched with a similar proportion of WC medium (P = 1.0 µM) with 100 µM C as glucose. These aliquots were incubated for 4 days at experimental light and temperature conditions, and enriched twice with increasing proportions of media prepared as above. During pre-conditioning, bacteria increased at rates of 0.39 day -1 (with glucose) and 0.48 day -1 (without glucose), while algal biomass increased at rates of 0.06 day -1 (with glucose) and 0.42 day -1 (without glucose). To inoculate continuous cultures, the aliquots with and without glucose were mixed, and 50 ml of the mixture were aseptically transferred to each chemostat culture vessel. Five replicated nutrient treatments were used in this experiment. All cultures received WC medium with P = 1.0 µM. One treatment received no organic carbon supplement; two treatments were supplied with 100 µM C (carbon) either as glucose or the substrate mixture described above; and two treatments received 600 µM C either as glucose or as a mixture. Experimental treatments thus consisted of three molar OC:P supply ratios of 0 (no OC supplied), 100 and 600, and two qualitatively different organic carbon sources. All five treatments were applied to each of the two blocks of continuous cultures. Dilution rates of all but two cultures ranged from 0.325 to 0.381 day -1 ; two cultures (one with no OC supply, the other with 100 µM C as glucose) had much lower dilution rates of 0.107 and 0.117 day -1 . Data from these cultures were included in all analyses because they were similar to their replicates and did not emerge as outliers in any analyses. All cultures ran for 35 days, during which time several turnovers of culture volume rendered negligible any residual nutrients from the lake water inoculum, ensuring that supply was governed by the media provided.

Model system experiment
A unialgal culture of S.quadricauda was used to inoculate continuous cultures. This strain has been in the personal collection of J.Grover since 1986, and was used previously in studies of phosphorus-dependent growth and competition (Grover, 1989(Grover, , 1991a. Incidental bacterial contaminants of the Scenedesmus stock culture served as a source of bacteria for the continuous cultures. In these experimental cultures, a single rod-shaped morphotype dominated bacteria, although other morphotypes were present. The dominant rod-shaped bacterium has since been isolated into pure culture and found to grow well in WC medium supplemented with glucose, and identified as a Pseudomonas sp. (Codeco, 1998). After running this model system of algae and bacteria until steady state was achieved, a mixotrophic flagellate, Ochromonas danica UTEX 1298, was added to some cultures, to construct a 'three component' model community.
As in the natural assemblage experiment, five replicated nutrient treatments were used with the model system. All cultures again received WC medium with P = 1.0 µM. One treatment received no organic carbon supplement; two treatments were supplied with 300 µM C as glucose or as a substrate mixture; and two treatments received 600 µM C as glucose or as a mixture. Experimental treatments thus consisted of three molar OC:P supply ratios, 0, 300 and 600, and two qualitatively different organic carbon sources.
All five treatments were applied to each of the two blocks of continuous

Resource competition in algae and bacteria
cultures. Dilution rates of all cultures ranged from 0.285 to 0.304 day -1 . Cultures ran for 44 days, except for one (receiving 300 µM C as glucose) which ran for 28 days. On day 32, while cultures were in steady state, one culture of each treatment was randomly selected for inoculation with Ochromonas, and the other culture was maintained as a reference system. Ochromonas stocks were maintained axenically in the medium recommended for its cultivation (Starr and Zeikus, 1987), but 4 days prior to inoculation, it was pre-conditioned to phagotrophy by transfer to WC medium supplemented with 600 µmol each of glucose and mixed organic substrates, prepared as for the chemostats. Bacteria from chemostats were collected by passing 3 ml of chemostat contents through a 1.2 µm Nuclepore filter, and these bacteria were added to the pre-conditioning culture of Ochromonas. Ochromonas grew well in this culture, and 2 ml of this suspension were inoculated into selected chemostats. The inoculation of Ochromonas could have delivered pulses of available P and DOC, in amounts of up to 0.2 and 4.8 µM, respectively. For the cultures receiving glucose at 300 µM, the culture which ran for 44 days was inoculated with Ochromonas at day 32, and the last 12 days of the shortened run for its replicate were used as a steady state reference condition. Available medium did not allow continuation until new steady states were achieved, but dynamics of Ochromonas and other species were documented over 12 days.

Sampling and analysis
Every 2 days, 30 ml samples were taken. Two 10 ml portions were filtered onto Whatman GF/F filters and frozen for later determinations of chlorophyll a. The remaining 10 ml were preserved with formalin for bacterial counts. During the natural assemblage experiment, and from days 36 to 44 in the model system experiment (after the inoculation of Ochromonas), an additional 10 ml sample was withdrawn every 4 days and preserved with Lugol's iodine and neutral formalin for algal counts. Every 4-12 days, samples were withdrawn for SRP and DOC analyses. SRP samples (50 ml) were filtered through 0.2 µm nitrocellulose filters and analyzed in duplicate. DOC samples (20 ml) were filtered through precombusted Whatman GF/F filters and frozen in acid-washed polypropylene bottles for later analysis in triplicate.
Chlorophyll a samples were extracted in 90% acetone and measured by fluorescence (Welschmeyer, 1994). Bacteria were stained with acridine orange and counted by epifluorescence microscopy (Daley and Hobbie, 1975). During days 0-32 of the model system experiment, cells of Scenedesmus were counted by epifluorescence, while counting bacteria. For the entire natural assemblage experiment, and after day 32 in the model system experiment, algae were counted by the inverted microscope method (Utermöhl, 1931). Cell volumes for most of the species observed in the natural assemblage experiment were determined from measurements of linear dimensions under similar experimental conditions, or from the literature for less common taxa. These were converted to estimates of carbon biomass with a conversion factor of 0.2 pg C µm -3 (Rocha and Duncan, 1985). SRP was determined colorimetrically (Strickland and Parsons, 1972), as were other nutrients in lake samples. To analyze DOC, samples were oxidized with persulfate in sealed ampoules, and the resulting CO 2 was captured in a stream of N 2 gas, and determined by IR spectroscopy (Menzel and Vaccaro, 1964;McDowell et al., 1987).
All data were plotted against time, and the steady state phase in each culture was visually identified. We considered cultures to have reached a steady state when transient exponential growth of cell numbers and chlorophyll a ceased, and SRP and DOC concentrations stopped consistently decreasing. In the natural assemblage experiment, algal biomass reached a steady state while slow trends of increase or decrease of particular species occurred. Such trends are not analyzed here. After identifying steady states, data from this phase were averaged within each culture. Nutrient treatment effects were then assessed with an ANOVA for a randomized complete block design, using the block ϫ treatment interaction as an error term, and assuming block effects to be random (Neter and Wasserman, 1974). To reduce skew and heteroscedasticity, population data (cell density and biomass, chlorophyll a) were transformed to logarithms prior to analysis.
When the ANOVA indicated that treatment effects were significant, five planned linear contrasts were calculated to dissect treatment effects. Three of these tested for effects of increased OC:P supply ratio: (1) cultures with no supply of organic carbon versus those with an intermediate OC:P supply ratio (100 or 300); (2) cultures receiving no supply of organic carbon versus those with an OC:P supply ratio of 600; and (3) cultures with an intermediate OC:P supply ratio versus those with an OC:P supply ratio of 600. A one-tailed test was appropriate for these contrasts because an increase was hypothesized a priori for some response variables (DOC, bacteria, phagotrophs), and a decrease for others (SRP, algae). Contrast (4) compared cultures receiving glucose versus those receiving mixed organic substrates, and contrast (5) tested for interactions between OC:P supply ratio and substrate type. For these contrasts, a two-tailed test was appropriate because there was no hypothesized direction of response. Variances of contrasts were computed from the contrasts calculated within blocks (Steel and Torrie, 1980), and a t-test applied, with one degree of freedom (d.f.). Using a test with low d.f. obviously pushes statistical analysis to its limits, but it does serve to identify objectively the strongest treatment effects.
In the model community experiment, trends in population densities after inoculating Ochromonas were analyzed. Natural logarithms of cell density were regressed against time to estimate the net specific rate of population change (i.e. the slope coefficient) and its standard error.

Natural assemblage experiment
By day 24, after eight turnovers of culture volume, steady states of chlorophyll a concentrations and bacterial densities were achieved in all cultures. Earlier, by day 13, SRP was reduced to <0.2 µM in all cultures, indicating uptake of >80% of supplied P by resident micro-organisms. SRP remained low throughout the Resource competition in algae and bacteria experiment. Measurements of SRP were made twice during the steady state phase, on days 25 and 35. Averaging these values within cultures (Figure 1a), ANOVA did not indicate significant differences between treatments (F = 4.95, P = 0.075).
Measurements of DOC were made once during the steady state phase, on day 31 (Figure 1b), and ANOVA indicated significant differences between treatments (Table II). Results were consistent with the hypothesis that increased OC:P supply ratio would increase DOC. Cultures with an OC:P supply ratio of 600 as mixed substrates had 3-4 times the steady state DOC of other cultures, and those with an OC:P supply ratio of 600 as glucose had up to about twice the DOC of cultures with lower OC:P supply ratios. Contrasts (2) and (3) tested the differences between cultures with an OC:P supply ratio of 600 and cultures with lower supply ratios, and were significant (Table II). The influence of OC:P supply ratio on DOC concentration was complex, however. First, there was no difference between cultures with OC:P supply ratios of 100, compared with those with no organic C supply [contrast (1) insignificant, Table II]. Second, the type of C substrate appeared to be important; the significance of contrast (4) suggests that cultures supplied with mixed C substrates had higher DOC than those with glucose only (Table II). Finally, this latter difference essentially occurred only in the cultures with an OC:P supply ratio of 600, leading to significance of contrast (5) for interaction of OC:P supply ratio and type of C substrate (Table II).
At steady state, chlorophyll a was several-fold lower in cultures with an OC:P supply ratio of 600 than in other cultures (Figure 1c). Chlorophyll a concentrations were highly variable between replicates of these high OC:P cultures, however, and treatment effects were not significant (Table II). Bacterial densities were higher in three of four cultures with an OC:P supply ratio of 600 than in other cultures (Figure 1d), but variance between replicates in this treatment group was large, and treatment effects insignificant (Table II).
Flagellates capable of phagotrophy were common in these cultures. Most of them were small (<10 µm greatest linear dimension) and appeared to belong to chrysophyte genera such as Ochromonas, Poteriochromonas, Spumella and Paraphysomonas, which are principally or exclusively phagotrophic (Sandgren, 1988). Definitive identification could not be made during routine counts, however. Cryptophyte flagellates, also potential phagotrophs (Tranvik et al., 1989), were present in many cultures at much lower densities than chrysophytes. Ciliates were seen only rarely and were not counted. Phagotrophic flagellates sometimes composed a large fraction of algal biomass (Figure 1e). Phagotrophs dominated the two cultures receiving mixed organic substrates at an OC:P supply ratio of 600, and were also relatively abundant (though highly variable) in one culture receiving glucose at an OC:P supply ratio of 600. In this latter culture, eukaryotic organisms were sparse and their counts highly variable. Phagotrophs had low relative abundance in the other culture receiving glucose at an OC:P supply ratio of 600, and they ranged up to ~50% of algal biomass in cultures with lower OC:P supply ratios. Overall, there were no significant treatment effects on the relative biomass of phagotrophs analyzed after an arcsin-square root transformation (Table II).
The relative biomass of phototrophic algae is the complement of Figure 1e. In  (4) and (5) ANOVA Contrast (1) Contrast (2) Contrast (3) Contrast (4) Contrast (5)  all but one culture, chlorophytes composed at least 60% of algal biomass (excluding mixotrophs). Despite the insignificance of treatment effects for the biomass of phototrophs in aggregate, at species level, the three most common phototrophs were strongly affected by treatments. Oocystis pusilla was present in all cultures and was the most abundant phototroph (by biomass) in six of 10 cultures. In the four exceptions, either Mougeotia thylespora or Synedra acus was the most abundant phototroph, but as these species were found in only a few cultures their responses to experimental treatments could not be examined. In addition to O.pusilla, Chlorella vulgaris and Chlorella ellipsoidea occurred in all cultures, though as relatively minor components of biomass. These three phototrophic species all showed similar responses to treatments (Figure 1f-h). Biomasses at steady state were lower in cultures with an OC:P supply ratio of 600 than in other cultures, and were often lower when glucose, rather than mixed organic substrates, was supplied. For O.pusilla, treatment effects were significant (Table II). Its biomass was significantly lower in cultures with OC:P supply ratios of 600 than in other cultures [contrasts (2) and (3), Table II], and was also significantly lower in cultures supplied with glucose rather than mixed organic substrates [contrast (4), Table II]. There was no interaction between OC:P supply ratio and substrate type [contrast (5) insignificant, Table II]. For C.vulgaris, treatment effects were also significant ( Table II). The pattern of contrasts was similar to that for O.pusilla; steady state biomass of C.vulgaris was significantly lower in cultures with an OC:P supply ratio of 600 than in other cultures, and was also significantly lower with glucose rather than mixed substrates [contrasts (2)-(4), Table II]. Treatment effects were significant for C.ellipsoidea, and the pattern of contrasts was similar to that for the other phototrophs (Table II), although insignificance of contrast (4) suggests only a weak influence of glucose versus mixed organic substrates.

Model system experiment
By day 22, after six turnovers of culture volume, steady states of chlorophyll a concentrations and bacterial densities were achieved in all cultures. Earlier, by day 7 for all cultures, SRP was reduced to <0.2 µM and SRP remained at these levels subsequently, indicating that resident micro-organisms took up >80% of supplied P. For analyzing steady state responses to nutrient treatments, the period from day 22 to 32 was used (on day 32, half the cultures were inoculated with Ochromonas, disturbing the steady state). The one culture that ran only 28 days was in steady state by day 18, so the last 10 days of its run were used in analysis. SRP was measured twice during steady state. Averaging these two values, SRP declined as the OC:P supply ratio increased (Figure 2a). Treatment effects were significant, and contrasts (1)-(3) indicate that cultures with an OC:P supply ratio of 600 had lower SRP than all other cultures, and that cultures with an OC:P supply ratio of 300 had significantly lower SRP than cultures with no organic C supply (Table III). Contrasts (4) and (5) for the effect of organic substrate type and the interaction of substrate type and OC:P supply ratio were both insignificant (Table III).

Resource competition in algae and bacteria
DOC was also measured twice during steady state. Averaging these two values, DOC increased as the OC:P supply ratio increased, and was also higher when mixed organic substrates were supplied rather than glucose alone, and especially so at an OC:P supply ratio of 600 (Figure 2b). This pattern led to significant treatment effects (Table III), and to significant effects of OC:P supply ratio [contrasts (1)-(3)], organic substrate type [contrast (4)] and interaction between substrate type and OC:P supply ratio [contrast (5)].
Steady state concentrations of chlorophyll a were at least 10-fold lower in cultures with a supply of organic C than in cultures without (Figure 2c). In cultures with an OC:P supply ratio of 600, Scenedesmus was essentially excluded; chlorophyll a concentrations were near experimental blank values, and algal cells were virtually absent (a few cells at most seen during microscopic counts). In the cultures with an OC:P supply ratio of 300, chlorophyll a was lower in cultures receiving mixed organic substrates than in those receiving glucose alone. Similar patterns were observed in cell densities of Scenedesmus (data not shown).
Steady state densities of bacteria were ~5-fold higher in cultures with an OC:P supply ratio of 600 than in cultures with no organic C supply, and intermediate in the cultures with an OC:P supply ratio of 300 (Figure 2d). Treatment effects, however, were insignificant (Table III).  On day 32, an inoculum of Ochromonas was delivered to a randomly selected replicate of each treatment type. Data from days 34-44 were then used to estimate trends in population variables as specific rates of change. In all cultures, the estimated trend for Ochromonas was positive (Figure 3a), suggesting successful invasion by this organism. Averaged over all cultures, the rate of increase of Ochromonas was 0.31 day -1 . Given a dilution rate of ~0.3 day -1 , Ochromonas thus appeared capable of sustaining a reproductive rate of ~0.6 day -1 under the range of culture conditions used. There was large variation between cultures, and the trend estimates are based on only three samples. Nevertheless, these trends suggest that the mixotrophic strategy of Ochromonas was at least viable, and perhaps highly beneficial in these cultures.
Results also suggest reliance by Ochromonas on phagotrophic nutrition in at least some cultures. After inoculation with Ochromonas, rapid bacterial declines (at net rates <-0.17 day -1 ) were observed in the culture with an OC:P supply ratio of 600 and that with an OC:P supply ratio of 300 as mixed organic substrates, while bacteria remained unchanged or increased slightly in corresponding reference cultures without Ochromonas (Figure 3b). After inoculation with Ochromonas, the bacterial decline was relatively modest (-0.07 day -1 ) in the culture with an OC:P supply ratio of 300 as glucose, and bacteria increased slightly in the culture with no supply of organic C. Either Ochromonas relied less on phagotrophic nutrition than on phototrophy in these cultures, or the bacteria reproduced at rates sufficient to compensate losses to phagotrophy, or both. In all cultures inoculated with Ochromonas, there was an increase in chlorophyll a, and there were no corresponding increases in reference cultures without Ochromonas (Figure 3c). Some of this increase is due to the growth of Ochromonas and indicates a potential for increased phototrophy associated with the invasion of the mixotroph. Scenedesmus might also increase in cultures inoculated with a mixotroph, as bacterial competitors would be selectively ingested and perhaps some of their P content recycled (Rothhaupt, 1996b). There was some evidence for such an increase. In four cultures, the net growth rate of Scenedesmus was higher in the presence of Ochromonas than in its absence, and in three cultures with Ochromonas, Scenedesmus had positive net growth, while net growth was zero or negative in all cultures without Ochromonas (Figure 3d).

Discussion
These experiments tested theoretical predictions of how nutrient concentrations and microbial abundances would vary in response to defined supplies of DOC relative to inorganic P (Bratbak and Thingstad, 1985;Thingstad and Pengerud, 1985). The experiments supported the predictions that as OC:P supply ratio increases, (i) DOC increases, (ii) dissolved inorganic P decreases and (iii) algal biomass decreases. Taken together, these results support hypotheses that bacteria and phototrophic algae compete for inorganic P, and that bacteria are superior competitors for inorganic P. Given these assumptions, a high OC:P supply ratio relieves C-limitation of bacterial growth, allowing them to deplete inorganic P to levels at which algae cannot replace themselves. With bacterial growth P-limited, unconsumed DOC accumulates. This interpretation is strengthened by further work with purified cultures of Scenedesmus and Pseudomonas obtained from the model system (Codeco, 1998). At steady state, chemostats of pure Pseudomonas supplied with OC:P from 126 to 629 had an approximately linear decline of SRP with OC:P, from 0.98 µM to 0.02 µM. After inoculation at low density into these cultures, Scenedesmus invaded those with OC:P ≤ 422, but not the culture with OC:P = 629. In those cultures where Scenedesmus invaded, it reached steady state biomasses that declined with OC:P supply ratio, and the new steady state SRP ranged from 0.05 to 0.2 µM. These dynamics strongly suggest that C-sufficient Pseudomonas reduces P below the level required by Scenedesmus at steady state, while C-limited Pseudomonas do not. Other experiments with bacteria and algae grown in chemostats also support this scenario (Currie and Kalff, 1984a;Pengerud et al., 1987).
Broadly, the results obtained in this study agree sufficiently with earlier similar studies to suggest a general response of planktonic algae and bacteria to gradients of OC:P supply, mediated through a competitive mechanism. Nevertheless, the results of such laboratory studies depend partially on conditions chosen by the experimenter (e.g. temperature, irradiance). The conditions adopted in this study are representative of the habitat where the natural assemblage was obtained (Joe Pool Lake, Texas) at the time of sampling (autumn) (Sterner, 1994), but may be less representative of other times of year or other habitats. In particular, dominance by chlorophytes of natural assemblages cultured continuously with low P Resource competition in algae and bacteria supply has not often been observed [cf. (Sommer, 1983;Tilman and Kiesling, 1984;Tilman et al., 1986)], and could be a particular outcome of these experimental conditions.
The experiments reported here provide evidence that commensalism is also an important mechanism in algal-bacterial interactions. In chemostats with no external supply of DOC, bacteria persisted at relatively high abundance. Many others have observed this [e.g. (Currie and Kalff, 1984a;Bratbak and Thingstad, 1985;Rothhaupt, 1992)] and indeed, the difficulty of maintaining bacteria-free cultures of algae is well known. In the study presented here, cultures without a supply of organic C had DOC concentrations of ~50-100 µM at steady state, most of which can only have originated with algae. Dilution rates pose a lower bound to turnover on this organic matter and on algal biomass, and given this, the minimum DOC production by algae can be roughly estimated. On an absolute basis, this was 15-30 µmol C l -1 day -1 in both the natural community and model system experiment. Because the model system had lower algal biomass, its exudation rates relative to estimated net production are ~65%, as opposed to ~12% in the natural community. These are comparable with the range of relative rates reported from similar chemostat experiments [6-50% (Bratbak and Thingstad, 1985)], and in physiological studies [1-55% (Zlotnick and Dubinsky, 1989)]. Algal communities that excrete large fractions of fixed carbon as DOC may be especially susceptible to competition from bacteria for inorganic nutrients. Some theoretical predictions were not supported by the results reported here. Bacteria did not consistently increase with the OC:P supply ratio. Such an increase is predicted by theory that does not include phagotrophic predators of bacteria (Bratbak and Thingstad, 1985;Thingstad and Pengerud, 1985). Possibly, the failure of bacteria to show the predicted response in the natural assemblage experiment is due to top-down control of bacteria by phagotrophs, which is predicted by theory including the latter (Thingstad and Pengerud, 1985). However, the predicted increase of phagotrophs with OC:P supply ratio was not seen in the natural assemblage experiment. In the model system experiment, during the time when phagotrophs were absent, there was a statistically insignificant tendency for bacteria to increase with OC:P supply ratio. Possibly, the predicted increase in bacterial abundance was simply too weak to detect with an experiment having few replicates.
On the other hand, genuine complexities in the biology of bacteria and phagotrophs could frustrate the theoretical predictions. Bacteria make a number of physiological adjustments to resource conditions (Vadstein and Olsen, 1989;Chrzanowski and Kyle, 1996). Moreover, strong interactions based on allelopathy or commensalism often modify the action of resource competition in bacteria (Grover, 1997). The internal dynamics of bacterial populations and communities could complicate the response of total cell numbers to resource conditions in ways not anticipated by current theory.
Our understanding of the ecology of phagotrophs, including mixotrophs and obligate heterotrophs, is currently limited. The model system of algae and bacteria was open to invasion by mixotrophs, but it is not clear whether mixotroph abundance at steady state would have shown any pattern in response to OC:P supply ratio. Pengerud et al. found co-existence of bacteria, phototrophic algae and an obligate phagotroph over a range of OC:P supply ratios in chemostats, with little influence of the supply ratio on their abundances (Pengerud et al., 1987). Because DOC was high in their experiments, and dissolved P undetectable, it appears that the OC:P supply ratios were not low enough to induce C-limitation of bacteria. However, co-existence of bacteria, phototrophic algae and an obligate phagotroph in chemostats with no organic C supply has also been observed (Rothhaupt, 1992). Co-existence of mixotrophs, phototrophs and bacteria, as well as co-existence of bacteria, mixotrophs and obligate phagotrophs, have both also been observed under varied conditions of mixed culture (Rothhaupt, 1996a). However, chemostat experiments with mixotrophs present across a range of OC:P supply ratios appear not to have been carried out. Were such experiments to be carried out, they could help answer a lingering question: can obligate phototrophs persist in systems where C-limitation of bacteria does not occur, or is the only viable strategy to eat one's bacterial competitors for inorganic nutrients as suggested by Nygaard and Tobiesen (Nygaard and Tobiesen, 1993)? Theory does predict that obligate phototrophs could persist if phagotrophic organisms sufficiently restrain bacteria (Thingstad et al., 1996), but the conditions for such co-existence are complex and as yet insufficiently tested.
The experiments reported here and those of Pengerud et al. (Pengerud et al., 1987) show that in the absence of phagotrophic consumers of bacteria, rather high OC:P supply ratios (>250-300 by moles) are required for the competitive exclusion of obligate phototrophs by bacteria. At high OC:P supply ratios, bacteria switch from C-to P-limitation, and become formidable competitors with algae. The switch point occurs at an OC:P supply ratio higher than the cellular OC:P ratio of P-limited bacterial cells [up to 150-160 (Vadstein and Olsen, 1989;Chrzanowski and Kyle, 1996)] because slow-growing, P-limited bacteria have high respiratory demands for C (Thingstad, 1987).
The switch from C-to P-limitation of bacteria may be influenced by the qualitative nature of DOC substrates. For most responses measured here, the effect of substrate type (glucose versus mixed substrates) was small relative to that of OC:P supply ratio. However, DOC concentrations at steady state were much higher when mixed substrates rather than glucose were supplied at the highest OC:P ratio. The mixed substrates may include material more refractory than glucose. Alternatively, bacteria receiving mixed substrates might have had a lower C demand, relative to P demand, than those metabolizing glucose. The substrate mixture contained substantial reduced N, while the bacteria supplied with glucose were forced to rely on nitrate. The energetic costs of reducing nitrate could account for the higher C demand of bacteria supplied only with glucose. If DOC substrates have an N composition suited to bacterial needs, the switch to P-limitation might occur at an OC:P supply ratio lower than ~300.
The results found here and in similar studies suggest that high supply ratios of labile DOC to inorganic P will characterize any natural habitats in which phototrophs and bacteria are simultaneously P-limited and compete. Such habitats are predicted to have bacterial dominance over algae and an accumulation of DOC, despite such heterotrophic dominance. Crude calculations suggest that some fresh waters may have sufficient organic C supply for this to occur. Along a gradient from oligotrophic to eutrophic lakes, total P concentration ranges from 0.1 to 30 µM (Rast et al., 1991), while DOC concentrations range from 300 to 2000 µM C (Jørgensen, 1986). Conservatively assuming that 15% of DOC is readily available to bacteria (Jørgensen, 1986), the molar ratio of labile DOC:total P ranges up to ~500 in oligotrophic lakes and tends to be lower in eutrophic lakes. If supply ratios are similarly high, bacterial dominance over algae due to competition for P could result. Chrysophytes, an algal group in which mixotrophy is common, are most abundant in oligotrophic lakes with low P supplies (Kalff and Watson, 1986;Sandgren, 1988). This pattern may be consistent with the hypothesis that obligate phototrophy is a less viable strategy than mixotrophy or heterotrophy when the OC:P supply ratio is high. Marine coastal sites with high riverine inputs also seem to have P-limited bacteria and accumulation of DOC (Zweifel et al., 1995;Thingstad et al., 1998), and this process also appears seasonally in oceanic sites (Cotner et al., 1997). The gradient of resource conditions produced in laboratory studies such as those presented here may help illuminate the mechanisms structuring microbial communities in such natural habitats.